Convergent Surface Water Distributions in U.S. Cities
Earth’s surface is rapidly urbanizing, resulting in dramatic changes in the abundance, distribution and character of surface water features in urban landscapes. However, the scope and consequences of surface water redistribution at broad spatial scales are not well understood. We hypothesized that urbanization would lead to convergent surface water abundance and distribution: in other words, cities will gain or lose water such that they become more similar to each other than are their surrounding natural landscapes. Using a database of more than 1 million water bodies and 1 million km of streams, we compared the surface water of 100 US cities with their surrounding undeveloped land. We evaluated differences in areal (A WB) and numeric densities (N WB) of water bodies (lakes, wetlands, and so on), the morphological characteristics of water bodies (size), and the density (D C) of surface flow channels (that is, streams and rivers). The variance of urban A WB, N WB, and D C across the 100 MSAs decreased, by 89, 25, and 71%, respectively, compared to undeveloped land. These data show that many cities are surface water poor relative to undeveloped land; however, in drier landscapes urbanization increases the occurrence of surface water. This convergence pattern strengthened with development intensity, such that high intensity urban development had an areal water body density 98% less than undeveloped lands. Urbanization appears to drive the convergence of hydrological features across the US, such that surface water distributions of cities are more similar to each other than to their surrounding landscapes.
Key words:surface water urbanization cities convergence urban streams urban water bodies hydrography
Surface water plays contrasting roles in cities, both supporting and hindering development of urban landscapes. Streams, rivers, lakes, ponds, and wetlands within urban environments can be important components of water supply and flood control systems; and they provide additional ecosystem services including recreational and aesthetic value (Bolund and Hunhammar 1999), species habitat (Hamer and Parris 2011; McKinney and others 2011), microclimate moderation (Sun and others 2012; Sun and Chen 2012), and removal of pollutants (Grimm and others 2008b). However, when urban land is scarce relative to demand for development, the space occupied by surface water is valuable and often subject to alteration to promote development (Du and others 2010). In addition, surface water can create flooding risks and other hazards to surrounding land, structures, and populations. To increase the area of urban land and control water supply and drainage, humans drain, fill, and bury surface water features. This practice has resulted in substantial loss of stream channels in cities in mesic regions (Elmore and Kaushal 2008; Roy and others 2009; Pataki and others 2011). Alternatively, construction of lakes, ponds, and canals can also increase surface water abundance, particularly of cities in arid regions (Larson and Grimm 2012). However, little attention has been paid to hydrographic change at broader spatial scales, where distinct patterns and processes may emerge (Heffernan and others 2014; Thorp 2014). A broader assessment of how urbanization changes the channel and water body abundance, distribution, and form (hydrography) in cities is essential to understanding both the environmental constraints on land-use change (Dunne and Leopold 1978) and the environmental and societal consequences of rapid and ongoing urbanization of human populations and landscapes (Paul and Meyer 2001; Cohen 2003; Foley and others 2005; Kareiva 2007; Grimm and others 2008a).
We contend that the value of surface water features in urban environments, relative to dry land, increases as a function of their scarcity. Accordingly, we hypothesize that this relationship drives patterns in the alteration of urban surface water distributions. If true, the difference in surface water coverage between urban and surrounding undeveloped land should be positive (urban > undeveloped), where surface water is rare, but negative (urban < undeveloped) where it is abundant. As a result, urbanization at continental scales should lead to lower variation in surface water abundance across cities relative to variation among their surrounding undeveloped hydroscapes. This pattern would show that urban surface water converges on a specific form, as observed in the constructed components of the urban environment (Seto and Fragkias 2005). Moreover, these patterns should be strongest in intensely developed urban areas, where land is most highly altered.
The objective of this study was to compare the characteristics of surface water (hydrography) of urban land cover to that of surrounding undeveloped land. We investigated urban hydrography using a database of more than 1 million water bodies (that is, lakes, ponds, wetland, marshes, reservoirs) and 1 million km of stream channels from 100 Metropolitan Statistical Areas (MSAs). Hydrographic measures of % water body coverage (A WB), water body density (N WB), and channel density (D C) for each land cover type were calculated using surface water features from the National Hydrography Dataset. These hydrographic characteristics were compared to the surrounding land cover in each MSA.
