Vegetation fire is one of the most significant ecological hazards, exerting substantial impacts on ecosystem structure and functioning. This study aimed to map burned areas resulting from a wildfire event in Golestan Province, Iran, through the integration of multi-sensor remote sensing data and machine learning algorithms within the Google Earth Engine platform. The datasets used included Sentinel-2 imagery, the MODIS burned area product (MCD64A1), and a suite of spectral and differential indices, namely NDVI, NBR, GNDVI, dNDVI, dNBR, dGNDVI, MIRBI, BAI, and RBR, derived for pre-fire, fire, and post-fire periods. For classification, two machine learning algorithms, Random Forest (RF) and Gradient Tree Boosting (GTB), were employed. Their performances were evaluated using a confusion matrix, overall accuracy (OA), and the Kappa coefficient. The results demonstrated that both models achieved a high classification performance, with an overall accuracy of 0.9960 and a Kappa coefficient of 0.9921, indicating excellent agreement with reference data. The analysis of spectral indices revealed a pronounced reduction in vegetation cover following the fire event, with dNBR identified as the most sensitive indicator for discriminating burn severity. Furthermore, the Random Forest algorithm produced more stable outputs, whereas Gradient Boosting showed higher sensitivity to boundary and transitional pixels. Overall, the findings indicate that the integration of Sentinel-2 and MODIS data, combined with spectral indices and machine learning algorithms, provides a robust and reliable approach for rapid burned area detection and post-fire impact assessment.