After I had finished my PhD I started to work on an exciting project: improving the maps that help the state agency of the environment in Baden-Württemberg, Germany (LUBW) assess the state of the state-wide groundwater resources.
The European Water Framework Directive and related national law mandates that the water resources are monitored. The state of Baden-Württemberg maintains an extensive observation network for this and other purposes in which about 300 solutes are monitored at least annually. Maps are produced for each of those solutes and these maps form the basis for decisions related to water resource management, e.g., if a certain area should be monitored more closely.
Previous methodologies used to create these maps had various problems: Land-use categories used as secondary information imposed unrealistic sharp boundaries in the solute maps, Kriging based approaches led sometimes to unrealistic negative-valued estimates, stationary within the large domain (~40,000km2) was an unrealistic assumption, among others.
Hence, we developed the concept of locally mixed distributions that are based on the composition of the land cover within the neighbourhood of each interpolation location. The local distribution varies at each interpolation location and is then used as secondary information for spatial interpolation. Categorical secondary information such as land cover can be included in a quantitative model of spatial dependence for interpolation purposes.
I'd like to thank my colleagues Theresia Heißerer, Michael Eisele, and András Bárdossy who have been instrumental in advancing the geostatistical methodology and bringing the software into operational use.
At the same time, I'd also like to thank my colleagues at the state agency of the environment, Burkhard Schneider, Wolfgang Feuerstein, and Emil Hildenbrand for their wonderful cooperation, their recognition of the problem, and their eagerness to solve this real-world problem with innovative geostatistical methods!