Current Research 2002-2008

 Current Initiative III - Forecasting Ecosystem Responses to Variation in the Socio-Natural Template

C-IIIA1. Conceptual and Technical Development of the Forecasting Framework.
Summary:  Forecasting is the formation of expectations about future states or processes of specific historical entities (Duncan 1969, Henschel 1976, Land and Schneider 1987). Forecasts can be used for their explanatory value in a scientific sense to the fulfillment of statutory policy requirements. Our objective is to isolate through our research the patterns that allow us to predict how the cycle of land-use change, ecological change, and human response will vary over space and time (Veldkamp and Lambin 2001).
Ecological systems have intrinsic temporal rhythms and patterns on characteristic spatial scales, but also bear the signature of human institutions that act directly or indirectly to alter the dominant spatial and temporal modes or introduce new ones (Pyne 1997, Carpenter and Gunderson 2001, Scheffer et al. 2001, Turner et al. 2002). The human institutions in turn are shaped and influenced by the environmental rhythms and ecological arrangements of the biogeographic region in which they emerged (Cronon 1983, Dove and Kammen 1997, Berkes and Folke 1998, Ostrom et al. 1999) leading to reciprocal imprinting of scales so that it becomes impossible to parse landscapes into “natural” and “human” components. Rather, they must be studied as integrated wholes (NRC 1999, Kinzig et al. 2000, Mitchener et al. 2001). Forecasting is the formation of expectations about future states or processes of specific historical entities (Duncan 1969, Henschel 1976, Land and Schneider 1987). Forecasts can be used for their explanatory value in a scientific sense to the fulfillment of statutory policy requirements. Our objective is to isolate through our research the patterns that allow us to predict how the cycle of land-use change, ecological change, and human response will vary over space and time (Veldkamp and Lambin 2001). We drawn on the common and comprehensive datasets we will build over the course of the research that includes: 1) characteristics and change over time in land use and land cover, 2) ecological processes associated with diversity, water quality, sedimentation, climate and organic flux, and 3) social processes associated with the formation of choice, development, agricultural practices, and political institutions. Landscape trajectories will be based on scenario analysis under different suites of assumptions (i.e., business as usual, major conservation efforts, planned development with some land-protection measures, massive population growth and economic development). The future is uncertain, but scenario planning can help anticipate a variety of contingencies (Bunn and Salo 1993, Schoemaker 1995). This work builds on our current research (Clark et al. 1998, Clark et al. 1999a, Clark et al. 1999b, LaDeau and Clark 2001, Pearson et al. in review) directed at understanding how changes in habitat abundance or quality affect population persistence by altering schedules of survival and reproduction. It also incorporates novel elements from our proposed decadal choice-based analysis (II.B.2) and disturbance regimes (II.B.4) in order to define a standard 29 discrete choice model with variables that incorporate land-use alternatives as well as characteristics of the unit of observation. We anticipate specifying land-use choice according to McFadden’s Discrete Choice model (Maddala 1983), and using logistic cumulative distribution functions based on relative rents, per capita income, and other social and economic factors. We will hold a workshop in year 4 in which all Coweeta investigators will participate in developing the final parameterization and design of the scenarios to forecast land-use probabilities for the Little Tennessee and French Broad drainages in 2010, 2020, and 2030. At this point we will also determine the means by which we will evaluate the relative importance of particular spatial predictions (Gardner et al. 1981, Costanza 1989, Turner et al. 1989, Caswell and Trevisan 1994). We will eventually develop a stochastic simulator for generating future land use maps to assess the relative impacts and uncertainty of distinct inputs, e.g., how commodity prices, transportation costs, or relative land values lead to particular land uses. Our stochastic simulator will also provide land-use forecasts by decade in order to evaluate the potential impacts of future land uses, e.g., how forest structure and sediment erosion suggest future policy alternatives. The previously noted pending collaborations with the UVT Gund Institute for Ecological Economics and the ASU Center for Environmental Studies will be complementary to our forecasting initiative activities. In all three cases, beyond the intrinsic interest our approach to forecasting may have in the scientific literature, our efforts are directed toward having an impact on the planning and management of real landscapes of the future (e.g., Hobbs 1997).

C-IIIA2. Partial Validation of Land-Use Forecasts.
There is no uniform procedure for validation, and ultimately no forecast can ever be thoroughly validated (Levins 1966, Greenberger et al. 1976). Validation is simply an indication of the level of confidence in a model’s behavior given its purpose, its desired performance, and the context for its use (Rykiel 1996, Ford 1999, Turner et al. 2001). Data does provide, however, a tangible link between a model and its reference system thus providing a central means for gaining confidence in the results of the model (Greenberger et al. 1976). The final test of a model’s usefulness is whether a forecast, for example, leads to implementation of policies that produce the results predicted by the model (Wear et al. 1998). The following three research activities address some of the issues of operational and conceptual validation (Sargents 1988, Rykiel 1996) that derive from the purpose, desired performance, and context of use for our forecasting.

C-IIIA3. Stream Hazard Site Project.
Summary:  We are documenting in the Stream Hazard Site Project the response-trajectory of streams to changing land-use patterns in multiple watersheds.

