Purpose and Scope

The purpose of this report is to document HWM data, flood-peak magnitudes, and flood-mapping products generated by the USGS in support of the FEMA response and recovery operations after the September 1–2, 2021, flooding of eastern and south-central Pennsylvania caused by the remnants of Hurricane Ida. This report includes (1) a description of the atmospheric conditions during the remnants of Hurricane Ida and the temporal and spatial patterns of rainfall that triggered the flooding, (2) a narrative of the flood and its effects, (3) a description of how the 338 HWMs were documented, (4) an analysis of flood-peak magnitudes and their statistical probabilities at 52 USGS streamgages, and (5) a geographic information system (GIS) analysis of HWM locations and elevations to produce flood-documentation maps (areal extent of the maximum depth of flooding) for 5 heavily flooded areas in southeastern Pennsylvania. Data from this study are in tables and a separate data release (Stuckey and Conlon, 2023). The wording and presentation of the material in this report are based on previous USGS reports (Watson and others, 2017, 2018; Austin and others, 2018); the contents of each section are modified from these previous reports based on the specific flooding event.

Description of Study Area

The study area encompasses eastern and south-central Pennsylvania (fig. 2). Parts of 5 of the 6 physiographic provinces in Pennsylvania were flooded by the remnants of Hurricane Ida: the Appalachia Plateau, the Atlantic Coastal Plain, New England, the Piedmont, and Ridge and Valley (Pennsylvania Department of Conservation and Natural Resources, 2023); the Piedmont experienced most of the flooding across multiple watersheds during the event. The Delaware River Basin flooded, from north to south, in the Appalachian Plateau, Ridge and Valley, New England, Piedmont, and Atlantic Coastal Plain provinces. The northern part of the basin consists of rounded hills and valleys, and flat upper terrace surfaces cut by shallow valleys in the southern part with a mix of rolling lowlands and broad to shallow valleys in between. Land-surface-elevations within the Delaware River Basin range from 0 feet (ft) in the Atlantic Coastal Plain province to 2,690 ft in the Appalachian Plateau province, referenced to the North American Vertical Datum of 1988 (NAVD 88). The Lower Susquehanna and Potomac River Basins flooded in the north and west in the Appalachian Plateau and Ridge and Valley provinces and the south and east in the Piedmont province. Land-surface-elevations within the Lower Susquehanna and Potomac River Basins range in elevation from 20 ft in the Piedmont province to 3,210 ft in the Appalachian Plateau province, referenced to NAVD 88. In Pennsylvania, the 30-year normal annual rainfall (1991–2020) varied from 36.38 inches in Tioga County at Tioga Hammond Dam to 55.93 inches in Somerset County at Hidden Valley (National Centers for Environmental Information, 2023b).

Collection of High-Water-Mark Data

High-water marks are the evidence of the maximum water levels observed during a flood, and they are valuable for efforts to understand flood events. The USGS followed the guidance provided in Koenig and others (2016) to identify and document HWMs. Small seeds or debris carried by flood waters that adhere to smooth surfaces or lodge in tree bark to form a distinct line are the best HWMs. Stain lines on buildings, fences, and other structures also provide excellent HWMs. HWMs are best identified immediately after the peak stage of a flood event because they may blow or wash away from the weather (wind, rain, or sun) or fade away with time.

Working with FEMA under the direction of a Mission Assignment immediately after the flood, USGS field crews identified 255 HWMs in Pennsylvania on stream and river reaches that were selected using input obtained from FEMA, the NWS, U.S. Army Corps of Engineers, Delaware and Susquehanna River Basin Commissions, and Pennsylvania Emergency Management Agency. Examples of the identified HWMs are shown in fig. 4. Of those 255 HWMs, 198 were surveyed to determine the elevation above land surface using NAVD 88. HWMs were identified and marked beginning September 5, 2021, and continuing through September 16, 2021. Written descriptions, sketches, photographs, and global positioning system horizontal measurements (obtained with a hand-held global positioning system unit) were made so these marks could be easily found later and surveyed to the standard vertical datum (NAVD 88). Descriptions and photographs of these HWMs are available for viewing or download from the USGS Flood Event Viewer (U.S. Geological Survey 2021). In addition, 83 HWMs were identified and surveyed by the USGS as part of indirect computations of peak flow at select streamgages, bringing the total number of HWMs documented from the flooding that occurred on September 1–2, 2021, to 338. Indirect computations of peak flows are sometimes needed when a direct measurement of streamflow cannot be made because of unsafe conditions, lack of access, or timing issues and are made after a flood. The number of identified and surveyed HWMs and associated water bodies are listed in table 2 and are individually shown in fig. 5. A full listing of all HWMs is in Stuckey and Conlon (2023).

