| The large-area monitoring of crop leaf area index generally need the support ofthe remote sensing technology. With the satellite remote sensing data of differentspatial resolution, the LAI of different spatial scales can be accessed. However,because of the heterogeneity of the remote sensing data and the non-linearity of theLAI inversion models, the transformation of LAI from one to another scale induces ascaling bias. Researchers usually call it scale bias for short.Taking scale bias as the study subject, this paper completes the LAI spatial scaletransformation modeling work. Following is the main conclusions which are drownduring the analysis and modeling process:(1) The fractal relation between the Haar wavelet decomposing high frequencycoefficient and the scale bias is discovered by the wavelet-fractal analysis.Their double logarithmic linear fitting coefficients, fractal dimension (FD)and intercept, are the quantitative representation of the scale bias;(2) With the support of multiple datasets, it is obvious that none of the varioustypes of the inversion models shares the same features of the FD distribution.The statistical ellipses, which are constructed by the second ordercombinations of the mean and std value, cover the most parts of the FD data.The model characteristics can be acquired to some extent;(3) This paper completes the extension from1D to2D on the scale bias analysis.A new analysis plan is proposed which takes the LAI as the study object.And the method of the2D analysis is determined, receiving the fractal fittingcoefficients as the result;(4) Based on the exponential and logarithmic inversion models, the feasibilityanalysis towards the construction of the spatial scale transformation model isundergone. With the usage of the exponential inversion model the scaletransformation model is constructed. The result of the validation whichinvolves other crop remote sensing products indicates the great generality ofthis model. |
PDF Full Text Request