In 2021, the Tajogaite Volcano erupted along the western slope of the Cumbre Vieja on the island of La Palma, Canary Islands, Spain. Volcanic tephra blanketed a substantial proportion of the island. By our estimations, approximately 23,000,000 m³ of pyroclastic ashes and more coarse-grained particles were deposited unto La Palma's land surface in addition to the lava flow. Five months following the initial eruption, we measured the depth of the new ash layer across the island. We combined this data with drone-based observations to compile a dataset comprising the point distribution of ash depth. A spatial interpolation was then performed using Inverse Distance Weighting (IDW) to estimate the ash depth across the island at a 2 m spatial resolution. The interpolation performed well, yielding a root mean squared error (RMSE) value of 0.34 and thus, the dataset offers immense reuse potential for spatial inquiries related to evolutionary traits, vegetation patterns, and vegetation response to disturbance on oceanic islands. In addition, the data can be used to test different spatial interpolation techniques in an effort to improve the accuracy achieved using IDW.

Over the course of industrialization in the 20th century, vast emissions of air pollutants have occurred. The exhaust gasses contain sulfur and nitrogen oxides, which are converted to sulfuric acid and nitric acid in the atmosphere. This causes acid rain to enter aquatic and terrestrial ecosystems, the most serious consequence of which is large-scale forest dieback across Europe and North America. However, through various political measures, the exhaust gasses have been reduced and, thus, acid rain and forest dieback were stopped. Nevertheless, the lingering effects of this pollution are still present today and are reflected in hydrochemistry. More recently, fluctuating precipitation regimes are causing additional stress to ecosystems in Central Europe. Climatic extremes are becoming more pronounced with climate change. Substantial differences between drought years and years with regular precipitation are directly altering the discharge of springs. Now, two overlapping and interacting syndromes of environmental pressures can be studied in these small catchments at a landscape scale: (1) acidification and (2) climate change. In this long-term study, the waters of 102 forest springs, located in two neighboring forest landscapes in north-eastern Bavaria, Germany (Frankenwald and Fichtelgebirge), were investigated over 24 years (1996 to 2020). By linking changes in pH values with changes in precipitation and spring discharge, we found that pH increases with decreasing discharge and decreasing precipitation. This effect was strongest in the Frankenwald compared to the Fichtelgebirge. We hypothesize that this temporal pattern reflects the longer residence time and, in consequence, the increased buffering of acidic interflow in small catchments during periods of drought. However, this should not be misinterpreted as rapid recovery from acidification because this effect fades in times of enhanced precipitation. We recommend that fluctuations in weather regimes be considered when investigating biogeochemical patterns throughout forest landscapes.

In the last decade, German forests have been decimated because of extreme events such as drought and windthrow, and bark beetle infestations that occur in the aftermath, primarily in monoculture Norway spruce stands. It is essential for decision makers in forest management to have an educated estimation of potential future loss. We have developed a model to predict future canopy cover loss in German spruce forests. Since, past canopy cover loss is a key predictor, we adapt the spatio-temporal matrix (STM) method used for predicting urban growth, to work with a canopy-cover-loss time-series product based on earth observation data. We configure a hybrid neural network model using the STM, its percentiles along with climatic and topographic data to produce the probability information of canopy cover loss in German spruce forests in the next year. The prediction results from the model show a good capacity of prediction, as validation results present an AUC of the ROC space as high as 82.3%. Our results show that future canopy cover loss can be predicted with reasonable accuracy using open-access earth-observation time-series data supplemented by environmental data without the need for site specific in situ data collection.

