The United States National Weather Service Real-Time Flood Forecasting
Summary and Keywords
The U.S. National Weather Service (NWS) is the agency responsible for flood forecasting. Operational flow forecasting at the NWS is carried out at the 13 river forecasting centers for main river flows. Flash floods, which occur in small localized areas, are forecast at the 122 weather forecast offices.
Real-time flood forecasting is a complex process that requires the acquisition and quality control of remotely sensed and ground-based observations, weather and climate forecasts, and operation of reservoirs, water diversions, and returns. Currently used remote-sense observations for operational hydrologic forecasts include satellite observations of precipitation, temperature, snow cover, radar observations of precipitation, and airborne observations of snow water equivalent. Ground-based observations include point precipitation, temperature, snow water equivalent, soil moisture and temperature, river stages, and discharge. Observations are collected by a number of federal, state, municipal, tribal and private entities, and transmitted to the NWS on a daily basis.
Once the observations have been checked for quality, a hydrologic forecaster uses the Community Hydrologic Prediction System (CHPS), which takes care of managing the sequence of models and their corresponding data needs along river reaches. Current operational forecasting requires an interaction between the forecaster and the models, in order to adjust differences between the model predictions and the observations, thus improving the forecasts. The final step in the forecast process is the publication of forecasts.
This article describes the current practices to the development of hydrologic forecasts at the U.S. National Weather Service (NWS). It provides an overview of the process and a description of the main components of the forecasting process, and provides a number of references to allow interested readers access to more in-depth information on the different aspects, models, procedures, and components of the NWS hydrologic forecasting process.
Until 2016, the NWS forecasting services were organized around the River Forecast Centers (RFCs) and Weather Forecast Offices (WFOs), with the RFCs being responsible for forecasting major river reaches, and the WFOs being responsible for flash flood forecasts.
There are 13 RFCs and 122 WFOs across the United States. Figure 1 shows a map of the area of responsibility of the 13 RFCs, and Figure 2 shows the map of the 122 WFOs. Notice that the areas of responsibility of the RFCs lie around natural catchment boundaries, while that of the WFOs lie along political (county or parish) boundaries. This is because the WFOs are primarily responsible for the issuing of watches and warnings to the public, and for the main coordination with county and state emergency managers. Both the RFCs and the WFOs report administratively to one of the NWS regional directors.
Since the 2016 reorganization of the NWS, there is a two-prong approach. One prong of the approach continues to provide official forecasts by the RFCs at over 4,000 points at major river reaches, and minor streams by the WFOs. The other prong of the NWS forecasting service is a National Water Model, being developed at the National Water Center (NWC), which is part of the NWS Office of Water Prediction. The NWC is following a centralized forecasting approach, using a nation-wide model, to be run on one of NWS’s supercomputers. As of this writing, the model is still under development and no official forecasts are issued to the public. An experimental forecast is available at http://water.noaa.gov/tools/nwm-image-viewer As of the date of this article the only official forecasts are those issued by the RFCs; therefore, the description of the NWS hydrologic forecasting process contained herein is based solely on the RFC operational procedures.
Stallings and Wenzel (1995), Larson and colleagues (1995), and Fread and colleagues (1995) have excellent descriptions of the organizational structure of the hydrologic forecasting at the NWS. Following the devastating 1997 floods on the Red River of the North (Shelby, 2003), Congress appropriated funding for the implementation of the Advanced Hydrologic Prediction Service (AHPS). AHPS didn’t modify the organizational structure of the hydrologic forecasting services, but considerably improved the quality of the service, by providing funding to increase the number of forecasting points to the current number of about 4,000, and, as importantly, developed the Hydrologic Ensemble Forecasting System (HEFS, which will be mentioned). AHPS is described by McEnery, Ingram, Duan, Adams, and Anderson (2005). An evaluation of AHPS was conducted by the National Research Council (2006).
Organizational Structure of the River Forecast Centers
The head of each RFC is the Hydrologist-in-Charge (HIC). This person is responsible for the overall operation of the center, including administrative, personnel, and budgetary issues. It is also common that the HIC takes part in forecasting duties.
There are two deputies at each RFC. The Development and Operations Hydrologist (DOH) is responsible for the technical aspects of the forecasting process, training, and development of new tools. The second deputy is the Service Coordination Hydrologist (SCH). This is a relatively new position, created in response to requests for the RFCs to have a more active role in communicating with NWS customers, such as water resources managers, utilities, and emergency managers, etc. Most of the RFCs have an Information Technology Officer (ITO), who is in charge of maintaining the software, troubleshooting issues with the forecasting machines or web servers, etc. Those that do not have an ITO rely on the collocated WFO for software support.
The next level of personnel at the RFCs is occupied by senior hydrologic forecasters and a senior Hydrometeorological Analysis and Support (HAS) person. The senior HAS forecaster provides guidance and tutoring to junior HAS personnel. This person also participates in hydrologic forecasting duties. The senior hydrologic forecasters are responsible for providing leadership and guidance to junior forecasters, taking care of hands-on calibration and development and research of new models or RFC products. The next level of personnel includes the hydrologic forecasters and HAS personnel. They participate in the forecasting duties and data quality control, and they work with the senior forecasters on model development, calibration, and research.
Finally, the administration of the RFC, including timekeeping, travel arrangement, detailed bookkeeping to manage the budget, and assisting the entire personnel to handle administrative tasks is on the shoulders of the Administration Support Assistant.
Flow Diagram of the NWS Hydrologic Forecasting Process
The flow of information leading from collecting data to issuing forecasts to the public by the NWS is a complex process. Figure 3 shows a somewhat simplified diagram. Missing from the diagram are international sources of information and collaboration with some international agencies, notably those from Canada. The reminder of this article will focus on explaining the overall forecasting process.
Data Collection and Transmission
A critical component of a hydrologic forecasting system is the observations. The observations can be classified as “forcings” (i.e., the drivers for the hydrologic models, such as precipitation), or the river flow and stage observations that serve to verify the quality of the forecasts. Collection and quality control of observations are the responsibility of a number of federal, state, and local agencies. Furthermore, many volunteer observers across the country collect and transmit daily observations of temperature. Close collaboration among all agencies has been fundamental in the successful operation of the hydrologic forecasting at the NWS.
Rainfall is the major source of flooding in the United States, with other sources, such as snow melt, storm surge, or dam breaks, being less frequent, although not less important. In the United States, as in most temperate regions, precipitation can be either frozen (snow or hail) or liquid. While hail could be a major source of disasters for agriculture and property (roofs, vehicles), it is not a major source of flooding. Snowfall, by itself, does not contribute to immediate flooding, but its accumulation during fall and winter, followed by melt during the spring time, has caused some of the major losses in U.S. flooding, such as the disastrous 1997 flooding of the Red River of the North, which caused losses in excess of $3.5B (Shelby, 2003).