Peak-Flow Data and Streamgage Selection

During the Hurricane Ida remnant storm that resulted in flooding across eastern and southcentral Pennsylvania on September 1–2, 2021, USGS personnel made over 50 streamflow measurements using direct methods at streamgages in the study area (Rantz, 1982; Turnipseed and Sauer, 2010). In addition, in the months after the storm, USGS personnel computed streamflow measurements using indirect methods (Benson and Dalrymple, 1967; Rantz, 1982) at 12 locations. Many of the streamflow measurements were made to verify the accuracy of the stage-streamflow rating curve or extend the stage-streamflow rating curve for a given streamgage (Rantz, 1982). The stage-streamflow rating curve for a given streamgage is used to produce instantaneous streamflow values for a given streamgage, and the maximum instantaneous value for a given year is termed the annual peak streamflow (U.S. Geological Survey, 2023c). Streamgages with a long, annual peak streamflow record (more than 40 years) are considered the most reliable for estimating AEPs.

Magnitude and frequency statistics for floods were determined from annual peak streamflows measured or recorded at 52 USGS streamgages (fig. 2). Streamgages with at least 10 years of continuously recorded data were included in this study if those datasets met at least one of the following criteria: (1) the instantaneous peak from the September 1–2, 2021, flood event was the largest peak recorded during the period of record (peak of record) or (2) the recurrence interval of the instantaneous peak of September 1–2, 2021, was estimated to be at least 10-percent AEP and 1 of the 5 largest peak streamflow magnitudes recorded at the site over the period of record. If streamflow was regulated by a flood control reservoir during the period of record at a site, only the period after regulation was included in the analysis. The periods of record for the selected 52 streamgages ranged from 10 to 107 years, with a mean of 45 years. Annual peak-flow data for these streamgages were retrieved from the USGS National Water Information System (NWIS) website and carefully reviewed for completeness of record and appropriateness of historic peaks and qualification codes (U.S. Geological Survey, 2023c).

The annual peak flows of the 52 streamgages were analyzed to determine whether trends existed in their respective periods of record using a Mann-Kendall statistical test. Trends in annual peak flows can affect a flood frequency analysis by violating the assumption that the annual peak-flow data are a reliable and representative time sample of random homogeneous events (England and others, 2018). The results of the Mann-Kendall test showed that 17 of the 52 streamgages (33 percent) exhibited significant trends at the 90-percent confidence interval (p<0.10) (fig. 6). Data were evaluated for autocorrelation to confirm there was no serial correlation in the annual time series that would affect the results of the trend test (England and others, 2018). All significant trends were positive, indicating annual peak flows were increasing over time. Of the remaining 35 streamgages without a significant trend, 27 had a positive trend and 8 had a negative (decreasing streamflow) trend. A comparison of streamgages that were evaluated for trends by Roland and Stuckey (2020) and this current study show five streamgages that had changes in trend significance (at the 90-percent confidence interval); four streamgages now have a significant positive trend and one streamgage no longer has a significant positive trend. These trends in annual peak flows may be characterized by several factors, including length of the period of record, variations in precipitation over time, land use changes, or single causal factors (for example, the construction of a dam). Because the cause and longevity of the significant trends are uncertain to continue in the future, no streamgages were excluded from this flood-frequency analysis because of significant trends.

