The MVP is a natural gas transmission pipeline which originates in northwestern West Virginia and ends in south-central Virginia (fig. 1). The purpose of the pipeline was to connect West Virginia’s Marcellus Shale region near Wetzel County, West Virginia to the existing Transcontinental Gas Pipe Line Company, LLC gas transmission pipeline in Pittsylvania County, Virginia (Parfomak and Vann, 2022). The 42-inch diameter pipeline spans approximately 303 miles and crosses the Appalachian Mountains. The pipeline was initially proposed in 2014, and construction began on January 1, 2018 (Parfomak and Vann, 2024); however, construction was suspended numerous times because of permitting issues in federal court. Construction was completed on June 10, 2024, and the MVP was approved to begin commercial service on June 11, 2024 (Parfomak and Vann, 2024).

The study sites are in the Appalachian Basin and the Valley and Ridge and Blue Ridge Physiographic Provinces, characterized by varying topography, including areas of high slope and highly erodible soils (Lévesque and Dubé, 2007; New River Geographics, LLC, 2017; Betcher and others, 2019; U.S. Geological Survey, 2025a). The karst topography in this region is characterized by vast caves and other subsurface water conduits and can pose an additional risk during construction activities (Lévesque and Dubé, 2007; Zerga, 2024).

The climate in this region is subtropical, with warm, humid summers and cold, often snowy winters, and an annual average temperature of 14.2 degrees Celsius (°C; H Brothers, Inc., 2025). The City of Roanoke, Virginia, (near the center of the study area) receives an average annual rainfall of 42.8 inches (in.) and an average snowfall of 14.7 in. (Coats and Jackson, 2020; H Brothers, Inc., 2025). The study area is predominantly forested, featuring deciduous, coniferous, and mixed stands composed of northern hardwood, cove hardwood, and various mixed deciduous species (Turner and others, 2003; Virginia Department of Conservation and Recreation, 2021). Mature trees in this area typically reach about 100 feet in height, creating a dense canopy over the mountain range. (Coats and Jackson, 2020).

The MVP intersects approximately 139 streams and 61 wetlands across Virginia and West Virginia, primarily in rural and previously undisturbed areas (Betcher and others, 2019; Downstream Strategies and West Virginia Rivers, 2024). Various stream crossing methods were used during the construction, which are categorized in this report as either trenched or trenchless crossings. The MVP selected the stream crossing methods on the basis of the characteristics of the landscape, including stream width, the steepness of the adjacent slopes, local geology, available space for construction activities, and proximity to residential areas, roads, or environmental resources (Tetra Tech, 2021).

The trenched crossing method used was a conventional dry-ditch open-cut method (Tetra Tech, 2021). Trenched crossing methods isolate the construction site by diverting streamflow using a temporary coffer dam, allowing work to proceed in dry or nearly dry conditions (Castro and others, 2015; Reid and others, 2002). The workflow includes excavating the trench, installing the pipe in the trench, and backfilling the trench. Trenched methods can disrupt sediment during installation and after removal of the dam (Lévesque and Dubé, 2007; Reid and others, 2002).

Trenchless methods install the pipeline by drilling boreholes beneath the stream, completing the installation without disturbing the surface or streambed (Castro and others, 2015; Yan and others, 2018; New Jersey Department of Environmental Protection, 2021). The pipeline is typically installed as an arc several feet below the streambed, with suitable pipeline entry and exit locations potentially outside of the streambank margins (Castro and others, 2015). Entry and exit points are initially drilled as a pilot hole by a drill rig, then drilling fluids are used to transport drill cuttings to the surface, maintain the borehole stability, and cool the drill bit (Yan and others, 2018). Trenchless methods for stream crossings can avoid the aquatic risks by boring outside the streambanks and underneath the stream, which is particularly used in sensitive or urban areas (Castro and others, 2015). However, even trenchless methods have associated risks, including groundwater infiltration and disruption of underground stream networks.

The pipeline required a temporary and permanent right-of-way (ROW) to accommodate construction and maintenance vehicles, equipment, and material stockpiles (Tetra Tech, 2021). In the upland areas, a 125-foot-wide temporary ROW for construction and a 50-foot-wide permanent ROW for long-term maintenance were constructed (Tetra Tech, 2021). Within 50 feet of a stream crossing, the construction ROW narrows to 75 feet to preserve the riparian buffer (Tetra Tech, 2021).

