Study area

Our study was conducted in Hamilton County, a 1074-km² area in the southwestern corner of Ohio, USA (Fig. 1). The county includes a variety of land covers, from high-density urban land use in the center of Cincinnati to mature forests in parks, nature preserves, and old estates. Based on the 2001 USGS National Land Cover Data (NLCD), the county has 50.2% developed land, 36.1% forested land, and 11% agricultural land. The city has a mixture of combined sanitary and storm sewers (built starting in the 1800s), and separate sanitary sewers routed to wastewater treatment plants and storm sewers routed to streams (built since the 1960s). The urban water infrastructure in the county includes various combinations of private wells, public drinking-water pipes, and private septic tanks that outlet to leach fields.

The region has a continental climate, with cold winters and hot summers. Mean annual precipitation in the county is ∼100 cm (Lerch et al. 1982) and evapotranspiration is 73 cm (Shuster et al. 2007). Hamilton County is in the Central Lowland physiographic province, and dominant features are gently rolling glacial uplands, steep hillsides, glacial river terraces, and flood plains (Lerch et al. 1982). About ⅔ of the county (southwestern part) is in the Interior Plateau level III ecoregion, and dominant features are rolling to rugged terrain underlain by sandstone, siltstone, shale, and limestone (Omernik 1987). The northeastern part of the county is in the Eastern Corn Belt Plains ecoregion and has rolling till plains set on glacial deposits (Omernik 1987).

Field data collection

We began with a stream network for Hamilton County that had been created by Cincinnati Area GIS (CAGIS) from DEMs (0.61-m contours). The network was fitted to approximate most channels, and CAGIS personnel had hand-edited part of the stream network on or near culverts and eliminated all streams in downtown Cincinnati that were known to have been piped, but the map had not been ground-truthed. We selected a random subset of 150 channel origins from the 6686 channel origins in the CAGIS stream network. We visited each randomly selected channel origin twice: 1) during the wet season (winter/spring 2006 and 2007), and 2) during the dry season (summer/autumn 2006). We waited at least 48 h after large rainfall events to ensure that ephemeral streams were not flowing. We used a handheld personal digital assistant (PDA) and geographical positioning system (GPS; Axim X5 with World Navigator Global Positioning System; Dell, Boston, Massachusetts) with navigator software (Pocket Navigator 4.0; Maptech, Amesbury, Massachusetts) and USGS 1:24,000 topographic maps to aid field surveys.

During the first visit, we walked or drove upstream or downstream of the point of channel origin identified from the map until we found the actual channel origin. Channel origins or heads are the upslope boundary between the hillslope and channel, where a channel has defined banks and bed material (Dietrich and Dunne 1993, Fritz et al. 2006). If the channel split into multiple tributaries, then we recorded the coordinates of all channel origins for tributaries >10 m long. If we found no channel at the point of channel origin selected from the map (i.e., the channel was piped or buried), then we moved downslope until we reached the actual channel origin (e.g., where the pipe returned to the surface or the stream began) and recorded the coordinates. We recorded the diameter of the pipe at channel origins that had pipe outlets. If we were unable to access a channel origin (e.g., no landowner permission), we surveyed the nearest accessible channel origin.

We also mapped and recorded ephemeral, intermittent, and perennial flow origins in each channel. We classified channel origins as ephemeral, intermittent, or perennial flow origins, depending on when they were flowing. Intermittent flow origins had continuous flow in spring but not in summer. Perennial flow origins had continuous flow in spring and summer. Ephemeral flow origins had no flow in spring or summer. We walked downstream of ephemeral flow origins until we reached the intermittent and perennial flow origins in the channel. We identified intermittent or perennial origins based on presence of continuous surface flow for a distance longer than the length of the adjacent dry section. We recorded coordinates, pipe diameter(s) (when applicable), bankfull width, bankfull height, and flood-prone width (i.e., width at 2× bankfull height; Fritz et al. 2006) of all intermittent and perennial flow origins. If intermittent or perennial flow origins were inaccessible, we did not survey additional points. We resurveyed all channels and identified flow origins in both seasons. For example, when we returned in summer to resurvey channels that had been surveyed in spring, we started at the mapped intermittent flow origin and walked downstream to locate the perennial flow origin. We compared permanence designations from the 2 visits in 2006 to designations based on surveys made 5 times/y for 4 to 6 y between 2003 and 2008 to assess the accuracy of our method for determining flow category in 4 intermittent and 3 perennial stream reaches.