We selected 100 cities with a defined MSA from the continental United States. The United States Office of Management and Budget defines an MSA as an urban core with population of at least 50,000 and associated counties with a high degree of social and economic integration (as measured by commuting to work) with the urban core (Census Bureau 2012). To ensure a representative sample of continental US cities, cities were categorized by their designated ecological regions as established by the National Ecological Observatory Network (NEON 2010). The number of MSAs selected from each ecological region was weighted based on the proportion of MSAs present. Cities were also categorized by population size into five groups. Within each eco-region, cities were randomly selected to generate roughly equal representation from each population group, so the MSAs were dispersed through the eco-region. Six cities were specifically included as part of a broader study of urban homogenization (Groffman and others 2014). Otherwise, cities were chosen “blind” to avoid biases (that is, the selector did not know the name of the MSA/city during selection). This method was chosen over a completely randomized selection process because cities with small populations in the US are more abundant than larger cities. Our selection process provided a more balanced distribution across population gradients.
We categorized land cover in each MSA using the 2006 National Land Cover Data (NLCD). The NLCD classes were grouped into five categories: urban open area (NLCD = 21), urban low intensity (NLCD = 22), urban medium intensity (NLCD = 23), urban high intensity (NLCD = 24), agriculture (NLCD = 81, 82), and “undeveloped” (all remaining NLCD). Using ArcGIS (v10) we calculated the majority land cover of every census block group (CBG) in the MSA. In the few instances where water (a land-cover category in the NLCD) was calculated as the majority land cover in a CBG, the non-water land cover was assigned by hand based on the classification of the surrounding CBGs. The water coverage from NLCD was not used in any further analysis.
Surface water data layers were obtained from the National Hydrography Dataset (NHD). Features included inland water bodies (that is, lakes/ponds, reservoirs, swamps/marshes), those in coastal and estuarine systems, and surface flow lines. The maximum resolution of the NHD high-resolution data is 1:5,000, but varies by state. Although there is no known estimate of the percentage of missing water bodies from this dataset, we recognized that the number of missing stream channels may be as high as 78% at this scale (Roy and others 2009). The length of streams converted to storm drains or buried underground are likely more dense in urbanized areas and no longer appear in this data as streams; however, there is no evidence that the errors of omission for streams are greater in urban areas. Despite its limitations, NHD remains the most comprehensive coverage at the national scale. NHD uses feature codes (FCodes) to categorize water bodies in to different feature domains. FCodes in this study were grouped into several different classes: perennial lakes and ponds, intermittent lakes and ponds, swamps and marshes, water storage reservoirs, other reservoirs, perennial streams and rivers, intermittent and ephemeral streams and rivers, and canals/ditches.
The hydrography data layers were intersected with the CBG land-cover layer to categorize each feature by its surrounding land cover. For the undeveloped land we used the NHD feature point data to locate all of the water bodies with dams in the undeveloped land class. The NHD data contains the location of dams in the United States, primarily based on the National Dam Inventory; however, only larger dams are included in this study (see Army Corps of Engineers (2013) for full description of criteria). Dammed water bodies were identified by the points that intersected the water bodies with a 10 m buffer to account for minor mapping variations. Intermittent water bodies were also removed so that only perennial features were included. Otherwise, for each land-cover category, we summed the number and area of water bodies and then divided by the total land area in each category to calculate the density of water bodies and the percentage of area covered by water. We summed the channel lengths and divided by land area to calculate channel density.
Water Body Size Distributions
To assess the convergence of water body characteristics, we analyzed the effect of land cover on water body size distributions (SD WB). We calculated the median water body size in each land-cover type for all 100 MSAs. However, because the frequency distribution of water body sizes is poorly described by means or medians, we also described and compared these distributions based on Pareto distribution parameters. The shape of the Pareto distribution can be described by its slope (β Size), which becomes increasingly negative as the number of small water bodies increases (Downing and others 2006). We calculated β Size as the linear slope of the log–log transformed frequency distribution for all n > size class (x).
Data analysis was conducted on the aggregate of the census block groups in each land category for every MSA. For example, to calculate percent water coverage we summed the area of all water bodies classified as “urban” in the MSA, and divided that sum by the total urban land area in the MSA. We compared the % water body coverage (A WB), water body density (N WB), and channel density (D C) of land-cover classes with a univariate analysis of variance and Tukey’s post-hoc mean comparison. The A WB, N WB, and D C data were log-transformed prior to analysis to normalize distributions.
