The Suitability of Artificial Neural Network Application to Predict Sekayam River Discharge in West Kalimantan, Indonesia

Abstract

Data availability of on a river discharge is the key to waterworks planning. Unfortunately, not all rivers have long and complete historical data records to support the planning. Therefore, a hydrological model capable of predicting long-term river discharge is needed. There are many hydrologic models that have been developed, ranging from the simplest ones by using empirical black-box model, to complex ones with physical white-box model. This study used ANN application due to its data requirement that is applicable to be met in study area, Sekayam River, a part of Kapuas Subwatershed, namely Kembayan Watershed. Although the available data is relatively minimal, which is only rainfall and evaporation data, the ANN application can predict river discharge that is close to the measurement in the field, with a mean square error (MSE) of 0.25. The results show that ANN application was able to predict river discharge reasonably with climate and rainfall data as the input. Deviation may occur due the broad scope of the research area, Kembayan Watershed, a Kapuas Subwatershed which amounted to 2,290 km2.