Characterization of gully erosion by remote sensing in Haiti: The case of the Grande Rivière du Nord watershed

Gully erosion is a widespread form of land degradation throughout the world. In Haiti, it is primarily driven by deforestation and erosive farming practices. Natural factors such as heavy rainfall, rugged terrain, and soil characteristics further exacerbate gully erosion process. Gully erosion results in significant soil loss and increased transported sediment load, which has a negative impact on soil quality and causes damage to downstream infrastructures. Therefore, accurate mapping of gullied areas is crucial for erosion mapping, monitoring, and control, particularly in the Grande Rivière du Nord watershed, which is highly susceptible to water erosion. In this study, an object-based approach was developed to map gully-affected areas in the Grande Rivière du Nord watershed using high-resolution Pléiades satellite imagery. The methodology involved data preparation, a field survey in Haiti, and remote sensing data processing. Various metrics such as spectral indices (NDVI, SAVI, NDSI) and elevation-derived data were computed using eCognition software. Image segmentation was performed using a scale parameter of 100, combined with shape (0.7), compactness (0.5), and spectral indices. For object-based image classification, both Decision Trees (DT) and Random Forest (RF) models were used. The results showed that the RF model outperformed the DT model in terms of overall accuracy (86%), precision (94%), and F-score (92%). The estimated extent of gullied areas using RF was 3.4 km2, corresponding to 3.2% of the total study area of 106 km2. Spectral indices were found to be the most influential features for gully mapping, followed by shape features and the textural homogeneity index. Elevation data were useful in image segmentation but did not contribute significantly to classification results. By combining field surveys and remote sensing data acquisition, the approach developed can be replicated for mapping gully erosion in the Grande Rivière du Nord watershed. However, it may need to be adapt before application in other regions. The results highlight the importance of remote sensing data and object-based analysis for accurate gully mapping in Haiti, providing valuable information for erosion monitoring and management.