Geostatistical Optimization of a Piezometer Network

By COLENCO, for the Research Centre for Alpine Environments (CREALP)

Optimisation of a piezometer network for an alluvial aquifer using intrinsic kriging (1998)

Since 1994, CREALP has been monitoring the alpine Rhône valley aquifer from the town of Sierre to Lake Geneva. Two complementary networks of piezometers were used: a system of 449 observation points where the water level is recorded manually twice a year, and a network of 114 piezometers equipped with automated recording devices. Having been implemented originally for the purpose of dam and highway construction the automated network was to be optimised in terms of its spatial distribution of the observation points in order to reduce the costs of monitoring and to depict the most representative image of the aquifer’s potentiometric surface on a regional scale.

In collaboration with CREALP, Colenco Power Engineering developed a methodology for the optimisation of the piezometer network based on geostatistical techniques. It was necessary to derive a specific geostatistical model to account for the drift in hydraulic gradient in the direction of flow. Intrinsic kriging was the tool of choice, allowing for the interpolation of the hydraulic potential. A kriging variance was assigned to each grid node to enhance subsequent mapping.

The variance quantifies the interpolation uncertainty and as such represents the basis for the optimisation process: observation points may be eliminated in zones of high information density, thereby reducing the total number of data points required in order to adequately characterise the spatial variability of the hydraulic potential. The increase in uncertainty caused by the interpolation in a network with a reduced number of data points was compensated for by incorporating data from the manually maintained piezometer network, processed to represent the average behaviour of the aquifer over a period of 3 years.

The geostatistical model for the alluvial aquifer of the plaine du Rhône covered an area of approx. 350 km2 and contained about 2700 kriging nodes. The geostatistical interpolations were performed with ISATIS.

The optimisation process resulted in a reduction of the total number of automated piezometers by 20% without diminishing the quality of the potentiometric surface map.