The quality of rainfall observations has been greatly enhanced since the availability of the Doppler radar network (Krajewski, Villarini, & Smith, 2010; Polger, Goldsmith, Przywarty, & Bocchieri, 1994). The Next Generation Radar Program (NEXRAD) of the U.S. Departments of Commerce, Transportation, and Defense, deployed a network of 155 Weather Surveillance Radar-1988 Doppler (WSR-88D). The radars are owned by the Department of Commerce’s NWS (122), Department of Transportation’s FAA (12), and the Department of Defense (21). This network of radars is commonly known as the “NEXRAD network.” Figure 4 shows the conterminous U.S. component of the NEXRAD and the FAA’s Terminal Doppler Weather Radar (TDWR) network. There is an abundant list of references about the WSR-88D algorithms and the most recent upgrade to the radars and algorithms, allowing for polarimetric measurements of precipitation characteristics. Some references to the original algorithms are Crum and Alberty, 1993; Fulton, Breidenbach, Seo, Miller, and O’Bannon, 1998; Johnson et al., 1998; and Kitzmiller, Miller, Fulton, and Ding, 2013. Information about the polarimetric enhancements to the WSR-88D algorithms can be found in Park, Ryzhkov, Zrnic, and Kim, 2009; Ryzhkov, Giangrande, and Schuur, 2005; and Zhang et al., 2011.
Specific applications to hydrology are discussed in Shedd and Fulton, 1993; Bonnin, Kitzmiller, Seo, Smith, and Restrepo, 2007; Kitzmiller et al., 2013; Krajewski et al., 2011; Shedd and Fulton, 1993; and Seo and Krajewski, 2010.
Ground-based precipitation observations serve to complement radar observations. They are also used for removal of biases in radar observations, and as one of the components of multisensor precipitation. The NWS uses both automated rain gauges (typically at 1-hr reading intervals), and manually reported precipitation observations on a daily basis. A large number of these daily reports come from the NWS Cooperative Observer Network (COOP); state mesonet networks; and the Community Collaborative Rain, Hail & Snow network (CoCORAHS), organized in 1998 at Colorado State University, and sponsored by the NOAA and the National Science Foundation of the United States. At the time of this writing, CoCORAHS networks are available in the United States, Canada, and the Bahamas.
Quality control of precipitation is of the utmost importance for weather, climate, and flow forecasts. This quality-control process is partly automated, but, given the natural spatial variability inherent in precipitation, a manual quality control (QC) procedure complements the automatic one (see data quality control).
Satellite-based sensors provide another source of precipitation observations. In the United States, satellite-based precipitation estimates are used primarily over areas not well covered by the radar network in the United States, or over tributary basins in Mexico, such as the Río Conchos, a tributary of the Rio Grande.
The RFCs use a number of sources of temperature observations. Notable among them are the NWS Cooperative Observer Network, the CoCORAHS network, the Climate Reference Network, and several state-operated mesonets. More recently, NCEP Central Operations is distributing the Real Time Mesoscale Analysis grids (RTMA). These are high-resolution grids (5 and 2.5 km across the conterminous United States, and 6 km for Alaska). These grids provide high-quality hourly analysis of near-surface weather conditions. RFCs are now able to use these temperature estimate grids to compute mean area temperature for streamflow forecasting.
Snow Depth and Water Equivalent
The NWS RFCs currently use two snow properties as input to flood forecasting models: snow fall (i.e., the area within a watershed that is covered in snow), and the snow water equivalent (SWE) that indicates how much water is available in the snow). Other properties of the snow, such as snow pack temperature, snow depth, and snow cover area are not needed by the models used at the NWS RFCs, although they are typically consulted. There are a number of sources for snow observations: SWE in-situ measurements by NWS Weather Forecast Offices and collaborating agencies, notably the U.S. Army Corps of Engineers and the USGS; airborne SWE estimates from Gamma-Ray detectors on-board NOAA aircraft; USDA SNOTEL sites; and satellite observations of snow-covered areas. Furthermore, in addition to managing the airborne survey, the National Operational Hydrologic Remote Sensing Center (NOHRSC), a branch of the NWS’s National Water Center in Chanhassen, MN, produces National Snow Analyses covering the United States and Canada (up to 53° N Latitude), which are periodically consulted by the RFCs in preparation for their forecasts. Figure 5 shows the SWE estimate corresponding to March 12, 2013.
River Discharge and Stage
River discharge and stage observations in the United States are compiled and distributed by the United States Geological Survey (USGS). The observations are taken in collaboration with various federal, state, local and tribal agencies, and subject to rigorous quality control by the USGS. Furthermore, the NWS relies very heavily on the USGS for additional observations during high-flow conditions, especially when the river level is above the rating curve. Upon request by the NWS, the USGS will send crews to measure flows even during flooding conditions.
River regulation presents some of the most difficult factors involved in streamflow forecasting. According to the National Inventory of Dams of the U.S. Army Corps of Engineers, there are over 87,000 dams in the United States. Most of these dams are small (49% lower than 25 ft, and 43% between 25 ft and 50 ft), and, therefore, most likely present a very small regulation capacity (regulation is the ability of dams with large associated reservoirs to modify considerably the natural flow regime, which is a challenge for hydrologic forecasting). Many dam operators share their operation plan freely with the NWS. Some share it under a confidentiality agreement, while some, including some government agencies, do not share their operating plans at all. In some cases, common in hydropower utilities, a reason for not sharing release plans is due to the competitive nature of power generation in the United States, which has the potential for competing generators to take advantage of that information to the detriment of the disclosing utility. But even when the NWS has access to the planned operational releases, those plans may change due to a number of reasons. For instance, unplanned transmission lines or generating station outages or changes in the spot pricing of power sales may induce considerable unforeseen changes in the operational releases.
In many cases, there is a need to iterate between the NWS forecasters and the reservoir operators. The latter may need information on forecast inflows to the reservoirs in order to decide on a release schedule. The NWS forecasters send a preliminary forecast based on natural inflows and expected releases, and communicate that information to the reservoir operators. Once the reservoir operators enter that inflow forecast into their operational procedure, a new release schedule is produced and forwarded to the NWS that prepares a final forecast.
Weather and Climate Forecasts
In addition to weather, water, and snow observations, weather and climate forecasts are vital for the production of hydrologic forecasts. Responsibility for the precipitation and temperature forecasts used by the hydrologic forecasting models, (0–15 days), falls to the Weather Prediction Center (WPC) of the National Centers for Environmental Prediction (NCEP), and to the Climate Prediction Center (CPC) for the corresponding forecasts over 15 days.
