The main biochemical components of vegetations, such as chlorophyll and water, are directly involved in the major ecological processes and function of terrestrial ecosystem．Remote sensing of vegetation biochemical components may have important applications in the fields of agriculture and forestry. Many nonlinear or linear models have been developed for quantitatively retrieving vegetation biochemical components from satellite remotely sensed data. But it is often difficult to retrieve the biochemical components with satisfied accuracy, especially in sparsely vegetated areas because of the influence of the soil background. To select the better models that can resistance soil influence, in this paper, LOPEX’93(Leaf Optical Properties Experiment) dataset was used to analyze the anti-soil capacity of the spectral models in retrieving the vegetation chlorophyll and water content. The parameters used in the model include reflectance and its variants, spectral position variables and vegetation indices. The correlation coefficient between the vegetation biochemical component and the mixed spectra, which were produced by weighting vegetation spectra and the soil spectra with area ratio, has been analyzed. The results show that the models composed of the spectral parameters of the reflectance and its variants to inverse vegetation chlorophyll, the reflectance and the logarithm of reciprocal reflectance of 730 nm and 400 nm combination can keep a high correlation coefficient while the area ratio of soil component changes from 10 percent to 90 percent, the correlation coefficient between the reflectance and chlorophyll was around 0.645, and the correlation coefficient between the logarithm of reciprocal reflectance and chlorophyll was 0.650. To inverse water content, the combination of 1100 nm, 1170 nm, 1000 nm, 1040 nm, 1080 nm reflectance, and the combination of 1170 nm, 960 nm, 1210 nm,1090 nm, 1080 nm, 950 nm, 1220 nm, 1210 nm logarithm of reciprocal reflectance show a strong anti-soil capacity, the correlation coefficients between the two models and water content were all larger than 0.99. In the models composed of the spectral parameters of the spectral position variables, the parameter of red edge-green peak-red valley was selected as the strongest Anti-soil capacity parameter, the correlation coefficients were distributed around 0.530. In the models composed of the vegetation index to retrieve vegetation chlorophyll,, the anti-soil capacity was poor but when the model is used to retrieve vegetation water content the correlation is stable though the soil area ratio is changable and the correlation coefficients are as high as 0.980 and 0.960 at the two typical water indices of Ratio975 and Ratio1200, respectively. These conclusions can be used to guide the vegetation biochemical component inversion for sparsely vegetated regions.