Abstract: Iron oxide is the main form of iron element existing in the soil. In subtropical areas, the high-content iron oxide constitutes the soil’s important coloring components, or its mineral substances, such as goethite and hematite, making the soil color apparently different from that in other climatic zones. The present paper, with the Pearl River Delta, a typical subtropical area, as illustration, and through analysis of the correlation between different spectral forms and the content of soil iron oxide, created inversion models of soil iron oxide by extracting characteristic spectral bands. The findings showed that there was a negative correlation between the content of soil iron oxide and the reflection spectrum, and the sensitive bands were mainly found in such visible near-infrared regions such as 404, 574, 784, 854 and 1204 nm. The correlation between the spectrum through differential processing and the soil iron oxide was significantly improved. On the basis of the correlation-prominent bands, the methods of both multiple linear regression and principal component analysis were adopted so as to remove collinear bands, and finally, characteristic bands were selec-ted to serve as the input parameters of inversion models. A comparison of the results revealed that the best inversion model of soil iron oxide content in the Pearl River Delta was BP artificial neural network (i.e., RMSEC=0.22, RMSEP=0.81, R²=0.93, RPD=12.20). It was applicable with excellent stability to the fast estimation of the iron oxide content in the soil and could hopefully serve as the research basis for the measure of the spatial distribution of the soil iron oxide.