COUPLING TWO HYDROLOGICAL MODELS TO COMPUTE RUNOFF IN UNGAUGED BASINS

Transcription

1 COUPLING TWO HYDROLOGICAL MODELS TO COMPUTE RUNOFF IN UNGAUGED BASINS ENG. A. ARCELUS National Directorate of Hydrography- Ministry of Transport and Public Works of Uruguay Rincón 575 2do Piso. C.P. 000 Montevideo, Uruguay Abstract The extrapolation of direct runoff from gauged watersheds to ungauged ones is not a simple task. There are several methodologies and models to obtain direct runoff from rainfall data. This paper discusses the combination of an event model (HEC-HMS) and a continuous water accounting model (NAM) to obtain discharge series from ungauged basins. An application of the methodology to Cebollatí River basin (Uruguay) is presented. Considering the lack of detailed spatial rainfall information, the applied procedure showed acceptable results. Resumen La extrapolación de escorrentía desde cuencas aforadas a cuencas que no lo son, no resulta una tarea sencilla. Existen varias metodologías y modelos para obtener datos de escorrentía a partir de los de precipitación. El siguiente trabajo analiza la combinación de un modelo por eventos (HEC-HMS) y uno de balance continuo de agua (NAM) de manera de obtener series de caudales en cuencas no aforadas. Para ilustrar lo anterior se presenta la aplicación de la metodología en la cuenca del río Cebollatí (Uruguay). Teniendo en cuenta la poca disopnibilidad de información espacial de precipitación, el procedimiento mostró resultados aceptables.

2 INTRODUCTION Hydrological models are simplified representations of the natural hydrologic system. In each case, the choice of the model to be applied depends mainly on the objective of the modelling but also on the available information. For instance, in order to study flooding frequency and classification of a wetland when discharge information is missing, the application of a continuous model to obtain long-term discharge series flowing from upstream basins into the wetland is appropriated. Many times the lack of data impedes the direct application of a model. In this case, there are also alternatives to overcome this situation. The application of numerical models as an indirect method to generate needed data for further modelling is a useful alternative procedure. This paper presents a methodology to couple a deterministic, lumped, event and empirical model (SCS approach from the HEC-HMS package) to a deterministic, lumped, continuous and conceptual model (DHI-NAM), in order to obtain long-term discharge series from ungauged basins. Results from its application to Cebollatí River basin in Uruguay are presented to illustrate the procedure. THE MODELS The HEC-HMS (Hydrologic Engineering Centre Hydrologic Modelling System) is a modelling package designed to simulate precipitation-runoff processes of dendritic watershed systems. The model offers different methods to simulate infiltration losses, transform excess precipitation into surface runoff, open-channel routing and compute baseflows. In order to obtain single event discharges, the utilised methods from the HEC-HMS package include the user defined rainfall gauge weights, SCS curve number, Unit Hydrograph, constant monthly baseflows, and Lag method for open channel routing. The parameters considered by each one of these methods are well known. The NAM model is a rainfall-runoff model that operates by continuously accounting for the moisture content in three different and mutually interrelated storages, which represent physical elements of a catchment (surface, root zone and ground water storages). Being a lumped model, it treats each sub-catchment as one unit, therefore the parameters and variables considered represent average values for the entire sub-basins. A short description of each one of the seven model parameters is presented. U max and L max define the maximum water content in the surface and root zone storages respectively. CQ OF is the overland flow runoff coefficient, small values are expected for catchments having coarse, sandy soils and values near for low-permeable soils like clay or bare rocks. CK is the time constant for overland flow routing, this is an important parameter and it depends on the size of the catchment and how fast it responds to rainfall. CKIF is the interflow routing parameter, is not usually a very important parameter since interflow is not the

3 dominant streamflow component. CK BF is the time constant for routing baseflow. Finally, Sy is the specific yield, which depends on the soil type. THE STUDY BASIN The study area (presented in fig. ) was set up on the Corbo, Pirarajá and Laureles Cebolltaí River sub-basins, which discharge flows into the Corbo wetland. These basins have total area km 2. Eighty four percent (2.850 km 2 ) is located upstream of a hydrological station named Paso Corbo, from which information on daily discharges and mean water levels is available. This area was named Corbo basin. The remaining 6% is composed of two ungauged basins presented as Pirarajá (47 km 2 ) and Laureles (376 km 2 ). The Corbo, Pirarajá and Laureles basin slopes computed according to Horton (94) are 9.22%, 6.4 % and 0.68% respectively. Following the USDA slope classification, the basins receive the same denomination: steep slope. The average altitude of the study area is 67m a.m.s.l. Concentration times computed according to the Kirpich formula (Chow et al., 988) are 28hs, 2hs, and 3hs respectively. For modelling purposes the Corbo basin was divided into four sub-basins named De los Chanchos (72 km 2 ), Molles (585 km 2 ), Barriga Negra (934 km 2 ) and Tapes Grande (69 km 2 ). The meteorological information used in this research was daily rainfall data, measured from two stations located in and near the basin (Pirarajá and Mariscala). The land use description most suitable for the study area is pastures with fair hydrologic condition. The prevailing soils are brunosols, and the association into hydrological groups is presented in fig. and table. METHODOLOGY AND RESULTS As a first step, the gauged basin was considered. An event model (HEC-HMS) and a continuous one (DHI-NAM) were applied. For the first model, parameters were obtained as presented later in this section, and the computed results were compared to the measured values to verify the validity of the model. The second model was calibrated to adjust the computed values to measured data. Once the above analysis produced reliable results, the ungauged basins were considered. Application of the HEC model was carried adjusting its parameters according to the basins characteristics. Later, NAM model was calibrated to that ungauged basins using those event hydrographs generated by HEC model as measured values. This last task involves calibration of two significant NAM parameters (overland flow runoff coefficient and time constant for overland flow routing) related directly with the shapes and peaks of the discharge hydrographs. The main concept presented is to link NAM results to HEC computations for each considered event, being NAM the model that will be used to calculate continuous discharge