After the sites were backfilled and hydrostatic tested, cleanup and restoration were completed. These activities include restoring the grade cuts to pre-construction topography and ground cover to regrow native plants, minimize erosion, and stabilize stream banks (Tetra Tech, 2021).

The USGS monitored six perennial streams in cooperation with the VDEQ along the MVP. These sites were selected on the basis of the presence of threatened or endangered species, proximity to- or designation as- Virginia Tier III waters (Exceptional State Waters), high ecological integrity based on previous aquatic community, VDEQ habitat and water-quality surveys, public water supplies, upland construction areas, site access, and suitable waterflow (Virginia State Water Control Board, 2024). The monitoring sites varied in watershed area, elevation, and proximity to the pipeline crossing construction (fig. 2; table 1). Bottom Creek was crossed twice and monitored upstream and downstream at both crossing sites. In this report, upper Bottom Creek refers to the location in the headwaters near the top of the watershed, and lower Bottom Creek is located approximately 2.5 miles downstream.

The three smallest watersheds of the study area, Little Stony Creek, upper Bottom Creek, and lower Bottom Creek, all contain parts designated as Exceptional State Waters and are noted for wild trout populations, diverse benthic macroinvertebrate communities, and high-quality instream and riparian habitat conditions (Burton and Gerritsen, 2003; J. Hill, VDEQ—Blue Ridge Regional Office, oral commun., 2017; Virginia State Water Control Board, 2024). This designation identifies, maintains, and protects the waterbody from permanent or long-term degradation (Virginia State Water Control Board, 2024). Sinking Creek, the third largest watershed, is distinguished by high benthic macroinvertebrate diversity and karstic geology. Upper Bottom Creek was crossed using trenched methods, but Little Stony Creek, Sinking Creek, and lower Bottom Creek were crossed using trenchless methods.

The two largest watersheds, Blackwater River and Roanoke River, are prominent western Piedmont Physiographic Province streams and are the water sources for Smith Mountain Lake, a reservoir used for drinking water and recreation by thousands of residents and visitors each year. Although the Blackwater River has a smaller watershed, it includes two pipeline crossings within the study area: one on the tributary Maggodee Creek, upstream of the monitoring area, and another between the two monitoring stations on the Blackwater River. The Roanoke River is a vital public water supply for the City of Roanoke, Roanoke County, and the City of Salem in Virginia. Although the Roanoke River was crossed using trenchless methods and the Blackwater River was crossed using trenched methods, the pipeline at both locations was installed at a minimum depth of six feet below the streambed (Tetra Tech, 2021).

The purpose of this study is to assess whether water temperature and turbidity regimes were affected by pipeline stream-crossing construction and to evaluate the utility of real-time stream monitoring alert networks during pipeline construction. These objectives were addressed through the following questions:

The purpose of this report is to present the water temperature and turbidity data and anomalous conditions revealed during the study. The analysis herein is limited to answering those specific questions for the six monitored stream reaches. Potential construction-related anomalous conditions outside those specifically evaluated are possible but fall outside the scope of this report. Additionally, the study design and analysis address potential reach-specific effects only—potential overall or watershed-wide effects of pipeline construction are not evaluated from monitored data in this report.

The continuous water-quality and water-level monitoring network for the MVP was established at six pipeline stream crossings in 2017 (fig. 1). Each site had a paired monitoring design, with stations upstream and downstream from the pipeline stream crossing. Watershed areas ranged from less than 1 square mile (mi²) to 260 mi² (table 1). Paired monitoring stations were used to assess downstream changes in water temperature and turbidity relative to the upstream conditions. Water levels were measured at one station per site, except for along the Roanoke River, where water level measurements were taken at upstream USGS site 02054500 (table 1). Data were reported in near real-time to support and inform the real-time compliance alert application.

Data were analyzed to assess water temperature and turbidity conditions before, during, and after pipeline stream crossing (table 2). The “before” period includes all dates from the establishment of the monitoring station to the day prior to stream crossing (2,123 to 2,289 days). Because of intermittent construction caused by permitting issues, activities not associated with the stream crossing construction, such as tree clearing and ROW construction, are included in the before period. The “during” period includes the pipeline stream crossing construction and the immediate post-crossing restoration work (12 to 208 days). The “after” period is at least one-year post-stream crossing construction (366 to 595 days).

