Purpose and Scope

This report documents hurricane damages to the USGS hydrologic monitoring network in Puerto Rico and the U.S. Virgin Islands and summarizes selected activities conducted by the USGS in response to Hurricane Maria flooding in Puerto Rico. These activities include the following:

This report is focused primarily on USGS hurricane-response activities in Puerto Rico. However, the documentation of damages to the USGS hydrologic monitoring network includes both Puerto Rico and the U.S. Virgin Islands because of their integration in terms of network operation and maintenance.

Description of Study Area

Puerto Rico and its three principal offshore islands (Isla de Vieques, Isla de Culebra, and Isla de Mona) are the easternmost islands of the Greater Antilles, bounded on the north by the Atlantic Ocean and on the south by the Caribbean Sea (Ramos-Ginés, 1999). The main island of Puerto Rico is about 110 miles (mi) long and 40 mi wide; the combined area of the islands of Puerto Rico, Isla de Vieques, Isla de Culebra, and Isla de Mona is about 3,471 square miles (mi²; Olcott, 1999). In 2019, the approximate population of Puerto Rico was 3,194,000 (U.S. Census Bureau, 2020).

Based on studies of Puerto Rico and its offshore islands conducted in the 1960s by Picó, (1964), Gómez-Gómez and others (2014, table 1) identified 11 geographic regions (fig. 2) with distinct landform (physiography) and climate characteristics: (1) north coastal plain, (2) humid valleys of the east coast, (3) Caguas Valley, (4) valleys of the west coast, (5) south coastal plain, (6) semi-arid southern foothills, (7), humid northern foothills, (8) humid mountains of the east, (9) rainy mountains of the west, (10) Sierra de Luquillo, and (11) Outlying islands (Isla de Vieques, Isla de Culebra, and Isla de Mona). Although finer-scale landform and climate variability may influence site-specific flood characteristics during large hurricane events, these 11 geographic regions provide an effective framework to document Hurricane Maria flooding and its effect on watersheds in different parts of Puerto Rico.

Physiography

The principal physiographic feature of Puerto Rico is an east-west mountain range that effectively divides the island into two areas: a northern area that covers two-thirds of the island and a southern area that covers the remaining third of the island (Ramos-Ginés, 1999). The range is composed of the Cordillera Central in the western and central parts of the island, and the Sierra de Cayey in the eastern part (fig. 2); peak elevations range from 3,000 to more than 4,000 feet (ft) above sea level, with the highest point being the Cerro de Punta in the Cordillera Central, having a summit elevation of about 4,390 ft above Puerto Rico mean sea level. A smaller mountain range in northeastern Puerto Rico, the Sierra de Luquillo, supports dense tropical rain forests, whereas semiarid conditions prevail in the southern and southwestern parts of the island. The island’s coastal plains extend inland 8–12 mi on the northern part of the island and 2–8 mi on the southern part; the northern and southern coastal plains are separated from the interior mountains by the northern and southern foothills, respectively.

The Cordillera Central and Sierra de Cayey form the principal drainage divide of the larger watersheds whose stream channels generally are deeply incised into the mountain topography. Stream gradients in the mountains are steep, but streams that drain the northern side of the island are longer and their slopes gentler than those that drain the southern side. Flooding often occurs as the result of intense rainfall in the upper (mountainous) parts of the basins; however, streams on the southern side of the island are more susceptible to flash floods because of their shorter length and steeper upper-basin gradients (Murphy and Stallard, 2012; Ramos-Ginés, 1999). A summary of the principal streams in Puerto Rico and their characteristics is available in Puerto Rico Department of Natural and Environmental Resources (2007).

The USGS delineates watersheds by using a nationwide system based on surface hydrologic features (U.S. Geological Survey and Natural Resources Conservation Service, 2013). This system divides the United States and its territories into 21 regions (2-digit), 222 subregions (4-digit), 370 basins (6-digit), 2,270 subbasins (8-digit), ~20,000 watersheds (10-digit), and ~100,000 subwatersheds (12-digit). Throughout this report, watersheds and USGS streamgage sites in Puerto Rico are referenced to their 10-digit hydrologic unit codes (HUCs) as defined by this watershed accounting system. Figure 3 shows the distribution of the watersheds, referenced to their 10-digit HUCs, for the main island of Puerto Rico and the Isla de Vieques and Isla de Culebra offshore.

Climate

Puerto Rico has a tropical marine climate, and its trade-wind patterns are predominantly from the east-northeast (Murphy and Stallard, 2012). Seasonal temperature variability during the year is minor, with the largest differences of up to nearly 3 degrees Celsius (°C) occurring between the months of December and July; the island-wide average temperature is 25 °C (Jennings and others, 2014). The 30-year mean annual precipitation (period 1991–2020) is shown in figure 4A, B. Mean annual precipitation ranges from 32.49 in. in the southwestern part of the island to more than 150 in. in El Yunque National Forest (National Weather Service [NWS], 2021b).

Differences in rainfall patterns across the island reflect the influence of diverse topography (Jennings and others, 2014). The Cordillera Central and Sierra de Cayey mountains form an orographic control for the island’s rainfall (Bush and others, 2009). Rainfall amounts generally are higher in the northeastern part of the island (Sierra Luquillo), whereas drier conditions exist in the southern and southwestern parts. These regional variations in climate and rainfall patterns are characterized as part of the 11 geographic regions identified by Gómez-Gómez and others (2014).

Storm fronts, referred to as “tropical waves,” bring precipitation to Puerto Rico and its neighboring islands during a pronounced rainy season from May to November. Other more isolated rainfall events occur as the result of cold-front systems passing through the region from November to April (Gómez-Gómez and others, 2014). The hurricane season generally lasts from June to November, with hurricanes and tropical storms occurring most frequently between mid-August and late October (National Oceanic and Atmospheric Administration [NOAA] National Hurricane Center and Central Pacific Hurricane Center, 2020).

Weather Conditions Associated With Hurricane Maria

Originating from a tropical wave that formed east of the Lesser Antilles, Hurricane Maria began as a tropical storm on September 16, 2017 (fig. 1). Favorable environmental conditions allowed the storm to undergo rapid intensification as it approached the Lesser Antilles island arc. Hurricane Maria reached category 5 strength on September 18, just before making landfall on the island of Dominica. After weakening slightly as it passed over Dominica, Hurricane Maria achieved its peak intensity over the eastern Caribbean Sea with maximum sustained winds of 175 mi/h (NWS, 2021a) and a pressure of 908 millibars (26.81 in. of mercury), the lowest pressure on record for this part of the Atlantic Basin (Pasch and others, 2019). Hurricane Maria made landfall on September 20, 2017, in the Yabucoa Municipio, Puerto Rico, as a strong category 4 hurricane with maximum sustained winds of 155 mi/h. As the center of the storm moved over southeastern Puerto Rico then into the interior and northwestern Puerto Rico, hurricane force winds spread over the entire island. Heavy rainfall produced major to catastrophic flooding and flash flooding across most of the island.

