Environ. Sci. Technol. 2010, 44, 329–334
Development of Resource Shed Delineation in Aquatic Ecosystems D A V I D F . R A I K O W , * ,† JOSEPH F. ATKINSON,‡ AND THOMAS E. CROLEY II§ U.S. Environmental Protection Agency, National Exposure Research Laboratory, Ecological Exposure Research Division, 26 West Martin Luther King Dr., Cincinnati, Ohio 45243; Great Lakes Program, University at Buffalo, 202 Jarvis Hall, Buffalo, New York 14260; and National Oceanic and Atmospheric Administration, Great Lakes Environmental Research Laboratory, 2205 Commonwealth Blvd., Ann Arbor, Michigan 48105
Received February 20, 2009. Revised manuscript received October 26, 2009. Accepted November 5, 2009.
We apply a concept derived from food web ecology to largescale spatial patterns of material supply within and between watersheds and coasts by generalizing the definition “resource shed” to source areas for materials supplied to a receptor (e.g., a point location) over a specified time interval. Independent hydrologic and hydrodynamic models, coupled with a particle tracking model, were used to delimit resource shed total spatial extent and relative contributory importance for selected receptors in Lake Erie (North America) over varying time intervals. One resource shed was extended into the Maumee River watershed (OH) by integrating the lake and hydrologic models. Model validation was achieved through comparison with data from the 2005 International Field Years on Lake Erie (IFYLE) study. Resource shed size, orientation, and internal structure varied with receptor location, in-lake circulation, terrestrial precipitation, time interval, and season. River plume extent and interaction were illustrated, and model integration revealed the relative contributory importance of subwatershed catchments to an off-shore receptor.
Introduction Environmental issues in aquatic ecosystems often involve spatially explicit phenomena that occur over large geographic scales. In the Laurentian Great Lakes (North America), for example, transport of sewage overflows and nonpoint source nutrients from watersheds (1, 2) and within the Great Lakes (3) promote microbial contamination that cause beach closings at distant locations (4). Such connectivity between watersheds and coasts necessitates an understanding of how physical forcing affects the movement of materials at the landscape level. Indeed, accounting for material flow across the landscape between otherwise intuitively distinct ecosystems is widely recognized as necessary for the understanding of ecological processes within ecosystems (5). Hence conditions measured in continuously flowing systems, as * Corresponding author phone: 513-569-7383; E-mail:
[email protected]. † U.S. Environmental Protection Agency, National Exposure Research Laboratory. ‡ Great Lakes Program, University at Buffalo. § National Oceanic and Atmospheric Administration, Great Lakes Environmental Research Laboratory. 10.1021/es900562t
2010 American Chemical Society
Published on Web 12/03/2009
with sampling lake water, are in part a function of processes acting while materials were traveling to the site. Analyses of ecological phenomena such as harmful algal bloom formation and transport account for such landscape-level influences by using large-scale circulation models (6). Another useful way to study the flow of material arriving at sites, or receptors, is to delineate source areas. Sophistication of source area delineation methods, however, varies greatly between disciplines. For example, methods to delineate source areas within airsheds called “areas of influence” are well-developed in atmospheric chemistry (7). Methods for the reconstruction of the spatial and temporal history of groundwater contaminant plumes are also welldeveloped and form the basis of subsurface environmental forensics (8). Study of subsurface flow processes within variable source hydrology has led to models allowing prediction of spatial sources of stormwater runoff within terrestrial watersheds at scales of meters to several kilometers (9). Source area-delineation in surface waters is less well developed. Using particle back-tracking simulations, or “backwards-in-time trajectory” calculations, Batchelder (10) estimated potential source area configuration for fish larvae arriving at receptors near idealized Pacific coastlines. Kasai and Komatsu (11) estimated and mapped source areas in the East China Sea for fish larvae arriving at receptors in the Sea of Japan. In the most comprehensive source-area analysis of open-water systems, Roberts (12) evaluated coral larvae drift in the Caribbean Sea, where source areas called “transport envelopes” were evaluated over 1- and 2-month intervals for several receptors, but did not include estimates of relative importance within the total spatial extent. Here we approach this concept from a food-web ecology perspective, and adapt “resource sheds”, originally defined as “source areas for resources consumed by individuals over their lifetimes” (13), to represent source areas from which materials (including but not limited to suspended nutrients, particulate organic matter, sediments, propagules, prey, or pollutants) are derived for a receptor (an individual, population, or point location) over a specified time interval [see ref (14) for mathematical definitions]. We can define resource sheds further as source locations for all materials originating (departing) during one time interval and arriving at a receptor: (a) during another time interval (type 1); (b) at the end of the time interval (type 2); or (c) during the same time interval (type 3) (14). A time interval must be specified to account for material movement (14), and is further guided by practical limitations on what is biologically relevant within material cycles that are global if considered with unlimited time (13). Moreover, natural disjunctions in material cycles serve to confine spatiotemporal consideration of resource sheds. For example, precipitation patterns are a logical starting point for resource sheds of water-borne materials in watersheds, because prior supply of materials to the watershed occurs through atmospheric deposition affected by other physical forcing variables (air movement). Importantly, if time intervals are extended back far enough, the resource shed for an off-shore receptor will eventually intersect with the coast and include tributaries. Hence resource sheds for water-borne resources in coastal and lake ecosystems are functions of both watershed inputs and circulation patterns. Uncommon in ecological literature, and subordinate to discussions of related ecological processes such as ecological subsidy (15), ecological studies explicitly using the term “resource shed” have been dominated by theoretical evaluVOL. 44, NO. 1, 2010 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