National Centers for Environmental Prediction
The National Centers for Environmental Prediction (NCEP) comprise nine centers in addition to the Director’s office. In addition to the WPC and CPC already mentioned, NCEP centers include the Aviation Prediction Center; the National Hurricane Center; NCEP Central Operations, which runs the numerical analyses and forecasts models; the Ocean Prediction Center; the Space Weather Prediction Center; and the Storm Prediction Center, which focuses on severe weather.
Storm Surge Forecasts
Storm surges are an important factor in severe flooding in coastal areas. Total water level forecasts used by the RFCs are issued by NCEP’s Ocean Prediction Center. There are two models available. The first model is the Extratropical Storm Surge model (ET-SURGE or ETSS), which is a model derived by the NWS Meteorological Development Laboratory from the original SLOSH model (Sea, Lake and Overland Surges from Hurricanes; Jelesnianski, Chen, & Shaffer, 1992). The model is driven by NCEP’s Global Forecasting System forecasts of winds and pressure. Additional information on the Meteorological Development Laboratory can be found at http://www.nws.noaa.gov/mdl/etsurge/.
The second model is the Extratropical Surge and Tide Operational Forecast System (ESTOFS), developed in collaboration between NOAA’s National Ocean Survey Coast Survey Development Laboratory and NCEP. It uses the ADCIRC model. Initial coarse resolutionwe Oregon State University’s TPXO* model.
Data Quality Control
A strict control of the data used in hydrologic forecasting is important, not only because of its impact on the quality of the forecast, but also because that data will become part of the climate record. The RFCs are responsible for the quality control of precipitation observations.
At the NWS, hourly and daily rain gauge observations of precipitation are used to remove biases from radar observations. Observations from the state mesonets and the CoCoRAHS network are quality controlled at their respective sources. Observations from the COOP network are quality controlled by the WFOs and RFCs. The first step in correcting the precipitation errors is the responsibility of the WFO where the observers are located. The RFCs double-check the accuracy of precipitation observations by using either XNAV, a tool developed at the Arkansas-Red Basin RFC (ABRFC); or Mountain Mapper (Schaake, Henkel, & Cong, 2004). The graphical tool displays the gridded information resulting from the merged radar, hourly rain gauge and satellite information, or Multi-Sensor Precipitation Estimator (MPE). Daily totals observations, in inches, are displayed at the gauge location, using the same color scale as that of the gridded information. Figure 6 shows an actual example from the North Central River Forecast Center (NCRFC). Gridded information corresponds to 24-hour precipitation accumulation, and the gauge information is shown numerically. The areas encircled in ovals indicate those gauges in which the 24-hour observation report was considerably different from the gridded MPE values.
Figure 7 shows the estimates of mean areal precipitation for August 21, 2016, in a portion of the NCRFC in Illinois before corrections of precipitation. The upper basin highlighted in red (Rock River at Rockton, IL) had an estimate of 0.378" Mean Areal Precipitation (MAP) before corrections. The lower basin (Rock River at Dixon, IL) had an estimated 0.123" of MAP, also before corrections. Figure 8 shows the results of the data correction. At Rockton, the MAP went down to 0.178", and at Dixon, it went down to 0.034".
The NWS River Forecasting System (NWSRFS)
The NWSRFS (Larson, 2002) was a forecasting system initially developed in the 1970s at the then-named Hydrologic Research Lab of the Office of Hydrology of the NWS. The system was well designed and engineered to make optimum use of the computers of that era, which were limited in internal memory as well as in processing speed. The software techniques used to minimize the memory usage and improve computational speed were very effective, but at the expense of maintainability and support. Often, adding a small change to NWSRFS implied that some parts of the system stopped working as originally intended. At different times, it operated on mainframe computers as well as on individual workstations. However, as hardware and software advanced, increasing the memory capacity and processing speed of modern computers, it became a priority to improve the extensibility and maintainability of the system, rather than minimizing memory use or improving processing speed. At the same time it became increasingly difficult to enhance the system to take advantage of the newer tools, both in terms of software and hardware, available to the users.
Community Hydrologic Prediction System (CHPS)
Starting at around 2004, the Hydrology Laboratory of the Office of Hydrologic Development started a search for a suitable substitute to replace NWSRFS. Together with the River Forecast Centers, it developed a number of requirements that a new hydrologic forecasting system, to be known as the Community Hydrologic Prediction System (CHPS), should meet. An extensive search of possible candidates eventually focused on Deltares’ Flood Early Warning System (FEWS) as the basis for CHPS.
FEWS is a widely used flow forecasting system. It is an open-source system, well-supported by its development organization in the Netherlands, Deltares. CHPS documentation can be accessed at: http://www.nws.noaa.gov/ohd/hrl/general/indexdoc.htm.
An important requirement for CHPS was the ability to operate in a client-server mode, to allow parties outside of the RFCs to run models or to, at least, examine the model results. Other parties include the WFOs, other federal agencies such as the U.S. Army Corps of Engineers, or local agencies such as county offices. A proof-of concept demonstration was carried out involving Yuba County Water Agency in California, the U.S. Army Corps of Engineers, and the California-Nevada River Forecast Center. This proof-of-concept was based on demonstrating the ability to run the Corps reservoir operations model, ResSim, within an operational hydrologic forecasting system.
The Office of Hydrologic Development chartered a CHPS Acceleration Team (CAT) to steer the CHPS development effort. A chronology of the CHPS development up to the acceptance of the ResSim demonstration system is available at http://www.nws.noaa.gov/ohd/hrl/chps/ResSim_proj.html.
CHPS is essentially FEWS, plus the NWS models and databases. Deltares maintains the FEWS wikipage, and periodically updates the list of models under FEWS, which includes those models of the NWS.
Deltares worked under contract with the NWS to enhance FEWS, such that the NWS’ requirements for CHPS and the NWS models not existing in FEWS could be incorporated. A notable requirement for CHPS that needed to be included in FEWS was the ability to do manual modifications to the model results. In the NWS, those modifications are commonly known as the “Mods,” and they are essentially a rudimentary form of data assimilation. More detail on the Mods may be found under “Forecaster modifications” in the next section.
The original NWSRFS had a suite of 44 “operations” that included hydrologic, hydraulic, and routing models, as well as time series manipulation. Not all the models or operations were used by all RFCs, and only those that were actually used were ported to CHPS. This section describes the main models.
Soil Moisture Accounting
The core of the NWS hydrologic forecast process is the Sacramento Soil Moisture Accounting model (SAC-SMA; Burnash, Ferral, & McGuire, 1973; Burnash, 1995). The NWS has an online documentation of the model, including a description of the components and the governing equations. Despite its age, the Sacramento model performance continues to be unsurpassed in intercomparison studies. Reed and colleagues (2004) and Smith and colleagues (2004, 2012) describe the results of the Distributed Model Intercomparison Projects (DMIP and DMIP-2).
