Contribution of Sentinel-1 data for the monitoring of seasonal variations of the vegetation P.-L. Frison, S. Kmiha, B. Fruneau, K. Soudani, E. Dufrêne, T. Koleck, L. Villard, M. Lepage, J.-F. Dejoux, J.-P. Rudant, Th. Le Toan Bruges, May 27 2017
Spaceborne Radar Sensors ERS1 SAR Diffusiomètre Bande C JERS SAR Bande L RADARSAT SAR Bande C ERS2 SAR Scatterometer Bande C Quikscat Scatterometer Bande Ku RADARSAT2 SAR Bande C TerraSAR X SAR Bande X ENVISAT SAR Bande C METOP Scatterometer Bande C ALOS AR Bande L COSMO-SkyMed SAR Bande X Tandem X SAR Bande X Sentinel-1 SAR Bande C ALOS 2 SAR Bande L 1991 1992 1995 1999 2002 2006 2007 2010 2014 2016
Vegetation monitoring with ERS SAR Crops (Spat. Res. 20 m - Temp. Res. 35 days, VV pol.) Temperate Forest o: Winter wheat : rape Wooding et al., 1993 -.: oaks : beeches pine grassland 1994 1995 1996 Day of Year Proisy et al., 1999
ERS Scatterometer (50 km / 4 days) Global image of Radar backscattering coefficient s 0
ERS scatterometer (Spat. res. 50 km - Temp. res. 5 days) Temporal monitoring over large homogeneous areas (Sahelian steps) Dry year Wet year ESCAT scatterometer - incidence: 45 - l = 6 cm High variability of the inter annual yearly amplitude ==> linked to annual biomass (annual grasses)
SAR SENTINEL-1 Scatterometers Acquisitions period: 12 days (S1-A) 6 days( S1- A+B) Planned mode over land surfaces: Interferometric Wide (IW) 2 Polarisations: VV VH Swath: 250 km (3 sub-swaths) GRD Products : Spatial resolution: Pixel: 10 m 20 m SLC Products Source: ESA Spatial resolution: 3 x 20 m; Pixel: 2 x 14 m (rge x az.) Temporal monitoring of seasonal variations of land surfaces Radar Backscattering Coefficient s 0 Interferometric Coherence r
SAR SENTINEL-1 Scatterometers Acquisitions period: 12 days (S1-A) 6 days( S1- A+B) Planned mode over land surfaces: Interferometric Wide (IW) 2 Polarisations: VV VH Swath: 250 km (3 sub-swaths) GRD Products : Spatial resolution: Pixel: 10 m 20 m SLC Products Source: ESA Spatial resolution: 3 x 20 m; Pixel: 2 x 14 m (rge x az.) Temporal monitoring of seasonal variations of land surfaces Radar Backscattering Coefficient s 0 Interferometric Coherence r
Cohérence ρ = z 1 z 2 z 1 2 z 2 2 Complex radar data: ==> Geometrical stability ( l) of el. scat. within res. cell between 2 acquisitions (6 or 12 days) Im Im Im z 1 z 1 Re Re Re z 2 z 2 z 1 z 2 1 1 1-1 1-1 0.9 1-1 0.1 1-1 -1-1 r = 1 r = 0.9 r = 0.1 Vegetation phenological stages?
IW mode Acquisition: VV + VH Radar backscattering coeffcient s 0 cohérence March 18 2015 18-30 March 2015 s 0 VV - s 0 VH s 0 VH / s 0 VV r VV - r VH r VV / r VH
s 0 multitemporal color composite image May 5 - Sept. 2 Dec. 19 2015 VV Polarisation VH Polarisation
r multitemporal Color composite image May 5 / 17 - Sept. 2 / 14 Dec. 19 / 31 2015 VVPolarisation VHPolarisation
Orbite 59 Acquisitions over Parisian Region Fontainebleau Forest Orbite 110 ONF management, oaks, beeches, pines plots Coll. ESE / Paris Sud Remote Sensing + forest models: ==> better understanting functioning + interaction with climate / hazard
Precipitations (mm) Temperature ( C) Oaks Radar Backscat. Coeff. s 0 Coherence r Asc. orbit Desc. orbit VV/VH Day of Year Day of Year Seasonal cycles 0 VH ==> s 0 VV / s 0 VH (yearly amplitude 3 db) signal low and constant (Marchs-Nov.) r VV et r VH Identical higher values for low temperatures
Precipitations (mm) Oaks Radar Backscat. Coeff. s 0 Asc. orbit Desc. orbit ERS (VV) temporal signature -.: Oaks : Beeches Pine VV/VH Grassland 1994 1995 1996 no seasonal cycle in VV pol. Seasonal cycle s 0 VH ==> s 0 VV / s 0 VH Proisy et al., 1999
Oaks Radar Backscat. Coeff. s 0 Asc. orbit Desc. orbit NDVI VV/VH s 0 VV / s 0 VH Day of Year Day of Year s 0 VV / s 0 VH and NDVI in phase C band sensitive to foliar activity
Precipitations (mm) Oaks Radar Backscattering Coefficient s 0 Asc. orbit Desc. orbit Day of Year Low correlation with precipitations C band: low soil contribution over forest
Precipitations (mm) Grassland Radar Backscattering Coefficient s 0 Asc. orbit Desc. orbit Day of Year highercorrelation with precipitations
Precipitations (mm) Temperature ( C) Pine trees Radar Backscat. Coeff. s 0 Coherence r Asc. orbit Desc. orbit VV/VH Day of Year Day of Year
Pine trees Radar Backscat. Coeff. s 0 Coherence r NDVI VV/VH Day of Year Day of Year
High temporal frequency... to detect erroneous data!