329
FIGURE 1. Type 3 resource sheds for receptors in western Lake Erie (North America) based on 20-year average meteorology. Time periods for resource sheds are 31 August (1-day sheds), 25 August to 31 August (1-week sheds), 1 August to 31 August (1-month August sheds), and 1 May to 31 May (1-month May sheds). Seasonal comparisons are 1-month sheds for May and August. (A) 1-day August resource sheds with density. (B) 1-week August resource sheds with density. The southern extent of the resource shed for site 881 is shown as a black line south of the site. The southern extent of the resource shed for site EPA60 is shown as a black line south of the site. (C) Site 881, 1-month August resource shed with density. (D) Site EPA60, 1-month August resource shed with density. (E) Site WT2, 1-month August resource shed with density. (F) Site EPA59, 1-month August resource shed with density. (G) Site WT2, 1-month August resource shed with density. (H) Seasonal comparison of resource sheds for site 881. (I) Seasonal comparison of resource sheds for site EPA60. (J) Seasonal comparison of resource sheds for site WT2. (K) Seasonal comparison of resource sheds for site EPA59. (L) Seasonal comparison of resource sheds for site WT1. ations. For example, Finlay et al. (16) found that as benthic stream insects consumed phytoplankton, carbon sheds for scrapers consisted of the surrounding riffle microhabitat, while carbon sheds for filterers consisted of unspecified upstream pools. Indeed, the original study defining resource sheds only schematically diagramed theoretical off-shore and in-watershed resource sheds (13). To advance source-area delineation for off-shore receptors, we apply new methods to delineate resource sheds. We evaluated spatially explicit patterns of material supply by modifying and integrating a hydrodynamic lake-transport model and a hydrologic runoff model (14). We present the first examination of general off-shore resource shed behavior, first comparisons of resource sheds under variable time periods, seasons, and years, first determinations of the relative contributory importance of areas within the total spatial extent, and first integration of off-shore and terrestrial watershed resource sheds.
Materials and Methods Resource sheds were modeled for 7 receptors coinciding with International Field Years on Lake Erie (IFYLE) (17) sample sites in the western basin of Lake Erie, a Laurentian Great Lake (North America) (Figure 1A). The western basin of Lake Erie is dominated by inputs from the Detroit River, actually 330
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 44, NO. 1, 2010
a strait through which the upper Great Lakes (Lakes Superior, Huron, and Michigan) drain, to the north, and the Maumee River to the southwest. Study placement minimized potential influence of Sandusky River outflow to the east, thus simplifying interpretation of results. Type 3 resource sheds were modeled for material originating in its source area and intersecting a receptor over three time intervals (1-day, 1-week, and 1-month) chosen for several reasons: (1) to examine changes in resource shed extent over increasing durations, (2) to allow for reasonable computation time, and (3) to explore resource sheds over time periods relevant to at least some in-lake ecological processes such as planktonic turnover. Baseline circulation patterns were developed using particle tracks based on a 20-year (1980-1999) averaged meteorological data set. Resource sheds were also calculated for the two periods immediately preceding the IFYLE fixedstation cruises in May and late July to early August 2005. To validate the model, general resource shed behavior was examined for potential hypotheses to be tested using spatial patterns of epilimnetic nutrients. Field data consisted of measurements taken during the May and August IFYLE fixed-station cruises, and mapped using inverse distance weighting in ArcMap 9.3 (ESRI Inc., Mountain View, CA). Of the parameters measured in IFYLE, nitrate (NO3) showed the greatest disparity in concentrations between Detroit River
and Maumee River inputs during May and August 2005, thus differentiating river inputs. Nitrate leaches from agricultural watersheds into streams, and this nutrient load is ecologically important in receiving water bodies (18). Nitrate in the western basin of Lake Erie is controlled by river inputs, and large reductions in concentrations typically occur after water has exited the western basin and entered the central basin (unpublished IFYLE data). In the lake, a combined hydrodynamic and particle tracking model (PTM) was used to generate particle paths as an integrator of velocity fields and to delineate resource sheds (6). The Princeton Ocean Model [POM (19, 20)] was used to calculate circulation patterns in response to imposed meteorological conditions. Bathymetric data were available on a 2-km grid (21). Long-term meteorological data were found from the Buffalo, Erie, Cleveland, Toledo, and Rondeau weather stations, with eight additional Great Lakes Coastal Forecast System stations available for the 2005 calculations. These data were interpolated to provide input for each POM computational grid. Inflows for the Detroit and Maumee Rivers were set at 5000 and 150 m3 per second, respectively. The velocity and diffusivity fields produced by the POM, also calculated on a two-km square grid, were interpolated to generate values for the PTM, where “particles” were considered neutrally buoyant, nonreactive points that traveled with the velocity and diffusivity fields. Particle tracks were calculated as random walks, with the random component incorporating the turbulent diffusive motion within the mean advective flow field (6). Although the model is capable of incorporating vertical movement, for the present discussion only horizontal transport was considered. The POM is a sigma coordinate model, and velocities and diffusivities were taken from the surface layer to drive the PTM. The study area is relatively shallow (mean depth
