Spiking regional vis-NIR calibration models with local samples to predict soil organic carbon in two High Arctic polar deserts using a vis-NIR probe

• Main Authors: Amanda L. Guy, Steven D. Siciliano, Eric G. Lamb
• Format: Text
• Language: English
• Published: Canadian Science Publishing 2015
• Subjects: Arctic; polar desert
• Online Access: https://doi.org/10.1139/CJSS-2015-004

Guy, A. L., Siciliano, S. D. and Lamb, E. G. 2015. Spiking regional vis-NIR calibration models with local samples to predict soil organic carbon in two High Arctic polar deserts using a vis-NIR probe. Can. J. Soil Sci. 95: 237-249. In situ visible and near-infrared (vis-NIR) spectroscopy is a potential solution to the logistic constraints limiting the accuracy and spatial resolution of soil organic carbon (SOC) estimates for Arctic regions. The objective of our study was to develop a calibration model based on field-condition soils for in situ applications to predict SOC in High Arctic polar desert soils from vis-NIR spectra. Soils (n=240) for calibration models were collected from three regional Canadian Arctic sites in 2010 and two local target sites in 2013. Local and regional calibration models were developed using partial least squares regression (PLSR). We assessed whether spiking or spiking and extra-weighting, regional models with calibration samples from local sites improved prediction of the local sites. The local model yielded successful prediction of target sites (R2=0.91) whereas unspiked regional models had poor prediction accuracy (R2=0.07 to 0.36; n=4). Spiking regional models with as few as 12 local samples greatly improved the SOC prediction of target sites; the best spiked models had R2 between 0.69 and 0.86. Extra-weighting spiking subsets in regional models yielded limited improvements in prediction performance. These results suggest that regional vis-NIR calibration models can be successfully used to predict SOC in High Arctic polar desert soils. The in situ application of these calibration models using field-portable instruments in remote areas, relative to traditional laboratory methods, can achieve higher sample sizes and the ability to characterize the spatial variability of SOC.