Flood-Frequency Analysis

The magnitude and frequency of flood flows at the 52 streamgages were computed using the expected moments algorithm (EMA) methodology (England and others, 2018), which fits a log-Pearson Type III probability distribution curve to annual peak streamflow data. The Interagency Advisory Committee on Water Data provides standard methods for computing peak streamflow frequency in its “Bulletin 17C” (England and others, 2018). Although Bulletin 17C maintains the moments-based approach of the previous Bulletin 17B procedures (Interagency Advisory Committee on Water Data, 1982), the method outlined in Bulletin 17C uses the EMA procedure (Cohn and others, 1997), which incorporates a generalized version of the multiple Grubbs-Beck test for the detection of low outliers (Cohn and others, 2013). EMA can also incorporate interval peak-flow data, which simplifies the analysis of datasets containing censored observations, historical data, low outliers, and uncertain data points, while providing enhanced confidence intervals for the estimated streamflows (Veilleux and others, 2014).

The log-Pearson Type III analyses were performed using the USGS program PeakFQ version 7.4 (Veilleux and others, 2014), following methods outlined in Bulletin 17C (England and others, 2018). Estimates of the annual peak-flow probabilities were computed in PeakFQ by determining the mean, standard deviation, and skew of the log-transformed annual peak-flow data and inserting the three statistics of the frequency distribution into the following equation:

As recommended in Bulletin 17C, a weighted average of the station skew and regional skew was used to determine flood frequencies for most streamgages, unless noted below. A regional skew of 0.343 (mean standard error of 0.178) was determined for Pennsylvania by Roland and Stuckey (2020) and used in this analysis. Note, potential computational issues related to the weighted skew calculation in PeakFQ remain unresolved at the time of preparation for publication (February 2023) (U.S. Geological Survey, 2023b). In some cases, the regional skew may be given too much weight over the station skew, particularly if the regional and station skew have a large difference (greater than 0.50). In this study, large differences between the regional and station skew were observed at eight streamgages (01467042, 01472198, 01472199, 01472620, 01472810, 01473000, 01474000, 01572025; table 3. One of these eight streamgages (01473000) had over 40 years of record, and the station skew coefficient option was used to avoid the potential weighting issue; the PeakFQ output for the remaining seven streamgages was evaluated for reasonableness and included in the analysis. The station skew coefficient option was used for streamgages with flow regulated by an upstream flood control reservoir because the regional skew was developed using unregulated watersheds and is not appropriate for use in regulated watersheds. A regulated watershed has at least 10 percent of the area upstream of its streamgage controlled by a reservoir.

The resulting flood frequencies for unregulated streamgages were weighted using the variance of independent estimates to reduce uncertainty as described in Bulletin 17C (England and others, 2018). Observed at-site flood-frequency estimates and predicted regression estimates from Roland and Stuckey (2020) are weighted in inverse proportion to their respective independent variances. Flood frequencies for regulated streamgages are not weighted because regression equations are not applicable. Flood frequencies associated with the 50-, 20-, 10-, 4-, 2-, 1-, 0.5-, and 0.2-percent AEPs for the 52 streamgages in Pennsylvania are available in Stuckey and Conlon (2023).

Flood-Probabilities of Peak Streamflow

After a substantial flood, emergency managers and water resource engineers often need to know the expected frequency of peak streamflow for the observed streamflow to provide information on the rarity of the event. The September 1–2, 2021, peak streamflows were analyzed using PeakFQ software version 7.4 (Veilleux and others, 2014) following guidance provided in the USGS Office of Surface Water Technical Memorandum 2013.01 (U.S. Geological Survey, 2012). The AEP for the flood event was estimated through the log-linear interpolation of flood frequencies determined from PeakFQ using the following equation:

Assigning an AEP to an observed flood results in greater uncertainty than determining traditional flood frequencies because the lack of reference data for the extreme flood being assessed. Limits based on the 66.7-percent confidence level were computed for each estimated AEP to convey this uncertainty.