GIS calculations

We used ArcGIS® (version 9.2; Environmental Systems Research Institute, Redlands, California) to calculate catchment areas, land cover, and other geographical information for catchments of channel and ephemeral, intermittent, and perennial flow origins. We used ArcHydro® tools with a National Elevation Dataset DEM (⅓ arc second, seamless, ∼10-m resolution) to identify catchments for each mapped origin. We verified all catchments against hand-drawn maps to check for correspondence to the correct location and redigitized, as necessary, from 3-m (10-ft) contour maps. We calculated % land cover as forest (deciduous, evergreen, and mixed forest; emergent and woody wetlands), agriculture (cultivated/exposed, pasture hay), and urban (high-, medium-, and low-intensity urban; open space developed) within 1) catchments and 2) 100-m buffers around randomly selected channel origins from the USGS 2001 NLCD (30-m pixels). Urban land cover increased between the collection of NLCD data and our field collection (5–6 y later), but we used NLCD because of its regional coverage and widespread use around the country. We also calculated % impervious surface area from the NLCD imperviousness layer (created from 2001 Landsat 7 ETM+ data). We calculated road and sewer densities (m/km²) and septic densities (no. tanks/km²) for each catchment with data obtained from the Department of Transportation and Cincinnati Metropolitan Sewer District. We determined elevation and relief ratio (i.e., difference in elevation between highest and lowest point in the catchment divided by the catchment area) from 3-m contour maps. We used the National Resources Conservation Service Soil Survey Geographic (SSURGO) Database to determine the % area of each soil type and multiplied % area by average soil depth for each type (Lerch et al. 1982) to obtain average catchment soil depth.

We used ArcHydro® (Environmental Systems Research Institute) to determine the flow accumulation coefficient (FAC) for the mapped stream network that corresponded to each mapped origin. The flow accumulation grid records the number of cells that drain into an individual cell (∼10 m × 10 m), and the coefficient is the threshold number of cells required for initiation of a stream. Our goal was to determine the unique FAC associated with each origin, so we created stream networks with various FAC values and matched them to the mapped origins. The FAC for an origin was the value (within increments of 25) that created a stream network that reached but did not extend upslope of the mapped origin. Where a network could not be created from the DEM for the channel, we considered the FAC to be 0. FAC values were less precise than catchment areas, particularly for smaller catchments, for which FAC  =  0 and catchments were hand-drawn. Therefore, we used the relationship between FAC and catchment areas for origins with FAC > 0 to convert catchment areas to FAC values:

Data analysis

We tested all variables for normality and transformed them as necessary. We used an arcsine(√[x])-transformation for all percentage variables. We used t-tests (2-tailed, assuming unequal variances) to compare catchment characteristics for channel and ephemeral, intermittent, and perennial flow origins between the 2 ecoregions (Eastern Corn Belt Plains and Interior Plateau). We used t-tests to compare % urban, % forested, and % agricultural land cover within the 100-m buffer around randomly selected channel origins at sites where a channel was present and where a channel was absent (i.e., the channel was buried). We used t-tests to compare catchment areas of channels with ephemeral, intermittent, and perennial flow origins between primarily forested (75–100% forest, depending on origin type) and primarily urban (75–100% urban cover) catchments. We also used t-tests to compare bankfull width:depth ratios and entrenchment ratios (i.e., flood-prone width divided by bankfull width) at intermittent and perennial flow origins between sites with and without pipes.

We developed linear regression models to predict mean catchment area of channel and ephemeral, intermittent, and perennial origins using catchment characteristics. Catchment area served as a surrogate for location of origins relative to ridge tops. Independent variables included measures of anthropogenic disturbance (road density, sewer density, septic density, % urban, % agriculture, % imperviousness) and natural factors (relief, elevation, soil depth, and % forest cover) that were expected to influence catchment area. We included relief ratio, which captures the gradient of the entire catchment, in all models because several studies had indicated the importance of local or valley slope in predicting channel origins (e.g., Montgomery and Dietrich 1989, Tucker and Bras 1998). We expected positive relationships between road density, % urban, % agriculture, and % imperviousness and catchment areas because these variables indicate decreased infiltration and routing of runoff to downstream locations. We also expected a positive relationship between sewer density (a combination of both sanitary and storm sewers) and catchment area, although water can leak both into and out of unpressurized pipes. On the other hand, we expected an inverse relationship between septic tanks and catchment area because septic systems contribute novel water sources to the catchment and should increase flows of headwater streams (thereby decreasing catchment area).