Hurricane Maria’s center moved offshore from northwestern Puerto Rico on the afternoon of September 20, 2017. Although hurricane-force winds began to diminish when the system moved offshore, tropical-storm-force winds continued into the evening and overnight hours across the island (NWS, 2021a).

Estimated 48-hour rainfall totals for September 19–21, 2017, are shown in figure 5A, B (NWS, 2021a); individual station totals shown on the figure were compiled and published by the National Hurricane Center based on data from various rainfall monitoring sites in Puerto Rico.

Moderate to heavy rains continued to fall over various parts of the island in October and November 2017, causing additional flooding in watersheds still saturated by Hurricane Maria rainfall (NOAA, 2017). As a result, some streams in Puerto Rico experienced peak streamflows in October and November that were higher than those caused by the hurricane.

Damage to the USGS Hydrologic Monitoring Network Caused by Hurricane Maria

Prior to Hurricane Maria, the USGS maintained and operated 300 hydrologic monitoring stations in Puerto Rico and the U.S. Virgin Islands. Specifically, the pre-Hurricane Maria hydrologic monitoring network consisted of the following (appendix 1):

Plate 1 shows the pre-Hurricane Maria USGS hydrologic monitoring network in Puerto Rico, Isla de Vieques, and the U.S. Virgin Islands. Damage assessments by USGS personnel determined that 85 hydrologic monitoring stations were damaged or destroyed by high winds and (or) flooding in Puerto Rico (appendix 1). This total includes 48 streamgages, 7 canal gages, 9 lake/reservoir gages, 1 water-quality sampling site, 9 standalone precipitation gages, 10 groundwater wells, and 1 coastal monitoring station. Costs associated with the infrastructure and equipment damage (including gage replacements and other repairs) were estimated at more than 1 million U.S. dollars.

USGS personnel began repairing and rebuilding affected monitoring stations in Puerto Rico as soon as they could be accessed. Personnel prioritized repairs according to needs identified by local and Federal cooperators involved with recovery activities. Some stations were relocated, and several were discontinued; several new stations also were installed as part of the restoration activities. The post-Hurricane Maria USGS monitoring network was fully operational by mid-2018.

Hurricane Maria not only had an economic impact on the USGS hydrologic monitoring network in terms of the cost of reconstruction or repair of monitoring stations, but also produced historical record peak streamflows that caused substantial morphologic changes to stream channels throughout Puerto Rico. The accuracy of computed streamflow from many streamgages was severely compromised because of channel aggradation or degradation. At-site streamflow measurements made over periods of months or years in a variety of streamflow conditions, often in conjunction with hydraulic modeling, were needed to reestablish the accuracy of streamflow records. Hydraulic modeling also was needed at many sites to determine peak streamflows associated with observed flood events.

High-Water Mark Identification and Documentation

High-water marks (HWMs) are evidence of the highest water levels that occurred during a riverine flood or coastal storm-tide event and are critical for reliable determination of peak streamflow by indirect methods described in subsequent sections of the report. USGS field crews followed guidance provided by Koenig and others (2016) to identify and document HWMs, noting the characteristics of each mark and assessing the uncertainty associated with each mark. Documenting HWM uncertainty is important because streamflow computations rely on analyst interpretation of the most reliable HWMs to estimate the peak water-surface profile in the study reach. HWM uncertainty, accompanied by a detailed description and photographs of the mark, is quantified according to the ranges identified in table 1.

Site Topographic Surveys and Characterization of Roughness

Post-flood analyses of peak streamflows and stream-channel conditions require detailed field surveys to characterize HWMs and peak water-surface profiles, channel cross-section geometry, and any relevant hydraulic structures such as bridges or roadways. USGS field crews used total stations (that is, precision electronic/optical instruments with a tiltable telescope and automated data processing and storage capabilities) and real-time kinematic Global Navigation Satellite System (RTK–GNSS) equipment to document the location and elevation of HWMs and hydraulic structures, and for conducting channel cross-section surveys. Whenever possible, surveyed elevations were referenced to both the designated vertical datum used at the streamgage and to PRVD02, geoid 12B (National Geodetic Survey, 2014). Further information describing the horizontal and vertical datums used for the topographic surveys of indirect measurements and step-backwater analysis (SBA) is provided in Gómez-Fragoso and Smith (2022). Surveying protocols adhered to guidance in Benson and Dalrymple (1967), Kenney (2010), and Rydlund and Densmore (2012).

Post-flood field studies also required characterization of channel and floodplain features in terms of a roughness coefficient, which quantifies the degree to which these features resist gravitational forces propelling the streamflow downstream. Roughness typically is characterized by Manning’s roughness coefficients or n-values (hereafter referred to as “Manning’s n-values”), which apply to a longitudinal reach of the channel and floodplain (Chow, 1959).

In practical applications of post-flood analytical methods described in this section, the Manning’s n-values for a given site and flood event are not known and must be estimated by experienced hydrographers for observed site features and conditions. Several methods are available to assist with the estimation process, including the Cowan method (Cowan, 1956) and photographic comparison (for example, Barnes, 1967, for channel roughness; and Arcement and Schneider, 1989, for floodplain roughness).

USGS personnel used one or more of these estimation methods to estimate Manning’s n-values at sites where one-dimensional hydraulic analyses were used to compute peak streamflow or to assess post-flood stage-streamflow relations. Where two-dimensional hydraulic modeling was used to compute peak streamflow, surface roughness was characterized in terms of the drag coefficient. This coefficient is a dimensionless quantity that characterizes the drag or resistance of an object in water as a function of sediment grain-size and vegetation characteristics through the study reach (Barton and others, 2009). Although the drag coefficient can be related computationally to the Manning’s n-values, the underlying physical assumptions associated with the two coefficients differ slightly.

