HYDROLOGICAL DROUGHT INDEX AT THE IRRIGATION AREA IN PEMALI-COMAL RIVER BASIN Waluyo Hatmoko 1, R. Wahyudi Triweko 2 and Iwan K. Hadihardaja 3 ABSTRACT Drought index is an important tool to detect the onset, ending, and severity of a drought event. Until recently there is no hydrological drought index has been accepted internationally. In search of promising hydrological drought index, this paper analyses the performance of several hydrological drought index at five irrigation weirs in the Pemali-Comal River Basin, Indonesia. The performance is measured by the correlation between the streamflow drought index and the irrigation areal affected by drought. This paper examines different kinds of hydrological drought index based on the Theory of Run. The combination of the index consists of: 1) moving averaged data of 1, 3, 6, and 12 months; 2) Normal, Log-Normal and Gamma statistical distribution; 3) fixed and seasonal threshold; and 4) threshold level of mean flow and dependable flow. The performance of the hydrological drought index is examined through their correlation with the historical data of irrigation area affected by drought. It is concluded that good correlation between index and the impact data generally: a) Index with 3 months moving average; b) Log-Normal distribution; c) Fixed threshold level; and threshold level of mean flow, except for original data of one month the dependable flow is better. Finally, the decision making impact of selecting a particular type of drought index from among others is not significant, for all of the index are be able to identify the drought events. Keywords: drought, drought index, hydrological drought index, irrigation 1. INTRODUCTION 1.1 Background Drought is a natural disaster that threatens life and cause enormous damage. Report of the Intergovernmental Panel on Climate Change (IPCC, 2007) states that the world is more vulnerable to drought in the next 25 years, and climate projections indicate that this will get worse in the future. Droughts differ from other natural hazards (Wilhite, 2000): 1) the onset and the end of a drought are difficult to determine, the impacts of a drought increase slowly, often accumulate over a considerable period and is often referred to as a creeping phenomenon; 2) It is difficult to have a universal definition of drought; 3) Drought impacts are non-structural and spread over large geographical areas than damages that may result from other natural hazards e.g. flood and earthquakes, and it seldom results in structural damage. The quantification of the impact is more difficult than for other natural hazards; 4) human activities can directly trigger a drought, such as excessive irrigation, deforestation, and overexploiting available water (Mishra and Singh, 2010). 1 Research Professor, Research Center for Water Resources, Ministry of Public Works and Housing, Jalan Juanda 193, Bandung 40135, Indonesia. E-mail: whatmoko@yahoo.com 2 Professor in Civil Engineering, Parahyangan Catholic University, Jalan Ciumbeuluit, Bandung, Indonesia 3 Professor in Civil Engineering, Bandung Institute of Technology, Jalan Ganesa 10, Bandung, Indonesia 1
Drought is generally classified into meteorological drought, hydrological drought, agricultural drought, and socio-economic drought. The origin of drought is meteorological drought of extreme low rainfall. Hydrological drought related to the river discharge, lake, reservoir, and ground water level. Agricultural drought refers to the soil moisture in agricultural area. Hydrological drought is related to a period with inadequate surface and subsurface water resources for established water uses of a given water resources management system. Panu and Sharma (2002) defines hydrological drought as periods in which the water in the rivers, lakes, and aquifers are below the average