TransCanada wanted more accurate estimates of the probable maximum flood (PMF) and the probable maximum precipitation (PMP) at four hydroelectric projects in the Upper Connecticut River Basin. A site-specific study showed significantly lower estimates than those produced by a generalized estimate for the basin.
By Damian M. Gomez, Jerry A. Gomez, and Justin Donaghy
In the course of designing and evaluating the safety of dams, it is often necessary to make an estimate of the probable maximum flood (PMF). To make such an estimate, it is first required to have an accurate estimate of the probable maximum precipitation (PMP) for input into a hydrologic model.
In most of the eastern U.S., the PMP is estimated based on guidance given by Hydrometeorological Report (HMR) Numbers 51 and 52, published by the National Oceanic and Atmospheric Administration. HMRs 51 and 52 produce generalized PMP estimates that may be deficient in certain regions where terrain has a pronounced effect on rainfall.
These regions, known as the stippled regions, include the Appalachian Mountains and a strip between the 103rd and 105th meridians, where upwind moisture barriers can cause a significant reduction in precipitation depth. Alternatively, some mountainous basins can experience an orographic increase in rainfall due to their high elevation. Often both of these phenomena play a role in the PMP. For areas within the stippled regions, HMR 51 suggests that the PMP be considered on a case by case basis, known as a site-specific study.
Within the northeastern U.S., four site-specific PMP studies have been performed prior to this study that have been approved by the Federal Energy Regulatory Commission (FERC). These previous studies were performed for Harriman Dam, 1987; FPL Upper and Middle Dams, 2002; Great Sacandaga Lake/Stewart’s Bridge Dam, 2003; and Blenheim Gilboa Pumped Storage Project, 2008. Each of these studies found that a site-specific methodology produced a lower estimate of PMP than the generalized estimate based on HMR 51.
Each of these previous site-specific studies was performed under the review of a board of consultants, ensuring that the methodologies used are consistent with those utilized in HMR 51 and other PMP estimations. These studies laid the groundwork for future site-specific PMP analyses by investigating and developing datasets for extreme storms which are transposed to the basin of interest to build the PMP envelope.
TransCanada owns six hydroelectric facilities on the Connecticut River, four of which are in the Upper Connecticut Basin: Moore, Comerford, McIndoes Falls and Wilder projects.
At each of these four sites, it was desired to have a more accurate estimate of the PMF and the ability of the structures to safely pass that flow. To this end, a site-specific PMP/PMF study was undertaken for these sites.
Due to the geographic distribution of the sites, two separate PMP estimates and subsequent PMF analyses were undertaken. One PMP estimate was made for the Moore and Comerford drainage basins. The second estimate was made for the Wilder drainage basin, and was also utilized for McIndoes Falls. The PMP/PMF analysis was completed for the Moore and Comerford sites before beginning the analysis for the McIndoes Falls and Wilder sites.
Site-specific PMP development
The site-specific PMP follows the same methodologies used in developing the generalized PMP estimates for the various hydrometeorological reports and the site-specific PMPs for the studies previously mentioned.
This process begins with identifying historic extreme storms which could reasonably occur in the basin of interest. For each of the identified short list storms, a series of adjustments is made to maximize and transpose the storm to the basin. The maximization and transposition adjustments are made based on the amount of available moisture, which is dependent on the dewpoint or, as a surrogate, the sea surface temperature. The orographic adjustment factor is based on correlations of historic rainfall versus elevation.
The transposition and maximization process utilizes various moisture ratios to construct a depth-area-duration table for a storm that is functionally similar to the short list storm, but occurs within the basin of interest and under the extreme of expected meteorological conditions. The maximization, transposition, barrier and elevation adjustment factors account for changes in storm moisture content due to dewpoint, moisture source, upwind removal and depletion due to elevation rise. These four adjustment factors follow the generalized PMP methodology.
In developing generalized PMP estimates, terrain effects are evaluated over large regions. This reduces their potential effect, particularly in the designated stippled regions. For site-specific PMP studies, the effect of the terrain features is studied in greater detail. For each of the transposed storms, an upwind moisture barrier was identified and the effective elevation was evaluated. This barrier elevation affects both the barrier adjustment factor and the orographic adjustment factor.
The adjusted depth-area-durations, produced by the transposition and maximization process, were utilized in developing the site-specific PMP. Envelope curves of transpositioned storms were produced to estimate the PMP tables.
In comparison to the estimates given by HMR 51, the site-specific PMP analysis gives lower estimates of precipitation for both study sites.
FIGURE 1 Upper Connecticut River Basin
|Vermont’s Upper Connecticut River Basin, where TransCanada performed a site-specific study of probable maximum flood (PMF) and probable maximum precipitation (PMP) estimates for several hydroelectric projects.|
All-season PMF development
The PMF was estimated from the PMP using a calibrated hydrologic model of the basin. GIS methods were used in the initial development of the model, while calibration routines were used in refining the model. The HMR 52 computer program was used to spatially distribute the PMP for use in the PMF simulation, and three storm centers were used to find the critical storm.