Indirect Measurements of Peak Streamflow and Associated Uncertainty

USGS personnel used standard USGS techniques and methods for one-dimensional indirect measurement of peak streamflows (Benson and Dalrymple, 1967), slope-area measurements (Dalrymple and Benson, 1968), culvert-control measurements (Bodhaine, 1968), width-contraction (bridge) measurements (Matthai, 1967), and road-overflow (weir) measurements (Hulsing, 1967). Access to streamgage sites by USGS field crews during Hurricane Maria flooding was unsafe and (or) physically impossible. Indirect (or post-flood) methods were used to characterize peak streamflows that could not be determined using direct methods such as current-meter or hydro-acoustic measurements, or from preflood stage-streamflow relations. Indirect determination of peak streamflow involves several steps: (1) site reconnaissance to identify an appropriate study reach; (2) onsite identification and flagging of flood HWMs; (3) topographic surveys of HWMs, channel cross-section geometry, and affected structures (such as bridges, culverts, weirs, or flooded roadways); (4) estimates of Manning’s n-values; and (5) hydraulic modeling to compute the peak streamflow using field data collected at the site.

Based on field reconnaissance efforts in October 2017, USGS personnel implemented the most appropriate indirect method(s) to compute peak streamflow at selected sites. Analysts used public-domain software tools developed by USGS or collaborating organizations to facilitate the following types of computations:

Flooding characterized by complex two-dimensional hydraulics cannot be analyzed using the one-dimensional methods described above. Estimation of peak streamflow for complex streamflow conditions requires more advanced modeling tools, some of which are available in the iRIC software platform (Shimizu and Nelson, 2020). Developed by USGS researchers, the iRIC System for Transport and River Modeling (SToRM) is a numerical model that uses the two-dimensional depth-averaged equations (also referred to as the “shallow water equations”) to simulate complex surface-water streamflows (Shimizu and Nelson, 2020). The iRIC SToRM model was used to estimate the Hurricane Maria peak streamflow at one site where flood hydraulics were characterized by valley-wide, two-dimensional streamflow components, low-gradient topography, and coastal backwater conditions.

Characterization of indirect-measurement uncertainty generally is qualitative, based on empirical guidelines and sensitivity analyses conducted for each measurement (comparison of computed streamflows for the feasible range of water-surface profile interpretation, Manning’s n-values, and other computational variables). Benson and Dalrymple (1967) provided an empirical framework for assessing the uncertainty of each measurement:

Several factors can affect the quality and reliability of an indirect measurement (Costa and Jarrett, 1981; Flaxman, 1974; Jarrett, 1986; Randall and Humphrey, 1984). HWM information and flood-peak water-surface profiles derived from observed HWMs are critical for reliable streamflow computations; HWM composition, quality, and density along each bank of the study reach are among the primary factors in assessing measurement quality. Selection of Manning’s n-values is another source of uncertainty. Manning’s n-values for a given site and flood peak are not known and thus are estimated using established methods and analyst experience. Sediment loads carried by the floodwaters also affect measurement quality. Ideal study reaches having stable, uniform channels with little erosion or sedimentation often do not exist. At sites where substantial sediment movement occurs, channel geometry can be significantly different during and after the flood peak; unstable channels subject to erosion and (or) sedimentation are therefore subject to additional uncertainty in the computed streamflow.

USGS analysts use sensitivity analyses over the feasible range of possible water-surface profile interpretations (elevation and slope), channel roughness (Manning’s n-values), and, in some cases, channel-bed geometry in combination with computational diagnostics, to characterize the uncertainty of each measurement.

SBA Method for Development of Theoretical Stage-Streamflow Relations at Streamgages

The standard-step hydraulic method, often referred to as the “step-backwater analysis method,” is a widely accepted one-dimensional hydraulic analysis used to determine theoretical water-surface elevations at a location of interest for specified streamflows (Chow, 1959; Davidian, 1984). Some applications of the SBA method include the establishment of a stage-streamflow relation (rating curve), definition of flood-inundation extent under specified hydrologic conditions, and computation of water-surface profiles needed for the analysis of hydraulic structures (Davidian, 1984). The SBA method was used by the USGS in Puerto Rico to develop stage-streamflow rating curves for newly installed streamgages, or for existing streamgages where Hurricane Maria caused substantial changes to channel morphology and invalidated the pre-hurricane stage-streamflow relation.

Application of the step-backwater method involves three main components: (1) data collection to characterize the channel and any associated hydraulic structure geometries, (2) selection of boundary conditions at which the model will initiate hydraulic calculations, and (3) development of the stage-streamflow relation for a range of user-selected streamflows and associated computed water-surface elevations at the desired location (Federal Emergency Management Agency, 2016). Convergence tests are conducted for each model to ensure water-surface profiles for a given streamflow have converged to the normal depth at the streamgage cross section, regardless of the water-surface elevation assumed at the downstream boundary; if tests show the model does not converge at the point of interest, the study reach must be extended further downstream. The resulting theoretical relation is particularly useful over the range of medium and high streamflows for which the collection of direct or indirect streamflow measurements is difficult or not possible; ideally, the theoretical relation subsequently will be verified or adjusted using the results of direct or indirect measurements.

Following completion of the field surveys, USGS analysts processed the survey data and developed one-dimensional hydraulic models for a target range of streamflows using the Hydrologic Engineering Center’s River Analysis System (HEC–RAS) computer model developed by the U.S. Army Corps of Engineers (2016). USGS personnel used HEC–RAS version 5.0 to develop theoretical stage-streamflow relations following Hurricane Maria and, in some cases, to estimate the Hurricane Maria peak streamflow for streamgage sites at which indirect measurements could not be made but peak gage heights were available. The term “gage height” (as opposed to the more general term “stage”) refers to the water-surface elevation relative to an arbitrary gage datum and is used in this report when documenting gage readings or water-surface elevations relative to the datum established for a given streamgage.

Additional modeling steps included (1) extension of the field surveying by using light detection and ranging (lidar)-derived data as necessary; (2) adjustment of the survey data, including processing of the Online Positioning User Service solution, rotating surveyed points (if needed), and straightening of cross sections before data input to the model; (3) modifications of original (at-site) channel-roughness subdivisions based on subarea ratio tests (Davidian, 1984); and (4) use of vertical variations in Manning’s n-values. Hydraulic structures, such as weirs and bridges, were surveyed and included in the models where necessary. Additional cross sections were generated using geographic information system (GIS) tools and lidar data when needed to improve model diagnostics and model accuracy.

CSA Method for Indirect Determination of Peak Streamflow and Theoretical Stage-Streamflow Relations

The CSA method is an extension of the conventional slope-area method used for indirect determination of streamflow over a complete hydrograph (Smith and others, 2010). The CSA method has been used extensively by the USGS in Arizona for indirect measurement of streamflows on ephemeral streams (Stewart and others, 2012). Because the CSA method computes streamflows that vary with time over the flood hydrograph, steady-flow assumptions of the conventional slope-area method introduce some error in computed velocities (Wiele and others, 2015); however, other studies assert that these errors become less significant in reaches with steeper slopes (Smith and others, 2010).