The SAC-SMA belongs to the “conceptual” class of hydrologic models. In these models, the physics of the movement of water through the sub-soil are simplified in a way that the soil is divided into a series of “tanks” or “reservoirs,” with heuristic equations governing the transfer of soil moisture from the surface to the tanks, among the tanks, and to the outlet. Figure 9 shows a schematic diagram of the original Sacramento model.
The model considers the soil to be divided into upper and lower zones. The upper zone has two reservoirs: one for tension water, (water that cannot be depleted except for evaporation); and free water, which is the source of interflow and also drains to the lower zone. Precipitation input is divided between that falling on impervious areas, causing direct runoff, and that falling into pervious areas thus infiltrating into the upper zone tension reservoir. Once the upper zone tension water is saturated, infiltration proceeds to the upper free zone reservoir, which immediately produces interflow and percolation to the lower zone. If the upper zone free water reservoir gets saturated, the model produces surface runoff.
The percolation function is considered to be the heart of the Sacramento model. It is an exponential function that controls the rate of percolation based on the relative saturation of the lower zone and the upper zone. Some of the percolation goes directly to the free water reservoirs, as controlled by a parameter that needs to be calibrated. These two reservoirs control the groundwater simulation of the model. The primary reservoir corresponds to base flow, and the supplementary reservoir is a faster draining reservoir to allow for non-linearity in the observed groundwater flow. Interflow, primary, and supplementary base flows are controlled by a first-order linear differential equation, such that those discharges are directly proportional to the contents of the reservoirs. The model allows for groundwater losses to other watersheds (SIDE), as well as channel losses to subsurface discharge. Evapotranspiration is controlled by the potential evapotranspiration input to the model, and it is satisfied from the upper layers to the bottom layers until the potential evapotranspiration demand is met.
Researchers at the Office of Hydrologic Development made some considerable enhancements to the model, including a procedure for estimating a priori parameter values based on soil types and land cover (Anderson, Koren, & Reed, 2006; Koren, Smith, Wang, & Zhang, 2000; Smith, Laurine, Koren, Reed, & Zhang, 2003), consideration of the effect of frozen ground (Koren, 2006; Koren et al., 2010, Koren, Smith, & Cui, 2014), an explicit consideration of vegetation (Koren et al., 2010), and a distributed version of the model (HL-RDHM; Koren, Reed, Smith, Zhang, & Seo, 2004). The heat transfer enhancement to the Sacramento model is operational within CHPS, although HL-RDHM is not.
There are at least two enhancements or modifications to HL-RDHM, although they are not used in operational practice at the NWS. One, by Livneh and colleagues (2011) combined some of the features of the Sacramento model with some features of the Noah model (Ek et al., 2003). A separate work enhanced the groundwater approach in the distributed Sacramento model to allow transfer of flows between adjacent cells (Khakbaz et al., 2011).
CHPS also makes available a different rainfall-runoff model, the Continuous API model, although it is used only in the Middle Atlantic River Forecast Center (MARFC). This model performs well for the hydrologic conditions in their area, but because of its limited use within the NWS, for further information, the reader is referred to the model documentation at: http://www.nws.noaa.gov/oh/hrl/nwsrfs/users_manual/part2/_pdf/23apicont.pdf.
The UNIT-HG is run after the rainfall-runoff models to estimate the instantaneous flow at the catchment outlet. The parameters of the model are described in:
Snow Accumulation and Depletion
Another major model for NWS forecast operations is the conceptual Snow Accumulation and Depletion model, better known as Snow-17. The model was originally developed in 1973 for the new at the time NWSRFS. A technical report (Anderson, 1976) describing the original model is available online.
Snow-17 was ported to CHPS as part of the NWS suite of models. The model has had some minor enhancements since it was originally deployed in NWSRFS, including the addition of snow depth calculations. An excellent document explaining the updated model, as well as the “insights into the reasoning and logic used by the author in developing the model” is available at: http://www.nws.noaa.gov/oh/hrl/nwsrfs/users_manual/part2/_pdf/22snow17.pdf. A short model description, adapted from that reference, is included next.
Similarly to the Sacramento model, Snow-17 is a conceptual model. This means that the physical processes governing the accumulation and melting of snow are represented in a simplified form. Figure 10 shows a high-level flowchart of the Snow-17 calculations.
Due to the scarcity of meteorological observations different from precipitation and temperature back in the 1970s when the model was developed, the model only uses those two inputs. It provides for ways to modify the temperature as a function of elevation based on the lapse rate, since it is rare to have weather observations at all elevations in a watershed.
Snow accumulation in Snow-17 is computed implicitly taking into consideration gauge catch losses due to wind, sublimation, or redistribution on the catchment, by means of a parameter that can be calibrated. Likewise, snow depth is computed based on an empirical function that relates snow density to air temperature.
For the calculation of the energy exchanges at the snow-air interface, the model keeps track of the type of events, differentiating whether there is melt or not. In the case of a snow-melt event the model differentiates between melt with rain or without it. The snowpack states include calculation of the ripeness of the snow, density, and depth computations; refreezing of melted snow, etc. For the transmission of water through the snow cover, Snow-17 uses empirical equations derived by the U.S. Army Corps of Engineers (1955). Another exchange of heat happens at the snow-soil interface.
Another important factor taken into consideration by Snow-17 is the computation of the snow areal cover, since the model equations are based on a 100% snow cover. Therefore, calculations in the case that the snow cover is less than 100% need to be adjusted. These computations are based on the Depletion curve, which relates the snow water equivalent to the areal cover.
Streamflow Routing and Reservoir Operations
The NWS hydrologic forecast operations make use of two types of streamflow routing: Hydraulic and Hydrologic. Hydraulic routing uses the de Saint Venant equation (see, for instance, http://www.efm.leeds.ac.uk/CIVE/CIVE3400/stvenant.pdf), commonly applied in one dimension, although there are currently river dynamics models that apply the equation in two dimensions. Hydrologic routing models are considerably simplified versions of routing that execute in a fraction of the time a hydraulic routing requires, but can’t consider some important effects such as the impact of downstream water elevation on upstream locations.
The NWS originally used two hydraulic routing models, DWOPR and FLDWAV, both developed at the NWS by Fread.
With the transition from NWSRFS into CHPS, a recommendation was issued to abandon any further development and maintenance of FLDWAV and DWOPR in favor of the U.S. Army Corps of Engineers HEC-RAS. See http://www.nws.noaa.gov/oh/hrl/hsmb/hydraulics/documents/fldwav_to_hecras_overview.pdf for details of the recommendation. Since HEC-RAS is well-supported and documented, the interested reader should examine the HEC-RAS page. The several hydrologic routing models currently in used at the NWS are shown in Table 1.
