Air temperature (Ta) is a fundamental variable in climatological and hydrological research. However, surface-based observations are often limited in regions like the Cananéia-Iguape Coastal System (CICS), Brazil. This study estimated monthly Ta using MODIS Aqua and Terra Land Surface Temperature (LST) and vegetation indices from 2007 to 2022. Two modeling approaches were used: Multiple Linear Regression (MLR) and Random Forest (RF). Validation results indicated that RF models outperformed MLR, achieving a Root Mean Square Error (RMSE) of approximately 1.00 °C and a Mean Absolute Error (MAE) of 0.68 °C. MODIS Aqua models provided more accurate estimates, likely due to the 13:30 local overpass time. These findings offer a robust alternative for generating Ta spatial datasets in regions with sparse meteorological networks. This visual summary serves as a pivotal entry point into the research, offering a concise overview of the study’s core findings on air temperature (Ta) estimation in the Cananéia-Iguape Coastal System (CICS), Brazil. Designed to provide a rapid understanding of the methodology, the infographic illustrates the integration of data from meteorological stations, MODIS Aqua/Terra Land Surface Temperature (LST), vegetation indices (NDVI/EVI), and geographic controls from 2007 to 2022. The central flow depicts the analytical comparison between Multiple Linear Regression (MLR) and Random Forest (RF) models. The visual results highlight that the Random Forest algorithm achieved superior performance with lower MAE and RMSE metrics compared to MLR. Furthermore, the graphics emphasize that MODIS Aqua models outperformed Terra, identifying ‘Month’ and ‘LSTmean’ as the most influential predictors. Finally, the conclusion panel visually reinforces the study’s main takeaway: satellite-based modeling is a reliable, low-error solution for filling meteorological data gaps in coastal regions with sparse monitoring networks, with potential for future improvements. Random Forest outperformed Multiple Linear Regression regarding accuracy (RMSE ≈ 1.00 °C). Seasonality (Month) and LST were the most influential variables, outweighing topographic factors. MODIS Aqua models outperformed Terra, likely due to alignment with daily maximum temperatures. Including “Month” as a categorical factor was essential for correcting systematic seasonal bias. The method provides a robust solution for filling data gaps in regions with sparse networks.
Baratto et al. (Mon,) studied this question.