Introduction

The 2.4 million ha Adirondack State Park in northern New York State, USA, encompasses the world’s largest intact temperate forest which contains a globally unique landscape of wetlands, northern hardwood and boreal forests, alpine tundra, and vast fresh water resources (Jenkins and Keal 2004). This large rural landscape is exposed to multiple environmental stressors operating at varying spatial scales and often in interacting ways that are poorly understood (Allan 2004; Palmer and Yan 2013). Historically, the principal environmental stressors studied in the region have been acid deposition and climate change (Ito et al. 2002; Stager et al. 2009). In recent years, road runoff has been identified as a significant threat to aquatic and terrestrial ecosystems in the region, comparable to acid deposition (Kaushal et al. 2005; Kelting and Laxson 2010). Over 8000 km of paved roads traverse the Adirondack Park; therefore, the potential land and surface water area receiving road runoff pollutants may be significant and likely also occurs on protected state lands which represent 45 % of the area.

The main categories of pollutants found in road runoff are road de-icing salts, heavy metals, herbicides, and polycyclic aromatic hydrocarbons (Trombulak and Frissell 2000; Fay and Shi 2012; Tang et al. 2013). Road runoff pollutants have been shown to induce tree mortality (Fan et al. 2014), shift the composition of floral and faunal communities (Kaspari et al. 2010; Ke et al. 2013; Neher et al. 2013; Snell-Rood et al. 2014), promote establishment of terrestrial and aquatic invasive species (Johnston and Johnston 2004; Crooks et al. 2011), and elevate Na concentrations in roadside soils, displacing base cations (Ca, Mg, K) essential to ecosystem function (Norrström and Bergstedt 2001). Flux of displaced cations and Cl into streams has been reported downslope of salted roads (Daley et al. 2009; Price and Szymanski 2013; Corsi et al. 2015), and Na and Cl concentrations in lakes correlate positively with paved road density (Kelting et al. 2012). Road runoff pollutants in lakes and rivers can increase fish and amphibian mortality (Yousef et al. 1983; Karraker et al. 2008) and may induce top-down (Sloman et al. 2003) and bottom-up trophic cascades (Hodkinson and Jackson 2005). Road runoff pollutants can affect ecosystem function similarly to other regional stressors (Norrström and Bergstedt 2001) and can often interact synergistically, enhancing their effects on terrestrial and aquatic communities (Holmstrup et al. 2010). Pollutants commonly found in road runoff have been shown to exacerbate the effects of warming temperatures (Qiang et al. 2012), drought (de Silva et al. 2012), acid deposition (Rosfjord et al. 2007), eutrophication (Ferreira et al. 2008), and disease (Kiesecker 2002).

Although the impacts of road runoff on terrestrial and aquatic ecosystems are well known, no studies have been published that model the spatial extent of road runoff over large geographic areas. Accurate knowledge of the spatial extent of road runoff is essential to understanding and managing the interacting impacts of road runoff and other regional stressors on aquatic and terrestrial ecosystems (Soranno et al. 2014). Our objectives were to develop an efficient method of estimating the spatial extent of road runoff impacts on lands and waters over large geographic extents and to quantify the land area and surface water areas, number of water bodies, and total length of streams potentially impacted by road runoff from state and federal, county, and local roads within the Adirondack Park.

Methods

Study area

The Adirondack Park (Park) is located in New York State, USA (Fig. 1). The Park is 2.4 million ha in size and is roughly divided equally between private and public lands that are dominated by natural forest cover. The Park contains over 8000 km of paved roads, broken out into 1965 km of state and federal roads, 1803 km of county roads, and 4421 km of local roads (Table 1). The Park also contains about 103,000 ha of surface water (lakes and ponds) and over 13,000 km of streams.

Fig. 1
figure 1

Location of Adirondack Park in New York State, USA

Full size image
Table 1 Total road length by category, total area of surface water, and total length of streams in the Adirondack Park, New York
Full size table

