Apport des données Sentinel-1 pour la cartographie des milieux humides

Cécile Cazals 1
1 COGIT - Cartographie et Géomatique
LaSTIG - Laboratoire des Sciences et Technologies de l'Information Géographique
Abstract : Wetlands are threatened by climate change and the anthropization of natural environments. Satellite remote sensing is useful for environmental monitoring at large areas. However, when it comes to the study of hydrological dynamics, a significant temporal resolution is essential. The latter is difficult to reach with optical satellite imagery because of the cloud cover that masks the ground. Radar sensors are well suited to the characterization of hydrological dynamics thanks to the sensitivity of their measurements in the presence of water, whatever the vegetation in place. As a result, all Synthetic Aperture Radar (SAR) acquisitions are available, both day and night, regardless of cloud cover.Satellite radar remote sensing has gone through a revolution with the launch of the Sentinel-1A satellite, followed by its twins Sentinel-1B by the European Space Agency as part of the Copernicus program in 2014. These sensors acquire C-band data (λ = 5.6 cm) on a regular basis on Europe and their distribution is free for all users. Their temporal frequency initially of 12 days has decreased to 6 days from the end of 2016. This work aims at evaluating the potential of these data with high temporal resolution for the monitoring of water bodies and wetlands.The first part of this thesis focuses on water bodies mapping. We found confusion in the C-band radar response between water surfaces and that of some bare soils. We showed that the winter period is the least ambiguous and that the VH polarization is the most suitable for the mapping of water surfaces. Four methods of water detection have been compared. It appears that the use of unsupervised methods without a priori data is not conceivable and that the methods taking into account the spatial neighborhood give better results. Temporal filtering has been developed and has improved detection and avoided confusion between bare soil and permanent water surfaces. Water surfaces of more than 0.5 ha are more than 80% likely to be detected.A second part of this thesis is devoted to the monitoring of wet grasslands by radar remote sensing. The use of fully polarimetric data has shown that the VV/VH partial polarimetry configuration available on the Sentinel-1 sensor is able to characterize the prairial floods with or without vegetation. A method taking into account the temporal neighborhood allowed to process a series of 14 Sentinel-1 images to obtain 14 flood maps. The accuracy of floods maps at the intra-parcel scale has been estimated, it appears that if the precision is relatively good (80%), but the recall is rather low (40%). This method allow to establish intra- and inter-annual monitoring.This thesis has shown the potential of high temporal resolution radar images for the mapping of the water surfaces and for the monitoring of a wetland meadow
Document type :
Theses
Complete list of metadatas

https://tel.archives-ouvertes.fr/tel-01757083
Contributor : Abes Star <>
Submitted on : Tuesday, April 3, 2018 - 1:26:14 PM
Last modification on : Thursday, February 27, 2020 - 4:28:03 PM

File

TH2017PESC1226.pdf
Version validated by the jury (STAR)

Identifiers

  • HAL Id : tel-01757083, version 1

Collections

Citation

Cécile Cazals. Apport des données Sentinel-1 pour la cartographie des milieux humides. Géographie. Université Paris-Est, 2017. Français. ⟨NNT : 2017PESC1226⟩. ⟨tel-01757083⟩

Share

Metrics

Record views

665

Files downloads

1222