Agricultural water impoundments are crucial for agricultural production in arid regions, but their small size has hindered large-scale monitoring with traditional remote sensing techniques. This study presents a tailored mapping framework for annual monitoring of agricultural water impoundments (AWIs) in arid regions, where conventional methods fail due to their field-scale size. By introducing a novel mapping framework that couples inter-annual water persistence analysis with vector-derived morphometric features within an ensemble learning architecture, our approach overcomes the limitations of conventional remote sensing methods and accurately distinguishes AWIs from natural water bodies (overall accuracy = 0.89)-enabling the first high-resolution (10 m), multi-year inventory (2018-2023) across the Hexi Corridor. In 2023, we identified 2,037 impoundments, with their areas ranging from 1000 to 10000 m2. In addition, 26.5% of them were constructed in 2022, when large-scale AWIs became construction priority, accounting for 63.5% of the newly-built AWIs that year. Furthermore, for the first time, three distinct types of river-reservoir-AWI distribution patterns in Hexi Corridor have been identified through an upstream-downstream analysis. AWIs, closer to rivers than wells, serve as immediate irrigation distribution points for riverine water sources. Our study provides a new perspective for monitoring agricultural water infrastructure in arid regions, offering critical insights into sustainable water resource management.