Estimation of total iron in soil using a water-absorption-peak-based color reconstructing machine (WCRM) method
Soil iron (Fe) is one of the necessary trace elements for plant growth. An iron deficit will affect the growth of plants, but excess iron can pollute soil and water and harm human health through the food chain and water sources. Effective monitoring of the dynamic change of soil Fe is important to guarantee soil and water quality, and human health. Soil spectroscopy technology has recently received extensive attention and has been used in the estimation of soil properties, because it has the advantages of real-time and dynamic operation when compared with the traditional chemical measuring approaches. And many researchers in the soil color research field have found that Fe is highly correlated with color information. Color is frequently used to infer soil properties such as total iron (Fe). However, soil water affects the soil color and severely limits the estimation accuracy for Fe.
Based on this, to improve the Fe estimation accuracy. this article, a research team from Nanjing University of Information Science and Technology and Henan University of Technology designed a multicolor-adapted dewatering method, and developed a new technology named the water-absorption-peak-based color reconstructing machine (WCRM) method. They collected topsoil (0-0.1 m) samples from several land-use types (including woodland, cropland, and bare land) from Liuhe (32.22 N, 118.88 E) and Pukou (32.20 N, 118.70 E), two regions in Nanjing, Jiangsu province, China. And then conducted soil Fe chemical analysis and obtained color and water data for the soil samples using an ASD FieldSpec 3 spectrometer under laboratory conditions.
Results:
(a) Pre- and (b) post-reconstructed L*a*b* (i = 1.5, j = 3.5, k = 1.5) for the soil sample with Fe of 36.22 g/kg. L*a*b*: lightness, redness, yellowness; Circles, air-dried; triangles, water-added.
Plots of measured versus predicted for soil Fe using (a) the water-absorption-peak-based color reconstructing machine (WCRM) method and (b) the adaptive neuro-fuzzy inference system (ANFIS) method. Circles: validation samples; × : calibration samples.
Conclusions:
Soil water seriously affects the accuracy of Fe content estimation from soil color. In this paper, to address this issue, the WCRM method has been proposed to improve the Fe estimation accuracy by reducing the influence of soil water. Based on the two water absorption peaks around 1400 and 1900 nm, a series of reconstructed L*a*b* data at different i, j, and k values were obtained and were subsequently used to produce the corresponding Fe models. The model for the 1900 group when i = 0.5, j = 4.5, and k = 2.5 produced the optimal performance among all 2000 Fe models, and was therefore treated as the final WCRM model. The ANFIS model without reconstruction processing was also established to evaluate the effectiveness of the WCRM model. The results showed that the ANFIS model provided unsatisfactory results because of the influence of soil water. The WCRM model performed much better than the ANFIS model, with an R2 of 0.603, an MREv of 6.6%, an RMSEv of 2.914 g/kg, an RPDv of 1.483, and an RPIQv of 2.492. The comparison results confirmed the dewatering effect of the WCRM model. Thus, the WCRM method represents a new tool for the accurate estimation of Fe content in soil. In the future study, the authors will investigate whether the WCRM method remains a valuable tool when studied with onsite soil samples.