4 series. Meteorological and hydrological data from 990 to 995 was used for the application of both models. For study carried with the HEC-HMS model, empirical relations were considered, most of them related to the basin soils and morphological characteristics. No model parameter calibration was performed since parameter adjustment was impracticable in the ungauged basins. In the gauged basin, most of the NAM model parameters were obtained by calibration or from soil physical measurements. For the model application in the ungauged basins, some parameters were extrapolated considering a sensibility analysis and the existing similarities among basin characteristics. The described methodology was applied considering the Corbo basin as the gauged basin, and Piraraja and Laureles as the ungauged ones. This application was as follows. Events modelling. As it was mentioned, the HEC-HMS package was used to model single storms and Corbo basin was divided into four sub-basins. Precipitation information was processed applying rainfall gauge weighting computed according to the Thiessen polygons. The SCS Curve number was used to compute basin loss. The applicability of this procedure in the study area showed an acceptable correspondence (Arcelus, 200). Hydrological groups and CN values were obtain following Duran s classification (Duran, 996), and Table 9. of the NEH-4 (National Engineering Handbook, Section 4 Hydrology) respectively. Considering the existing hydrologic soil complexes, one weighted CN for each sub-basin was computed. Table shows information and parameters utilised in the modelling procedure. The SCS unit hydrograph was applied as basin transform method. Concentration times were used to estimate the SCS Lag time. For baseflows computations constant monthly values were employed. These values were computed taking into account the mean monthly specific baseflows obtained in Paso Corbo for the entire basin (q= Q/A). As reach routing procedure, since other computations required parameters not available at this point, the lag method was used. (see table, Basin Lags) For the time analysed, thirteen events were considered. Table 2 presents date, peak discharge, precipitation, direct runoff measured, runoff coefficient and discharge return period for each event. Figure 2 shows the comparison between computed and measured values for each storm. Hydrological data from 990 to 995 were used to contrast and verify if the computed values fitted the measured ones. As it was presented, the mentioned comparison was carried with no calibration or adjustment of the model parameters. The same procedure for model parameter estimation was applied to Pirarajá and Laureles basins to obtain event discharge hydrographs for the mentioned period of time. This information was later used for calibration of NAM model to those ungauged basins.

5 BASIN THIESSEN WEIGHTS PIRARAJA MARISC CHANCHOS 00 0 MOLLES 00 0 BARRIGA NEGRA TAPES GRANDE PIRARAJA 00 0 SOIL UNIT HYDR GROUPS WEIGHT (%) CN(II) SCS Reach Lmax Lag (hs) Lag(hs) (mm) SCl B Za C SP C Sag D TOTAL BASIN CN 79 SCl B Cch B 84.2 Za C 84.2 SP C JPV C 98.5 Sag D TOTAL BASIN CN 78 SCl B Za C 84.2 SP C JPV C 98.5 TOTAL BASIN CN 72 SCl B Ca B 40 SP C N/A JPV C Af C 02.5 TOTAL BASIN CN 76 SP C 77 JPV C Af C.4 N/A 02.5 TOTAL BASIN CN 79 Table Sub-basins Soils, hydrologic groups and CNs DATE PEAK DISCHARGE (CMS) PRECIP (mm) RUN OFF (mm) RUN-OFF COEF RETURN PERIOD 07/0/90 6// /06/9 22/07/ /09/ /05/ /06/ /06/92 3/08/ /2/93 /05/ /08/ /02/ Table. Measured events to compare HEC-HMS computations in Paso Corbo Continuous modelling.