Flood-Documentation Mapping

Flood-documentation maps were created using a geographic information system (GIS) approach to show the extent of inundation from the September 1–2, 2021, flooding along five selected river reaches in southeastern Pennsylvania that contained HWMs in sufficient number and density (fig. 7). These flood-documentation maps are intended to be estimates of the areal extent and the maximum depth of flooding that correspond to the HWMs identified and surveyed by USGS field crews after the flood event. The community, county, waterbodies, reach lengths, and number of HWMs used to generate the flood-documentation maps are listed in table 4.

A flood-elevation raster surface was created in GIS to generate the flood-documentation maps. GIS surfaces documenting the areal extent of the maximum depth of flooding were created independently for each area that was mapped using the point elevations of the HWMs and streamgages by using a GIS interpolation technique—the ArcGIS Topo to Raster tool (Esri, 2021) described by Musser and others (2016). A geographic areal boundary limit was placed on the generated surface because of the distribution of HWMs and an understanding of the natural hydrologic flow in each area that was mapped.

The flood-elevation raster surface that was created using GIS interpolation was then combined with 1-meter digital elevation model (DEM) data with a 10-meter cell size. The DEM was derived from light detection and ranging (lidar) data that had a 10-centimeter vertical root-mean-square-error and a nominal point spacing of 2 points per square meter or better (U.S. Geological Survey, 2023a). A DEM is a digital point representation of a surface; in this instance, the DEM is a topographical representation of the earth’s surface in the study area. An inundated area was depicted where the flood-elevation surface was higher than the DEM land surface. The maximum depth of flooding was calculated as the difference between the flood-elevation surface and the DEM land surface. Because of the many bridges in the mapped reaches (common in Pennsylvania), the inundation surfaces were not modified to show any bridges that were not inundated.

The flood-elevation interpolation method was used to develop the water-surface elevation layer from points represented by the HWM dataset and the USGS streamgage peak stage water-surface elevation values (where available). Interpolation was conducted between these elevations to generate the water-surface elevation using the ArcGIS Topo to Raster tool with the point interpolation procedure (Esri, 2021).

Uncertainties in the mapped extent and calculated depth of flooding within the maps result from the interpolation methods described above and the number and spatial distribution of HWMs within each mapped reach. Hydraulic models were not used to determine the extent or depth of flooding. Without hydraulic models, factors that can affect the extent or depth of flooding are not accounted for using this GIS interpolation methodology. Factors such as land-surface features, changes in flood plains, as well as the timing of the flooding event in the smaller inflow tributaries versus the larger main-stem tributaries and intermingling flows from adjacent streams, likely contribute to the extent and impact of flooding. In locations where HWMs are at a comparably greater distance of separation, spatial interpretations of the extent and depth of flooding may be less accurate. Within a given mapped area, some extrapolation was used beyond the most upstream and downstream HWMs. Often, the mapping boundary was extended to some anthropogenic structure, such as a road or bridge crossing.

The flood-documentation maps as described and shown in this report should not be used for navigation or regulatory, permitting, or other legal purposes. The USGS provides these maps “as is” for a quick reference, emergency planning tool but assumes no legal liability or responsibility resulting from the use of the maps. The maps shown in figs. 8–12 only depict the September 2021 remnant Hurricane Ida flood event on select stream reaches in southeastern Pennsylvania and may not be representative of other flooding conditions.

