Reconceptualizing Urban Land Cover: The HERCULES Model
|
(Much of this text is taken from: Cadenasso, M.L., S.T.A. Pickett, and K. Schwarz. 2007. Spatial heterogeneity in urban ecosystems: reconceptualizing land cover and a framework for classification. Frontiers in Ecology and the Environment. 5(2): 80-87.)
|
Summary
|
In order to investigate the link between ecological structure and function, researchers need to first accurately quantify structure. Therefore, ecologists depend heavily on classification systems for their research. Existing urban classification systems neglect important features of spatial heterogeneity in cities and suburbs. We have developed a new classification, High Ecological Resolution Classification for Urban Landscapes and Environmental Systems (HERCULES), which complements existing classifications, but works at spatial scales appropriate for urban landscapes and keeps system structure and function separate. This new tool balances detail and efficiency and is flexible, allowing it to be used for interdisciplinary research, with ancillary datasets, and across urban systems.
|
Urban Ecosystems and Spatial Heterogeneity
|
Figure 1. Urban heterogeneity. False color infrared air photo of Baltimore City taken in 1999 at submeter resolution.
|
There are two reasons for investigating heterogeneity in urban systems. First, urban land cover is clearly heterogeneous (Figure 1). For example, types and densities of buildings, presence and density of vegetation, types of infrastructure, and presence of remnant and natural areas can vary greatly over short distances. The integration of built and non-built (including biological and physical) structures yields complex urban landscape configurations. Therefore, the concept of heterogeneity provides a link between the emerging area of urban ecology and ecological theory in general.
Second, heterogeneity is a core ecological concept and plays a role in the functioning of systems. A goal of ecology is to describe features of the landscape that influence ecological processes. However, because there are many ways to characterize the heterogeneity of a system, there may be alternative descriptions of heterogeneity for the same area, each hypothesized to influence different functions. Consequently, there is no single description of ecological heterogeneity and individual research projects must articulate the scale and elements of heterogeneity within their study area.
|
A New Classification (HERCULES)
|
We present a new land-cover classification that accounts for the complex nature of urban systems and fills the under-represented space in the conceptual framework of urban land cover. We address this complexity by challenging the assumptions embodied in Anderson et al. (1976), rather than by simply adding greater detail within its classes. The new classification is known as High Ecological Resolution Classification for Urban Landscapes and Environmental Systems (HERCULES). This classification involves more than just a greater spatial resolution; it also refines the characterization of ecological features of the urban landscape. HERCULES focuses on the biophysical structure of urban environments and incorporates the recognized elements of urban land heterogeneity – buildings, surface materials, and vegetation. These three elements are divided into six features in HERCULES: (1) coarse-textured vegetation (trees and shrubs), (2) fine-textured vegetation (herbs and grasses), (3) bare soil, (4) pavement, (5) buildings, and (6) the building typology (Figure 2). The type of vegetation, surface material, and buildings are hypothesized to influence ecosystem function because of their differential influence on the amount and distribution of organisms, materials, and energy.
Figure 2. Hierarchy of urban landscape structure. The three basic elements of land cover can be divided into more detailed features.
|
Though surface water may be an additional element influencing ecosystem function, HERCULES addresses only the terrestrial component of the landscape. Patches in the landscape are classified by a human operator evaluating the shift in the proportional cover of each of the first five features listed above. The proportional cover is divided into four ranges: (0) absent, (1) present to 10% cover, (2) 11–35% cover, (3) 36–75% cover, and (4) > 75% cover. The sixth feature, building typology, has five recognized types: (1) single structures in rows or clusters, (2) connected structures that share a wall or are associated with multiple walkways while sharing the same roofline, (3) mixed, ie with multiple wings, connection by courtyards or arcades, or a group of buildings with different structural footprints, (4) highrises that are between 4–10 stories, and (5) towers, which are greater than 10 stories (Figure 3).
Table 1. Examples of patches classified using HERCULES. The proportional cover of coarse and fine vegetation, bare soil, pavement, and buildings is scored into 5 categories (0 = none, 1 = present - 10%, 2 = 11 - 35%, 3 = 36 - 75% and 4 = > 75%). Building types are identified as N = none, S = single, C = connected, or M = mixed.
|
Figure 3. Two of the most common building types identified by HERCULES in metropolitan Baltimore. The first is connected structures that share a wall. In the air photos this type is identified by multiple walkways entering the same roof line (panel A). A view of this building type from the street is in panel B. The second is single structures in rows or clusters as seen in the false color infrared air photos and at ground level (panels C and D).
