Stockholm Environment Institute, Technical Report

Size: px
Start display at page:

Download "Stockholm Environment Institute, Technical Report"

Transcription

1 Stockholm Environment Institute, Technical Report Application of SWAT and a Groundwater Model for Impact Assessment of Agricultural Water Management Interventions in Jaldhaka Watershed: Data and Set Up of Models Devaraj de Condappa, Jennie Barron, Sat Kumar Tomer and Sekhar Muddu

2

3 Application of SWAT and a Groundwater Model for Impact Assessment of Agricultural Water Management Interventions in Jaldhaka Watershed: Data and Set Up of Models Devaraj de Condappa, Jennie Barron, Sat Kumar Tomer and Sekhar Muddu

4 Stockholm Environment Institute Kräftriket 2B SE Stockholm Sweden Tel: Fax: Web: Head of Communications: Robert Watt Publications Manager: Erik Willis Layout: Richard Clay Cover Photo: This publication may be reproduced in whole or in part and in any form for educational or non-profit purposes, without special permission from the copyright holder(s) provided acknowledgement of the source is made. No use of this publication may be made for resale or other commercial purpose, without the written permission of the copyright holder(s). Copyright March 2012 by Stockholm Environment Institute

5 Abstract This study contributes to the understanding of potential for Agricultural Water Management (AWM) interventions in the watershed of Jaldhaka river, a tributary of the Brahmaputra river, located in Bhutan, India and Bangladesh. An application of the Soil Water Assessment Tool (SWAT) and of a simple lumped groundwater model was developed for the Jaldhaka watershed. The first stage of this work was to collect a large dataset to characterise the natural and agricultural contexts of the Jaldhaka watershed. The watershed has a contrasting topography, with mountains upstream and large plains downstream. It experiences high rainfall with a monsoonal pattern and an average of 3,300 mm/year. The river flow is seasonal, with a sustained flow during the dry season, high flows during the monsoon and recurrent flood events. The soils are sandy loam (upstream) to silty loam (downstream), with little permeability. The aquifers in the region are alluvial and the groundwater levels in the watershed are shallow and stable. This study contributed to the development of a precise landuse map which identifies the natural vegetation, the water bodies, the settlements / towns, the tea plantations and the different cropping sequences in the agricultural land. Agricultural statistics were gathered at administrative levels for cropping sequences and crop yields. The irrigation in the watershed is predominantly from groundwater, with diesel pumps, to irrigate rice during summer and potatoes during winter. SWAT and the groundwater model were adjusted in an interactive manner: SWAT was calibrated against the observed streamflows while the groundwater model was calibrated against the observed groundwater levels and the interaction aimed at the convergence of both models. The performance was satisfactory for modelling the watershed on an average monthly basis. However, the model set-up failed to reproduce adequately the crop yields. This paper ends with a discussion of the modelling setup and data collection for agro-hydrological modelling. This set-up was applied in an accompanying research report to study the current state of the hydrology in the Jaldhaka watershed and the impacts of two types of AWM scenarios.

6

7 Contents Abstract List of abbreviations 1 Introduction 1 2 Introduction to the modelling softwares Soil and Water Assessment Tool (SWAT) Groundwater model 3 3 Biophysical data of the Jaldhaka watershed Digital Elevation Model Streamflow data Climate data Soils Groundwater data Land-use Agricultural Irrigation 37 4 Modelling set up Initial setting of SWAT Calibration of the groundwater model and SWAT 45 5 Discussion on the input dataset on the model set up 57 6 Conclusion 58 Acknowledgements 60 Annex 62 References 70 iii viii

8 LIST of figures Figure 1: Location of the Jaldhaka / Dharla river watershed (in purple). The delineation of the Jaldhaka / Dharla watershed were generated in this work. 1 Figure 2: Scheme of the modelling 3 Figure 3: Digital Elevation Model from the Shuttle Radar Topography Mission and locations where climatic and streamflow data was available 7 Figure 5: Topographic profile of the transect defined in Figures 3 and 4 7 Figure 4: Slope derived from the DEM, the two local meteorological and streamflow gauge stations 7 Figure 7: Available time-series for streamflows measured at Taluk-Simulbari and Kurigram stations, unfiltered (left) and average monthly streamflow, filtered (right); the vertical error bars indicate the statistical standard deviation of daily streamflows 9 Figure 8: Zoom around Kurigram on Google Earth where are visible the infrastructures for water diversion as well as the neighbouring rivers, in particular the massive Brahmaputra. 10 Figure 9: Representative average rainfall for the Jaldhaka watershed, as calculated by SWAT, and average streamflow at Kurigram (period ) 11 Figure 10: Rainfall at Jalpaiguri and Cooch Behar stations (period ). Top: daily rainfall. Middle: annual rainfall. Bottom: average monthly rainfall, the vertical error bars in red indicate the statistical standard deviation of daily rainfall (in mm/day) 13 Figure 11: Average climatic data at Jalpaiguri and Cooch Behar stations (period ). Top: temperature. Middle: wind. Bottom: humidity. The vertical error bars indicate the statistical standard deviation of daily data 15 Figure 12: Distribution of the average annual rainfall in the sub-watersheds, as represented in SWAT (period ) 16 Figure 13: The georeferenced soil map in the region of the Jaldhaka watershed 17 Figure 14: The Harmonised World Soil Database and its soil units in the region of the Jaldhaka watershed. 17 Figure 15: Plot in soil textural triangle of the United State Department of Agriculture 19 Figure 16: Location of the observation wells for groundwater level measurement. CGWB stands for Central Ground Water Board and SWID for State Water 21 Figure 17: Measured groundwater levels in the Jaldhaka watershed. In pale: level of different wells. In black: average of all the wells 22 Figure 18: Typical groundwater levels in the Jaldhaka watershed. The wells are located on Figure 16. The vertical error bars indicate the statistical standard deviation 22 Figure 19: Interpolation of average piezometric levels observed by the State Water Investigation Directorate (SWID) (period ). 23 Figure 20: Satellite images acquired for high resolution landuse mapping. Note the demarcation between the north and south view 24 Figure 21: Location of the groundtruthing sites visited in April 2010 and draft unsupervised classification of the landuse. Right: zoom on the transect (note on this view the discrepancy 25 Figure 22: Calendar of the main cropping sequences in the Jaldhaka watershed 26 Figure 23: High resolution (10 m) landuse map of the Jaldhaka watershed (year 2008). 27 Figure 24: Photos of the spots identified on the landuse map (Figure 23) 28 Figure 25: Modified version of the landuse map (Figure 23, year 2008) entered in SWAT (90 m resolution) 29 vi

9 Figure 26: Area of the major crops in administrative blocks containing the Jaldhaka watershed 32 Figure 27: Yield of the major crops in administrative blocks containing the Jaldhaka watershed. Mind the different vertical scale 33 Figure 29: Average monthly reference evapotranspiration calculated from difference sources 46 Figure 30: Calibration with respect to the actual evapotranspiration ETa. Monthly value of the different landuse vegetation categories (average over the calibration period, ). 47 Figure 31: Piezometric levels simulated at a monthly time-step by the groundwater model vs. observations 48 Figure 32: Calibration with respect to the recharge of the shallow aquifer (GW_RCHG), average for the Jaldhaka watershed over the calibration period ( ) 49 Figure 33: Calibration with respect to the shallow groundwater baseflow (GW_Q), average for the Jaldhaka watershed over the calibration period ( ) 50 Figure 34: Streamflow simulated (FLOW_OUT) at Kurigram in the initial run over the calibration period ( ) 51 Figure 35: Streamflow simulated (FLOW_OUT) in the final calibration (calibration run n 100) over the calibration period ( ). 53 Figure A.1: Example of the groundtruthing form (site GT 35) filled by the field assistants 69 LIST of TAbLES Table 1: Topographic regions of the Jaldhaka watershed 7 Table 2: Available number of measurements at Taluk-Simulbari and Kurigram stations. Source of data: Bangladesh Water Development Board. 8 Table 3: Available climatic time-series and gaps in the datasets. RMC stands for Regional Meteorological Centre (Kolkata) and NCC for National Climate Centre. 12 Table 4: Annual rainfall at Jalpaiguri and Cooch Behar stations (period ) 14 Table 5: Available measured groundwater levels in the Indian part of the watershed. CGWB stands for Central Ground Water Board and SWID for State Water Investigation Directorate. 20 Table 6: Distribution of the landuse categories (Figure 23) within the Jaldhaka watershed. 28 Table 7: Distribution of the landuse categories entered in SWAT (Figure 25). 30 Table 8: Available agricultural statistics. 30 Table 9: Average yields in the administrative blocks containing the Jaldhaka watershed, period In bracket the average dry yield of rice for period 2004 to Source of data: Bureau of Applied Economics and Statistics and Directorate of Agriculture. 31 Table 10: Typical cropping sequences and associated irrigation schedules in the Jaldhaka watershed. 34 Table 11: Indicative distribution per sub-watershed of the cropping sequences within the landuse units AAAJ and AWJJ. Derived with data from the Bureau of Applied Economics and Statistics and the Directorate of Agriculture. 36 Table 12: Estimated irrigation per crop. Sources of data: groundwater pumping duration from Mukherji (2007) and diesel pump discharge from TERI (2007) 38 vii

10 Table 13: Indicative areas irrigated from surface sources in each sub-watershed, derived from DPDWB (2005) for year 2004/5 38 Table 14: HRUs generation stages. 40 Table 15: Landuse distribution considered in SWAT after pre-processing by ArcSWAT, with respect to the discretisation in HRUs, and management operations for each category. 42 Table 16: Estimation of irrigation areas and amount for 2008, with respect to the discretisation in HRUs. 43 Table 17: Initial values for undetermined SWAT s parameters. 44 Table 18: Average annual reference evapotraspiration calculated from different sources. 46 Table 19: Calibration with respect to the evapotranspiration ETa. Annual values of the ratio ETa / ET0 for the different landuse vegetation categories (average over the calibration period, ). Aman: monsoon rice, Boro: summer rice, Aus: pre-monsoon rice. 47 Table 20: Values of the calibration indicators defined by Eq. (10) to (13). 52 Table 21: Watershed-average dry crop yields simulated by SWAT in the final calibration (calibration run n 100) over the calibration period ( ). 53 Table 22: Simplifications and limitations of the modelling. 55 Table A.1: Soil parameters. Light orange: data from the original soil map from the Indian National Bureau of Soil Survey and Landuse Planning. 62 Table A.2: SWAT vegetation / crop parameters. 65 Table A.3: Sources of irrigation per administrative blocks containing the Jaldhaka watershed, year 2004/5 66 Table A.4: SWAT calibration steps. 68 LIST of AbbREvIATIonS ALAI_MIN SWAT parameter, minimum LAI for plant during dormant period [L2/L2] ALOS Advanced Land Observing Satellite ALPHA_BF SWAT parameter, baseflow alpha factor [-] ArcSWAT ArcGIS interface for SWAT AVNIR Advanced Visible and Near Infrared Radiometer AWC SWAT parameter, available water capacity (AWC, [L3/L3]) AWM Agricultural water management B Baseflow into the streams [L/T] BLAI SWAT parameter, maximum potential LAI [L2/L2] CGWB Central Ground Water Board CH_K(1) SWAT parameter, effective hydraulic conductivity in tributary channel alluvium [L/T] CH_K(2) SWAT parameter, effective hydraulic conductivity in main channel alluvium [L/T] CH_N(1) SWAT parameter, Manning's value for the tributary channel [-] CH_N(2) SWAT parameter, Manning's value for the main channel [-] CHTMX SWAT parameter, maximum canopy height [L] CN2 SWAT parameter, initial soil curve number for moisture condition II [-] DEEPST SWAT parameter, initial depth of water in the deep aquifer [L] DEM Digital elevation model Dnet Net groundwater draft [L/T] DPDWB Development & Planning Department - West Bengal EPCO SWAT parameter, plant uptake compensation factor [-] ESCO SWAT parameter, soil evaporation compensation factor [-] viii

11 ETa Actual evapotranspiration [L/T] ET0 Reference evapotranspiration [L/T] FLOW_OUT SWAT ouput, average daily streamflow out of reach during time step [L/T] GIS Geographical Information System GPS Global Positioning System GT Groundtruthing GW_DELAY SWAT parameter, groundwater delay time [T] GW_Q SWAT ouput, groundwater baseflow contribution to streamflow [L] GW_RCHG SWAT ouput, recharge entering the shallow aquifer [L] GW_REVAP SWAT parameter, groundwater evaporation coefficient [-] GWQMIN SWAT parameter, threshold depth of water in the shallow aquifer required for baseflow to occur [L] h Groundwater piezometric level [L] HRU Hydrologic response unit HWSD Harmonised World Soil Database I Irrigation [L/T] IDC SWAT parameter, land cover / plant classification IWMI International Water Management Institute LAI Leaf area index [L2/L2] M Bias indicator [-] masl Meter above sea level [L] mbgl Meter below ground level [L] NS Nash and Sutcliffe (1970) efficiency [-] NShigh Modified version of NS to emphasise on high flows [-] NSlow Modified version of NS to emphasise on low flows [-] O Net groundwater underflow [L/T] P Rainfall [L/T] PGIS Participatory GIS PHU SWAT parameter, total number of heat units or growing degree days needed to bring plant to maturity RCHRG_DP SWAT parameter, deep aquifer percolation fraction [-] RDMX SWAT parameter, maximum root depth [L] REVAPMN SWAT parameter, threshold depth of water in the shallow aquifer required for evaporation or percolation to the deep aquifer to occur [L] RG Total groundwater recharge [L/T] SEI Stockholm Environment Institute SHALLST SWAT parameter, initial depth of water in the shallow aquifer [L] SOL_K SWAT parameter, Soil conductivity, [L/T] Sol_Z SWAT parameter, depth from soil surface to bottom soil layer [L] SOL_ZMX SWAT parameter, Maximum rooting depth [L] SRTM Shuttle Radar Topography Mission SURLAG SWAT parameter, surface runoff lag coefficient [-] SWAT Soil Water Assessment Tool SWID State Water Investigation Directorate (West Bengal) Sy Specific yield [L3/L3] t Time [T] T_BASE SWAT parameter, minimum (base) temperature for plan growth [ C] T_OPT SWAT parameter, optimal temperature for plan growth [ C] WISE World Inventory of Soil Emission Potentials WTF Water Table Fluctuation ix

12 x

13 Stockholm Environment Institute 1 InTRoDUcTIon Agricultural Water Management (AWM) interventions are often a first step towards increasing smallholder farmers yield levels, their incomes and household food security, in many developing countries. Globally, smallholder farming systems may have the potential to increase current yield levels 2-4 times, and water productivity gains potentially more than double (Rockström 2003). The AgWater Solution project ( is systematically assessing opportunities to invest in agricultural water management interventions at local to continental scale, to enhance smallholder farmers livelihoods. However, agricultural development and intensification can also unintentionally impact various social and environmental dimensions where the interventions are adopted. This report considers the AgWater Solution project watershed of the Jaldhaka river, also known as Dharla, a tributary of the Brahmaputra river. It is a transboundary river originating in Bhutan, flowing through India and joining the Brahmaputra in Bangladesh (Figure 1). The Jaldhaka watershed is one of four project watershed sites, subject to a suite of assessments on agro-hydrological, livelihood and institutional contexts undertaken to identify what potential opportunities there are at a local (watershed) scale and how potential interventions may impact the environment, in particular water resources, and livelihoods. Figure 1: Location of the Jaldhaka / Dharla river watershed (in purple). The delineation of the Jaldhaka / Dharla watershed were generated in this work. Images adapted from Google Earth 1

14 Application of SWAT and a Groundwater Model for Impact Assessment The focus of this work is the development of an application of the Soil Water Assessment Tool (SWAT) and of a simple lumped groundwater model to study the impacts on hydrological balance and crop production under different scenarios of agricultural interventions. This working paper presents the methodology deployed for data collection and agro-hydrological modelling of the surface and groundwater resource of the Jaldhaka watershed. The accompanying research report de Condappa et al. (2011) applies the modelling to analyse the current state of the hydrology and agricultural water management scenarios. The following sections introduce the chosen modelling software (section II.), the input dataset (section III.), the set up of the models (section IV.) and will end with a discussion (section V.). 2

15 Stockholm Environment Institute 2 InTRoDUcTIon To ThE MoDELLInG SofTWARES The primary hydrological model selected for this purpose was the Soil and Water Assessment Tool (SWAT) developed by the United State Department of Agriculture and Texas A & M University. SWAT simulates the different surface and ground hydrological components as well as crop yields. Since its modelling of the groundwater is extremely simplified and the groundwater is a prominent water resource for agriculture in the Jaldhaka watershed, the groundwater model developed by Tomer et al. (2010) was also employed to specifically describe the groundwater processes. The groundwater model interpreted the available groundwater levels, which is not possible with the used version of SWAT, and guided subsequently the setting of SWAT s groundwater parameters. The strategy of the modelling that will be detailed in the following sections is illustrated in Figure Figure 2: Scheme of the modelling 2.1 Soil and Water Assessment Tool (SWAT) For the application to the Jaldhaka watershed, version 433 of SWAT 2009 was used and it was operated through the interface ArcSWAT version General information on this model can be found on the website and in the references Arnold et al. (1993), Srinivasan and Arnold (1994), Arnold et al. (1995) and Arnold et al. (1998) 2.2 Groundwater model The groundwater model considered here was developed by Tomer et al. (2010). It is based on a combination of groundwater budget and the Water Table Fluctuation (WTF) technique. The WTF technique has widely been applied to link the change in ground water storage with resulting water table fluctuations through the storage parameter (specific yield). The WTF is a lumped model based approach suited when limited hydraulic head measurements made at a finite number of observation wells and also little hydrological, geological and meteorological information is available. It was first used to estimate ground water recharge (e.g., in West Africa by Leduc et al. (1997), in Korea by Moon et al. (2004)) and has been extended to estimate change in groundwater storage (e.g., in California by Ruud et al. (2004)) or the ground water recharge and the specific yield (e.g., in India by Maréchal et al. (2006)) with the combined use of groundwater budget. 3

16 Application of SWAT and a Groundwater Model for Impact Assessment The main limitations of the WTF modelling method are: (i) it requires the knowledge of the specific yield of the saturated aquifer at a suitable scale, (ii) its accuracy depends on both the knowledge and representativeness of water table fluctuations and (iii) it does not explicitly take into account the spatial variability of inputs, outputs, or parameters and considers the catchment as an undivided entity and uses lumped values of input variables and parameters. This approach was however relevant to this work as observed time-series of groundwater levels were available at different locations while very few other hydrogeological data were obtained. As SWAT s groundwater module does not simulate piezometric levels, the groundwater model enabled the use of the available measured groundwater levels. The mathematical expressions at the core of the groundwater model developed by Tomer et al. (2010) is the groundwater budget: (1) where S y [L 3 /L 3 ] is the specific yield, h [L] is the groundwater piezometric level, t [T] is the time, R G [L/T] is the total groundwater recharge due to rainfall and other sources including irrigation and recharge from streams, D net [L/T] is the net groundwater draft, B [L/T] is the baseflow into the streams and O [L/T] is the net groundwater underflow from the area across the watershed boundary. The term represents the total discharge (Tomer et al., 2010). In this work, we assumed that O represented regional deep aquifer processes and was nil at the scale of the Jaldhaka watershed. Moreover, following the approach of Park and Parker (2008), a linear relationship was assumed between the baseflow B and the level h: (2) where λ [1/T] is a rate coefficient. Replaced in Eq. (1) it gives: (3) The equation (3) is a linear ordinary differential equation, which can be solved analytically. Following the guidance of Simon (2006), the analytical solution was converted into a discrete equation for the ease of modelling, which can be written as: (4) where A [-] is called the discharge parameter, k is the index for time and the discharge was equal to: (5) As commonly assumed, the recharge R G is calculated linearly from rainfall P [L/T] and irrigation I [L/T]: 4

17 Stockholm Environment Institute (6) where r [-] is the recharge factor. Note that the irrigation I is only equal to Dnet if groundwater is the only source of irrigation water (e.g., no irrigation from river). As in the Jaldhaka watershed the groundwater data did not show a constant recharge factor, a time varying recharge factor was assumed, calculated from the rainfall P and irrigation I: (7) where a [-] and b [T/L] are recharge parameters. Finally, the total recharge RG is expressed as: (8) and Eq. (4) and (8) are used during the calculations of the groundwater model. 5

18 Application of SWAT and a Groundwater Model for Impact Assessment 3 biophysical DATA of ThE JALDhAkA WATERShED Biophysical data were gathered by conducting field work, request to relevant organisations and using publicly available data on internet. 3.1 Digital Elevation Model 3.1.a Source The source of the Digital Elevation Model (DEM) is the Shuttle Topography Mission (SRTM), as pre-processed by Jarvis et al. (2008). 3.1.b Analysis In Figures 3 and 4, three topographic regions can be identified in the watershed (Table 1): mountainous upstream (18 per cent of the watershed), where elevation ranges from 500 to more than 4,000 meter above sea level (masl) and slope from 3 to 40 degrees (within the watershed), piedmont upstream (22 per cent of the watershed), where elevation ranges from 100 to 500 masl and slope from 1 to 3 degree, and plain middle and downstream (60 per cent of the watershed), where elevation ranges from 100 to 18 masl and slope less than 1 degree. The profile of transect defined on Figures 3 and 4 is placed in Figure 5. A striking characteristic of this watershed is the flatness in the plain which: makes the delineation of the watershed boundary downstream highly uncertain, and entails invasive floods from neighbouring rivers, in particular from the Teesta river bordering the watershed in the west; during these events, the delineation is further uncertain as rain falling outside of the watershed s border contributes to the flood of the Jaldhaka river, hence the defined watershed boundaries varies with rainfall and flood events. 3.2 Streamflow data 3.2.a Source Obtaining streamflow from Indian organisations (Central Water Commission, Irrigation and Waterways Department) was impossible during the span of this study. Instead the Bangladesh Water Development Board provided flow of the Jaldhaka at two gauges stations, downstream in the Bangladeshi part: Taluk-Simulbari and Kurigram (Figure 6). The measurements were instantaneous readings of height, i.e., no average over a time-span, from August 1998 to June 2009 with variable frequencies: at Taluk-Simulbari: almost daily up to June 2002, then weekly, at Kurigram: about twice a month. 6

