Integrating remote sensing, GIS, and machine learning for zoonotic cutaneous leishmaniasis modelling

Zoonotic Cutaneous Leishmaniasis (ZCL) is a vector-borne disease (VBD) characterized by distinct spatiotemporal patterns. Accurate evaluation of ZCL risk patterns necessitates the utilization of comprehensive epidemiological and ecological data. This study proposes a hybrid model that integrates the advantages of geographic information systems (GIS) for trend analysis, remote sensing (RS) for environmental data extraction, and machine learning (ML) for ZCL risk assessment in Ilam Province, Iran. Utilizing data from 2014 to 2019, spatial and temporal patterns are investigated using Moran’s I, Getis-Ord Gi* statistics, and the Mann-Kendall (MK) test, while high-risk ZCL maps are generated through Extreme Gradient Boosting (XGBoost) and Random Forest (RF) models. The proposed model harnesses high-precision disease and environmental geospatial monitoring to address limitations of previous systems through robust, data-driven insights. The results reveal significant patterns, with a Moran’s I statistic of 0.68 (p < 0.01) and MK values of –2.254 for annual data (p = 0.024) and 3.340 for monthly data (p = 0.001). Temporal analysis indicates a declining trend, with peak incidence observed in late fall and early winter. Consequently, due to the incubation period, the critical infection period occurs during summer. The risk maps demonstrate high levels of accuracy (area under the curve of 0.96 for RF and 0.98 for XGBoost), pinpointing high-risk areas in the western and southern hot deserts and low-risk regions in the northeastern mountainous areas. Moreover, there is an increasing trend in high-risk zones, corresponding to rising temperatures across different cities and seasons. These findings highlight a significant relationship between ZCL spread and temperature-related factors, offering valuable insights for future research.