Author(s): Seunghyun Hwang; Jeemi Sung; Seoyeong Ku; Jongjin Baik; Changhyun Jun

Linked Author(s): Jongjin Baik, Changhyun Jun, Changhyun Jun

Keywords: Wetland biodiversity; Explainable artificial intelligence; SHAP; Remote sensing; Environmental monitoring

Abstract: This study aimed to identify key environmental determinants of wetland biodiversity using satellite-based remote sensing and explainable artificial intelligence (XAI). To this end, a long short -term memory (LSTM) model was developed using a biodiversity in dex derived from the composite biodiversity rankings provided by the National Institute of Ecology in the Republic of Korea as the target variable. A total of 15 variables were considered as model inputs, including 13 remote sensing–based environmental indicators associated with topography, hydrometeorology, vegetation, and carbon dynamics, along with wetland area and wetland type. The XAI framework was then applied to quantify the contribution of each feature to the estimation of wetland biodiversity. The results demonstrated that wetland area, digital elevation model (DEM), wetland type, and fraction of absorbed photosynthetically active radiation (FPAR) were the most influential determinants of biodiversity. The findings of this study provide empirical evidence supporting the use of remotely sensed environmental data for large-scale monitoring of wetland biodiversity, and a scalable and repeatable framework for assessing ecological conditions in regions where field-based surveys are logistically challenging.

DOI: https://doi.org/10.64697/iahr.proc.ise2026.abs.160

Year: 2026