Characterization of nitrate long term trends at regional scale using statistical and geostatistical tools
Résumé
European Water Framework Directive requires the Member States to characterize the quality of their ground waters, toidentify trends and to take measures to achieve a good chemical and quantitative status by 2015. This is why, together with two water agencies, the BRGM (France) conducted a study on two large basins (Loire-Brittany basin, 157,000 km² and Seine-Normandy Basin, 95,300 km²) in order to understand the current status of groundwater nitrate contamination and describe trends at punctual and regional scales. Time variograms of water table level and nitrate concentrations (around 400 time series collected from the national database in each basin for each variable) were automatically computed and fitted and a hierarchical ascendant classification (HAC) of the variogram models was performed. Results were combined with hydro-geological map, to obtain homogenous sectors for spatial trend detection. A modified CUSUM algorithm was applied to determine the date of the major change of trend in nitrate concentrations. The Man Kendall and the Kendall Regional non parametric tests, coupled with the Sen slope estimator procedure, were used to estimate the "post-change" trends both at punctual and regional scales. At last, cross covariances between water table level and nitrate concentration time series were computed in order to i) determine correlations, ii ) detect possible time shift between time series and iii) identify any spatial distribution in this shift. The application of statistical and geostatistical semi-automated procedures appears relevant to describe groundwater nitrate concentration trends at different scales. The proposed methodology is robust, allows the rapid processing of a large number of data and helps stakeholders to achieve the requirements of the Water Framework Directive.