(1) Background: The recent decade appears to be the hottest since the beginning of modern measurements. Changes in climate patterns related to extreme events and disturbances in forest ecosystems are well documented. Six prominent protected areas (PAs), mountainous forest ecosystems in Serbia, were assessed from the perspective of species potential distribution and vulnerability. (2) Methods: Seven different machine learning models were employed, evaluated using AUC, the maximum F-measure, and TSS and joined into an ensemble model for each of the eight tree species/groups taken from the National Forest Inventory. Representatives from four groups of environmental variables were included: 1. climate (Ellenberg’s Climate Quotient), 2. soil (soil organic carbon), 3. topography (elevation), and 4. remotely sensed indices (NDVI). Future climate was derived from four scenarios (SSP1-2.6, SSP2-4.5, SSP3-7.0, and SSP5-8.5). Stable/gain/loss areas and species vulnerability were calculated with a focus on the end of the 21st century. (3) Results: By the 2090s, generally, contraction of Silver fir, Norway spruce, and European beech is expected, together with the promotion of Downy oak and Sessile oak, in all climate scenarios at all PAs. Two high-mountain PAs expect to see promotions in average forest suitability, one PA both a promotion and a reduction in two scenarios, and three PAs reductions in forest ecosystems in general. (4) Conclusions: National parks “Kopaonik” and “Tara” appear to be the least endangered, followed by “Golija”, while “Stara planina”, “Djerdap”, and “Fruska gora” are expected to experience overall reductions in forest habitats.

Benchmarking deep learning forecasters against classical and physically based numerical baselines remains uncommon in the time-series forecasting literature. Meteorological station networks offer an under-exploited evaluation environment, uniquely providing a physically based climate-model comparator alongside standard baselines. We evaluated eight forecasting approaches—climatology, SARIMA, Random Forest, and five deep learning architectures (TFT, N-HiTS, PatchTST, TiDE, xLSTM)—against bias-corrected output from a five-member CMIP6 ensemble, on 100 meteorological stations across four Western Balkan countries (monthly temperature and precipitation, 1961–2020), using non-parametric significance testing, a rolling-origin backtest (five windows, 2011–2020), and a five-seed robustness check. For temperature, all five deep learning architectures achieved lower MAE than the classical baselines (p < 10⁻⁹⁹), though PatchTST’s advantage over climatology was not significant; the best-performing architecture varied across seeds and evaluation windows, so we characterise a leading cluster (N-HiTS, TFT, TiDE, PatchTST) rather than a single winner. The primary temperature advantage was geographically broad-based, while the comparison against the physical-model baseline was robust to the choice of comparator GCM. For precipitation, by contrast, a simple climatological-mean baseline outperformed all five deep learning architectures with no exception across all five rolling-origin windows. The deep learning advantage over classical and physical baselines is thus variable-specific rather than universal. Meteorological station networks, combined with a physically based climate-model comparator, constitute a well-suited evaluation environment for the broader time series forecasting community.

The mesic grasslands of the Molinio-Arrhenatheretea Tx. 1937 in Vojvodina could play a crucial role in biodiversity conservation, but also in local economies, providing essential ecosystem services, such as habitats for diverse species and resources for agricultural and pastoral activities. However, they face growing threats from unsustainable land use, urbanization and climate change. In this study, a database comprising 716 relevés and 636 plant species was created. All meadow plots were classified into seven habitat types and evaluated for their sustainable use potential using the index developed in this study, based on economically notable species, their status of protection and total cover. Through this index, moist or wet mesotrophic to eutrophic pasture demonstrates the highest potential, whereas temperate and boreal moist or wet oligotrophic grassland shows the lowest. This index offers a decision-support tool, optimizing economic benefits while minimizing environmental impact and offering guidelines for sustainable grassland management and policy recommendations tailored to local conditions. It also serves as a framework for other regions facing similar challenges, contributing to the advancement of grassland ecosystem service valuation and its preservation.