Complementarity of Linear Polarizations in C-Band SAR Imagery to estimate LAI for Maize and Winter Wheat Crop Monitoring
Leonard, Aline; Beriaux, Emilie; Defourny, Pierre
Universite catholique de Louvain, BELGIUM

In agricultural applications, monitoring crop growth and development in order to obtain early estimates of yearly yields is of common purpose. Remote sensing allows collecting frequent information about crop development, such as the Leaf Area Index (LAI). LAI is a biophysical variable of major importance for crop monitoring as well as for coupling Earth Observation with crop growth models in the perspective of yield forecasting. The RADARSAT-2 sensor presents a quad-polarization mode that multiplies information in one image by the simultaneous acquisition of the four linear polarizations (HH, VV, VH, and HV). Furthermore, SAR signal is very sensitive to plant water content, a variable highly correlated with the LAI during the growing phase.

Taking advantage of a large multi-year data set of SAR and ground observations collected in Belgium and in the Netherlands, this research aims at improving the understanding of the SAR signal sensitivity to crop growth. Although the interaction between linear polarized microwaves and agricultural targets has been studied widely, far less is undertaken in the definition of a method that takes benefits of all linear polarizations to optimize the LAI estimation.

From March 2009 to July 2011, four different SAR time series including 50 RADARSAT-2 images, quad-polarizations and at different incidence angles, were acquired under the SOAR-EU project in Belgium and the AgriSAR project in the Netherlands. Over the SAR data acquisition time series, we gathered ground data (LAI, volumetric soil moisture and phenological stages) for a large set of fields during the 2009, 2010, and 2011 crop growth seasons. The LAI was either measured with LAI-2000 sensor (LiCor) or by taking hemispherical photographs. Surface volumetric soil moisture was estimated thanks to the Soil, Water, Atmosphere and Plant model (SWAP) which simulates transport of water, solutes and heat in interaction with maize and winter wheat crop development by integrating a generic crop module WOFOST. The field measurements have been carried out according to the same protocol every 2 weeks during the crop growing seasons.

The semi-empirical Water Cloud Model (WCM) is implemented to derive maize LAI values from each linear polarization and surface soil moisture information. The cross-polarization was found the most relevant polarization to retrieve maize LAI through this model. A combination of the retrieved LAI and their associated errors for each polarization is then computed to improve the LAI estimation. Results indicate that this LAI estimation is often enhanced and its uncertainty always reduced by comparison with the ones retrieved from a single polarization. Moreover, the time series report reliable information about crop growth stage, such as booting and heading in winter wheat. The results show the influence of the acquisition time and of the local incidence angle in SAR time series.