Forecasting Zoonotic Cutaneous Leishmaniasis Risk Through Integrated Simulated Remote-Sensing Data and Epidemiological Records

ABSTRACT

Background and objectives:

Zoonotic cutaneous leishmaniasis (ZCL), one of the most significant vector-borne parasitic diseases, poses a major public health challenge in many semi-arid and warm regions of the world, particularly in the Middle East and Iran. The disease follows a distinct spatiotemporal pattern that is strongly influenced by environmental and ecological factors such as temperature, vegetation cover, and soil moisture. In recent years, climate change, habitat degradation, and the expansion of human settlements into environmentally sensitive areas have altered the distribution of vectors and reservoir hosts, thereby increasing the risk of emergence and expansion of new transmission foci. Consequently, anticipatory estimation of future disease risk plays a critical role in designing targeted interventions, optimizing resource allocation, implementing preventive planning, and establishing early-warning systems. The present study aimed to predict the spatiotemporal risk of ZCL in Ilam Province, western Iran, up to the year 2030, using simulated remote-sensing data and available epidemiological records.

Materials and methods:

In the initial phase, a total of 5,353 reported ZCL cases from 2014 to 2019 were subjected to data cleaning, quality control, and georeferencing procedures to construct a baseline disease-intensity map within a Geographic Information System (GIS) framework. This map served as the primary epidemiological layer reflecting the historical spatial distribution of the disease. Subsequently, three key environmental predictors—Soil-Adjusted Vegetation Index (SAVI), Land Surface Temperature (LST), and Soil Moisture Index (SMI)—were derived from Landsat-8 imagery for the 2014–2024 period, using the Google Earth Engine platform. Pixel-wise time series spanning ten years were analyzed through linear regression to characterize temporal trends for each variable. Based on these trends, pixel-level values were extrapolated to 2030, and simulated future environmental layers were generated. The cumulative epidemiological surface and the projected ecological layers were then integrated as input features within a Support Vector Machine (SVM) framework to produce a spatially explicit ZCL risk-prediction model for 2030. Model performance was rigorously assessed using multiple statistical metrics, including the Root Mean Square Error (RMSE), Mean Absolute Error (MAE), Area Under the Receiver Operating Characteristic Curve (AUC), Cohen’s Kappa coefficient, overall accuracy, and k-fold spatial cross-validation. Furthermore, a model-uncertainty map was constructed based on the standard deviation of ensemble predictions and visualized as an independent spatial layer.

Results and discussion:

The results demonstrated that the proposed framework exhibits a high capability in capturing and representing the spatiotemporal patterns of ZCL risk. The SVM model achieved reliable performance (RMSE ≈ 18.9, MAE ≈ 7.5, AUC ≈ 0.81, Kappa ≈ 0.68, and an overall accuracy of approximately 0.91), indicating strong discriminative power between high-, moderate-, and low-risk areas. The risk map projected for 2030 revealed a higher concentration of risk in the dry, warm, and low-altitude regions of the southern and western parts of Ilam Province, whereas the higher-elevation and mountainous regions in the northern, central, and eastern parts were predominantly classified as low-risk zones. Co-location analysis between environmental layers and the risk map indicated a significant association between increasing land surface temperature, decreasing vegetation greenness, reduced soil moisture, and the intensification and expansion of high-risk areas. Furthermore, the uncertainty map demonstrated relatively low levels of uncertainty in most critical zones, with slightly higher values in marginal and transitional areas, confirming the robustness and stability of the model predictions.

Conclusion:

The findings of this study indicate that integration of simulated remote-sensing data, spatial epidemiological information, and advanced machine-learning algorithms provides a powerful, efficient, and generalizable framework for predicting the future risk of zoonotic cutaneous leishmaniasis in endemic regions. The projected expansion of high-risk clusters toward the southern parts of Ilam Province by 2030 highlights the urgent need for continuous environmental monitoring, targeted interventions in critical hotspots, and the development of climate-adaptive preventive strategies. The proposed framework can serve as a scientific foundation for establishing intelligent monitoring and early-warning systems for environmentally driven diseases and can play a significant role in supporting data-driven policymaking, optimized resource management, and the enhancement of public-health resilience.