Applicability of ABCD water balance model for the assessment of water resources in Kelani basin, Sri Lanka

Abstract

Water resources management in watersheds has become increasingly important due to rapid expansion of human settlements while pollution caused by industrial development has led to the part of the available precious water resources unusable for consumption, thus aggravating scarcity of fresh water resources. The impacts are further exacerbated due to global warming. The use of the multi-parameter, distributed hydrologic models for water resources assessment in the local basins are hindered due to scarcity of data and other resources. The lumped parameter rainfall runoff hydrologic models are widely applied to predict watershed response of small watersheds by simulating rainfall runoff generation and thus useful in water resource management in ungauged basins. This study aims at identifying distinct characteristics of one such widely used model, ABCD Water Balance Model, and studying its applicability to a selected sub basin in Kelani River Basin for simulating catchment response in terms of rainfall runoff. The model was subsequently applied to analyze surface and groundwater resources available in the basin, targeting effective and sustainable water resources development and management. The data required for the ABCD water balance model were precipitation, evapotranspiration, average temperature and minimum and maximum temperatures. The model was developed in Excel spread-sheet format focusing on the data period from 1994~2011 in the Kelani basin. For model calibration, precipitation and potential evapotranspiration data during the period 1994 to 2001 were used. The generated model streamflow was compared with observed streamflow at Glencorse station for the same period. For the validation of the model, the precipitation and potential evapotranspiration data in the latter 10-year period were used. For estimating the goodness-of-fit, Nash-Sutcliff efficiency coefficient method was used, while model response to four distinct parameters were assessed based on sensitivity analysis and parameter optimization. The calibrated model has shown that the model is less sensitive to parameters a (0.9) and b (20) while on the other hand, the model was highly sensitive to parameter c (0.68) and d (0.01). It was noted that even with the lesser amount of moisture infiltration from the upper soil zone, the aquifer was able to produce runoff. Hence, it proved that in the wet zone, the propensity of the area to produce runoff was largely independent of rainfall intensity. For the model calibration runs, the correlation or coefficient of determination (R2) between model flow and observed flow was 0.77 with NASH coefficient value of 0.71 and MRAE of 0.27. The model produced a better response to medium flows between 5% ~ 82% with NASH value of 0.78 and good response for high flows below 5% of percent exceedance with acceptable results (NASH = 0.62). The model could not response well for low flows (NASH = 0.45). This model with four parameters could adequately simulate the rainfall runoff response of the selected sub-watershed area in Kelani Basin (at Glencorse). Hence, this lumped parameter model was deemed suitable for streamflow forecasting and water resources assessment in Kelani basin and it can also be applied in areas elsewhere with similar hydrological characteristics.