Table 1. Streamflow and Reservoir Routing Models Used by the National Weather Service
This routing technique had its origins in a graphical routing technique. The lag component addresses the time of travel, and the K coefficient applies to the dampening of the flood wave. For more details and model parameters, see:
This is a routing method that attenuates the discharge on a reach by applying different routing coefficients depending on the flow level. For model parameters, see:
This routing model is based on the Streamflow Simulation and Reservoir Regulation System (SSAR) model, developed by the Northwest River Forecast Center and the U.S. Army Corps of Engineers. For model parameters see:
This is the classical Muskingum routing. For model parameters, see:
This routing method could be thought of as a layered unit hydrograph, in which to different levels of flow at the upstream point correspond different transfer functions. For parameters, see:
This is one of the three reservoir operations models available to the NWS forecasters under CHPS. This was developed by the U.S. Army Corps of Engineers. As is the case with HEC-RAS, HEC-ResSim is well documented:
This model is based on the Streamflow Synthesis and Reservoir Regulation System developed by the U.S. Army Corps of Engineers and Northwest River Forecast Center of National Weather Service. The parameters are described here:
This is a model for multiple or single reservoir operations developed at the NWS. The model parameters for CHPS operations are described in:
This model simulates a single reservoir operation. It is used when the forecaster does not receive a release plan from the reservoir operators. The parameters are described in:
Raw Forecast Process
After the observations have been quality controlled, the forecaster can proceed to the actual forecasting process. Figure 11 shows the CHPS screen that details the status of the observations datasets.
The values under the Last Import Time column are color coded to provide the forecaster immediate feedback on the status of the different datasets. Green means that datasets are ready to go. The other colors may mean that the datasets are not quite ready when compared to the time of the forecast. This is relative, however. Some observations are normally received on weekly or even irregular intervals. The forecaster is ultimately responsible to decide whether the information is all ready to be processed.
The forecasting flow in CHPS is controlled by configuration files, written in XML. Some of the XML files describe the models that are used for each watershed. For instance, it may have a snow-melt model, followed by the soil moisture accounting model and the unit hydrograph. If the basin is downstream of another basin that is independently modeled, then the routing scheme is included in the configuration file. Likewise, the operation of any reservoirs within the basin is the result of one of the reservoir operations models, and its order of execution is defined in the configuration file.
Forecast points are processed in forecast groups. Each group has also its own configuration file, which indicates to CHPS in what order the basins should be processed. Figure 12 shows a typical CHPS forecast screen with two windows. The right window controls the forecasting process. The panel on the left side of the right window shows the forecast group (next to the green check marks). The forecast groups can then be expanded to show each of the forecast points (or sub-basins) within the group. Basins that are not processed yet are shown in red. So, if the forecaster makes a change to an upstream basin (for instance, releases from a reservoir changed), all downstream basins are automatically shown in red as well.
The right panel allows the forecaster to control the forecast and to modify the forecast results, as needed (see Forecaster Modifications). The left windows display model results. The bottom-left panel displays model output results. Immediately above it, the smaller panel shows the model forcings: precipitation and temperature. The plots include simulated and observed flows, or stages, or reservoir pool elevation. Also, the forecaster can display routed flows, snow and soil moisture states from the corresponding models, and flows that have been adjusted based on the forecaster modifications.
Figure 13 shows the screen panel that allows the forecaster to enter modifications to the model results. Again, these modifications are based on the forecaster’s own judgment as to whether the model results are not tracking the streamflow observations well, due to a number of reasons. For instance, the river response to a storm centered close to the watershed outlet as opposed to being on the headwaters, will be different. The forecaster then will use one modification to change the shape of the unit hydrograph to better reflect actual conditions. If the model is over forecasting, based on observations, the forecaster may decide on whether to decrease the mean areal precipitation over the basin or its distribution in time, or to “blend” the model results with the observations. Another possibility is to change the model states, or to change the type of precipitation from snow to rain or vice versa.
Ensemble Streamflow Prediction
The Ensemble Streamflow Prediction (ESP) was originally known as the Extended Streamflow Prediction System. It was described by Day (1985). ESP is used at all RFCs to provide a probabilistic hydrologic forecast from months to seasons up to a couple of years. It is based on the concept that the forecasting skill of precipitation and temperature provided by climate prediction models at the resolution required to run hydrologic forecast models very quickly approaches the climatological values. Therefore, ESP precipitation and temperature forcings for the hydrologic models are based on their corresponding historical time series.
The model is initialized with a warm-up run using actual observations of precipitation and temperature, and initial conditions saved from a previous run. Then, using the current model state values saved at the end of the warm-up run, the model is driven with the precipitation and temperature observed at the watershed for each of the historical years, and for the duration of the forecast. For instance, assume that on a particular date (for example, July 24th, 2016) an RFC needs a probabilistic hydrologic forecast for the next 90 days. Assume also that there are historical observations of precipitation and temperature dating from 1930 to 2015. CHPS is then executed starting a few months earlier, for instance from March 1st, 2016 through July 23rd, 2016. Then, the model is executed for each of the years 1930 through 2015, using the July 24th–October 29th actual observations of each year, while keeping the state variables estimated on July 23rd, 2016, as initial conditions for each series. The resulting series of simulations are the conditional simulations (conditional to the current initial conditions). Notice that the streamflow forecasts from July 24th through July 30th are not used, since the historical observations are not reliable forecasts in the short term. Once ranked, those simulations provide a probabilistic view of the flow conditions for the 90 days ahead.
To provide a way to compare those forecasts with historical observations, the NWS also executes a continuous run from 1930 through 2005. In other words, the model is not initialized at every July 24th, but executed as a continuous stream from the first day of the simulation. Then, the values of that historical simulation for the 90 days following July 24th of each year are used to prepare the historical simulation probability of exceedance chart. Figures 14 and 15 show the historical and conditional probability of exceedance of river stages and flows for the Minnesota River at Jordan, MN. Figures 16 and 17 show the probability of exceeding certain levels of stage or flow at the same location, respectively, for each week in the period of forecast.
Over the years, several authors have enhanced the ESP procedure. For instance Werner, Brandon, Clark, and Gangopadhyay (2004) compared several methods to incorporate climate information into ESP. Werner and colleagues (2005) developed a procedure to use actual climate forecasts for days 1–10, in coordination with ESP for days 11–seasonal. This procedure is used at the Colorado Basin River Forecast Center.
The Meteorological Model-Based Ensemble Forecast System (MMEFS) was developed by the NWS Eastern Region RFCs (Adams, 2015) to produce short-term hydrologic forecasts. It uses as input Global Ensemble Forecast System (GEFS), North American Ensemble Forecast System (NAEFS), and Short Range Ensemble Forecast System (SREF), all produced by the NWS’s NCEP. One shortcoming of MMEFS is the presence of biases in the meteorological forcings and the lack of explicit consideration of hydrologic model errors. Nevertheless, the Eastern Region RFCs and the Southeast RFC have found the MMEFS products worthwhile. Figure 18 shows the different traces of precipitation forecasts obtained from the various models, and, with the temperature traces (Figure 19) used as input to the hydrologic models. Model results include the traces of Snow Water Equivalent (Figure 20), runoff, river stage traces (Figure 21), river stage spread (Figure 22) among others, and exceedance probability chart (Figure 23).
