Arequipa Peru Agricultural Soil Chemistry with pXRF and VNIR spectroscopy

<p>These data are a subset from research conducted by the Soil Health Impairment, and Vulnerability (SHIVA) Team within the Arequipa Nexus Institute. The objective was to test these rapid spectroscopy techniques to quantify elemental content (pXRF) and predict organic carbon content (VNIR).</p>

<p>The data are summarized and analyzed within a manuscript entitled "Soil chemical assessment with portable XRF and VNIR spectroscopy: A case study of irrigated desert Agriculture in Arequipa, Peru". The abstract explaining methods and objectives in further detail is below.</p>

<p>The expansion of irrigated agriculture into desert systems can rapidly change soil chemistry and require shifts in management to maintain crop production. Interacting soil chemical properties such pH, electrical conductivity (EC), ionic composition, and soil organic carbon (OC), can all influence crop yield, but regular assessment of multiple soil metrics is time and cost intensive. This study evaluates low-cost soil assessment using paired analyses of traditional soil chemical assays (pH, EC, total carbon, total nitrogen, available phosphorus, and organic matter) with Visible Near Infrared (VNIR) spectroscopy to predict soil OC and portable X-Ray Fluorescence (pXRF) analysis to quantify additional elemental contents. For an initial case study in the Arequipa region of Peru, surface sandy loam soils (n=83) were sampled from drip irrigated vineyards within the hyperarid Majes District. Additional loamy soils from traditional agricultural fields around the City of Arequipa (AQP, n=29) serve as reference points with higher OC contents.  In the desert soils, OC ranged between 0.18-2.31 weight % (averaging 0.79 weight % OC), resulting in initially poor model prediction from VNIR spectra. Therefore, to maximize accuracy of the VNIR partial least squares regression, we included the higher OC soils from AQP. We then determined the optimal pre-processing procedure: an international soil standard correction, a robust baseline correction, multiplicative scatter correction, and Savitzky-Golay smoothing (R<sup>2</sup> = 0.81, RMSE = 0.12 (equivalent to 0.09% OC), RPD = 2.5). We achieved prediction power similar to previous VNIR spectroscopy analyses of low OC soils. For validation of the pXRF method we used the soil standard SRM 2711a. The elements detected with reliable accuracy (80-120% recovery) and precision (coefficient of variation <0.02) were Bi, Cr, Cu, Hg, Mn, Ti, and Zn. Patterns in the pXRF data correlated with other soil chemical properties at the site level, particularly with Ca, suggesting interactions with calcium carbonates. Analysis of soil metal content revealed elevated elemental Cr (198 mg kg<sup>-1</sup>), probably from a geologic source. Together these initial tests of pXRF and VNIR spectroscopy demonstrate their utility as screening techniques for rapidly changing soil chemistry under different agricultural management strategies and arid land development.</p>