Brandywine Creek

Brandywine Creek generally flows southeast through Chester County in southeastern Pennsylvania to its confluence with the Christina River in Delaware. Brandywine Creek’s two main branches, the East Branch Brandywine Creek and the West Branch Brandywine Creek, both begin in northern Chester County. Separate flood-documentation maps for the East Branch Brandywine Creek, West Branch Brandywine Creek, and the main stem Brandywine Creek were created to show the areal extent of flooding in the Brandywine Creek Basin (figs. 8–10). The map extent for the East Branch Brandywine Creek includes 5 miles (mi) through several municipalities, including the borough of Downingtown, Pennsylvania. (fig. 8). The map extent for the West Branch Brandywine Creek includes a short distance (about 0.5 mi) as it flows through the borough of Modena (fig. 9). The main-stem Brandywine Creek includes 9 mi as it flows through several municipalities, including Chadds Ford Township (fig. 10). A total of 105 HWMs surveyed along the Brandywine Creek were used to create flood-documentation maps: 43 in the East Branch Brandywine Creek, 40 in the West Branch Brandywine Creek, and 22 in the main stem Brandywine Creek. The measured height above ground surface at the HWMs ranged from 0 to 7.5 ft, and the peak water-surface elevations ranged from 47.8 to 84.5 ft above NAVD 88. The USGS operates two streamgages on the Brandywine Creek that were also used in the creation of the flood-documentation maps:

The USGS operates four precipitation gages in the Brandywine Creek Basin, only three of the four undergo review and approval of the data. Approved rainfall totals from the USGS precipitation gages ranged from 6.68 inches (01480399) to 8.18 inches (01480617) on September 1, 2021 (U.S. Geological Survey, 2023c). The NWS reported rainfall totals in Chester County to be from 10.1 inches in Downingtown to 1.68 inches in Octoraro Creek Nottingham during the event (National Weather Service, 2023b).

Perkiomen Creek

Perkiomen Creek generally flows southeast through Berks, Bucks, Lehigh, and Montgomery Counties in southeastern Pennsylvania to its confluence with the Schuylkill River. Perkiomen Creek consists of two main branches: the Perkiomen Creek, which begins in Berks County, and East Branch Perkiomen Creek, which begins in Bucks County. A flood-documentation map was created to show the areal extent of flooding in the Perkiomen Creek Basin near Graterford and Collegeville, Pennsylvania, which is downstream from the confluence of the two branches at Schwenksville, Pennsylvania (fig. 11). The map extent includes 8 mi of the Perkiomen Creek, flowing through the boroughs of Graterford and Collegeville, Pennsylvania. (fig.11). A total of 23 HWMs surveyed along the Perkiomen Creek were used to create the flood-documentation map. The measured height above ground surface at the HWMs ranged from 0 to 9.7 ft, and the peak water-surface elevations ranged from 35.6 to 49.7 ft above NAVD 88. The USGS operates two streamgages on the Perkiomen Creek that were also used to create the flood-documentation map:

The NWS reported rainfall totals in Montgomery County to be 8.4 inches in Hatfield to 3.48 inches in Glenside during the event (National Weather Service, 2023b).

Schuylkill River

The Schuylkill River generally flows southeast through the Schuylkill, Berks, and Montgomery Counties of southeastern Pennsylvania to its confluence with the Delaware River near Philadelphia, Pennsylvania. The Schuylkill River flows for 135 mi from the confluence of the East and West Branch Schuylkill River to Philadelphia and is joined by several major tributaries. One flood-documentation map was created to show the areal extent of flooding in the Schuylkill River Basin near Norristown and Bridgeport, Pennsylvania, which is downstream from the confluence of the Schuylkill River and Perkiomen Creek at Phoenixville, Pennsylvania (fig. 12). The reach includes 4 mi of the Schuylkill River, flowing through the boroughs of Norristown and Bridgeport, Pennsylvania (fig. 12). A total of 26 HWMs surveyed along the Schuylkill River were used to create flood-documentation maps. The measured height above ground surface at the HWMs ranged from 0 to 9.1 ft above ground, and the peak water-surface elevations ranged from 22.3 to 24.0 ft above NAVD 88. The USGS operates one streamflow-gaging station on the Schuylkill River that was used in the creation of the flood-documentation maps:

The USGS operates a precipitation gage at 01473500 Schuylkill River at Norristown. These precipitation data are not reviewed and approved and offered as provisional data for 120 days only (U.S. Geological Survey, 2023c). The provisional rainfall total recorded at 01473500 during the event was 7.00 inches. The NWS reported rainfall totals in Montgomery County to be 8.41 inches in Hatfield to 3.48 inches in Glenside during the event (National Weather Service, 2023b).