Prediction of iron content in soil based on microspectrophotometry analysis

Soil is a very important type of trace evidence. The iron content of soil is of great significance in distinguishing soil types, discriminating among different soils, and tracing soils. However, conventional methods for analyzing the iron content of soil are expensive, laborious, and time-consuming. Previous studies have shown that the color of soil correlates well with its hematite content. This article thus deals with the indirect determination of iron content using soil color as a proxy. Soil color measurements were conducted using microspectrophotometry (MSP), and resulting data were transformed into chromaticity value (L*, a*, and b*). Predictions using the redness index in conjunction with a linear regression model were compared with those using the chromaticity value and a back propagation neural network (BPNN) model. The influences of different modeling conditions on the modeling accuracy were compared, and more accurate predictions were achieved when the iron content was higher than 2.13%. The BPNN model produced predictions with R2 and RMSE values of 0.955 and 0.336%, which were better than the predictions of the linear regression model (R2: 0.859, RMSE, 1.07%). We thus demonstrated that MSP can be used for fast, accurate, and non-destructive measurements of soil color and prediction of its iron content. Although the results may not be as precise as conventional laboratory analysis, they still provide more information with acceptable accuracy, which should have promising applications in forensic applications.