Evaluation of Multiorbital SAR and Multisensor Optical Data for Empirical Estimation of Rapeseed Biophysical Parameters

Abstract

This paper aims to evaluate the potential of multitemporal and multi-orbital remote sensing data acquired both in the microwave and optical domain to derive rapeseed biophysical parameters (crop height, dry mass, fresh mass and plant water content). Dense temporal series of 98 Landsat-8 and Sentinel-2 images were used to derive Normalized Difference Vegetation Index (NDVI), green fraction cover (fCover) and Green Area Index (GAI), while backscattering coefficients and radar vegetation index (RVI) were obtained from 231 mages acquired by Synthetic Aperture Radar (SAR) onboard Sentinel-1 platform. Temporal signatures of these Remote Sensing Indicators (RSI) were physically interpreted, compared each other and to ground measurements of biophysical parameters acquired over 14 winter rapeseed fields throughout the 2017-2018 crop season. We introduced new indicators based on the cumulative sum of each RSI that showed a significant improvement of their predictive power. Results particularly reveal the complementarity of SAR and optical data for rapeseed crop monitoring throughout its phenological cycle. They highlight the potential of the newly introduced indicator based on: the VH polarized backscatter coefficient to estimate height (R 2 = 0.87), plant water content (R 2 = 0.77, from flowering to harvest) and fresh mass (R 2 = 0.73) and RVI to estimate dry mass (R 2 = 0.82). Results also demonstrate that multi-orbital SAR data can be merged without significantly degrading the performance of SAR-based relationships, while strongly increasing the temporal sampling of the monitoring. These results are

Assimilation of Multisensor Optical and Multiorbital SAR Satellite Data in a Simplified Agrometeorological Model for Rapeseed Crops Monitoring

“…A. From SAR data to dry mass estimates DM has been derived from Sentinel-1 SAR-derived RVI using polynomial regression as proposed by [21]:…”

“…For each field GAI estimates were derived as the median of pixels included in the field for images with more than 80 % of cloud free pixels. Overland's GAI estimates from both Sentinel-2 and Landsat-8 have already been validated against ground measurements acquired over 18 winter rapeseed fields all along the rapeseed growth cycle (R 2 = 0.78 and RMSE ≤ 0.41 m 2 .m -2 ; see [21] for details). VH were extracted for each field from Sentinel-1 A and B data using Google Earth Engine website [44].…”

“…Finally, potential modifications of the model to take into account temporal variation in both SLA and ELUE necessary require adding parameters raising possible new equifinality question. Moreover the uncertainty in remote sensing derived GAI and DM is non-negligible (see [21]) and a better physiological representativeness of the model does not guarantee more accurate simulations resulting of the assimilation of these remote sensing data.…”

“…Rapeseed growth (biomass, yield, duration of phenological stages…) is affected by climatic conditions (soil moisture, temperature warming…) [27] and crop management (date of sowing, fertilization…) [28]. Only few studies addressed the sensitivity of backscatter coefficients to biophysical parameters [23], [29]- [31], but [21] have recently showed that the accuracy of rapeseed DM retrievals from Radar Vegetation Index (RVI) derived from Sentinel-1 backscatter coefficients was sufficiently satisfying (R 2 > 0.75 and RMSEr < 50%) to consider their assimilation into an agrometeorological model. They have also demonstrated that the angular impact of multiorbital acquisitions of Sentinel-1 images was low for the polarization ratio and RVI.…”

Assessing the Potential of Fully Polarimetric Mono- and Bistatic SAR Acquisitions in L-Band for Crop and Soil Monitoring

“…This makes the retrieval of soil moisture with such systems challenging. Recent studies investigated the potential of combining radar and optical images, from Sentinel-1 and Sentinel-2 respectively, for predicting soil moisture and crop biophysical parameters [9]- [11]. The results of these studies show that combining the microwave and optical domains is a promising technique for finer crop and soil monitoring thanks to its increased temporal sampling and retrieval accuracy.…”