Frédéric Frappart (INRAE, UMR1391 ISPA, 33140 Villenave d’Ornon, France, France); Luc Bourrel (GET, UMR 5563, Université de Toulouse, CNRS/IRD/OMP/CNES, Toulouse, France., France); Bertrand Ygorra (INRAE, UMR1391 ISPA, 33140 Villenave d’Ornon, France, France); Cassandra Normandin (INRAE, UMR1391 ISPA, 33140 Villenave d’Ornon, France, France); Luis Huaraca (SCA, Ministerio de Ambiente Agua y Transición Ecológica, Quito, Ecuador, Ecuador); Jean-Pierre Wigneron (INRAE, UMR1391 ISPA, 33140 Villenave d’Ornon, France, France)

The Surface Water Ocean Topography (SWOT) mission was originally launched to provide continuous spatio-temporal monitoring of water levels on oceans and land with unprecedented spatial resolution that surpasses the limitations of conventional altimetric missions (Fu et al., 2024). This study shows that the advanced capabilities of SWOT pave the way for new applications. Thus, we showed that SWOT data is also sensitive to LULC changes in Pacific Coast of Ecuador. The three study areas chosen according to a north-south gradient allow to represent climatic and landscape diversity. In the north, with a humid tropical climate and limited human presence, the area is home to mostly wet tropical forests (Area A). In the middle of the Pacific Coast, the region is home to seasonal tropical forests, more affected by anthropogenic pressures characterized mainly by the presence of the Peripa Daule hydroelectric dam (Area B). The last zone in the south, largely anthropized, is characteristic of the cultivated areas of the Pacific coast of the country (Area C). In this study, we analyzed the backscatter coefficient (noted Sig0, σ0) available in the SWOT gridded product at 100m spatial resolution for the year 2024. First, the SWOT data was pre-processed to obtain two layers: the number of occurrences of each pixel and the average value of the backscatter coefficient over the year 2024. Then, these two layers are used as input of a machine learning model (Support Vector Machine - SVM) to classify the SWOT detections. The SWOT detections were automatically classified with an SVM into three categories: city (1), river (2) and no forest (3) for Area A and four categories: city (1), river (2), road (3), cropland (4) for Area B and C. Finally, this classification of the SWOT data was then compared to three other data sources: LULC maps from the Ecuadorian Ministry of the Environment (MAATE, 2017), the RADD alerts (Reiche et al., 2021) and output results of the CuSum change detection algorithm (Ygorra et al., 2021). The first results show a good correlation with the maps of the Ministry with accuracy between 0.81 and 0.90 depending on the area considered. The results of the data classification also show that SWOT observations can provide information on agricultural parcel boundaries and road networks (Area C), on the deforested areas and the opening of deforestation fronts mainly along communication routes (Zone A), as well as on the state of agricultural parcels (Zone B). The results also highlight SWOT detections in areas affected by large and difficult-to-access topography, which would make it possible to fill out a lack of information on LULC changes in these regions. The SWOT data could therefore be a complementary source of information to existing LULC change products.



References:

Fu, L. et al. 2024. The Surface Water and Ocean Topography Mission: A Breakthrough in Radar Remote Sensing of the Ocean and Land Surface Water. Geophysical Research Letters 51(4), p. e2023GL107652. doi: 10.1029/2023GL107652.

Ministerio del Ambiente. 2017. Deforestación del Ecuador continental periodo 2014-2016. Quito-Ecuador.

Reiche, J. et al. 2021. Forest disturbance alerts for the Congo Basin using Sentinel-1. Environmental Research Letters 16(2), p. 024005. doi: 10.1088/1748-9326/abd0a8.

Ygorra, B. et al. 2021. Monitoring loss of tropical forest cover from Sentinel-1 time-series: A CuSum-based approach. International Journal of Applied Earth Observation and Geoinformation 103, p. 102532. doi: 10.1016/j.jag.2021.102532.