Uncertainty analysis of spectra simulated using crop models driven by remote sensing data observed in the field

Multitemporal analysis has been expected an effective tool for crop identification and monitoring with the easy access of regularly recorded remote sensing data. To make this approach feasible, an important study is to build crop models based on the observed measurements and simulate the crop spectra in the key growing stages. While a few vegetation indices have been developed, such as Normalized Difference Vegetation Index (NDVI) and the slope of red edge, little attention has been given to the problem of uncertainty contained in the data due to wavelength excursion, sun angle variation, and the background spectral influence, etc. The propagation and accumulation of the uncertainty in the recorded data and developed crop models need to be taken into account for reliably crop spectra modeling. How data´s uncertainty affects spectra simulated in crop models driven by remote sensing data and decision-making is a critical issue. In this paper, several uncertainty sources are discussed. The contrast between the uncertainties of simulation spectral using Kimes model and Gap model is given at the same time. The result indicates that sun zenith, growth stage of winter wheat and observed angle have a closely relation with the spectral uncertainty. The experimental data sets were collected by the hyperspectral instrument SE590. The wavelength range is from 400 nm to 1100 nm and spectral resolution is about 3 nm. The two models were tested using the data recorded in ShunYi, Beijing, China (40°00N´-40°18´N, 116°28´E-116°58´E) in 2001.