A Preliminary Study on the Impact of Marcellus Shale Drilling on Headwater Streams


The vast deposits of natural gas in the Marcellus Shale of Pennsylvania promise a significant source of domestic energy that—for a fossil fuel—is relatively clean. However, hydrofracturing, the method for recovering the natural gas from this deep geologic formation, has generated a great deal of controversy and is considered by many to pose serious environmental threats to aquifers and surface waters.


Natural gas drilling rig in Lycoming County, Pennsylvania
Natural gas drilling rig in Lycoming County,
Pennsylvania. Photo by Ruhrfisch


Hydrofracturing involves the injection of huge volumes of water supplemented with additives that promote rock fracturing and the release of natural gas; the composition of these additives are proprietary, but many of the identified constituents are known to present health and/or environmental hazards. Flowback, the return of the water to the surface, is highest immediately following hydrofracturing but continues for the duration natural gas collection.

Regulatory oversight and the adoption of best practices can help reduce surface water contamination by flowback, but spills will continue to occur. One of the unresolved questions concerning the Marcellus Shale is the cumulative effects of large scale development on surface waters. How much drilling and extraction can be sustained without seriously degrading local waters?

This preliminary study, conducted by Frank W. Anderson, Patrick Center Staff Scientist, sought to address this question by examining the relationship of the intensity of natural gas drilling, as expressed by well density per watershed area, on local stream health. Mr. Anderson was a graduate student at the University of Pennsylvania's Department of Earth and Environmental Studies. He was advised by faculty member Dr. Fred Scatena and by Dr. Jerry Mead, Section Leaderr of the Watershed and Systems Ecology Section of the Patrick Center for Environmental Research at the Academy of Natural Sciences.

Study Sites


headwater stream in northeastern Pennsylvania


During July 2010, Mr. Anderson studied nine headwater streams located within, or adjacent to, Susquehanna County in northeastern Pennsylvania. All of the streams are tributaries of the North Branch of the Susquehanna River.

The sites were similar in watershed and physical stream characteristics, but differed in the density of natural gas wells in their catchments. Three of the sites are classified as high-density (0.75-2.38 wells/km2). Three other sites have lower well densities (0.39-0.61 wells/km2), and three reference sites have no drilling within their catchments (Table 1).

Table 1. Physical characteristics of the study sites
Site Well
% Forest
% Riparian
(See methods for explanation of units)
HD1 2.52 2.38 34% 75% 1.3/1.8 0.4 2.89
HD2 1.82 4.96 50% 87% 2.0/3.0 1.1 6.35
HD3 0.75 9.33 52% 79% 2.3/2.5 0.6 6.56
LD1 0.61 3.30 49% 84% 1.5/1.6 0.5 4.21
LD2 0.57 1.75 36% 82% 1.3/1.4 0.5 3.08
LD3 0.39 2.59 42% 88% 1.8/1.8 0.3 1.83
R1 0 1.84 53% 90% 1.4/3.1 0.3 2.68
R2 0 3.33 63% 88% 1.6/3.3 0.4 3.54
R3 0 5.63 62% 90% 2.5/3.5 1.3 6.80


Well densities (wells/km2) were determined using Google Maps, Arc GIS, and IDRISI. Natural gas well locations were determined by using map data from Fracktracker.org and PAGasLease.com and in most cases were confirmed with on-site observations. GIS was used to determine watershed area (km2), topography, and percent forest cover, while percent riparian canopy cover was measured with a spherical densiometer. Streambed substrates were assessed by performing a 100 point Wolman Pebble Count (D-50); each particle was also observed for embeddedness in fine substrate. Active channel and bankfull widths and stream depth (in meters) were recorded. Stream velocities (feet/second) were measured with a Mash McBriney flowmeter.

Macroinvertebrates communities were assessed following the EPA protocol for monitoring stream macroinvertebrates. Samples were taken from riffle zones using a kick-net, stored in 89% ethanol, and then identified to the family level. Three indices were used to assess community structure: family richness (the number of macroinvertebrate families); the percent of individuals represented by Ephemeropera, Plecoptera and Trichoptera (ETP); and Shannon Diversity. Aquatic amphibians were were quantified by performing 5 minute timed-searches in addition to collection obtained during kick-net sampling. In addition, algal presence was recorded during the Wolman Pebble Count. Specific conductance (µS/cm), total dissolved solids (TDS, mg/l), and pH were measured using a YSI 557.