We compared models with Akaike's Information Criterion adjusted for small sample size (AIC_c). We considered the model with the highest Akaike weight (w_i) for each model set to be the best supported model, but we regarded all models with ΔAIC_c ≤ 2 as plausible alternatives (Burnham and Anderson 2002). We report adjusted r² values and parameter estimates for the best supported models for the Eastern Corn Belt Plains, Interior Plateau, and combined ecoregions. We did all analyses with JMP (version 5.1; SAS Institute, Cary, North Carolina).

Mapping headwater streams

Commonly used stream maps greatly underestimate actual stream length (Mueller 1979, Meyer and Wallace 2001). Inaccuracies are biased toward underrepresentation of the smallest, intermittently flowing headwater streams. In our study (in the eastern US), 1:100,000-scale maps identified only ½ of perennial channel length, and ∼1/7^th of the total channel length, whereas 1:24,000-scale maps identified 22% of the total channel length. These underestimations are comparable to those reported for a field assessment in the Chattooga River watershed in the Blue Ridge Mountains of Georgia, South Carolina, and North Carolina (Hansen 2001). However, Colson et al. (2008) found that accuracy of maps can vary by ecoregion. USGS 1:24,000-scale maps underestimated stream length in the Southwestern Mountain ecoregion of North Carolina by 54% and overestimated stream length in the Coastal Plain ecoregion by 31% (Colson et al. 2008). This difference could be caused by higher drainage densities in mountains than in lower elevations, or cartographic guidelines that require minimum lengths and distances from ridges before identifying streams (Drummond 1974, Colson et al. 2008). The substantial underestimation of streams based on nationally available maps can lead to inaccurate scientific understanding of ecosystem processes. For example, riparian buffers calculated from coarse-scale maps overestimated buffer prevalence and led to artificially high estimates of watershed-scale nutrient retention potential (Baker et al. 2007). In regions where headwater streams have retained their forested buffers, coarse-scale maps could underestimate riparian buffers. Accurate delineation of stream networks also is a critical first step for identifying anthropogenic impacts within catchments and developing watershed management plans that protect headwater streams and, thus, downstream ecosystems.

We used catchment area and corresponding DEM-derived FACs to estimate channel length and to indicate relative locations of origins and channels in the landscape. Thus, our method depended on our ability to predict catchment area. Relief ratio can accurately predict catchment areas of ephemeral, intermittent, and perennial flow origins. Other researchers have reported strong relationships between local slope or valley gradient, which are comparable to relief ratio, and catchments areas of channel origins (Montgomery and Dietrich 1989, Tucker and Bras 1998, Vogt et al. 2003, Colombo et al. 2007; but see Jaeger et al. 2007). These results suggest that DEMs can be effective tools for mapping stream networks. In our study, soil depth also was included in the best-supported models for catchment areas of ephemeral and intermittent flow origins, a result suggesting that interflow might be critical for maintaining stream flows. These 2 geographic variables are easily obtained from topographic and soil maps and predicted 71 to 77% of the variation in catchment area for channel origins. Our predictive models were relatively strong, but we emphasize that channel and flow origins were highly variable across sites (even in natural settings) because locations of origins are related to local headcuts, bedrock, springs, and other site-specific variables (Adams and Spotila 2005). Thus, precise locations of origins are not likely to be mapped accurately and consistently from regional landscape variables. However, locations of origins might be easily and more accurately predicted in some regions, such as areas strongly influenced by groundwater.

