Comparison and detection of total and available soil carbon fractions using visible/near infrared diffuse reflectance spectroscopy

Abstract

Available carbon (C), measured as hot water extractable carbon (HC), has been suggested as an indicator to assess management effects on soil C fractions, transformations among C fractions, and carbon sequestration potential. Their upscaling to coarser scales has been hampered due to the cost and labor to measure C fractions. With the advent of visible/near-infrared-diffuse reflectance spectroscopy (VNIR-DRS) to infer on soil C fractions, new opportunities exist to combine sensor-based and geostatistical methods to estimate them across landscapes. Our research objectives were to (i) assess the accuracy of VNIR-DRS for estimating available C (HC) in soil, and (ii) compare interpolated HC estimates derived from laboratory measurements and VNIR-DRS derived estimates. The study site was within the Rio Grande floodplain in Quemado, Texas with mixed acreage of Arundo donax L. and Cynodon dactylon (L.) pers. Soil samples were taken in 10-cm depth increments from 0- to 50-cm depth, at 125 locations across a 34.5 ha field. Total C (TC, dry combustion), inorganic C (IC, modified pressure calcimeter), and HC (extraction at 80 °C) were measured and organic C (OC) was calculated by difference (TC − IC). Partial least squares regression (PLSR) and boosted regression tree (RT) were used to relate TC, HC, IC, and OC to VNIR-DRS spectra. Ordinary kriging was used to interpolate HC derived from analytical measurements and the values derived from the best spectral estimation model (PLSR). All four C pools were estimated well using VNIR-DRS and PLSR (R 2 > 0.77, residual prediction deviation > 2), ratio of performance to inter-quartile distance > 1.62. Overall, the maps estimating HC derived from VNIR-DRS estimates of HC were comparable to those from analytical measurements. In conclusion, VNIR-DRS reliably predicted soil C, as measured by HC.