See more from this Session: Spatial Predictions In Soils, Crops and Agro/Forest/Urban/Wetland Ecosystems: III (Includes Graduate Student Competition)
Tuesday, October 18, 2011
Henry Gonzalez Convention Center, Hall C, Street Level
In order to derive a digital map of the soil organic matter (SOM) content of surface layer (0-30 cm) in the Monteregie area, near Montreal (Canada), soil morphological and analytical databases (MSDB and ASDB) were used with kriging and cokriging methods. The main objective of this study was to assess the usefulness of MSDB in terms of reducing the root mean square error of prediction (RMSEP) of SOM measured by the Walkley-Black method on 2524 soil samples. This MSDB contains 48481 soil profile data collected during soil survey works (1982-2009) realized by Agriculture and Agri-Food Canada. Soil color has been described using the Munsell system: hue (H: 2.5YR-5Y, recoded as 2.5 to 15), value (V) and chroma (C). A color indice combining these three parameters (IC=(16-H) (10/(V*C)) as well as soil surface texture groups (TEXA: 1 to 9) were used as ancillary variables for cokriging SOM. Anisotropic and isotropic semivariograms, block kriging (2 x 2) and cokriging have been computed using GS+ (V5) and the Geostatistical Analyst toolbox of ArcGIS (V9.3). The semivariance map showed no significant anisotropy. As indicated by the low nugget (C0) to sill ratio (C0/C0+C < 0.1), a strong structure with a spatial range (A0) of 6.1 km, was observed for SOM. A cross-validation test (Jackknife method) showed moderately good prediction accuracy (r2=0.67; RMSEP=3.89%) when using ordinary kriging for mapping SOM. Using IC and TEXA as co-variables and cokriging method for producing the digital map of SOM resulted in very small reduction of RMSEP (gain(-)=1-3%). When available, morphological soil data can improve the accuracy of digital soil maps, but reduction of RMSEP was lower for SOM than sand, silt and clay contents determined from laboratory analysis and field estimation of texture. Future works will assess earth observation data as co-variables for cokriging SOM.