Geostatistics for Environmental Issues

The application of geostatistics to environmental questions is a key issue that cannot be ignored by those concerned by quality control and risk analysis. The data are spread out in geographic space and recorded at specific times. Hence geostatistical data analysis provides efficient and consistent models of the variability of the phenomena under investigation.

The kriging interpolator calculates a weighted average of values measured on scattered points. The weights are chosen optimally to take into account the structure of spatial correlation of the phenomenon (i.e. the variogram). In Isatis, kriging also deals with data provided with measurement errors and copes with stationary, intrinsic or non-stationary hypotheses.

(GIF) Whatever the scale, map quality is enhanced by the integration of additional information. Auxiliary information can easily be integrated for cokriging interpolation. When the auxiliary variable is a cofactor known everywhere, collocated cokriging or external drift makes the map of the primary variable “resemble” the cofactor map.

Because of the probabilistic framework of geostatistics, the variance of the interpolation error can be calculated. This estimation variance only depends on the variogram and the data configuration. Therefore, it is possible to tune additional sample locations to reduce the uncertainty in under-sampled areas.

(JPEG) Geostatistical Conditional Simulations (TB, SGS, SIS, ...) compute equally probable images of the reality. From numerous simulations, risk analysis provide relevant information like the probability to overcome a threshold.


(GIF) When rehabiliting a polluted site, errors in spoiled areas assessment may lead to serious sanitary and financial consequences. Because geostatistics are based on probabilistic models, they are appropriate for assessing contamination risks. They provide accurate and reliable estimations of the volumes to be cleaned up.

The stochastic simulations allow to calculate a probability to exceed a threshold of pollution. This probability accounts for the mesh size of the cleaning technique and the rehabilitation objectives. They provide the probabilistic distribution of the contaminated volumes. The amount of soil to clean up as well as its location are directly deduced from the risk one accepts of leaving spoiled blocs in place.