Unlike the conventional slope-area method, which determines the water-surface slope from profiles defined by numerous HWMs associated with a flood peak, the CSA method determines water-surface slopes by using a time series of gage-height data recorded at three or more fixed cross-section locations during a streamflow event. Gage-height data are collected using continuously recording pressure transducers (PTs) installed at preselected cross-section locations along the study reach (Wiele and others, 2015).

CSA streamflow computations are made using the CSA2SAC program (Wiele, 2015), an automated algorithm developed by the USGS in Arizona that runs the SAC computer program (Fulford, 1994) during each interval of the recorded gage-height time series. The final product is a complete stage-streamflow hydrograph composed of the series of individual SAC output files.

Indirect Measurements of Peak Streamflows Caused by Hurricane Maria and Subsequent Flooding in 2017

Because of ongoing emergency response activities and the extensive island-wide damage caused by Hurricane Maria, USGS field crews were unable to begin collecting flood HWM information until the middle of October 2017, nearly 4 weeks after passage of the hurricane. Moderate to heavy rainfall continued in many areas during and after this 4-week period, causing additional flooding—with some post-Hurricane Maria flood peaks exceeding those caused by the hurricane itself—and degrading or washing away much of the HWM evidence of interest. As a result, HWMs generally were of poor quality and subject to large uncertainty in terms of determining which HWMs were associated with each flood event.

Figure 6A, B shows the locations of 20 sites for which indirect measurements were surveyed, computed, and published, and 4 sites for which planned indirect measurements either were not surveyed or were later rejected; site locations are shown relative to the geographic regions of the island (fig. 6A) and to the watersheds of Puerto Rico (fig. 6B). Tables 2 and 3 summarize peak streamflow information, physiography, and hydraulic characteristics for the 20 published indirect measurements. Surveyed cross-section and HWM data (horizontal coordinates, elevations, and survey coordinate system) for each study site are available in a USGS data release (Gómez-Fragoso and Smith, 2022). Appendix 2 documents the implementation of indirect-measurement methods after the hurricane and provides detailed site and computational information for selected indirect measurements of peak streamflow resulting from Hurricane Maria and subsequent flooding in 2017.

The estimated uncertainty of each computed peak streamflow is shown in tables 2 and 3. As noted previously, deteriorated HWM evidence and multiple flood peaks were among the largest sources of uncertainty for indirect measurements made after the passage of Hurricane Maria. HWMs at many sites were scarce or exhibited large variability in surveyed elevations, making profile interpretation difficult. When multiple peaks occurred during Hurricane Maria flooding and subsequent flooding into October, HWMs recovered by USGS field crews in October potentially reflected a mix of different flood events.

Despite substantial uncertainties associated with indirect measurements made to characterize Hurricane Maria flooding, the measurements generally were judged to be more reliable than estimates made using SBA methods, because they incorporate observed HWM information and water-surface profiles at the time of the flood peak. As such, the indirect measurements were used as calibration data for SBAs made at corresponding streamgages.

The four sites listed below and shown in figure 6A, B were targeted for indirect measurement but ultimately were rejected because of unfavorable site conditions or computational issues; peak streamflows for these sites were estimated from SBAs:

Stage-Streamflow Relations Developed Using SBA and CSA Methods

Table 4 summarizes selected reach and hydraulic-modeling characteristics of sites where step-backwater field surveys were conducted for use in developing new stage-streamflow relations. Figure 7A, B shows the locations of the 58 surveyed SBA sites and identifies the sites at which both indirect measurements and SBAs were conducted in 2017 and 2018; site locations are shown relative to the geographic regions of the island (fig. 7A) and to the watersheds of Puerto Rico (fig. 7B).

Table 4 indicates which sites required cross-section extension using lidar data. Cross-sectional data for each site listed in table 4 are available as a USGS data release (Gómez-Fragoso and Smith, 2022). Roughness conditions varied widely for streams in the Puerto Rico streamgage network, resulting in assigned Manning’s n-values ranging from 0.020 to 0.120 in the main channel and from 0.024 to 0.210 in the overbank (floodplain) areas (table 4).

Assigned energy-loss coefficients varied for each model, with initial values set to the USGS recommended coefficients and HEC–RAS modeling guidelines for contraction (k_c) and expansion (k_e). As part of the model-calibration process, some of the initial energy-loss coefficients were adjusted to smooth section-to-section conveyance ratios or to ensure computed water-surface elevations were increasing in the upstream direction. Table 4 shows the range of energy-loss coefficients used for each study reach and the slope used for normal depth calculations. Normal depth was used as the starting water-surface elevation at the modeled cross section furthest downstream, and the slope used for normal depth calculations usually was the thalweg slope of the reach segment furthest downstream; for very low gradient study reaches, however, the slope used for normal depth calculations was based on a reach segment further upstream.

Available post-Hurricane Maria direct and indirect streamflow measurements (that is, those associated with Hurricane Maria or other large flood events that occurred through November 2017) were used to calibrate each model whenever possible (table 5). Historical high-streamflow measurements also were used for model verification at streamgage sites having no recent measurements or having a limited range in magnitude of measured streamflows. In some cases, model results were favorable compared to available streamflow measurements; however, in other cases, modeled water-surface elevations could not be calibrated within a feasible range of modeling parameters to match water-surface elevations associated with the streamflow measurements. For those sites where calibration could not be achieved by making physically reasonable adjustments to the model, future direct and indirect measurements will be used to verify or revise the existing SBA. Estimates of Hurricane Maria peak streamflows were made at 39 streamgage sites using SBA modeling and available peak-stage information. Most estimates were considered poor (rated uncertainty of ±25 percent or greater) unless an indirect measurement could be used to calibrate the model. Each SBA model and computed Hurricane Maria peak streamflow was reviewed and approved according to USGS data-processing standards.

Appendix 3 documents the implementation of SBA methods in 2018 and provides details of SBAs conducted at 4 sites, including a description of reach characteristics, model results, and comparison of the pre-Hurricane Maria stage-streamflow relation with the post-Hurricane Maria stage-streamflow relation (developed based on SBA).

