<div> <div> <p><span data-contrast="auto">Seewinkel salt pans are a unique wetland ecosystem in eastern Austria that serves as habitat for a diverse range of e.g. birds and halophilic species. Due to groundwater drainage by channels and wells, the salt pans are in an increasingly vulnerable state as they are decisively conditioned by duration and timing of water abundance. However, water gauge data are merely given for three salt pans. The dynamics of</span><span data-contrast="auto"> salt pans in Seewinkel, locally referred to as </span><em><span data-contrast="auto">Salzlacken</span></em><span data-contrast="auto">, remain insufficiently understood in the context of continuously changing seasonal and long-term hydrological, meteorological, and climatological patterns. Based on previous results on salt pan mapping and monitoring, this work advances inundation state prediction for 34 salt pans by using high-resolution remote sensing data and machine learning methods. The random forest classification models build on hydrological and meteorological predictors in 12-monthly temporal resolution, as, e.g., reduced precipitation sums during the preceding winter season affect the recharge rates of salt pans and groundwater and, as a result, drying state in summer. Four models predict summer drying state at respective four points in time, namely in March, April, May, and June of each year between 1984 and 2022. We first show that remotely sensed water extent products, retrieved from Landsat data can serve as a target variable for data-driven modelling of small-scale salt pan water-dynamics. Secondly, we show that the applied models can successfully predict summer drying state and inundation periods of individual salt pans achieving a maximum F1-score of 0.81. Finally, it is demonstrated that very similar model results can be attained without in-situ groundwater measurements. Research based on water gauge measurements with similar model-designs has been done in the context of lakes, whereas the combination of satellite-derived water extent and salt pans, especially for ecosystems of small size, remains underrepresented. As the data retrieval in this work is based on global and freely available remote sensing data, this method is transferable to comparable salt pan ecosystems in other parts of the world.</span><span data-ccp-props="{"> </span></p> </div> </div>