ASSESSMENT OF WATER RESOURCES IN UNGAUGED CATCHMENTS, KERALA

Abstract

Streamflow estimation is a crucial challenge required in assessments of water resources, planning, decision-making and improvement. Tools and data needed to carry out such assessments are often limited, especially in developing countries having limited technical capacity and funding. Due to this, there is a need to develop methods, techniques and tools for assessing water resources that works with limited data or in ungauged catchments. Hence, the objectives of this study are (a) to assess the catchment characteristics of gauged catchments, (b) to estimate the flow quantiles to establish flow duration curve of gauged catchments, and (c) to regionalize the information obtained from gauged catchments to estimate flow characteristics of ungauged catchments. The study uses 19 gauged catchment data and in 110 ungauged catchments located in Kerala, India. ALOS DEM was used to obtained various geomorphological characteristics and delineation of catchments. Precipitation data was obtained from IMD. Temperature, relative humidity and wind data was obtained from Global weather data provided by Climate Forecast System Reanalysis (CFSR) data. FAO Local Climate Estimator (New_LocClim) was used to obtained sunshine hours. The discharge data of gauged catchments was obtained from Central Water Commission portal i.e. WRIS. With MODIS land cover data, crop coefficient (Kc) values were calculated based on FAO guidelines. CROPWAT was used to get potential evapotranspiration. A climatic classification based on Budyko curve was used and it was found that 19 gauged catchments fall in the wet category with aridity index values between 0.38 and 0.69. They also have comparable evaporative indices (between 0.24 and 0.64) for regionalization of the catchments. The gauged catchments were then used as the donor catchments. Regression models were developed for all parameter using a forward stepwise-regression considering non-transformed and log-transformed data. Leave-one-out cross validation was utilized as the basic criteria for selecting the best performing models. The results showed that the log-transformed models outperformed the non-transformed ones. For high flows (q5), it was observed that precipitation (PREC), potential evapotranspiration (ETpot), drainage density(D), catchment area (CA) and percent of urban (LU) area are the explanatory variables. For median flow (q50), precipitation (PREC), minimum elevation (H-), percent of grasses/crop cover (LGC), catchment area (CA) and drainage density(D) were observed as the dominant explanatory variables. For low flows (q90) prediction, precipitation (PREC), percent of water bodies (LW), minimum elevation (H-), potential evapotranspiration (ETpot), percent of iv evergreen needle leaf forest (LEN) and maximum elevation (H+) appears in the stepwise model. And for low flows (q95), precipitation (PREC), percent of water bodies (LW), minimum elevation (H-), potential evapotranspiration (ETpot) and percent of evergreen needle leaf forest (LEN) are the dominant explanatory variables. The identified most substantial variables for the regionalization of FDCslp are elevation range (HR), percentage of broadleaf forest (LB) and waterbodies (LW), drainage density (D), catchment perimeter (Cp) and potential evapotranspiration (ETpot). The assessment of q5 and q50 in ungauged catchments shows that a larger predicted value found along the coastal area mostly draining in to the Arabian sea. The spatial pattern of low flows i.e. q90 and q95 demonstrates a decreasing trend alongside the eastern parts which is mostly mountainous. FDCslp is higher in the south eastern catchments demonstrating a high variable streamflow owing to direct runoff. The lower values of FDCslp are found in the western catchments or along the coastal area indicating higher contribution and presence of groundwater in these catchments.