CSA methods were implemented in 2018 at USGS streamgages 50044810 Río Guadiana near Guadiana and 50093000 Río Marín near Patillas. At the time of this report, no CSA analyses of peak streamflows were available for streamgage 50044810 Río Guadiana near Guadiana. CSA analyses were conducted for 18 peak streamflows observed at streamgage 50093000 Río Marín near Patillas, for use in the development of a new stage-streamflow relation. All events occurred in the 2018 water year, and computed peak streamflows varied from 963 cubic feet per second (ft³/s) (elevation of 10.10 ft above gage datum) to 2,250 ft³/s (elevation of 12.47 ft above gage datum). The computations were rated as poor (uncertainty of ±25 percent or greater) because of large uncertainties associated with the measurements (for example, no observed HWMs for verification of sensor readings and substantial scatter of 2-section computations versus the accepted, 4-section computation of peak streamflow). Appendix 4 documents the implementation of CSA methods at both pilot sites and provides a detailed discussion of peak-streamflow computations for CSA measurements made at streamgage 50093000 Río Marín near Patillas.

Peak Gage Heights, Peak Streamflows, and Annual Exceedance Probabilities for Hurricane Maria Flooding, September 20-22, 2017

USGS personnel computed peak streamflows for flood events across the island of Puerto Rico from (1) indirect measurements associated with Hurricane Maria flooding, (2) step-backwater studies conducted in 2018 to characterize Hurricane Maria and post-hurricane stage-streamflow relations, and (3) existing stage-streamflow rating curves that provided valid streamflow data for Hurricane Maria flooding. Table 6 summarizes peak gage heights, peak streamflows, periods of record (PORs), number of annual peak streamflows used for analysis,⁴ and period of comparable-record ranks (series of annual-maximum peak streamflows) for flooding that occurred September 20–22, 2017, at 73 streamgages in Puerto Rico. The table includes unregulated, regulated, and urbanized sites; regulated and urbanized sites are identified in the “Remarks” column. Hurricane Maria peak streamflows at four sites were exceeded by different flood events during the 2017 water year and therefore are not ranked in the series of annual-maximum peak streamflows (shown as “NR” in table 6).

The number of years of annual peak streamflows ranges from 4 to 69 (table 6), with a median of 30 years. Hurricane Maria peak streamflows ranked highest for the POR at 28 sites (38.4 percent), second highest for the POR at 17 sites (23.3 percent), and third highest for the POR at 9 sites (12.3 percent). For the remaining 19 sites, Hurricane Maria peak streamflows either ranked from 4th to 20th or were not ranked.

Figure 8A, B shows the spatial distribution (island-wide and within each geographic region and watershed, respectively) of sites in each of the four ranking categories of highest, second highest, third highest, and 4th–20th highest or not ranked. POR peaks ranked highest and second highest occurred throughout the island, consistent with Hurricane Maria’s trajectory across the entire island from southeast to northwest. However, it is apparent from figure 8A, B that peaks ranked third highest and smaller occurred mainly in the eastern and southern parts of the island. Although several factors (watershed characteristics, station periods of record, regional physiography, and long-term precipitation, or the meteorological dynamics of the hurricane itself) might have influenced the spatial distribution of lower-ranked peak streamflows, investigation of those factors is beyond the scope of this report.

Fifty-eight of the 73 streamgages in table 6 have PORs with 20 or more published annual peaks. Because several large historical floods occurred in the 1990s (specifically related to Hurricane Hortense in 1996 and Hurricane Georges in 1998) and earlier, an evaluation of Hurricane Maria peak streamflow ranking for streamgages with at least 20 annual peak streamflows⁵ provides some historical context. For this subset of 58 sites, 19 (32.8 percent) Hurricane Maria peaks were the highest, 14 (24.1 percent) were the second highest, and 8 (13.8 percent) were the third highest for their respective PORs.

Ryan and others (2021) documented updated flood-frequency analyses and new regional equations for Puerto Rico, including peak-streamflow data through the 2017 water year. The magnitude of an observed peak streamflow relative to the series of annual-maximum peak streamflows for a given site is characterized statistically in terms of the following concepts:

AEP estimates (developed using the new flood-frequency analyses and regional equations for Puerto Rico), corresponding recurrence intervals, and associated confidence intervals for 53 unregulated Hurricane Maria peak streamflows that met computational criteria for flood-frequency and AEP analysis are provided in appendix 5. Specific computational criteria were the following:

Estimates were determined using computational procedures documented in USGS Office of Surface Water Technical Memorandum 2013.01 (USGS, 2012b) and England and others (2018); source data for the computations are documented in the report by Ryan and others (2021) and are incorporated into the USGS Web application StreamStats (Ries and others, 2017; https://streamstats.usgs.gov/ss/)

AEPs for the 53 unregulated peak streamflows listed in appendix 5 are summarized by range of exceedance probability in figure 9; they ranged from greater than 50.0 percent (recurrence interval of less than 2 years) to 0.3 percent (recurrence interval of 333 years), with the majority (28 of 53) in the range of 10.0–2.1 percent (recurrence interval of 10–48 years).

Hurricane Maria peak streamflows at 16 sites were subject to undetermined effects of regulation from upstream dams; Hurricane Maria peak streamflows at two sites were affected by watershed urbanization (see remarks in table 6). AEP estimates can be computed for both regulated and urbanized sites (USGS, 2012b; England and others, 2018), but the computational process for regulated sites requires (1) evaluation of dam operational protocols over the entire regulated period of record and (2) detailed analysis of peak streamflows generated by flow releases to determine whether or how reservoir storage affected the flood hydrograph. Such analyses were beyond the scope of this study. Historical annual peak streamflows for regulated and urbanized sites in Puerto Rico are publicly available from NWISWeb (USGS, 2021a) for use by readers interested in conducting site-specific AEP analyses.

Comparison of Flooding Caused by Hurricane Maria and Past Hurricanes in Puerto Rico

Hurricane San Ciriaco (1899) is the earliest hurricane for which USGS peak streamflow information in Puerto Rico is available. Table 7 summarizes the chronology, storm category, trajectory, and rainfall totals of the six major (category 3, 4, or 5) hurricanes that directly impacted Puerto Rico between 1899 and 2017. Table 7 also includes Hurricane Hortense (1996), which directly impacted Puerto Rico as a category 1 storm. Although it was not a major hurricane when it passed over Puerto Rico, this storm did cause extensive flooding and either record or near-record historical flood peaks at several USGS streamgage.

Although the effects of Hurricane Irma were overshadowed by those of Hurricane Maria, Hurricane Irma indirectly affected Puerto Rico just 2 weeks earlier (September 6–7, 2017) when it passed 50 nautical miles to the north of the island. As a category 5 storm, Hurricane Irma produced 10–15 in. of rainfall in parts of Puerto Rico and 2017 water-year peak streamflows that exceeded those caused by Hurricane Maria at a few USGS streamgages.

