Digital Soil Mapping At the Microwatershed Scale (Order 2) Using Soil Legacy Data and Fine-Resolution Satellite Images.

Cokriging models have been successfully employed to improve the estimation accuracy of a number of soil physical properties. High spatial resolution satellite imagery has the potential for mapping. The objective of the project was to improve the prediction of digital soil surface texture maps using high spatial resolution imagery. The project was carried out on an experimental sub-watershed (470 ha) located near Quebec City. Soil particle size fractions were determined by the hydrometer method. The sand, silt and clay of compositional data were transformed by using isometric log-ratio (ILR). The dataset was split into two groups, one was used as the training dataset and the other one was used as the validation dataset. Reflectance data and multiples algorithms band ratios were used as ancillary variables to improve the accuracy of soil surface texture maps. The ancillary variables were extracted from two high resolution images acquired with the IKONOS and Quickbird satellites. Anisotropic and isotropic semivariograms, ordinary kriging and cokriging were computed using Gesotatistical Analyst toolbox of ArGIS. No significant anisotropy was detected. Variographic analyses revealed the presence of a high spatial structure (C/Co+C > 0.9) allowing DSM created by kriging. Thereafter, covariates were used for the cokriging approach to see if it improve the accuracy of the maps create by kriging. A back transformation was applied to kriging and co-kriging models to obtain sand, silt and clay maps. Then, the models performance was assessed using the validation dataset by comparing the root-mean-square error (RMSE) of the kriging and the cokriging models. Cokriging models give lower RMSE between estimated and measurements than kriging B3, BI, SCI and IOR of the Quickbird. The most significant improvement of the estimation is with B3 of the Quickbird, the improvement is up to 12.44% for clay, 5.02% for sand and 3.54% for silt.