Author(s): Kenta Hakoishi; Masayuki Hitokoto; Shingo Zenkoji; Ryota Nishiguchi

Linked Author(s):

Keywords: Convolutional Neural Network; Explainable AI; Runoff analysis; Dam inflow prediction

Abstract: In this study, we developed the dam inflow prediction model using a 2D-CNN, taking the reanalysis precipitation from the Japan Meteorological Agency as the input condition. We applied XAI technology to visualize which input data influence the prediction results. The objective areas are the Haji Dam basin in the Gonokawa river system and the Nomura Dam basin in the Hijikawa river system in Japan. As the results of the case study, we could confirm that the degree of influence of each input data was reasonable in the sense of the rainfall-runoff mechanism.

DOI: https://doi.org/10.3850/iahr-hic2483430201-333

Year: 2024