conditions. Hydrological drought is closely related with the water demand and water allocation management. Drought Index is a useful tool to indicate the onset and ending of a drought event, as well as drought duration, severity and intensity. A hydrological drought index is an essential tool in water allocation supporting the decision on water allocation. Unlike meteorological drought that already widely accepted of using Standardized Precipitation Index (SPI), until recently there is no hydrological drought index has been accepted internationally. In search of promising hydrological drought index, this paper analyzes the performance of several hydrological drought index at some irrigation weirs in the Pemali-Comal River Basin, Indonesia. 1.2 Hydrological Drought Index To enable identification of drought onset, duration and severity, a theory of run approach developed by Yevjevich (1967) is applied. The drought indicator such as rainfall or discharge time-series Xt is truncated to a certain threshold level X0 that usually is arithmetic mean, median, any percentile, or any level, fixed or variable. Drought is defined at the time when the indicator value is below threshold level, or if the truncated value is negative, as presented in Figure 1. Figure 1. The Theory of Run applied to the truncated series of drought indicator (Mishra and Singh, 2010) 2
The duration of a drought event is length of time between successive crosses of X0 when the truncated value is negative. Drought severity is the cumulative deviation from X0, and drought intensity or magnitude is the average deviation from X0. Drought severity can also be computed as a multiplication between drought intensity and its duration. The illustration in Error! Reference source not found. shows three drought events. Drought event 1 is the drought with the highest severity; event 2 is drought with the longest duration; and event 3 represents drought with the highest intensity. This theory of run approach is applied in SPI meteorological drought index, and its application to river discharges data is named as Standardized Runoff Index or SRI (Shukla and Wood, 2008). The truncation threshold level can be either the average flow or dependable flow (Edossa et al., 2010). The hydrological index SRI can be constructed using four combinations of: 1) moving averaging flow data; 2) assumption of the statistical distribution (Normal, Log-Normal and Gamma); 3) the truncation method of fixed and seasonal level; and 4) threshold level of mean flow and dependable flow. The different combination of hydrological drought index based on the Theory of Run is presented in Error! Reference source not found.. Moving Average 1 month 3 months 6 months 12 months Statistical Distribution Normal Log-Normal Gamma Fixed Threshold type Seasonal Figure 2. Mean Flow 1.3 Research Questions Threshold Level Dependable Flow Different combination of hydrological drought index This paper is expected to answer the following questions: 1) Which SRI is having the highest correlation with the drought impact data? 1) raw data or moving averaged data of 3, 6, or 12 months; 2) Fixed or seasonal threshold? 3) Threshold of average flow or dependable flow 80%? 2. METHODS 2.1 Case study: Pemali-Comal River Basin The study area is Pemali-Comal River Basin at the Northern part of Central Java, Indonesia (Figure 1). Five irrigation areas are selected in the river basins: Notog, Sukowati, Kaliwadas, Asemsiketek, and Kramat. The location of the five irrigation areas is presented in Figure 2. 3