19 Stockholm Environment Institute Figure 3: Digital Elevation Model from the Figure 4: Slope derived from the DEM, the two Shuttle Radar Topography Mission and locations local meteorological and streamflow gauge where climatic and streamflow data was stations available Elevation (masl) 4,500 4,000 3,500 3,000 2,500 2,000 1,500 Crest of the watershed Mountains Piedmont Plains Table 1: Topographic regions of the Jaldhaka watershed Topography Area (km²) Share (%) Mountains 1, Piedmont 1, Plains 3, , Boundary of the watershed Total 5, Distance (km) Figure 5: Topographic profile of the transect defined in Figures 3 and 4 7

20 Application of SWAT and a Groundwater Model for Impact Assessment 3.2.b Inspection of the data The number of measurements is higher at Taluk-Simulbari, especially during the low flow season (Table 2). The hydrographs of both stations were compared and some measurements at Kurigram appeared suspicious (Figure 7); these were removed. In total, 878 measurements at Taluk-Simulbari and 197 at Kurigram were used to calibrate the SWAT application for the Jaldhaka watershed. Table 2: Available number of measurements at Taluk-Simulbari and Kurigram stations. Source of data: Bangladesh Water Development Board. original data filtered data kurigram Taluk-Simulbari kurigram Taluk-Simulbari # % of days # % of days # % of days # % of days Low flows November to April High flows May to October 112 6% % 97 5% % 112 5% 204 9% 100 5% 204 9% Total 224 5% % 197 5% % As Kurigram is located downstream of Taluk-Simulbari, the flow at Kurigram should be greater than at Taluk-Simulbari. The average monthly flow during the months July to March seems to satisfy this criteria (Figure 7). However during April to June, when the first major rain events occur, this is not any longer the case. Without a detailed knowledge of the local conditions, we speculated that this may be due to diversion infrastructures that are visible on Google Earth, which divert part of the raising water (Figure 8). Another apparent anomaly is that peaks of the discharge at Kurigram are much greater than at Taluk- Simulbari. Three possible explanations could be: there are errors in measurements of peak flows at Taluk-Simulbari and / or at Kurigram; we think that these errors are most likely at Kurigram where the standard deviation is very important for measurements in July (Figure 7), in this flat zone the tributaries of the Brahmaputra (in particular the Teesta and Torsa rivers) converge, hence it is possible that flood water from these neighbouring rivers invade the part of the watershed between Taluk-Simulbari and Kurigram, creating much higher flows at Kurigram, Kurigram is just 20 km upstream from the massive Brahmaputra, hence it could be that flood water from the Brahmaputra travel upstream; however, according to the DEM, the difference in elevation from Kurigram and the confluence is about 9 m, which does not favour this possibility. 3.2.c Analysis of the filtered data The flow of the Jaldhaka is perennial at both locations, although there is an important seasonal contrast (Figure 7). The low flow season is between November to April, the average base discharge is 8

21 Stockholm Environment Institute 4,000 3,500 Kurigram Taluk-Simulbari 3,000 Discharge (m3/s) 2,500 2,000 1,500 1, Jun-98 Oct-99 Mar-01 Jul-02 Dec-03 Apr-05 Aug-06 Jan-08 May-09 3,000 Kurigram Taluk-Simulbari 2,500 Discharge (m3/s) 2,000 1,500 1, Month Figure 7: Available time-series for streamflows measured at Taluk-Simulbari and Kurigram stations, unfiltered (left) and average monthly streamflow, filtered (right); the vertical error bars indicate the statistical standard deviation of daily streamflows Source: Bangladesh Water Development Board 9

22 Application of SWAT and a Groundwater Model for Impact Assessment Figure 8: Zoom around Kurigram on Google Earth where are visible the infrastructures for water diversion as well as the neighbouring rivers, in particular the massive Brahmaputra. 114 m 3 /s (coefficient of variation of 52 per cent) at Kurigram and about 11 per cent of the annual flow occurs during this period. The high flow season occurs between May to October with a sharp raise of the discharge in June. Occurrence of flood is common, often in the month of July, with a maximum of 3,700 m 3 /s measured at Kurigram in July According to a personal communication from the Irrigation and Waterways Department Cooch Behar, the maximum flow measured at Mathabhanga (the inlet of sub-watershed 13 on Figure 6) is 10,120 m 3 /s, about three times more the maximum observed during the time period August 1998 to June 2009 (3,700 m 3 /s). The force of the floods can be such that the main river of the region, the Teesta, shifted its course from the Ganges to the Brahmaputra only in the last three hundred years (Kundu and Soppe 2002). The comparison of rainfall and streamflow patterns (Figure 9) shows that both signals reach their maximum in the same month (July) but there is a lag: the increase in streamflow at the beginning of the monsoon is not as rapid as for rainfall and the decrease of flows is buffered at the end of the rain season. Moreover, there is a noticeable baseflow during the dry season. 3.2.d Processing for modelling generation of the sub-watersheds The DEM (section III.1.) was used as such in ArcSWAT and the Automatic Watershed Delineation routine of ArcSWAT generated the set of sub-watersheds for the modelling. As the downstream part of the watershed is very flat, the tributaries downstream were digitised on Google Earth and included in the Automatic Watershed Delineation so as to carve their river beds in the DEM. The same was done with tributaries of the neighbouring Teesta river system (Figure 1) for a correct demarcation. As we had no data on the flow of the Jaldhaka at the confluence with the Brahmaputra, the outlet of the watershed chosen in this work was not this confluence (6,140 km²) but the Kurigram station (5,795 km²) (Figures 2 and 3). 10

23 Stockholm Environment Institute 1,200 Jaldhaka watershed rainfall Streamflow at Kurigram 1,000 (mm / month) Month Figure 9: Representative average rainfall for the Jaldhaka watershed, as calculated by SWAT, and average streamflow at Kurigram (period ) The vertical error bars indicate the statistical standard deviation of monthly values (in mm/month). An important parameter of the delineation routine is the minimum area of the sub-watersheds. As sub-watersheds carry the climatic characteristics of the watershed, it is advisable to consider a sufficient number of sub-watersheds (S. L. Neitsch et al. 2005). However the observed streamflow was only available at two gauge stations and climatic data about every 0.5 (cf. section III.3. and Figure 3), the minimum area threshold was chosen equal to 200 km², which generated a sufficient number of 14 sub-watersheds. Four additional sub-watersheds were manually created (Figure 6): two having the Indian streamflow gauge stations as outlet (NH-31 and Mathabhanga), in case data from these stations become available at a later stage (sub-watersheds n 3 and 10), one for the Taluk-Simulbari station (sub-watershed n 17), and a last to represent climatic data from Jalpaiguri meteorological station (sub-watershed n 8). 11

24 Application of SWAT and a Groundwater Model for Impact Assessment 3.3 climate data 3.3.a Sources Two sources of climate data were obtained (Figure 3): daily rainfall, min and max temperatures, wind and humidity at Jalpaiguri and Cooch Behar stations, period , from the Regional Meteorological Centre, Kolkata; gridded daily rainfall at a resolution of 0.5, period , from the National Climate Centre, Pune (Rajeevan and Bhate 2008); 8 pixels from this dataset are within and near the watershed. The coverage of these dataset is summarised in Table 3. Table 3: Available climatic time-series and gaps in the datasets. RMC stands for Regional Meteorological Centre (Kolkata) and NCC for National Climate Centre. Time period variable Resolution Available measurements (days) Missing measurements Location Source (days) ( per cent) Rainfall Daily 6,040 1, Cooch Behar 7, Jalpaiguri RMC Rainfall Daily 12, Whole India, resolution 0.5 NCC Min temperature Daily 6,036 1, Cooch Behar 5,868 1, Jalpaiguri RMC Max temperature Daily 6,039 1, Cooch Behar 7, Jalpaiguri RMC Humidity Daily 6,017 1, Cooch Behar 6,353 1, Jalpaiguri RMC Wind Daily 3,656 4, Cooch Behar 4,467 3, Jalpaiguri RMC 3.3.b Analysis The annual rainfall is important in the Jaldhaka watershed with values fluctuating from 2,000 to almost 5,000 mm/year, with an average of 3,500 mm/year at Cooch Behar station, during period 1988 to 2008 (Figure 10 and Table 4). The rainfall has a monsoonal seasonal pattern, with a relatively dry season from November to March and a rainy season from April to October (Figure 10). Almost all of the annual rain falls in the rainy season (98 per cent), especially between June to September (80 per cent). Daily rainfall intensities can be very high during the peak of the monsoon, with in average about 33 mm/day per rain event, occurrence every year of events greater than 100 mm/day and a maximum recorded in the magnitude of 470 mm/day in Some daily rain events usually occur every month of the dry season, but of much lesser magnitude. 12

25 Stockholm Environment Institute Daily rainfall (mm / day) Jalpaiguri Year Daily rainfall (mm / day) Cooch Behar Year Annual rainfall (mm / year) 5,000 4,500 4,000 3,500 3,000 2,500 2,000 1,500 1, Jalpaiguri Year Annual rainfall (mm / year) 5,000 4,500 4,000 3,500 3,000 2,500 2,000 1,500 1, Cooch Behar Year Monthly rainfall (mm / month) 1, Monthly Daily event Jalpaiguri Month Average daily rain event (mm / day) Monthly rainfall (mm / month) 1, Monthly Daily event Cooch Behar Month Average daily rain event (mm / day) Figure 10: Rainfall at Jalpaiguri and Cooch Behar stations (period ). Top: daily rainfall. Middle: annual rainfall. Bottom: average monthly rainfall, the vertical error bars in red indicate the statistical standard deviation of daily rainfall (in mm/day) Source: Regional Meteorological Centre, Kolkata Temperature in the watershed is moderately warm with a winter season from November to February, where minimum monthly temperature is about 10 C in the plain, and a summer season from March to October, where maximum monthly temperature is about 32 C in the plain (Figure 11). Wind measurements from both local stations show some differences and after comparing with average data available 13

26 Application of SWAT and a Groundwater Model for Impact Assessment Table 4: Annual rainfall at Jalpaiguri and Cooch Behar stations (period ) Source: Regional Meteorological Centre, Kolkata Station Range (mm/year) Average (mm/year) cv ( per cent) Jalpaiguri 2,000 4,800 3, Cooch Behar 2,500 4,900 3, on the website of the Indian Meteorological Department, measurements from Jalpaiguri station appear erroneous. Values from Cooch Behar indicate that there is a clear variation of wind with time, with a maximum of about 6 m/s in April followed by a continuous decrease to 2 m/s in December. Monthly humidity is high and does not vary much monthly, taking a minimum of 60 per cent in March and a maximum of 85 per cent during several months in winter and summer. High humidity and moderately warm temperature imply that the reference evapotranspiration (R. G. Allen et al. 1998) is modest in the watershed, with an average value of 1,300 mm/year (period ). As a consequence, the ratio [rainfall] / [reference evapotranspiration] is particularly high as it equals 2.5, which is a distinguishable characteristic of this watershed as compared to the other watersheds studied by the AgWater Solution Project. With respect to daily variability of the climate data within a month, unsurprisingly rainfall is the most variable followed by the wind, and less variable is the humidity and almost stable monthly-wise are the temperatures (Figures 10 and 11). 3.3.c Processing for modelling Modelling requires a continuous climate input dataset. As the data from the two local stations has some gaps (Table 3) and the daily rainfall gridded data from NCC ends in 2005, we tried to complement both dataset to have a continuous coverage in the period , which was the calibration period of the groundwater model and SWAT. More precisely we used multivariate regression method: during period to fill gaps in rainfall data from the two local stations Jalpaiguri and Cooch Behar using gridded data as predictor, without considering any time lag in the rainfall of neighbouring stations, during , on the contrary, gridded data at the 8 pixels were predicted from measurements at the two local stations. As for temperatures, humidity and especially wind, a daily missing value was replaced by the average of contiguous days. Gaps of several days were replaced by the average value of the given month. During the processing of input data, the ArcSWAT interface selects for each sub-watershed the meteorological station which is the closest to the centroid of the sub-watershed. In our case, ArcSWAT chose for the whole Jaldhaka watershed the two local stations and three pixels of the gridded data. The equivalent distribution of annual rainfall is mapped on Figure 12. There is no clear pattern of the rainfall with this method of distributing rainfall, with a succession of higher or lower rainfall amount while one moves from upstream to downstream. 14

27 Stockholm Environment Institute 40 Jalpaiguri Max Min 40 Cooch Behar Max Min Temperature ( C) Temperature ( C) Month Month Jalpaiguri Cooch Behar Wind speed (m/s) Wind speed (m/s) Month Month Jalpaiguri Cooch Behar Humidity (%) Humidity (%) Month Month Figure 11: Average climatic data at Jalpaiguri and Cooch Behar stations (period ). Top: temperature. Middle: wind. Bottom: humidity. The vertical error bars indicate the statistical standard deviation of daily data Source: Regional Meteorological Centre, Kolkata 15

28 Application of SWAT and a Groundwater Model for Impact Assessment Figure 12: Distribution of the average annual rainfall in the sub-watersheds, as represented in SWAT (period ) 3.4 Soils 3.4.a Sources Five sources were used: the scanned soil map of West Bengal prepared by the Indian National Bureau of Soil Survey and Landuse Planning, courteously provided by the Indian Space Research Organisation; general data on soil texture and fertility for Cooch Behar district, courteously provided by the Principal Agricultural Officer, Cooch Behar; the Harmonised World Soil Database which is an international database that combines regional and national soil information worldwide (SOTER, ESD, Soil Map of China, WISE) with the information contained within the FAO-UNESCO Soil Map of the World (HWSD 2009); 16

29 Stockholm Environment Institute the qualitative description in Kundu and Soppe (2002); and discussions with local soil scientists from the Agricultural University of Cooch Behar. 3.4.b Analysis The soil map of West Bengal was clipped to the zone of the Jaldhaka watershed, georeferenced and digitised in GIS (Figure 13). Unfortunately we did not have the detailed notice attached to the soil map and only the brief qualitative information from the legend of the map was available (Table A.1, Annex). Three topographical zones are defined in the soil map: the mountainous soils, W001 to W004, shallow, coarse sandy loam, soils in the piedmont, W006 to W008, deep, sandy loam to loam, soils in the plain, W010 to W028, deep, sandy loam to silty loam. The unit Riv additionally describes soils of the river beds. Figure 13: The georeferenced soil map in the region of the Jaldhaka watershed Source: soil map of West Bengal, Indian National Bureau of Soil Survey and Landuse Planning Figure 14: The Harmonised World Soil Database and its soil units in the region of the Jaldhaka watershed. 17

30 Application of SWAT and a Groundwater Model for Impact Assessment The soil spatial units defined by the HWSD in the region of the Jaldhaka watershed (Figure 14) are those of the Digital Soil Map of the World, from the FAO-UNESCO. The HWSD reports additionally some quantitative information from the WISE database (FAO/IIASA/ISRIC/ISS-CAS/JRC 2009), like the textural percentage in sand, silt, clay (Table A.1 in Annex). The information from the Principal Agricultural Officer, Cooch Behar, were used to interpret the soil map (Table A.1, Annex). Local soil scientists indicated the same texture for these soils (Loamy Sand) and insisted that generally the clay percentage is low. Kundu and Soppe (2002) report that the mountainous soils (units W001 to W004) are sandy with high infiltration rates this information was used while entering the soil data in SWAT s database. They also mention that in the plains top-soils are usually Sandy Loam with rather low infiltration rates and that there is a textural transition at about 50 cm in depth for a sandier sub-soil. 3.4.c Pcrocessing for modelling The soil information had to be processed before entering it in SWAT. In particular, the qualitative description had to be transformed into equivalent quantitative values for the soil database of SWAT. The Table A.1 in Annex summarises the values entered in the soil database. The textural percentages from HWSD were plotted in the United State Department of Agriculture soil textural triangle to check if they were in accordance with the information from the soil map (Table A.1) and Kundu and Soppe (2002) (Figure 15). This was not the case as the top soils texture from HWSD was finer (Loam) and sub soils were even finer instead of getting coarser. Corrections were as follows: For soils in mountainous regions (W001 to W004), the percentages of clay and silt from the HWSD for top soils were too high while percentage for sand too low, hence 10 per cent was deducted to the percentage of clay and silt and 20 per cent was added to the percentage in sand. For sub-soils, original values from the HWSD were ignored and instead the corrected values of top-soils were considered, reducing further the percentages of silt and clay in favour of the sand content. For soils in piedmont and plains (W006 to W028), the HWSD clay percentage for top soils were too high (more than 20 per cent) compared to the information from local soil scientists, hence the top soil textures of Loamy Sand units were modified by deducing 10 to the clay percentage and adding it to the sand content; silt content was not modified. For the sub-soil, original values from the HWSD were ignored as well and the textures were calculated by deducing 10 per cent and 5 per cent to the silt and clay content of the top soil, adding it to the sand percentage. The corrected soil textures were indeed matching with description from the various sources (Figure 15). The texture of the unit Riv was not modified. Eventually the qualitative information of the soil map (Figure 13) was translated into equivalent quantitative values using the HSWD and local soil knowledge. The soil hydrologic group, required by the ArcSWAT interface, was derived from the soil texture: group A: coarse sandy loams, units W001, W002, W004 and Riv, group B: fine sandy loams, units W003, W006, W008 to W025, group C: loamy soils, units W007, W026 and W

31 Stockholm Environment Institute Top soil Original texture from the HWSD Top soil Corrected texture Sub soil Original texture from the HWSD Sub soil Corrected texture Figure 15: Plot in soil textural triangle of the United State Department of Agriculture After the soil texture, another important soil characteristic in SWAT is the soil Available Water Capacity (AWC, [L 3 /L 3 ]). The HWSD provides approximate AWC but we preferred to enter values considering the texture of each soil unit (Figure 15) and adapting the capacities advised by Kundu and Soppe (2002). Moreover, there is a sort of continuity between of the soil AWC and the specific yield used by the groundwater model. The value of specific yield was often 0.15 in the plain, less in mountainous sub-watersheds and this was reflected in the AWC, as it will be detailed in section IV.1.d.. Ultimately: unit Riv: top soil AWC = 0.08, unit W001: top soil AWC = 0.10, units W002 and W004: top soil AWC = 0.15, sub-soil AWC = 0.10, units W003, W006, W008 to W025: top soil AWC = 0.20, sub-soil AWC = 0.15, 19

32 Application of SWAT and a Groundwater Model for Impact Assessment units W007 and W026: top soil AWC = 0.25, sub-soil AWC = 0.20, unit W028: top soil AWC = 0.30, sub-soil AWC = The soil depths were also specified combining the information from the soil map and Kundu and Soppe (2002): shallow soils: top soil 50 cm, no sub-soil, moderate shallow: top soil 30 cm, sub-soil 70 cm, deep: top soil 40 cm, sub soil 100 cm, very deep: top soil 50 cm, sub-soil 200 cm. The parameter SOL_ZMX was given the value of 300 cm. The soil conductivity (SOL_K) was chosen with respect to the soil hydrologic group as advised by Neitsch et al. (2010). Finally, the soil map (Figure 13) was approximatively extended using the satellite imageries to cover all the delineation of the watershed. 3.5 Groundwater data 3.5.a Source Groundwater levels in the Indian part of the watershed, from Jalpaiguri and Cooch Behar districts, were obtained from two organisations (Table 5): IWMI provided reading from the Central Ground Water Board (CGWB), and the State Water Investigation Directorate (SWID) of West Bengal. Table 5: Available measured groundwater levels in the Indian part of the watershed. CGWB stands for Central Ground Water Board and SWID for State Water Investigation Directorate. Monitoring organisation number of observation wells Total Jaldhaka watershed period Total CGWB SWID Jaldhaka watershed Measurement frequency Every 3 months Every 3 months 3.5.b Analysis These measurements are mainly located in the plain, with some few in the piedmont (Figure 16). The spatial resolution of the measurement in the Indian part of the watershed is satisfactory with a sufficient number of the observation wells. However, the time resolution is scattered as measurements are not monthly but almost 3 months with some irregularities, hence we may miss some time variation 20

33 Stockholm Environment Institute Figure 16: Location of the observation wells for groundwater level measurement. CGWB stands for Central Ground Water Board and SWID for State Water of the groundwater levels. In particular we may not know exactly when the groundwater levels are the deepest and the shallowest. Leaving aside the groundwater system in the mountains upstream where no observations are available, the groundwater regime can be categorized in two categories: (i) in the plain downstream and (ii) in the piedmont area middle stream. In the plain downstream, the aquifer system is alluvial and composed of ancient sediments from succession of the Ganga Brahmaputra river systems (Kundu and Soppe 2002; CGWB 2009). The groundwater levels are shallow. According to the scattered timeseries of groundwater levels, the water table is apparently deepest in April before the monsoon, fluctuating from 1 to 5 meter below ground level (mbgl), and shallowest in August during the monsoon, fluctuating from 0 to 3 mbgl (Figure 17). Typical groundwater levels are those of wells PTC-25 and PTC-9 (Figure 18). In the piedmont area, the aquifer is composed of more recent sediments carried by the tributaries from the mountainous upstream areas (Kundu and Soppe 2002; CGWB 2009). The groundwater levels are also shallow, although some wells show deeper level. The levels are apparently the deepest between February to April before the monsoon, fluctuating from 2 to 15 mbgl, and the shallowest in August during the monsoon, fluctuating from 1 to 11 mbgl (Figure 17). Typical groundwater levels are those of wells D-10 and D-12 (Figure 18). Watershed-wise, the average groundwater level is between 2 to 4 mbgl (Figure 17). The last 16 years of the groundwater levels time-series show little inter-annual water table fluctuations (Figure 17 and 21