Estimating spatial extent of road runoff

Spatial data for depth to bedrock and groundwater flow was lacking for the Adirondack region, so we used surface topography to delineate the potential flow of runoff downslope of paved roads. We used ArcGIS -ArcMap 10.2 (Esri, Redlands, CA) with a hydrologic terrain analysis toolset TauDEM 5.1.1 (TauDEM Toolbox 5.1.1, www.hydrology.usu.edu/taudem) to develop a series of parameters: a hydrologically relevant surface based on the D-infinity flow direction algorithm (Tarboton 1997), indicator grids of rasterized road networks, effective precipitation weights, and decay multipliers (Fig. 2). TauDEM 5.1.1’s “D-Infinity concentration limited accumulation” tool used these parameters to delineate flow downslope of the rasterized road networks. We then used a series of ArcGIS processes to quantify the total land area represented by the downslope accumulations, as well as to identify the hydrologically connected water bodies, and to quantify the length of rivers and streams downstream of road runoff. We used Adirondack Park Agency (APA) land use classification shapefiles to determine the estimated lands and waters impacted by road runoff in the Park’s forest preserves. Our methods are briefly described here, a more detailed description is provided in the Appendix.

Fig. 2
figure 2

Flow chart of TauDEM hydrologic modeling and ArcGIS processes used to delineate surface flow downslope of paved roads and to estimate the land area, surface water area, and total stream length impacted by road runoff from paved roads

Full size image

Producing a hydrologically relevant surface

We obtained 202 USGS 10 m New York State 7.5-min Digital Elevation Models (DEMs) from the Cornell University Geospatial Information Repository (cugir.mannlib.cornell.edu). These DEMs were mosaicked; then, sinks were removed by raising their elevation to the lowest pour point using the TauDEM 5.1.1 Pit Remove tool (TauDEM Toolbox 5.1.1, www.hydrology.usu.edu/taudem). Pits often occur due to deficiencies in DEM production (Jenson and Domingue 1988). A D-infinity flow direction grid was derived from the pit removed DEM using the TauDEM D-infinity flow direction algorithm (Tarboton 1997).

Producing road disturbance grids

Road disturbance grids indicate the spatial location of road runoff inputs in the flow delineation process. Road polyline shapefiles were created by extracting road polylines from NYSDOT Local Highway Inventory datasets by three separate road categories: state and federal (SF), county (C), and local (L) roads (apa.ny.gov/gis). The ESRI “buffer” tool was used to create 100 m road spray influence shapefiles for each category, including a combined dataset of all paved roads (SFCL). The buffer added 100 m to each side of the road. These eight shapefiles (SF, C, L, and SFCL and 100 m SF, 100 m C, 100 m L, and 100 m SFCL) were converted into 10 m cell size raster datasets using the ESRI “polyline to new raster” tool and the “polygon to new raster” tool. Field experiments have shown vehicular spray, and wind can transport pollutants hundreds of meters from the road (e.g., Zechmeister et al. 2005; Bernhardt-Roemermann et al. 2006), so the 100 m buffer was selected as a conservative estimate of road spray.

Producing input weights

An effective runoff grid and a first-order decay multiplier grid were required model parameters. Effective runoff grids represent the precipitation over an area. For simple contributing area delineations such as this, the weighting field was a constant raster set to 1 (Tarboton et al. 2009). The decay multiplier grid value was an exponential decay factor. The decay factor for road runoff was set to a constant raster of 1; this value was arbitrary for we were only interested in the spatial extent of runoff, not any quantitative accumulation values.

Delineating overland flow grids

TauDEM’s “D-Infinity concentration limited accumulation” tool was used to delineate substance accumulation downhill of the eight disturbance indicator grids, SF, C, L, and SFCL and the 100 m buffered SF, C, L, and SFCL.

Quantifying impacted land area

Cells with a value greater than zero represented overland flow. The >0 cells were extracted using the ArcGIS “extract by attribute” tool. To exclude overland flow cells that traversed open water areas, extracted cells were multiplied by a park wide open water raster dataset (cell values: open water = NoData, land = 1) using the ArcGIS “times” tool.

Identifying simultaneous impacts of multiple road categories

Lands impacted by multiple road sources were found by multiplying the overland flow rasters by category using the ArcGIS “times” tool (SF × C, SF × L, SF × C × L, et cetera). Lands not impacted by the multiple roads in question resulted in a NoData output value.

Identifying impacted lakes and ponds

The original eight overland flow grid outputs (those without open water cells removed) were converted to polygons using the ArcGIS “raster to polygon” tool. ArcGIS’s “select by location” tool was used to select lakes and ponds from the NYS area hydrography shapefile that intersected with the overland flow polygons. Intersected lakes and ponds were extracted according to the road category they intersected with. These intersected lakes and ponds were hydrologically connected to road runoff based upon the DEM-derived flow direction.

