نوع مقاله : مقاله پژوهشی
نویسندگان
بخش علوم خاک، دانشکده کشاورزی، دانشگاه شیراز، ایران
چکیده
کلیدواژهها
موضوعات
عنوان مقاله [English]
نویسندگان [English]
Iron is an essential plant nutrient whose availability and environmental role depend on the distribution of its chemical forms in soil, making accurate measurement important. Conventional laboratory methods are difficult, costly, and time-consuming, so new approaches are needed. Visible and near-infrared (Vis-NIR) spectroscopy is rapid, non-destructive, and cost-effective. This study aimed to estimate soil iron forms using Vis-NIR reflectance data with partial least squares regression (PLSR) and stepwise multiple linear regression (SMLR) to develop spectral transfer functions. After measuring iron forms and spectral reflectance of 130 soil samples, PLSR was used to predict iron forms and identify effective spectral bands. Selected bands were then entered into SMLR to develop simple spectral functions. Models were calibrated with training data and validated with test data using R², RMSE, and mean error. Validation R² values for PLSR models were: carbonate-bound iron (0.37), exchangeable iron (0.63), organic-bound iron (0.53), amorphous iron oxides (0.67), crystalline iron oxides (0.68), manganese oxide-bound iron (0.59), residual iron (0.51), and total iron (0.54). For SMLR-based functions, values were 0.41, 0.55, 0.44, 0.41, 0.65, 0.53, 0.48, and 0.48, respectively. Although PLSR outperformed SMLR, the difference was not substantial for some properties. Thus, Vis-NIR spectroscopy combined with regression models, especially PLSR, is an effective, low-cost, non-destructive method for estimating soil iron forms. The developed spectral functions should be validated before use in non-calcareous soils.
کلیدواژهها [English]
Iron is one of the most important essential nutrients for plant growth. In soil, iron can exist in various forms depending on conditions, ranging from soluble forms to precipitates or other forms bound to different compounds. The distribution of these different iron forms in soil is important for both plant growth and environmental considerations. Numerous soil-related and external factors directly and indirectly influence the chemical forms of iron in the soil. One factor affecting the distribution of iron's chemical forms is land use type, as well as the type of vegetation present in the soil. Chemical forms of iron can have various effects on soil properties and the environment. For instance, iron oxides contribute to the protection of organic matter from decomposition and enhance the stabilization of soil organic carbon. Plant-available iron depends on the relative distribution of its different chemical forms in soil. Accurately measuring bioavailable iron in soils is crucial for assessing soil conditions and enabling precision agriculture. Traditional laboratory methods for determining the chemical forms of iron are often laborious, expensive, and time-intensive. Therefore, there is a clear need to develop new, simple, and rapid alternatives. Compared with traditional laboratory methods, visible (Vis) and near-infrared (NIR) spectroscopy is rapid, non-destructive, and cost-effective. Therefore, the objective of this study was to estimate the chemical forms of soil iron using Vis-NIR spectral reflectance data via partial least squares regression (PLSR) and stepwise multiple linear regression (SMLR), and ultimately to develop simple, applicable spectral transfer functions (STFs).
The present study was conducted in an Acacia victoriae plantation forest near Galkuyeh Village, within the Zarindasht County area, in the southeastern part of Fars Province. The study area is part of the larger Banaruyeh watershed, a small portion of Tang-e Charkhi (Tammab-e Fars) and Kol-e Mehran, and a small part of the Mond River basin (Tammab-e Iran). In the present study, a total of 130 surface soil samples (2 kg each) were collected from a depth of 0–30 cm using a random-systematic sampling method. Out of these, 50 samples were taken from within the canopy cover (shadows) and 50 samples from outside the canopy cover of Acacia victoriae forest trees aged 18–20 years. Additionally, 30 samples were collected from surrounding land uses adjacent to the forest, including rangeland, rainfed farmland, irrigated cropland, and irrigated orchard land (10, 10, 5, and 5 samples, respectively). The location of each sampling point was recorded using a Global Positioning System (GPS). The soil samples were packaged in plastic bags, labeled, and transferred to the laboratory. After air-drying and passing the soil samples through a 2-mm sieve, they were used to determine physical and chemical properties, chemical forms of iron, and to perform Vis-NIR spectroscopy using standard methods. In this study, PLSR modeling was initially applied to predict different chemical forms of soil iron using spectral reflectance bands in the Vis-NIR range. The results of this model were analyzed to identify and select the most influential and significant single spectral bands. Regression coefficients ± standard deviation (SD) were used to standardize and select the most important wavelengths. Subsequently, all identified effective single spectral bands for each target variable were entered into multiple linear regression (MLR) models as independent predictor variables for developing simple and practical STFs. Following data normalization using transformation methods, PLSR and MLR models were developed with training datasets to predict different soil iron forms, and the models were validated using independent test datasets. Model performance was evaluated using the coefficient of determination (R²), root mean square error (RMSE), and mean error (ME). All regression modeling steps were carried out using Unscrambler X software, version 7.9.
The results showed that the inverse carbonate-bound iron, logarithm of exchangeable iron, inverse organically-bound iron, logarithm of amorphous iron oxides, logarithm of crystalline iron oxides, logarithm of manganese oxide-bound iron, logarithm of residual iron, and logarithm of total iron were estimated with R² values of 0.37, 0.63, 0.53, 0.67, 0.68, 0.59, 0.51, and 0.54 using PLSR; and 0.41, 0.55, 0.44, 0.41, 0.65, 0.53, 0.48, and 0.48 using STFs based on MLR models and Vis-NIR spectral bands. The study indicated that although PLSR generally outperformed STF-based MLR models in estimating most soil iron chemical forms, the performance difference was not substantial in some cases. Therefore, the developed STFs can be used to provide an acceptable estimate of certain iron fractions.
The findings of this research demonstrate that combining Vis-NIR spectroscopy with advanced regression models, such as PLSR, is an effective tool for estimating the different chemical forms of soil iron. Vis-NIR spectroscopy can serve as an indirect, cost-effective, and non-destructive method for estimating soil iron fractions. It is recommended that the STFs developed in this study be tested and validated before application in other environments (e.g., non-calcareous soils).
The study was funded by Shiraz University, Iran.
Conceptualization, M.B., S.A.A., A. R., A.A.M. and M.Z.; methodology, M.B., A.A.. A. R., and A.A.M.; software, A.R., H. M., and A.A.M.; validation, A. R., H.M., A.A.M., M. B., and S. A. A.; formal analysis, A. R., H.M., A.A.M., M. B., and S. A. A.; investigation, A. R., H.M., A.A.M., M. B., and S. A. A., and M.Z.; resources, M. B., S. A. A., A.A.M., A.R.; data curation, A.R., and H.M.; writing-original draft preparation, A.R., A.A.M., and H. M.,; writing-review and editing, M.B., S.A.A., A.R., A.A.M., M.Z., and H.M.; visualization, M.B., S.A.A., and A.A.M.; supervision, M.B., S.A.A., and A.A.M.; project administration, M.B.; funding acquisition, M.B., S.A.A., and A.A.M..
All authors have read and agreed to the published version of the manuscript.
All data generated or analyzed during this study are included in this published article.
The authors would like to thank Shiraz University for providing all the needed facilities.
The authors avoided data fabrication, falsification, and plagiarism, and any form of misconduct.
The authors declare no conflict of interest.
)