A coupled eco-hydrological model to simulate vegetation condition from soil moisture deficit in a data-scarce semi-arid wetland

Abstract

Environmental flows are increasingly being used to restore or maintain wetland vegetation condition, with vegetation response being one of the primary considerations when determining environmental water releases. However, it is challenging and expensive to evaluate vegetation condition over time and at scale from on-ground monitoring to support such decisions. This study presents a model that predicts the Leaf Area Index (LAI) used as a proxy for vegetation response to climate and river flow drivers. The proposed coupled eco-hydrological model simulates floodplain inundation extent using a mass balance equation and estimates soil moisture deficit by modifying the IHACRES-CMD hydrological model to account for both climate and inundation. The eco-hydrological model finally predicts LAI via a linear regression using simulated soil moisture deficit as a predictor. The conceptual-empirical approach is the first attempt to use soil moisture-related information to predict vegetation condition in an environmental flow management context, compared with the more commonly used flow- and inundation-based metrics in floodplain wetlands. The model achieves promising performance in a semi-arid wetland, the Narran Lakes in Australia. The results based on cross-validation on different periods show that: (1) inundation influences vegetation response significantly; (2) soil moisture deficit, when estimated via a modified hydrological model, is a useful predictor of vegetation response, especially for drought periods when flow releases are most needed; and (3) the optimal frequency of flood events required to maintain vegetation health in the Narran Lakes is one event every 4-6 years. The simple structure of our eco-hydrological model, which utilizes regularly monitored hydro-climatic variables such as precipitation, evapotranspiration, and inflow, could potentially be used in vegetation monitoring and management in similar semi-arid wetland systems. This requires processing of remote-sensed vegetation information and calibration of parameters in the model from the hydro-climatic data.