Calculating downstream length of impacted streams

The converted polygons of the eight overland flow categories were intersected with linear hydrography polylines from a NYS hydrography dataset of rivers and streams (gis.ny.gov) using the ArcGIS “intersect” tool. Point shapefiles were created and then converted to raster datasets using the ArcGIS “point to raster” tool and were treated as disturbance indicator grid parameters for the TauDEM “concentration limited accumulation” tool. The eight subsequent flow grids were then converted to polylines. The eight polylines shape files (SF, C, L, and SFCL and 100 m SF, C, L, and SFCL impacted streams) represented impacted stream channels.

Quantifying impacted lands and waters in APA forest preserve lands

The ArcGIS “zonal statistics as table” tool was used to categorize total impacted terrestrial area for each of the 15 APA land use types in the Adirondack Park Land Use and Development Plan dataset (apa.ny.gov/gis).

Results

A total of 156,561 ha of land area was estimated to be impacted by road runoff, which constituted 6.5 % of the land area of the Park (Table 2). Of this total, runoff from state and federal roads was estimated to impact 36,991 ha of land area, while runoff from county and local roads was estimated to impact 44,182 and 96,550 ha of land area, respectively. The total land area estimated to be impacted by road runoff increased to 254,224 ha when including the 100 m buffer on all roads, which constituted 11 % of the land area of the Park. The relative contribution to impacted land area from state and federal, county, and local roads was the same with and without the 100 m buffer and was proportionate to the length of each road type in the Park (e.g., local roads represented 54 % of total road length and 61 % of the impacted land area). An example of the relationship between the three road types and land area impacted by road runoff is provided for a 5000-ha area in the vicinity of the Village of Saranac Lake within the Park (Fig. 3). In this example, runoff from state and federal roads impacted 6 % of the land area (Fig. 3a), runoff from county roads impacted 5 % of the land area (Fig. 3b), runoff from local roads impacted 11 % of the land area (Fig. 3c), and all three road types together impacted 18 % of the land area (Fig. 3d). The close association between the road network and impacted land and surface waters is also clearly seen.

Table 2 Land and surface water area, number of lakes, and stream length impacted by surface runoff from paved roads with and without a 100 m buffer in the Adirondack Park, New York
Full size table
Fig. 3
figure 3

Estimated land area impacted by road runoff from state and federal (a), county (b), local (c), and the total paved road networks (d) in a 5000-ha area centrally located within the Adirondack Park, New York. Location is indicated by the star symbol on map inserted into Fig. 3a

Full size image

We found a total of 1787 ha of land to be impacted by state, federal, county, and local roads simultaneously (Table 3). This value increased to 4042 ha when including the 100 m buffer. State, federal, and county roads simultaneously impacted 3255 ha without the buffer and 6980 ha with the buffer. State, federal, and local roads simultaneously impacted 8378 ha without the buffer and 19,009 ha with the buffer. County and local roads simultaneously impacted 11,228 ha without the buffer and 20,837 ha with the buffer.

Table 3 Land and surface water area, number of lakes, and stream length impacted by surface runoff from multiple road categories with and without a 100 m buffer in the Adirondack Park, New York
Full size table

A total of 77,860 ha of surface waters was estimated to be impacted by road runoff, which constituted 76 % of the surface water area of the Park (Table 2, Fig. 4d). Note that two large lakes, Great Sacandaga Reservoir and Lake George, constitute 26 % of this total. Including all lakes, runoff from state and federal roads was estimated to impact 62,656 ha (61 %, Fig. 4a) of surface water area, while runoff from county and local roads was estimated to impact 62,218 ha (61 %, Fig. 4b) and 72,982 ha (71 %, Fig. 4c) of surface water area, respectively. The total surface water area estimated to be impacted by road runoff increased slightly to 78,792 ha when including the 100 m buffer on all roads, which constituted 77 % of the surface water area of the Park. There were a total of 820 waters impacted by road runoff, with 326 waters impacted by state and federal roads, 299 waters impacted by county roads, and 644 waters impacted by local roads (Table 2). The relative contribution to impacted surface water area from state and federal, county, and local roads was the same with and without the 100 m buffer.