This report focuses solely on water temperature, turbidity, and water-level data for the entire period of record. All data were collected and quality assured in accordance with USGS Guidelines and standard procedures (Wagner and others, 2006; Sauer and Turnipseed, 2010). Continuous water-quality data were measured using YSI, Inc. (Yellow Springs, Ohio) 6-series Multiparameter Water Quality Sondes at 5-minute intervals for water temperature, specific conductance, pH, dissolved oxygen, and turbidity. Water temperature has a measurement precision of 0.01 °C and a measurement accuracy of plus or minus (±) 0.15 °C (YSI, 2012). The measurement precision for turbidity is 0.1 formazin nephelometric units (FNU) with an accuracy of ±2 percent of a measurement or 0.3 FNU, whichever is greater. Water levels were also measured at 5-minute intervals at the six monitored pipeline crossings using a TE Connectivity (Galway, Ireland) KSPI 500 transducer or OTT HydroMet (Kempten, Germany) PLS pressure transducer. Peak-stage verification was accomplished with additional water-level measurements with an Onset HOBO (Bourne, Massachusetts) water-level data logger. Turbidity measurements exceeding 1,000 FNU are outside the manufacturer’s stated measurement range and were therefore censored and treated as values greater than 1,000 FNU. This affected less than 0.15 percent of all measurements (app. 1, table 1.1). This small amount of right-censored turbidity data was inconsequential to the analytical methods used and described in the “Statistical Analysis of Water Temperature and Turbidity” subsection and did not affect the interpretation of potentially anomalously high upstream-to-downstream differences. All data in this report are publicly available on the USGS Water Data for the Nation database (WDFN; U.S. Geological Survey, 2025b).

The monitoring network was used by VDEQ as a real-time alert tool for regulators of the pipeline construction. From January 2018 through April 2025, water temperature, pH, dissolved oxygen, turbidity, and specific conductance were evaluated in real-time using an automated script for potential anomalous conditions. An alert provided by the automated data review system happened when water-quality measurements were below the minimum values or above maximum values set by the thresholds defined in table 3.

A field visit could lead to one of the following four response categories by VDEQ (J. Hill and E. Jones, VDEQ—Blue Ridge Regional Office, oral commun., 2017):

Water temperature and turbidity were analyzed by graphical inspection, descriptive statistics, magnitude and duration analysis, and time-matched paired-difference testing. At each site and for each period, bootstrapped resampling and confidence intervals were computed to test if the observed median of all time-matched paired differences between upstream and downstream was significantly different from zero. Water temperature and turbidity were each further analyzed by methods suited for each specific parameter in the “Continued Statistical Analysis of Water Temperature and Continued Statistical Analysis of Turbidity” subsection.

The raw data were initially plotted over time for graphical inspection and used to calculate the minimum, first quartile (Q1), median, third quartile (Q3), and maximum for each site, station, and period. An hourly median was computed to account for lag time between the upstream and downstream stations. The downstream hourly median data were plotted against the upstream data for visual inspection. The hourly median data were used to analyze the magnitude and duration of the time-matched paired differences to identify potential short- and long-term anomalous conditions for both parameters. If there were not a complete concurrent pair, the difference was not calculated, and the given date time was removed from analysis. The hourly values were not calculated and removed from the analysis if there were less than 80 percent of the data in a given hour.

The instrument’s uncertainty for water temperature is ±0.15 °C, but turbidity measurements have an uncertainty of ±0.3 FNU or 2 percent, whichever is greater (YSI, 2012). Difference calculations required accounting for the combined uncertainty from the upstream and downstream instruments and were computed as averages for each site, station, and period as in Joint Committee for Guides in Metrology (2020; eq. 1):1n∑i=1nxDSi2+xUSi2,(1)where

A bootstrapping of median differences approach was used to determine if the difference between the hourly median time-matched paired observations between downstream and upstream stations differed significantly from zero. A nonparametric bootstrapping of medians approach is well-suited for large, nonnormal, and asymmetrical datasets (Mudelsee and Alkio, 2007; Johnston and Faulkner, 2021), which do not meet the assumptions of other statistical tests for dependent samples. Additionally, Johnston and Faulkner (2021) reported that bootstrapping differenced medians resulted in fewer Type I errors than the rank-based Mann-Whitney-Wilcoxon test, even though the analysis involved testing two independent samples.

