Caeiro

Optimization of Sediment Estuarine Monitoring Program Using a Contamination Index

Sandra Caeiro (1), Luis Nunes (2), Pierre Goovaerts (3), Marco Painho (4),
Helena Costa (5)

(1) IMAR Department of Exact and Technological Sciences of the Portuguese Learning Distance University, Lisbon (PT)
(2) CVRM, Faculty of Marine and Environmental Sciences, University of Algarve (PT)
(3) Biomedware, Ann Arbor (US)
(4) ISEGI/CEGI, Institute for Statistics and Information Management of the New University of Lisbon (PT)
5. IMAR, Faculty of Science and Technology of the New University of Lisbon (PT)

Estuarine areas are usually highly populated and industrialized, which result in important pressures on the environment. These pressures may inflict severe negative environmental impacts if not carefully managed. It is then necessary to manage them in an integrated perspective, considering among other things the impact of activities discharging effluents into the estuary. Environmental management of these ecosystems cannot be conducted effectively without reliable information on changes in the environment and on the causes of those changes. Ecological monitoring programs represent an important source of that information. Monitoring should be planned in order to provide quantitative assessments of pollutants' complex effects. Sampling designs that provide statistically unbiased estimates of the status, trends, and relationships, with quantitative confidence limits on spatial scale are then crucial. Variance reduction techniques fulfill these criteria and have proved to be good methodologies for choosing the best spatial design.

The team has been working on the development of an environmental management system integrated in a Geographical Information System for the Sado Estuary in the south of Portugal. In this estuary the human activities must be managed and balanced with conservation practices to assure safety of the existing protected natural areas. This management system is developed based on spatially contiguous areas of sediment structure. Nineteen homogeneous areas were delineated using a multivariate geostatistical analysis (cluster analysis of dissimilarity matrix function of geographical separation followed by indicator kriging of the cluster data) of an extensive stratified random sampling campaign of estuarine sediments (153 stations) (1). At each location, three sediment parameters (Organic matter, Fine fraction and Redox potential) were measured. In this article, the focus is on the design of the sediment monitoring network. To avoid information redundancy as well as due to budget constraints, a smaller subset of the most representative stations inside each homogeneous area was selected for contaminant assessment of the estuary. These subsets were selected by an iterative approach whereby one station is removed at a time and values of sediment parameters are estimated by indicator kriging using the remaining stations in the sub-set. For a given sample size the optimal subset is the one that minimizes the mean square error of estimation and it was identified using simulated annealing, within a controlled non-exhaustive looping scheme. The model results indicated a 60 station design to be optimal. Beyond this number, each new added station had little effect on the monitoring spatial accuracy. However 17 additional stations were added based on expertise criteria of proximity to point sources and characterization of all homogenous areas (2). These 77 stations still represent a high cost and it is time consuming for any future long-term monitoring program.

The aim of this work is to select a new subset of monitoring sampling stations, representative of each homogenous area, based on the 77 locations and information on the level of contamination (evaluated through concentrations in 7 heavy metals: Cd, Cr, Cu, Hg, Pb, As and Zn). The heavy metals data were summarized in an index of contamination (mean SQG-quotient) and then integrated in the optimization model. This index works as a central tendency indicator of adverse biological effects due to heavy metal concentrations. Mean SQG-quotient was calculated dividing first each heavy metal by the Probable Effect Level (PEL) and then summing this quotient (PEL-Q) for individual substances and dividing by the number of PEL-Qs that were calculated (3). Simulated annealing was used again to identify the set of optimal stations. Different sub-set cardinalities were tested in order to determine the optimal cost-benefit relationship between the number and location of stations and monitoring costs. Results indicated 30 stations as the optimal number to be used in the future long term monitoring program.

References
1. Caeiro, S., P. Goovaerts, M. Painho, and M. H. Costa. Delineation of Estuarine management areas using multivariate geostatistics: the case of Sado estuary. Journal of Environmental Science and Technology. Accepted.
2. Caeiro, S., L. Nunes, P. Goovaerts, H. Costa, M. C. Cunha, M. Painho, and L. Ribeiro. Optimisation of an estuarine monitoring program: selecting the best spatial distribution. Ruiz, M. Gould, M. Ramon, J. (ed.). GeoENV IV Geostatistical for Environmental Applications. Kluwer Academic Press, Dordrecht . in press.
3. MacDonald, D. D., R. A. Lindskoog, D. E. G. H. Smorong, R. Pribble, T. Janicki, S. Janicki, S. Grabe, G. Sloane, C. G. Ingersoll, S. Eckenrod, and E. R. Long. Development of an Ecosystem-Based Framework for Assessing and Managing Sediment Quality Conditions in Tampa Bay, Florida. Tampa Bay Estuary Program. Florida 2000.