34 Application of SWAT and a Groundwater Model for Impact Assessment Source: Central Groundwater Board (CGWB) Source: State Water Investigation Directorate (SWID) Figure 17: Measured groundwater levels in the Jaldhaka watershed. In pale: level of different wells. In black: average of all the wells In piedmont In piedmont In plain In plain Figure 18: Typical groundwater levels in the Jaldhaka watershed. The wells are located on Figure 16. The vertical error bars indicate the statistical standard deviation Source of data: State Water Investigation Directorate (SWID) 22

35 Stockholm Environment Institute Figure 18): there is no noticeable trend to increase or to decrease, i.e., watershed-wise the groundwater levels have been stable during the period This is consistent with observations from Shamsudduha et al. (2009). Interpolation of observed groundwater piezometric levels in India using the Inverse Distance Squared Weighting method shows that the levels follow the topography, which is expected for an alluvial aquifer system (Figure 19). The gradient is along the North to South orientation and direction in piedmont region and changes towards the South East in the plain. The groundwater flow is therefore in the same orientation and direction of the surface water and converges towards plains in Bangladesh, in particular towards the Brahmaputra river. This is again consistent with Shamsudduha et al. (2009). If we compare the cases when the available observed piezometric levels are the shallowest (August) versus the deepest (April), there is a general shift of the contours along the flow direction, i.e., the topography, but the relative distribution does not change significantly. 3.5.c Pre-processing The dataset from SWID contained more measurements than the one from CGWB (Figure 17), hence only readings from SWID were considered afterwards. Moreover the groundwater model focused on the wells located within the watershed, which amounted to 33 wells. 3.6 Land-use A high spatial resolution landuse map of the Jaldhaka was generated within the context of the AgWater Solution Project. Satellites images were acquired and we contributed to the development of the landuse map by conducting the groundtruthing and helping in the landuse classification. Figure 19: Interpolation of average piezometric levels observed by the State Water Investigation Directorate (SWID) (period ). 23

36 Application of SWAT and a Groundwater Model for Impact Assessment 3.6.a Satellite images Six images from the Advanced Visible and Near Infrared Radiometer type 2 (AVNIR-2) aboard the Advanced Land Observing Satellite (ALOS) were acquired by IWMI. The images are of 10 m spatial resolution and have four bands, three in visible and one in near infrared. The images do not cover the Jaldhaka watershed completely (Figure 20). The six images were for two locations with three replications for each location. The purpose of having multiple images from different seasons is to enable to better extract crop type and rotation information (Cai, personal communication, 2010). Among the three dates, October (31/10/2008) have the clearest views. The two views of January (31/01/2009) are slightly hazy while views of March (i) are from different dates (15/03/2008 and 18/03/2009) and (ii) there is a spatial discrepancy between both views. This combination of dates was representative of the year 2008 and thus an additional difficulty was that groundtruthing was carriedout in April 2010, at a date different to the satellite images. 3.6.b Groundtruthing Before the field work, a draft unsupervised classification was produced on a southern view, to choose the location the groundtruthing (GT) sites (Figure 21). In total, 90 GT sites were visited in April 2010 and two types of observations were carried-out: precise in 40 (GT points 1 to 40) of these sites, brief in the remaining 50 sites (GT points 41 to 90). Taking inspiration from Cai and Sharma (2010), a GT form pertaining to the vegetation distribution and the crop system (sequence, growing period etc.) was prepared. The choices of the sites were as follows (Figure 21): Date North view: 31/01/2009 Date South view: 31/01/2009 Date North view: 15/03/2008 Date South view: 18/03/2009 Date North view: 31/10/2008 Date South view: 31/10/2008 Figure 20: Satellite images acquired for high resolution landuse mapping. Note the demarcation between the north and south view 24

37 Stockholm Environment Institute Figure 21: Location of the groundtruthing sites visited in April 2010 and draft unsupervised classification of the landuse. Right: zoom on the transect (note on this view the discrepancy location of sites was decided in the field while driving on roads referring (i) downstream to the different classes of the unsupervised classification and (ii) upstream to the visible landuse as observed on original satellite image and GoogleEarth, sites were visited while walking along a transect of about 8.5 km; this transect was chosen on the draft classification so as to be representative of the downstream part of the watershed. Precise observations meant a 360 assessment of the land cover (e.g., urban, natural vegetation, agricultural land) and interview of farmers on the spot to distinguish the cropping systems within agricultural lands. Two local field assistants helped for the survey by interviewing farmers and filling the form. An example of this form is shown in Figure A.1 (Annex). Brief observations were quick observation without stopping the car, noting the main land cover categories (e.g., forest, settlement, light forest, tea plantation, agricultural land). Some photos of different landuse are placed in Figure 24. The distribution of GT sites is greater middle and downstream as the draft classification was only available for this part. This is acceptable as most of the agricultural land lies in this part of the watershed. From the GT observations, main cropping sequences were identified (Table 10). The cropping calendar of these sequences is illustrated on Figure c Generation of the landuse map The landuse map was generated in collaboration with IWMI. The first task was to geo-rectify the satellites as some gaps have been observed as compared to GPS measurements. The GIS / remote sensing expert from IWMI ran an unsupervised classification on two separate sets: the first set containing the 3 northern views and the second set containing the 3 southern views. The advantage of processing separately images of different seasons is the changes in vegetation conditions across both 25

38 Application of SWAT and a Groundwater Model for Impact Assessment Rainfed Monsoon_Rice Rainfed Monsoon_Rice Jute Rainfed Monsoon_Rice Rainfed Monsoon_Rice Jute Pre-monsoon_Rice Irrigated Monsoon_Rice Irrigated Monsoon_Rice Pre-monsoon_Rice Potato Jute Irrigated Monsoon_Rice Potato Irrigated Monsoon_Rice Potato Jute Summer_Rice Wheat Irrigated Monsoon_Rice Irrigated Monsoon_Rice Irrigated Monsoon_Rice Summer_Rice Irrigated Monsoon_Rice Wheat Month Figure 22: Calendar of the main cropping sequences in the Jaldhaka watershed views (mainly crops) are taken into account while the pixels are clustered. The northern and southern classifications created each 20 classes (Cai, personal communication, 2010). As this study focuses primarily on agricultural land, IWMI s GIS / remote sensing expert tried to identify crop types as well as crop rotations. Temporal spectral changes of each agricultural class were analysed and the trend in vegetation development was assessed. They were then compared to the crop calendar (Table 10) to match with dominant crop types. GT points and Google Earth were used for to aid crop type determinations (Cai, personal communication, 2010). The fact that groundtruthing was conducted at a date different from the original satellite images hindered the classification. This generated 40 classes: 20 for the northern view and another 20 for the southern view. In an attempt to validate, this classification was queried in a buffer of about 100 m radius around each GT location and extracted landuse was compared with GT observations. Discrepancies were important with, for instance, an over-representation of Wheat, an under-representation of Jute and the absence of a category for towns. Hence we tried to reduce the number of classes and enhance their representativeness by: merging equivalent northern and southern classes, using Google Earth to recognise the GT observations and associate them with classification clusters, differentiate more precisely Tea from Shrublands, creating manually a new class for towns. 26

39 Stockholm Environment Institute Eventually, a high resolution (10 m) landuse map with 11 classes representative of the year 2008 was generated (Figure 23): 8 classes for non-agricultural lands with in particular three categories of habitation zones, from the smallest to the largest: (i) settlements, with few habitations around trees, (ii) villages with a greater number of habitations and sparse vegetation and (iii) towns with urbanised areas, and 3 categories of agricultural practices: - Monsoon_Rice [Pre-monsoon_Rice], i.e., Monsoon_Rice possibly followed by Premonsoon_Rice, - Monsoon_Rice [Winter Crop] [Jute or Pre-monsoon_Rice], i.e., Monsoon_Rice, possibly followed by a Winter Crop, possibly followed by Jute or Pre-monsoon_Rice, - Monsoon_Rice [Winter Crop] Summer_Rice, i.e., Monsoon_Rice, possibly followed by a Winter Crop, followed by Summer_Rice. Locally, the monsoon, summer and pre-monsoon rices are called Aman, Boro and Aus respectively. The schedule of each crop is placed in Figure 22. The photos taken at the locations identified on the Figure 23 are placed in Figure 24. Figure 23: High resolution (10 m) landuse map of the Jaldhaka watershed (year 2008). Photos of the identified observation sites are placed in Figure

40 Application of SWAT and a Groundwater Model for Impact Assessment Table 6: Distribution of the landuse categories (Figure 23) within the Jaldhaka watershed. Landuse category Area (km²) ( per cent) Forest Tea or Light Forest Small Trees or Shrubland or Settlement Monsoon_Rice > [Pre-monsoon_Rice] Monsoon_Rice > [Winter Crop] > [Jute or Pre-monsoon_Rice] 1, Monsoon_Rice > [Winter Crop] > Summer_Rice Village or Fallow Town Water River bed Cloud Total 4, Figure 24: Photos of the spots identified on the landuse map (Figure 23) 3.6.d Processing for modelling The landuse map (Figure 23) cannot be used as such by the interface ArcSWAT as its cloud category has to be replaced by a relevant landuse and it does not cover all of the watershed. The cloud category was replaced referring to the landcover visible on Google Earth, ie., generally Forest upstream in Bhutan and Monsoon_Rice [Winter Crop] [Jute or Pre-monsoon_Rice] in India and Bangladesh. Before extending the landuse map to part of the watershed not covered, as the category River bed is a mixture of sand and gravels bare soil class (visible on satellite images) where hardly any vegetation grows, it cannot be matched to a SWAT land use class, hence River bed was replaced by the category Water. Finally, the map was extended using Google Earth and matching with the soil map: 28

41 Stockholm Environment Institute upstream it was expanded as Forest, while downstream it was a mixture of Monsoon_Rice > [Winter Crop] > Summer_Rice, Monsoon_Rice > [Winter Crop] > [Jute or Pre-monsoon_Rice] and Water. As the GIS raster of the landuse map was to be processed by ArcSWAT, the raster was matched spatially with the DEM, which is the fundamental GIS information for ArcSWAT. This implied in particular that the spatial resolution of the enlarged landuse map (10 m) was downgraded to 90 m. The resulting landuse map is placed in Figure 25 and the tabulated areas are in Table 7: these are the landuse information eventually entered in SWAT. We defined 9 landuse categories for SWAT (e.g., FRSJ for Forest, AAAJ for Monsoon_Rice [Pre-monsoon_Rice]) that were associated to a crop/ vegetation of SWAT s database. The agricultural units AAAJ, AWJJ and AWBJ were generic crop classes, their associated crop sequences, that will be presented in section III.7., were entered in SWAT at a later stage, while defining SWAT s management tables (section IV.1.b.). 3.7 Agricultural 3.7.a Sources Data on crop yields, area extent and productivity were obtained from 3 sources (Table 8): the website of the Development & Planning Department - West Bengal (DPDWB 2005) the Bureau of Applied Economics and Statistics, Kolkata, India, Figure 25: Modified version of the landuse map (Figure 23, year 2008) entered in SWAT (90 m resolution) 29

42 Application of SWAT and a Groundwater Model for Impact Assessment Table 7: Distribution of the landuse categories entered in SWAT (Figure 25). Landuse category In SWAT Area (km²) (%) Forest FRSJ Tea or Light Forest TEAB Small Trees or Shrubland or Settlement FRMJ Monsoon_Rice > [Pre-monsoon_Rice] AAAJ Monsoon_Rice > [Winter Crop] > [Jute or Pre-monsoon_ Rice] AWJJ 1, Monsoon_Rice > [Winter Crop] > Summer_Rice AWBJ Village or Fallow VIFA Town URMD Water WATR Total 5, Table 8: Available agricultural statistics. Source Spatial resolution Timeperiod variables crops Development & Planning Department - West Bengal Administrative Blocks of Cooch Behar and Jalpaiguri districts Area, Yield, Production Monsoon_Rice, Pre-monsoon_Rice, Summer_Rice, Potato, Jute, Wheat, various pulses Bureau of Applied Economics and Statistics Administrative Blocks of Cooch Behar and Jalpaiguri districts Area, Yield, Production Monsoon_Rice, Pre-monsoon_Rice, Summer_Rice, Jute, Wheat, Maize, various pulses Directorate of Agriculture Administrative Blocks of Cooch Behar and Jalpaiguri districts Area, Yield, Production Potato and the Directorate of Agriculture, Kolkata, India. Additionally, typical crop growing seasons were noted during the field work described in section III.6.b. (Figure 22) and are reported in Table b Analysis The major crops in the region of the Jaldhaka watershed are (Figure 22): Summer_Rice, called locally Boro: irrigated rice grown before the onset of the monsoon, from February to May. Pre-monsoon_Rice, called locally Aus: partly irrigated rice grown at the onset of the monsoon, from April to June. 30

43 Stockholm Environment Institute Monsoon_Rice, called locally Aman: rice grown during the rain season, from June/July to September/October, rainfed or partly irrigated depending on the case. Jute: rainfed vegetable fibre grown at the onset of the monsoon, from April to June. Winter Crop: irrigated crop following the rain season, which is Potato (predominantly), Tobacco or Vegetables, from November to February; from now we will consider this crop to be Potato. Wheat: irrigated during winter, from January to April. The irrigation schedule of these crops are described in the following section III.8.. Although Tobacco is an important cash crop in the watershed, in particular in Cooch Behar district, no data could be gathered as this crop is not monitored by governmental organisations. The data from the Development & Planning Department - West Bengal were ignored as these were only for the year 2003/04, however their utility were to assure the homogeneity with statistics from the Bureau of Applied Economics and Statistics and the Directorate of Agriculture. Area and yield of these crops during the period in the administrative block containing the Jaldhaka watershed is shown on Figures 26and 27. There is a gap in the data from the Bureau of Applied Economics and Statistics for the years 2001/02. A striking feature is that the area under Monsoon_Rice is much greater than the other cultivations and is quite stable. The tendency for area under Pre-monsoon_Rice was to gently decrease while potato to increase. The area under Summer_Rice increased sharply in the recent years in the administrative blocks located downstream. Extent of Jute and Wheat cultivation is relatively stable with a small area under Wheat as compared with the other crops. Among the yields, those of Potato are the most varying. The average crop yields in the watershed for the period is estimated by calculating the average yields of the administrative blocks containing the watershed (Table 9). It is noteworthy that the yield of Summer_Rice is greater than Monsoon_Rice and Pre-monsoon_Rice. The agricultural statistics obtained for rice from the Bureau of Applied Economics and Statistics were in two parts: one for the period 1998/ /04, which only mentioned for the rice the clean yield, and another for the period 2004/05 to 2008/09, which mentioned for the rice the dry yield and clean yield. Hence only the clean rice yield was available throughout the period but the dry yield was also reported in Table 9 as it was required to compare with SWAT s outputs (cf. section IV.2.f.). 3.7.c Processing for modelling We derived typical cropping sequences that will be considered in SWAT (Figure 22 and Table 10) from (i) observations during the landuse groundtruthing (cf. section III.6.b.), (ii) Participatory GIS Table 9: Average yields in the administrative blocks containing the Jaldhaka watershed, period In bracket the average dry yield of rice for period 2004 to Source of data: Bureau of Applied Economics and Statistics and Directorate of Agriculture. Monsoon rice (Aman) clean yield (T/ha) pre-monsoon rice (Aus) clean yield (T/ha) Summer rice (boro) clean yield (T/ha) Jute Dry yield (T/ha) Wheat Dry yield (T/ha) potato yield (T/ha) 1.5 (2.4) 1.4 (2.0) 2.2 (2.9)

44 Application of SWAT and a Groundwater Model for Impact Assessment Area (1,000 ha) Years Monsoon_Rice (a.k.a. Aman) Rajganj Mal Matiali Nagrakata Madarihar Kalchine Kumargram Alipuduar I Alipuduar II Falakata Dhupguri Mainaguri Mekhliganj Mathabhanga I Mathabhanga II Cooch Behar I Dinhata I Dinhata II Sitai Sitalkuchi Area (1,000 ha) 35 Rajganj 30 Mal Matiali Nagrakata 25 Madarihar Kalchine Kumargram 20 Alipuduar I Alipuduar II Falakata 15 Dhupguri Mainaguri Mekhliganj Mathabhanga I 10 Mathabhanga II Cooch Behar I Dinhata I 5 Dinhata II Sitai Sitalkuchi Years Pre-monsoon_Rice (a.k.a. Aus) Area (1,000 ha) Years Summer_Rice (a.k.a. Boro) Rajganj Mal Matiali Nagrakata Madarihar Kalchine Kumargram Alipuduar I Alipuduar II Falakata Dhupguri Mainaguri Mekhliganj Mathabhanga I Mathabhanga II Cooch Behar I Dinhata I Dinhata II Sitai Sitalkuchi Area (1,000 ha) 35 Rajganj 30 Mal Matiali Nagrakata 25 Madarihar Kalchine Kumargram Alipuduar I 20 Alipuduar II Falakata Dhupguri 15 Mainaguri Mekhliganj Mathabhanga I 10 Mathabhanga II Cooch Behar I Dinhata I 5 Dinhata II Sitai Sitalkuchi Years Jute Area (1,000 ha) Rajganj 30 Mal Matiali Nagrakata 25 Madarihar Kalchine Kumargram Alipuduar I 20 Alipuduar II Falakata Dhupguri 15 Mainaguri Mekhliganj Mathabhanga I 10 Mathabhanga II Cooch Behar I Dinhata I 5 Dinhata II Sitai Sitalkuchi Years Wheat Area (1,000 ha) Rajganj 30 Mal Matiali Nagrakata 25 Madarihar Kalchine Kumargram Alipuduar I 20 Alipuduar II Falakata Dhupguri 15 Mainaguri Mekhliganj Mathabhanga I 10 Mathabhanga II Cooch Behar I Dinhata I 5 Dinhata II Sitai Sitalkuchi Years Potato Figure 26: Area of the major crops in administrative blocks containing the Jaldhaka watershed Source of data: Bureau of Applied Economics and Statistics and Directorate of Agriculture 32

45 Stockholm Environment Institute Clean rice yield (T/ha) Clean rice yield (T/ha) Dry yield (T/ha) Average of administrative blocks in the Watershed Years Monsoon_Rice (a.k.a. Aman) Average of administrative blocks in the Watershed Years Summer_Rice (a.k.a. Boro) Average of administrative blocks in the Watershed Wheat Years Watershed Rajganj Mal Matiali Nagrakata Madarihar Kalchine Kumargram Alipuduar I Alipuduar II Falakata Dhupguri Mainaguri Mekhliganj Mathabhanga I Mathabhanga II Cooch Behar I Dinhata I Dinhata II Sitai Sitalkuchi Watershed Rajganj Mal Matiali Nagrakata Madarihar Kalchine Kumargram Alipuduar I Alipuduar II Falakata Dhupguri Mainaguri Mekhliganj Mathabhanga I Mathabhanga II Cooch Behar I Dinhata I Dinhata II Sitai Sitalkuchi Watershed Rajganj Mal Matiali Nagrakata Madarihar Kalchine Kumargram Alipuduar I Alipuduar II Falakata Dhupguri Mainaguri Mekhliganj Mathabhanga I Mathabhanga II Cooch Behar I Dinhata I Dinhata II Sitai Sitalkuchi Clean rice yield (T/ha) Dry yield (T/ha) Yield (T/ha) Average of administrative blocks in the Watershed Years Pre-monsoon_Rice (a.k.a. Aus) Average of administrative blocks in the Watershed Jute Years Average of administrative blocks in the Watershed Potato Years Watershed Rajganj Mal Matiali Nagrakata Madarihar Kalchine Kumargram Alipuduar I Alipuduar II Falakata Dhupguri Mainaguri Mekhliganj Mathabhanga I Mathabhanga II Cooch Behar I Dinhata I Dinhata II Sitai Sitalkuchi Watershed Rajganj Mal Matiali Nagrakata Madarihar Kalchine Kumargram Alipuduar I Alipuduar II Falakata Dhupguri Mainaguri Mekhliganj Mathabhanga I Mathabhanga II Cooch Behar I Dinhata I Dinhata II Sitai Sitalkuchi Watershed Rajganj Mal Matiali Nagrakata Madarihar Kalchine Kumargram Alipuduar I Alipuduar II Falakata Dhupguri Mainaguri Mekhliganj Mathabhanga I Mathabhanga II Cooch Behar I Dinhata I Dinhata II Sitai Sitalkuchi Figure 27: Yield of the major crops in administrative blocks containing the Jaldhaka watershed. Mind the different vertical scale Source of data: Bureau of Applied Economics and Statistics and Directorate of Agriculture 33

