CONNECTING GROUNDWATER INFLUXES WITH FISH SPECIES DIVERSITY IN AN URBANIZED WATERSHED

Luanne Y. Steffy1, Angela L. McGinty2,3,

Claire Welty2,4,5 and Susan S. Kilham1

1. Department of Bioscience and Biotechnology

Drexel University, Philadelphia, PA 19104

2. Department of Civil, Architectural and Environmental Engineering

Drexel University, Philadelphia, PA 19104

3. Now at: The Environmental Careers Organization, 30 Winter Street, Boston, MA


02108-4720

4. Now at: Center for Urban Environmental Research and Education and Department of Civil and Environmental Engineering, University of Maryland Baltimore County,

Baltimore, MD 21250

5. Corresponding Author, 410-455-1766,

17


Abstract

Valley Creek watershed is a small stream system that feeds the Schuylkill River near Philadelphia, Pennsylvania. The watershed is highly urbanized, including over 17% impervious surface cover (ISC) by area. Imperviousness in a watershed has been linked to fish community structure and integrity. Generally, above 10-12 % ISC there is marked decline in fish assemblages with fish being absent above 25 % ISC. This study quantifies the importance of groundwater in maintaining fish species diversity in sub-basins with over 30 % ISC. Valley Creek contains an atypical fish assemblage in that the majority of the fish are warmwater species, and the stream supports naturally reproducing brown trout, which were introduced and stocked from the early 1900s to 1985. Fish communities were quantified at thirteen stations throughout the watershed and Simpson’s species diversity index was calculated. One hundred and nine springs were located and their flow rates measured. A cross-covariance analysis between Simpson’s species diversity index and spring flow rates upstream of fish stations was performed to quantify the spatial correlation between these two variables. The correlation was found to be highest at lag distances up to about 400 m and drop off significantly beyond lag distances of about 800 m.

Key Terms

Ground Water Hydrology, Aquatic Ecosystems, Urban Watersheds, Springs, Fish Communities, Species Diversity

Background

Valley Creek is a 64 km2 (23 mi2) spring-fed urbanized stream system in southeastern Pennsylvania about 35 km (20 mi) northwest of Philadelphia. The watershed lies in the Piedmont physiographic province. Carbonate and crystalline formations are located in geographically distinct areas within the watershed boundary (Figure 1). Generally, crystalline formations are found in the north and south hills, encompassing nearly all headwater streams in the watershed. Carbonate rocks, specifically limestones and dolomites, comprise the largest portion of the geology within the watershed. These lie in the valley through which the main branches of both Valley Creek and Little Valley Creek flow before discharging to the Schuylkill River at Valley Forge National Historic Park (VFNHP).

The watershed, as defined by its topographic boundaries, overlies and is well connected to a productive system of fractured rock aquifers. The hydrogeology has been previously well characterized, as reported by Sloto (1990). The aquifer system is phreatic which is reflected by water levels that respond to evapotranspiration and recharge, rising in spring and dropping in late summer. Groundwater divides do not precisely coincide with surface water divides on the western and eastern ends of the watershed, with the southwest – northeast groundwater flow direction being dominated by the geologic structure of the Chester Valley carbonate rocks. Springs discharge where the groundwater potentiometric surface is at or above the land surface. Springs occur in and above the stream’s channel, and in times of low precipitation, Valley Creek relies largely on springs to supply base flow. One hundred and seventy two springs were found within this watershed, one hundred and nine of which could be sampled (Figure 1). Several portions of Valley Creek do not maintain a continuous base flow and the streambeds dry out intermittently. The reaches with no surface flow lie within carbonate rock regions where the stream is flowing only in the subsurface. Consequently, the influence of the geology can contribute to the patchy pattern of fish distribution in Valley Creek in some areas, but does not explain varying fish diversity at adjacent stations having continuous flow in between.

