Variography analysis is the core of geostatistics. The accuracy of a geostatistical characterization depends strongly on the representativeness of the variogram model. Variography analysis also helps in designing a consistent interpolation process.
Understand and model the spatial behavior of the contaminants
Geostatistics, beyond conventional statistical analyses, allows studying data spatial variability. This is performed through the variogram which measures how dissimilar the neighboring data are. More specifically, it quantifies how much the contamination level measured on two samples varies depending on the distance between the samples. The variogram is a function that is required for the geostatistical interpolation process.
From no spatial structure (pure nugget effect) as it is generally found with heterogeneous wastes in trenches (as on the left side of the picture), to very continuous contamination in groundwater plume (as on the right side of the picture), with contamination of soils and concrete in between, the variogram reveals and models the spatial correlation of the phenomenon.
The spatial structure is crucial for the overall geostatistical approach and strongly impact sampling strategies.
Break down the different variability contributions
The variogram quantifies the variability at any scale. Thus the spatial structure can be directly compared to :
– the variance of sample duplicates, which can be significant if the sampling protocol is not relevant;
– the variance of measurement replications, which is generally negligible.
In addition, coordinate errors (with poor GPS accuracy for example) can artificially generate a nugget effect and therefore compress the real spatial structure of the phenomenon.
Not to mention the impact of data support (punctual value in comparison to average value on a given surface or mass) impacts the variogram behavior at very small scales and can be analyzed separately.