The bootstrapping of median differences test used in this analysis was a one-sample test applied to the dependent sample of time-matched paired differences between the upstream and downstream stations. The median of the observed distribution of differences was computed and used as the test statistic and then compared to 10,000 bootstrapped resampled distributions with replacement. Recent studies emphasized that calculating p-values for classic null hypothesis statistical tests when testing large sample sizes is not useful or appropriate because the tests are heavily influenced by and dependent on sample size (Lin and others, 2013; Krawczyk, 2015; Gómez-de-Mariscal and others, 2021). Alternatively, recent studies suggest evaluating confidence intervals directly (Lin and others, 2013; Balkin and Lenz, 2021; Gómez-de-Mariscal and others, 2021). Upper and lower 99-percent confidence intervals were calculated for each station and period from 10,000 bootstrapped resampling, with replacement. The confidence intervals were evaluated for the hourly median differences of time-matched pairs between the downstream and upstream monitoring stations.

Results from the bootstrapping median differences tests were evaluated according to the following criteria:(1)

If the test’s confidence interval overlapped zero, or the observed median of the time-matched paired differences was less than or equal to the combined uncertainty from instrument accuracy, then the observed median of the time-matched paired differences was not considered significantly different from zero; and

The upstream and downstream temperatures showed similar seasonal variation throughout the study period, following typical East Coast seasonal temperature patterns (fig. 3A). Across all paired sites, the hourly median temperatures ranged from −2 to 30.4 °C (table 4). January consistently had the coldest median temperatures at upstream and downstream sites throughout the study. The warmest median temperatures occurred during July for all sites, except for upper Bottom Creek, where peak median temperatures were recorded in August.

The hourly median water temperature differences between the upstream and downstream stations were used to visually observe thermal variations on an interannual and seasonal timescale for each of the six sites (fig. 3B). A negative hourly median temperature difference indicated that the upstream station had a higher hourly median temperature than the downstream station. A positive hourly median temperature difference indicated that the downstream station had a higher hourly median temperature than the upstream station.

Sinking Creek, upper Bottom Creek, and lower Bottom Creek demonstrated the most pronounced seasonal variation throughout the monitoring period. However, upper and lower Bottom Creek showed a different pattern from Sinking Creek. At the Bottom Creek sites, the downstream station’s hourly median temperature had warmer temperatures than the upstream station during the summer (fig. 3B). In contrast, the Bottom Creek upstream station’s hourly median temperatures were higher than those at the downstream station during summer (fig. 3B). The hourly median temperature differences between upstream and downstream stations on Little Stony Creek, Roanoke River, and Blackwater River had less seasonal and interannual variation.

To further explore these patterns, temperature differences were evaluated at all sites before, during, and after construction of pipeline stream crossing. Within each site and each period, the upstream and downstream stations generally had similar temperatures (fig. 4; table 4), although there were site-specific differences. At Sinking Creek, there were higher temperatures at the upstream station compared to the downstream station at warmer overall temperatures in the before and after periods (fig. 4B). Upper and lower Bottom Creek showed warmer downstream temperatures as the temperature increased during all three periods (fig. 4D, E). In the during stream crossing period, upper Bottom Creek notably diverged from the 1-to-1 line (that is, the line that shows when downstream temperature equivalent is to upstream temperature) (fig. 4D), with a higher temperature at the downstream station compared to the upstream station.

The finding that, in general, the upstream and downstream hourly water temperature differences were similar was supported by the bootstrapping of median differences analysis of the hourly median differences (table 5). The observed median differences ranged from the upstream warmer by 0.15 °C and the downstream warmer by 0.6 °C. Sixteen of the 18 tests had hourly median differences, less than the combined uncertainty from the instrumentation. Of these, three tests had confidence intervals that overlapped zero—lower Bottom Creek in the before and after period and upper Bottom Creek in the after period. Sinking Creek and upper Bottom Creek in the during period are the only two tests that have potential for a significant test result because the observed median of differences is greater than the combined instrument uncertainties and the confidence intervals do not overlap zero. For these two tests that yielded a significant result, the maximum median of differences is less than 1°C when uncertainty is accounted for, ranging from 0.24 to 0.66 °C at Sinking Creek and from 0.39 to 0.81 °C at upper Bottom Creek in the during period.