We tested for convergence indicators using two different estimates from the surrounding landscape to answer different questions regarding convergence. First, we compared urban versus non-urban land to test if urban hydroscapes are converging relative to all surrounding land, regardless of human alterations to the surrounding land. The non-urban land in this comparison included water bodies that originated from dams and those that occurred within agricultural development. Second, to test if urban hydroscapes are converging relative to minimally disturbed hydroscapes, we compared urban versus undeveloped land cover without dams. Although any space for time substitution limits the inferences regarding change, this comparison provides the best method for understanding how urban hydroscapes may have changed. We made this comparison without dams because we assume the hydrographic parameters (specifically A WB) without dams to be a more accurate estimate of the antecedent hydrography. Convergence of A WB (the parameter most likely to be impacted by dams) was tested both with and without dams. We compared the degree of convergence in different urban land covers (open area, low, medium, and high intensity) using the undeveloped land cover without large dams. MSAs were also categorized by population size into three categories: small (<250,000 people, n = 33), medium (250,000–1 million people, n = 34), and large (>1 million people, n = 33). The convergence indicators were calculated for each population category.
To assess whether changes in the area of surface water were associated with changes in the number of water bodies, we examined the relationship between ΔA WB and ΔN WB across the 100 cities. A positive relationship between these variables would suggest that gain or loss of surface water occurs simply through the addition or removal of water features, but other combinations are possible. We categorized each city into one of four groups based on the values of these two measures: wetter (ΔA WB > 0, ΔN WB > 0), drier (ΔA WB < 0, ΔN WB < 0), fragmented (ΔA WB < 0, ΔN WB > 0), and consolidated (ΔA WB > 0, ΔN WB < 0). Cities in the “wetter” category had larger area and numbers than their undeveloped landscapes. Conversely, cities in the “drier” category had smaller area and numbers than their undeveloped landscapes. Cities in the “fragmented” and “consolidated” categories had more complex differences. Cities in the “fragmented” category had smaller water body area, but larger numbers of water bodies, whereas cities in the “consolidated” category had a larger water body area, but were fewer in number. IBM SPSS v20 was used for all statistical procedures.
Hydrography of Undeveloped Land Cover
Evidence of Convergence
As urban development intensity increased the variance in A WB and N WB decreased. For example, the variance of A WB decreased by 76% in open area and 98% in high intensity urban development compared to undeveloped lands. The decrease in variance coincided with a decrease in the median surface water abundance, suggesting the abundance on which hydrography is converging is lower than the median surface water abundance of the undeveloped land (Figure 3). Compared to undeveloped land, the median A WB decreased by 47, 64, and 74% in low, medium, and high intensity urban land, respectively. A parallel trend was observed for D C, which decreased by 28% in the low intensity and 48% in medium and high intensity urban land cover relative to undeveloped land. The variance of D C remained relatively constant across the urban intensity classes, but was much less than the undeveloped D C for all urban development intensities.
City size and the exclusion of either dams or agricultural land cover in the surrounding land had minimal impact on the results. The variance of all non-urban land was lower than that of the undeveloped land by 24, 6, and 18% for A WB, N WB, and D C, respectively; however, the variance of all non-urban land was still larger than the urban land covers. The variation in undeveloped land parameters was also very similar with and without dams. Likewise, no differences were identified between the city size categories for any of hydrographic parameters, nor for the indicators of convergence (data not shown).
Converging Urban Hydroscapes
Based on patterns in US cities, we conclude that urbanization causes the convergence of surface water abundance at broad scales. Surface water abundance in the undeveloped land of this study followed previously observed continental scale patterns in hydrography (Downing and others 2006); however, the surface water abundance of the urban landscapes did not conform to these hydrographic patterns. Continental scale variation among urban hydroscapes was far less than variation between undeveloped hydroscapes. In addition, the extent and direction of differences between urban and surrounding undeveloped hydrography appear to be related to the broad geographic patterns in climate and topography. Urban land in dry regions had greater abundances of surface water than surrounding undeveloped land, whereas urban land in very wet regions had lower abundances of surface water than their surroundings. Only by examining patterns at broad spatial scales are we able to understand the role of climatic, physiographic, and anthropogenic drivers of surface water distributions (Heffernan and others 2014; Thorp 2014).
