Estimation of leaf area index and crop height of sunflowers using multi-temporal optical and SAR satellite data

摘要

The objective of this study was to improve the understanding of radar signals for sunflower and to assess the potential of microwave and optical satellite data to monitor crop parameters (leaf area index (LAI) or crop height (CH)) from sowing to harvest by determining the best suitable antenna configurations (i.e. frequency, polarization, and incidence angle). These parameters have been targeted since they are considered key parameters to derive some important agronomical or physical indicators of the crop (i.e. grain yield, leaf nitrogen concentration, radiation interception). This study is based on the Multispectral Crop Monitoring experimental campaign conducted by the CESBIO laboratory in 2010 (MCM'10) for an agricultural region located in southwestern France. From sunflower emergence to harvest, satellite images were regularly acquired by TerraSAR-X, Radarsat-2, Alos, Formosat-2, and Spot-4/5, quasi-synchronously with in situ measurements (combining soil and vegetation observations). The time series of synthetic aperture radar (SAR) images demonstrate a wide range of configurations, with different frequencies, polarization states (X-HH, C-HH/VV/VH/HV, and L-HH), and incidence angles (from 24 to 53 degrees). The first results have shown that the angular radar sensitivity is greater in the C-band than in the X-band for the co-polarized signals (HH or VV) and for low normalized difference vegetation index (NDVI<0.4), and decreases with the increase of vegetation (from 0.4dB/degree for bare soils to 0.05dB/degree for fully vegetated fields). Among the wide ranges of compared frequency and polarization states, the best results were obtained with the C-HH and L-HH satellite signals for LAI estimates (coefficient of determination or R-2>0.82, relative root mean square error or rRMSE<13%), and with C-HH, L-HH, and NDVI for CH retrieval (R-2>0.75, rRMSE<15%). Further analyses are needed to confirm the promising results observed in the L-band. Finally, a modified version of the water cloud model (WCM) allows analysis of the signal components from the soil and/or the different vegetation layers.