With the exception of forest cover, few of the variables describing anthropogenic disturbance were included in the best supported models explaining catchment areas of origins. General indicators of urban anthropogenic disturbance (e.g., % urban land cover, % impervious surface area) encompass factors that could increase (e.g., via landscape irrigation) or decrease (e.g., via reduced infiltration) base flows. These opposing effects might prevent variables indicating urban disturbance from predicting catchment areas of origins. Sanitary sewer lines commonly are positioned near headwater channels and have the potential to infiltrate local soil water or exfiltrate wastewater, potentially confounding our ability to detect a consistent pipe effect across all streams. Depending on the location of pipe outlets relative to flow origins, storm sewers could increase (if upstream) or decrease (if downstream) flow in channels. Septic tanks, which are likely only to increase water supply to headwater streams, were not included in the best supported models. The only urban disturbance variable included in the best supported models was road density, which was positively related to catchment area of ephemeral channels and reflected burial or piping of channels in areas with many roads. Road density was inversely related to catchment area of perennial channels in the Interior Plateau. This relationship might reflect a sustained water supply from gullies adjacent to roads. Despite regulations, individual landowners decide whether to fill, pipe, dredge, channelize, impound, or otherwise alter stream channels. In areas where most property is privately owned (as in this study), these personal decisions probably are an important source of variability of stream burial in urban environments and decrease our ability to predict catchment area based on anthropogenic disturbance variables. In catchments with relatively uniform, publicly operated water infrastructure, relationships between anthropogenic disturbance variables and catchment area might be stronger.

We used DEMs (10-m resolution) to generate stream lines. DEM-derived stream networks are likely to be more accurate than other national stream network sources, such as USGS NHD and topographic maps (depending on map scale), but DEMs also have potential inaccuracies. Advanced Light Detection and Ranging (LiDAR) technology probably would have provided a more accurate account of ground elevation than DEMs and, consequently, improved our ability to generate stream network maps. However, elevation maps based on 10-m DEMs and LiDAR are comparable except where little topographic relief exists in the Coastal Plain (Colson et al. 2006). Depending on the source and year of the data, DEMs might not reflect recent landscape transformations associated with urbanization, which can occur quickly and can drastically change valley slopes. Even when streams are buried or the landscape is recontoured to route water to detention basins, new stream channels (especially ephemeral channels) are likely to be carved into the landscape. Thus, stream maps created from DEMs should not be seen as static, but rather should be continuously updated with the most recent and accurate GIS data.

Given the variability in locations of channel and flow origins observed in our study, we think that field surveys are necessary to produce accurate maps of ephemeral, intermittent, and perennial stream networks. Field surveys are particularly important in catchments where researchers want to understand processes, such as nutrient or sediment dynamics, occurring downstream that might depend on upstream drainage density, connectivity, and flow. For particularly large catchments, researchers could consider mapping streams initially based on empirical formulas (e.g., from our study for eastern US streams) and conducting field surveys to adjust actual locations.

Field-survey methods come with caveats. First, observations of hydrologic permanence are highly dependent on season and precipitation. We avoided sampling immediately after large rain events, but the amount of rain, time since rainfall, level of ground saturation, and depth of the water table all affect whether and for how long a particular stream reach is flowing after a storm (Hunter et al. 2005). In addition, interannual differences in climate will affect mapping results (see Paybins 2003). Our study was conducted in a year with slightly higher than average rainfall (125 cm) for the years 2000 to 2007 for the study region (average 120 cm; National Oceanic and Atmospheric Administration National Climatic Data Center, Cheviot, Ohio). Consequently, the 2 observations in 2006 correctly classified 3 of the 4 intermittent reaches and all 3 perennial reaches according to data collected on 7 streams over 6 y. However, had our study been in summer 2007, a drier year (104 cm rainfall), many of the perennial streams would have been considered intermittent (HRL, unpublished data). An alternative to point surveys is long-term, continuous monitoring of the presence and duration of flow (as done by Svec et al. 2005), which should accurately determine hydrologic permanence of headwater streams and eliminate some of the problems of intra- and interannual variability. However, gauges placed at single points in a catchment cannot capture longitudinal differences in permanence within stream networks. Geomorphic and biological indicators also could be used as surrogate measures of hydrologic permanence. For example, the North Carolina Division of Water Quality assigns point values to geomorphic, hydrologic, and biologic characteristics of the stream reach to classify streams as intermittent and perennial (NCDWQ 2005). However, these alternatives are more time-consuming and expensive than the approach we used, and these problems minimize their usefulness for large geographic areas. Last, our methods did not account for situations in which long sections of channel downstream of flowing reaches are dry (i.e., discontinuous or interrupted surface flow). In a temporal assessment of flow in headwater streams, Hunter et al. (2005) found no predictable pattern of longitudinal drying along a stream reach. Instead, dry-season recession typically was characterized by a transition from long segments of surface flow to short segments of dry areas with scattered pools. Despite these caveats, the random sampling and methods of on-the-ground observation used in our study were cost-effective, enabled us to gather data efficiently over a wide area, and allowed us to estimate the locations of ephemeral, intermittent, and perennial streams. However, the methods should be tested elsewhere to determine their appropriateness and to test the generality of the predictive models in other regions of the world.