The convergence of surface water abundance may also cause the convergence of surface water characteristics. Previous studies on urban stream loss document a preferential loss of smaller water features (Elmore and Kaushal 2008; Roy and others 2009; Larson and Grimm 2012). Urban water bodies tend to be more moderately sized and less connected to surface channels than their counterparts in undeveloped lands (Steele and Heffernan 2014). These differences in physical form may reflect the relative change in removal costs and value with size, certain functional needs, or aesthetic preferences. As a whole, the convergence of water body sizes suggests a reshaping of urban hydroscapes that is more pervasive and nuanced than simply adding or removing surface water from the landscape.
This study adds to the growing evidence that urbanization is a homogenizing process (Groffman and others 2014). Although cities are highly heterogeneous at local scales (Cadenasso and others 2007), the standardization of design, construction, and land use creates urban ecosystems that are broadly similar across large regions (Pouyat and others 2003, 2007; Pickett and others 2011; Seto and others 2012). Such similarities lead to urban growth and development patterns that converge toward a specific form (Seto and Fragkias 2005; Batty 2008). Ecological communities in cities are homogenized by the intentional and accidental shuffling of flora and fauna by commerce and through planting, which introduces a common pool of cosmopolitan species to similarly structured urban environments (McKinney 2006; Lososová and others 2012). Our data suggest that homogenization of urban environments extends beyond changes in land cover and species composition to include transformation of the land’s shape, and indeed, whether it is even dry land at all.
Urban Development Intensity and City Size
As predicted, increasing development intensity led to greater convergence of surface water abundance. We primarily attribute this pattern to the increased value of dry land for building, where land is scarce and demand for development is high. Alternatively, this pattern may also reflect a legacy of land use. Most urban cores in the US were established as centers of business and industrial production, activities for which water bodies and streams are of little use or value. As manufacturing industries relocate and city centers develop service oriented economies with larger residential populations and mixed land uses, we may observe a reincorporation of surface water into high density development. Surface water, such as park ponds and streams, serve as recreational amenities for urban residents and tourists and therefore are likely more highly valued in this type of urban center (Abbott and Klaiber 2013). Regardless of the drivers, it is important to note that the difference between urban and undeveloped land was far greater than the differences between the intensity classes for both water body coverage and channel density. Given that most urban development is low intensity, the majority of surface water in a city may still be substantially different than its original condition.
This study provides a unique examination of urbanization across a large number of cities and range of population sizes. Although many aspects of cities vary predictably with population (Bettencourt and others 2007), we find the population of cities only indirectly affects hydrographic characteristics. Cities with smaller populations tended to lack the high intensity urban development, where observed changes were most pronounced and covered smaller areas; however, the hydrography of other land cover types (open, low, and medium intensity) was indistinguishable between large and small cities. This finding is important given that the number of small towns far exceeds the number of large cities. Thus any positive or negative effects resulting from urban hydrographic convergence will affect a much larger, extensive area.
Shaping the Urban Hydroscape: Regional Patterns and Mechanisms
In both absolute terms and relative to lands surrounding cities, urban hydroscapes are both structurally and spatially complex, and exhibit strong geographic patterns. When the differences in both water body number and areal coverage for each MSA were observed in tandem, cities with minimal surface water abundance tended to have both a larger areal coverage and numeric density, although a few exceptions did not follow this pattern. Across the rest of the US, a large number of cities fell into the fragmented or consolidated categories. Cities in the fragmented category (ΔA WB < 0, ΔN WB > 0) were mostly located along the southern coastline, where expansive areas of wetlands are prominent in undeveloped lands. In comparison, the urban water bodies covered less area, but were greater in number. This fragmented pattern may reflect the drainage and breakup of these wetland tracts, as observed in case studies (Du and others 2010), or may reflect selection of sites around or between those features. The compensatory mitigation of wetland losses predominantly leads to wetland reconstruction outside of the affected watersheds (Kettlewell and others 2008). Similarly, large reservoirs that serve as municipal water supplies and flood control are often located outside of developed areas. In effect, relocation accumulates and consolidates water lost within the urban core elsewhere, but at this time its contribution to the observed patterns is unknown.