Fig. 4
figure 4

Estimated surface waters impacted by road runoff from state and federal (a), county (b), local (c), and the total paved road networks (d) within the Adirondack Park, New York

Full size image

We found a total of 59,893 ha of surface water area to be impacted by the four road types (state, federal, county, and local roads) simultaneously, with this value increasing slightly when including the 100 m buffer (Table 3). The other multiple road categories had similar areas of surface waters impacted. The four road types impacted 184 lakes and 922 km of stream length simultaneously, with these values increasing to 194 lakes and 1222 km of stream length when including the 100 m buffer. The number of lakes and stream length impacted by the other multiple road categories was more variable compared to surface water area; for example, county and local roads simultaneously impacted over twice the stream length (2056 km) compared to the four road types together (922 km).

A total of 5934 km of stream length was estimated to be impacted by road runoff, which constituted 45 % of stream length in the Park (Table 2, Fig. 5d). Of this total, runoff from state and federal roads was estimated to impact 2795 km (21 %, Fig. 5a) of stream length, while runoff from county and local roads was estimated to impact 2859 km (22 %, Fig. 5b) and 4915 km (37 %, Fig. 5c) of stream length, respectively. The total stream length estimated to be impacted by road runoff increased to 6830 km (a 15 % increase) when including the 100 m buffer on all roads, which constituted 52 % of stream length the Park.

Fig. 5
figure 5

Estimated streams impacted by road runoff from state and federal (a), county (b), local (c), and the total paved road networks (d) within the Adirondack Park, New York

Full size image

Of the total 156,561 ha of land area estimated to be impacted by road runoff, 20,474 ha (13 %) was in the forest preserve, including the 100 m buffer which increased the impacted land area to 34,889 ha (Table 4). Of the total 78,153 ha of surface water area estimated to be impacted by road runoff, 67,496 ha (86 %) was in the forest preserve. Of the total 820 lakes estimated to be impacted by road runoff, 334 (40 %) were in the forest preserve. Of the total 5934 km of stream length estimated to be impacted by road runoff, 1152 km (19 %) was in the forest preserve. Within the forest preserve, 12 % of the impacted land area and 26 % of the impacted surface water area were located in lands classified as Wilderness, which are the most protected lands in the Park.

Table 4 Land and surface water area, number of lakes, and stream length impacted by surface runoff from paved roads with and without a 100 m buffer located in state-owned forest preserve lands within the Adirondack Park, New York
Full size table

Discussion

Impacted lands and waters

Our results show a high degree of connectivity between paved road networks and lands and waters within the Park. Due to this high connectivity and the basic hydrologic connectivity between water and its watershed (Hynes 1975; Vannote et al. 1980), road runoff pollutants may flow through substantial land and surface water area in the Park. This area may be significantly greater than what we estimated with our 100 m buffer, as field experiments have shown vehicular spray and wind can transport pollutants hundreds of meters from the road (Zechmeister et al. 2005; Bernhardt-Roemermann et al. 2006); thus, our estimates are likely conservative. The results further show that local roads impact substantially more land area than state and federal roads; however, the impact of state and federal roads may be more intense. For example, road salt loading is largest for state and federal roads in the Park (Kelting and Laxson 2010) and the loading of heavy metals and PAH pollutants has been shown to increase with road use intensity, giving evidence to a disproportionate impact of high-use roads (Klimaszewska et al. 2007). The impacts of road runoff may be compounded in areas hydrologically connected to multiple road categories, which represents a challenge to monitoring and management as the road categories are managed in different ways. While state and federal roads are all salted, county, town, and local roads in the Park receive a variety of winter road management treatments from salting to just plowing (Kelting et al. 2012). Modeling flow downslope of high-use roads will identify high-risk land areas for effective monitoring regimes.

A substantial area of land downslope of roads may be at risk from decreasing soil fertility and increasing toxins. Road salt (NaCl) is widely used on roads throughout the region for snow and ice control, for example the New York State Department of Transportation applies 28 t of road salt per kilometer of state and federal road annually (NYSDOT 2006). Increased Na concentrations in road runoff have similar base cation depleting effects as acid deposition (Norrström and Bergstedt 2001) and may become a more relevant stressor as public policy has greatly reduced levels of acid deposition in the northeast (Waller et al. 2012). Given that 67,496 ha of land may be impacted by runoff from 1965 km of state and federal roads, we can assume that this land area is at risk of base cation depletion and may also be experiencing the toxic effects of other associated road pollutants such as heavy metals and PAHs. Road salt discharge compounded by the northeasterly gradient of acid deposition in the Park (Ito et al. 2002) can also have implications for soil fertility (Gałuszka et al. 2011), tree physiology (Fan et al. 2014), and forest community dynamics (McEathron et al. 2013).