Urbanization effects on headwater stream length

Catchment areas of ephemeral and intermittent channel origins were larger in urban than in forested areas. We estimate that urban areas have lost 93% of historical ephemeral channel length and 46% of historical intermittent channel length (Fig. 7). These estimates might be slightly high because most forested catchments in Hamilton County were logged (Bryant and Held 2004), and logging can cause headcut migration up tributaries and increase channel length in forested catchments relative to unlogged natural conditions (Schumm 1999). The differences in area between urban and forested catchments might have arisen because developments typically occur on flatter landscapes, which have naturally larger catchments. However, the correlation between relief ratio and urban land cover was weak (r  =  −0.33, n  =  240). The loss of ephemeral and intermittent stream length also might be a consequence of increased flow in urban areas. For example, naturally ephemeral or intermittent streams could become perennial if flow increased. However, our study was not designed to distinguish among the causes of reductions of channel length, and hydrologic monitoring would be necessary to understand changes in baseflow hydrology with urbanization.

Catchment area of perennial channels was smaller in urban than in forested areas, despite the fact that the remaining forests in Hamilton County are on steep slopes where catchments are small. We estimate that urban areas gained 22% of perennial channel length relative to historical perennial channel length, a result suggesting that some naturally intermittent streams had been converted to perennial streams (Fig. 7). Our predictive models indicate that the increase in perennial channel length in urban catchments is related to deforestation. In the eastern US, deforestation increased water yield (i.e., surface-water discharge) because of decreased evapotranspiration (Hewlett and Hibbert 1961, Bosch and Hewlett 1982, Sun et al. 2005). Increased water yield in urban catchments might explain why perennial channels have smaller catchments in urban than in forested landscapes. Alternatively, urban and forested land cover are strongly negatively correlated (r  =  −0.83, n  =  240) in Hamilton County, and more urban land cover (rather than less forest land cover) could have contributed to increased flows in urban catchments. However, the likely sources of increased flows in urban catchments (e.g., leaky sewers, septic tanks; Lerner 2002) did not improve the ability of our models to predict catchment area of perennial flow origins in urban catchments.

The increased perennial channel length in urban catchments does not indicate that perennial streams were not piped. In fact, 18 of the 64 channel origins with pipes had perennial flow, and 40% of perennial flow origins were in pipes. Thus, piping of perennial streams was widespread in urban catchments. Moreover, we did not account for the many streams (perennial and otherwise) in downtown Cincinnati that have been converted to sewer lines. Multiple mechanisms affect the length of perennial channel in urban areas, including piping of channels and conversion of ephemeral and intermittent channels to perennial channels.

Regulatory implications in the US

Our study highlights the inaccuracies of existing stream maps and the risks of using USGS NHD or 1:24,000-scale topographic maps as a basis for stream regulation and landuse planning. USGS topographic maps typically are used as a source when defining regulatory boundaries, such as riparian buffer requirements (Russell 2008) and forestry best-management practices (Svec et al. 2005). Thus, regulations are not likely to provide any protection for ephemeral and intermittent streams, and insufficiently protect perennial streams. To our knowledge, North Carolina was the first state to have developed field protocols for mapping headwater streams. The new, field-based maps have been used for stream regulation in North Carolina and for jurisdictional determinations in other states (Russell 2008). Regulations frequently are based on estimated catchment area for intermittent (e.g., 4.9 ha, West Virginia permitted valley fills; Paybins 2003) or perennial (e.g., 5.3 ha, Washington State forest practices; Jaeger et al. 2007) streams. However, differences in catchment area associated with different types of flow origin mean that use of any standardized area as a cutoff is likely to result in incomplete protection of headwater streams. Regardless of the method used to map streams, the smallest headwater streams must be included in planning and management plans to ensure comprehensive stream protection.

