Using geostatistical methods to help optimizing an existing groundwater monitoring network.

Nathalie Courtois (1) , Claire Faucheux (2)

1 LMTE - Laboratoire de Modélisation des Transferts dans l'Environnement
2 Geovariances

The studied site is a research center, located in the South of France. Three superimposed aquifers are in presence in, from surface to the bottom, Quaternary, Miocene and Cretaceous Formations. Since its creation in the 60's, the center started to constitute a groundwater monitoring network dedicated to the survey of both groundwater levels and quality which has continuously evolved as a function of different needs: survey of new facilities, knowledge of flow rate and directions, improvement of the 3D hydrogeological model used for flow and transport simulations, etc. The monitoring network is now composed of about 400 wells distributed in the three superimposed aquifers. Geostatistical methods are used to help optimizing this network in terms of number and spatial distribution of the wells. An original and specific geostatistical methodology is developed. First, variograms are calculated on hydraulic heads surveys at different dates, covering a large panel of hydrological conditions. Corresponding head distributions are then constructed by kriging. For some aquifers, as hydraulic heads and elevations are correlated, a smoothed digital elevation model is used as external drift. Then, trajectories starting from specific zones (facilities, buildings, etc.) are calculated, in order to highlight the downstream positions. Finally, a network optimization is conducted in two parts: (i) sequential addition of new wells, allowing to decrease the uncertainty on hydraulic head in zones with few information, (ii) sequential removal of existing wells, on a criteria of geometrical redundancy. During the calculation process, several constraints are imposed such as a minimal thickness of geological formation to add a new well and a minimal distance to existing or added wells. This sequential and automated process allows testing different configurations (number of additions/removals, minimal distance between wells, etc.). As a result, the study leads to an optimized list of new wells to add and existing wells to remove for each of the three aquifers. This list is a precious base for optimization that has to be further discussed, taking into account other criteria that cannot be included in the geostatistical analysis, e.g. presence of faults, available space, access conditions for drilling machines, etc.

Domaines: Environnement et Société

Nathalie Courtois, Claire Faucheux. Using geostatistical methods to help optimizing an existing groundwater monitoring network.. geoENV2022 - the 14th International Conference on Geostatistics for Environmental Applications, Jun 2022, Parme, Italy. ⟨cea-04551773⟩