Prediction of groundwater level using satellite image and multivariate GRU

Groundwater depletion, driven by excessive extraction and climate change, necessitates accurate prediction models. This study proposes a groundwater level prediction model using GPM Core satellite precipitation data combined with a GRU (Gated Recurrent Unit) time series prediction model. The model's performance will be compared with Korea's LDAPS (Local Data Assimilation and Prediction System) to develop an optimal groundwater level prediction model for Korea. The research focuses on a monitoring station in Shinpung-myeon, Gongju-si, Chungcheongnam-do, covering the period from 2013 to 2021. Input data includes GPM Core precipitation, Day of Year (DOY), and cumulative precipitation data. The 20-day cumulative precipitation showed high importance, improving the model's performance. This model is expected to effectively monitor groundwater level changes and predict decreases due to reduced precipitation, facilitating proactive groundwater management. This work is supported by the Korea Agency for Infrastructure Technology Advancemen(KAIA) grant funded by the Ministry of Land,Infrastructure and Transport (Grant RS-2022-00155763).