46 Application of SWAT and a Groundwater Model for Impact Assessment Table 10: Typical cropping sequences and associated irrigation schedules in the Jaldhaka watershed. cropping systems Landuse category in SWAT Start End Growing days Total pumped (mm) Irrigation duration (days) chosen irrigation events Rainfed Monsoon_ Rice AAAJ and AWJJ June September Rainfed Monsoon_ Rice AAAJ and AWJJ July October Jute April June 91 - Irrigated Monsoon_ Rice AAAJ and AWJJ July October mm every 5 days Pre-monsoon_ Rice April June mm every 2 days Irrigated Monsoon_ Rice AAAJ and AWJJ July October mm every 5 days Potato November February mm every 4 days Jute April June 91 - Irrigated Monsoon_ Rice AWBJ June September mm every 5 days Summer_ Rice Wheat AAAJ and AWJJ February May 120 1, mm every 2 days January April mm every 5 days Irrigated Monsoon_ Rice June September mm every 5 days 34

47 Stockholm Environment Institute (PGIS) analysis conducted by SEI (de Bruin et al. 2010) and (iii) the landuse map (Figure 25). The irrigation of these sequences will be explained in the forthcoming section III.8. The next critical steps was to estimate the area of each cropping sequences from the agricultural statistics at administrative block level. These areas were indeed used prior to the modelling to differentiate the agricultural units AAAJ and AWJJ of the landuse map (Figure 25); the unit AWBJ was assigned the sequence Irrigated Monsoon_Rice - Summer_Rice, hence no winter crop was eventually supposed to grown in this unit. As the area growing Monsoon_Rice is much greater than the other crops (Figure 26), we took the assumption that all the fields of units AAAJ and AWJJ grow Monsoon_Rice during the rain season, which may be followed by fallow or crop(s). We did as follows in each administrative block: We calculated the average area of each crops (i.e., Monsoon_Rice, Pre-monsoon_Rice, Summer_Rice, Jute, Wheat, Maize, various pulses, Potato) over the last 5 years. The areas of Maize and various pulses were small, hence Maize was combined with Wheat and various pluses with Potato (approximative same growing season). As this differentiation concerns units AAAJ and AWJJ, where there is no Summer_Rice, for each administrative block we subtracted to the Monsoon_Rice area the Summer_Rice extent. We took the additional assumption that each cropping sequence introduced in Table 10 are distinct, hence the area calculate in step 3 is composed of all the cropping sequences defined in Table 10, except Irrigated Monsoon_Rice - Summer_Rice. We expressed the area growing Pre-monsoon_Rice, Jute, Wheat and Potato (calculated in step 1 & 2) as a percentage of Monsoon_Rice s area estimated in step 3. This percentage was always less than 100 per cent. Then it was decided to distribute the cropping sequences as follows: Area ( per cent) of... was equal to... Irrigated Monsoon_Rice Winter Crop Jute Percentage of Potato computed in step 5 Irrigated Monsoon_Rice Wheat Percentage of Wheat computed in step 5 Irrigated Monsoon_Rice Pre-monsoon_ Rice Rainfed Monsoon_Rice Jute Rainfed Monsoon_Rice Percentage of Pre-monsoon_Rice computed in step 5 Maximum [0,Percentage of Pre-monsoon_Rice computed in step 5 - percentage of Irrigated Monsoon_ Rice Winter Crop Jute] Remaining to 100 per cent Subsequent to this allocation per administrative block, the percentages were proportionally distributed per sub-watersheds (Table 11). It is reminded that these shares concern SWAT s landuse units AAAJ and AWJJ, and that the unit AWBJ was assigned the sequence Irrigated Monsoon_Rice Summer_Rice. This table is indicative as the distribution actually considered during the modelling depends on the spatial discretisation in SWAT (cf. forthcoming section IV.1.b.). Referring to the agricultural statistics (Figure 26), Rainfed Monsoon_Rice is logically the predominant cropping sequence, fol- 35

48 Application of SWAT and a Groundwater Model for Impact Assessment lowed by Irrigated Monsoon_Rice Potato Jute. Table 11 was utilised while determining SWAT s management table (cf. section IV.1.b.). The last step of the crop / landuse processing was to define the relevant SWAT crop / vegetation parameters. The Table A.2 (Annex) summarises the value of the crop parameters of the landuse map. Except for rice, we created a specific category for the vegetation of the Jaldhaka watershed from an existing category of the SWAT database, modifying some parameters as mentioned in the Table A.2 : Large trees (FRSJ): we created from the existing SWAT class Forest-Evergreen, modifying the optimal temperature from 30 to 25 C and the maximum height from 10 to 15 m. Tea (TEAB): from Range-Brush, modifying the maximum LAI from 2 to 3 (cf. We used the defaults crop parameters for rice, the minimum LAI from 0 to 0.7 as it s a perennial crop, the maximum root depth from 2 to 1 m. Table 11: Indicative distribution per sub-watershed of the cropping sequences within the landuse units AAAJ and AWJJ. Derived with data from the Bureau of Applied Economics and Statistics and the Directorate of Agriculture. Subwatershed Rainfed Monsoon_ Rice Irrigated Monsoon_ Rice premonsoon_ Rice Rainfed Monsoon_ Rice Jute Irrigated Monsoon_Rice potato Jute Irrigated Monsoon_Rice Wheat 1 55% 7% 0% 16% 22% 2 46% 11% 0% 23% 20% 3 39% 18% 0% 30% 13% 4 18% 27% 0% 38% 17% 5 11% 35% 0% 25% 29% 6 7% 34% 0% 38% 21% 7 37% 16% 1% 34% 12% 8 49% 19% 0% 22% 10% 9 53% 12% 5% 20% 9% 10 35% 19% 0% 33% 11% 11 61% 7% 1% 25% 6% 12 53% 11% 8% 19% 9% 13 53% 10% 13% 17% 7% 14 52% 10% 19% 12% 7% 15 55% 10% 1% 27% 8% 16 39% 5% 12% 33% 11% 17 52% 18% 7% 14% 10% 18 39% 5% 12% 33% 11% Watershed 43% 15% 5% 25% 12% 36

49 Stockholm Environment Institute Medium trees (FRMJ): from the Large trees category above, reducing the maximum LAI from 5 to 4, the maximum root depth from 3.5 to 2 m, the maximum height from 15 to 10 and the optimal temperature from 25 to 30 C as these trees are more in the plains. Potato and Wheat: from Potato and Spring Wheat, adjusting the temperature factors to match with the Jaldhaka region. Jute: no matching crop was found hence the Jute class was created from the generic agricultural class, taking a maximum LAI equal to 5 (it s a leafy crop) and maximum root depth equal to 1 m. The last parameter to be estimated for the agricultural crops (non perennial) was the total heat units required for plant maturity (PHU) [ C] defined as (S. L. Neitsch et al. 2005): (9) calculated over the growing period, where T i [ C] is the average daily temperature and T base [ C] is the plant base temperature. PHU was assessed using average temperature from the two local meteorological stations (cf. section III.3.) and the crop growing periods (Table 10). 3.8 Irrigation 3.8.a Sources There were three sources of data. The first was Mukherji (2007) which provides typical duration of pumping for Summer_Rice, Monsoon_Rice and Potato, from diesel pumps (Table 12). The author coined these figures by conducting field works in various region of West Bengal, in particular in Cooch Behar district. The second source was DPDWB (2005) which identifies the source of water for irrigation in each administrative blocks for year 2004/5 (Table A.3, Annex). The third source was the irrigation frequency per crop (Table 10) obtained during the groundtruthing field work for landuse mapping (cf. section III.6.b.). 3.8.b Analysis Participatory GIS analysis conducted by SEI within the AgWater Solution Project in the downstream part of the Jaldhaka watershed (de Bruin et al. 2010) reveals that farmers largely pump groundwater with diesel pumps from shallow tubewells. Figures from DPDWB (2005) confirm this characteristic and more precisely that (i) the main source for irrigation is groundwater pumped from shallow tubewells and (ii) this is especially true in Cooch Behar district, where lies the downstream part of the watershed. According to this same reference, the next irrigation source is deep tubewells but the statistics for this category seem suspicious if the number is deep tubewells is compared with the area these wells are supposed to irrigate; hence we ignored this category. The following irrigation source is surface water, from canals or river lifts, especially in Jalpaiguri district, where lies the upstream and middlestream part of the watershed. It is noteworthy that there is schematically a spatial differentiation of irrigation source in the watershed: the upstream and middlestream part of the watershed mainly irrigates from surface water source, more precisely by diverting and lifting the river (canals and river lift); 37

50 Application of SWAT and a Groundwater Model for Impact Assessment while the downstream part uses mainly groundwater (diesel pumps with shallow tubewells). 3.8.c Processing for modelling We considered two types of Monsoon_Rice cultivation: rainfed Monsoon_Rice when Monsoon_ Rice is not followed by any crop or is preceded by Jute, and irrigated Monsoon_Rice when it is followed by a crop which is irrigated: if there is a provision for irrigating winter or summer crop, we supposed this same facility is used to partly irrigate the Monsoon_Rice. We assessed the irrigation requirement of the crops Summer_Rice, Potato and Irrigated Monsoon_Rice in Table 12. We assumed that the requirement for Pre-monsoon_Rice is half of Summer_Rice and that of Wheat is the same as Potato or Irrigated Monsoon_Rice. Using the irrigation frequencies and crops growing period noted during field works, we derived in Table 10 the irrigation requirement for each crop identified in the landuse map (Figure 25). Figure 28 shows the scaling of the crops with respect to irrigation requirement. After assessing irrigation requirement, we had to distribute the source of irrigation water, whether it is from diversion of the river or groundwater pumping. We used the values from DPDWB (2005). As (i) these figures for source of irrigation water are for year 2004/5, (ii) assessing areas under surface irrigation is perhaps more precise than those under groundwater and (iii) total irrigated area is expected to vary from year to year, we only considered the areas for surface sources (canal, tank, river lift), agglomerated them and assumed that the total area irrigated from surface sources did not vary since 2004/5. The spatially proportionate values calculated per sub-watersheds (Table 13) are indicative as the areas actually considered in SWAT depend on the landuse map (Figure 25) and the spatial discretisation in SWAT (cf. forthcoming section IV.1.c.). As Summer_Rice is an irrigated crop, the variation of its area provides an indication of irrigation trends in the watershed. The last 10 years agricultural statistics show a trend to increase and this change in irrigated area is supposed to be due to the variation in number of shallow tubewells. From now we only used this indicative dataset for assessing the Table 12: Estimated irrigation per crop. Sources of data: groundwater pumping duration from Mukherji (2007) and diesel pump discharge from TERI (2007) crop Groundwater pumping duration (hr/ biga a ) Diesel pump discharge (m3/hr) Total groundwater irrigation (m3/ha) Summer_Rice ,334 1,233 Potato , Total groundwater irrigation (mm) Irrigated Monsoon_Rice , a 1 biga ~ ha Table 13: Indicative areas irrigated from surface sources in each sub-watershed, derived from DPDWB (2005) for year 2004/5 Sub-watershed Area (ha) 5,030 1, ,187 3,532 1,105 1,789 1,176 1,551 Sub-watershed Area (ha) 1, ,991 3, , ,466 38

51 Stockholm Environment Institute irrigated areas under surface water source, the remaining irrigated area assumed to be irrigated from groundwater (cf. section IV.1.c.). Table 16 summarises the figures eventually considered in SWAT. Table 13 was used with Table 10 while defining SWAT s management tables (forthcoming section IV.1.b.). 39

52 Application of SWAT and a Groundwater Model for Impact Assessment 4 MoDELLInG SET Up 4.1 Initial setting of SWAT 4.1.a Generation of the HRUs The Hydrologic Response Units (HRUs) were generated with the ArcSWAT interface in two stages. In the first stage, ArcSWAT intersects the GIS layers of the sub-watersheds, landuse, soil and slopeclasses. The first GIS layers were those were presented above, in particular the landuse of Figure 25. The map of slope classes is created by ArcSWAT after the user enters the desired slope classes. Referring to the three typical topographical regions of the watershed and the histogram of Figure 4, the following 3 classes were selected: [0 3 per cent] or [0 1.7 ], [3 43 per cent] or [ ], and greater than 43 per cent or As this first step generates a large number of geographical entities (Table 14), the second step aims to remove small units for computational efficiency. The user enters a threshold for landuse, soil and slope-classes and the simplification is done per sub-watershed: if in a given sub-watershed the coverage of a unit of landuse, soil or slope-classes is less than the associated threshold (for landuse, soil or slope-classes), ArcSWAT ignores the geometrical features generated with this unit in the first step in the considered sub-watershed. As explained by Romanowicz et al (2005), a disadvantage of this simplification is the loss of spatial information which are: small in extent but has a singular characteristic (i.e., Towns), or dispersed in the watershed (i.e., Monsoon_Rice [Pre-monsoon_Rice], Monsoon_Rice [Winter Crop] Summer_Rice). Moreover, spatial information is lost in this second step as ArcSWAT may agglomerate spatially distinct entities which have the same combination of Sub-watershed/landuse/soil/slope-classes. One may consequently question the usefulness of this second step but Romanowicz et al (2005) and Geza and McCray (2008) report that the ability of SWAT to reproduces observed signals is not at best if all the units generated in the first step are kept. Table 14: HRUs generation stages. first stage number of units Second stage Threshold Exempted landuse classes number Landuse Soil Slopeclasses of hrus 970 5% 10% 10% Towns (URMD) Monsoon_Rice [Pre-monsoon_Rice] Monsoon_Rice [Winter Crop] Summer_Rice

53 Stockholm Environment Institute Since the landuse map spatial precision is greater than the soil map, we chose the following thresholds: 10 per cent for soil and slope-classes, 5 per cent for landuse. The landuse categories Towns (VIFA), Monsoon_Rice [Pre-monsoon_Rice] (AAAJ) and Monsoon_Rice [Winter Crop] Summer_Rice (AWJJ) were exempted from the threshold analysis. This second step generated 293 HRUs (Table 14). The number of these HRUs is greater upstream, where the elevated terrain is more contrasted compared to the plain which is more uniform. Distribution of the cropping sequences among the agricultural HRUs Once the HRUs have been generated, a fundamental task is to set the management practices of the two agricultural landuse units. As presented above in section III.7.c., the two agricultural units Monsoon_Rice [Pre-monsoon_Rice] (AAAJ), Monsoon_Rice [Winter Crop] [Jute or Pre-monsoon_Rice] (AWJJ), have to be differentiated with respect to the various cropping sequences occurring in the watershed. For this purpose, the cropping sequences were distributed among the HRUs pertaining to the landuse category AAAJ or AWJJ so as to reproduce as much as possible the percentages of Table 11. While doing so: cropping sequences Irrigated Monsoon_Rice Pre-monsoon_Rice and Monsoon_Rice Winter Crop Jute were distributed in priority within the unit AAAJ and AWJJ respectively to concur with the landuse map, attention was given to the value of the HRU slope so that Rainfed Monsoon_Rice would rather be grown on uneven HRU while the other cropping sequences would be cultivated on flat HRUs. The cropping sequence of the category Monsoon_Rice > [Winter Crop] > Summer_Rice (AWBJ) was chosen to be Monsoon_Rice followed by Summer_Rice. The distribution of all the landuse categories presented in section III.6.d. among the HRUs and the management operations are summarised in Table 15: this is the input landuse information considered by SWAT after pre-processing of ArcSWAT s interface. The landuse information read as input by Arc- SWAT (Table 7) is slightly different from the output of the pre-processing (Table 15), which is due to the HRUs generation routine. 41

54 Application of SWAT and a Groundwater Model for Impact Assessment Table 15: Landuse distribution considered in SWAT after pre-processing by ArcSWAT, with respect to the discretisation in HRUs, and management operations for each category. Landuse category Area Management operations (table mgt2) (km²) (%) Date planting or beginning growing season Date harvesting or End growing season Irrigation Forest (FRSJ) 1, st March 30th November None fertilisation None Tea or Light Forest (TEAB) st March 30th November Auto-irrigation Auto-fertilisation Small Trees or Shrubland or Settlement (FRMJ) st January 31th December None None Rainfed Monsoon_Rice 1, As per Table 10 As per Table 10 As per Table 10 Auto-fertilisation Irrigated Monsoon_ Rice Pre-monsoon_ Rice As per Table 10 As per Table 10 As per Table 10 Auto-fertilisation Irrigated Monsoon_ Rice Wheat As per Table 10 As per Table 10 As per Table 10 Auto-fertilisation Irrigated Monsoon_ Rice Potato Jute As per Table 10 As per Table 10 As per Table 10 Auto-fertilisation Irrigated Monsoon_ Rice Summer_Rice As per Table 10 As per Table 10 As per Table 10 Auto-fertilisation Village or Fallow (VIFA) Modelled as bare soil Town (URMD) Modelled as urban zones Water (WATR) Modelled as Water Total 5, c Definition of the cropping sequences in SWAT s management table Once the cropping sequences have been distributed among all the agricultural HRUs, their schedule in planting, irrigation, fertilisation and harvest were detailed in the management table (table mgt2), following what was summarised in Table 10. The PHU values compiled in Table A.2 were entered for each crop. For trees and tea, as (i) temperature measurements were only available at two stations (Cooch Behar and Jalpaiguri) located in the plain and (ii) Forests and Tea plantations located in hilly / mountainous zones experience low temperatures during winter, these two perennial categories were set to hibernate during months with lowest minimum temperatures (Figure 11), i.e., from beginning of December to end of February. The water source for irrigation (i.e., river or groundwater) tried to reproduce as much as possible the indicative Table 13 for river source while the remaining area was under groundwater irrigation (table mgt1). The Table 16 summarises the consequent irrigated areas and amount per sub-watershed. These figures were partly derived from the landuse map, hence they are representative of the year

55 Stockholm Environment Institute Table 16: Estimation of irrigation areas and amount for 2008, with respect to the discretisation in HRUs. (ha) from surface from groundwater Total (mm/ (% of (ha) (mm/ (% of (ha) (mm/ year) total) year) total) year) 1 3, , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , Subwatershed Watershed total / weighted equivalent 33, , , Surface water irrigation is predominant in upstream sub-watersheds and the contrary for groundwater irrigation. Irrigated areas and amount is higher from groundwater source than river. With SWAT s spatial representation, i.e., with respect to the discretisation with HRUs, the irrigated areas and amount representative of the whole watershed are 138,424 ha and 187 mm / year respectively. As much as 78 per cent of the irrigation (145 mm / year) comes from groundwater and the remaining 22 per cent (42 mm / year) are obtained by river diversion. This amount is very small when compared to the rainfall (Figure 10) and the terms of the water budget, as shown in de Condappa et al. (2011). The sub-watershed 17 has a very high amount of groundwater irrigation as this sub-watershed is located downstream on the west border of the watershed (Figure 6 and Figure 25), where lies most of the area under Irrigated Monsoon_Rice Summer_Rice. It will be explained in section IV.2.a. that the calibration period was During this period the amount of irrigation was generally increasing, reaching the values compiled in Table 16. Viewing the variation in Summer_Rice area during period , which is a crop totally irrigated (Figure 43

56 Application of SWAT and a Groundwater Model for Impact Assessment 28), we assumed that (i) in 1998 the amount of irrigation was 50 per cent of the quantum in 2008 and (ii) irrigation increased linearly between the two dates. This variation was entered in the groundwater model. Considering an inter-annual increase of irrigation is however cumbersome in SWAT s management table (mgt2), hence we assumed that the irrigation was constant during the calibration period and equal to 75 per cent of values compiled in Table 16. For tea, auto-irrigation was chosen. Without specific data on fertilisation, auto-fertilisation was applied to all the crops and Tea. The correct simulation with SWAT of a paddy field with impounded water requires that the concerned HRU is declared as being a pothole (S. L. Neitsch et al. 2005). Unfortunately, only one HRU per sub-watershed can be declared as being a pothole in the version SWAT 2009 used in this work. This would mean that only one HRU per sub-watershed can correctly model rice cultivation, which would not be acceptable in our case as all the agricultural units have a rice cultivation. Hence, we did not declare any pothole, which entailed that rice fields were not simulated as being impounded. This is a limitation of this modelling work as estimations of rice yields and water budget terms (evapotranspiration, groundwater recharge, surface and sub-surface runoff) may be incorrect in agricultural HRUs. 4.1.d Initial values To initiate the calibration, initial values were entered for parameters in SWAT which were not determined in preceding sections (Table 17). The values of CN2 were chosen as advised by Neitsch et al. (2010) with respect of the soil hydrologic group (Table A.1, Annex) and the landuse category. The groundwater parameters were guessed considering typical characteristics of alluvial shallow aquifer. Finally, the hydraulic conductivity of the main and tributary channels were taken equal to 0 mm/h, as suggested by Neitsch et al. (2010) for perennial rivers. Regarding the groundwater model, the specific yield S y is required but values measured in the Jaldhaka basin were not available. Instead, we followed the range of values mentioned by Chatterjee and Purohit (2009) for whole India. Initial value of specific yield was about 0.15 in the in the alluvial system of the plains and less than 0.1 in mountainous sub-watersheds. Table 17: Initial values for undetermined SWAT s parameters. category parameter Initial value Management cn2 As advised by neitsch et al. (2010) Groundwater SHALLST 3,000 mm DEEPST GW_DELAY 3,000 mm 10 days ALPHA_BF 0.5 GWQMIN 2,500 mm GW_REVAP 0.1 REVAPMN 300 RCHRG_DP 0 Channels CH_K(1) 0 mm/h CH_K(2) 0 mm/h 44

