Drinking water quality in rural areas is impacted by industrial and agricultural runoff, water treatment infrastructure, and household economic conditions. This study explores the relationship between drinking water quality, water sources, and land cover types in northeastern Kazakhstan. The Water Quality Index (WQI) was calculated for each household and village using the Horton Equation. Land cover was mapped using Sentinel-2 Level-2A imagery. Statistical differences among villages were analyzed through one-way ANOVA and t-tests. A Structural Equation Model (SEM) was built using Maximum Likelihood estimators, with significance set at p < 0.05. Significant variations in manganese, hydrocarbonates, and chlorides were observed based on the distance from the Irtysh River and water sources. Grasslands had the greatest influence on water parameters (−14.89), followed by croplands (5.96), urban lands (2.15), and other land types (2), with forests having the least effect. Biological indicators, such as Actinomycetes sp., were significantly correlated with forests (2.32) and other land cover types. Grasslands reduce mineral content in groundwater, while croplands and forests contribute to mineral enrichment, particularly nitrates from croplands. Urban areas increase chemical loads in groundwater, and manganese levels decrease with distance from the Irtysh River. Chlorides and hydrocarbonates are highest near the river. Rural water treatment infrastructure should be improved, stricter pollution controls should be enforced, and sustainable land use practices should be promoted to reduce agricultural and urban runoff. Additionally, economic incentives for household filtration, regular water quality monitoring, and a coordinated watershed management approach can enhance long-term water security.

It remains challenging to map grassland above-ground biomass (AGB) over Eurasia because in-situ measurements are sparse, grassland types are highly diverse, and grasslands are highly sensitive to climate change and human activities. This study compiled extensive in-situ AGB measurements from the past 20 years and combined them with MODIS imagery, meteorological, soil, and topographic data to map grassland AGB over Eurasia for 2000-2024. We used a stacking ensemble framework to generate Eurasian grassland AGB maps at 500 m resolution for 2000–2024 and quantified pixel-wise errors and map reliability. The optimized stacking ensemble framework improved the accuracy of grassland AGB estimation and reduced estimation bias compared to single machine learning (ML) models, with RMSE reduced by 13.7% and bias reduced by 20.9%. The saturation effect under high AGB conditions was reduced. The 25-year AGB maps had a mean pixel-wise uncertainty of 18.44 g/m², with relative errors below 15% for all grassland types. The error analysis indicates that the ensemble framework has good extrapolation ability, with errors well controlled in regions with sparse in-situ measurements. The AGB maps showed an overall increasing trend in AGB across Eurasia, with an increase in 23.4% of the Eurasian grasslands and a decrease in 3.1% of the region. Semi-humid regions contributed most of the increase, but the instability of these regions in this region also increased. This study produced a long-term and reliable Eurasian grassland AGB dataset, which can be useful for future ecological and climate change studies and inform grassland management.