The consolidated pattern was particularly strong in the Midwest and Plains cities. Here water bodies in the urban land covered a larger area, but were fewer in number compared to those in undeveloped landscapes. Although we tried to minimize the presence of other land covers within the undeveloped area, it is possible that the consolidation pattern of the mid-section of the US reflects a legacy of agricultural development in the undeveloped land cover. Artificial drainage and the presence of agricultural farm ponds could generate a pattern of small areal coverage and high numbers of water bodies (Skaggs and others 1994; Downing and others 2006). Though we cannot determine agriculture’s contribution to these patterns with certainty, we still observe convergence at the US scale even when agriculture is included in the surrounding land estimate. This result suggests that urban hydrography is not simply inherited from an agricultural legacy, and highlights the need for investigation of agricultural hydrography at macro-scales, as well as how and where urban development incorporates or alters those characteristics.
Urban hydrographic patterns likely reflect two main mechanisms: alteration and choice of location for building. Alteration includes construction, removal, or relocation of surface water features during urban development. Direct observation from previous, more detailed case studies of urban hydrographic change (Elmore and Kaushal 2008; Roy and others 2009; Du and others 2010; Larson and Grimm 2012) suggests that hydrographic alteration is a dominant mechanism of change in some cities. An alternative explanation for the hydrographic patterns we observed is that cities are selectively built in areas with particular hydrographic characteristics. Arid cities, for example Phoenix, are often built along rivers and other locally “wet” areas within the arid landscape that could account for the apparent increase in surface water abundance. Many cities are built on major rivers or the shorelines of major water bodies and the tendency to build in very similar locations may contribute to overall similarities in hydrographic pattern.
The relative role of the mechanisms of hydrographic change, like other land use and land cover changes, is likely spatially and temporally heterogeneous. Spatial heterogeneity in the mechanisms may be correlated with the figurative distance of a city from the convergence point. For example, cities built in regions at the extremes of surface water abundance may require more substantial alteration (for example, Phoenix, Miami, Houston). In addition, the role of these mechanisms is also likely temporally heterogeneous. The distribution of the types of water bodies in cities of different sizes suggests selection may be more important in the initial stages of cities and changes as cities grow outward from their original settlements (Steele and Heffernan 2014). A case study of water bodies in Phoenix, wherein the initial settlement was located along a locally wet riparian area, supports the idea of temporal variation. As the city has grown outward, the density of built water bodies has increased, and the structure of watered lands has changed from one linearly centered along rivers and riparian areas toward more expansive watered landscapes that serve urban and residential purposes (Larson and Grimm 2012). It is important to mention, however, that using a space for time substitution limits our ability to draw inferences regarding how the hydrography of cities changes.
Implications for Ecosystems and Management
A number of ecological and ecosystem processes are mediated by the abundance of surface water and its geomorphology. The abundance of lakes, ponds, wetlands, and streams, as well as the size of water bodies and the connectivity between these features, mediates watershed fluxes of carbon, nitrogen, and phosphorus (Cole and others 2007; Fraterrigo and Downing 2008; Downing 2010) in addition to species abundance and distribution (Dunham and Rieman 1999; Dahlgren and Ehrlen 2005). Further research will be needed to determine if the convergence of urban hydrography leads to a convergence of biogeochemical and ecosystem properties related to the surface water abundance and characteristics. Some urban watersheds, for example those in coastal areas, may function very differently than the original ecosystem. For others, the differences in the hydroscape structure may be of little consequence to local ecosystems relative to the effects of pollution and other changes associated with urbanization.
Though the urban hydroscape is converging, this does not indicate that the management of urban watersheds should be likewise homogeneous. Cities manage surface water to supply a host of services as well as mitigate the negative consequences of an overabundance of water (for example, flooding, mosquitoes) or aridity (for example, scarcity, heat). We can only speculate about the extent to which the convergent pattern represents an optimization of services provided by surface water; however, increasing these benefits is likely to remain a continuing goal of landscape design and development. Depending on the service, some cities may benefit from returning to a hydroscape more closely resembling the antecedent one, whereas others may benefit from a further departure. For example, small water bodies in the urban landscape more efficiently regulate microclimates (Sun and others 2012; Sun and Chen 2012) and incorporation or reestablishment of small water bodies in cities may provide cooling benefits within the urban heat island and energy savings. For some cities that may have removed a significant portion of the small water bodies, this re-establishment would return the city to a hydroscape more similar to the antecedent one; however, for cities such as Phoenix, such perennial features would present a further departure.