Aquatic ecosystems in the southern Blue Ridge tend to have low productivity due to low light, low temperature, high gradient, and low nutrient levels. Human activities are rapidly transforming the landscape and altering these important drivers of ecosystem structure and function. Our sampling design consists of collecting data in streams draining watersheds identified as “high risk” for development in the near future (Figure 13). We predict that disturbance will affect sites differently depending on the starting point of each site. Eight medium-sized watersheds (10-40 sq. km) were selected in 1998 in the Little Tennessee and French Broad drainages in three categories: (1) forested “reference”, (2) forested, but changing land use, and (3) agricultural, but changing land use. Land use in categories (2) and (3) is anticipated to move respectively toward second home development and suburban land use based on projections from 1993-2000 land use data. Site-selection was based on the work of Wear and Bolstad (1998), and the regression models of Gardiner (in prep). Abiotic and biotic parameters will be measured at each Hazard Site on a five-year interval over the next 30 years. 30 We completed our first sampling of algae, insects, and fishes in 2000 and propose resampling in 2005 with renewal funds. Coinciding with the 2005 stream sampling we will map land use, building location, road location and type, and impervious surface from geometrically corrected aerial photographs and/or high-resolution satellite imagery. The recently completed ordination of data on fish relative abundance collected in 2000 from the Hazard Sites and other Figure 14. Land use data were used to classify watersheds into four land use categories corresponding to the groups identified with biological and physical ecosystem assessments (Gardiner et al. in prep). sites (Scott 2001) suggest study sites form two distinct clusters related to land use in 1970 and in 1993. In effect, stream assemblages are starting from different points as a result of differences in land use in each watershed. Sampling over time at the Hazard Sites will allow us to determine if the trajectory of ecosystem response to disturbance will similarly differ, and as such help validate our integrated forecasts of ecosystem response to variation in the socio-natural template.

C-IIIA4. Effects of Development on Stream Ecology.
Summary:  Our objective is to determine how streams respond to contrasting rates of suburbanization.

Many parts of the southern Appalachian region are experiencing rapid population growth tied to suburbanization of formally agricultural lands and leading to negative impacts on stream structure (Sponseller et al. 2001) and function (Sponseller and Benfield 2001). Suburbanization of sub-basins in the Little Tennessee and French Broad watersheds is proceeding at varying rates (e.g., slow vs. fast) that we hypothesize are a function of factors such as elevation, road density, market access, land-use history, and others. In earlier work, we identified the end-members of a continuum of private land holdings in agricultural use distinguished by their suburbanization rate (Wear et al. 1998). We propose to distinguish among local effects of particular land-use practices in order to explore the cumulative effects of these practices on steam structure and function over the longitudinal axis of the study basins from headwater to mouth. We will attempt to replicate stream types at the level of stream order within sub-basins. An example stream type would be a 2nd order stream with a 500 m lateral buffer zone from mouth-to-headwater that was approximately 90% in agricultural use and had 10 houses in 1950, that by 2000 was approximately 40% in agricultural use and had 100 houses. While we realize there will be difficulties in identifying replicate streams, we are experienced in dealing with the problem. We are currently studying 30 streams (6 classes of 5 replicates each) recovering from agriculture that were selected using the method described below. Streams will represent a land-use continuum over space and time selected by interpreting chrono-sequenced (oldest available to most recently available in 10 y increments), high resolution aerial photographs or ortho-images at spatial scales from m to km. Using GIS, we will quantify land-use variables in the sub-basins into zones reaching laterally from the stream channels and longitudinally to the headwaters. Buffer zones will then be defined (Harding et al. 1998) by: % land cover type (forest, agricultural row crop/pasture/fallow, recreational, industrial, suburban, urban); miles of improved and unimproved roads; bridge crossings; building density; impervious surface area, and others. We will assess stream responses in land-use categories through structural (macroinvertebrate and fish biodiversity), functional (decomposition and stream metabolism) and geomorphic (cross-section and sedimentation) measurements obtained using standard methods (e.g., Benfield et al. 2000, Benfield et al. 2001, Sponseller and Benfield 2001, Sponseller et al. 2001, McTammany et al. in prep).

C-IIIA5. Ecosystem Valuation and Social Dynamics.
Summary:  Valuation is the measurement of the contributions of natural services to human objectives and well-being.

In the psychological literature, the determinants of well-being are most often linked to the satisfaction of basic human needs (Max-Neef 1992), but it should be more broadly linked to the satisfaction of particular sets of goals. Some goals are shared among all individuals (e.g., basic human needs) while others are specific to different ethnic, socioeconomic, and regional populations. We propose assessing the degree to which our research results and forecast scenarios contribute to the achievement of specific management goals. 32 We will use a procedure developed by Gregory and Wellman (2001) to determine perceptions and relative importance of management options, acceptable trade-offs in management, and implied valuations of ecosystem services. The procedure relies on a variety of value-structuring tools. After determining the fundamental objectives of stakeholders, meansends analyses will be used to specify and evaluate alternatives, e.g., limiting livestock access to streams relative to upgrading forest management roads. Participants assess trade-offs using a workbook to (a) rank management options irrespective of formally presented information on their costs and benefits, and (b) detail choice tasks (i.e., “plans”) for each management option individually. Costs are varied for each individual until they switch their preference for given plans, thus providing information on acceptable trade-offs for a given objective. The notable feature of the Gregory and Wellman (2001) procedure is that it does not rely on individual valuations as might be derived from willingness-to-pay measures, but on the public costs implied by federal or state activities. The individual is thereby induced to consider the valuation as a public citizen (Sagoff 1998). The basis for the procedure is multi-attribute utility theory (Keeney and Raiffa 1993), which is a variation of conjoint analysis (Farber and Griner 2000). Understanding valuation and the links between social dynamic and value formation will provide independent tests of key assumptions behind the choice-based modeling framework. In addition, this study will provide insights into the potential long run structural change in individuals’ perspectives on and uses of land and natural resources in southern Appalachia. This understanding is critical to building our scenarios and validating our forecasts.