Largest Three Historical Flood Events Based on Period-of-Record USGS Streamflow Data

Table 8 summarizes POR ranks 1, 2, and 3 for 58 streamgage sites with at least 20 years of annual peak streamflows. Inspection of the years in which annual peak streamflows ranked 1, 2, and 3 occurred at each streamgage provides a framework to compare the magnitude of Hurricane Maria flooding with past large floods.

Table 8.    

Compilation of water years in which rank 1, 2, and 3 period-of-record streamflows occurred at 58 streamgage sites in Puerto Rico with at least 20 years of annual peak-streamflow record.

[blw, below; PR, Puerto Rico; nr, near; abv, above; Hwy, Highway; ft³/s, cubic foot per second]

Table 8.    Compilation of water years in which rank 1, 2, and 3 period-of-record streamflows occurred at 58 streamgage sites in Puerto Rico with at least 20 years of annual peak-streamflow record.

Station number	Station name	Period of record (water years)1	Number of peak streamflows used2	Water year for ranked period-of-record streamflow			Remarks
				Rank 1	Rank 2	Rank 3
50011200	Rio Guajataca blw Lago Guajataca, PR	1970-1971, 1984-1993, 2005-2017	25	2017	1986	2011
50014800	Rio Camuy nr Bayaney, PR	1984-2017	34	1998	2017	2002
50021700	Rio Grande de Arecibo abv Utuado, PR	1989-1993, 1995-2017	28	1998	2017	2008
50024950	Rio Grande de Arecibo blw Utuado, PR	1996-2017	22	1998	2017	2011
50025155	Rio Saliente at Coabey nr Jayuya, PR	1990-1996, 1998-2017	27	1998	2017	1996
50026025	Rio Caonillas at Paso Palma, PR	1996, 1998-1999, 2001-2017	20	2017	1998	1996
50028000	Rio Tanama nr Utuado, PR	1960-2017	58	1998	2017	1985
50028400	Rio Tanama at Charco Hondo, PR	1969-1971, 1982-2017	39	1998	2017	1985
50029000	Rio Grande de Arecibo at Central Cambalache, PR	1899, 1928, 1955, 1969-1983, 1997-20173	37	2017	1955	1979	Streamflows regulated 1955-2017. Unregulated peak streamflow of 1899 was higher than peak streamflow of 2017.
50031200	Rio Grande de Manati nr Morovis, PR	1965-1973, 1975-1996, 1998-2017	51	2017	1985	1996
50034000	Rio Bauta nr Orocovis, PR	1970-1982, 1989-2017	42	2017	1998	1996
50035000	Rio Grande de Manati at Ciales, PR	1947-1953, 1956-2017	69	2017	1996	1971
50038100	Rio Grande de Manati at Hwy 2 nr Manati, PR	1928, 1932, 1945, 1959-1963, 1965-1966, 1968-2017	60	1996	2017	1928
50038320	Rio Cibuco blw Corozal, PR	1970-2017	48	2017	2002	1996	Rank 3 peak streamflow (water year 1996) occurred November 9, 1995—not associated with Hurricane Hortense.
50039500	Rio Cibuco at Vega Baja, PR	1959-2017	59	2017	1987	1982
50043800	Rio de La Plata at Comerio, PR	1989-2017	29	1992	1996	2017
50045010	Rio de La Plata blw La Plata Damsite, PR	1989-1996, 1998-2017	28	1996	2017	1992
50046000	Rio de La Plata at Hwy 2 nr Toa Alta, PR	1899, 1928, 1943, 1960-20173	44	1996	2017	1992	Streamflows regulated 1974-2017. Unregulated peak streamflow of 1899 was higher than peak streamflow of 2017.
50047535	Rio de Bayamon at Arenas, PR	1993, 1996-2017	23	1998	1999	2008
50047850	Rio de Bayamon nr Bayamon, PR	1965-1971, 1989-2017	36	1996	1971	2013
50049100	Rio Piedras at Hato Rey, PR	1970, 1972-1974, 1976-1982, 1988-2017	41	2010	2017	1996
50050900	Rio Grande de Loiza at Quebrada Arenas, PR	1978-2017	40	1996	1998	1992	Rank 1 and rank 2 peak streamflows are both published as greater than 21,500 ft3/s; the first occurrence (1996) is listed as rank 1.
50051310	Rio Cayaguas at Cerro Gordo, PR	1978-2017	40	2008	2001	1979
50051800	Rio Grande de Loiza at Hwy 183 San Lorenzo, PR	1991-2017	27	1996	1992	1998
50053025	Rio Turabo abv Borinquen, PR	1991-2017	27	1996	1998	1992
50055000	Rio Grande de Loiza at Caguas	1945, 1960-2017	59	1945	1996	1960
50055225	Rio Caguitas at Villa Blanca at Caguas, PR	1991-2017	27	1996	2000	2017
50055750	Rio Gurabo blw El Mango, PR	1991-2017	27	2017	1996	1998
50056400	Rio Valenciano nr Juncos, PR	1960, 1971-2017	48	1988	1960	1985
50057000	Rio Gurabo at Gurabo, PR	1960-2017	58	2017	1960	1971
50058350	Rio Canas at Rio Canas, PR	1990-2017	28	2005	1996	2017
50059050	Rio Grande de Loiza blw Loiza Damsite, PR	1987-2017	31	1996	1998	2017	Rank 2 peak streamflow (water year 1998) occurred October 14, 1997—not associated with Hurricane Georges.
50061800	Rio Canovanas nr Campo Rico, PR	1968-2017	50	1998	2008	2017
50063800	Rio Espiritu Santo nr Rio Grande, PR	1967-2017	51	2017	1990	1988
50064200	Rio Grande nr El Verde, PR	1968-1982, 1991-2017	42	1998	2004	2017
50065500	Rio Mameyes nr Sabana, PR	1968-1973, 1983-2008, 2011-2017	39	1989	2014	1973
50067000	Rio Sabana at Sabana, PR	1980-2017	38	1992	1983	2008
50071000	Rio Fajardo nr Fajardo, PR	1960, 1962-2000, 2002-2017	56	1992	1989	1975
50075000	Rio Icacos nr Naguabo, PR	1946, 1948-1952, 1954-1955, 1957-1962, 1980-2017	52	1983	2003	1989
50081000	Rio Humacao at Las Piedras, PR	1960, 1969, 1971-1977, 1988-2017	39	1960	1988	1976
50090500	Rio Maunabo at Lizas, PR	1971-1984, 1992-2008, 2011-2017	38	1994	2008	1996
50092000	Rio Grande de Patillas nr Patillas, PR	1966-2017	52	1998	1992	1996
50100200	Rio Lapa nr Rabo del Buey, PR	1971, 1989-2017	30	1996	1992	2017
50100450	Rio Majada at La Plena, PR	1989-2017	29	2017	1996	1992
50106100	Rio Coamo at Hwy 14 at Coamo, PR	1988-2017	30	2017	1998	1992
50110900	Rio Toa Vaca abv Lago Toa Vaca, PR	1990-2017	28	1998	1996	2017
50111500	Rio Jacaguas at Juana Diaz, PR	1984-2017	34	2017	1986	1998
50112500	Rio Inabon at Real Abajo, PR	1965-1969, 1972-2017	51	1986	2017	1992
50113800	Rio Cerrillos abv Lago Cerrillos nr Ponce, PR	1990-2017	28	2017	1998	1996
50114000	Rio Cerrillos blw Lago Cerrillos nr Ponce, PR	1965-1986, 1991-20173	27	2017	2011	2008	Streamflows regulated 1991-2017. Five unregulated peak streamflows were higher than the peak streamflow of 2017.
50114900	Rio Portugues nr Tibes, PR	1998-2017	20	1998	2017	2008	Rank 2 peak streamflow (water year 2017) occurred October 7, 2016—not associated with Hurricane Maria.
50124200	Rio Guayanilla near Guayanilla, PR	1981-2017	37	2013	1998	2001
50126150	Rio Yauco abv Diversion Monserrate nr Yauco, PR	1978-1984, 2003-2017	22	2013	2017	2011
50136400	Rio Rosario nr Hormigueros, PR	1986-2017	32	1998	2017	1988
50138000	Rio Guanajibo nr Hormigueros, PR	1975-2017	43	1975	2017	1979
50144000	Rio Grande de Anasco nr San Sebastian, PR	1964-2017	54	2017	1998	1975
50147800	Rio Culebrinas at Hwy 404 nr Moca, PR	1968-2017	50	2017	1988	1975
50148890	Rio Culebrinas at Margarita Damsite nr Aguada, PR	1998-2017	20	2017	1998	2000	No peak streamflow published for 1996 water year.

