Assessing the Hydrological Response to Land Use Changes Linking SWAT and CA-Markov Models

Abstract

Land use change, as a major driving factor of watershed hydrological process, has a significant influence on watershed hydrological change. In addition, a series of hydrological models, as important tools for simulating hydrological impacts, are widely employed in studying land use change. However, when employing hydrological model to analyse the hydrological impacts of land use changes, most previous studies focused on the evolution of historical land use change and lacked reasonable predictions of future land use. Therefore, it is necessary to extend such studies to future scenarios to cope with possible future hydrological variations in the basin. Given this, this paper making the Wuwei section of Shiyang River Basin as the study area, coupled the SWAT (Soil and Water Assessment Tool) model for hydrological simulation with the CA-Markov (cellular automata-Markov chain) model for future land use prediction to analyse the regional hydrological effects caused by historical and future land use changes. In addition, the general CA-Markov model directly uses a system-generated suitability atlas. In contrast, this study applied logistic regression and Multi-criteria evaluation (MCE) methods to construct the suitability atlas, thereby establishing the Logistic-CA-Markov and MCE-CA-Markov models. Based on the model results, the main results are as follows: (1) The land use in study area is mainly grassland and barren, accounting for more than 80%. Additionally, forest is changing at the highest rate among all land use types. (2) In terms of the percentage of grassland and forest, the future land use predicted by MCE-CA-Markov (Multi-criteria evaluation-cellular automata-Markov chain) has the largest forest and grassland coverage (57.78%), whereas the future land use predicted by Logistic CA-Markov has the lowest (54.69%), indicating that the former pays more attention to the sustainable development of ecological environment. (3) The study area's R² = 0.83, NSE = 0.79, PBIAS = −18.6%, and validation R² = 0.81, NSE = 0.76, PBIAS = −17.8% demonstrate the favourable application of the SWAT model. (4) Based on simulated runoff results under historical and future land use scenarios, the amount of increasing grassland and forest coverage in the study area would eventually rise water yield (WYLD) by increasing lateral runoff (LATQ), increasing subsurface runoff (GWQ), and reducing surface runoff (SURQ). The study contributes to a better understanding of the impact of land use change on regional water resources and water balance, thus guiding regional water resources management and sustainable development.