Key points are not available for this paper at this time.
The interactions between landscape structure, land use intensity (LUI), climate change, and ecological processes significantly impact hydrological processes, affecting water quality. Monitoring these factors is crucial for understanding their influence on water quality. Remote sensing (RS) provides a continuous, standardized approach to capture landscape structures, LUI, and landscape changes over long-term time series. In this study, RS-based indicators from Landsat data (2018–2021) were used to assess landscape structure, LUI, and land use change for a study area in northern Germany, applying the ESIS/Imalys tool. These indicators were then used to model and predict water quality (Chl a) in 119 standing waters. Various machine learning methods, including Generalised Linear Models, Support Vector Machines, Deep Learning, Decision Trees, Random Forest, and Gradient Boosted Trees, were tested. The Random Forest model performed best, with a correlation of 0. 744 ± 0. 11. Indicators related to landscape structure, such as diversityₘean (0. 376) and relationₘean (0. 292), had the highest global correlation weights, while LUI and land use change indicators like NirV2ₘean (0. 369) and NirVᵣegme (0. 284) were also significant. All indicators and their effects on water quality (Chl a) are discussed in detail. The study highlights the potential of the ESIS/Imalys tool for quantifying landscape structure, LUI, and land use change with RS to model and predict water quality and suggests directions for future model improvements by incorporating additional influencing factors. • Currently, there is no RS-based tool to derive hybrid raster and zonal RS-based indicators. • We use our recently developed and freely available ESIS/Imalys RS tool to quantify feature-based landscape structures, land use intensity and landscape change. • Based on the ESIS/Imalys tool, indicators are derived to model the importance of landscape indicators for predicting the water quality (Chl a) of still waters • Random forest is the most suitable machine model for predicting the water quality (Chl a) of still waters. • The importance of the integrated indicators of landscape structure, land use intensity and landscape change is determined.
Lausch et al. (Wed,) studied this question.