See more from this Session: Sensor-Driven Digital Soil Mapping: I
Monday, November 1, 2010: 1:15 PM
Hyatt Regency Long Beach, Seaview Ballroom B, First Floor
In Hawaii there is a need for techniques that can quantify variability of soil total carbon (Ct) across agricultural landscapes in space and time. However, traditional laboratory analysis of the large number of samples needed for accurate assessment of the variability of soil Ct is so time-consuming and expensive that it limits the degree to which this can be accomplished. New technologies such as visible/near-infrared (VNIR)- and mid-infrared (MIR)-diffuse reflectance spectroscopy (DRS) allow for samples to be scanned rapidly, inexpensively, and non-destructively. While DRS technologies have been used extensively in continental areas, they have yet to be used in the Pacific Islands. Thus, the objectives of this study were to: (1) employ VNIR- and MIR-DRS to scan archived Pacific Island samples from the NRCS National Soil Characterization Database; (2) collect new samples from agricultural fields in Hawaii under different soils types and management systems; and (3) calibrate and validate chemometric models for soil Ct prediction with the DRS spectral data. The project tested the following hypotheses: (1) fundamental wavelength regions in the MIR range result in better predictions of Ct than combination or overtone regions in the VNIR range; and (2) due to the high diversity of Pacific Island soils, subsetting the data (i.e. by taxonomy, texture) improves predictions of soil Ct. In spring 2010, we scanned 215 samples (150 pedons) from Hawaii and 99 samples (63 pedons) from Guam, Micronesia, Palau, Samoa, and the Marianas. In summer 2010, we collected >100 soil samples from agricultural soils on Kauai, Oahu, and Maui across a range of soil orders and management systems. In these diverse soil datasets, Ct values range from < 0.1 g kg-1 to > 60 g kg-1. Preliminary chemometric models indicate that the MIR spectral range is more effective at predicting Ct than the VNIR range.