2.2 Hydrological Drought Index From monthly time-series data of the weirs covering period from the year of 1991 until 2013, the hydrological drought index SRI are computed using four combinations of: 1) moving averaging flow data; 2) assumption of the statistical distribution (Normal, Log- Normal and Gamma); 3) the truncation method of fixed and seasonal level; and 4) threshold level of mean flow and dependable flow. Figure 1 Location of the study area, Pemali Comal River Basin in Indonesia CIREBON KUNINGAN KOTA TEGAL BREBES TEGAL Notog Notog KOTA PEKALONGAN Sungapan Kramat Asemsiketek PEMALANG Sokawati PEKALONGAN Kaliwadas BATANG PURBALINGGA Figure 2 Location of the five weirs in Pemali-Comal BANJARNEGARARiver Basin CIAMIS WONOSOBO The irrigation weirs with their catchment areas, and the districts location of the CILACAP BANYUMAS irrigation area are presented. Table 1. Description on irrigation diversion structures Diversion structures River Catchment Area (km 2 ) District location of irrigation area Notog Weir Kali Pemali 1,276 Brebes Sukowati Weir Kali Comal 527 Pemalang Kaliwadas Weir Kali Genteng 225 Pekalongan Asemsiketek Weir Kali Sengkarang 121 Pekalongan Kramat Weir Kali Sambong 120 Batang 4
SRI 2 nd World Irrigation Forum (WIF2) 3. RESULTS AND DISCUSSION 3.1 Hydrological Drought Index For each of the 5 irrigation weirs, the hydrological index is computed, as combination of: 1) assumed statistical distribution, Normal, Log Normal and Gamma distribution; 2) fixed or seasonal threshold; 3) threshold of mean or dependable flow 80% discharge; 4) moving average of 1, 3, 6, and 12 months. Example of Hydrological Drought Index assuming Normal Distribution with fixed threshold of mean discharge for 1, 3, 6, and 12 months moving average is presented in Figure 5. Longer moving averaged data stabilize the graph and eliminating seasonal fluctuation. 4.00 3.00 2.00 1.00 0.00 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013-1.00-2.00 Figure 5. 1 month 3 months 6 months 12 months Hydrological Drought Index SRI ND.F.M at Notog Weir for 1, 3, 6, and 12 months 3.2 Impact of Drought in Irrigation Area Drought impact data available in the study area is the annual time-series of areal affected by drought in the district corresponding to the rice field in the irrigation area commanded by the irrigation structures. This drought impact data reflects the severity of the drought events, and its coefficient of correlation with the index represent the performance of the index, how good the index describes the drought events. 3.3 Correlation between Drought Index and Drought Impact Performance of hydrological drought index is reflected in their coefficient of correlation (r) with the data on rice field affected by drought. Table 1, 2 and 3 present the r values from different possible combination of SRI with the drought impact data, for original data, moving average 3 months and 6 months. 5
Table 1. Correlation between Index and Drought Impact Data for 1 month (no moving average) data Methods Distribution Threshold Notog Sokawati Kaliwadas Asemsiketek Sungapan Kramat Average ND.F.M Normal Fixed Mean 62% 56% 51% 48% 56% 51% 54% ND.F.80 Normal Fixed Q80% 51% 56% 61% 73% 44% 57% 57% LN.F.M LogNormal Fixed Mean 52% 33% 50% 78% 42% 68% 54% LN.F.80 LogNormal Fixed Q80% 53% 14% 36% 87% 19% 76% 47% GM.F.M Gamma Fixed Mean 57% 46% 54% 68% 49% 62% 56% GM.F.80 Gamma Fixed Q80% 53% 33% 53% 81% 23% 67% 52% ND.S.M Normal Seasonal Mean 85% 60% 68% 57% 59% 65% 66% ND.S.80 Normal Seasonal Q80% 91% 42% -4% 79% 59% 49% 53% LN.S.M LogNormal Seasonal Mean 87% 46% 67% 81% 61% 79% 70% LN.S.80 LogNormal Seasonal Q80% 91% 26% 10% 88% 26% 84% 54% GM.S.M Gamma Seasonal Mean 86% 57% 60% 62% 51% 62% 63% GM.S.80 Gamma Seasonal Q80% 88% 27% 1% 62% 33% 62% 45% Table 2. Correlation between Index and Drought Impact Data for 3 months data Methods Distribution Threshold Notog Sokawati Kaliwadas Asemsiketek Sungapan Kramat Average ND.F.M Normal Fixed Mean 71% 55% 53% 46% 58% 50% 56% ND.F.80 Normal Fixed Q80% 56% 61% 62% 67% 56% 61% 61% LN.F.M LogNormal Fixed Mean 64% 54% 59% 72% 57% 70% 63% LN.F.80 LogNormal Fixed Q80% 