57 Stockholm Environment Institute Most of the parameters of Table 17 were adjusted during the calibration. 4.2 calibration of the groundwater model and SWAT 4.2.a Method It is reminded that the modelling set-up is illustrated on Figure 2. The general calibration strategy was to reproduce (1) the evapotranspiration (section IV.2.b.), (2) the measured groundwater levels with the groundwater model so as to assess the groundwater recharge and subsequently reproduce this signal with SWAT (section IV.2.c.), (3) the same as step 2 with groundwater baseflow (section IV.2.d.) and (4) the streamflow (section IV.2.e.). A last stage, which was not part of the calibration process, attempted to validate the crop yields simulated by SWAT (section IV.2.f.). The procedure followed these steps but as calibrating on one variable may change the result of the previously calibrated variable (e.g., tuning the soil parameters to calibrate the groundwater recharge may change the evapotranspiration as compared to the values calculated in the previous calibration step), the procedure was completed several time until variation in all the four fluxes (evapotranspiration, groundwater recharge and baseflow and streamflow) was negligible. In particular, any transformations in steps 1 to 3 impacted the streamflow at the outlet of the watershed (Kurigram gauge station) as this flow is an integrated signal of the whole watershed. A total of 100 manual calibration runs were necessary and following sections will only present results of the initial run, using parameters defined above in section IV.1.d., and of the final calibration (calibration run n 100). The Table A.4 (Annex) summarises the calibration steps. The modelling period of each calibration run was imposed by the availability of observed streamflow, i.e., 1998 to We calibrated on an average monthly time step, except for the streamflows as time-series of daily observed flow was available. To stabilise the water budget and avoid effects of initial conditions, each calibration run was initiated 4 years earlier, i.e., from 1994 to 2008 (climate data were available since 1988); the outputs of period 1994 to 1997 were ignored. The irrigation, as explained in section IV.1.c. above, was equal to 75 per cent of values compiled in Table 16 during this calibration phase. The groundwater model was ran on the 33 wells from SWID within the watershed and the outputs were then extrapolated to the sub-watersheds. 4.2.b Evapotranspiration Average monthly time step We started to check that the calculated reference evapotranspiration as defined by Allen et al. (1998) and referred to as ET0 [L/T], was satisfactory. SWAT can calculate ET0 with three formulas: Priestley- Taylor, Penman-Monteith or Hargreaves. The three options were selected and their outputs were compared with values reported in Kundu and Soppe (2002) and Raghuwanshi et al. (2007) (Figure 29 and Table 18). First it can be noticed that Kundu and Soppe (2002) and Raghuwanshi et al. (2007) provide similar trends. Hargreaves over-estimated ET0, Priestley-Taylor on the contrary under-estimated it. Although solar radiation data from the region was not provided as input climatic data, Penman-Monteith formula yielded the closest values to the two references and was therefore selected to calculate ET 0. We then examined values calculated for the actual evapotranspiration, referred to as ETa [L/T]. More precisely, we inspected for each vegetative landuse category the values taken by the ratio and checked that its monthly average value took sensible values (Figure 30 and Table 19). The initial values for the landuse category Forest (FRSJ), an important category for the watershed, were judged to be too small, hence we aimed at increasing the values taken by this ratio by tuning some soils parameters. 45

58 Application of SWAT and a Groundwater Model for Impact Assessment ET0 (mm/month) 180 Kundu and Soppe (2002) Raghuwanshi et al. 160 (2007) SWAT Penman- Monteith 140 SWAT Hargreaves SWAT Priestley- Taylor Month Figure 29: Average monthly reference evapotranspiration calculated from difference sources Table 18: Average annual reference evapotraspiration calculated from different sources. Source ET0 (mm/year) Kundu and Soppe (2002) 1,360 Raghuwanshi et al. (2007) 1,284 SWAT Penman-Monteith 1,296 SWAT Hargreaves 1,476 SWAT Priestley-Taylor 1,163 With the aim of increasing ETa, sensitivity analysis showed that ETa calculated by SWAT is mainly sensitive to soils AWC, ESCO, EPCO and to a less extent soil thickness (Sol_Z). ETa increases with the parameter EPCO but this parameter was already equal to 1, its maximum value in the initial run. Therefore in this calibration we mainly increased the AWC and Sol_Z (Table A.4, Annex), in coordination with next calibration step (section IV.2.c.) as these two soil parameters also influence on SWAT calculation of groundwater recharge. It was not required to tune ESCO parameter. 4.2.c Groundwater recharge Average monthly time step This stage saw an interaction between the groundwater model and SWAT so as to model the groundwater recharge and baseflow. Indeed, the groundwater model interpreted the groundwater levels, rainfall and streamflows during the low flow season to provide an indication of the magnitude of the recharge and baseflow. At the same time SWAT gave an indication of the spatial and temporal variation of this recharge, with respect to properties of the overlaying soil cover. The aim was that both models converge to similar outputs. 46

59 Stockholm Environment Institute 1.0 FRSJ FRMJ TEAB Aman Aman Aus Aman Wheat Aman Winter crop Jute Aman Boro 1.0 FRSJ FRMJ TEAB Aman Aman Aus Aman Wheat Aman Winter crop Jute Aman Boro ETa / ET ETa / ET Month Month Figure 30: Calibration with respect to the actual evapotranspiration ETa. Monthly value of the different landuse vegetation categories (average over the calibration period, ). Left: initial run. Right: after calibration (calibration run n 100). Aman: monsoon rice, Boro: summer rice, Aus: premonsoon rice. Table 19: Calibration with respect to the evapotranspiration ETa. Annual values of the ratio ETa / ET0 for the different landuse vegetation categories (average over the calibration period, ). Aman: monsoon rice, Boro: summer rice, Aus: pre-monsoon rice. Landuse category frsj frmj TEAb Aman Aman Aus Aman Wheat Aman potato Jute Aman boro Whole watershed Initial run Calibrated (calibration run n 100) The groundwater model was ran on the 33 wells from SWID located within the Jaldhaka watershed. It is reminded that the groundwater draft D net (cf. section II.2.) for irrigation increased linearly from 50 per cent to 100 per cent of the quantum of 2008 between the years 1998 to 2008 (cf. section III.8.c.). At each observation well, the groundwater model utilised the Kalman filter (Kalman 1960) to fit the Eq. (4) and (8) to the observed groundwater levels, using the rainfall P, the groundwater draft D net and the irrigation I which are known (D net and I are different if irrigation is also provided by a river source, in addition to groundwater). This estimated the total recharge R G, the baseflow B and the groundwater level h at each of the 33 wells. The simulated levels compared with measurements are shown on Figure 31 for wells located in different topographical positions. It is noteworthy to remind that the groundwater model was required to reproduce the observed groundwater levels, which was not possible with the version of SWAT used in this work (SWAT 2009). At well D-12, the model simulated groundwater level surfacing at some time step. Simulations suggest that the measurements may have missed moments when groundwater levels were the deepest or the shallowest. The recharge representative of each sub-watershed was then calculated by averaging the values of wells within the sub-watershed. Since the observed groundwater levels were not available at a monthly time step (cf. section III.5.b.) and the groundwater model used in this work is lumped, i.e., a simpler approach than the semi-distributed of SWAT, we aimed at reproducing with SWAT 47

60 Application of SWAT and a Groundwater Model for Impact Assessment Piezometric level (m) Date 01/96 10/98 07/01 04/04 01/ Soil level Simulated Observed 166 In piedmont (well D-10) Piezometric level (m) Date 01/96 10/98 07/01 04/04 01/ Soil level Simulated Observed 86 In piedmont (well D-12) Piezometric level (m) Date 01/96 10/98 07/01 04/04 01/07 71 Soil level Simulated Observed In plain (well P-25) Piezometric level (m) Date 01/96 10/98 07/01 04/04 01/07 36 Soil level In plain (well PTC-9) Simulated Observed Figure 31: Piezometric levels simulated at a monthly time-step by the groundwater model vs. observations The wells are located on Figure 16 the simulated groundwater recharge at an average monthly time step over the calibration period ( ), instead of a monthly time step. Eventually the groundwater model enabled to translate the variation in groundwater levels into recharge, which is reproducible with SWAT. The groundwater recharge of the shallow aquifer calculated by SWAT (GW_RCHG) was sensible to GW_DELAYS, the surface curve number CN2, Sol_Z, to a lesser extent AWC and Sol_K where slope is important. Initially the recharge calculated by SWAT was too small in mountainous sub-watersheds while too high in the plains as compared to the estimation from the groundwater model. Overall SWAT s recharge was too hight (Figure 32). We increased CN2 to reduce infiltration at the soil surface, decreased Sol_K in mountainous region to reduce sub-surface later flow and controlled the increase in Sol_Z in the previous calibration step (section IV.2.b.). We also fixed GW_DELAYS (the time that percolation out of the soil cover becomes shallow groundwater recharge) equal to 0 as the groundwater levels are shallow. 48

61 Stockholm Environment Institute Shallow aquifer recharge (mm / month) Average annual recharge: * Groundwater model: 409 mm / year * SWAT: 967 mm / year Groundwater model SWAT Month Shallow aquifer recharge (mm / month) Average annual recharge: * Groundwater model: 569 mm / year * SWAT: 571 mm / year Groundwater model SWAT Month Figure 32: Calibration with respect to the recharge of the shallow aquifer (GW_RCHG), average for the Jaldhaka watershed over the calibration period ( ) Left: initial run. Right: after calibration (calibration run n 100) The calibrated shallow aquifer recharge (GW_RCHG) is plotted on Figure 32. Note that output of the groundwater model also changed during the numerous calibration runs as this is an interactive and iterative process. The annual sum calculated by SWAT is very close to the value simulated by the groundwater model (569 mm/year vs. 571 mm/year) but the monthly figures are different. This is due to the different approach that both models follow to compute the recharge from rainfall and irrigation: the groundwater model uses a linear relationship, while SWAT calculates the recharge with non-linear rules function of soils properties. Therefore we focused on reproducing the annual amount and optionally match as much as possible the monthly trends. Calibrated values of CN2 are very high. Its maximum value is 95 for the agricultural unit Monsoon_Rice Summer_Rice: in the absence of any pothole, this value was purposely chosen to simulate the rather impermeable soils of this agricultural practice. Others agricultural units in soils of the plain were given the value 90, which is also a high value, which is consistent with the low conductivity reported by Kundu and Soppe (2002) for soils of the region and the fact that values reported in Neitsch et al. (2010) are appropriate for a 5 per cent (a little less than 3 ) slope, while the slope is mostly less than 1 in the plain (Figure 4). 4.2.d Groundwater baseflow Average monthly time step Similarly to the groundwater recharge, this calibration step aimed at converging the simulations by both models of the groundwater baseflow. Using the simulations of the groundwater model, the baseflow representative of each sub-watershed was calculated by averaging the values of wells within the sub-watershed. Eventually the groundwater model enabled to translate the variation in groundwater levels into baseflow that can be in turn reproduced with SWAT. As was the case for the recharge, the calibration in SWAT was also realised on an average monthly time step over the calibration period ( ), instead of a monthly time step. The groundwater baseflow of the shallow aquifer calculated by SWAT (GW_Q) was sensible to the 49

62 Application of SWAT and a Groundwater Model for Impact Assessment baseflow recession constant (ALPHA_BF), the deep aquifer percolation fraction (RCHRG_DP) and the shallow aquifer threshold for baseflow (GWQMIN). The observed streamflow at the outlet of the watershed (Kurigram gauge, cf. Figure 7) and the groundwater model demonstrate that the groundwater baseflow is buffered all along the year and in particular baseflow occurs during the relative dry season. This buffered groundwater baseflow was absent in the first run (Figure 33) and the solution was to decrease drastically ALPHA_BF and play with GWQMIN. This was in coordination with next calibration step (section IV.2.e.) as these two groundwater parameters also influence on SWAT calculation of streamflow. For simplification purpose, RCHRG_DP was taken equal to 0. Indeed, processes of the deep aquifer in such alluvial context were assumed to be governed by lateral transfers between regions outside of the Jaldhaka watershed. Baseflow from shallow aquifer (mm / month) Average annual baseflow: * Groundwater model: 427 mm / year * SWAT: 313 mm / year Groundwater model SWAT Month Baseflow from shallow aquifer (mm / month) Average annual baseflow: * Groundwater model: 421 mm / year * SWAT: 418 mm / year Groundwater model SWAT Month Figure 33: Calibration with respect to the shallow groundwater baseflow (GW_Q), average for the Jaldhaka watershed over the calibration period ( ) Left: initial run. Right: after calibration (calibration run n 100) In the final calibrated run, the matching is greatly improved (Figure 33). Note that output of the groundwater model again changed during the numerous calibration runs. The final value of ALPHA_BF (0.002, cf Table A.4, Annex) is extremely small with respect to values presented in Neitsch et al. (2010), hence this watershed setting of Jaldhaka may be an extreme case reaching the limitation of SWAT modelling. Having this in mind and similarly to the case of groundwater recharge, we focused on reproducing the annual amount and optionally match as much as possible the monthly trends. The calibrated value of shallow groundwater baseflow is less than the recharge: this is logical since the modelling accounted for irrigation groundwater pumping. 4.2.e Streamflows Daily time step The critical step of the calibration was to reproduce the observed streamflows at the stations Taluk-Simulbari and Kurigram in Bangaldesh. As it could be expected, there was a gap between modelled and observed values in the initial run (Figure 34). More precisely (i) simulated flows were too low during the low season, as the groundwater baseflow was greatly underestimated during this period (cf. section IV.2.d.) and (ii) the signal of the streamflow was too sharp hence it was required to delay the flows, i.e., the runoff. 50

63 Stockholm Environment Institute 5,000 4,500 Simulated Observed 1,600 1,400 1,200 Simulated Observed Streamflow (m3/s) 4,000 3,500 3,000 2,500 2,000 1,500 Streamflow (m3/s) 1, , /98 10/99 03/01 07/02 12/03 04/05 08/06 01/ Month Figure 34: Streamflow simulated (FLOW_OUT) at Kurigram in the initial run over the calibration period ( ) Left: daily time-series. Right: monthly averages Since the streamflow at Kurigram and Taluk-Simulbari is an integrated signal representative of the areas upstream, the simulated streamflow (FLOW_OUT) was sensitive to all the parameters modified in the previous calibration steps (sections IV.2.b. to IV.2.d.) plus the values of Manning s roughness coefficient for the channels (CH_N(1) and CH_N(2)) and the surface lag coefficient (SURLAG). Only CH_N(1) and CH_N(2) were modified in this section to introduce lag in the runoff and it was not required to tune SURLAG. To quantitatively assess the quality of the calibration, we used four metric indicators. The first checked that there is no bias in the modelling by making sure that the magnitude of the simulations over the calibration period is close to the values measured. For this purpose we calculated the following bias indicator M [-]: (10) with Q SWATi [L 3 /T] and Q Obsi [L 3 /T] respectively the simulated and observed on day i; the sum is calculated over the calibration period ( ). The second indicator was the Nash and Sutcliffe (1970) efficiency NS [-]: (11) with the average of Q Obsi over the calibration period ( ). The third indicator was a modified version of NS to emphasise on low flows NS low [-] (Krause, Boyle, and Bäse 2005): 51

64 Application of SWAT and a Groundwater Model for Impact Assessment (12) The forth and last indicator was a modified version of NS to emphasise on high flows Ns high [-]:. (13) The aim of the calibration was quantitatively to have all the 4 indicators closest to 1 and qualitatively to improve the visual fit. We paid greater attention to values of NS low, in particular at Taluk- Simulbari, as measurement of high flows are suspected to be less precise than low flows and number of readings is more at Taluk-Simulbari (section III.2.b.). To achieve this, we increased CH_N(1) and CH_N(2) to create roughness, hence lag, along respectively the tributaries and the main channel. The final value of CH_N(1) and CH_N(2) (respectively 0.5 and 0.3, cf Table A.4, Annex) are high (S. Neitsch et al. 2010) and can be explained by the particular configuration of the watershed middle and downstream which is extremely flat with various depressions. Values of the 4 indicators at the initial run and final calibration (calibration run n 100) are showed in Table 20 and the calibrated hydrographs are shown on Figure 35. Unfortunately values of the 4 indicators are not available for Taluk-Simulbari for the initial run. The value of M for Taluk-Simulbari is very close to 1, which shows that there is no bias for the simulation down to this station. The values of NS, NS high and NS low are lower for Taluk-Simulbari than for Kurigram which is due to the much greater number of available observations at Taluk-Simulbari. Overall values of these three parameters are judged to be satisfactory at both gauge stations. The average flows during the low season are apparently slightly over-estimated at both stations but this final calibration configuration yielded the values of M, NS, NS high and NS low closest to 1. Table 20: Values of the calibration indicators defined by Eq. (10) to (13). Station Sub-watershed Initial run calibrated (calibration run n 100) M ns ns low ns high M ns ns low ns high number of observations Taluk- Simulbari NA NA NA NA Kurigram Incalculable The gap between simulation and observation is greater for high flows, in particular at Kurigram where M takes the low value of This is acceptable referring to section III.2.c.: the greater the flow, the greater the error in measurement; floods from other river systems may invade the most downstream part of the watershed hence create flow-peaks or higher magnitude; this cannot be modelled with the current modelling set-up. 52

65 Stockholm Environment Institute 4,000 3,500 Simulated Observed Taluk-Simulbari 4,000 3,500 Simulated Observed Kurigram 3,000 3,000 Streamflow (m3/s) 2,500 2,000 1,500 1,000 Streamflow (m3/s) 2,500 2,000 1,500 1, /98 10/99 03/01 07/02 12/03 04/05 08/06 01/ /98 10/99 03/01 07/02 12/03 04/05 08/06 01/08 1,600 1,400 Simulated Observed Taluk-Simulbari 1,600 1,400 Simulated Observed Kurigram 1,200 1,200 Streamflow (m3/s) 1, Streamflow (m3/s) 1, Month Month Figure 35: Streamflow simulated (FLOW_OUT) in the final calibration (calibration run n 100) over the calibration period ( ). Top: daily time-series. Bottom: monthly averages. 4.2.f Crop yields Annual time step This stage did not involve any calibration. We attempted instead to validate the dry crop yields simulated by SWAT in each agricultural HRU by comparing them to the administrative agricultural statistics. The average yields (period ) simulated for each cropping sequence defined in Table 15 and for the whole watershed are placed in Table 21; the average yield of the Monsoon_Rice from all the cropping sequences is mentioned as well. Table 21: Watershed-average dry crop yields simulated by SWAT in the final calibration (calibration run n 100) over the calibration period ( ). Aman: monsoon rice, Boro: summer rice, Aus: pre-monsoon rice. cropping sequence Aman Aman Aus Aman Wheat Aman potato Jute Aman boro All Aman Aman Aus Aman Wheat Aman potato Jute Aman boro Aman Dry yield (T/ha)

66 Application of SWAT and a Groundwater Model for Impact Assessment When compared with the agricultural statistics (Table 9), those values do not match well for at least four reasons. First, the crop parameters presented in section III.7.c. for Rice, Wheat and Potato are not specific for the region of the Jaldhaka watershed. Moreover, Jute was not present in SWAT s database and no agronomic information was available on this crop, hence its characteristics were arbitrarily derived from SWAT s generic agricultural class. Second, yields can be calculated in different ways. SWAT estimates the dry yield while the agricultural statistics mention: for rice, the clean yield, which does not match with SWAT s results, and the dry yield, to which SWAT s estimation are closer, for potato, the wet yields as potato contains a significant amount of water after harvest; if the water content is assumed to be about 80 per cent, hence the dry matter is 1/5th of the total potato mass, SWAT s yield is closer to agricultural statistics. Third, the management practices in SWAT considered auto-fertilisation, which results in greater application of fertilisers as compared to reality. This may entail an over-estimation of the yields. Fourth, as was emphasised in section IV.1.c., a correct simulation of paddy fields by SWAT would require to use the pothole characteristic, which was not possible in this application in the Jaldhaka watershed. Hence, in SWAT rice fields were simulated as regular fields which are not impounded. This should obviously lead to wrong yield calculations, in particular for the Summer_Rice and Pre-Monsoon_Rice which are irrigated, hence impounded. This limitation may not impact as much the yield simulation of rainfed Monsoon_Rice. All these reasons imply that the yields simulated by SWAT are not reliable, although calculated values are close to the administrative statistics for clean rice and potato. Analyses of modelling output should not utilise the crop yields simulated by SWAT. 4.2.g Simplifications and limitations of the modelling The Table 22 summarises the simplifications and limitations of the modelling. As the dataset was limited (no monthly groundwater levels, only the streamflow was on a daily time step and this only for 2 gauges) and the model was not validated with an independent dataset, the current version is only valid to show monthly or annual average trends at the scale of the whole watershed. In particular, it should not be used to analyse outputs of individual month, year or sub-watershed. Application of this modelling is reported in de Condappa et al. (2011). 54

67 Stockholm Environment Institute Table 22: Simplifications and limitations of the modelling. category Limitation Why? consequence Ponding of paddy field should be modelled in SWAT with a pothole Ignored Only one HRU per sub-watershed can be a pothole in SWAT 2009 The hydrological functioning of a paddy field is not modelled correctly, which could affect calculations of: rice yield, water budget terms around the paddy field (evapotranspiration, groundwater recharge, surface and sub-surface runoff). Domestic & industrial water consumption Ignored No data Should not be an issue as there is no major city nor industries along the Jaldhaka. Moreover these demands usually consume little water as compared to irrigation. Deep aquifer Ignored Deep aquifer processes in such alluvial context were assumed to be governed by lateral transfers between regions outside of the Jaldhaka watershed. Should not affect the hydrological modelling of the Jaldhaka river network. Crop yield not simulated correctly Possible bad representation in SWAT of the crops and agricultural practices in the Jaldhaka watershed. Analysis of the current and scenario contexts ignored SWAT's simulations for crop yields. Snow accumulation and melting upstream Ignored No data on snow accumulation and lapse in temperature Should be negligible as main rain occurs during the warm period. Groundwater pumping for irrigation Considered constant in SWAT Cumbersome to enter a varying irrigation in management table (mgt2) Streamflow Available only at 2 locations downstream Data restriction in India Not analyses are possible at subwatershed scale but only for the whole watershed. 55

