The hydrochemical analysis method was used to reveal the sources and spatiotemporal variations of carbon and nitrogen elements in the Pingzhai Reservoir, and the C–N coupling cycle and its influence on the karst carbon sink are discussed. The results show the following: (1) The hydrochemical type of the study area is HCO₃-Ca. (2) From the river to the reservoir and then to the reservoir outlet, the values of HCO₃⁻ and δ¹³C_DIC showed an opposite trend. The values of NO₃⁻, δ¹⁵N-NO₃⁻, and δ¹⁸O-NO₃⁻ were different in each stage of the river. (3) HCO₃⁻ mainly comes from the weathering of carbonate rocks and the oxidative decomposition of organic matter. Nitrate mainly comes from chemical fertilizers, soil organic nitrogen, sewage, and livestock manure. (4) The average proportion of HCO₃⁻ produced by HNO₃ dissolving carbonate rock is 8.38%, but this part does not constitute a carbon sink. Compared with rivers, the proportion of HCO₃⁻ and (Ca²⁺ + Mg²⁺) produced by HNO₃ dissolving carbonate rock in reservoir water is relatively large. The input of nitrate not only pollutes the water body with NO₃⁻ but also changes the carbon source/sink pattern of the water–rock interaction.

The concentrations of carbon, nitrogen, and phosphorus in water bodies significantly influence aquatic ecological conditions. By collecting multitemporal hyperspectral data and water quality parameter data from water bodies and through systematic preprocessing of hyperspectral data combined with multimethod sensitive band selection, an optimal spectral feature subset was determined. Within a machine learning framework, multiple combined remote sensing inversion models were constructed to identify the optimal inversion model for each water quality parameter, along with corresponding preprocessing methods and sensitive bands. The results indicate that differential processing of remote sensing reflectance enhances model accuracy. Sensitive band selection effectively eliminates redundant bands, significantly improving the computational efficiency of inversion models. XGBoost demonstrated superior accuracy in constructing 240 water quality parameter inversion models because of its unique algorithmic design. However, model accuracy is not solely determined by algorithmic complexity or predictive capability but rather by the combined effect of algorithm performance and input feature quality. Verification of the inversion model’s generalization ability via an independent dataset demonstrated its capacity for generalization. These findings provide valuable insights for the reliable application of hyperspectral data in aquatic environmental remote sensing and offer support for regional water quality conservation efforts.

Chlorophyll a (Chla), total phosphorus (TP), total nitrogen (TN), and turbidity (Turb) are key indicators for assessing water eutrophication. To overcome the limitations of conventional regression methods, this study developed and compared inversion models for these parameters using Landsat-8 OLI imagery and field data, comparing multiple linear regression and seven machine learning algorithms: Genetic Algorithm- and Particle Swarm-optimized Backpropagation Neural Networks (BPNNs), Convolutional Neural Network (CNN), Extreme Learning Machine (ELM), Random Forest (RF), XGBoost, and Support Vector Regression (SVR). The results revealed that traditional regression performed better for optically active parameters (Chla and Turb) than for non-optically active ones (TP and TN), whereas machine learning models significantly improved accuracy, particularly for TP and TN. The XGBoost model achieved the highest performance (R² > 0.90 for all parameters). Post-calibration analysis further delineated the spatial distributions and inter-parameter correlations in Pingzhai Reservoir, providing a robust method for water quality monitoring and assessment.

