Journal of Remote Sensing & GIS

Exploring radiative transfer model inversion of canopy properties of oilseed rape by numerical optimization based on simulation and field experiments

Joint Event on 3^rd International Convention on Geosciences and Remote Sensing & Geochemistry, Environmental Chemistry and Atmospheric Chemistry

October 19-20, 2018 | Ottawa, Canada

Shanqin Wang

Huazhong Agricutural University, China

Posters & Accepted Abstracts: J Remote Sensing & GIS

Abstract:

Inversion of radiative transfer model may provide a generally applicable tool for assessment of crop status in precision agriculture. The potential of the PROSAIL inverted by means of hyper-parameter searching algorithms to find optimal values as an estimation of target crop variables was investigated based on a simulation and a field experiments of oilseed rape (Brassica npus L.). First, one thousand of canopy reflectance spectra were simulated with PROSAIL using crop biophysical variables generated in the reasonable range. By the inversion technique based on the Tree-structure Parzen Estimators (TPE), the simulated leaf area index (LAI), leaf dry mass (LMA) and leaf chlorophyll a+b (Cab) were used as measured variables to predict carotenoid content (Car), brown material (Cb) and equivalent water thickness (Cw) with high accuracy. Even using LAI as an only measured variable with optimization of acceptable iterations, other properties could be estimated with reasonable accuracy. Then for practical application, the field experiment was conducted with one oilseed rape cultivar under eight nitrogen (N) rates (45- 300 kg N ha-1) using a randomized complete block design with three replications for each treatment. Canopy reflectance, LAI, Cab, and LMA were measured on four critical days after transplanting (DAT). The Cab was predicated with high accuracy(R2=0.72, nRMSE=0.12) for each sample by using LAI and LMA as known crop variables to invert PROSAIL based on TPE of maximum 3000 iterations. The higher estimation of Cab (R2=0.82, nRMSE=0.10) can be obtained with maximum 10000 iterations of optimization. Unfortunately, neither LMA nor Cab was successfully estimated by only using LAI as known variable, indicating the limits of the PROSAIL to deal with an ill-posed problem if without enough parameter. Our results confirm the potential of RTM inversion by means of TPE for estimating vegetation variables at the canopy scale using hyperspectral measurements for crop precision management.

Biography :

Shanqin WANG has completed his PhD at the age of 33 years from Wuhan University. He is one lecturer of Huazhong Agricutural University now. He has published more than 15 papers in reputed journals. His research focuses on the quantitative remote sensing of environment and agriculture, the geostatistics and the digital soil mapping.

E-mail: 253896575@qq.com