Aquatic ecosystems and groundwater resources hydrologically connected to paved roads may also be at risk from the toxic effects of road runoff pollutants. Regional salinization of lakes in the Park has already been documented by Kelting et al. (2012), who found state and federal road densities explained 86 % of the variation in both Na and Cl concentrations in lakes. Similar relationships between road density and salt concentration in surface waters were reported for studies in New Hampshire (Daley et al. 2009) and Rhode Island (Nimiroski and Waldron 2002). Now, through this new work, we have estimated that 76 % of surface water area and at least 45 % of stream length (dependent on buffer width) in the Park may be receiving road runoff pollutants such as salt. Salt is one of the most widely studied road runoff pollutants, and the effects of salinization on freshwater ecosystems are well documented; for example, studies have reported reduced aeration and water circulation at depth (Fay and Shi 2012), decreased spotted salamander (Ambystoma maculatum) and wood frog (Rana sylvatica) survival (Karraker et al. 2008), shifts in community structure (Collins and Russel 2009), reduced copepod density and changes in algal resources (Meter et al. 2011), and decreased productivity at all trophic levels in model freshwater communities (Dalinsky et al. 2014). A significant amount of road salt also enters groundwater where it can accumulate over years and exceed thresholds for potable water and aquatic organisms (Perera et al. 2012).

Given the large regional extent of road runoff and the potential for negative ecological effects at a large scale, the effects of this stressor should be more deliberately evaluated, particularly with its potential to interact with other regional stressors. Road salt, and other runoff pollutants, may confound the recovery of acidified lakes, complicating the assessment of lake management and policies on air pollution (Rosfjord et al. 2007). Jensen et al. (2014) found the biological recovery of an acidified lake receiving road salt discharge lagged behind the biological recovery of a similarly acidified lake not receiving road salt discharge. Thus, the application of road salt may create a mosaic of varying levels of lake recovery from acid deposition in the Park. Additionally, climate change may exacerbate the effects of road runoff pollutants on aquatic ecosystems (Schiedek et al. 2007). Warming spring temperatures break the diapause of aquatic invertebrates (Goddeeris et al. 2001) which corresponds temporally with high stream salt loads in snow melt (Oberts 1994). Also considering that the variability and intensity of spring storms are projected to increase in the northeastern USA (Hayhoe et al. 2007), the likelihood of physiologically active aquatic invertebrates being exposed to road salt discharge and other pollutants must increase (Silver et al. 2009). Our method for identifying areas receiving road runoff is vital to understanding the extent to which these multiple stressors of water quality interact across a large geographic area.

We can conceptualize road runoff from the road network as a geographically extensive stressor with spatially explicit inputs, and from this perspective, the ecological implications of road runoff are largely unknown. Acid deposition and climate change in the northeast are characterized at the regional scale by a smooth distribution of continuous stress inputs over an extensive geographic area, or in other words, the stress inputs of acid deposition and climate change do not vary much from hectare to hectare and thus are monitored at coarse scales (Ito et al. 2002; Hayhoe et al. 2007). Conversely, eutrophication is characterized at the watershed scale by discontinuous, but spatially explicit stress inputs (Agha et al. 2012) and therefore monitored at finer scales. Road runoff shares characteristics of each of the abovementioned stressors. Total runoff from a road network spans the entire region similar to that of acid deposition and climate change, but the stress inputs occur in spatially explicit distributions similar to eutrophication, creating a mosaic of environmental stress defined by topography and road position which is never the same from hectare to hectare.

Application of methods

Our methods create a spatially defined sampling unit that provides the framework for regional monitoring of road runoff as well as provides an estimate of aquatic and terrestrial ecosystems that may benefit from reductions in road salt applications (McDonald 2002; Kilgour et al. 2014). As mentioned above, road runoff pollutants often share similar environmental impacts as other regional stressors and often act synergistically, enhancing their effects on ecosystem function (Palmer and Yan 2013). A solid understanding of the spatial extent of road runoff stress allows researchers to focus regional monitoring efforts as well as develop study designs for experiments investigating the interactions of our significant regional stressors (Fancy et al. 2009).