A correlation matrix of stream quality indicators with physical characteristics was generated and paired t-tests were conducted to compare differences among high-density, low-density, and reference sites.


Data from this preliminary study demonstrate an association between increases in natural gas well density with decreases in water quality indicators (Table 2). Despite the small sample sizes involved, a number of statistically significant correlations emerged for bothboth natural gas well density and riparian canopy cover. Specific conductance and total dissolved solids were positively correlated with well density; greater well density led to higher levels of specific conductivity and total dissolved solids. HD sites showed on average a 60% increase in specific conductivity values compared to LD sites.

Table 2. Stream quality indicators for the study sites
Site Well
TDS Algae EPT Shannon
HD1 2.52 277 0.180 31% 13% 1.18 6
HD2 1.82 178 0.116 26% 34% 1.49 10
HD3 0.75 162 0.105 40% 42% 1.95 12
LD1 0.61 134 0.087 22% 41% 2.31 19
LD2 0.57 126 0.083 27% 42% 2.30 18
LD3 0.39 110 0.072 14% 40% 2.00 14
R1 0 131 0.085 27% 38% 2.28 13
R2 0 124 0.081 22% 46% 2.34 16
R3 0 110 0.085 19% 39% 2.20 14

Macroinvertebrates, such as mayfly nymphs, are widely regarded by biologists as effective water quality indicators. Macroinvertebrate indicators (% EPT, Shannon diversity, and family richness) were negatively correlated with well density indicating decreasing water quality with increasing well density (Table 3). Specific conductance, total dissolved solids, and algal presence were negatively correlated with riparian canopy coverage.

Table 3. Correlation matrix (r2) of selected physical characteristics and stream quality indicators. Significant correlations are in bold.
TDS Algal
EPT Shannon
(‡ = p > 0.05; † p > 0.1)
Watershed Area .01 .04 .52 .19 -.11 .18
Forest Cover -.523 -.481 -.17 .56 .48 .26
Riparian Canopy -.77‡ -.74‡ -.68‡ .59† .56 .40
Active Channel -.27 -.23 -.02 .27 .00 -.11
Bank Width -.22 -.17 -.02 .22 .13 -.18
Stream Velocity -.07 .03 -.02 .00 -.16 -.14
Well Density .68‡ .67‡ .23 -.59† -.86‡ -.66†

Paired T-tests revealed significant differences between High Well Density sites compared to Low Well Density sites and References (R) sites. LD and R sites were not found to be significantly different. Specific conductance and total dissolved solids were elevated in HD sites, but two indicators of macroinvertebrate community quality, Shannon Diversity and family richness, were depressed (Table 4). Testing indicated that assumptions were met to conduct t-tests.

Table 4. Paired T-tests comparing High Well Density (HD), Low Well Density (LD) and Reference (R) sites. Significant values are in bold.
TDS % Algal
% EPT Shannon
(Paired data are stream quality indicator means; ‡ = p > 0.05; † p > 0.1)
HD and LD 205|123† 0.134|0.081† 32|21 30|41 1.54|2.20 9|17‡
HD and R 205|122† 0.134|0.084† 32|23 30|41 1.54|2.28‡ 9|14†
LD and R 123|122 0.081|0.084 21|23 41|41 2.20|2.28 17|14


The results present significant correlations with respect to natural gas well density and riparian canopy coverage. The significant correlations involving the latter are unexpected given the narrow 15% range recorded from the nine stations (75-90%). Moreover, while substantial differences in riparian cover are known to influence macroinvertebrate and algal populations, they are not considered important factors in determining stream chemistry. Considering the small differences in the riparian zones and the more substantial differences in well density, correlations with riparian coverage may be the result of cross-correlations with natural gas drilling density.

On the other hand, significant relationships between natural gas well density and indicators of stream health were demonstrated in this study despite its preliminary nature and the small sample sizes involved. Increased well density is associated with elevated levels of chemical contaminants (specific conductance and total dissolved solids) and the degradation of macroinvertebrate community structure. Moreover, the negative impacts were only evident in sites with high drilling densities; there were no statistically discernable differences between sites in catchments with low drilling densities and those with none. This last finding suggests that there is an operational threshold of drilling intensity below which the impacts on surface waters are sustainable. Increasing well density increases the cumulative impacts of extraction as well as increases the probability of an environmentally damaging event like a blowout or large volume leak occurring in a given watershed.