Development of Watershed Characteristics-Based Synthetic Unit Hydrograph

This paper intends to build a Synthetic Unit Hydrograph based on watershed characteristics, not from rainfall and runoff data. The geomorphology parameter from the synthetic unit hydrograph model is the most useful approach for predicting run-off and the simplest method for understanding the different hydrology behaviors of watersheds, mainly in ungauged watersheds or those with a lack of data. This research was conducted in 10 sub-watersheds in Indonesia that were completed using an automatic water level recorder (AWLR) and an automatic rainfall recorder (ARR). The methodology consists of data collection of watershed characteristic parameter hydrology, feature selection using genetic algorithm (GA), processing the initial data, assessment of SVR model by optimal subset model, hyperparameter tuning using particle swarm optimization (PSO), building Model of Support Vector Regression (SVR), and evaluation model. The results are as follows: 1) formulation of peak discharge: f(X) = 1.1627 X1 + (-0.5060 X2) + (-0.0315 X3) + (-0.1493 X4 + 0.0775 X5 + (-0.0340 X6); 2) formulation of time to peak: f(X) = (-0.8013) X1 + 0.9282 X2 + 0.0162 X3 + 0.5461 X4; and 3)formulation of time base: f(X) = 0.1771 X1 + 0.3323 X2 + 0.1128 X3 + 0.7149 X4; explanation: X1 = watershed area; X2 = river length; X3 = river slope; X4 = river drainage density; X5 = watershed shape; and X6 = run-off coefficient.