Hydrographic patterns reflect the interaction between geophysical constraints and the policies and practices of diverse decision makers whose decisions shape water management, stream and wetland conservation, and urban land use and land cover change at a variety of scales (Roy Chowdhury and others 2011). At broad scales, these multi-scaled decisions lead to a configuration of urban development in the United States that is less dense and sprawling relative to cities elsewhere (Huang and others 2007; Schneider and Woodcock 2008). Higher density development in this study increased the convergence of urban surface water on a “drier” hydroscape relative to lower densities. Whether similar patterns occur in more consolidated, densely built regions of the globe might indicate how different priorities and constraints shape relationships among people, water, and the built environment across a wide range of political, economic, and environmental conditions.
Differences between urban and undeveloped hydrography reveal the environmental context of surface water abundance. We observed lower variance in surface water abundance parameters amongst urban lands than amongst their undeveloped counterpart, converging on a central value. Urbanization results in a tightly constrained range of urban hydrographic characteristics, at least within US cities. As such, the loss of spatial heterogeneity across the country indicates that the built environment may be a stronger predictor of the abundance of surface water features than climate and topography, two of the most fundamental drivers of large-scale patterns in nature.
We thank Emily Bernhardt, Martin Doyle, Jacqueline Aitkenhead-Peterson, Allison Appling, and the anonymous reviewers for helpful comments on earlier drafts of this manuscript. This research was supported by grants from the National Science Foundation’s Macrosystems Biology program (NSF EF #1065785).
- Army Corps of Engineers. 2013. National inventory of dams. Sausalito: US Army Corps of Engineers.Google Scholar
- Census Bureau. 2012. Metropolitan and micropolitan statistical areas. Washington, DC: Census Bureau.Google Scholar
- Downing JA. 2010. Emerging global role of small lakes and ponds: little things mean a lot. Limnetica 29:9–24.Google Scholar
- Dunne T, Leopold LB. 1978. Water in environmental planning. San Francisco: WH Freeman.Google Scholar
- Groffman PM, Cavender-Bares J, Bettez ND, Grove JM, Hall SJ, Heffernan JB, Hobbie SE, Larson KL, Morse JL, Neill C, Nelson K, O’Neil-Dunne J, Ogden L, Pataki DE, Polsky C, Roy Chowdhury R, Steele MK. 2014. Ecological homogenization of urban America. Frontiers Ecol Environ 12(1):74–81.Google Scholar
- Heffernan JB, Soranno P, Angilletta M, Buckley L, Gruner D, Keitt T, Kellner J, Kominoski J, Rocha A, Xiao J, Harms TK, Goring S, Koenig L, McDowell WH, Powell H, Richardson A, Stow C, Vargas R, Weathers KC. 2014. Macrosystems ecology: understanding ecological pattern and process at continental scales. Frontiers Ecol Environ 12(1):5.CrossRefGoogle Scholar
- NEON. 2010. Field site and sampling strategy.Google Scholar
- Pouyat RV, Russell-Anelli J, Yesilonis ID, Groffman PM. 2003. Soil carbon in urban forest ecosystems. In: Kimble J, Ed. Potential of United States forest soils to sequester carbon and mitigate the greenhouse effect. Boca Raton: CRC Press. p 347–60.Google Scholar
- Roy Chowdhury R, Larson K, Grove JM, Polsky C, Cook E. 2011. A multi-scalar approach to theorizing socio-ecological dynamics of urban residential landscapes. Cities Environ 4:1–19.Google Scholar
- Steele MK, Heffernan JB. 2014. Morphological characteristics of urban water bodies: mechanisms of change and implications for ecosystem function. Ecol Appl. doi: 10.1890/13-0983.1.
Open AccessThis article is distributed under the terms of the Creative Commons Attribution License which permits any use, distribution, and reproduction in any medium, provided the original author(s) and the source are credited.