Stochastic modeling of space-time variability of nitrate pollution using indicator geostatistics and transition probability

Abstract:

Public supply to Beja, a major town of Alentejo region (Portugal), depends partially on groundwater resources of a gabbro-dioritic aquifer. Due to agricultural activities the aquifer is vulnerable to nitrate contamination. The median concentration fluctuates between 53 and 86 mg/l, however the values can easily reach seasonally seasonally over 100 mg/l and a maximum around 200 mg/l. For these reasons there is a growing need to assess nitrate pollution in order to optimise urban and agriculture water management as well as ecological aspects related to large scale diffuse pollution. For this purpose a hydrochemical monitoring was carried out between July 1997 and July 2000 to assess spatial and temporal variability of nitrate contents in the aquifer due to seasonal fertilization and rainfall episodes. The study of the space-time variability of nitrate diffuse pollution, during 36 months, was carried out using advanced geostatistical techniques. The study encompassed the following steps: 1-Several thresholds and indicators of the nitrate were build, and a structural analysis was performed, 2-The indicator structural analysis has shown a phenomenon with gradual variations, i.e. a transition through neighboring values. The fact suggests the use of a diffusion-type model for kriging purposes. 3- Iso-probability contour maps of the nitrate content exceeding a specific threshold value were determined 4-Transition probabilities in function of the distance, i.e. the probabilities to exceed a specific threshold when entering the domain of a lower threshold, were calculated on the basis of the ratios between the cross-variograms of two indicators and their simple variograms. These transition curves represent the probabilities with which, getting into the domain of the values ≥ Z, one meets a value ≥ Z′ (upwards) and the probabilities with which, getting into the domain of the values < Z′, one meets a value < Z (downwards). This methodological approach provides a good image of the spatial correlation patterns where the diffuse phenomenon is well characterized by continuity/non continuity models.

Año de publicación:

2002

Keywords:

Fuente:

scopus