Across the six sites, 7,284 hourly median water temperature anomalies were detected out of the 766,504 total hourly median differences, representing approximately 0.95 percent of the data. Roanoke River, lower Bottom Creek, and Blackwater River had the highest percentage of data that detected anomalies, at 1.72 percent (n [number of observations]=2,171 of 126,450), 1.55 percent (n=2,010 of 129322), and 1.45 percent (n=1,794 of 123682) of the data, respectively. Anomalies were detected at Little Stony Creek in 0.89 percent (n=1,152 of 129,656) of the data. Two sites contained fewer than 100 anomalies detected in the time series and consist of about 0.06 percent of the data—Sinking Creek (n=80 of 128,984) and upper Bottom Creek (n=76 of 128,410).

The following three types of anomalies were detected: downstream-warmer, which accounted for 3,218 anomalies (44.2 percent), upstream-warmer, which accounted for 4,030 anomalies (55.3 percent), and same-temperature anomalies, which accounted for 36 anomalies (0.9 percent). Throughout the study period, the median duration of all anomaly types was 2 hours, and the median temperature difference was 0.73 °C, with the downstream station being warmer. This report further investigates the magnitude and duration of downstream-warmer and upstream-warmer anomalies.

At all monitored locations, the upstream and downstream site exhibited a similar thermal behavior within each evaluated period. The differences between the paired sites were minimal, with the range between the lower and upper 99-percent confidence intervals from 0 to 0.1 °C. Twelve of 18 tests had a difference in the upper and lower 99-percent confidence intervals of 0 °C. The median of the hourly median differences was also small, from −0.15 to 0.6 °C for all sites and periods. Of the 18 tests performed, only Sinking Creek in the during period and upper Bottom Creek in the during period were significant. At both sites in the during period, the medians of hourly median differences exceeded the combined uncertainty of the instrumentation, and the confidence interval did not include 0. Although these tests are statistically significant, temperature differences at both sites are less than 1 °C: 0.24–0.66 °C at Sinking Creek and 0.39–0.81 °C at upper Bottom Creek.

To distinguish any potential differences between the upstream and downstream temperatures, water temperature anomaly detection was used to remove the effects of seasonality and long-term changes to detect potential anomalous events. There were downstream-warmer and upstream-warmer anomalies across all study sites; however, these anomalies constituted approximately 1 percent of the total data collected. Most of the downstream-warmer and upstream-warmer anomalies were identified in the before period because of the length of the before period in comparison to the during and after. Sites that used the trenchless construction stream crossing method detected downstream-warmer anomalies in the during period, whereas the two sites using trenched stream crossing methods did not detect any anomalies; however, this disparity is more closely linked to the duration of the stream crossing period than the crossing technique itself. The trenched construction period lasted 12 days at Blackwater River and 21 days at upper Bottom Creek; however, the trenchless stream crossing period had a duration ranging from 141 to 208 days. Additionally, some of the sites showed a notable seasonality that was captured in the during period. Specifically, at lower Bottom Creek and the Roanoke River, downstream-warmer and upstream-warmer anomalies occurred in the late-spring to early-summer months.

The magnitude and duration of the downstream-warmer and upstream-warmer anomalies were low in magnitude and short in duration. The median duration across all anomaly types was 2 hours, and the median magnitude of 0.73±0.21 °C, with the downstream station warmer than the upstream. Little Stony Creek, Roanoke River, lower Bottom Creek, and Blackwater River yielded comparable ranges for upstream- and downstream-warmer anomalies, demonstrating consistent patterns across different periods. The magnitudes and duration of the anomalies in the during stream crossing period aligned closely to those of the before and after stream crossing periods.

The median duration of upstream-warmer and downstream-warmer anomalies was 2 hours, indicating that, in general, temperature anomalies in the streams did not persist for long durations. The longest downstream-warmer anomaly lasted 18 hours, while the longest upstream-warmer anomaly persisted for 134 hours. Although longer duration anomalies exist, 61 percent of the upstream-warmer anomalies (n=380 of 610) and downstream-warmer anomalies (n=441 of 722) were less than or equal to 2 hours. About 89 percent of the anomalies lasted 4 hours or less, including 649 of 722 upstream-warmer anomalies (90 percent) and 539 of 610 downstream-warmer anomalies (88 percent), suggesting that most of the anomalies were short in duration.

