Search

The nonlinear remote sensing ecological index (nRSEI) is a recently proposed eco⁃ logical index, which used the kernel principal component analysis (kPCA) algorithm to integrate four indica⁃ tors of the existing remote sensing ecological index (RSEI) rather than using the traditional principal compo⁃ nent analysis (PCA) technique. The main reason for using kPCA was that in the Beijing area the correla⁃ tions between wetness, greenness, dryness, and heat that are four indicators used in RSEI were generally weak, so the kPCA that is specially developed to deal with variables with nonlinear relationship was needed to handle these four non-linear indicators. This paper aims to examine the correlation strength of these four indicators in the Beijing area to see whether their relationship is strong or weak and analyzes the effective⁃ ness of the accuracy assessment method that was used for the validation of the new index. Methods: Through examining correlation coefficients and scatter diagrams, the correlation between the four indicators is investigated to find out whether the relationship between the indicators is linear or non-linear. Also, the effectiveness of the validation for the new index is analyzed. Results: The results show that the four indica⁃ tors are strongly linearly correlated with each other, therefore, the kPCA was not suitable for the intergra⁃ dation of the four indicators. The methods used for the accuracy assessment of the new index also have obvious defects and thus failed to effectively validate the accuracy of the new index. In addition, some important issues related to prepare remote sensing papers are also discussed. These include the feasibility of remote sensing modeling, the robustness of the model, the scale consistency of the sub-indicators in the model, the validation method for the model, the selection of reference images for the validation, and the required sample size for accuracy validation. Conclusions: The nRSEI is not a suitable index for the assessment of regional ecological status as it mistakenly employed kPCA to intergrade the four indicators that are linearly related. [ABSTRACT FROM AUTHOR]