Abstract A geostatistical analysis assumes some form of stationarity of the variable under study, but different types of stationarity exist and often spatial data exhibit some form of nonstationarity. However, most studies assume one type of nonstationarity and consequently apply one type of interpolation method within the study area. A study area of 8×18-km area was selected because it was expected to contain complex nonstationary conditions in soil texture. Therefore, four geostatistical interpolation methods were evaluated in their ability to account for different types of nonstationarity in the topsoil silt content: two univariate interpolation methods, ordinary kriging (OK) and universal kriging (UK), and two bivariate methods, simple kriging with varying local means (SKlm) and ordinary cokriging (OCK). A digital elevation model (DEM) was used as the exhaustive secondary information for the bivariate methods. Two kinds of nonstationary conditions were identified inside the study area: (1) a large-scale trend in both the silt content and elevation, with a strong correlation between them, and (2) a very strong local fluctuation around a mean value, representing a local nonstationarity. Consequently, different techniques were applied in different parts inside the study area: the global trend was best accounted for by OCK and UK could best account for the local nonstationarity. After combining the results of the two prediction methods, it was found that the overall estimation of the silt content was more precise than when any single method was used over the entire study area.
[1]
A. Journel,et al.
When do we need a trend model in kriging?
,
1989
.
[2]
Boucneau,et al.
Integrating properties of soil map delineations into ordinary kriging
,
1998
.
[3]
Alex B. McBratney,et al.
Further Comparison of Spatial Methods for Predicting Soil pH
,
1990
.
[4]
D. Myers.
Matrix formulation of co-kriging
,
1982
.
[5]
Donald E. Myers,et al.
Basic Linear Geostatistics
,
1998,
Technometrics.
[6]
N. Cressie.
Kriging Nonstationary Data
,
1986
.
[7]
B. Efron.
The jackknife, the bootstrap, and other resampling plans
,
1987
.
[8]
Richard Webster,et al.
Is soil variation random
,
2000
.
[9]
Marc Voltz,et al.
A comparison of kriging, cubic splines and classification for predicting soil properties from sample information
,
1990
.
[10]
Gary W. Hergert,et al.
Incorporating Spatial Trends and Anisotropy in Geostatistical Mapping of Soil Properties
,
1997
.
[11]
Timothy C. Coburn,et al.
Geostatistics for Natural Resources Evaluation
,
2000,
Technometrics.
[12]
Pierre Goovaerts,et al.
Using elevation to aid the geostatistical mapping of rainfall erosivity
,
1999
.