Fish assemblages in non-urbanized streams typically follow a continuous pattern of distribution from headwater reaches to mouth (Vannote et al., 1980). As stream width, depth and habitat complexity increase downstream, fish species diversity and richness tend to increase as well. In urban streams there can be a patchier pattern of fish distribution related to local land use and water quality than is normally found in more pristine systems. In urban systems, species diversity for the watershed as a whole can be much higher than at any individual sampling location and community similarity between upstream stations is often low (Steffy, 2003). Fish that have strict temperature and substrate requirements for survival and reproduction such as brown trout (Salmo trutta) are typically not present in urban streams (Spotila et al., 1979; Carlander, 1969). Brown trout were stocked in Valley Creek beginning in the early 1900s to promote recreational fishing in the area. Although stocking was halted in 1985, there remains a naturally reproducing population of brown trout. Before the intense pressures of urbanization in the watershed, Valley Creek was a high-quality, spring-fed coldwater stream that supported the naturally reproducing brown trout population. However, because of the high amount of degradation which has taken place in the stream as a direct result of urbanization and land use changes, present day Valley Creek fish assemblages more closely resemble a warmwater stream system. It is now dominated by warmwater, eurythermal species. This is typical of highly urbanized, highly degraded coldwater streams (Wang 2003). The population of brown trout continues to decline owing to increased urbanization, land use changes, and increased stream temperatures (Kemp and Spotila, 1997; Steffy, 2003). Some of the eurythermal species include white sucker, creek chub, northern hognose, blacknose dace, bluegill, and pumpkinseed.

Impervious surface cover (ISC) can be used as a measure of urban land use and there is a definitive link between fish assemblages and ISC (Schueler and Galli, 1992). Increased imperviousness often results in increased stream water temperatures (Pluhowski, 1970). During storm events, surface water run-off from impervious surfaces (i.e., rooftops and parking lots) is generally warmer and accounts for much of the flow in the stream. Base-flow stream temperatures can be elevated from the absence of riparian canopy cover, which is often associated with developed areas. Increased stream temperatures and high annual temperature fluctuation have a negative impact on fish communities, particularly for fish that thrive at cooler water temperatures.

Valley Creek watershed has greater than 17 % ISC by area. Generally, it has been observed that between 10-12 % ISC there is a decline is fish communities and above 25 % ISC fish are usually absent (Paul and Meyer, 2001). More specifically, in Maryland, fish diversity decreased dramatically in warmwater streams having 12-15% ISC, with fish being absent above 30-50% impervious area (Klein 1979). At another site in Maryland, fish diversity decreased above 10-12% impervious area (Schueler and Galli 1992). Increases in ISC also affect species richness. Schueler and Galli (1992) report a significant reduction in species richness above 12% ISC. In Wisconsin, fish coldwater index of biotic integrity (IBI) decreased rapidly at 10% ISC (Wang et al. 1997) and 8-12% connected ISC resulted in major changes in stream condition (Wang et al. 2001). In Canada, fish IBI decreased sharply above 10% ISC but streams with high riparian cover were less affected (Steedman 1988).

Knowing and understanding land use patterns is helpful in interpreting fish community data. Valley Creek has an unusual pattern of urbanization in that traditionally the most intense urbanization has been in the western, upstream end of the watershed. The most downstream three kilometers of Valley Creek lie in VFNHP and are relatively undeveloped. Upstream of VFNHP, the land use is primarily residential, including some dense housing areas and some large-lot residential properties. Even further upstream, the watershed contains most of the heavy industrial land uses along with dense commercial properties and less residential area.

Very little research has been published documenting the connections between the quantity and quality of groundwater flowing into a stream and the distribution and survival of local fish assemblages. Traditionally, ecologists and hydrologists do not work closely together and therefore the connection between stream base flow and fish assemblages is not well documented. Some older studies indicating a positive relationship between groundwater and trout populations were carried out (Benson 1953, Boussu 1954); however, those studies did not quantify fish communities and the influence of groundwater. This paper quantitatively connects groundwater influxes to fish species diversity by using a standard geostatistical method, cross-covariance analysis. The cross-covariance reveals the spatial correlation between fish species diversity and spring flow rate as the average distance between a fish station and an upstream spring increases.

Methods

One hundred nine springs were sampled for temperature and their flow rates were measured in one sampling round. Most of the springs found in this watershed had an opening less than four inches in diameter. The springs were most often located within a few feet of the stream channel and could be seen flowing directly into the stream. Temperature was measured when the flow rates were taken in the field with a Horiba Conductivity Meter ES-12. The flows of the springs were measured at their source with a bucket, stopwatch, and graduated cylinder (McGinty, 2003). Thirteen stations were sampled for fish assemblage and species composition. Station one and station two were located in crystalline formations, while stations three through thirteen were located in carbonate rock formations (Figure 1).

