A Dynamic Matching Algorithm (DMA) within an Integrated Evaluation Framework: Artificial Precipitation Enhancement and Its Effect on Vegetation Change

Accurate evaluation of artificial precipitation enhancement (APE) is vital for optimizing operational strategies and managing water resources in arid regions. However, traditional fixed-domain evaluation methods are limited by static target-control domains and reliance on spatiotemporally averaged data, which obscure localized, dynamic seeding effects amid natural precipitation variability. This study developed a dynamic matching algorithm (DMA) to construct weather-driven adaptive domains and use optimized bootstrap testing. In a localized comparative experiment, the DMA increased the mean target-control Spearman correlation from 0.90 to 0.96 and reduced the RMSE of natural precipitation prediction by 31.8% relative to a fixed-domain scheme. Applied to 1693 seeding operations in the Hexi Corridor during 2019–2022, the evaluation revealed a mean per-operation absolute increase of 0.47 mm in APE and a relative increase of 18.0% in operational efficiency. Leveraging the DMA’s event-scale resolution, analysis identified the relationship between operational frequency and efficiency, and a spatially differentiated strategy of increasing frequency in the Shule River basin was proposed, while introducing effectiveness monitoring and refining seeding windows in southeastern high-output areas. Furthermore, APE generated a 3.5% mean annual net increase in the Normalized Difference Vegetation Index across 85.9% of the Hexi Corridor. Based on the uncertainty bounds quantified by the DMA, findings indicate that APE might need to be coupled with in situ ecological restoration measures in environments that are severely moisture-constrained to counter the diminishing marginal efficacy of increasing operational frequency. The proposed integrated framework effectively enhances the accuracy and robustness of APE evaluation, providing a spatially explicit tool for strategic planning. Its computational efficiency and structural scalability support its application in nonrandomized APE programs.