DMCA
Statistical validation and spatio-temporal modelling of river monitoring networks (2007)
BibTeX
@MISC{Clement07statisticalvalidation,
author = {Lieven Clement},
title = {Statistical validation and spatio-temporal modelling of river monitoring networks},
year = {2007}
}
Share
 |
 |
 |
 |
||
OpenURL
Abstract
The European Water Framework Directive (WFD)(EC, 2000) is one of the driving forces in environmental policy in the European Union (EU). The WFD's overall environmental objective is the achievement of a 'good status' for all of Europe's surface- and ground waters within a 15-year period. Its implementation is a big challenge for the European environmental managers. The WFD triggered the wa- ter authorities to design monitoring programmes. Thus, large amounts of environmental data are being collected, processed and stored throughout Europe. They are for instance needed for a coherent and comprehensive overview of the water status, to identify pressures on water systems, as a warning system for detecting negative changes in the water quality and to detect trends. Like other environmental data, water quality data have a complex nature. They contain a considerable amount of noise, due to their natural variability and the measurement error. They often contain missing values, are often censored due to the detection limits of the measuring methods, and are commonly gathered on irregular time intervals. They also may be mutually dependent, non-normally distributed, possess cyclic variations and show nonlinear trends (e.g. Hirsch et al., 1982, Van Belle and Hughes, 1984, Cai and Tiwari, 2000 and McMullan, 2004). Due to the large amount of data and their complex nature, modelling has become an essential tool to extract information from these observations. Within the research community, monitoring and modelling have now become generally accepted to be interlinked activities (e.g. Parr et al., 2003; Højberg et al., 2007). From this perspective, models can be used for a number of different purposes. For instance, they can be useful to assure data quality, for inter- and extrapolation in time and space, to increase the conceptual understanding of the underlying processes, to evaluate the impact of (future) management strategies, to assess the effect of anthropogenic activities and to design monitoring programmes (Højberg et al., 2007). High quality data is essential for an adequate management of the water resources. Therefore, quality assurance is specifically mentioned as an important activity in the WFD guidance document on monitoring (EC, 2003; Højberg et al., 2007). Thus before the data can be used in an assessment, they have to be validated. Errors might be introduced during the analysis in the laboratory, wrong calibration of the equipment or while entering the data. It is, however, also possible that there is a change in the system that causes changes in the water quality. The purpose of the validation procedure is thus twofold: it should act as a tool to provide a quality check and as a warning system to detect negative changes. The large amount of water quality data and its complex nature, however, make it difï¬cult for the environmental agencies to validate all incoming data. An ICT tool could be of great help to assist experts with the maintenance of monitoring databases compelled by the WFD. Such a tool should be able to deal with the complex nature of the water quality data and it also should be adaptive because the environmental system is likely to change, e.g. due to more stringent environmental legislation. Once the environmental agencies have a consistent database at their disposal, the data should be used to assess the evolution of the water status and to evaluate the impact of their management strategies. Such an assessment should be possible at the level of individual sampling locations as well as on a more regional scale. Many classical statistical techniques cannot be used for these purposes because data originating from environmental monitoring networks are clearly not independent. They are sampled from a dynamic process that evolves over space and time. Therefore, the methodology should incorporate this spatio-temporal dependence structure in order to provide valid statistical inference. Until recently, researchers mainly fo-cussed on the assessment of water quality at the level of the individual sampling locations. There have been some attempts in the past to provide techniques to perform an analysis on a spatial scale, but they used rather ad hoc methods to account for the spatial dependence. Only the last couple of years spatio-temporal models have been developed to take the specific spatio-temporal dependence structure of river networks explicitly into account (Gardner et al., 2003; Monestiez et al., 2005; Cressie et al., 2006; Ver Hoef et al., 2006). But they are all related to spatial prediction in river networks. Our aim, however, is to enable an assessment on the data that is observed at the sampling locations. Therefore the observations of the monitoring network at a certain time instant can be considered as the realisation of a finite-dimensional multivariate random variable with each dimension corresponding to each of the p sampling locations. Here, the spatio-temporal dependence also has to be taken into account to provide valid statistical inference. Both the data validation problem and the development of spatio-temporal models for river networks have become the major themes of this dissertation. Before we give the outline of this dissertation, we will introduce the data that were used throughout the work to test and illustrate the developed methodology.
Keyphrases
complex nature water quality data river monitoring network statistical validation spatio-temporal modelling water quality large amount river network warning system management strategy valid statistical inference negative change environmental data environmental agency jberg et al monitoring programme individual sampling location water status spatio-temporal model important activity van belle detection limit certain time instant quality assurance stringent environmental legislation specific spatio-temporal dependence structure european environmental manager natural variability data validation problem european union sampling location ad hoc method validation procedure consistent database water resource overall environmental objective irregular time interval driving force good status ground water different purpose adequate management essential tool developed methodology data quality monitoring database year spatio-temporal model spatial scale measuring method research community spatio-temporal dependence spatio-temporal dependence structure conceptual understanding dynamic process posse cyclic variation environmental policy interlinked activity environmental monitoring network big challenge high quality data ver hoef last couple water system spatial dependence monitoring network quality check finite-dimensional multivariate random wa ter authority ict tool spatial prediction underlying process wfd guidance document major theme anthropogenic activity incoming data many classical statistical technique measurement error considerable amount 15-year period european water framework directive environmental system comprehensive overview great help nonlinear trend wrong calibration regional scale
