Hydrological model application under data scarcity for multiple watersheds, Java Island, Indonesia

Study regionJava Island, Indonesia.Study focusThe Indonesian island of Java is home to more than half of Indonesia's population and routinely experiences water related natural disasters. This study represents a first step towards skillful hydrologic prediction and hydrologically-informed mitigation strategies. This is the first study to collate a comprehensive suite of hydrometeorological observations and systematically identify Variable Infiltration Capacity (VIC) Land Surface Model (LSM) parameters on Java to create a set of benchmark simulations.New hydrological insights for the regionQuality control procedures revealed inconsistencies between precipitation and streamflow with only five watersheds possessing data of suitable quality. Simulations and observations confirmed that both precipitation and streamflow variability increase eastward on the island and that rainfall-runoff response was most frequently dominated by baseflow, rather than surface runoff. The most sensitive VIC parameters were identified and then calibrated with an automatic calibration procedure. In the calibration period, model performance was generally deemed satisfactory with Nash Sutcliffe Efficiency (NSE) between 0.31 to 0.89, whereas the validation period exhibited poorer performance than expected (0.07 < NSE < 0.79). This drop in performance was attributed to a combination of inconsistent data quality, hydrometeorological outliers during the validation period, and over-fitting parameters during the calibration period. The model indicated that direct runoff exhibits more spatial and temporal variability than both rainfall or baseflow, the latter being associated with variability of soil thickness.