Unveiling the spatiotemporal patterns and predictive associations of human settlement quality in the Hunan-Jiangxi peripheral region

Peripheral regions remain underrepresented in analyses of between-county disparities in human settlements. Using 24 counties in the Hunan-Jiangxi peripheral region, this study integrates Spatial Markov Chains (SMC), Boosted Regression Trees (BRT), and the Optimal Parameters-based Geographical Detector (OPGD) to examine HSQ patterns and predictive spatial associations in 2010, 2015, 2020, and 2022. Results indicate that mean HSQ increased by 90.75%; the cross-county standard deviation increased by 68.03%, whereas the Gini coefficient declined from 0.182 to 0.167. Alternative weighting schemes yielded highly concordant HSQ scores (Pearson r = 0.919–0.943) and preserved the direction of the temporal trend. In the primary SMC analysis of two equal five-year transitions, 56.25% of transitions moved upward, 37.50% persisted, and 6.25% moved downward; the neighborhood-conditioned permutation test was not significant ( p = 0.8254). The separate 2020–2022 sensitivity analysis showed similarly upward or persistent mobility, but its significant neighborhood-conditioned result ( p = 0.0065) was based on a short interval and sparse conditional cells and was therefore interpreted cautiously. Residential life factors had the largest category-level predictive importance (50.17%); Tran (47.37%), Reta (25.52%), and PM2.5 (8.87%) had the largest individual point estimates, with overlapping county-cluster bootstrap intervals. Annual OPGD analyses found a significant factor-level association for Tran in 2020 (q = 0.55, p = 0.01) and 2022 (q = 0.50, p = 0.02); pairwise interactions were generally enhanced but were treated as exploratory. These findings provide study-area-specific empirical evidence for adaptive urban-rural planning in comparable interprovincial or administratively fragmented peripheral regions, while external application requires local validation.