Hydrologic Ensemble Forecasting System
Around 2005 the NWS Office of Hydrologic Development responded to requests from the field offices to start the development of a hydrologic ensemble forecast system that would allow the generation of short term, mid- and long-term ensemble members in a seamless fashion. The project was named the Experimental Ensemble Forecasting System (XEFS; Demargne et al., 2010; Demargne et al., 2014; Demargne, Wu, Regonda, & Brown, 2010; Schaake, 2008; Schaake et al., 2007; Seo, Bonnin, Roe, & Restrepo, 2009; Seo, Demargne, Wu, Brown, & Schaake, 2007; Seo et al., 2008; Seo, 2011). Once it became operational, the name was changed to Hydrologic Ensemble Forecasting System or HEFS. Figure 24 shows a schematic diagram of HEFS. The major components are the Meteorological Ensemble Forecasting Preprocessor (MEFP), and its parameter estimation process (MEFPPE), the Ensemble Post Processor (EnsPost) and its parameter estimation process (EnsPostPE), the Ensemble Verification System (EVS), and the graphics generator (GraphGen). The reader is referred to the above references for detailed description of the HEFS components.
MEFP works on the basis of the correlation between past forecasts and the corresponding observations. For precipitation, it uses a probability mapping technique to map observed and forecast precipitations into normally distributed variables. The forecast period is initially divided into a series of canonical time steps, which increase in duration as the forecast period increases. For instance, the initial eight time steps may be six hours, followed by time steps of one day, then three days, then one week, then two weeks, then month long, etc. That way, the system is able to capture any skill the meteorological forecasting models may have, since increasing the level of aggregation typically increases the model skill.
The user is free to select the sources of forecasts, and they can be combined. For instance, the short-term forecasts (i.e., the first few canonical steps) may be based on the NWS’s Global Forecasting System (GFS), and longer periods may use the NWS’s Climate Forecasting System Version 2 (CFSV2), or any other model for which a historical reforecast has been completed.
Estimates of the MEFP parameters are obtained by looking at the correlation between past observed and forecast values during the reforecast period, looking at a user-defined window around each canonical time step. Then, give a deterministic forecast from the same model that was used in the calibration, the system samples from the conditional probability of the observations, given the forecasts. This guarantees that any biases in the weather forecast models are automatically removed. Because it uses the conditional distribution, if there is a good correlation between forecasts and observations the resulting conditional distribution will have a lower standard deviation than the marginal distribution. If the correlation is low, the model ends up producing the marginal distribution, or, in other words, the historical values.
Once each of the individual values at each canonical period is obtained, the system assembles each time series using a technique known as the Schaake Shuffle (Clark, 2004). The ensemble series is then input to the hydrology model, where it is run. A second process (Ens-Post) is executed. Ens-Post was designed to remove biases in the produced hydrologic series. Finally, an ensemble verification system (EVS) can then be used to assess the performance of the HEFS (Brown, Demargne, Seo, & Liu, 2010).
Forecasts are communicated via a number of means: Internet, NOAA radio, specific arrangements for major government users, etc.
The primary distributor of NWS hydrologic forecasts on the Web are the WFOs. The home page for hydrologic forecasts is http://water.weather.gov/ahps/, (Figure 25). The forecast locations are color coded to indicate either the current or the forecaster state of the river at each point.
The legend on the right side of the page indicates the meaning of each color, and the number of gauges under each condition. The drop menus above the legend allow the user to select the view by state, WFOs, RFCs, or Water Resources Regions. Figure 26 shows a portion of the web page with the status of the forecast pointing at the North Central River Forecast Center on August, 25, 2016. Notice that users can choose between river observations (selected) or river forecasts. Notice also that not all points are forecast every time. Some points are forecast only during flooding conditions, based on requirement from emergency managers or other interested parties. Hovering the mouse pointer over one point with forecasts will show a pop-up window with the forecast (Figure 27). If the user clicks on a point, a specific web page to that point containing considerably more information is displayed. Figure 28 shows just the hydrograph chart on that page.
In addition to the web-based information dissemination, the NWS also operates the NOAA weather radio. This is a network of Very High Frequency (VHF) radios that cover the 50 states, adjacent coastal waters, Puerto Rico, the U.S. Virgin Islands, and the U.S. Pacific Territories, although not 100% of the territory is completely covered. For a coverage map, see http://www.nws.noaa.gov/nwr/resources/NWR_Propagation.pdf.
This network is an all-hazards system that broadcasts not only weather and flooding watches and warnings, but all other hazards including earthquakes, fires, avalanches, environmental hazards (such as chemical spills), or public safety (AMBER alerts, for instance). It requires specialized receivers that are commercially available.
Verification and Performance Evaluation
As of the date of this writing, there is no agency-wide requirement for verification and validation of flow forecasts at each RFC. However, forecast performance is assessed internally by the RFCs and used to decide on what watersheds are in need of recalibration due to significant changes to the basin characteristics, such as urban development, deforestation, reforestation, etc.
The responsibility for official flood forecasting in the United States falls on the National Weather Service (NWS), a line office of the National Oceanic and Atmospheric Administration (NOAA), an agency of the U.S. Department of Commerce. Forecasts at places with over 6 hours of concentration time are carried out at 13 river forecast centers (RFCs) of the National Weather Service. These forecasts are coordinated with the 122 Weather Forecast Offices, which are responsible for flash flood forecasts (those with concentration times less than 6 hours) and for issuing all forecasts to the public. The NWS recently created the Office of Water Prediction that, among other things, seeks to establish a centralized, high-resolution forecasting system for the United States. As of this writing, the process and the plans for integration with the 13 river forecast centers are still under development.
The hydrologic forecasting process at the NWS involves a series of coordinated steps leading to the publication of the final product. The first step is the acquisition of observations and forecasts. These always include precipitation and temperature observations and forecasts, and river flow and stage observations, including reservoir discharge observations and proposed operations, where available. Additionally, some RFCs also use additional information in a qualitative way, such as snow water equivalent, snow areal coverage, soil moisture, and soil temperature.
The next step in the process is the data quality control. Although some of the quality control is performed automatically, a considerable amount of effort is dedicated to a manual quality-control process of precipitation and temperature observations. Precipitation observations from rain gauges are used jointly with radar observations to arrive at mean areal quantitative precipitation estimates over individual watersheds.
The initial source of quantitative precipitation and temperature forecasts (QPF and QTF) is the numerical guidance that is prepared by the National Centers for Environmental Prediction (NCEP). That numerical guidance is refined further at the RFCs in coordination with the Weather Forecast Offices (WFOs). With the precipitation and temperature observations and forecasts on hand, it is possible to execute the hydrology models and issue forecasts but only at non-regulated watersheds.