6 Calibration of NAM model in the Corbo basin was carried out adjusting the already described parameters. Some of them were found considering an annual water balance (U max, CQ OF ) and physical measured data using values recommended by the literature (L max, Sy). The others were settled trying to fit the measured discharges using the trial-anderror methodology (CK, CKIF, CKBF). Once again, hydrological data from 990 to 995 were used for this purpose. Data on potential evapotranspiration (E t ) were not available, therefore information to compute it was processed as follows. Class A pan monthly evaporation data (E a ) representative of the study basin were obtained (Treinta y Tres meteorological station). For each month, using Evapotranspiration Studies in Uruguay (J. Henry, 973) a relation Et/Eo was established where Eo is open water evaporation. Finally, E a was transformed into E t considering the relation Eo= 0.7*Ea. L max, can be estimated multiplying the available water holding capacity of the soil, by the root zone depth. For each soil unit, the parameter was computed considering data from C. Alvarez et al (989). Since the model is lumped, in order to find one representative value for the basin, these values were weighted according to each soil extension. The average value used in the model was L max = 78.5 mm (see Table ). An average value of Sy= 0. was considered according to the soil types (clay) presented in the area (DHI, 990). The calibrated model parameters were U max = 20mm, CQ OF = 0.9, CK = 30 hs, CK IF = 750 hs, CK BF = 000 hs. Figure 3 presents, for the six years period, the adjustment between the modelled and the measured values at Paso Corbo station. NAM model parameters were established for the modelling of Pirarajá and Laureles basins discharges. L max and Sy were computed as described in the Corbo basin, but using the corresponding local data. Considering that the examined basins do not have important differences regarding vegetation and topography, since U max is related to moisture intercepted on the vegetation as well as to water trapped in depressions, this parameter was kept unchanged. A sensibility analysis was performed to examine uncertainties existing for those parameters that could not be obtained from field data. The study showed that variations in CK IF and CK BF did not have significant impact in the computed discharges. Therefore, the extrapolation of the parameters with the same value was thought as a reasonable step. Finally, the calibrated parameters to fit the previous computed discharges were the overland runoff coefficient (CQ OF = 0.9) and the time constant for overland flow routing (CK=0hs for Pirarajá, and 0hs for Laureles). Figure 4 presents the adjustment between results from NAM model and those values used as measured (given by HEC-HMS), for the six years period in Pirarajá sub-basin. DISCUSSION AND CONCLUSIONS

7 For the modelled events, an exact agreement between simulations and observations was not achieved. One of the reasons for that may be the use of the SCS unit hydrograph method in basins where some of its hypothesis are not fulfilled (mainly basin size). Nevertheless, considering the above-mentioned and the fact that information of only two raingauges was available for the study, the results presented in fig. 2 show a suitable relation between the measured and the computed values. In this sense the methodology was considered valid for application in the near ungauged basins with similar characteristics (soils and topography). In the case of continuous modelling the results indicate more accurate outcomes. Nevertheless, it is considered that some of the disagreements are due the lack of more detailed spatial rainfall information. In this instance, the basins size did not constrain the model application since other studies showed its employment in greater catchments (DHI, 990). In the NAM model calibration for the ungauged sub-basins, a reduction in CK was observed. This seemed logical, since a minor basin size implies faster rainfall-runoff responses. A sensibility analysis showed relative importance of two routing parameters (CKIF, CKOF). This was useful for the case analysed since for that reason the two values were extrapolated with the same values. Nevertheless similar studies should be performed in order to discuss if this step can always be applied. A procedure to obtain discharges from rainfall in ungauged basins was developed based in the characteristis and advantages presented by each hydrological model (HEC-HMS and DHI-NAM). In this manner it was possible to adjust the overland flow runoff coefficient and the time constant overland flow routing parameter, since both values are difficult to extrapolate to ungauged basins. The methodology cannot be used in areas without any discharge measures at all, since hydrological data is needed for event model testing and continuous model calibration. Further research considering other basins, with more detailed spatial rainfall information as well as utilisation and comparison among the different models provided in the HEC-HMS package should be addressed to confirm or improve the proposed methodology. REFERENCES Army Corps of Engineers U.S.- HEC-HMS, Technical Reference Manual. Army Corps of Engineers U.S.- HEC-HMS, User s Manual. Chow, V.T., Maidment, D.R. and Mays, L.W. (988) Applied Hydrology. Mc Graw-Hill, Inc., New York, USA. Doti, R., Durán A. And López Taborda, O. (979) Carta de Reconocimiento de Suelos del Uruguay. Danish Hydraulic Institute, NAM Documentation and User s guide. Durán, A. (995) Clasificación en Grupos Hidrológicos de los Suelos del Uruguay. Facultad de Agronomía, Area de Suelos y Aguas- Cátedra de Edafología.

8 Sherman, L.K., Streamflow from rainfall by the unit-graph method, Eng. News Rec., Vol. 08, pp , April 7, 932. Soil Conservation Service, Hydrology, Sec. 4 of National Engineering Handbook, U.S. Dept. of Agriculture, Washington, D.C., 972.