A Study of Rain Station Network Distribution Using Artificial Neural Networks

Hydrological analysis is an important component in water resources management, especially for planning and controlling water infrastructure. This study evaluates the effectiveness of the rain station network in the Upper Ciliwung Watershed and identifies rain station with maximum accuracy in representing the study area conditions. Rainfall and discharge data were tested using statistical tests to ensure the absence of trends, stationary, persistence, and outliers. The evaluation of the rain station network density was conducted based on WMO guidelines, which determined the Upper Ciliwung Watershed met the criteria with a density of 37.981 km² per rain station. Analysis of rain station network distribution patterns using Artificial Neural Networks (ANN) was conducted with three data divisions (70-20-10, 60-25-15, 50-30-20) and tested at 100, 500, and 1000 epochs. The best results were obtained at 70-20-10 composition with 1000 epochs, showing the smallest relative error of 9.880% and NSE value of 0.983. The most effective rain station combinations are Gadog, Cilember, and Gunung Mas. This research provides recommendations in rain station network optimization to improve the accuracy of hydrological data.