Ju Zhao

A key challenge in precision agriculture is acquiring reliable spatial soil information under varying sampling densities, from sparse surveys to intensive monitoring. The individual predictive soil mapping (iPSM) method performs well in data-scarce conditions but neglects spatial proximity, limiting its predictive accuracy where spatial autocorrelation exists. To overcome this, we developed an enhanced framework, iPSM-Spatial Distance (iPSM-SD), which systematically integrates spatial proximity through multiplicative (MUL) and additive (ADD) strategies. The framework was validated using two contrasting cases: sparse soil organic carbon density data from Yunnan Province (n = 118) and dense soil organic matter data from Bayi Farm (n = 2511). Results show that the additive model (iPSM-ADD) significantly outperformed the original iPSM and benchmark models, including random forest, regression kriging, geographically weighted regression, and multiple linear regression, under sufficient sampling, achieving an R² of 0.86 and reducing RMSE by 46.6% at Bayi Farm. It also maintained robust accuracy under sparse sampling conditions. The iPSM-SD framework thus provides a unified and adaptive tool for digital soil mapping across a wide range of data availability, supporting scalable soil management decisions from regional assessment to field-scale variable-rate applications in precision agriculture.

Recognizing the evolution of landscape patterns in the Chishui River source region is essential for protecting ecosystems and sustainable growth in the Yangtze River Basin and other similar areas. However, knowledge of landscape pattern evolution within the primary channel zone remains insufficient. To address this gap, the current study used 2000–2020 land-use, geography, and socio-economic data, integrating landscape pattern indices, land-use transfer matrices, dynamic degree, the GeoDetector model, and the PLUS model. Results revealed that forest and cropland remained the prevailing land-use types throughout 2000–2020, comprising over 85% of the landscape. Grassland had the highest dynamic degree (1.58%), and landscape evolution during the study period was characterized by increased fragmentation, enhanced diversity, and stable dominance of major forms of land use. Anthropogenic influence on different landscape types followed the order: construction land > cropland > grassland > forest > water bodies. Land-use change in this region is a complex process governed by the interrelationships among various factors. Scenario-based predictions demonstrate pronounced variability in various land types. These findings provided a more comprehensive understanding of landscape patterns in karst river source regions, provided evidence-based support for regional planning, and offered guidance for ecological management of similar global river sources.