An important aspect of site selection and the up- and down-scaling approach to integration of models, observations, and process studies is the estimation of representativeness. The multivariate data-mining methodologies described above for landscape characterization offer useful metrics for indicating the representativeness of sites, measurements, and model parameters. Hargrove et al. (2003) described a technique for understanding the representativeness of a sampling network based on a suite of environmental gradients considered to be useful proxies for the characteristics being measured. Maps indicating poorly represented regions can be produced, suggesting where new measurements should be taken. While Hargrove et al. (2003) calculated representativeness in the context of customized ecoregions, this same approach can be applied to every map cell projected onto the hypervolume of environmental gradients (Hutchinson 1957) used to perform the cluster analysis that produced those ecoregions (Hargrove and Hoffman 2004), providing a continuously varying metric describing the representativeness of every location with respect to one or more than one sampling location.
F. Hoffman (ORNL) will lead this team.
As illustrated for the example landscape-scale regionalization in the figure, representative sampling sites can be determined for each region or sampling domain. These sites are the realized map locations that
|
|
most closely correspond to the idealized centroids of the clusters composed of member map cells in the 37-dimensional data space formed by the environmental characteristics. Similarly, the representativeness of any location selected a priori can be calculated for a sampling domain or across a larger region. Figure 7 contains a map depicting the representativeness of a location in Barrow for the state of Alaska. This unitless, relative representativeness metric is the Euclidean distance between every cell in the map and the Barrow location within the 37-dimensional data space. Therefore, low values imply high similarity and high values imply high dissimilarity between any location and Barrow. In this map, white to light gray land areas are well represented by the Barrow location; dark gray to black land areas are poorly represented by Barrow. If a field researcher were attempting to select one additional sampling location in order to provide optimal coverage of the environments within the state of Alaska, that next site should be chosen within the darkest land areas shown in the map. Once a new candidate site has been selected, a new map of representativeness can be generated with simultaneous consideration of both sites. Using this relative representativeness metric, optimal sampling locations can be chosen to maximize the coverage of environmental conditions for any domain at any scale for which sufficient data are available.
Because of a lack of data availability at a pan-Arctic scale, additional relationships between landscape characteristics and broadly observable features will likely need to be developed in order to assess the representativeness of the study area and to scale model parameters to the larger Arctic region. Moreover, statistical relationships between biogeomorphic units at different scales may be required to bridge observations across scales. This effort must be performed in concert with the data collection teams and the modeling teams.
Perform representativeness analysis within the Barrow domain to determine sampling locations and fine-scale measurement representativeness. In support of fine-scale sampling and modeling, the biogeomorphic unit characteristics will be used to generate maps of representativeness for candidate and chosen sampling locations within the Barrow Environmental Observatory area and with respect to the larger Barrow region. This same framework will be applied to understand the representativeness of measurements and model results in support of up-scaling the observations and model parameters to the larger Barrow region. This task depends upon acquisition of field measurements and the landscape characterization tasks defining and mapping the biogeomorphic unit characteristics.
Define the relationships for up-scaling for fine-to-intermediate and intermediate-to-landscape or climate scales. Representativeness analysis will be used to determine the most important environmental gradients at the fine, intermediate, and landscape/climate scales. These characteristics are likely to be the most significant in controlling processes at each of the scales. This analysis will use the biogeomorphic unit characteristics to inform the creation of the multiscale grids used in the fine-, intermediate-, and climate-scale models. In addition, this analysis will test a metric multidimensional scaling methodology for interpolating and extrapolating model parameters. This task depends upon the landscape characterization tasks defining and mapping the biogeomorphic unit characteristics and links to the fine- and intermediate-scale modeling tasks developing multiscale grids and interpolating and extrapolating model parameters.
Perform representativeness analysis for NGEE Arctic Phase 2 sampling and site selection. Biogeomorphic unit characteristics acquired and computed at the landscape scale will be applied to perform representativeness analyses to determine optimal sampling and manipulation site locations. Maps will be produced showing the sampling network coverage offered by various candidate sites. This task depends upon the landscape characterization tasks defining and mapping the biogeomorphic unit characteristics.
Develop pan-Arctic representativeness analysis. The development of biogeomorphic unit characteristics, scaling methodologies, and representativeness analyses will culminate in an initial pan-Arctic characterization that can be applied to understand important processes and vulnerabilities of current Arctic ecosystems. A pan-Arctic representativeness analysis will be performed to locate under-represented biogeomorphic regions that may be critical to incorporate into future measurements or modeling activities. This same framework will be applied to scale existing measurements and to model results to the pan-Arctic domain. This task depends upon the landscape characterization tasks defining and mapping the biogeomorphic unit characteristics and the modeling tasks producing fine-, intermediate‑ , and climate-scale results.
Scaling Framework Development Deliverables
- Develop and apply network analysis and representativeness methodologies suitable for selection of sampling locations, scaling of measurements, and integration of model parameters across the fine, intermediate, and climate model scales.
- Develop maps of biogeomorphic characteristics and integrated field data for the Barrow region.
- Develop spatially distributed collections of model driver data, parameters, and model evaluation data for the fine, intermediate, and climate model scales.
- Publish results from investigations of scaling approaches, biogeomorphic characteristics data, and model driver and evaluation data.
Empty