Our results also have implications for federal regulations pertaining to stream protection. The larger catchment areas for ephemeral and intermittent channels in urban landscapes might indicate that small streams have been piped and filled, perhaps because of misperception of these temporary waters as valueless ecosystems and minimal legal protection. In the US, the Army Corps of Engineers (USACE) oversees permits for discharge of dredge or fill material through Section 404(b) of the Clean Water Act (CWA), but permits are required only for “waters of the United States”, which might not include ephemeral and intermittent channels. Moreover, nationwide permits allow disturbance of streams if the length of stream or catchment area buried or disturbed is small (e.g., 91 m of stream or 0.2 ha area for residential, commercial, and industrial development), and these limits can be waived for streams classified as intermittent or ephemeral. The jurisdictional authority to protect headwater streams under the CWA has also recently come under scrutiny in the US Supreme Court in Rapanos (consolidated Rapanos v. United States and Carabell v. United States, 126 S. Ct. 2008; Nadeau and Rains 2007, Leibowitz et al. 2008). Following the split decision in Rapanos, the US Environmental Protection Agency (EPA) and US Army Corps of Engineers (ACE) issued a joint memorandum describing how “waters of the United States” will be assessed for inclusion under the CWA (USEPA and USACE 2007). In accordance to Justice Scalia's opinion, the agencies state that surface water connection (i.e., perennial flow) warrants protection under the CWA. Based on the increase in perennial channel length with urbanization, more stream length is likely to be covered by this criterion in urban areas than before. Nonetheless, the requirement for perennial flow would leave many streams unprotected in US arid zones where many large rivers are intermittent (Nadeau and Rains 2007). A “significant nexus” to waters of the US is an alternate criterion for protection based on Justice Kennedy's opinion. However, it is not clear what data will be required for determining this connection (Leibowitz et al. 2008). The lack of criteria for determining a significant nexus is likely to lead to inconsistent protection, at best, and, at worst, a tendency for managers to ignore disturbances in temporary waters.

Ecological implications

Changes in the hydrologic permanence of streams can greatly affect stream communities and ecosystem functioning. Decreased flow caused by reduced infiltration in urbanized catchments can result in a loss of habitat for plants and animals that require water for all or part of their life cycle (Freeman et al. 2007, Meyer et al. 2007). The pools that remain are likely to support fewer taxa (Boulton and Hancock 2006), partially because of increased competition and predation in an unnaturally restricted habitat (Power et al. 1988). Increased flow (i.e., making streams more perennial) might increase habitat for some organisms and could lead to additional processing of nutrients and C (Richardson 1990, Stanley et al. 2004, Chadwick and Huryn 2005), but would negatively affect organisms that rely on temporary habitats for protection from fish and other predators. For example, perennial streams are more likely to harbor fish that can reduce the abundance of predatory invertebrates and salamanders, which are typical top predators in intermittent streams (Wilkins and Peterson 2000, Meyer et al. 2007). Wigington et al. (2006) reported that juvenile coho salmon are common in intermittent reaches, and overwintering smolts were larger than those from perennial reaches. The cumulative loss of ephemeral and intermittent streams also could negatively affect population recovery dynamics via reduced potential for recolonization (Fritz and Dodds 2004, Meyer et al. 2007). Last, urbanization might alter the variability of base flows, which could negatively affect organisms that require a certain number of continuously wet days (e.g., for breeding) or wet days during certain seasons (e.g., for nursery ground; Boulton and Hancock 2006). However, flow in temporary streams is stochastic by nature, and many organisms in these streams are adapted to drying and might be able to adjust to altered flow regimes (Lytle and Poff 2004, Boulton and Hancock 2006). Several studies have examined effects of drying on stream ecosystems, but most such studies addressed large droughts or were in areas where organisms are adapted to intermittent flow regimes (e.g., Lytle and Poff 2004). More research is needed on the effects of urban-induced hydrologic alteration on ecosystem processing and the ability of intermittent stream biota to adapt to altered timing, frequency, and duration of drying.

In urban streams, the challenge will be to isolate the effects of hydrologic permanence from the myriad of other urban impacts. We also need to understand better the cumulative effects of stream burial on downstream ecosystems. Some estimates of altered ecosystem processing with stream channel loss have been made using data in small streams (Meyer and Wallace 2001, Freeman et al. 2007). However, to our knowledge, no studies have assessed communities and ecosystem functions in water bodies receiving urban headwaters with and without piping. An understanding of the locations of headwater streams in the landscape is critical as scientists work to determine indicators of hydrologic permanence (e.g., Fritz et al. 2008) and the cumulative effects of headwater stream loss on downstream waters.