The Response Of Remotely Sensed Retrieval Of Wheat Leaf Area Index To Spectral And Spatial Scale In Different Growth Stages

Leaf area index (LAI) is a significant structural parameter for quantifying the energy and mass exchange characteristics of terrestrial ecosystems such as photosynthesis, respiration, transpiration, the carbon and nutrient cycle, and rainfall interception. For the crop research, LAI is an important indicator for crop growth dynamic and production predicting. Remote-sensing retrieving is an important method of producting regional and global LAI products. However, spectral and spatial scale effects result in inconsistency among remotely sensed LAI products from different sources. This limits their use in unified applications and affects the precision of ecological models that utilize them as input parameters. However, the systematic study on the spectral scale effect is relatively absent, and the mechanism of spatial scale effect is still weak at present. As a typical continuous distributed vegetation, wheat has a different canopy structure and physiological characteristics in different growth stages, resulting in different spectral characteristics and spatial heterogeneity of remote sensing images. Therefore, this thesis aims to take wheat canopy LAI as the the research object, comprehensively using the radiative transfer models, various information processing and mathematical statistics method, to study the response characteristics between remotely sensed LAI and the spectral and spatial scale. This research could provide theoretical reference for the producing and validation of LAI remote sensing products and the design of the related sensors. It could also provide support for the application of wheat growth monitoring and productivity forecast. In order to achieve the objectives mentioned above, based on the forward simulation of the canopy radiative transfer model PROSAIL and continuous wavelet transform method, the response characteristics of canopy LAI in different growth stages to spectral scale was analyzed. Based on the numerical simulation method and PROSAIL model inversion, the response characteristics of the LAI at pixel scale in different growth stages to spatial scale was analyzed. The main research contents and conclusions are as follows:1. The simulated and measured canopy LAI and reflectance in different growth stages were analyzed, as well as the correlations between canopy LAI and reflectance, the first derivative of reflectance, and the common multispectral and hyperspectral vegetation indices. The results showed that the change trend in different growth stages of wheat canopy LAI was consistent with the change trend of the nutrition growth, reproductive growth and senescence law of wheat. The change of the wheat canopy spectrum in the visible region was consistent with the changes of pigments content of wheat leaves, and the change in the near infrared region was related to the structure of the canopy and leaves. LAI and the reflectance in the visible region were significantly negatively correlated, and LAI had a significant positive correlation with the near infrared reflectance. LAI was positively correlated with the first derivative of the spectral reflectance in the band range of 400-450nm,500-540nm, and 690-780nm. EVI and MCARI/OSAVI were the multi-/hyper-spectral vegetation indices having the highest correlations with LAI, respectively.2. The method to analyze the response patterns of LAI to different spectral scale (spectral band position and band width) was established based on the continuous wavelet transform method. The canopy reflctance and its first derivative were decomposed by the continuous wavelet transform method based on the Mexican Hat and Haar wavelet basises. The wavelet coefficients were related to the different spectral band positions and band width. The correlation and sensitivity between the wavelet coefficients and LAI were analyzed by Pearson’s correlation analysis and the sensitivity analysis method based on generalized regression neural network (GRNN), respectively. The results showed that the correlation of wavelet coefficients and LAI increased with the increase of wavelength, and the high value of R2 mainly appears in the red and near infrared spectral regions. The sensitivity of wavelet coefficients based on reflectance and LAI was highest in the red region. The sensitivity of wavelet coefficients based on the first derivative of reflectance was highest in the red edge position and increasing with the increase of wavelet scales.3. The optimal spectral band positions and band width were extracted for the LAI estimation in different growth period. The spectral response function based on the Gaussian function was proved to be the optimal spectral response function for fitting the wide band reflctance. A variety of spectral indices are constructed based on the wide band reflectance to obtain the optimal spectral indices. The corresponding optimal LAI estimation models were established. The results showed that the optimal band position based on the reflectance in different growth periods basically located in the red and near infrared regions. The optimal band widths of the bands in the red region were in 10 nm, and the optimal band widths of the bands in the near infrared region were about 100 nm. The optimal band positions related to LAI based on the first order derivative of reflectance were extracted in the red and near infrared region and the red edge position. The optimal band positions in the blue and orange region were also extracted in the late period of wheat growth. The difference of the optimal band widths in different growth stages is relatively small. The ability of the Haar wavelet to extract the canopy spectral features related to LAI was proved to be weaker than that of Hat Mexican. The optimal spectral indices in this thesis had a higher correlation with LAI, compared with the multispectral indices based on the existing satellite sensors. The correlation between the new optimal spectral indices and LAI was close to but not higher than that between the existing hyperspectral indices and LAI, but the former had the advantage of improving the signal to noise ratio.4. Based on numerical simulation method, the mixed pixels of different endmember (wheat canopy and soil background) proportion were simulated for different growth stages. The semivariogram and two-dimensional wavelet analysis method were used to quantitatively describe the spatial heterogeneity of the mixed pixels. The LAI of the mixed pixels was inversed by the inversion method and the simulated pixel reflectance based on linear model. The correlation and sensitivity between the heterogeneity indices and LAI were analyzed by Pearson’s correlation analysis and the sensitivity analysis method based on GRNN, respectively. The results show that the spatial scale of wavelet variance was consistent with the spatial resolution of the pixels, so it is more suitable for the interpretation of the spatial scale of the pixel. The spatial heterogeneity of pixels described by the wavelet variance increased first and then decreased with the increase of the ratio of the canopy endmembers in the mixed pixel and the spatial scale. For the different growth periods, the LAI values at pixel scale in the booting stage had the highest correlation with the spatial heterogeneity, while the milking stage had the lowest correlation. With the increase of the spatial scale, the correlation between LAI at pixel scale and the spatial heterogeneity first increased and then decreased. The differece of the sensitivity of LAI and spatial heterogeneity in different growth stages was small, which first decreased and then increased with the increase of spatial scale.