|
Each of the six patch features can vary independently, and variation in any feature can define a new patch. For example, patches with the same proportion of coarse vegetation can differ based on the percentage cover of buildings. In addition, two patches containing the same building type and building density can be distinguished if one contains more coarse vegetation. Each patch is therefore classified based on its position in the six-dimensional space (Table 1). We have applied HERCULES using high-resolution digital aerial photographs. Patches are classified based on the uppermost layer of land cover; for example, bare soil under a tree canopy would not be recognized. Because the six features vary independently, percent cover data can be sorted and analyzed based on any single feature or combination of features. For example, if research focuses on the coarse vegetation in the landscape, the data can be sorted and analyzed to emphasize the variation in that feature, independent of heterogeneity in the other features. Alternatively, if research focuses on the density of buildings, the data can be resorted and analyzed to address this question.
|
Preliminary Analysis of Nitrate Yield using HERCULES
|
Figure 4a.
|
|
Figure 4b.
|
|
Figure 4c.
|
|
Figure 4d.
|
|
Figure 4e.
|
|
Figure 4f.
|
|
Figure 4. Comparing Anderson (Maryland Department of Planning 1999) and HERCULES for predicting nitrate yield (1999-2001; Groffman et al. 2004) from residential lands in the 17,150 ha Gwynns Falls watershed, metropolitan Baltimore, Maryland. A-C. Mainstem catchments. D-F. Addition of tributary catchment as an independent sample (shown in red). HERCULES is a better predictor of nitrate yield than Anderson in both analyses.
|
|
Evaluating links between system structure and function can be accomplished by combining data on landscape function with HERCULES for further analyses. In a preliminary test, we compared the utility of HERCULES to Anderson-derived land cover MRLC (Maryland Department of Planning 1999) for predicting nitrate yield, a pollutant of concern in the Chesapeake Bay. Gauged catchments along the main stem of the 17,150 ha Gwynns Falls watershed (n = 4) in metropolitan Baltimore, MD, were used (Figure 4a). Weirs maintained by USGS and sampled by the Baltimore Ecosystem Study (BES) are located at the junction of two catchments and at the bottom of the watershed. Based on these weirs, nitrate yield was calculated from weekly water samples and flow measurements for the years 1999–2001. Nitrogen yield was regressed against percent residential land from the Anderson derived classification (Figure 4b; r2 = 0.61, P =0.21). Figure 4c uses the same nitrate yield data, but regresses against the percent cover of patches containing single and connected structures in the HERCULES classification (r2 = 0.98, P = 0.01). Patches containing single and connected structures were used because they most closely resemble residential areas identified in the other classifications. The four catchments used may not be statistically independent, because they all represent the main stem of the Gwynns Falls stream. The addition of the Dead Run catchment, which is a tributary to Gywnns Falls, provides samples independent from those taken on the main stem (n = 5; Figure 4d). Nitrate yield regressed against percent residential land as identified by the Anderson-derived classification remained insignificant when the Dead Run tributary was added to the analysis (Figure 4e; r2 = 0.40, P = 0.25), while nitrate yield regressed against HERCULES, incorporating the independent sample of Dead Run, was highly significant (Figure 4f; r2 = 0.81, P = 0.03). Thus, HERCULES can predict nitrate yield in a set of gauged urban catchments. Still greater predictive power might be obtained by combining HERCULES with ancillary data such as soil type, topography, and residential land uses.
|
Future Directions
|
HERCULES has been applied to the entire Gwynns Falls Watershed using false color infrared aerial photos collected in 1999. HERCULES has also been applied to Baltimore City using false color aerial photos collected in 2004. We are in the process of applying HERCULES to the entire Gwynns Falls Watershed using the 2004 imagery. This additional dataset will allow us to observe changes in the landscape between 1999 and 2004. We will continue to add current and historic aerial photos to the dataset.
In addition, HERCULES has been applied to other research questions. HERCULES has been used to investigate bird diversity in urban forest patches (Carlson et al. 2006). HERCULES has also been used to predict nitrate yields in urban watersheds (Cadenasso et al. 2007) and is currently being used to investigate the spatial distribution of lead in urban soils (Schwarz et al. 2006).
|
Publications
|
Cadenasso, M.L., S.T.A. Pickett, and K. Schwarz. 2007. Spatial heterogeneity in urban ecosystems: reconceptualizing land cover and a framework for classification. Frontiers in Ecology and the Environment. 5(2): 80-87.
Carlson, C., M.L. Cadenasso, and G. Barrett. 2006. The relationship between breeding bird diversity in urban forest patches and the human-mediated resources located in the surrounding residential matrix. Ecological Society of America, 91st annual meeting abstracts. August 6-11. Memphis, Tennessee.
Schwarz, K., S.T.A. Pickett, M.L. Cadenasso, R.V. Pouyat, and I.D. Yesilonis. 2006. The Spatial Dynamics of Lead Levels in Urban Soil and Correlations with Land Cover in Baltimore, MD. BES Annual Meeting. October 18-19. Baltimore, Maryland.
|
