The impact of environmental variables on reed stands of the intermittent Lake Cerknica, Slovenia: 40 years of change

• The M5P algorithm holds potential as a predictive management tool. • MSAVI 2 is suitable for long-term change detection of reed stand dynamics. • Significant changes in reed stands at Lake Cerknica occur near annually. • Yearly dynamics highlight significant differences between observed locations. • Reed stands most affected by disturbances showcase the highest amplitude of changes. The common reed ( Phragmites australis ) is a highly adaptable perennial grass that thrives in diverse wetland environments, capable of forming dense stands that are crucial for ecosystem services such as soil stabilization and habitat provision. At Lake Cerknica, long-term data on water levels and temperatures, combined with four decades of Landsat satellite imagery, allowed for a comprehensive analysis of reed stand dynamics of three different reed stands using the modified soil adjusted vegetation index (MSAVI 2 ) and an index differentiating technique. Decision tree-based data mining using the M5P algorithm effectively identified significant changes in MSAVI 2 . Previous MSAVI 2 values were key predictors for detecting increases in the following year, while water level data, particularly high summer water levels, were critical for identifying significant decreases. Smaller changes were less reliably predicted, but temperature data emerged as a more influential predictor in such cases. Long-term field measurements of common reed biomass confirmed the accuracy of remotely sensed data. Yearly dynamics of reed stands showed a consistent alternation between significant changes in MSAVI 2 with short periods of smaller changes. The largest amplitude of change occurred at the reed stand most affected by disturbances. Despite these annual fluctuations, reed stands maintained long-term equilibrium but require continued monitoring to identify years where a decrease in reed vitality is expected. Consequently, mowing events in such years should be minimised. The findings highlight the utility of combining remote sensing with environmental data to develop predictive models that recognize changes of reed stand vitality, providing crucial information for effective reed management.