The median magnitude of the downstream-warmer anomalies was 0.74 °C, and the median magnitude of the upstream-warmer anomalies was 0.43 °C. Some sites had a consistent typical range in the upstream-warmer and downstream-warmer anomalies, such as Little Stony Creek, Roanoke River, and lower Bottom Creek (fig. 5). All sites showed similar temperature magnitudes throughout the before, during, and after periods. Most temperature differences between the upstream and downstream sites were small, with about 63 percent of downstream-warmer anomalies (n=385 of 610) and 68 percent of upstream-warmer anomalies (n=493 of 722) measuring less than or equal to 1 °C. Ninety percent of the downstream-warmer anomalies were less than or equal to 2 °C difference between upstream and downstream, and an even higher percentage, 97 percent of the upstream-warmer anomalies (n=702 of 722). During the 7-year study period, the magnitude of the temperature anomalies remained low; most detected anomalies were within a 2 °C difference.

One criterion applied in the real-time alert application included flagging instances when the temperature differences between upstream and downstream exceeded the natural temperature variability threshold, defined as 1–3 °C depending on the stream class (Virginia Administrative Code 9VAC25–260–50; Virginia State Water Control Board, 2023). None of the study sites recorded hourly temperature differences exceeding 3 °C, indicating that the temperature anomalies generally remained within the expected range for Virginia. The single temperature anomaly difference had a median temperature difference of exactly 3 °C. This anomaly occurred during the before period at lower Bottom Creek from May 23, 2023, at 12:00 to May 23, 2023, at 17:00. However, the small number of hourly median temperature anomalies with respect to the entire dataset, combined with the low magnitude and brief duration of anomaly events, suggests that the pipeline stream crossing had minimal consequences on the thermal regime downstream.

The anomaly detection method used in the analysis removes the effect of long-term trends in the water temperature but does not remove the effect of short-term changes in temperature. Temperature is a highly variable metric that is largely influenced by air temperature and environmental conditions, including channel geometry, water depth, tributary confluences, groundwater seeps, stream discharge levels, and microhabitat characteristics (Caissie, 2006; Siegel and Volk, 2019; Mejia and others, 2020; Gendaszek and Appel, 2022). Although air temperature was not directly measured in this study, stream water temperature increases about 0.6–0.8 °C for every 1 °C increase in air temperature (Morrill and others, 2005). However, higher elevations or groundwater-dependent watersheds are less sensitive to temperature changes (Siegel and Volk, 2019). For example, direct field observations indicated that Sinking Creek receives groundwater input downstream from the upstream monitoring station. Others have also observed groundwater inputs throughout the Sinking Creek watershed (Christensen and others, 2022). Groundwater discharge cools surface water during the summer (Gendaszek and Appel, 2022), contributing to more upstream-warmer temperature anomalies. As a result, the upstream station was warmer during the summer months (fig. 3). The confluence of tributaries could affect temperature differences at the upstream or downstream site. At the Blackwater River, two tributary confluences are near the Blackwater River upstream station (fig. 2F), which could affect conditions observed at the upstream or downstream location. The confluence upstream from the Blackwater River upstream site, Maggodee Creek, was also crossed by the MVP. Maggodee Creek, which has a watershed area of 45.5 mi² (U.S. Geological Survey, 2019) was crossed using trenched methods and overlapped construction on the Blackwater River for 10 days.

Upstream median turbidity ranged from 0.5 to 7.2 FNU across all sites and periods, and the downstream turbidity ranged from 0.6 to 8.3 FNU (table 6). Higher turbidity at upstream and downstream stations was typically associated with increased gage height throughout the study period and lower turbidity at low gage height, as expected (figs. 6, 1.1, 1.2, 1.3, 1.4, 1.5 and 1.6). The Blackwater River had the highest magnitude turbidity values across all summary statistics; upper Bottom Creek had the lowest (table 5). At all sites and during all periods, the minimum, Q1, median, and Q3 are close together in magnitude compared to their maximums.

Little Stony Creek, Roanoke River, upper Bottom Creek, and Blackwater River showed general agreement in the time-series data throughout the study period. Sinking Creek turbidity peaks during storms were consistently higher at the upstream stations before the stream crossing and in the during and after crossing periods were closer in magnitude. Lower Bottom Creek had consistently higher turbidity at the downstream station compared to the upstream station throughout the study period and anomalously high magnitude turbidity at the downstream station when compared to the upstream station in the during crossing period.