At both the Moore and Comerford Dams, the hydrologic model provided an estimated peak outflow of about 123,000 cubic feet per second (cfs) for the all-season PMF. The flood of record for these two sites occurred in March 1936 with an estimated peak flow of 50,000 cfs at Moore Dam and 55,000 cfs at Comerford Dam. Since construction of these projects, the highest flow was recorded on April 1, 1998, with a peak flow of 42,000 cfs measured at the USGS gage at Dalton, NH.
At the McIndoes Falls and Wilder Dams, the model produced estimated all-season PMF flows of about 166,000 cfs and 208,000 cfs respectively. At these sites, the flood of record is the March 1936 flood, with recorded peak flows of about 68,000 cfs and 91,000 cfs at McIndoes Falls and Wilder respectively.
It is expected in most cases that the PMF will be significantly higher than any recorded flood at the site. At these sites, the PMF predictions were significantly lower than those provided by prior estimates utilizing generalized PMP estimates and were not as high compared to floods of record as might generally be expected. In addition, the floods of record for these sites were associated with a combination of rainfall and snowmelt. Due to these factors, an investigation of a rain-on-snow PMF event was undertaken.
Rain-on-snow screening analysis
For prior site-specific PMF studies, where rain on snow has been considered, it has been ruled out as a controlling event through a screening level analysis.
Such an analysis was performed for the Moore/Comerford drainage basin, conservatively assuming that a seasonally adjusted PMP occurs coincident with a 100-year snowpack, causing it to melt entirely. This analysis can effectively eliminate the rain-on-snow event as controlling the PMF if the upper limit it defines is lower than the all-season PMF.
For this screening analysis, monthly ratios contained in HMR 33 were used to develop the seasonal PMP from the site-specific all-season PMP. Snowpack records were used to determine during which month the worst case rain-on-snow event could occur.
For the basin, the snow water content typically peaks near the end of March or beginning of April. Since the April seasonal PMP is higher than March, this was used for the screening analysis. The April seasonal PMP is estimated to be 64 percent of the all-season PMP, based on the HMR 33 ratios.
For this basin, the 100-year snowpack was estimated to have an average water content of 16.5 inches based on a Cornell University study. For the screening analysis, the snowmelt was assumed to occur with the same distribution as the rainfall. The combined rainfall and snowmelt was then input as precipitation to the hydrologic model. The resulting outflows predicted by the model were higher than the flows predicted by the all-season PMF runs. Therefore, the screening analysis failed to rule out the rain-on-snow event as the critical PMF for the Moore and Comerford sites and a full rain-on-snow analysis was performed.
FIGURE 2 Moore Reservoir Inflow
|This chart compares the inflow probable maximum flood (PMF) hydrograph to Moore Reservoir produced from model runs for the Hydrometeorological Report (HMR) 51 probable maximum precipitation (PMP) and the site-specific all-season PMP and the site-specific all-season PMP and seasonal PMP on snow.|
In-depth rain-on-snow analysis
The full rain-on-snow analysis required calibration of the snowmelt equation, re-calibration of the hydrologic model to cold season conditions, and identification of a maximum coincident wind and temperature series. For the Moore/Comerford drainage basin, three rain-on-snow events were used in calibrating the snowmelt and hydrologic models.
FERC, the governing body for these projects, prefers that the energy budget method be utilized for estimating snowmelt. That method was utilized for these projects.
The energy budget equations can take several forms, depending on the assumptions made regarding the forest cover at the location and the precipitation occurring during the snowmelt. For the Moore/Comerford drainage basin, snowmelt equations applicable to areas with 10- to- 60-percent forest cover were utilized. These same equations are utilized in the HEC-1 hydrologic model.
Within HEC-1, it is implicitly assumed that the forest cover equals 0.5, the basin wind exposure coefficient equals 0.6 and the basin shortwave radiation melt factor equals 1. Since the HEC-HMS hydrologic model does not include the energy budget snowmelt, these computations were performed outside the hydrologic model allowing the forest cover, basin exposure coefficient and shortwave melt factor to be varied.
The forest cover and shortwave melt factor were estimated based on measured properties of each sub-basin. The basin exposure coefficient was calibrated to best estimate measured snowmelt for three historic storms.
There are few weather stations near the basin that collect all of the data necessary to perform the computation of snowmelt, particularly wind speed, dewpoint temperature and solar radiation. Data for these parameters was acquired from the National Oceanic and Atmospheric Administration’s (NOAA) National Center for Environmental Prediction North American Regional Reanalysis (NCEP NARR). The NARR data is produced from an analysis/forecast model which estimates environmental variables from measured data and weather patterns every three hours at a 0.3 degree (latitude/longitude) grid.
