|WANG, G - University Of Arizona|
|WEST, A - University Of Arizona|
Submitted to: Remote Sensing of Environment
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 5/19/2012
Publication Date: 8/1/2012
Citation: Thorp, K.R., Wang, G., West, A.L., Moran, M.S., Bronson, K.F., White, J.W., Mon, J. 2012. Estimating crop biophysical properties from remote sensing data by inverting linked radiative transfer and ecophysiological models. Remote Sensing of Environment. 124:224-233.
Interpretive Summary: Remote sensing can rapidly map crop growth status over large areas based on reflectance of radiation from the crop canopy. However, this information must be linked with simulation models in order to quantify crop properties, forecast crop yield, understand the impacts of drought or global climate change, and guide agricultural resource management for water and nitrogen fertilizer. In this study, we developed techniques for using simulation models to increase the amount of useful information gleaned from remote sensing data. By implementing inverse modeling techniques (using remote sensing data to automatically parameterize the simulation model), we were able to improve estimates of several crop properties as compared to field-measured values. We also showed that data from newer 'hyperspectral' instruments, which collect reflectance information in many narrow wavebands, offered several advantages as compared to older instruments that collect information in only a few broad wavebands. These advantages included better estimation of crop traits and fewer required remote sensing observations. The results of the study will be useful to agricultural and remote sensing scientists who are using remote sensing solve agricultural problems related to crop yield forecasting and cycling of water and nutrients.
Technical Abstract: Remote sensing technology can rapidly provide spatial information on crop growth status, which ideally could be used to invert radiative transfer models or ecophysiological models for estimating a variety of crop biophysical properties. However, the outcome of the model inversion procedure will be influenced by the timing and availability of remote sensing data, the spectral resolution of the data, the types of models implemented, and the choice of parameters to adjust. Our objective was to investigate these issues using inverse modeling with linked radiative transfer and ecophysiological models to estimate leaf area index (LAI), canopy weight, plant nitrogen content, and yield for a durum wheat (Triticum durum) study conducted in central Arizona over the winter of 2010-2011. Observations of crop canopy spectral reflectance were obtained weekly using a GER 1500 spectroradiometer. Other field measurements were regularly collected to describe plant growth characteristics and plant nitrogen content. Linkages were developed between the DSSAT Cropping System Model (CSM) and the PROSAIL radiative transfer model (CSM-PROSAIL) and between the DSSAT-CSM and an empirical model relating NDVI to LAI (CSM-NDVI). The PEST parameter estimation algorithm was implemented to adjust the leaf area growth parameters of the CSM by minimizing error between measured and simulated NDVI or canopy spectral reflectance. A genetic algorithm was implemented to identify the optimum combination of remote sensing observations required to optimize simulations of LAI through inverse modeling. The relative root mean squared error (RRMSE) between measured and simulated LAI was 24.1% for the CSM-PROSAIL model, whereas the stand-alone PROSAIL and CSM models simulated LAI with RRMSEs of 40.7% and 27.8%, respectively. Wheat yield was simulated with RRMSEs of 12.8% and 10.0% for the lone CSM model and the CSM-PROSAIL model, respectively. Inverting CSM-PROSAIL using hyperspectral data offered several advantages as compared to the CSM-NDVI inversion using a simple vegetation index, including better estimates of crop biophysical properties and fewer required remote sensing observations for optimum LAI simulation. Only two observations, one at mid-vegetative growth and one at maximum vegetative growth, were required to optimize LAI simulations for CSM-PROSAIL. The inversion of linked DSSAT-CSM and PROSAIL models based on crop canopy spectral reflectance observations offered several advantages, including improved estimates of several crop biophysical properties and fewer required remote sensing observations, as compared to inversion of a linked DSSAT-CSM and NDVI model, inversion of the lone PROSAIL model, and forward simulations with the lone DSSAT-CSM model.