¹

Period of record reflects water years for which annual peak streamflow values are published. A water year is defined as the 12-month period from October 1 to September 30 and is designated by the calendar year in which it ends.

²

The stated number of annual peak streamflows used may not coincide with period of record, as indicated for specific sites.

³

Period of record includes unregulated and regulated flows. The number of annual peak streamflows used and the period-of-record rank for the Hurricane Maria peak streamflow are based on the regulated period, which includes the 2017 peak streamflow.

Water-year counts for ranks 1, 2, and 3 show, for the POR available,⁶ that Hurricane Maria peak streamflows during September 20–22, 2017, Hurricane Georges peak streamflows during September 21–22, 1998, and Hurricane Hortense peak streamflows during September 10, 1996, compose the majority of the top three rank categories (91 of 174 values, or 52.3 percent).⁷ Figure 10 shows the POR rank 1–3 occurrences for the three flood events. Hurricane Maria flooding generally exceeded that of both Hurricane Georges and Hurricane Hortense in terms of POR rank 1–3 occurrences (the only exception is rank 3, for which the number of Hurricane Maria and Hurricane Hortense occurrences is equal).

Results of this comparative analysis are limited by the relatively short PORs available for the 58 streamgages listed in table 8. Of the 58 sites, 33 have annual peak streamflows prior to 1980, but only 7 have annual peak streamflows prior to 1960. Of the seven sites with annual peak streamflows prior to 1960, three have published peak streamflows for the major hurricanes of 1899, 1928, or 1932:

• • Station 50029000 Río Grande de Arecibo at Central Cambalache, Puerto Rico (regulated 1955–2017) has published peak streamflows for Hurricane San Ciriaco in 1899 (195,000 ft³/s), Hurricane San Felipe II in 1928 (105,000 ft³/s), and Hurricane Maria in 2017 (187,000 ft³/s). Undetermined effects of regulation on the Hurricane Maria peak streamflow may preclude direct comparison with those in 1899 and 1928.

• • Station 50038100 Río Grande de Manatí at Highway 2 near Manatí, Puerto Rico (no regulation) has published peak streamflows for Hurricane San Felipe II in 1928 (148,000 ft³/s), Hurricane San Ciprián in 1932 (134,000 ft³/s), and Hurricane Maria in 2017 (167,000 ft³/s).

• • Station 50046000 Río de La Plata at Highway 2 near Toa Alta, Puerto Rico (regulated 1974–2017) has published peak streamflows for Hurricane San Ciriaco in 1899 (157,000 ft³/s), Hurricane San Felipe II in 1928 (120,000 ft³/s), and Hurricane Maria in 2017 (150,000 ft³/s). Undetermined effects of regulation on the Hurricane Maria peak streamflow may preclude direct comparison with those in 1899 and 1928.

Comparison of available historical peak streamflows is evidence that Hurricane Maria flooding was the largest in recent Puerto Rico history, likely since the 1960s. However, the hydrologic significance of Hurricane Maria flooding relative to extreme floods that occurred prior to the 1960s cannot be assessed effectively because of the general lack of historical peak-streamflow information for those floods.

Maximum Peak Streamflow Envelope Curve for Puerto Rico

Crippen and Bue (1977) developed maximum peak streamflow envelope curves for the conterminous United States; the curves are grouped and delineated by regions, generally based on physiography and variations in rainfall intensity. These envelope curves were derived using maximum observed peak streamflows from a selected set of streamgages across the country. Crippen and Bue (1977) limited their study to extreme floods (maximum observed peak streamflows), without reference to their frequency of occurrence. Envelope curves, constructed using graphs of site drainage area versus maximum observed peak streamflow, thereby provide reasonable upper limits for estimates of maximum floods in each region. Observed peak streamflows also can be normalized by site drainage area to provide a numerical means of comparing peak streamflows for sites with different drainage areas. Unit-peak streamflow is defined as peak streamflow divided by site drainage area.

Puerto Rico was not included in the report by Crippen and Bue (1977); however, data compiled as part of this study provided the opportunity to construct an envelope curve for Puerto Rico. Following the methodology described by Crippen and Bue (1977), maximum observed peak streamflows for 119 active streamgages, 56 inactive streamgages, and 6 miscellaneous sites in Puerto Rico (all unaffected by regulation, urbanization, or karst) were available for analysis; drainage areas ranged from 0.25 to 208 mi² (appendix 6). Sites with drainage areas less than 0.2 mi² were excluded from the analysis because, as noted by Crippen and Bue (1977), streamflow patterns for such small basins can be dominated by unique conditions not likely to be recognized in larger basins.