Temperate Forests Radar Backscattering Coeff. s 0 Dt = 12-6 days VV + VH Well suited for seasonal variation monitoring s 0 VV / s 0 VH in phase with foliar activity No seasonal cycle s 0 VV Low correlation with precipitations Coherence: r VV and r VH identical low ( r = 0.2 noise) and constant from March to Nov. higher values during winter with low temperatures
Crops monitoring Lamasquère region in situ survey (CESBIO) Winter crops: wheat, barley, rapeseed Summer crops: soybean, sorghum, maïze, sunflower
Lamasquère Region multi-temporal Color Composite images Radar Backscat. Coeff. s 0 Coherence r 10 Jun- 14 Sept. 7 Dec. 4-16 Jul.- 9-16 Aug. 7-19 Dec.
Harvest Harvest Sowing CROPS: TEMPORAL PROFILES s 0 Sowing Sowing Sowing Harvest Harvest WHEAT - BARLEY MAIZE SOYBEAN - SORGHUM RAPESEED SUNFLOWERS
Harvest Harvest Sowing CROPS: COHERENCE TEMPORAL PROFILES r Sowing Sowing Sowing Harvest Harvest WHEAT - BARLEY MAIZE SOYBEAN - SORGHUM RAPESEED SUNFLOWERS
Crops areas Strong seasonal variations due to contrasted landscapes (tillages, vegetation dev ) Radar Backscattering Coefficient s 0 : Strong seasonal cycle s 0 VV / s 0 VH Not visible in VV polarisation Coherence: low ( r = 0.2) when vegetation fully developped high contrasts for other stages s 0 VV / s 0 VH and r complementary onformations Accurate in situ surveys (tillage) for deeper interpretation Dt = 12 days (1 satellite) limitations for reliable information ==> Sentinel-1A+B
CONCLUSION Monitoring of seasonal variations of vegetation: Strong contribtion of Sentinel-1: DT = 6-12 days s 0 VH / s 0 VV forests deciduous species (foliar activity) low correlation with precipitations r no meaningful information Agricultural areas s 0 VH / s 0 VV r DT = 12 days nice, DT = 6 days better Strong seasonal variations Complemetary informations Results to be extended with additional observations + Modelisation!!! (sites + longer period)
Sentinel-1 RADAR BACKSCATTERING IMAGE : Acquisition 2015/03/02 Parisian region VV VH VH/VV
Sentinel-1 RADAR BACKSCATTERING IMAGE : Temporal average 2015/03/02-2017/01/26 Parisian region VV VH VH/VV
GoogleEarth Image Parisian region
Sentinel-1 RADAR BACKSCATTERING IMAGE : Acquisition 2015/03/02 Fontainebleau Forest VV VV VH VH VH/VV VH/VV
Sentinel-1 RADAR BACKSCATTERING IMAGE : Temporal average 2015/03/02-2017/01/26 Fontainebleau Forest VV VH VH/VV
GoogleEarth Image Fontainebleau Forest
Sentinel-1 RADAR BACKSCATTERING IMAGE : Acquisition 2015/03/02 Parisian region VV VV VH VH VH/VV VH/VV
Sentinel-1 RADAR BACKSCATTERING IMAGE : Temporal average 2015/03/02-2017/01/26 Parisian region VV VV VH VH VH/VV VH/VV
GoogleEarth Image
Sentinel-1 RADAR BACKSCATTERING IMAGE : Acquisition 2015/03/02 VV VH VH/VV
Sentinel-1 RADAR BACKSCATTERING IMAGE : Temporal average 2015/03/02-2017/01/26 VV VH VH/VV
GoogleEarth Image
Sentinel-1 RADAR BACKSCATTERING IMAGE : Acquisition 2015/03/02 VV VH VH/VV
Sentinel-1 RADAR BACKSCATTERING IMAGE : Temporal average 2015/03/02-2017/01/26 VV VH VH/VV
GoogleEarth Image
Sentinel-1 RADAR BACKSCATTERING IMAGE : Acquisition 2015/03/02 VV VH VH/VV
Sentinel-1 RADAR BACKSCATTERING IMAGE : Temporal average 2015/03/02-2017/01/26 VV VH VH/VV
GoogleEarth Image