Abstract

GRIMA-OLMEDO, Carlos et al. SURVIVAL ANALYSIS TO THE DETECTION AND CORRECTION OF POLLUTION IN SERIES WITH VALUES BELOW DETECTION LIMIT. Rev. Int. Contam. Ambient [online]. 2019, vol.35, n.1, pp.165-178.  Epub Aug 21, 2020. ISSN 0188-4999.  https://doi.org/10.20937/rica.2019.35.01.12.

Under article 17 of the Water Framework Directive the European Union was required to establish a framework to protect groundwater and set criteria for determining its chemical status and define environmental quality standards. Considering the specific characteristics of contaminants in groundwater, it is necessary to use adequate statistical tests. Annex IV of the Groundwater Directive establishes that all measurements below the limit of quantification (LOQ) must be replaced by half of the highest LOQ, with the exception of total pesticides. Although it is possible to estimate mean values in data censored with a single LOQ, its substitution distorts the confidence intervals used to verify tolerance with thresholds. The official monitoring networks lack series with values below the LOQ. In order to set up a methodological framework for its evaluation, nitrate concentration values have been taken in order to perform a subsequent artificial censorship. Two groundwater bodies with different monitoring characteristics and frequencies have been selected, Plana de Sagunto, in the eastern part of Spain, and a sector of the Central Valley (Chile), located around the systems of Tinguirica and Antivero rivers. Through the simulation of different percentages of censorship to data, several methods have been tested. Specific techniques for estimating censored values, such as Kaplan-Meier or robust regression have proven to be the most efficient to verify compliance with environmental standards.

Keywords : censored data; water framework directive; Kaplan-Meier; robust regression.