Geostatistical analysis of sampling uncertainty at the Tollesbury Managed Retreat site in Blackwater Estuary, Essex, UK: Kriging and cokriging approach to minimise sampling density

Soil and sediment samples were collected from an area of low value agricultural land which has been allowed to flood through tidal inundation as a part of a coastal management scheme. A grid-sampling strategy was applied before and after a sea wall was breached. Sample properties (Al, Ca, Cd, Cr, Cu, Fe, K, Li, Mg, Mn, Na, Ni, Pb, Zn, water content and particle size) were analysed as the original data sets. Both full and reduced data sets were used to generate experimental semivariograms or cross-semivariograms prior to kriging and cokriging analysis. The results indicated that reduction of the original sampling density of almost 50% in 1995 and by 40% in 1996 would have resulted in no loss of spatial information. The analysis of data indicated that a 50% reduction in the 1995 data set was possible, however, estimation from only 19 sample points maybe statistically unreliable, and a minimum sample size of 28 is suggested for subsequent kriging and cokriging analysis in the study area. Comparison of the kriging technique and traditional robust ANOVA method for estimating the uncertainty in sampling indicates that the former method is more suited to data exhibiting an anisotropic spatial distribution.