Wetland inflow simulation using artificial intelligence prediction model based on classification for water surface area identification

Wetland ecosystems have complex interactions of physical and biogeochemical processes, but the first step toward restoring the health of wetland ecosystems is an accurate understanding of the water cycle in wetland ecosystems. In addition, a quantitative understanding of the wetland water cycle is essential to utilize wetlands in regional water balance and ecosystem conservation. However, observational data essential for understanding the wetland water cycle are difficult to obtain through field measurements or are difficult to observe due to cost issues. Therefore, this study proposes a procedure for estimating wetland inflow using Sentinel-2 satellite data. To this end, a classification-based artificial intelligence model using data from major multi-purpose dams located on the Nakdong River in the southeastern part of the Korean Peninsula is designed. Input data for artificial intelligence learning is created by the following procedure. 1) Derivation of the water level-water surface area relationship curve using the water level-water volume relationship of the multi-purpose dam. 2) Using the water level-water surface area relationship curve and DEM, derive an identifier that distinguishes water and land areas. 3) Design a random forest model that compares Sentinel-2 satellite information and water-land identifiers. 4) Derivation of identifiers that can identify water and land in unmeasured wetland areas from water-land information of satellite information. By combining the water surface area of the wetland estimated through this process and the DEM of the wetland area, the wetland water level-water surface area-water volume relationship curve is calculated, and finally the wetland inflow is simulated. The simulated wetland inflow can be used to estimate the parameters of various hydrologic models, and it is expected that the understanding of the wetland water cycle can be improved by using the verified hydrological model.

 

Acknowledgement

This work was supported by Korea Environmental Industry&Technology Institute (KEITI) through Wetland Ecosystem Value Evaluation and Carbon Absorption Value Promotion Technology Development Project, funded by Korea Ministry of Environment (MOE). (2022003640001)