Modeling the regional water storage for geographically isolated wetlands

Abstract:

Geographically isolated wetlands (GIWs) are essential to regional hydrological modeling and vulnerable to future climate changes. The water storage volume (WSV) of GIWs is the key factor to understand GIWs’ ecosystem health and services. Therefore, estimating regional WSV of GIWs and projecting their future changes are the basis of building hydrological models and making regional environmental management policies. This study developed a hybrid Machine Learning (ML) - hydrologic modelling approach to estimate regional WSV of GIWs and to project the future regional WSV. To estimate the baseline WSV, 21 ML models were built based on 234 in-situ wetland survey data and a Gaussian progress regression model was chosen to estimate regional WSV for the study area. 4 out of 14 indicators that describe GIWs’ geometry, bathymetry, and the complexity of hydrologic relation to the surrounding drainage basin were selected to estimate WSV by a Multi Principal Component Analysis method. To project the future WSV changes (2018-2034) of regional GIWs in the study area, a conceptual model based on water balance was adopted, of which the future climate change data (precipitation, evapotranspiration, seepage) were implemented from comprehensive sources such as a Soil and Water Assessment Tool model, remote sensing data, and an ensemble of nine GCM under two emissions scenarios (RCP 2.6 and RCP 8.5). The spatiotemporal analysis of the WSV offers new insights and reveals (1) in total, the provincial GIWs hold a WSV of 550 million m^3, with an average of 5,508 m^3 per GIW, (2) a decreasing trend in the future WSV for both RCP 2.6 and RCP 8.5, (3) the minimum and maximum WSV loss rates occurring during October and February, respectively. This study provides a better insight into the connection between the WSV of GIWs and watershed-scale hydrology and offers an improved framework to understand climate change across multiple spatial scales.