Site drainage area versus maximum peak streamflow is plotted for all 181 unregulated sites with one or more years of published peak streamflows in figure 11. The envelope curve, fitted visually, represents an approximate upper limit for peak streamflows in Puerto Rico as a function of their respective drainage areas. Inspection of the envelope curve indicates that Hurricane Maria peak streamflows reflect a smaller range of drainage areas (4.73–134 mi²) than that for the entire dataset (0.25–208 mi²). It is notable that most Hurricane Maria peak streamflows lie below the envelope curve defined by all observed maximum peak streamflows. Only the Hurricane Maria peak streamflow for station 50035000 Río Grande de Manatí at Ciales, Puerto Rico provides definition for the curve, at its uppermost end. (The remarks in appendix 6 identify peak streamflows that define various parts of the curve.) This finding may appear to contradict results of the comparison of Hurricane Maria flooding to past hurricanes in Puerto Rico, presented in the previous section. However, that analysis was conducted on POR ranks of peak streamflows for streamgages having 20 or more years of record, and without consideration of site drainage area. In contrast, the envelope-curve analysis has no requirement for long-term continuity of streamgage operation, as reflected by the inclusion of all streamgages in figure 11 that have one or more years of peak-streamflow data. Moreover, the envelope-curve analysis effectively normalizes the observed POR maximum peak streamflows based on their respective drainage areas.

Unit-peak streamflow is a logical metric for characterizing peak streamflows generated at stream locations with different drainage areas and effectively summarizes the graphical information in figure 11. Figure 12 characterizes POR maximum (rank 1) unit-peak streamflows for all 181 peak streamflows listed in appendix 6, the subset of 161 non-Hurricane Maria unit-peak streamflows, and the subset of 20 Hurricane Maria unit-peak streamflows; table 9 provides dataset statistics in tabular form. Figure 12 and table 9 are meant to summarize characteristics of Hurricane Maria POR maximum unit-peak streamflows in the context of POR maximum unit-peak streamflows that have occurred historically in Puerto Rico and do not infer any direct statistical comparison of the datasets.

The curve was developed specifically for Puerto Rico and may or may not be applicable for the U.S. Virgin Islands and (or) other islands in the Greater Antilles; further analysis of island physiography and regional meteorological conditions would be needed to determine whether broader application of the Puerto Rico envelope curve is reasonable. Finally, future floods in Puerto Rico may lie outside the curve developed in this study. Peak streamflows for such extreme floods could be plotted on the graph in figure 11 and used to develop a new curve that supersedes this one.

Conversion of the USGS Hydrologic Network to PRVD02

From March 2018 to December 2019, USGS field crews conducted GNSS surveys at all sites in the Puerto Rico hydrologic monitoring network to determine and publish gage elevations relative to PRVD02. Some of these surveys were conducted in conjunction with SBA surveys used for post-Hurricane Maria streamgage rating curve development. Additional standalone surveys were conducted at all remaining stations in the Puerto Rico network so that the entire network could be referenced to PRVD02. Historically, USGS monitoring stations were referenced to a designated “gage datum” and (or) Puerto Rico mean sea level (selected relative to local topography and, for streamgages, with consideration of stream-channel depth at the site). The approximate true elevation of the station, based on contour elevations in PRVD02/North American Datum of 1983 (NAD 83) 1:24,000-scale topographic maps, often was published along with the local gage datum. In recent years, the USGS has prioritized the nationwide conversion of gage datums to a vertical datum consistent with the National Spatial Reference System through technologies such as Survey-Grade GNSS systems (Rydlund and Densmore, 2012). The USGS prioritized conversion of gage datums to PRVD02 as part of the network restoration activities conducted after the passage of Hurricane Maria.

The GNSS-surveying campaigns were conducted using two approaches: GNSS static and (or) Real-Time Network (RTN) surveys. RTN surveying is similar in concept to RTK surveying discussed earlier, as both GNSS-surveying methods use communication with a reference station to make real-time corrections to the data being collected by a rover station, which is a portable device that collects positioning data relative to a base station. Unlike RTK surveying, for which the reference station is physically located at a monumented or non-monumented benchmark, RTN surveying uses Continuously Operating Reference Stations (CORS) (USGS, 2020a), which continuously sends positioning information to a centralized server. A virtual reference station is created from these CORS, and real-time positioning data (raw data) are properly adjusted (satellite orbit, clock error, among other factors). These adjustments are made through a connection to the internet by linking the rover network to the CORS. The RTN-surveying campaign conducted for the USGS sites in Puerto Rico included checking monumented benchmarks at the start and end of each day to ensure vertical accuracy.

The GNSS-static and RTN-GNSS surveying methods were both completed according to USGS level-quality standards and protocols documented in Rydlund and Densmore (2012). The accepted minimum accuracy standard for the GNSS surveys conducted in Puerto Rico was Level II quality (accuracy ±0.26 ft); however, higher accuracy surveys were conducted at some sites.

Real-Time, High-Resolution Monitoring of Spillway Damage of the Lago Guajataca Dam

USGS streamgage 50010800 Lago Guajataca at damsite near Quebradillas is situated at the Lago Guajataca Dam (plate 1, index number 3 near northwestern corner of Puerto Rico), which impounds the main source of water supply for the northwestern part of Puerto Rico. During the passage of Hurricane Maria, severe flooding occurred in the Río Guajataca watershed, causing uncontrolled flow over the dam spillway and erosion of the spillway structure, making the dam vulnerable to potential failure (Carrasco, 2019). Local and Federal agencies immediately initiated a cooperative effort to address the emergency and mitigate erosion processes threatening the dam. In particular, the Puerto Rico Electric Power Authority and U.S. Army Corps of Engineers needed real-time information to implement temporary dam-safety measures, which required continuous monitoring of the spillway structure and its integrity (USGS, 2017; Carrasco, 2019).

USGS personnel, along with community volunteers, assisted in the logistics and installation of a telemetered high-resolution camera that collected targeted images of the affected area at 30-minute intervals from October to December 2017. Local and Federal agencies were able to access an internet webpage and view near-live imagery of spillway head-cut erosion and progress of mitigation efforts. Figure 13 shows a photograph captured during the emergency (fig. 13A) and a photograph of the components of the high-resolution camera equipment used to collect images (fig. 13B). The image in figure 14 is a panoramic view of the affected area showing the dam, spillway structures, and area of head-cut erosion.