<p>Accurate classification of gullies is essential for effective land management, but identifying these linear erosional features from a coarser spatial resolution image can be challenging. Resampling the spatial resolution of readily available satellites, such as PlanetScope, which is freely accessible for educational and research purposes, can enhance pixel resolution and improve classification accuracy without incurring additional costs. However, this cost-effective approach has yet to be adopted in gully classification, where a finer level of detail is often necessary. This study explored the potential of a resampled PlanetScope product for improving gully classification. More precisely, it evaluated whether resampled imagery outperforms original-resolution imagery and identified the optimal combination of resampling techniques and machine learning classifiers for gully mapping. Four resampling techniques, such as the bilinear interpolation, cubic convolution, majority rule, and nearest neighbor, were applied and their effect on gully classification was assessed using three classifiers: support vector machine (SVM), random forest (RF), and gradient boosting machine (GBM). Fifteen model variations, each defined by a unique combination of spatial resolution, resampling technique, and classifier, were developed and evaluated using an error matrix. Gully classification accuracy generally improved with cubic convolution and nearest neighbor resampling when using SVM and RF. In contrast, resampling did not improve classification accuracy with the GBM classifier. These findings demonstrate the potential of spatial resampling as a cost-effective strategy to enhance gully classification accuracy, particularly when applied with SVM and RF. However, the lack of improvement with GBM suggests that the effectiveness of resampling is classifier-dependent. Therefore, spatial resampling should be prioritized for gully mapping when using SVM and RF to maximize classification accuracy.</p>