The hydrologic forecasting system the NWS uses is known as the Community Hydrologic Prediction System, or CHPS. It is based on the Flood Early Warning System (FEWS) developed by the Dutch non-profit Deltares. The NWS forecasters make deterministic runs of CHPS by river groups, starting with the upstream-most watersheds and moving downstream. The forecasters then assess the model performance by comparing past forecasts with streamflow observations, and, optionally, modify model states, precipitation or temperature input, or even model parameters. This interaction between forecaster and model results is a key component of the forecasting process. Finally, the deterministic runs serve as “hot states” for the probabilistic forecasts.
There are three types of probabilistic hydrologic forecasts at the NWS, depending on the source of the precipitation and temperature forcings. The first and oldest technique is the Ensemble Prediction System. A more recent ensemble forecast approach is known as the Multi Model Ensemble Forecasting System (MMEFS). The most recent approach is the Hydrologic Ensemble Forecasting System (HEFS).
The final steps are the communication of forecasts and the performance evaluation. Forecasts are communicated via a number of means: Internet, NOAA radio, specific arrangements for major government users, etc. Forecast performance is assessed internally at each RFC and used to decide on what watersheds are in need of recalibration due to significant changes to the basin characteristics, such as urban development, deforestation, reforestation, etc.
The author wishes to thank the very insightful and constructive comments from an anonymous reviewer, which considerably improved this article contents; and to Ms. Holly Reckel of the North Central River Forecasting Center for providing the example of quality control of precipitation observations.
Adams, T. (2015). Verification of the NOAA/NWS MMEFS Operational Hydrologic Ensemble Forecasting System in the Ohio River Valley. In World Environmental and Water Resources Congress, 2015, 2688–2700.Find this resource:
Anderson, E. A. (1976). A point of energy and mass balance model of snow cover. NOAA Technical Reports NWS, 19, 1–150.Find this resource:
Anderson, R. M., Koren, V. I., & Reed, S. M. (2006). Using SSURGO data to improve Sacramento Model a priori parameter estimates. Journal of Hydrology, 320(1), 103–116.Find this resource:
Bonnin, G. M., Kitzmiller, D., Seo, D., Smith, M., & Restrepo, P. (2007). 2.1 QPE in NWS hydrologic applications: Present and future.
Brown, J. D., Demargne, J., Seo, D. J., & Liu, Y. (2010). The Ensemble Verification System (EVS): A software tool for verifying ensemble forecasts of hydrometeorological and hydrologic variables at discrete locations. Environmental Modelling & Software, 25(7), 854–872.Find this resource:
Burnash, R. J. C., Ferral, R. L., & McGuire, R. A. (1973). A generalized streamflow simulation system—Conceptual modeling for digital computers. Sacramento, CA: Joint Federal and State River Forecast Center, US National Weather Service/California State Department of Water Resources.Find this resource:
Burnash, R. J. C. (1995). The NWS river forecasting system. In V. P. Singh (Ed.), Computer models of watershed hydrology (pp. 395–442). Highlands Ranch, CO: Water Resources Publications, LLC.Find this resource:
Clark, M., Gangopadhyay, S., Hay, L., Rajagopalan, B., & Wilby, R. (2004). The Schaake shuffle: A method for reconstructing space–time variability in forecast precipitation and temperature fields. Journal of Hydrometeorology, 5(1), 243–262.Find this resource:
Crum, T. D., & Alberty, R. L. (1993). The WSR-88D and the WSR-88D operational support facility. Bulletin of the American Meteorological Society, 74(9), 1669–1687.Find this resource:
Day, G. N. (1985). Extended streamflow forecasting using NWSRFS. Journal of Water Resources Planning and Management, 111(2), 157–170.Find this resource:
Demargne, J., Liu, Y., Brown, J., Seo, D. J., Wu, L., Weerts, A., & Werner, M. (2010, May). Hydrologic ensemble hindcasting and verification in the U.S. National Weather Service. In EGU General Assembly Conference Abstracts (Vol. 12, p. 10604).Find this resource:
Demargne, J., Wu, L., Regonda, S., & Brown, J. D. (2010, December). Verification of experimental short-term streamflow ensemble forecasts produced by the U.S. National Weather Service. AGU Fall Meeting Abstracts.
Demargne, J., Wu, L., Regonda, S. K., Brown, J. D., Lee, H., He, M., . . . Schaake, J. (2014). The science of NOAA’s operational hydrologic ensemble forecast service. Bulletin of the American Meteorological Society, 95(1), 79–98.Find this resource:
Ek, M. B., Mitchell, K. E., Lin, Y., Rogers, E., Grunmann, P., Koren, V., . . . Tarpley, J. D. (2003). Implementation of Noah land surface model advances in the National Centers for Environmental Prediction operational mesoscale Eta model: GEWEX Continental-Scale International Project, Part 3 (GCIP3). Journal of Geophysical Research, 108(D22), GCP12-1.Find this resource:
Fread, D. L., Shedd, R. C., Smith, G. F., Farnsworth, R., Hoffeditz, C. N., Wenzel, L. A., . . . Day, G. N. (1995). Modernization in the National Weather Service river and flood program. Weather and Forecasting, 10(3), 477–484.Find this resource:
Fulton, R. A., Breidenbach, J. P., Seo, D. J., Miller, D. A., & O’Bannon, T. (1998). The WSR-88D rainfall algorithm. Weather and Forecasting, 13(2), 377–395.Find this resource:
Jelesnianski, C. P., Chen, J., & Shaffer, W. A. (1992). SLOSH: Sea, lake, and overland surges from hurricanes (Vol. 48). U.S. Department of Commerce, National Oceanic and Atmospheric Administration, National Weather Service, Silver Spring, MD.Find this resource:
Johnson, J. T., MacKeen, P. L., Witt, A., Mitchell, E. D. W., Stumpf, G. J., Eilts, M. D., & Thomas, K. W. (1998). The storm cell identification and tracking algorithm: An enhanced WSR-88D algorithm. Weather and Forecasting, 13(2), 263–276.Find this resource:
Khakbaz, B., Imam, B., Sorooshian, S., Koren, V. I., Cui, Z., Smith, M. B., & Restrepo, P. (2011). Modification of the National Weather Service Distributed Hydrologic Model for subsurface water exchanges between grids. Water Resources Research, 47(6).Find this resource:
Kitzmiller, D., Miller, D., Fulton, R., & Ding, F. (2013). Radar and multisensor precipitation estimation techniques in National Weather Service hydrologic operations. Journal of Hydrologic Engineering, 18(2), 133–142.Find this resource:
Koren, V. (2006). Parameterization of frozen ground effects: Sensitivity to soil properties. IAHS Publication, 303, 125.Find this resource:
Koren, V., Reed, S., Smith, M., Zhang, Z., & Seo, D. J. (2004). Hydrology Laboratory Research Modeling System (HL-RMS) of the U.S. National Weather Service. Journal of Hydrology, 291(3), 297–318.Find this resource:
Koren, V., Smith, M., & Cui, Z. (2014). Physically-based modifications to the Sacramento Soil Moisture Accounting model. Part A: Modeling the effects of frozen ground on the runoff generation process. Journal of Hydrology, 519, 3475–3491.Find this resource:
Koren, V. I., Smith, M., Wang, D., & Zhang, Z. (2000). 2.16 Use of soil property data in the derivation of Conceptual Rainfall-Runoff model parameters.
