ASA-ENVR
Session Slot: 8:30-10:20 Monday
Estimated Audience Size: 50-60
AudioVisual Request: overhead & slide projector only
Session Title: Design of Environmental Monitoring Networks
Theme Session: Yes
Applied Session: Yes
Session Organizer: Royle, J. Andy National Center for Atmospheric Research
Address: Geophysical Statistics Project, NCAR, Boulder, CO 80307
Phone: 303-497-1704
Fax: 303-497-1333
Email: rolye@ncar.edu
Session Timing: 110 minutes total (Sorry about format):
Opening Remarks by Chair - 0 minutes First Speaker - 35 minutes Second Speaker - 35 minutes Third Speaker - 35 minutes Floor Discusion - 5 minutes
Session Chair: Royle, J. Andy National Center for Atmospheric Research
Address: Geophysical Statistics Project, NCAR, Boulder, CO 80307
Phone: 303-497-1704
Fax: 303-497-1333
Email: royle@ncar.edu
1. Shared Network Redesign Decision Making
Haas, Timothy C., University of Wisconsin-Milwaukee
Address: School of Business Administration, University of Wisconsin-Milwaukee, P.O. Box 742, Milwaukee, WI 53201
Phone: 414-229-4360
Fax: 414-229-6957
Email: haas@uwm.edu
Abstract: Generally, if the public feels they have a say in determining how environmental monitoring will be conducted, they will be more willing to support long-term monitoring expenditures. Hence, publicly-accessible systems are needed that allow anyone to participate in monitoring network design and redesign decision making. This talk will describe how such a system can be implemented as a World-Wide-Web site.The proposed system will allow a web site user to perform the same site addition and site deletion optimization analyses that are used by governmental decision makers. The focus of this talk will be on the optimal removal of sites from a monitoring network. It will be argued that (a) network optimality is inherently multidimensional and hence requires multiple objectives to be satisfied, and (b) the public needs to be involved in the determination of the these objectives and their priority weights. It is also argued that there are three core objectives of any monitoring network. First and second, the network should deliver pollutant predictions that are as accurate and precise as possible. Third, the network should provide these predictions as inexpensively as possible.
For a network monitoring a large random field, stationarity is typically not present. Also, the problem of identifying the best subset of sites for deletion is combinatorially explosive. For these two reasons, Simulated Annealing is used to solve the nonlinear, integer, multi-objective mathematical programming problem using the author's Moving Cylinder Kriging spatio-temporal prediction algorithm to compute average prediction differences between the full and reduced network (network accuracy), and average increase in prediction relative error between the full and reduced network (network precision).
An example using NADP/NTN sulfate deposition across the conterminous U.S will be given.
2. Monitoring Network Design and Air Quality Standards
Sampson, Paul D., University of Washington
Address: Department of Statistics, Box 354322, University of Washington, Seattle, WA 98195-4322
Phone: 206-685-2664
Fax: 206-685-7419
Email: pds@stat.washington.edu
Guttorp, Peter, University of Washington
Thompson, Mary Lou, University of Washington
Cox, Larry, U.S. Environmental Protection Agency
Abstract: A number of issues are generally taken into account in designing an air quality monitoring network. Local siting criteria for individual monitoring sites are well-discussed in the literature, but formal statistical design criteria seem not to be much used. We discuss issues of spatial and temporal sampling and averaging in the definition of air quality standards, the consequent specification of violators of these standards, and the probability of detecting violators from the perspective of monitoring network design for spatio-temporal air quality processes.
3. Space-time Models and Dynamic Design
Wikle, Christopher K., National Center for Atmospheric Research
Address: NCAR, Boulder, CO 80307
Phone: 303-497-1722
Fax: 303-497-1333
Email: wikle@ucar.edu
Royle, J. Andy, National Center for Atmospheric Research
Abstract: An important area of spatial statistics is the design of spatial sampling plans for monitoring air pollution and other environmental processes. Many methods for constructing optimal spatial designs are widely applied in a large number of disciplines. These methods generally produce static designs which are optimal under models with no explicit temporal structure. However, most environmental processes often tend to exhibit both spatial and temporal variability, and hence static networks may not capture the essential space-time variability of the process. Although static designs are often necessary due to geopolitical and economic considerations, one could imagine the possibility of supplementing a static network with one or more mobile monitoring devices. Then, the design problem is where should the mobile monitor or monitors be located at time t+1 in order to provide the most information based on observations up to time t. We consider several approaches to this problem, including a simplistic approach which moves the mobile monitor to the location which is optimal at time t. We then propose a dynamical space-time model that allows prediction of future spatial correlation structure, and use this correlation structure to optimize over all possible monitoring sites. These approaches are demonstrated with a surface ozone data set.
List of speakers who are nonmembers: 0