Mountain snow is an important component of the cryosphere that directly affects downstream livelihoods. Accurate monitoring of mountain snow is crucial, yet remains challenging due to complex topography and frequent cloud cover. Although combining data from multiple Earth Observation (EO) satellites can improve spatial and temporal coverage, extracting Fractional Snow Cover (FSC) from sensors with different spatial and temporal resolutions remains difficult. AI-based Earth foundation models can process diverse sensor inputs and have been successfully applied to various remote sensing tasks; however, they often lack snow-specific design considerations. This paper presents SnowGalileo, a pre-trained transformer model designed to integrate EO data to map snow-covered areas. We evaluated SnowGalileo's potential for generating daily, gap-free FSC maps at 100m resolution in mountainous regions. SnowGalileo combines one week of satellite time-series data from multiple sensors, including Sentinel-1, Sentinel-2, Landsat, Sentinel-3, MODIS, and VIIRS, with topographic, land cover, and meteorological information to predict FSC for a given day. The model was pre-trained using masked autoencoding and fine-tuned using labeled data from various mountain ranges across the Northern Hemisphere. In addition to clear-sky conditions, SnowGalileo is evaluated under a wider variety of conditions than was previously possible, including cloud cover and lack of high-resolution (~10–30 m) satellite imagery. For the Canadian Rockies and the Swiss Alps, respectively, SnowGalileo achieves RMSEs of 0.099 and 0.124 on clear days, 0.144 and 0.209 on cloudy days, 0.185 and 0.262 on days without high-resolution imagery, and 0.200 and 0.299 on cloudy days without high-resolution imagery. SnowGalileo also consistently outperforms random forests, support vector regressors, and multi-layer perceptrons. While the current product is a proof of concept that has undergone limited validation across geographic regions, operates on 1km × 1km tiles rather than full maps, and has restricted capabilities in challenging conditions, this approach could ultimately enable the continuous, gap-free operational generation of FSC time series for mountain regions worldwide.

Mountain snow is an important component of the cryosphere that directly affects downstream livelihoods. Accurate monitoring of mountain snow is crucial, yet remains challenging due to complex topography and frequent cloud cover. Although combining data from multiple Earth Observation (EO) satellites can improve spatial and temporal coverage, extracting Fractional Snow Cover (FSC) from sensors with different spatial and temporal resolutions remains difficult. AI-based Earth foundation models can process diverse sensor inputs and have been successfully applied to various remote sensing tasks; however, they often lack snow-specific design considerations. This paper presents SnowGalileo, a pre-trained transformer model designed to integrate EO data to map snow-covered areas. We evaluated SnowGalileo's potential for generating daily, gap-free FSC maps at 100m resolution in mountainous regions. SnowGalileo combines one week of satellite time-series data from multiple sensors, including Sentinel-1, Sentinel-2, Landsat, Sentinel-3, MODIS, and VIIRS, with topographic, land cover, and meteorological information to predict FSC for a given day. The model was pre-trained using masked autoencoding and fine-tuned using labeled data from various mountain ranges across the Northern Hemisphere. In addition to clear-sky conditions, SnowGalileo is evaluated under a wider variety of conditions than was previously possible, including cloud cover and lack of high-resolution (~10–30 m) satellite imagery. For the Canadian Rockies and the Swiss Alps, respectively, SnowGalileo achieves RMSEs of 0.099 and 0.124 on clear days, 0.144 and 0.209 on cloudy days, 0.185 and 0.262 on days without high-resolution imagery, and 0.200 and 0.299 on cloudy days without high-resolution imagery. SnowGalileo also consistently outperforms random forests, support vector regressors, and multi-layer perceptrons. While the current product is a proof of concept that has undergone limited validation across geographic regions, operates on 1km × 1km tiles rather than full maps, and has restricted capabilities in challenging conditions, this approach could ultimately enable the continuous, gap-free operational generation of FSC time series for mountain regions worldwide.

Snow cover is a critical component of the earth’s climate and weather system, which exhibits high spatial and temporal variability. Therefore, we predict daily snow cover using spatio-temporal forecasting. Unlike traditional forecasting approaches, that require spatial or temporal aggregation, our approach employs deep learning models specifically designed for spatio-temporal data. Spatio-temporal predictive learning has primarily focused on nowcasting and sub-seasonal forecasts at a daily scale with a lead time up to 15 days. However, we implemented the capability of using such models with long multi-year satellite image time series to predict at a daily scale annually, specifically for daily snow cover. In our research, we use the historical snow cover data from the DLR Global SnowPack remote sensing product, which is a daily cloud free 500m snow cover representation on the ground. We generate high resolution daily snow cover forecasts for up to 365 days (one year ahead) beginning from 1st July. We implemented models such as Convolutional Long Short-Term Memory (ConvLSTM) networks, convolutional encoder decoder architectures with attention mechanisms, and Vision Transformer (ViT) based models and adapted them for our use case. To further enhance our predictions, we also made adaptations to the models to include multivariate spatial and temporal data which are key drivers of snow cover variability into the model. Topographical feature maps derived from elevation, and time series of climatological indices (atmospheric oscillation patterns) are two examples. Validation against reference data demonstrates exceptional accuracy and F1-scores exceeding 84% across forecasts.