68 Application of SWAT and a Groundwater Model for Impact Assessment 5 DIScUSSIon 5.1 on the input dataset Input data is a critical requirement for a successful modelling and are of two sorts: (i) data characterising the studied region and (ii) the forcing data to calibrate and validate the model. A fair dataset was gathered in this work and we tried to complement gaps. This enabled a satisfactory modelling of average trends at the watershed scale. There were nonetheless some shortcomings. In term of data characterising the studied region, climatic data, and in particular rainfall, is the primary input data. In the Jaldhaka watershed, the daily rainfall was the most varying variable within a month, followed by wind while the humidity was less variable and the temperature almost stable monthly-wise (Figures 10 and 11). Hence the priority to capture the climatic characteristics of a region is to gather rainfall and wind data at a time and space resolution as fine as possible, while humidity and temperatures could be compiled at a coarser scale. In this work, we collected daily climatic data at two local stations and complemented with a gridded daily rainfall dataset to include a spatial representation of the variability of rainfall. To characterise succinctly watershed, SWAT generates Hydrologic Response Units (HRUs) from the topographical, soil and landuse information. Obtaining topographical data, ie., the Digital Elevation Model (DEM), is usually not an issue for large watersheds / basins, as it was the case in this work with the Shuttle Radar Topography Mission (SRTM). Soil information was the second information required to create the HRUs and describe the soil hydrological processes and vegetation growth cycles. Qualitative information is usually accessible through international database, such as the FAO Digital Soil Map of the World or the World Reference Base, or national agencies, as was the case in this work with the soil map of the Indian National Bureau of Soil Survey and Landuse Planning. A agro-hydrological model like SWAT however requires quantitative soil data usually not as available, such as the thickness of the different horizons, the soil texture (e.g., percentage of clay, silt, sand), and the soil structure (e.g., soil water retention parameter such as the Available Water Capacity in SWAT AWC -, the soil conductivity parameter such as the empirical curve number CN2 in SWAT) in these horizons. In our case, we did not have specific local data for these quantitative soil parameters and we used instead the free Harmonised World Soil Database. This international database was not matching with the local soil knowledge and literature hence we had to adjust and complement to / with the local knowledge. During SWAT s calibration, we found that the most sensitive soil parameters were the AWC, the soil thickness and the curve number CN2. One may question the relevance of modifying these three last soil parameters as they should be estimated based on local data, which we try to attain before the calibration. Modifying the soil thickness could indeed be questioned as it is informed by the soil map, although in an approximative manner. However, the AWC and the curve number are much dependent of the soil structure, i.e., the arrangement of the soil particles in the soil medium, which is extremely variable in space and time, thus these two parameters can be considered as calibration parameters. The landuse is the third layer necessary to generate the HRUs. In addition to describing the land occupation, it should inform as much as possible on the cropping patterns in agricultural lands. Our case was ideal as we contributed to the generation of a fine resolution landuse map in the context the AgWater Solutions Project in the Jaldhaka watershed. In case no specific landuse map is accessible, an alternative could be to gather landuse map from international database and try to use available tools, such as Google Earth, to attempt to improve the landuse map. In this work without reservoirs in the watershed and in addition to generate HRUs, crop information and management practices are necessary to describe appropriately the cropping sequences. Crop data pertain to agronomic properties for simulating the growth of vegetation and the crop yields. We lacked the specific characteristics for the crops grown in the Jaldhaka and approximated them with values 56

69 Stockholm Environment Institute present in the ArcSWAT crop database. This was one of the causes of our failure to simulate correctly crop yields in the Jaldhaka basin. Finally, knowledge of management practices, such as irrigation and fertilisation applications, are essential as well. Though we collected typical irrigation time-schedules during the groundtruthing of the landuse map developed in AgWater Solution project, we had no systematic and continuous information on the current and past irrigation amount in the watershed. Alternatively, we approximated fairly well these irrigations using the precise landuse map and taking the trivial assumption that farmers growing an irrigated crop (e.g., Summer_Rice) have the mean to irrigate. Subsequently and in a further approximation, we chose for the irrigation amount values reported in the literature for West Bengal regions. Regarding the fertilisation, we possessed no data specific to the Jaldhaka watershed map and instead used the auto-fertilisation function of SWAT. This was not satisfactory and surely contributed to a wrong simulation of crop yields. With respect to forcing data, the two variables considered here were streamflows and groundwater levels. Availability of the measured streamflows at the outlet of the watershed was critical to calibrate SWAT, which guaranteed a reproduction of the integrated hydrological signature of the watershed. Moreover, the streamflows being highly variable within a month, it is advisable to collect daily measurements, which was feasible in this work. We lacked however measurements at intermediary location in the watershed and therefore we could not simulate sub-watershed hydrological processes. The groundwater data was particularly required in the Jaldhaka watershed as groundwater is the predominant source for irrigation and it was therefore important to model it appropriately. In India groundwater levels are more easily obtainable than streamflows and, in our case and thanks to the State Water Investigation Directorate of West Bengal, a good dataset of groundwater levels was available free of cost. Although this dataset was limited as it was not monthly, which means that we might have missed the months when groundwater levels are the deepest or the shallowest, it enabled the simulation of groundwater levels fluctuation. We missed in our dataset measured values of the specific yield in the Jaldhaka watershed, that we eventually approximated with range of values for whole India. 5.2 on the model set up One characteristic of this work is to have associated two different models: SWAT and the groundwater model developed by Tomer et al. (2010). The groundwater model enabled the interpretation of the measured groundwater levels, which was impossible with the version of SWAT used here (SWAT 2009, version 433). Subsequently, the groundwater model translated the variation in groundwater levels into fluctuation of groundwater recharge. SWAT also contributed to the determination of the recharge by improving its spatial and temporal variation, with respect to properties of the overlaying soil cover. Eventually both models interacted to simulate correctly the recharge with respect to the rainfall and the groundwater levels. An additional benefit of using the groundwater model was to guide SWAT in reproducing the buffered groundwater baseflow, which is essential to simulate the perennial nature of the Jaldhaka river. Indeed, the first calibration runs with SWAT failed critically to model the baseflow during the dry season (Figure 33). The calculations of the groundwater model motivated the reduction of SWAT s groundwater parameter ALPHA_BF to an extremely small value, which improved drastically the simulation of the streamflows. Since the final value of ALPHA_BF is smaller than the range mentioned by Neitsch et al. (2010), this case is extreme in term of SWAT s groundwater modelling and the groundwater model was critical to guide towards this singular setting. Application of this modelling is reported in de Condappa et al. (2011). 57

70 Application of SWAT and a Groundwater Model for Impact Assessment 6 conclusion This paper contributed to the understanding of potential for development of Agricultural Water Management (AWM) in the watershed of the Jaldhaka river, a tributary to the Brahmaputra river, located in Bhutan, India and Bangladesh. An application of the Soil Water Assessment Tool (SWAT) and of the lumped groundwater model of Tomer et al. (2010) was developed as a tool to study AWM development scenarios in the Jaldhaka watershed. The first stage of this work was to collect data / information to characterise the natural and agricultural contexts of the Jaldhaka watershed. The watershed has a contrasted topography, with mountains upstream and large plains downstream. It experiences high rainfall with a monsoonal pattern and an average of 3,300 mm/year. The river flow is perennial with recurrent occurrence of flood events during the monsoon. The aquifers are alluvial in the region and the groundwater levels are shallow and stable in the watershed. This study contributed to the development of a precise landuse map which identifies in particular the different cropping sequences in agricultural lands. Agricultural statistics were gathered at administrative levels and the irrigation in the watershed was found to be predominantly from groundwater, with diesel pumps, and to irrigate rice during summer and potato during winter. A fairly large dataset was gathered and we tried to complement gaps. There were nonetheless some shortcomings. In term of data characterising the studied region, climatic data, and in particular rainfall, are the most required input data. In the Jaldhaka watershed, the daily rainfall and wind were the critical climate data as they were the most varying variable within a month. Soil information is usually available in a qualitative form but quantitative information necessary for agro-hydrological modelling is rarer and in our case we combined local soil knowledge and literature with an international database. The landuse map describes the land occupation and should inform as much as possible on the cropping pattern in agricultural lands. Our case was ideal as we contributed to the generation of a fine resolution landuse map in the context the AgWater Solutions Project in the Jaldhaka watershed. Precise knowledge of irrigation patterns is required to account for anthropogenic water uses and simulate correctly crop cultivations. Though we had no systematic and continuous information on the current and past irrigation in the watershed, we approximated it fairly well using the precise landuse map and irrigation amount values reported in the literature for West Bengal regions. We lacked specific agronomic information on crops and agricultural practices in the Jaldhaka watershed and consequently failed to reproduce correctly crop yields. With respect to calibration data, we used streamflows and groundwater levels. Availability of the measured streamflow at the outlet of the watershed was critical to calibrate SWAT, which guaranteed a reproduction of the integrated hydrological signature of the watershed but we lacked however measurement at intermediary locations in the watershed and therefore we could not simulate sub-watershed hydrological processes. The groundwater data was particularly required in the Jaldhaka watershed as groundwater is the predominant source for irrigation and it was important to model it appropriately. We gathered a good dataset of groundwater levels and were able to simulate fluctuation of groundwater levels. A characteristic feature of this work was to have associated in an interactive manner the model SWAT with the groundwater model of Tomer et al. (2010). The last enabled the interpretation of the measured groundwater levels, which was impossible with SWAT and which was particularly important in the context of the Jaldhaka watershed. It also guided SWAT in reproducing correctly the buffered groundwater baseflow, which is critical to simulate the perennial nature of the Jaldhaka river. At the same time, SWAT interpreted the measured streamflows and improved the spatial and temporal description of the groundwater recharge. Eventually both models interacted to convergence to a satisfactory simulation of hydrological processes in the Jaldhaka watershed. However, the model set-up failed to reproduce adequately the crop yields. 58

71 Stockholm Environment Institute This modelling framework was applied in an accompanying report (de Condappa et al. 2011) to study the current state of the hydrology in the Jaldhaka watershed and the impacts of two types of AWM scenarios. 59

72 Application of SWAT and a Groundwater Model for Impact Assessment AcknoWLEDGEMEnTS This work was supported by the AgWater Solutions project, funded by the Bill and Melinda Gates Foundation. We are also thankful to the following persons for their contribution (by alphabetical order): Nyayapati Aakanksh, International Water Management Institute, for supporting request of streamflow data. Badrul Alam, International Development Enterprises - Bangladesh, for supporting request of streamflow data. U. S. Aich, Directorate of Agriculture (Kolkata), for providing agricultural statistics on potatoes. Saswati Bandyopadhyay, State Water Investigation Directorate (West Bengal), for providing groundwater data. Gopal Barma, Assistant Director of Agriculture (Administration, Mathabhanga), for providing agricultural statistics. Salim Bhuiyan, Bangladesh Water Development Board, for providing streamflow data. Suman Biswas, International Development Enterprises India, for support during field works. S. Biswas, Agricultural University of Cooch Behar, for providing information on soils. P. K. Biswas, Assistant Agricultural Meteorologist (Jalpaiguri), for providing a copy of Kundu and Soppe (2002). Aniruddha Brahmachari, International Development Enterprises India, for support during field works and data collection. Annemarieke de Bruin, Stockholm Environment Institute, for sharing results of the Participatory GIS in the Jaldhaka watershed. Xueliang Cai, International Water Management Institute, for contributing to the development of the landuse map. Howard Cambridge, Stockholm Environment Institute, for advices on SWAT. D. Dutta, Indian Space Research Organisation, for general advices. Sylvain Ferrant, Indo-French Center for Groundwater Research, for numerous advices on SWAT. Charlotte de Fraiture, International Water Management Institute, for initial advices. Victor Kongo, Stockholm Environment Institute, for advices on SWAT. Monique Mikhail, Stockholm Environment Institute, for sharing results of the Participatory GIS in the Jaldhaka watershed. 60

73 Stockholm Environment Institute K. K. Mondal, Director of the Bureau of Applied Economics & Statistics (Kolkata), for providing agricultural statistics. Aditi Mukherji, International Water Management Institute, for numerous advices on groundwater and data collection. Rajiv Pradhan, Director of International Development Enterprises - Bangladesh, for supporting request of streamflow data. Mala Ranawake, International Water Management Institute, for supporting data requests. Adam Regis, Stockholm Environment Institute, for administrative works. Bhaskar Roy, Assistant Director of Agriculture (Cooch Behar), for providing general information. Bharat Sharma, International Water Management Institute, for supporting request of streamflow data. Raghuwanshi Narendra Singh, Indian Institute of Technology Kharagpur, for providing data of Raghuwanshi et al. (2007). A. K. Sinha, Agricultural University of Cooch Behar, for providing information on soils. Jean Philippe Venot, International Water Management Institute, for support on ArcSWAT interface. Hua Xie, International Food Policy Research Institute, for advices on SWAT. 61

74 Application of SWAT and a Groundwater Model for Impact Assessment AnnEx Table A.1: Soil parameters. Light orange: data from the original soil map from the Indian National Bureau of Soil Survey and Landuse Planning. Light grey: parameters derived by crossing soil map data with information from the Principal Agricultural Officer, Cooch Behar. Light blue: data from the Harmonised World Soil Database. Light yellow: data eventually entered in SWAT. Riv W001 W002 W003 W004 W006 W007 W008 W010 W018 W025 W026 W028 USDA Taxonomy Lithic Udorthents Typic Udorthents Umbric Dystrochrepts Typic Dystrochrepts Umbric Dystrochrepts Fluventic Eutrochrepts Typic Haplaquents Aquic Ustifluvents Typic Ustorthents Aquic Ustifluvents Aeric Haplaquepts Typic Fluvaquents Position Jaldhaka River Hills and side slopes (brown forest soils) Hills and side slopes (brown forest soils) Hills and side slopes (brown forest soils) Hills and side slopes (brown forest soils) Piedmont plain (Terai soils) Piedmont plain (Terai soils) Piedmont plain (Terai soils) Active alluvial plain (Flood plain soils) Recent alluvial plain (most recent soils) Recent alluvial plain (most recent soils) Recent alluvial plain (most recent soils) Recent alluvial plain (most recent soils) Slope position Very steep slope Steep slope Steep slope Steep slope Gentle Very gentle Very gentle Very gentle, on active alluvial plains Very gentle, on recent alluvial plains Very gentle, on recent alluvial plains Very gentle, on recent alluvial plains Very gentle, on recent alluvial plains Depth Shallow Moderate shallow Deep Moderate shallow Very deep Very deep Very deep Very deep Very deep Very deep Very deep Very deep Drainage Excessive Excessive Well Well Imperfect Imperfect Poor Moderately well Poor Imperfect Poor Poor Erosion Severe Severe Moderate Moderate Moderate Moderate Moderate Flood Moderate - Moderate - - Texture Gravelly loamy Coarse loamy Fine loamy Gravelly loamy Coarse loamy Fine loamy Coarse loamy Coarse loamy Coarse loamy Coarse loamy Fine loamy Fine silty 62

75 Stockholm Environment Institute Riv W001 W002 W003 W004 W006 W007 W008 W010 W018 W025 W026 W028 Derived texture Sandy Loam Sandy Loam Sandy Loam Loam Sandy Loam Sandy Loam Loam Sandy Loam Sandy Loam Sandy Loam Sandy Loam Loam Silt Loam Matched HWSD units Top soil texture ( per cent) Gravel: Sand: Silt: Clay 5:78: 13:9 26:43: 34:23 20:42: 37:21 11:44: 33:23 10:41: 39:20 4:39: 41:20 9:34: 43:23 9:42: 36:22 4:37: 40:23 9:42: 36:22 4:49: 32:19 9:34: 43:23 4:33: 45:22 Soil depth (cm) AWC (mm/mm) Top soil dry bulk density (g/cm 3 ) Top soil organic content ( per cent) Sub soil texture ( per cent) Gravel:Sand: Silt:Clay :44: 35:21 26:35: 24:41 20:45: 35:20 8:41: 38:21 7:28: 38:34 12:40: 35:25 5:36: 36:28 12:40: 35:25 8:52: 28:20 7:28: 38:34 5:37: 35:28 Sub soil dry bulk density (g/cm 3 ) Sub soil organic content ( per cent)

76 Application of SWAT and a Groundwater Model for Impact Assessment Riv W001 W002 W003 W004 W006 W007 W008 W010 W018 W025 W026 W028 Corrected top soil texture ( per cent) Gravel:Sand: Silt:Clay 5:78: 13:9 26:68: 19:13 20:62: 27:11 11:44: 33:23 10:61: 29:10 4:49: 41:10 9:34: 43:23 9:52: 36:12 4:47: 40:13 9:52: 36:12 4:59: 32:9 9:34: 43:23 4:23: 55:22 Corrected sub soil texture ( per cent) Gravel:Sand: Silt:Clay :77: 17:6 26:59: 23:18 20:76: 19:5 8:64: 31:5 7:49: 33:18 12:67: 26:7 5:62: 30:8 12:67: 26:7 8:74: 22:4 7:49: 33:18 5:38: 45:17 Hydrologic group A A A B A B C B B B B C C N of layers Thickness layer 1 (cm) Thickness layer 2 (cm) SOL_ZMX (cm) AWC layer AWC layer SOL_K layer 1 (mm/h) SOL_K layer 2 (mm/h)

77 Stockholm Environment Institute Table A.2: SWAT vegetation / crop parameters. vegetation / crop Large trees FRSJ SWAT landuse unit created from FRSE (Forest-Evergreen) Modifications parameter original value Modified CHTMX 10 m 15 m T_OPT 30 C 25 C BLAI 2 3 Tea TEAB RNGB (Range-Brush) ALAI_MIN RDMX 2 m 1 m BLAI 5 4 Medium trees FRMJ FRSJ RDMX 3.5 m 2 m CHTMX 15 m 10 m T_OPT 25 C 30 C Rice AAAJ, AWJJ and AWBJ RICE (Rice) Potato AWJJ POTA (Potato) IDC Cold season annual T_OPT 22 C 25 C Warm season annual T_BASE 7 C 8 C Jute AWJJ AGRL (Agricultural Land-Generic) BLAI 3 5 RDMX 2 m 1 m Wheat AAAJ and AWJJ SWHT (Spring Wheat) IDC Cold season annual T_OPT 18 C 25 C Warm season annual T_BASE 0 C 5 C phu ( c) Monsoon_Rice: 2,220 Pre-monsoon_Rice: 1,540 Summer_Rice: 1,540 1,330 1,540 1,655 65

78 Application of SWAT and a Groundwater Model for Impact Assessment Table A.3: Sources of irrigation per administrative blocks containing the Jaldhaka watershed, year 2004/5 Source of data: DPDWB (2005) Block District Canal Tank River Lift Area No. Area No. Area No. Deep Tubewell Area (ha) (% of total) (ha) (% of total) (ha) (% of total) (ha) (% of total) Alipurduar I Jalpaiguri 2,000 39% % % % Dhupguri Jalpaiguri 3,020 40% % 40 1,200 16% % Falakata Jalpaiguri 3,000 44% % % % Kalchini Jalpaiguri 1,350 58% % % Madarihat Jalpaiguri 3,540 55% % % % Mal Jalpaiguri 2,240 49% % % % Maynaguri Jalpaiguri 1,600 23% % 56 1,800 26% % Metiali Jalpaiguri 1,000 50% % % Nagrakata Jalpaiguri 1,400 61% % % Cooch Behar I Cooch Behar 200 1% % 33 8,844 42% 13 5,440 26% Dinhata I Cooch Behar 150 1% % % % Dinhata II Cooch Behar 40 0% % 23 1,265 15% 22 1,730 20% Mathabhanga I Cooch Behar 200 4% % % % Mathabhanga II Cooch Behar 240 2% % % 10 3,802 39% Mekhliganj Cooch Behar 100 2% % % % Sitai Cooch Behar 50 1% % % % Sitalkuchi Cooch Behar 230 2% % % 10 8,230 56% Total 20,360 17% 169 3,677 3% ,355 16% ,948 19% 66

79 Stockholm Environment Institute Block District No. Shallow Tubewell Dugwell Area No. Area No. Others Total Area No. (ha) (% of total) (ha) (% of total) (ha) (% of total) (ha) (% of total) Alipurduar I Jalpaiguri % % 1,380 1,626 32% 1,696 5, % Dhupguri Jalpaiguri % % 1,900 1,715 23% 2,513 7, % Falakata Jalpaiguri % % 2,090 1,649 24% 2,732 6, % Kalchini Jalpaiguri % % 304 2, % Madarihat Jalpaiguri % % 30 1,200 19% 375 6, % Mal Jalpaiguri % % % 1,198 4, % Maynaguri Jalpaiguri % % 1,930 1,818 27% 2,545 6, % Metiali Jalpaiguri % % 215 2, % Nagrakata Jalpaiguri % % 111 2, % Cooch Behar I Cooch Behar 5,459 5,762 27% % % 5,643 21, % Dinhata I Cooch Behar 4,946 8,500 72% % % 5,276 11, % Dinhata II Cooch Behar 4,022 4,876 57% % % 4,205 8, % Mathabhanga I Cooch Behar 2,694 2,133 45% 1, % % 4,525 4, % Mathabhanga II Cooch Behar 5,183 4,581 47% % % 5,392 9, % Mekhliganj Cooch Behar 1,000 1,193 27% 3,423 1,590 36% % 4,779 4, % Sitai Cooch Behar 1,412 2,527 67% % % 1,520 3, % Sitalkuchi Cooch Behar 3,056 5,138 35% % % 3,173 14, % Total 29,051 37,268 30% 7,328 5,538 5% 9,083 13,363 11% 46, , % Area 67