Hourly precipitation data was available from NOAA for several stations throughout the basin. Due to the spatial coverage of precipitation stations and direct measurement of precipitation, the NOAA station data was considered a better estimate of precipitation throughout the basin than NARR data.
Rain-on-snow PMF development
The computed snowmelt and the measured rainfall were input to the hydrologic model for re-calibration to cold season events. For the Moore/Comerford basin, it was found that during the studied cold season events, a greater percentage of the precipitation and snowmelt ran off than did from the studied warm season events. Additionally, it was found that the basin demonstrated a somewhat longer “apparent” lag time for these events. For the Moore/Comerford hydrologic model, Dalton is the only calibrated sub-basin and contains nearly 93 percent of the drainage for the sites.
Calibrated curve numbers for the basin under cold season events were very close to antecedent moisture condition (AMC) III curve numbers for the soils in the basin, so much so that AMC III curve numbers were utilized in simulating the rain on snow PMF. The calibrated curve numbers for warm-season event were extremely close to AMC II curve numbers for the soils in the basin, thus the AMC II curve numbers were used in simulating the all-season PMF.
In order to estimate the snowmelt occurring during the seasonal PMP, estimates of wind speed, temperature, dewpoint and solar radiation must be made. In estimating these conditions, it is unreasonable to expect that the highest wind and temperature would occur at the same time and during the heaviest rain likely for the month.
The FERC Engineering Guidelines suggest that the peak wind and temperature series be chosen from historic records. Also, it was decided that only wind and temperature records should be considered for periods of significant rainfall with a snowpack on the ground. This is in recognition that high temperatures in March and April are unlikely to occur during heavy rainfall and that snow cover may affect temperatures.
The maximum recorded wind and temperature series was found to occur from March 29, 1987, at 7 a.m. to April 1, 1987, at 7 a.m. Once identified, the maximum wind and temperature series was paired with the seasonal PMP to produce the PMP snowmelt.
The combined seasonal PMP and resulting snowmelt was input into the hydrologic model in order to estimate the rain on snow PMF. The model produced an estimated peak discharge of about 145,000 cfs at both Moore and Comerford Dams. This result indicates that a rain-on-snow event controls the PMF for these two sites.
Results and conclusions
While the results of the rain on snow analysis are higher than those of the all-season PMF analysis, they were significantly lower than the results of the screening level analysis. For this basin, the 100-year snowpack did not limit the amount of snowmelt.
The PMF estimates that have been produced so far by this study have been significantly lower than those which would result from using generalized PMP estimates for the basin.
In the parts of the western U.S., where much of the water supply is derived from melting snowpack, the National Resource Conservation Service (NRCS) maintains measured snow courses as part of their study of snowpack and melt. Such sources of data are limited in the Northeast.
Fortunately, TransCanada has maintained decent records of snowpack and snow water equivalent within the basin to help predict spring runoff. Even so, the frequency of measurement meant that for some studied events it was difficult to determine exactly when and at what rate snowmelt occurred. To aid in correctly timing the snowmelt, measurements of the depth of snow on the ground were used.
Direct measurements of solar radiation, wind speed and dewpoint are quite scarce. To overcome this deficiency, NARR data, which is derived from meteorological models, was utilized. A benefit of the NARR data is that the models, from which the data are derived, are intended to account for spatial variation due to weather patterns, which cannot be accounted for in traditional spatial averaging techniques such as theissen polygon. To date, all of the FERC approved site-specific PMP studies, which have been performed in the stippled regions of the Northeast, have shown the generalized PMP estimates to be overly conservative in the basins studied. As a point of comparison, this site-specific study indicated a reduction in the 72-hour, 1,000-square mile PMP value of 25 percent from the HMR 51 value and the other referenced studies found a reduction ranging from 25 to 34 percent.
These reductions are influenced not only by the geographic location of the basin, but also by the effectiveness of topographic barriers between the basin and storm sources. For this site, an average reduction in precipitation for the short list storms of approximately 24 percent resulted from the upwind barriers.
The refined estimate of PMF from a site-specific PMP/PMF study can have significant implications for the dam owner.
In this case, previous hydrologic/hydraulic studies indicated that the Moore and Comerford Dams could not safely pass the PMF without costly modifications, whereas the site-specific PMP/PMF study indicates the projects can safely pass the PMF without modification.
Damian M. Gomez, P.E., a water resources engineer for Gomez and Sullivan Engineers, is the project engineer on TransCanada’s Site-Specific PMP/PMF project in the Upper Connecticut River Basin. Jerry A. Gomez, P.E., president of Gomez and Sullivan Engineers, is the principal in charge of the TransCanada project. Justin Donaghy, P.E., is project manager and reviewer for TransCanada Hydro Northeast.