<p>There are large omission errors in the estimation of burned area in map products that are generated at a global scale. This error is then inherited by other models, for instance, those used to report Greenhouse Gas Emissions using a “bottom up” approach. This study evaluates temporal methods to improve burned area detection using Landsat 5-TM and 8-OLI. In this process, the normalized burn ratio (NBR) was used to highlight burned areas and thresholds to classify burned and non-burned areas. In order to maximize the burned area detection two alternatives to the temporal dNBR method were evaluated: the relative form of the temporal difference RdNBR and the use of time series metrics. The processing, algorithm development and access to Landsat data was made on the Google Earth Engine GEE platform. Three regions of Latin America with large fire occurrence were selected: The Amazon Forest in Colombia, the transition from Chiquitano to Amazon Forest in Bolivia, and El Chaco Region in Argentina. The accuracy assessment of these new products was based on burned area protocols. The best model classified 85% of burned areas in the Chiquitano Forests of Bolivia, 63% of the burned areas of the Amazon Forests of Colombia and 69% of burned areas in El Chaco of Argentina.</p>

Accurate reference data to validate burned area (BA) products are crucial to obtaining reliable accuracy metrics for such products. However, the accuracy of reference data can be affected by numerous factors; hence, we can expect some degree of deviation with respect to real ground conditions. Since reference data are usually produced by semi-automatic methods, where human-based image interpretation is an important part of the process, in this study, we analyze the impact of the interpreter on the accuracy of the reference data. Here, we compare the accuracy metrics of the FireCCI51 BA product obtained from reference datasets that were produced by different analysts over 60 sites located in tropical regions of South America. Additionally, fire severity, tree cover percentage, and canopy height were selected as explanatory sources of discrepancies between interpreters’ reference BA classifications. We found significant differences between the FireCCI51 accuracy metrics obtained with the different reference datasets. The highest accuracies (highest Dice coefficient) were obtained with the reference dataset produced by the most experienced interpreter. The results indicated that fire severity is the main source of discrepancy between interpreters. Disagreement between interpreters was more likely to occur in areas with low fire severity. We conclude that the training and experience of the interpreter play a crucial role in guaranteeing the quality of the reference data.