Assessing the accuracy of remotely-sensed fire datasets across the Southwestern Mediterranean basin

Abstract

Abstract. Recently, many remote-sensing (RS) based datasets providing features of individual fire events from gridded global burned area products have been released. Although very promising, these datasets still lack a quantitative estimate of their accuracy with respect to historical ground-based fire databases. Here, we compared three state-of-the-art RS datasets (Fire Atlas, FRY and GlobFire) with high-quality ground databases compiled by regional fire agencies (AG) across the Southwestern Mediterranean basin (2005–2015). We assessed the spatial and temporal accuracy in estimated RS burned area (BA) and number of fires (NF) aggregated at monthly and 0.25° resolutions, considering different individual fire size thresholds ranging from 1 to 500 ha. Our results show that RS datasets were highly correlated with AG in terms of monthly BA and NF but severely underestimated both (by 38 % and 96 %, respectively) when considering all fires > 1 ha. Stronger agreement was found when increasing the fire size threshold, with fires > 100 ha denoting higher correlation and much lower error (BA 10 %; NF 35%). The agreement between RS and AG was also the highest during the warm season (May to October) in particular across the regions with greater fire activity such as the Northern Iberian Peninsula. The Fire Atlas displayed a slightly better performance, with a lower relative error, although uncertainty in gridded BA product largely outpaced uncertainties across the RS datasets. Overall, our findings suggest a reasonable agreement between RS and ground-based datasets for fires larger than 100 ha, but care is needed when examining smaller fires at regional scales.

Pyrogeography across the western Palaearctic: A diversity of fire regimes

“…In the future, this limitation will be even more relaxed. In relationship to the different sources, I evaluated fire regime parameters from a range of sources (Supporting Information Table S3), and the results were consistent, suggesting that this is not a key issue, at least at the scale of the present study (see also Artés et al, 2019; Galizia et al, 2021). Small fires are likely to be underrepresented (Galizia et al, 2021), but they are also the least important.…”

“…In relationship to the different sources, I evaluated fire regime parameters from a range of sources (Supporting Information Table S3), and the results were consistent, suggesting that this is not a key issue, at least at the scale of the present study (see also Artés et al, 2019; Galizia et al, 2021). Small fires are likely to be underrepresented (Galizia et al, 2021), but they are also the least important. Finally, remote sensing data are likely to fail to capture underground (peat) fires that occur mainly in Tundra (Turetsky et al, 2011), hence this type of fire needs to be added to the present results to gain a full understanding of fires in this ecoregion.…”

“…GlobFire is the most recently published and has the most years of data. It also includes a comparison with precedents and validations with independent data, and considerable agreement was found among the databases (see also Galizia et al, 2021), but some biases in the precedents were also found (e.g., the problem of using tildes that split fires in FireAtlas). In addition, GlobFire provides the perimeter of all fires in a GIS format, enabling me to study fire overlaps (hence recurrence and pyrodiversity; see below).…”