The purpose of this study is to describe spatial distribution of soil organic carbon (SOC) in mangroves of arid environment and its variability related to the species of mangrove using aerial imagery, supervised classification, and Generalized Additive model (GAM). Samples of soil were analyzed to quantify SOC. Aerial images were acquired using an unmanned aerial vehicle and two cameras (RGB [Red, Green, and Blue] and GRN [Green, Red, and Near-Infrared]). Random Forest was used to classify study area into eight classes (including three mangrove species: Rhizophora mangle, Laguncularia racemosa, and Avicennia germinans). Generalized Additive Model was used to describe relationship between SOC and predictor variables (mangrove’s species, height, and distance to water), then to predict SOC in the study area using predicted distribution of mangrove species from RF and distance to water. Estimated SOC reserve in the first 15 cm of sediment ranged between 96.95 and 125.43 Mg C in the study area. The highest SOC were present between 20 and 60 m from water bodies. Rhizophora manglehad higher SOC followed by Laguncularia racemosaand Avicennia germinans. A gradient of distribution by species was related to distance to the water in the following order: R. mangle, L. racemosa, and A. germinans. Both cameras (RGB and GRN) can be used to classify the study area with a high accuracy (> 90%). The methodology proposed was effective for describing the spatial variability of SOC in mangroves of arid environment. This method can be easily implemented in other ecosystems and represents a low cost and time procedure.