80 Application of SWAT and a Groundwater Model for Impact Assessment Table A.4: SWAT calibration steps. calibration step variable analysed 1 Evapotranspiration Simulation time step Average monthly output file considered output. hru 2 Shallow groundwater recharge Average monthly output. hru 3 Shallow groundwater baseflow Average monthly output. hru 4 Streamflows Daily output. rch calibration parameter change Table parameter applied Why? final values sol AWC Increase Increase ETa FRSJ, all soils: 0.25, 0.15 FRMJ, all soils: 0.30, 0.25 Agricultural units, soils group B & C: 0.25, 0.15 sol Sol_Z Increase FRSJ, soils group A: x 2 FRMJ, soils group A: x 1.5 Agricultural units, soils group B & C: Sol_ Z2 = 3000 mm mgt1 CN2 Increase Decrease recharge FRSJ, soils group A: 37 FRSJ, soils B &C: 77 FRMJ, soils group A: 39 FRMJ, soils group B & C: 80 TEAB, soils group B & C: 82 Agricultural units, soils group A: 63 Agricultural units, soils group B & C: 90 Irrigated Monsoon_Rice Summer_Rice, all soils: 95 VIFA, all soils: 93 mgt1 CN2 Option Varies with slope sol Sol_K Decrease Wherever slope > 3%: x 0.1 gw GW_DELAY Decrease 0 days gw ALPHA_BF Decrease Buffer the baseflow gw GWQMIN Decrease 2,150 mm gw RCHRG_DP Decrease 0 sub CH_N(1) Increase Buffer the 0.3 runoff rte CH_N(2) Increase

81 Stockholm Environment Institute Figure A.1: Example of the groundtruthing form (site GT 35) filled by the field assistants Adapted from Cai and Sharma (2010). 69

Impact assessment of agricultural water management interventions in the Jaldhaka watershed. Stockholm Environment Institute, Project Report

Impact assessment of agricultural water management interventions in the Jaldhaka watershed. Stockholm Environment Institute, Project Report Stockholm Environment Institute, Project Report - 2011 Impact assessment of agricultural water management interventions in the Jaldhaka watershed Devaraj de Condappa, Jennie Barron, Sat Kumar Tomer and

More information

Evaluation of Swat for Modelling the Water Balance and Water Yield in Yerrakalva River Basin, A.P. National Institute of Hydrology, Roorkee

Evaluation of Swat for Modelling the Water Balance and Water Yield in Yerrakalva River Basin, A.P. National Institute of Hydrology, Roorkee Evaluation of Swat for Modelling the Water Balance and Water Yield in Yerrakalva River Basin, A.P. By Dr. J.V. Tyagi Dr. Y.R.S. Rao National Institute of Hydrology, Roorkee INTRODUCTION Knowledge of water

More information

Application of a Basin Scale Hydrological Model for Characterizing flow and Drought Trend

Application of a Basin Scale Hydrological Model for Characterizing flow and Drought Trend Application of a Basin Scale Hydrological Model for Characterizing flow and Drought Trend 20 July 2012 International SWAT conference, Delhi INDIA TIPAPORN HOMDEE 1 Ph.D candidate Prof. KOBKIAT PONGPUT

More information

Hamid R. Solaymani. A.K.Gosain

Hamid R. Solaymani. A.K.Gosain Motivation An integrated management plan is required to prevent the negative impacts of climate change on social- economic and environment aspects Each of these segments is expected to be strongly impacted

More information

Assessing the impact of projected climate changes on small coastal basins of the Western US

Assessing the impact of projected climate changes on small coastal basins of the Western US Assessing the impact of projected climate changes on small coastal basins of the Western US William Burke Dr. Darren Ficklin Dept. of Geography, Indiana University Introduction How will climate change

More information

Evaluating the Reduction Effect of Nonpoint Source Pollution Loads from Upland Crop Areas by Rice Straw Covering Using SWAT

Evaluating the Reduction Effect of Nonpoint Source Pollution Loads from Upland Crop Areas by Rice Straw Covering Using SWAT SESSION J2 : Water Resources Applications - I New Delhi, India 2012 International SWAT Conference Evaluating the Reduction Effect of Nonpoint Source Pollution Loads from Upland Crop Areas by Rice Straw

More information

Simulation of stream discharge from an agroforestry catchment in NW Spain using the SWAT model

Simulation of stream discharge from an agroforestry catchment in NW Spain using the SWAT model Simulation of stream discharge from an agroforestry catchment in NW Spain using the SWAT model Rodríguez-Blanco, M.L. 1, 2, Taboada-Castro, M.M. 1, Arias, R. 1, Taboada-Castro, M.T. 1, Nunes, J.P. 2, Rial-Rivas,

More information

EVALUATION OF ArcSWAT MODEL FOR STREAMFLOW SIMULATION

EVALUATION OF ArcSWAT MODEL FOR STREAMFLOW SIMULATION EVALUATION OF ArcSWAT MODEL FOR STREAMFLOW SIMULATION 212 International SWAT Conference By Arbind K. Verma 1 and Madan K.Jha 2 1. Assistant Professor, SASRD, Nagaland University, Ngaland (India) 2. Professor,

More information

Institute of Water and Flood Management, BUET, Dhaka. *Corresponding Author, >

Institute of Water and Flood Management, BUET, Dhaka. *Corresponding Author, > ID: WRE 002 HYDROLOGICAL MODELING FOR THE SEMI UNGAUGED BRAHMAPUTRA RIVER BASIN USING SWAT MODEL S. Paul 1*, A.S. Islam 1, M. A. Hasan 1, G. M.T. Islam 1 & S. K. Bala 1 1 Institute of Water and Flood Management,

More information

Use of a distributed catchment model to assess hydrologic modifications in the Upper Ganges Basin

Use of a distributed catchment model to assess hydrologic modifications in the Upper Ganges Basin River Basin Management VI 177 Use of a distributed catchment model to assess hydrologic modifications in the Upper Ganges Basin L. Bharati 1, V. Smakhtin 2, P. Jayakody 2, N. Kaushal 3 & P. Gurung 1 1

More information

APPLICATION OF SWAT MODEL TO THE STUDY AREA

APPLICATION OF SWAT MODEL TO THE STUDY AREA CHAPTER 5 APPLICATION OF SWAT MODEL TO THE STUDY 5.1 Introduction The Meenachil river basin suffers from water shortage during the six non-monsoon months of the year. The available land and water resources

More information

Malfunctioning of streamgauge stations in the Chanza and Arochete rivers (Huelva, Spain) detected from hydrological modeling with SWAT.

Malfunctioning of streamgauge stations in the Chanza and Arochete rivers (Huelva, Spain) detected from hydrological modeling with SWAT. Malfunctioning of streamgauge stations in the Chanza and Arochete rivers (Huelva, Spain) detected from hydrological modeling with SWAT. L. Galván, M. Olías and A. Van Griensven Introduction The Odiel river

More information

Effects of land use change on the water resources of the Basoda basin using the SWAT model

Effects of land use change on the water resources of the Basoda basin using the SWAT model INDIAN INSTITUTE OF TECHNOLOGY ROORKEE Effects of land use change on the water resources of the Basoda basin using the SWAT model By Santosh S. Palmate* 1 (Ph.D. Student) Paul D. Wagner 2 (Postdoctoral

More information

CHAPTER 7 GROUNDWATER FLOW MODELING

CHAPTER 7 GROUNDWATER FLOW MODELING 148 CHAPTER 7 GROUNDWATER FLOW MODELING 7.1 GENERAL In reality, it is not possible to see into the sub-surface and observe the geological structure and the groundwater flow processes. It is for this reason

More information

Modelling of the Hydrology, Soil Erosion and Sediment transport processes in the Lake Tana Catchments of Blue Nile River Basin, Ethiopia

Modelling of the Hydrology, Soil Erosion and Sediment transport processes in the Lake Tana Catchments of Blue Nile River Basin, Ethiopia Modelling of the Hydrology, Soil Erosion and Sediment transport processes in the Lake Tana Catchments of Blue Nile River Basin, Ethiopia Combining Field Data, Mathematical Models and Geographic Information

More information

Model to assess the impacts of external drivers on the hydrology of the Ganges River Basin

Model to assess the impacts of external drivers on the hydrology of the Ganges River Basin 76 Evolving Water Resources Systems: Understanding, Predicting and Managing Water Society Interactions Proceedings of ICWRS2014, Bologna, Italy, June 2014 (IAHS Publ. 364, 2014). Model to assess the impacts

More information

TECHNICAL NOTE: ESTIMATION OF FRESH WATER INFLOW TO BAYS FROM GAGED AND UNGAGED WATERSHEDS. T. Lee, R. Srinivasan, J. Moon, N.

TECHNICAL NOTE: ESTIMATION OF FRESH WATER INFLOW TO BAYS FROM GAGED AND UNGAGED WATERSHEDS. T. Lee, R. Srinivasan, J. Moon, N. TECHNICAL NOTE: ESTIMATION OF FRESH WATER INFLOW TO BAYS FROM GAGED AND UNGAGED WATERSHEDS T. Lee, R. Srinivasan, J. Moon, N. Omani ABSTRACT. The long term estimation of fresh water inflow to coastal bays

More information

soil losses in a small rainfed catchment with Mediterranean

soil losses in a small rainfed catchment with Mediterranean Using SWAT to predict climate change effects on runoff and soil losses in a small rainfed catchment with Mediterranean climate of NE Spain Ramos MC, Martínez-Casasnovas JA Department of Environment and

More information

Joint Research Centre (JRC)

Joint Research Centre (JRC) Joint Research Centre (JRC) Marco Pastori and Faycal Bouraoui IES - Institute for Environment and Sustainability Ispra - Italy http://ies.jrc.ec.europa.eu/ http://www.jrc.ec.europa.eu/ CONTENT Introduction

More information

Comparative analysis of SWAT model with Coupled SWAT-MODFLOW model for Gibbs Farm Watershed in Georgia

Comparative analysis of SWAT model with Coupled SWAT-MODFLOW model for Gibbs Farm Watershed in Georgia 2018 SWAT INTERNATIONAL CONFERENCE, JAN 10-12, CHENNAI 1 Comparative analysis of SWAT model with Coupled SWAT-MODFLOW model for Gibbs Farm Watershed in Georgia Presented By K.Sangeetha B.Narasimhan D.D.Bosch

More information

Turbidity Monitoring Under Ice Cover in NYC DEP

Turbidity Monitoring Under Ice Cover in NYC DEP Turbidity Monitoring Under Ice Cover in NYC DEP Reducing equifinality by using spatial wetness information and reducing complexity in the SWAT-Hillslope model Linh Hoang 1,2, Elliot M. Schneiderman 2,

More information

A comparison study of multi-gage and single-gage calibration of the SWAT model for runoff simulation in Qingjiang river basin

A comparison study of multi-gage and single-gage calibration of the SWAT model for runoff simulation in Qingjiang river basin A comparison study of multi-gage and single-gage calibration of the SWAT model for runoff simulation in Qingjiang river basin Dan YU, Xiaohua DONG, Lei LI, Sanhong SONG, Zhixiang LV and Ji LIU China Three

More information

Parameter Calibration of SWAT Hydrology and Water Quality Focusing on Long-term Drought Periods

Parameter Calibration of SWAT Hydrology and Water Quality Focusing on Long-term Drought Periods 2017 SWAT June 28-30, 2017 Centrum Wodne SGGW, Warsaw, Poland Theme I3 Environmental Applications Room: Assembly Hall 2 2017 International SWAT Conference Parameter Calibration of SWAT Hydrology and Water

More information

Evaluation of Mixed Forest Evapotranspiration and Soil Moisture using Measured and SWAT Simulated Results in a Hillslope Watershed

Evaluation of Mixed Forest Evapotranspiration and Soil Moisture using Measured and SWAT Simulated Results in a Hillslope Watershed KONKUK UNIVERSITY Evaluation of Mixed Forest Evapotranspiration and Soil Moisture using Measured and SWAT Simulated Results in a Hillslope Watershed 5 August 2010 JOH, Hyung-Kyung Graduate Student LEE,

More information

1 THE USGS MODULAR MODELING SYSTEM MODEL OF THE UPPER COSUMNES RIVER

1 THE USGS MODULAR MODELING SYSTEM MODEL OF THE UPPER COSUMNES RIVER 1 THE USGS MODULAR MODELING SYSTEM MODEL OF THE UPPER COSUMNES RIVER 1.1 Introduction The Hydrologic Model of the Upper Cosumnes River Basin (HMCRB) under the USGS Modular Modeling System (MMS) uses a

More information

IMPACT OF LAND USE/COVER CHANGES ON STREAMFLOW:

IMPACT OF LAND USE/COVER CHANGES ON STREAMFLOW: IMPACT OF LAND USE/COVER CHANGES ON STREAMFLOW: THE CASE OF HARE RIVER WATERSHED, ETHIOPIA Kassa Tadele and Gerd Foerch University of Siegen July 06, 2007 Presentation outline 1. Introduction Study area

More information

Lecture 9A: Drainage Basins

Lecture 9A: Drainage Basins GEOG415 Lecture 9A: Drainage Basins 9-1 Drainage basin (watershed, catchment) -Drains surfacewater to a common outlet Drainage divide - how is it defined? Scale effects? - Represents a hydrologic cycle

More information

Analyzing water resources in a monsoon-driven environment an example from the Indian Western Ghats

Analyzing water resources in a monsoon-driven environment an example from the Indian Western Ghats Analyzing water resources in a monsoon-driven environment an example from the Indian Western Ghats 1, Shamita Kumar 2, Peter Fiener 1 and Karl Schneider 1 1,,, Germany 2 Institute of Environment Education

More information

Regionalization of SWAT Model Parameters for Use in Ungauged Watersheds

Regionalization of SWAT Model Parameters for Use in Ungauged Watersheds Water 21, 2, 849-871; doi:1.339/w24849 OPEN ACCESS water ISSN 273-4441 www.mdpi.com/journal/water Article Regionalization of SWAT Model Parameters for Use in Ungauged Watersheds Margaret W. Gitau 1, *and

More information

UNCERTAINTY ISSUES IN SWAT MODEL CALIBRATION AT CIRASEA WATERSHED, INDONESIA.

UNCERTAINTY ISSUES IN SWAT MODEL CALIBRATION AT CIRASEA WATERSHED, INDONESIA. UNCERTAINTY ISSUES IN SWAT MODEL CALIBRATION AT CIRASEA WATERSHED, INDONESIA. Sri Malahayati Yusuf 1 Kukuh Murtilaksono 2 1 PhD Student of Watershed Management Major, Bogor Agricultural University 2 Department

More information

Lanie A. Alejo & Victor B. Ella* University of the Philippines Los Baños. *presenter

Lanie A. Alejo & Victor B. Ella* University of the Philippines Los Baños. *presenter ASSESSING THE IMPACTS OF CLIMATE CHANGE ON DEPENDABLE FLOW AND POTENTIAL IRRIGABLE AREA USING THE SWAT MODEL: THE CASE OF MAASIN RIVER WATERSHED IN LAGUNA, PHILIPPINES Lanie A. Alejo & Victor B. Ella*

More information

Estimation of Actual Evapotranspiration at Regional Annual scale using SWAT

Estimation of Actual Evapotranspiration at Regional Annual scale using SWAT Estimation of Actual Evapotranspiration at Regional Annual scale using SWAT Azizallah Izady Ph.D student of Water Resources Engineering Department of Water Engineering, Faculty of Agriculture, Ferdowsi

More information

The effects of the short-term Brazilian sugarcane expansion in stream flow: Monte Mor basin case study

The effects of the short-term Brazilian sugarcane expansion in stream flow: Monte Mor basin case study The effects of the short-term Brazilian sugarcane expansion in stream flow: Monte Mor basin case study Hernandes, TAD; Scarpare, FV; Seabra, JEA; Duft, DG; Picoli, MCA; Walter, A Overview 12,000,000 10,000,000

More information

Estimation of transported pollutant load in Ardila catchment using the SWAT model

Estimation of transported pollutant load in Ardila catchment using the SWAT model June 15-17 Estimation of transported pollutant load in Ardila catchment using the SWAT model 1 Engineering Department Polytechnic Institute of Beja 2 Section of Environmental and Energy Technical University

More information

Hydrological Modelling of Narmada basin in Central India using Soil and Water Assessment Tool (SWAT)

Hydrological Modelling of Narmada basin in Central India using Soil and Water Assessment Tool (SWAT) Hydrological Modelling of Narmada basin in Central India using Soil and Water Assessment Tool (SWAT) T. Thomas, N. C. Ghosh, K. P. Sudheer National Institute of Hydrology, Roorkee (A Govt. of India Society

More information

Assessment of StreamFlow Using SWAT Hydrological Model

Assessment of StreamFlow Using SWAT Hydrological Model Assessment of StreamFlow Using SWAT Hydrological Model Sreelakshmi C.M. 1 Dr.K.Varija 2 1 M.Tech student, Applied Mechanics and Hydraulics, NITK, Surathkal,India 2 Associate Professor, Department of Applied

More information

Surface Soil Moisture Assimilation with SWAT

Surface Soil Moisture Assimilation with SWAT Surface Soil Moisture Assimilation with SWAT Eunjin Han, Venatesh Merwade School of Civil Engineering, Purdue University, West Lafayette, IN, USA Gary C. Heathman USDA-ARS, National Soil Erosion Research

More information

Application of the SWAT Model to the Hii River Basin, Shimane Prefecture, Japan

Application of the SWAT Model to the Hii River Basin, Shimane Prefecture, Japan Application of the SWAT Model to the Hii River Basin, Shimane Prefecture, Japan H. Somura, I. Takeda, Y. Mori Shimane University D. Hoffman Blackland Research and Extension Center J. Arnold Grassland Soil

More information

APPLICATION OF THE SWAT (SOIL AND WATER ASSESSMENT TOOL) MODEL IN THE RONNEA CATCHMENT OF SWEDEN

APPLICATION OF THE SWAT (SOIL AND WATER ASSESSMENT TOOL) MODEL IN THE RONNEA CATCHMENT OF SWEDEN Global NEST Journal, Vol 7, No 3, pp 5-57, 5 Copyright 5 Global NEST Printed in Greece. All rights reserved APPLICATION OF THE SWAT (SOIL AND WATER ASSESSMENT TOOL) MODEL IN THE RONNEA CATCHMENT OF SWEDEN

More information

R. Srinivasan, J.H. Jacobs, J.W. Stuth, J. Angerer, R. Kaithio and N. Clarke

R. Srinivasan, J.H. Jacobs, J.W. Stuth, J. Angerer, R. Kaithio and N. Clarke Impacts of reforestation policy and agro-forestry technology on the environment and food security in the Upper Tana river basin of Kenya R. Srinivasan, J.H. Jacobs, J.W. Stuth, J. Angerer, R. Kaithio and

More information

Assessing the benefit of improved precipitation input in SWAT model simulations

Assessing the benefit of improved precipitation input in SWAT model simulations Assessing the benefit of improved precipitation input in SWAT model simulations Ilyas Masih 1,2, Shreedhar Maskey 2, Stefan Uhlenbrook 2,3, Vladimir Smakhtin 1 1 International Water Management Institute

More information

Linking Soil Water and Groundwater Models to Investigate Salinity Management Options

Linking Soil Water and Groundwater Models to Investigate Salinity Management Options Linking Soil Water and Groundwater s to Investigate Salinity Management Options Carl C. Daamen a and Greg P. Hoxley a a Sinclair Kight Merz, P.O.Box 25, Malvern, VIC 3162, Australia. Abstract: Salinisation

More information

July, International SWAT Conference & Workshops

July, International SWAT Conference & Workshops Analysis of the impact of water conservation measures on the hydrological response of a medium-sized watershed July, 212 212 International SWAT Conference & Workshops ANALYSIS OF THE IMPACT OF WATER CONSERVATION

More information

M.L. Kavvas, Z. Q. Chen, M. Anderson, L. Liang, N. Ohara Hydrologic Research Laboratory, Civil and Environmental Engineering, UC Davis

M.L. Kavvas, Z. Q. Chen, M. Anderson, L. Liang, N. Ohara Hydrologic Research Laboratory, Civil and Environmental Engineering, UC Davis Assessment of the Restoration Activities on Water Balance and Water Quality at Last Chance Creek Watershed Using Watershed Environmental Hydrology (WEHY) Model M.L. Kavvas, Z. Q. Chen, M. Anderson, L.