Our results show that roads are hydrologically connected to a substantial area of land and water in the Park’s forest preserve, which includes four categories of land use: wilderness, primitive, canoe, and wild forest. The wilderness category is the most protected land and is managed to protect or restore its natural conditions (apa.ny.gov). Yet, we find that roads connect to over 5000 ha of land and nearly 70 % of surface water in the wilderness category. Road runoff may put at risk substantial forest preserve lands given the multitude of well-studied ecological impacts of runoff pollutants. Our methods may direct management to use alternatives to road salt in these sensitive areas (Jackson and Jobbagy 2005).

There is no New York State-wide monitoring program that chooses monitoring sites based on a single, spatially sound framework. The selection of monitoring sites in the state is predominately based on voluntary cooperation of lake associations (New York DEC 2009). New York State primarily utilizes the Citizen Statewide Lake Assessment Program (CSLAP; www.dec.ny.gov) to produce data for US Environmental Protection Agency Clean Water Act reports. CSLAP monitors sites on the basis of voluntary cooperation of lake associations, although the in-lake sampling protocols used adhere to the literature (Kishbaugh 1988). The New York State Department of Conservation’s Rotating Integrated Basin Studies and Lake Classification and Inventory programs monitor specific water bodies, but in a rotation of the State’s 17 major basins and the methodologies of choosing sampling sites are based on random and targeted selections, but are not based upon a regional framework and may produce sampling bias (New York DEC 2009; Larsen et al. 2007). Few lakes are monitored by CSLAP in the Park, though two other programs monitor the Park’s water quality: the Adirondack Lake Assessment Program (www.adkwatershed.org), which monitors lakes based on voluntary cooperation similar to CSLAP, and the Adirondack Long Term Monitoring Program (www.adirondacklakessurvey.org), which monitors only minimally impacted water bodies. Given that road runoff is hydrologically connected to the majority of surface water area in the Park and that road salt represents a major external pollutant load in our region (Kelting and Laxson 2010), our methods may provide a model for a useful study design that can answer real questions regarding statewide water quality.

Efficacy of methods

Subresolution variability is an inherent issue with DEM-based hydrologic modeling (Woods 2006), but for identifying impacted lakes and streams in this study area, the 10 m USGS DEMs are of sufficient resolution to meet this channel-based objective in this topographically variable study area (McMaster 2002). The total impacted area values are more subject to resolution biases. Coarser resolutions produce higher accumulation area values (Yang and Chu 2013), possibly overestimating the area impacted by overland flow. Zhang and Montgomery (1994) analyzed the effects of DEM resolution on geomorphic and hydrologic process in a moderate to steep gradient landscape. They found that 10 m resolution DEMs such as those used in this study should be sufficient to model surface flow processes, while DEMs greater than 30 m resolution produce erroneous estimates of hill slope and runoff processes.

This study quantifies total land area downslope of roads impacted by road runoff, but does not differentiate between surface and subsurface flows within this land area. Surface flow is a combination of direct runoff of salt-laden melt waters through road drainage networks (Labadia and Buttle 1996) and Hortonian flow (de Lima and Singh 2002). Hortonian flow is an important process during spring snow melt when soils are saturated and when frozen soils prevent infiltration of melt waters during thaws (Laudon et al. 2004). Surface flow removes about 50 % of road salt annually, and the remainder accumulates and moves through soils and groundwater as subsurface flow (Meriano et al. 2009), though a greater percentage of road salt may enter subsurface flow in the Adirondacks, as the region is dominated by coarse texture sandy soils with high infiltration rates (Sullivan et al. 2006; Kelting et al. 2012). Groundwater salt concentration increases every year with road salting as only a portion of salt entering in subsurface flow will flush annually (Perera et al. 2012), resulting in higher salt concentrations in streams during summer when biological activity is highest and perhaps most vulnerable to negative effects of salt (Jackson and Jobbagy 2005; Corsi et al. 2015). Given the importance of subsurface flow to road runoff impacts, landscape level modeling estimates would be enhanced by partitioning surface and subsurface flow paths when the additional soil and geologic data necessary to make these estimates exists.

Conclusions

The high degree of hydrologic association between paved roads and the lands and waters of the Park strongly suggests that the environmental impacts of road runoff should be evaluated along with other regional stressors currently being studied. Being able to estimate the spatial extent of road runoff is important for designing monitoring programs that can explicitly monitor this stressor while also providing opportunities to understand the interaction of multiple environmental stressors. Our methods provide an efficient way to estimate the spatial extent of road runoff for large geographic areas.