Fish were sampled by electroshock fishing (Smith-Root 110 V AC backpack electroshocker at 70 Hz every two ms) and dip-netting at each of thirteen stations in Valley Creek watershed in July of 2001. Stations were intentionally spread throughout the entire watershed (Figure 1) in order to include a variety of land uses and geology formations. Three passes were made through each sampling reach (mean length 35 m) which consisted of two pool and two riffle reaches. Captured fish were placed in buckets filled with stream water until all electroshocking was completed at a station. Fish were identified, weighed and measured in the field and a majority were returned to the stream. The remaining fish were taken to Drexel University laboratories for other analyses. Stream temperature was measured seasonally for two years at all fifteen station locations.

Fish species diversity was calculated using Simpson's species diversity index (Odum, 1971) at each of the thirteen stations (Steffy, 2003). Using a diversity index to quantify fish assemblages and characterize a particular stream reach allows for better comparison than abundance or species richness alone. Two stations (13 and 14) were dry during the sampling period and thus had no fish and no diversity index value.

In order to quantitatively assess the influence of the location and flow rate of springs on fish diversity, the spatial cross-covariance function was calculated for all fish-station /spring pairs, for springs upstream of each fish station. In situations where there was no base flow, the springs upstream of the dried-out reaches were not included in the cross-covariance calculation. The purpose of this calculation was to test the hypothesis that the species diversity at a sampled fish station was generally higher when the station was downstream of a productive nearby spring. This qualitative observation had been made from comparing field notes on the spring attributes to the fish diversity data, and a more quantitative measure was desired. The shortest distance between a spring and closest fish sampling station ranged from 20 to 1600 meters, with an average of 600 m.

The cross-covariance function is a measure of spatial variability or continuity between two attributes separated by a known lag distance. Using common notation such as that of Deutsch and Journel (1998) or Goovaerts (1997), the cross covariance as a function of lag separation is defined as:

(1)

where N(h) is the number of data pairs in a lag class, xi is the value of the attribute at the tail of the pair, yi is the value of an attribute at the head of the pair, h is the average separation distance or lag between the head and the tail, m-h = is the mean of the tail attribute values, and m+h = is the mean of the head attribute values.

In the case presented here, the coordinate of the “tail” of the lag separation vector is defined as the fish station location, having Simpson’s diversity index as its attribute; the “head” coordinate of the separation vector is defined as the spring location, having spring flow rate (L/s) as its attribute. The distances between each fish station and all springs upstream of each station were measured along the stream using Arcview 8.1. Because Simpson’s diversity index is dimensionless and the flow rate is reported in L/s, the resulting units on the cross covariance function are L/s.
Impervious surface cover data for the entire watershed were derived from a geographic information system (GIS) impervious layer (provided by Cahill and Associates) from 1995 and updated using aerial photographs from 2000. The sub-basins for each sampling station were derived using the HEC-GeoHMS 1.1 (U. S. Army Corps of Engineers, 2000) model running in conjunction with ArcView 3.2. The input data were from the USGS Digital Elevation Model at a 30m resolution (Clay Emerson, pers. comm.). Species diversity was then plotted as a function of percent impervious surface cover for each sub-basin to show the general trend of decreasing species diversity with increasing ISC (Figure 2).

Results

The average temperature for all of the springs did not vary much between the crystalline and carbonate rock types. The crystalline formation showed an average temperature of 10.4 °C and the carbonate showed the highest average temperature at 10.6 °C. The flow rates did vary largely between the two rock types. The mean flow rate for the crystalline rocks was 0.22 liters/sec (L/s) (3.45 gallons per minutes (gpm)). A larger average flow rate occurred in the carbonate rocks and was found to be 0.57 L/s (9.0 gpm). Temperature and flow rates of the springs were therefore both highest in the carbonate rock areas (McGinty, 2003).

The stream temperature in Valley Creek watershed is highly seasonally variable. The mean summer temperature is 20 ºC while the mean winter temperature is 5 ºC. Some in-stream temperatures were found to be affected by groundwater inputs. Station locations near large springs were 5-10 ºC cooler during the summer and had a smaller annual fluctuation in stream temperature. For example, station 10, located 200 m downstream of multiple springs, had a mean summer temperature of 13 ºC with a 5 ºC difference between summer and winter stream temperature. Other stations much further downstream (on the order of 1000 - 2000 m) of significant spring flow inputs (e.g. #1 and #2) fluctuated up to 20 ºC throughout the year (Steffy, 2003). Figure 1 contains station location information.