More information

Estimation of Infiltration Parameter for Tehri Garhwal Catchment

Estimation of Infiltration Parameter for Tehri Garhwal Catchment Estimation of Infiltration Parameter for Tehri Garhwal Catchment Ashish Bhatt 1 H L Yadav 2 Dilip Kumar 3 1 UG Student, Department of civil engineering, G B Pant Engineering College, Pauri, UK-246194 2,3

More information

Module 2 Measurement and Processing of Hydrologic Data

Module 2 Measurement and Processing of Hydrologic Data Module 2 Measurement and Processing of Hydrologic Data 2.1 Introduction 2.1.1 Methods of Collection of Hydrologic Data 2.2 Classification of Hydrologic Data 2.2.1 Time-Oriented Data 2.2.2 Space-Oriented

More information

The Fourth Assessment of the Intergovernmental

The Fourth Assessment of the Intergovernmental Hydrologic Characterization of the Koshi Basin and the Impact of Climate Change Luna Bharati, Pabitra Gurung and Priyantha Jayakody Luna Bharati Pabitra Gurung Priyantha Jayakody Abstract: Assessment of

More information

SWAT application to simulate the impact of soil conservation measurements on soil and nutrient losses in a small basin with mechanised vineyards

SWAT application to simulate the impact of soil conservation measurements on soil and nutrient losses in a small basin with mechanised vineyards Soil loss prediction using SWAT in a small ungaged catchment Soil loss prediction using SWAT in a small ungaged catchment with Mediterranean climate and vines as the main land use Ramos MC & Martínez-Casasnovas

More information

FLOOD Analysis And Development of Groundwater Recharge and Discharge Estimate Maps using GIS

FLOOD Analysis And Development of Groundwater Recharge and Discharge Estimate Maps using GIS FLOOD Analysis And Development of Groundwater Recharge and Discharge Estimate Maps using GIS Dr. Ghulam Nabi Assistant Professor Center of Excellence in Water Resources Engineering, g University of Engineering

More information

Sixth Semester B. E. (R)/ First Semester B. E. (PTDP) Civil Engineering Examination

Sixth Semester B. E. (R)/ First Semester B. E. (PTDP) Civil Engineering Examination CAB/2KTF/EET 1221/1413 Sixth Semester B. E. (R)/ First Semester B. E. (PTDP) Civil Engineering Examination Course Code : CV 312 / CV 507 Course Name : Engineering Hydrology Time : 3 Hours ] [ Max. Marks

More information

Field Observations and Model Simulations of an Extreme Drought Event in the Southeast Brazil

Field Observations and Model Simulations of an Extreme Drought Event in the Southeast Brazil Field Observations and Model Simulations of an Extreme Drought Event in the Southeast Brazil Leonardo Domingues, Humberto Rocha, Jonathan Mota da Silva Warsaw, Poland 2017 Jaguari Sub-basin Drainage area:

More information

Calibration and sensitivity analysis of SWAT for a small forested catchment, northcentral

Calibration and sensitivity analysis of SWAT for a small forested catchment, northcentral Calibration and sensitivity analysis of SWAT for a small forested catchment, northcentral Portugal Rial-Rivas, M.E. 1 ; Santos, J. 1 ;Bernard-Jannin L. 1 ; Boulet, A.K. 1 ; Coelho, C.O.A. 1 ; Ferreira,

More information

Study of Hydrology based on Climate Changes Simulation Using SWAT Model At Jatiluhur Reservoir Catchment Area

Study of Hydrology based on Climate Changes Simulation Using SWAT Model At Jatiluhur Reservoir Catchment Area Study of Hydrology based on Climate Changes Simulation Using SWAT Model At Jatiluhur Reservoir Catchment Area Budi Darmawan Supatmanto 1, Sri Malahayati Yusuf 2, Florentinus Heru Widodo 1, Tri Handoko

More information

Calibrating the Soquel-Aptos PRMS Model to Streamflow Data Using PEST

Calibrating the Soquel-Aptos PRMS Model to Streamflow Data Using PEST Calibrating the Soquel-Aptos PRMS Model to Streamflow Data Using PEST Cameron Tana Georgina King HydroMetrics Water Resources Inc. California Water Environmental and Modeling Forum 2015 Annual Meeting

More information

Flood forecasting model based on geographical information system

Flood forecasting model based on geographical information system 192 Remote Sensing and GIS for Hydrology and Water Resources (IAHS Publ. 368, 2015) (Proceedings RSHS14 and ICGRHWE14, Guangzhou, China, August 2014). Flood forecasting model based on geographical information

More information

The Impact of Climate Change on a Humid, Equatorial Catchment in Uganda.

The Impact of Climate Change on a Humid, Equatorial Catchment in Uganda. The Impact of Climate Change on a Humid, Equatorial Catchment in Uganda. Lucinda Mileham, Dr Richard Taylor, Dr Martin Todd Department of Geography University College London Changing Climate Africa has

More information

Development of Urban Modeling Tools in SWAT

Development of Urban Modeling Tools in SWAT 29 International SWAT conference August 5-7 Boulder, CO Development of Urban Modeling Tools in SWAT J. Jeong, N. Kannan, J. G. Arnold, R. Glick, L. Gosselink, R. Srinivasan LOGO Tasks in urban SWAT project

More information

What is runoff? Runoff. Runoff is often defined as the portion of rainfall, that runs over and under the soil surface toward the stream

What is runoff? Runoff. Runoff is often defined as the portion of rainfall, that runs over and under the soil surface toward the stream What is runoff? Runoff Runoff is often defined as the portion of rainfall, that runs over and under the soil surface toward the stream 1 COMPONENTS OF Runoff or STREAM FLOW 2 Cont. The types of runoff

More information

Flood forecasting model based on geographical information system

Flood forecasting model based on geographical information system doi:10.5194/piahs-368-192-2015 192 Remote Sensing and GIS for Hydrology and Water Resources (IAHS Publ. 368, 2015) (Proceedings RSHS14 and ICGRHWE14, Guangzhou, China, August 2014). Flood forecasting model

More information

EFFECTS OF IRRIGATION WITHDRAWAL AND CLIMATE CHANGE ON GROUNDWATER DYNAMICS IN A SEMI-ARID INDIAN WATERSHED

EFFECTS OF IRRIGATION WITHDRAWAL AND CLIMATE CHANGE ON GROUNDWATER DYNAMICS IN A SEMI-ARID INDIAN WATERSHED EFFECTS OF IRRIGATION WITHDRAWAL AND CLIMATE CHANGE ON GROUNDWATER DYNAMICS IN A SEMI-ARID INDIAN WATERSHED Rajendra Sishodia Sanjay Shukla, UF Suhas Wani, ICRISAT Jim Jones, UF Wendy Graham, UF GROUNDWATER

More information

Management Scenario for the Critical Subwatersheds of Small Agricultural Watershed using SWAT model and GIS Technique

Management Scenario for the Critical Subwatersheds of Small Agricultural Watershed using SWAT model and GIS Technique Management Scenario for the Critical Subwatersheds of Small Agricultural Watershed using SWAT model and GIS Technique Authors: M.P. Tripathi, N. Agrawal and M.K. Verma Indira Gandhi Krishi Vishwavidyalaya

More information

CHANGES ON FLOOD CHARACTERISTICS DUE TO LAND USE CHANGES IN A RIVER BASIN

CHANGES ON FLOOD CHARACTERISTICS DUE TO LAND USE CHANGES IN A RIVER BASIN U.S.- Italy Research Workshop on the Hydrometeorology, Impacts, and Management of Extreme Floods Perugia (Italy), November 1995 CHANGES ON FLOOD CHARACTERISTICS DUE TO LAND USE CHANGES IN A RIVER BASIN

More information

Hydrological And Water Quality Modeling For Alternative Scenarios In A Semi-arid Catchment

Hydrological And Water Quality Modeling For Alternative Scenarios In A Semi-arid Catchment Hydrological And Water Quality Modeling For Alternative Scenarios In A Semi-arid Catchment AZIZ ABOUABDILLAH, ANTONIO LO PORTO METIER Final Conference: Brussels, Belgium-4-6 November 2009 Outline Problem

More information

M.L. Kavvas, Z. Q. Chen, M. Anderson, L. Liang, N. Ohara Hydrologic Research Laboratory, Civil and Environmental Engineering, UC Davis

M.L. Kavvas, Z. Q. Chen, M. Anderson, L. Liang, N. Ohara Hydrologic Research Laboratory, Civil and Environmental Engineering, UC Davis Assessment of the Restoration Activities on Water Balance and Water Quality at Last Chance Creek Watershed Using Watershed Environmental Hydrology (WEHY) Model M.L. Kavvas, Z. Q. Chen, M. Anderson, L.

More information

Estimating catchment sediment yield, reservoir sedimentation and reservoir effective life using SWAT Model

Estimating catchment sediment yield, reservoir sedimentation and reservoir effective life using SWAT Model Estimating catchment sediment yield, reservoir sedimentation and reservoir effective life using SWAT Model Sanjeet Kumar a *, Ashok Mishra a, N.S. Raghuwanshi a a Department of Agricultural and Food Engineering.

More information

Hydrogeological Investigation and Analyzing Groundwater Scenario in Haringhata Block, West Bengal

Hydrogeological Investigation and Analyzing Groundwater Scenario in Haringhata Block, West Bengal Hydrogeological Investigation and Analyzing Groundwater Scenario in Haringhata Block, West Bengal Alivia Chowdhury Department of Soil and Water Engineering Faculty of Agricultural Engineering Bidhan Chandra

More information

Reservoir on the Rio Boba

Reservoir on the Rio Boba Reservoir on the Rio Boba Michael J. Burns II Guillermo Bustamante J. James Peterson Executive Summary The National Institute of Water Resources in the Dominican Republic (INDRHI) plans to construct a

More information

Improved Lower Mekong River Basin Hydrological Decision Making Using NASA Satellite-based Earth Observation Systems

Improved Lower Mekong River Basin Hydrological Decision Making Using NASA Satellite-based Earth Observation Systems Improved Lower Mekong River Basin Hydrological Decision Making Using NASA Satellite-based Earth Observation Systems Ibrahim Mohammed Ibrahim.mohammed@nasa.gov John Bolten R. Srinivasan Venkat Lakshmi AGU

More information

Event and Continuous Hydrological Modeling with HEC- HMS: A Review Study

Event and Continuous Hydrological Modeling with HEC- HMS: A Review Study Event and Continuous Hydrological Modeling with HEC- HMS: A Review Study Sonu Duhan *, Mohit Kumar # * M.E (Water Resources Engineering) Civil Engineering Student, PEC University Of Technology, Chandigarh,

More information

An Assessment of Climate Change Impacts on Streamflows in the Musi Catchment, India

An Assessment of Climate Change Impacts on Streamflows in the Musi Catchment, India 20th International Congress on Modelling and Simulation, Adelaide, Australia, 1 6 December 2013 www.mssanz.org.au/modsim2013 An Assessment of Climate Change Impacts on Streamflows in the Musi Catchment,

More information

Chapter 7 - Monitoring Groundwater Resources

Chapter 7 - Monitoring Groundwater Resources Chapter 7 - Monitoring Groundwater Resources Introduction Because of its hidden nature, virtually everything that is known about Marlborough s aquifers comes from indirect observations made at wells. The

More information

Modeling Status Update Review of Data and Documents

Modeling Status Update Review of Data and Documents SUSTAINABLE, JUST & PRODUCTIVE WATER RESOURCES DEVELOPMENT IN WESTERN NEPAL UNDER CURRENT & FUTURE CONDITIONS (DIGO JAL BIKAS DJB) Modeling Status Update Review of Data and Documents Tuesday, 1 ST August,

More information

Topography and the Spatial Distribution of Groundwater Recharge and Evapotranspiration:

Topography and the Spatial Distribution of Groundwater Recharge and Evapotranspiration: Topography and the Spatial Distribution of Groundwater Recharge and Evapotranspiration: A Need to Revisit Distributed Water Budget Analysis when Assessing Impacts to Ecological Systems. By M.A. Marchildon,

More information

Scale Effects in Large Scale Watershed Modeling

Scale Effects in Large Scale Watershed Modeling Scale Effects in Large Scale Watershed Modeling Mustafa M. Aral and Orhan Gunduz Multimedia Environmental Simulations Laboratory School of Civil and Environmental Engineering Georgia Institute of Technology

More information

AN INTEGRATED FRAMEWORK FOR EFFECTIVE ADAPTATION TO CLIMATE CHANGE IMPACTS ON WATER RESOURCES

AN INTEGRATED FRAMEWORK FOR EFFECTIVE ADAPTATION TO CLIMATE CHANGE IMPACTS ON WATER RESOURCES AN INTEGRATED FRAMEWORK FOR EFFECTIVE ADAPTATION TO CLIMATE CHANGE IMPACTS ON WATER RESOURCES A. K. Gosain, Professor & Head Civil Engineering Department Indian Institute of Technology Delhi: gosain@civil.iitd.ac.in

More information

Effects of climate change on streamflow in Kon Ha Thanh river watershed, Vietnam

Effects of climate change on streamflow in Kon Ha Thanh river watershed, Vietnam EPiC Series in Engineering Volume 3, 2018, Pages 2233 2240 Engineering HIC 2018. 13th International Conference on Hydroinformatics Effects of climate change on streamflow in Kon Ha Thanh river watershed,

More information

Assessment of sediment and carbon flux in a tropical watershed: the Red River study case (China and Vietnam)

Assessment of sediment and carbon flux in a tropical watershed: the Red River study case (China and Vietnam) Assessment of sediment and carbon flux in a tropical watershed: the Red River study case (China and Vietnam) Xi WEI Supervisors: Sabine SAUVAGE, Jose-Miguel SANCHEZ-PEREZ, Thi Phuong Quynh LE 2 th September

More information

Proposed Project. Integrated Water Resources Management Using Remote Sensing Data in Upper Indus Basin

Proposed Project. Integrated Water Resources Management Using Remote Sensing Data in Upper Indus Basin Proposed Project Integrated Water Resources Management Using Remote Sensing Data in Upper Indus Basin Background Snowmelt contributes more than 6% of water resources of Upper Indus Basin Most of the moisture

More information

Hydrologic Engineering Center Hydrologic Modeling System (HEC-HMS) Sunil KUMAR Director, National Water Academy

Hydrologic Engineering Center Hydrologic Modeling System (HEC-HMS) Sunil KUMAR Director, National Water Academy Hydrologic Engineering Center Hydrologic Modeling System (HEC-HMS) Sunil KUMAR Director, National Water Academy 22 April 2015 NWA, Pune Exercise Objective: To determine hydrological Response of the given

More information

Purpose of the web- based rainwater harvesting toolkit. What does the Rainwater Harvesting Toolkit DO and what does it NOT DO?

Purpose of the web- based rainwater harvesting toolkit. What does the Rainwater Harvesting Toolkit DO and what does it NOT DO? User s Manual Purpose of the web- based rainwater harvesting toolkit The Rainwater Harvesting Toolkit was developed to provide quantitative, locally specific information about water yield of rainwater

More information

CONTINUOUS RAINFALL-RUN OFF SIMULATION USING SMA ALGORITHM

CONTINUOUS RAINFALL-RUN OFF SIMULATION USING SMA ALGORITHM CONTINUOUS RAINFALL-RUN OFF SIMULATION USING SMA ALGORITHM INTRODUCTION Dr. R N Sankhua Director, NWA, CWC, Pune In this continuous rainfall-runoff simulation, we will perform a continuous or long-term

More information

Integrating wetlands and riparian zones in regional hydrological modelling

Integrating wetlands and riparian zones in regional hydrological modelling Integrating wetlands and riparian zones in regional hydrological modelling Fred Hattermann, Valentina Krysanova & Joachim Post Potsdam Institute for Climate Impact Research Outline Introduction Model concept:

More information

IJSER. within the watershed during a specific period. It is constructed

IJSER. within the watershed during a specific period. It is constructed International Journal of Scientific & Engineering Research, Volume 5, Issue 7, July-014 ISSN 9-5518 306 within the watershed during a specific period. It is constructed I. INTRODUCTION In many instances,

More information

The Texas A&M University and U.S. Bureau of Reclamation Hydrologic Modeling Inventory (HMI) Questionnaire

The Texas A&M University and U.S. Bureau of Reclamation Hydrologic Modeling Inventory (HMI) Questionnaire The Texas A&M University and U.S. Bureau of Reclamation Hydrologic Modeling Inventory (HMI) Questionnaire May 4, 2010 Name of Model, Date, Version Number Dynamic Watershed Simulation Model (DWSM) 2002

More information

Solutions towards hydrological challenges in Africa in support of hydropower developments Ms. Catherine Blersch, Civil Engineer, Aurecon, South

Solutions towards hydrological challenges in Africa in support of hydropower developments Ms. Catherine Blersch, Civil Engineer, Aurecon, South Solutions towards hydrological challenges in Africa in support of hydropower developments Ms. Catherine Blersch, Civil Engineer, Aurecon, South Africa Dr Verno Jonker, Civil Engineer, Aurecon, South Africa

More information

ICELANDIC RIVER / WASHOW BAY CREEK INTEGRATED WATERSHED MANAGEMENT PLAN STATE OF THE WATERSHED REPORT CONTRIBUTION SURFACE WATER HYDROLOGY REPORT

ICELANDIC RIVER / WASHOW BAY CREEK INTEGRATED WATERSHED MANAGEMENT PLAN STATE OF THE WATERSHED REPORT CONTRIBUTION SURFACE WATER HYDROLOGY REPORT ICELANDIC RIVER / WASHOW BAY CREEK INTEGRATED WATERSHED MANAGEMENT PLAN STATE OF THE WATERSHED REPORT CONTRIBUTION SURFACE WATER HYDROLOGY REPORT Disclaimer: The hydrologic conditions presented in this

More information

Dynamic groundwater-river interaction model for planning water allocation in a narrow valley aquifer system of the Upper Motueka catchment

Dynamic groundwater-river interaction model for planning water allocation in a narrow valley aquifer system of the Upper Motueka catchment Dynamic groundwater-river interaction model for planning water allocation in a narrow valley aquifer system of the Upper Motueka catchment Timothy Hong t.hong@gns.cri.nz Gilles Minni g.minni@gns.cri.nz

More information

Anthropogenic and climate change contributions to uncertainties in hydrological modeling of sustainable water supply

Anthropogenic and climate change contributions to uncertainties in hydrological modeling of sustainable water supply Anthropogenic and climate change contributions to uncertainties in hydrological modeling of sustainable water supply Roman Corobov The Republic of Moldova Several words about Moldova The Republic of Moldova

More information

Columbia, Missouri. Contents

Columbia, Missouri. Contents Supplemental Material to Accompany: Long-term Agro-ecosystem Research in the Central Mississippi River Basin, USA - SWAT Simulation of Flow and Water Quality in the Goodwater Creek Experimental Watershed

More information

Impact analysis of the decline of agricultural land-use on flood risk and material flux in hilly and mountainous watersheds

Impact analysis of the decline of agricultural land-use on flood risk and material flux in hilly and mountainous watersheds Proc. IAHS, 370, 39 44, 2015 doi:10.5194/piahs-370-39-2015 Author(s) 2015. CC Attribution 3.0 License. Impact analysis of the decline of agricultural land-use on flood risk and material flux in hilly and

More information

Comparison of Recharge Estimation Methods Used in Minnesota

Comparison of Recharge Estimation Methods Used in Minnesota Comparison of Recharge Estimation Methods Used in Minnesota by Geoffrey Delin, Richard Healy, David Lorenz, and John Nimmo Minnesota Ground Water Association Spring Conference Methods for Solving Complex

More information

Measuring discharge. Climatological and hydrological field work

Measuring discharge. Climatological and hydrological field work Measuring discharge Climatological and hydrological field work 1. Background Discharge (or surface runoff Q s) refers to the horizontal water flow occurring at the surface in rivers and streams. It does

More information

5.5 Improving Water Use Efficiency of Irrigated Crops in the North China Plain Measurements and Modelling

5.5 Improving Water Use Efficiency of Irrigated Crops in the North China Plain Measurements and Modelling 183 5.5 Improving Water Use Efficiency of Irrigated Crops in the North China Plain Measurements and Modelling H.X. Wang, L. Zhang, W.R. Dawes, C.M. Liu Abstract High crop productivity in the North China

More information

RAINFALL RUN-OFF AND BASEFLOW ESTIMATION

RAINFALL RUN-OFF AND BASEFLOW ESTIMATION CHAPTER 2 RAINFALL RUN-OFF AND BASEFLOW ESTIMATION 2.1 Introduction The west coast of India receives abundant rainfall from the southwest monsoon. The Western Ghats escarpment (Sahyadri mountain range)

More information

ACRU HYDROLOGICAL MODELLING OF THE MUPFURE CATCHMENT

ACRU HYDROLOGICAL MODELLING OF THE MUPFURE CATCHMENT ACRU HYDROLOGICAL MODELLING OF THE MUPFURE CATCHMENT Table of Contents: 1 INTRODUCTION... 1 2 CONFIGURATION OF ACRU... 2 2.1 RAINFALL DATA... 4 2.2 SOILS... 7 2.3 LAND COVER INFORMATION... 9 2.4 STREAM

More information

Flood Modelling and Water Harvesting Plan for Paravanar Basin

Flood Modelling and Water Harvesting Plan for Paravanar Basin International Journal of ChemTech Research CODEN (USA): IJCRGG, ISSN: 0974-4290, ISSN(Online):2455-9555 Vol.10 No.14, pp 01-08, 2017 Flood Modelling and Water Harvesting Plan for Paravanar Basin Dhinesh

More information

Development of a GIS Tool for Rainfall-Runoff Estimation

Development of a GIS Tool for Rainfall-Runoff Estimation Development of a GIS Tool for Rainfall-Runoff Estimation Ashraf M. Elmoustafa * M. E. Shalaby Ahmed A. Hassan A.H. El-Nahry Irrigation and Hydraulics Department, Ain Shams University, Egypt NARSS, Egypt

More information

TD 603. Water Resources Milind Sohoni sohoni/ Lecture 2: Water cycle, stocks and flows. () July 28, / 30

TD 603. Water Resources Milind Sohoni  sohoni/ Lecture 2: Water cycle, stocks and flows. () July 28, / 30 TD 603 Water Resources Milind Sohoni www.cse.iitb.ac.in/ sohoni/ Lecture 2: Water cycle, stocks and flows () July 28, 2013 1 / 30 The basic movement of water source: USGS. () July 28, 2013 2 / 30 The basic

More information

CLIMATE CHANGE IMPACT AND VULNERABILITY OF SURFACE WATER. Prof. A. K. Gosian Indian Institute of Technology, Delhi

CLIMATE CHANGE IMPACT AND VULNERABILITY OF SURFACE WATER. Prof. A. K. Gosian Indian Institute of Technology, Delhi CLIMATE CHANGE IMPACT AND VULNERABILITY OF SURFACE WATER Prof. A. K. Gosian Indian Institute of Technology, Delhi NATCOM MoEF (IIT Delhi) Climate Change and its Impact on Water Resources of India Tools

More information