60% 43% 63% 85% 53% 76% 63% GM.F.M Gamma Fixed Mean 68% 57% 57% 63% 57% 64% 61% GM.F.80 Gamma Fixed Q80% 58% 54% 63% 76% 46% 69% 61% ND.S.M Normal Seasonal Mean 92% 60% 62% 45% 64% 50% 62% ND.S.80 Normal Seasonal Q80% 92% 42% -6% 51% 61% 16% 43% LN.S.M LogNormal Seasonal Mean 93% 61% 70% 67% 66% 68% 71% LN.S.80 LogNormal Seasonal Q80% 92% 43% 41% 74% 42% 73% 61% GM.S.M Gamma Seasonal Mean 92% 65% 67% 55% 50% 55% 64% GM.S.80 Gamma Seasonal Q80% 91% 10% 3% 35% 32% 37% 35% Table 3. Correlation between Index and Drought Impact Data for 6 months data Methods Distribution Threshold Notog Sokawati Kaliwadas Asemsiketek Sungapan Kramat Average ND.F.M Normal Fixed Mean 88% 57% 60% 34% 65% 45% 58% ND.F.80 Normal Fixed Q80% 87% 65% 67% 59% 66% 64% 68% LN.F.M LogNormal Fixed Mean 89% 63% 66% 58% 66% 68% 68% LN.F.80 LogNormal Fixed Q80% 85% 64% 69% 74% 63% 76% 72% GM.F.M Gamma Fixed Mean 89% 63% 64% 50% 66% 63% 66% GM.F.80 Gamma Fixed Q80% 86% 63% 68% 67% 58% 70% 69% ND.S.M Normal Seasonal Mean 92% 48% 30% 20% 58% 18% 44% ND.S.80 Normal Seasonal Q80% 92% 33% -10% 18% 49% -10% 29% LN.S.M LogNormal Seasonal Mean 95% 49% 36% 30% 60% 32% 50% LN.S.80 LogNormal Seasonal Q80% 93% 37% 21% 37% 44% 24% 42% GM.S.M Gamma Seasonal Mean 94% 48% 40% 23% 42% 23% 45% GM.S.80 Gamma Seasonal Q80% 93% -7% -8% 3% 21% 5% 18% It is interesting to note that Notog Weir having relatively large basin is having good correlation for 6 months moving averaging, while Kramat Weir having small basin is good at 1 month. This phenomenon might be related to the propagation of water in the basin. 6
Table 4. Average Index Summary the Correlation between Index and Drought Impact Data Methods Distribution Threshold 1 month 3 months 6 months ND.F.M Normal Fixed Mean 54% 56% 58% ND.F.80 Normal Fixed Q80% 57% 61% 68% LN.F.M LogNormal Fixed Mean 54% 63% 68% LN.F.80 LogNormal Fixed Q80% 47% 63% 72% GM.F.M Gamma Fixed Mean 56% 61% 66% GM.F.80 Gamma Fixed Q80% 52% 61% 69% ND.S.M Normal Seasonal Mean 66% 62% 44% ND.S.80 Normal Seasonal Q80% 53% 43% 29% LN.S.M LogNormal Seasonal Mean 70% 71% 50% LN.S.80 LogNormal Seasonal Q80% 54% 61% 42% GM.S.M Gamma Seasonal Mean 63% 64% 45% GM.S.80 Gamma Seasonal Q80% 45% 35% 18% Average correlation of the hydrological drought index of irrigation weirs for moving averaged period of 1, 3, and 6 months are summarized in Table 4. It shows that for fixed threshold, the 6 months moving average gives the highest correlation with drought impact data, while for seasonal threshold the best results are in one and three months. Table 5. Average Parameter Summary of the Correlation between Index and Drought Impact Data Index Paramater 1 month 3 months 6 months Fixed Threshold 56% 61% 67% Seasonal Threshold 53% 56% 38% Normal Distribution 59% 55% 50% Log Normal Distribution 57% 64% 58% Gamma Distribution 56% 55% 49% Mean Threshold 54% 63% 55% Q80% Threshold 60% 54% 49% Overall Average 56% 58% 52% The average parameter summary of the correlation between hydrological drought index and areal affected by drought is presented in table 5. The best result is achieved for fixed threshold in 6 month moving average. However, good correlation generally appears for 3 months moving average. Concerning statistical distribution, Log-Normal distribution significantly higher than Normal and Gamma distributions. Fixed threshold level have better correlation than seasonal, and mean flow threshold better than dependable flow of 80% except for original data of one month. 7