Differences in hourly median turbidity values between downstream and upstream stations show anomalous behavior in the during crossing period at lower Bottom Creek, as shown by the red points ascending the y-axis without a corresponding signal of similar magnitude on the x-axis (fig. 7). Most observations at lower Bottom Creek in the during stream crossing period plot to the left of the 1-to-1 line, suggesting that the downstream station frequently recorded higher turbidity values than the upstream station. The Blackwater River in the during stream crossing period had a pattern similar to the Roanoke River in the after stream crossing period with downstream turbidity exceeding upstream turbidity in the during stream crossing period as did the Roanoke River in the after crossing period. The other sites and periods generally are close to the 1-to-1 line or are evenly distributed about it.

Of the 18 comparisons (6 matched sites×3 periods), 15 had time-matched differences of less than or equal to 5 FNU during 90 percent of the time series. The remaining three comparisons had differences less than or equal to 5 FNU for 80 to 89 percent of the time series (table 1.2). Differences between downstream and upstream stations that were greater than 100 FNU represented less than or equal to 0 to 0.8 percent of the time series across all matched sites (table 1.2).

Distributions of median hourly turbidity differences among stations show that most of the observations across all sites and periods were close to zero (fig. 8). At Sinking Creek, the largest magnitude differences occurred in the before crossing period, with more total observations showing higher turbidity at the upstream station. The Roanoke River had some larger magnitude differences in the after crossing period, when the downstream station exceeding upstream levels, likely attributable to the presence of the pipeline ROW. At upper Bottom Creek, a small number of observations showed larger differences in the before crossing period, when the downstream station recorded higher turbidity. In the after crossing period, a similarly small number of larger differences were observed when the upstream station recorded higher values. At lower Bottom Creek, the before and during crossing periods showed a consistent signal of higher turbidity at the downstream station, with slightly more and greater magnitude observations occurring in the during crossing period. Most differences at lower Bottom Creek were still close to zero. At Blackwater River, all periods showed most differences near zero, with a more normal distribution and roughly equal numbers of observations greater than and less than zero. At Little Stony Creek, turbidity differences were close to zero for all periods.

The bootstrapped median differences confidence intervals tests resulted in 3 of the 18 tests confidence intervals overlapping zero, suggesting strong evidence that the median does not differ significantly from zero. Nine additional tests (12 total tests) yielded observed medians of the time-matched paired differences that were smaller than the uncertainty associated with instrument accuracy, suggesting that these tests did not differ significantly from zero (table 7). The six remaining tests overcame the uncertainty from instrument accuracy and had confidence intervals that did not overlap zero, included Little Stony Creek in the during stream crossing, upper Bottom Creek in the during stream crossing, lower Bottom Creek in the before, during, and after stream crossing periods, and Blackwater River in the during stream crossing period, suggesting that the median of the time-matched paired differences differed significantly from zero for these tests. When the uncertainty from instrument accuracy is accounted for in the six tests that yielded a significant difference, only two tests had a median of the differences that was reliably over 1 FNU, the before and during stream crossing periods at lower Bottom Creek, which ranged from 1.19 to 2.11 FNU and 1.42 to 2.38 FNU, respectively. These two tests at lower Bottom Creek represent the largest magnitude upstream-to-downstream differences; however, the differences were minor in practical terms, were present in the before stream crossing period, and had the smallest observed median of upstream-to-downstream differences in the after crossing period.

At all stations and for all periods, the median hysteresis index (HI) and median slope were positive, indicating turbidity tended to increase during storm events, compared to ambient conditions, and peaked on the rising limb of the hydrograph (fig. 9). A similar pattern was observed for all individual storm events at each station and across all periods, which is consistent with other stream turbidity hysteresis studies (fig. 1.7; Lloyd and others 2016b; Waite and others, 2023). Throughout the study period, positive HI values were observed at upstream and downstream stations the following times: 182 times at Little Stony Creek, 137 times at Sinking Creek, 59 times at Roanoke River, 98 times at upper Bottom Creek, 193 times at lower Bottom Creek, and 109 times at Blackwater River. Negative HI values were observed 2 times at Little Stony Creek, 18 times at Sinking Creek, 37 times at Roanoke River, 0 times at upper Bottom Creek, 2 times at lower Bottom Creek, and 36 times at Blackwater River; however, many HI values were close to zero, indicating no hysteresis. A threshold of −0.05 to 0.05 HI was used following the guidance of Waite and others (2023) as a range of HI values within which hysteresis is not present. The lack of hysteresis was noted 12 times at Little Stony Creek, 25 times at Sinking Creek, 27 times at Roanoke River, 2 times at upper Bottom Creek, 13 times at lower Bottom Creek, and 47 times at Blackwater River.

