Colloquium

Digital soil mapping using legacy soil maps for predecting soil properties in Sanmatenga, Burkina Faso

Organisator Laboratory of Geo-information Science and Remote Sensing
Datum

wo 9 april 2014 13:00 tot 13:30

Locatie Gaia, building number 101
Droevendaalsesteeg 3
101
6708 PB Wageningen
+31 317 48 16 00
Zaal/kamer 1

by Charl Wong (Suriname)

Abstract

Digital soil mapping (DSM) is a method to model the relationship between observed soil profile data with environmental covariates to predict soil properties at unvisited locations. The aim of this study is to focus on the usability of legacy soil maps with different scales as environmental covariates in DSM to predict three target variables (pH, Cation Exchange Capacity (CEC) and soil depth) for the area of Sanmetanga province in Burkina Faso. Four methods were used to include legacy soil map information in DSM, namely: 1) Regression kriging with legacy soil map as categorical variable (CAT); 2) Stratified kriging using the delineations of the legacy soil map as map unit boundaries (STK); 3) Legacy soil map used as observed information obtained from the accompanying report (OBS); and 4) Combining the OBS result with kriged residuals (OBSRES). Regression kriging without use of legacy soil map information is used as a reference method (RK). The accuracy of all methods was assessed with the Correlation (r), Mean Error (ME) and the Root Mean Square Error (RMSE). The best method for predicting the soil properties pH and CEC was the CAT method using a legacy soil map with a scale of 1:100.000, where the pH had an r of 0.53, an ME of -0.004 and an RMSE of 0.76 and the CEC had an r of 0.59, an ME of 0.011 [cmol+/kg] and a RMSE of 4.54. The best method for predicting soil depth was the reference method, which had an r of 0.26, an ME of 0.32 [cm], and an RMSE of 61.42. For the reference method, the predicted pH had an r of 0.48, an ME of -0.009 and RMSE of 0.78, for CEC an r of 0.54, an ME of 0.047 and RMSE of 4.71. The obtained results show that the soil map provided disappointingly little information. However, the methods and data used in this thesis create a basis for a possibly more accurate DSM using legacy soil maps for the province in Burkina Faso, which should be extended to include a much denser soil profile dataset and more adequate covariates including more detailed soil information from the legacy soil maps.

Keywords: Digital Soil Mapping, Legacy Soil Map, Regression Kriging, Stratified Kriging, Variogram