SRI 2 nd World Irrigation Forum (WIF2) The most suitable hydrological drought index identified is LN.F.80(6), a Log-Normal distribution on moving average data in 6 months, using a fixed threshold of dependable flow 80%. This result is related with the fact that water supply for irrigation is designed based on water availability of dependable 80%. Fixed threshold is generally good for longer moving averaged data while seasonal threshold will be more appropriate with original data and shorter moving average of 3 months. 3.4 Decision making impact on selecting type of index To analyse the decision making impact on selecting a particular type of hydrological drought index described in this paper, Figure 6 clearly shows that the four different index consistently define the drought events in the year of 1991, 1994, and 1997 in the same order of severity magnitude. All of the four indices are good, and any one of them may be adopted as drought index in the region. 0.00 1991 1992 1993 1994 1995 1996 1997 1998-0.50-1.00-1.50-2.00-2.50-3.00-3.50-4.00 ND.F.M ND.S.M LN.F.M LN.S.M Figure 6. Different kind of index describing drought events in 1991, 1994, and 1997 4. CONCLUDING REMARKS This paper describes different kinds of hydrological drought index SRI based on the Theory of Run. The combination of the index consists of: 1) moving averaged data of 1, 3, 6, and 12 months; 2) Normal, Log-Normal and Gamma statistical distribution; 3) fixed and seasonal threshold; and 4) threshold level of mean flow and dependable flow. The performance of the hydrological drought index is examined through their correlation with the historical data of irrigation area affected by drought. It is concluded from the case study in Pemali-Comal River Basin, Indonesia that hydrological drought index based on the theory of run, are good drought index, enable to identify drought onset and ending, duration, and severity. Good correlation between index and the impact data generally: a) Index with 3 months moving average; b) Log-Normal distribution; c) Fixed threshold level; and threshold level of mean flow, except for original data of one month the dependable flow is better. 8
The most suitable hydrological drought index identified is LN.F.80(6), a Log-Normal distribution on moving average data in 6 months, using a fixed threshold of dependable flow 80%. This result is related with the fact that water supply for irrigation is designed based on water availability of dependable 80%. Fixed threshold is generally good for longer moving averaged data while seasonal threshold will be more appropriate with original data and shorter moving average of 3 months. However, the decision making impact of selecting a particular type of drought index from among others is not significant, for all of the index are be able to identify the drought events. It is proposed to use this hydrological drought index for the area having the same hydrological characteristics in the tropical area. REFERENCES Edossa, D. C., Babel, M. S., & Gupta, A. Das. 2010. Drought analysis in the Awash river basin, Ethiopia. Water Resources Management, 1441 1460. doi:10.1007/s11269-009-9508-0 IPCC, 2007. Intergovernmental Panel on Climate Change: Synthesis Report, An Assessment of the Intergovernmental Panel on Climate Change, WMO, Geneva. Mishra, A. K., & Singh, V. P. 2010. A review of drought concepts. Journal of Hydrology, 391(1-2), 202 216. doi:10.1016/j.jhydrol.2010.07.012 Panu, U. S., & Sharma, T. C. 2009. Analysis of annual hydrological droughts: the case of northwest Ontario, Canada. Hydrological Sciences Journal, 54(1), 29 42. doi:10.1623/hysj.54.1.29 Shukla, S., & Wood, A. W. 2008. Use of a standardized runoff index for characterizing hydrologic drought. Geophysical Research Letters, 35(2), L02405. doi:10.1029/2007gl032487 Wilhite, D. A., Sivakumar, M. V. K., & Wood, D. A. 2000. Early Warning Systems for Drought Preparedness and Drought Management, Proceedings of an Expert Group Meeting held 5-7 September, 2000, in Lisbon, Portugal Yevjevich, Vujica. 1967. An Objective Approach to Definitions and Investigations of Continental Hydrologic Droughts Hydrology Papers Colorado State University Fort Collins, Colorado (August). 9