Geostatistical Modeling and Visible and Near-Infrared Reflectance Spectroscopy to Estimate Soil Carbon Stocks in Agroecosystems.

A full greenhouse gas accounting for agroecosystems requires estimation of baseline soil carbon (C) stocks, which can be prohibitively expensive. In this study we evaluated geostatistical modeling techniques and Visible and Near-Infrared Reflectance Spectroscopy (VNIR) for reducing soil C assessment costs for a dairy farm in Harford, NY with multiple land uses, including cultivation of silage corn, alfalfa, hay, pasture, small grain, and forest. We collected 320 samples in a spatially balanced design over a 232 hectare area to a depth of 30 cm. Samples were measured for VNIR, total C and nitrogen (combustion analyzer), organic matter (OM) percent (loss-on-ignition), active C (permanganate oxidation), bulk density, and soil texture. Minimum sample number required for given levels of confidence and precision varied among the soil quality parameters measured, and proxy measures for total C, such as VNIR and OM were evaluated. Partial least squares regression was used to predict soil quality parameters from VNIR spectral data. Geostatistical methods compared included ordinary kriging, regression kriging, and cokriging, with both VNIR and permanganate oxidizable C values used as secondary variables. In addition, within a subset 1-hectare silage corn plot 52 samples were collected to a depth of 75cm, and analyzed at 0-10cm, 10-20cm, 20-30cm, 30-50cm, and 50-75cm. We found that the spatial structure of surface C levels could be used to predict soil C at lower depths, thus reducing the number of required deep samples and reducing costs at a minimal loss of accuracy.