In deep karst reservoirs, the internal environment is complex, and thermal stratification isnot the only factor controlling the vertical distribution of the DIC concentration. Previous studies have not fully understood the migration and transformation of DIC in a deep-water reservoir. In this study, a deep-water reservoir in southwest China was chosen, and the spatial and temporal characteristics of the DIC concentration, pCO₂, δ¹³C_DIC value, and SIc were investigated. It was found that the Pingzhai Reservoir is a double temperature leapfrog reservoir. The DIC concentration, pCO₂, Sic, and δ¹³C_DICvalues showed annual cycle variation. During the thermal stratification phase, the DIC concentration, pCO₂, Sic, and δ¹³C_DICvalues were significantly different between the surface layer and the lower layer. However, during the mixing and mixed phases, the differences were not significant. The vertical divergence of the DIC in the Pingzhai Reservoir was influenced by the subtemperate layer, human activities, and sources. The formation of the subtemperate layer was due to the submerged flow formed when river water enteredthe reservoir, which provides a channel for DIC from the river to enter the lower layer of the reservoir. Human activities increased the solubility of carbonate rocks in the reservoir, and the source of DIC was one of the factors contributing to the concentration stratification of DIC in the reservoir.

Studying the temporal and spatial evolution pattern and transformation rule of urban–rural construction land in karst mountainous areas has important guiding significance for urban development boundary planning, red lines for ecological protection, and cultivated land protection. The present study took 46 townships (streets) in Qixingguan District of Guizhou Province, southwest China, as the research area; collected the current status of four-phase land use data in 2009, 2013, 2017, and 2020; and used GIS spatial analysis models and geographical detectors to analyze the temporal and spatial evolution pattern characteristics and influencing factors of urban–rural construction. The results showed the following: (1) Since 2009, the total area of urban–rural construction land has continued to increase; the largest area is rural residential land, followed by urban land and transportation land, with relatively little urban industrial and mining land, scenic spots, and special land. The growth rate of land used for transport increased rapidly, and urban land grew faster than rural residential land. (2) More than 57.72% of the newly increased urban–rural construction land came from cultivated land, but the transformation of cultivated land for construction gradually slowed down; 57.48% of urban–rural construction land was transferred for reclamation as cultivated land. During the study period, the transformation of cultivated land to construction land was more intense (the transfer out of cultivated land was greater than the transfer in by 9541.94 hm²). (3) There are strong spatial differences in the density of urban–rural construction land, showing scattered agglomeration distribution, and the degree of aggregation in medium-high- and high- density areas is further strengthened, expanding to the east and southwest. (4) The growth of urban–rural construction land has been controlled by a variety of complex factors, the most influential of which are the completion of fixed asset investment in society as a whole and the total fiscal revenue, with explanatory power (PD) values of 0.819 and 0.607, respectively. Interactions between detection factors have a greater impact on the spatial differentiation of urban–rural construction land than single factors. The results of this study can provide basic research data and support the control and high-quality development of urban–rural construction land in Qixingguan District and karst mountain areas.

Investigating land cover patterns, changes in carbon stocks, and forecasting future conditions are essential for formulating regional sustainable development strategies and enhancing ecological and environmental quality. This study centers on Guiyang, a mountainous urban area in southwestern China, to analyze the dynamic changes in land cover and their effects on carbon stocks from 2000 to 2035. A carbon stocks assessment framework was developed using a cellular automaton-based artificial neural network model (CA-ANN), the InVEST model, and the geographical detector model to predict future land cover changes and identify the primary drivers of variations in carbon stocks. The results indicate that (1) from 2000 to 2020, impervious surfaces expanded significantly, increasing by 199.73 km². Compared to 2020, impervious surfaces are projected to increase by 1.06 km², 13.54 km², and 34.97 km² in 2025, 2030, and 2035, respectively, leading to further reductions in grassland and forest areas. (2) Over time, carbon stocks in Guiyang exhibited a general decreasing trend; spatially, carbon stocks were higher in the western and northern regions and lower in the central and southern regions. (3) The level of greenness, measured by the normalized vegetation index (NDVI), significantly influenced the spatial variation of carbon stocks in Guiyang. Changes in carbon stocks resulted from the combined effects of multiple factors, with the annual average temperature and NDVI being the most influential. These findings provide a scientific basis for advancing low-carbon development and constructing an ecological civilization in Guiyang.