Resumo

CHAMPO JIMENEZ, Omar; URIBE SALAS, María Dolores  e  ESPANA BOQUERA, María Luisa. Comparative validation of PROBA-V/MODIS/Sentinel-2 burn area detection in Michoacan, Mexico (2015-2020). Invest. Geog [online]. 2023, n.112, e60774.  Epub 22-Jan-2024. ISSN 2448-7279.  https://doi.org/10.14350/rig.60774.

Fires are a major environmental concern due to their potentially adverse effects on ecosystems, soil, and the atmosphere. Fires are increasing worldwide as a result of climate change. Field reports are the closest data on active fires and burned areas; however, they are not always reliable or sufficient, so global series generated from satellite information are useful. Satellite products available for burned areas include the 300m Copernicus PROBA-V series (2014-2020), which calculates indices that detect significant vegetation changes, and NASA’s MCD64A1 500m MODIS collection, based on the detection of heat spots. This study aimed to compare these products in the state of Michoacán, Mexico, where temperate and tropical dry forests (or low deciduous forests) are located, as well as cropland and other vegetation types; the study period covered from 2015 to 2020.

For each series, the pixels with a burned-area report were quantified by month and year for each vegetation cover, and matches were analyzed. To illustrate some cases, Sentinel-2 images were used as an independent baseline with a higher resolution than the compared series, and the dNBR index was calculated to assess the impact level.

During the study period, the number of PROBA-V detections was 16 times greater than that of MODIS (308699 versus 19061). PROBA-V identified burned areas from December to May, while MODIS did so in April and May, which better matched the actual conditions in the study area. In the PROBA-V series, 56.65 % of the detections corresponded to tropical dry forest, 27.25 % to agriculture, and 13.57 % to temperate forest; in MODIS, 50.02 % of the area was cropland, 31.31 % temperate forest, and only 17.31 % tropical dry forest. Between 37.90 % (2015) and 78.72 % (2020) of the burned areas noted by MODIS were also detected by PROBA-V, but MODIS detected less than 5 % of the areas identified by PROBA-V. In the tropical dry forest, the correlation between both systems was below 2 % in any year, indicating a significant overdetection by PROBA-V in this type of vegetation. The six burned areas visualized in Sentinel-2 images from January to June show that they usually occur in April or May in the transition zone between temperate and dry forests or until June in areas near avocado plantations. On the other hand, cropland fires occur at different times of the year. MODIS was generally more accurate than PROBA-V in the study area; the dNBR index correctly identified high-impact burned areas but overdetected low-impact areas.

The burned-area series facilitates general and systematic monitoring of fires in a territory, while archival data are suitable for historical studies. The accuracy of each series depends on the data and algorithms used to generate it, as well as on the characteristics of the study area. Burned-area series based on thermal anomalies, such as the MCD64A1 collection of MODIS, are more accurate than those identifying changes in vegetation, such as PROBA-V, particularly when deciduous vegetation is found in the area. However, false detections from highly reflective surfaces, such as clouds or water bodies, should be eliminated. Additionally, to study forest fires, it is necessary to overlay land-use maps.

The information provided by the series does not include fires that are small relative to the pixel size. If a detailed study of fires in small areas is required, the information in the series can be supplemented with field data and high-resolution satellite imagery, for example, to assess the impact of fire or to monitor the recovery process. In this case, the image needs to be classified beforehand to identify the burned areas.

Although the MODIS and PROBA-V satellites are approaching the end of their useful life, others are already taking over, both by NASA and ESA, so the availability of time series for burned area and other biophysical and climatic variables is warranted, recording increasingly higher-quality data over longer periods.

Palavras-chave : temperate forest; tropical dry forest; Copernicus; dNBR; time series.

        · resumo em Espanhol     · texto em Espanhol     · Espanhol ( pdf )