Turbidity analyses were used to evaluate whether pipeline stream crossing resulted in increased turbidity conditions in the during period relative to the turbidity regime measured using high-frequency data from sensors over the 7-year study period. Turbidity is a parameter of interest and importance because of its strong and well-established relation to suspended sediment concentration (Stubblefield and others, 2007; Rügner and others, 2013; Miller and others, 2023; Porter, 2025). Sedimentation in streams has been associated with pipeline stream crossings in previous studies (Lévesque and Dubé, 2007; Moyer and Hyer, 2009), and sedimentation has been shown to adversely affect instream biota and ecological functions (Henley and others, 2000; Wilber and Clarke, 2001). Measuring turbidity at a high frequency provided additional information about the short-term anomalous events observed. According to the EPA (2006), of the 29 States that had a threshold for turbidity, 10 FNU was the most common threshold. Fifteen States set thresholds based on comparison to background conditions, 12 have absolute turbidity thresholds, and 2 use a combination of both, with most States having stricter thresholds for streams that support colder water aquatic communities. All observed median upstream-to-downstream differences in this study were below established turbidity thresholds for all States that have a statute for turbidity, except for New Hampshire’s most protected streams that are classified as such whose threshold is 0 FNU above background conditions.

Results from the analysis of the high-frequency turbidity data at all six monitored stream crossing sites did not show evidence of long-term, persistent effects from pipeline stream crossing. Short-term, anomalously high turbidity conditions observed during stream crossing in this study were neither of sufficient magnitude nor duration to produce upstream-to-downstream median turbidity differences exceeding established States thresholds or levels known to alter ecological function. Across all sites, no significant turbidity difference was observed between the before and after stream crossing periods, which is consistent with similar studies in the region (Moyer and Hyer, 2009).

Storm turbidity hysteresis exhibited positive HI and positive slopes, indicating that most storms peak on the rising limb of hydrographs, which is consistent with similar studies (Lloyd and others, 2016a, b; Fovet and others, 2018; Waite and others, 2023). The positive HI indicates rapid flushing from a proximal source. The results of the hysteresis analysis; however, generally show consistency between upstream and downstream stations at all sites and periods. The storm hysteresis patterns add to the body of evidence that no persistent changes in turbidity were identified.

Previous studies of pipeline stream crossings have primarily focused on comparing crossing methods at a single downstream station and monitoring suspended sediment concentration and total suspended solids (Reid and others, 2002; Lévesque and Dubé, 2007). However, many of these studies are limited by low sample-collection frequency, short durations (less than 1 year), or a lack of monitoring across multiple stream crossings within the same study. However, Moyer and Hyer (2009) collected high-frequency turbidity data before, during, and after two pipeline stream crossings, which used two different crossing methods. Moyer and Hyer (2009) provided evidence that the trenchless crossing method had little effect on turbidity conditions; however, in this study, the largest magnitude anomalous turbidity observation occurred at lower Bottom Creek, which was crossed using the trenchless method. This might suggest that other factors might be driving these short-term anomalous turbidity conditions, such as ROW construction activities, failed sediment controls, high soil erosivity, or natural watershed processes like bank erosion. The median differences in Moyer and Hyer (2009) and in this report generally show low median turbidity differences between upstream and downstream stations across all streams and periods.

Out of the total 44,256,853 measurements from the USGS monitoring stations, a total of 137,263 (0.3 percent) flags were identified using the USGS real-time data when the parameter values exceeded parameter-specific water-quality standards or thresholds (table 8).

Alerts could be caused by a flag from the real-time alert system, self-reported by MVP, or a reported field complaint. A total of 464 alerts were produced from the real-time alert application from January 2018 to June 2025 among all six crossing sites. Out of the 464 alert notices, 253 (55 percent) resulted in a compliance action, and 211 (45 percent) resulted in no compliance issues.

Application of the real-time alert system provided timely information about stream conditions to regulators, installers, and the public, which can contribute to minimizing potential short- and long-term effects and improving regulatory agencies’ responses (Storey and others, 2011). Most of the real-time alerts led to compliance issues being identified on site, issues that likely would not have been detected without the alert system. Overall, the real-time alert system helped address and resolve compliance issues in a timely manner.