Vis-NIR-SWIR spectroscopy in mapping of soil physical-chemical attributes in areas cultivated with irrigated mango in the Brazilian semi-arid region

Abstract

The adoption of specific soil management in agricultural areas requires a series of soil analyses, which is time-consuming and costly. In this context, Vis-NIR-SWIR spectroscopy (visible - near infrared - short-wave infrared) emerges as an alternative to determine soil attributes quickly, with lower cost and few environmental impacts. Thus, the objective of this study was to map the physical-chemical attributes of the soil in areas cultivated with irrigated mango in different soil classes in the Brazilian semi-arid region using Vis-NIR-SWIR spectroscopy. In total 318 soil samples were used. For these samples, the reflectance spectra were obtained (350 to 2500 nm) and the values of pH, EC (electrical condutivicty), Ca2+, Mg2+, K+, Na+, Al3+, P, H + Al, TOC (total organic carbon), sand and clay were determined by standard analytical methods. For the development of predictive models, the techniques of Partial Least Squares Regression (PLSR) and Multiple Linear Regression (MLR) were used. For the predictive models that had R2 above 0.50, the semivariograms and maps of the soil attributes determined by the reference methods and by Vis-NIR-SWIR spectroscopy were constructed. The PLSR and MLR regression models provided strong predictions for sand, clay and TOC, moderate for Na+, Ca2+ and Mg2+, weak for pH, CE, K+ and Al3+ and very weak for H + Al and P. The maps of soil attributes showed the existence of spatial correlation with each other. Therefore, the Vis-NIR-SWIR spectroscopy is a potential tool for evaluation of soil and mapping of fruit growing areas.

Recent Trends on the use of Infrared Spectroscopy for Soil Assessment

“…Carbon stock or storage in soil presents relevance for both environmental and agricultural contexts, and has been explored by means of portable MIR spectroscopy in vineyard soils, achieving reliable results for varying geopedological conditions [24]. Soil carbon stock was also mapped via near-infrared spectroscopy in a desertifi cation region from northeastern Brazil, a work in which the authors indicated that this technology could be used by government agencies for environmental monitoring [25]. The estimation of Soil Inorganic Carbon (SIC) is also an interesting approach which has been undertaken in the Vis-NIR range, coupling variable selection algorithms and deep learning methods, detecting SIC content quickly and accurately [26].…”