Abstract: Background Terraced fields serve as vital agricultural resources but are prone to erosion during extreme rainfall events, leading to the formation of severe gullies. Traditional gully volume measurement methods are often time-consuming, labor-intensive, and costly. With the widespread application of drone technology, it has become feasible to establish relationships between gully morphology and volume based on aerial imagery. This study aims to develop a reliable model for estimating gully volume in terraced fields to facilitate rapid and accurate monitoring of erosion during heavy rainfall. Methods Following an extreme rainfall event on July 24th, 2025, in Xin'anbian Town, Dingbian County, Yulin City, Shaanxi Province, UAVs deployed to capture DOM (digital orthophoto map) of the Xuetianchi small watershed. Using stratified random sampling approach, 25 gullies were selected for analysis. Morphological parameters—including length, perimeter, projected area, and surface area—were extracted through visual interpretation, while gully volumes were measured with RTK (Real - time kinematic). An empirical model was established to relate morphological features to volume. Results The results show that: 1) Based on developmental location and morphology, terraced-field gullies can be classified into three types: field surface- ridge curved type, field surface- ridge straight type, field ridge straight type. Curved gullies were the most numerous (accounting for 56%), while straight gullies on the ridge exhibited the largest average volume (34.05 m³). 2) Among common morphological parameters, both projected area and surface area performed well in estimating gully volume, with surface area providing the best fit. The corresponding models were V = 0.28A^1.374 (R² = 0.95, NSE = 0.89) for projected area and V = 0.24 As^1.379 (R² = 0.95, NSE = 0.93) for surface area. Surface area is therefore recommended for gully volume and erosion estimation in Loess Hilly-Gully Region. 3) Predictive performance of area-volume empirical models varied notably across regions, with differing sensitivity to area changes. Therefore, it is necessary to develop empirical models for gully volume estimation based on the specific characteristics of each region.ConclusionThis study improves the gully erosion prediction model for loess hilly and gully regions. The established morphology-volume relationship offers a dependable approach for accurate gully erosion estimation and systematic assessment in terraced fields, thereby supporting effective soil and water conservation strategies.