| |
| In remote sensing community, a lot of researches have been carried out to reduce noise interferences from soil and thus more accurately predict plant biophysical parameters by developing numerous vegetation indices (VIs). However, few studies were focused on the effect of crop residue cover on Leaf Area Index (LAI) estimation by remote sensing. Based on WinSail model and ground measurements, the objectives of this study were to investigate the variation of residue cover induced by tilling practices, evaluate the effect of this variation on spectra and VIs of wheat canopy, and identify the optimum VIs in decreasing the effect of residue variation on LAI estimation. Results suggested that intensive-tilling and no-tilling practices induced large variation in crop residue covers, with average Relative Residue Cover (RRC, residue cover to whole background) 8.32% and 61.6%, respectively. Because of the difference in brightness between residue and soil, the variation of RRC from 8.32% to 61.6% caused 43.3-to-23.5% and 50.5-to-15.6% Variation Rates in spectral reflectance (VRs) when LAI was lower than 1.0 at red and NIR wavelengths, respectively. Compared with spectral reflectance, some VIs based on Red-NIR bands such as the Normalized Difference Vegetation Index (NDVI) and Ratio Vegetation Index (RVI) had significantly better abilities of resisting to residue cover variations. With proper adjustment coefficients, soil-adjusted VIs could further decrease the effect of residue. But the performance of soil-adjusted VIs was greatly influenced by the selection of background lines. Moreover, residue could influence LAI estimation by obscuring effect which can be measured by the percentage of Green Covered by Residue (GCR). At early growth stage of wheat, our measured average GCR between the plots with RRC being below 5% and these with RRC being over 65% was 10.9% which would directly cause the underestimation of LAI by remote sensing. By all the VIs used in this study, including ratio-based, soil-adjusted and hyperspectral VIs, LAI of no-tilling fields with large residue cover would be underestimated, with underestimation extent (Predicted Deviation, PD) ranging from 4.2% to 23.1%. By the selection of VIs and background lines, PD can be greatly decreased. Generally, soil-adjusted VIs with proper background line performed better than ratio-based VIs. Instead of soil line, residue line was more suitable for soil-adjusted VIs in minimizing the effect of residue. Although hyperspectral VIs of Red Edge Inflection Point (REIP) and the first Derivative at the Red Edge (dRE) were not necessarily better in correlating LAI for intensive-tilling fields with low residue cover, both dRE and REIP could decrease the effect of residue when compared with NDVI and RVI. In conclusion, it is important to carry out the research on the effect of residue cover on the estimation of crop parameters at the areas with different tilling practices. The Transformed Soil-Adjusted Vegetation Index (TSAVI) with proper background line and dRE can greatly minimize the effect of residue on LAI estimation. |
|
| Keywords:Remote sensing; Wheat; Residue; LAI; Vegetation index |
|