Koren, V. I., Smith, M. B., Cui, Z., Cosgrove, B. A., Werner, K. V., & Zamora, R. (2010). Modification of Sacramento Soil Moisture Accounting Heat Transfer Component (SAC-HT) for Enhanced Evapotranspiration. U.S. Department of Commerce, National Oceanic and Atmospheric Administration, National Weather Service, Silver Spring, MD. Available at http://amazon.nws.noaa.gov/articles/HRL_Pubs_PDF_May12_2009/New_Scans_June_2011/NOAA_Technical_Report_NWS_53.pdf.Find this resource:
Krajewski, W. F., Kruger, A., Smith, J. A., Lawrence, R., Gunyon, C., Goska, R., . . . Weber, J. (2011). Towards better utilization of NEXRAD data in hydrology: An overview of Hydro-NEXRAD. Journal of Hydroinformatics, 13(2), 255–266.Find this resource:
Krajewski, W. F., Villarini, G., & Smith, J. A. (2010). Radar-rainfall uncertainties: Where are we after thirty years of effort? Bulletin of the American Meteorological Society, 91(1), 87–94.Find this resource:
Larson, L. (2002). National Weather Service River Forecasting System (NWSRFS). In V. P. Singh & D. K. Frevert (Eds.), Mathematical models of small watershed hydrology and applications (pp. 657–706). Highlands Ranch, CO: Water Resources Publications.Find this resource:
Larson, L. W., Ferral, R. L., Strem, E. T., Morin, A. J., Armstrong, B., Carroll, T. R., . . . Johnson, D. E. (1995). Operational responsibilities of the National Weather Service river and flood program. Weather and Forecasting, 10(3), 465–476.Find this resource:
Livneh, B., Restrepo, P. J., & Lettenmaier, D. P. (2011). Development of a unified land model for prediction of surface hydrology and land–atmosphere interactions. Journal of Hydrometeorology, 12(6), 1299–1320.Find this resource:
McEnery, J., Ingram, J., Duan, Q., Adams, T., & Anderson, L. (2005). NOAA’s advanced hydrologic prediction service: Building pathways for better science in water forecasting. Bulletin of the American Meteorological Society, 86(3), 375–385.Find this resource:
National Research Council. (2006). Toward a new advanced hydrologic prediction service (AHPS). Washington, DC: National Academies Press.Find this resource:
Park, H. S., Ryzhkov, A. V., Zrnic, D. S., & Kim, K.-E. (2009). The hydrometeor classification algorithm for the polarimetric WSR-88D: Description and application to an MCS. Weather and Forecasting, 24(3), 730–748.Find this resource:
Polger, P. D., Goldsmith, B. S., Przywarty, R. C., & Bocchieri, J. R. (1994). National Weather Service warning performance based on the WSR-88D. Bulletin of the American Meteorological Society, 75(2), 203–214.Find this resource:
Reed, S., Koren, V., Smith, M., Zhang, Z., Moreda, F., Seo, D. J., & Participants, D. M. I. P. (2004). Overall distributed model intercomparison project results. Journal of Hydrology, 298(1), 27–60.Find this resource:
Ryzhkov, A. V., Giangrande, S. E., & Schuur, T. J. (2005). Rainfall estimation with a polarimetric prototype of WSR-88D. Journal of Applied Meteorology, 44(4), 502–515.Find this resource:
Schaake, J. (2008). Analysis of uncertainty in the distribution of wet precipitation events. Geophysical Research Abstracts. Vol. 10. EGU General Assembly 2008.Find this resource:
Schaake, J., Demargne, J., Hartman, R., Mullusky, M., Welles, E., Wu, L., . . . Seo, D. J. (2007). Precipitation and temperature ensemble forecasts from single-value forecasts. Hydrology and Earth System Sciences Discussions, 4(2), 655–717.Find this resource:
Schaake, J., Henkel, A., & Cong, S. (2004). 5.3 Application of prism climatologies for hydrologic modeling and forecasting in the western U.S. American Meteorological Society General Assembly 2008.
Seo, D. J. (2011). Ensemble prediction–The new paradigm in operational hydrologic forecasting.
Seo, D., Liu, Y., & Herr, H. (2008). Experimental Hydrologic Ensemble Forecast System for Collaborative R&D and Research-to-Operations Transition in NWS. AGU Fall Meeting Abstracts. Vol. 1.Find this resource:
Seo, B. C., & Krajewski, W. F. (2010). Scale dependence of radar rainfall uncertainty: Initial evaluation of NEXRAD’s new super-resolution data for hydrologic applications. Journal of Hydrometeorology, 11(5), 1191–1198.Find this resource:
Seo, D., Bonnin, G. M., Roe, J., & Restrepo, P. J. (2009, December). Hydrologic ensemble forecasting in the National Weather Service: Status, plans and challenges. In AGU Fall Meeting Abstracts (Vol. 1, p. 8).Find this resource:
Seo, D., Demargne, J., Wu, L., Brown, J. D., & Schaake, J. C. (2007, December). Toward integrative uncertainty accounting in operational hydrologic ensemble forecasting. AGU Fall Meeting Abstracts.Find this resource:
Shedd, R. C., & Fulton, R. A. (1993, July). WSR-88D precipitation processing and its use in National Weather Service hydrologic forecasting. In Proceedings of the international symposium on engineering hydrology San Francisco: ASCE.Find this resource:
Shelby, A. (2003). Red River rising: The anatomy of a flood and the survival of an American city. St. Paul: Minnesota Historical Society Press.Find this resource:
Smith, M. B., Koren, V., Zhang, Z., Zhang, Y., Reed, S. M., Cui, Z., . . . Anderson, E. A. (2012). Results of the DMIP 2 Oklahoma experiments. Journal of Hydrology, 418, 17–48.Find this resource:
Smith, M. B., Laurine, D. P., Koren, V. I., Reed, S. M., & Zhang, Z. (2003). Hydrologic model calibration in the National Weather Service. In Q. Duan, H. V. Gupta, S. Sorooshian, A. N. Rousseau, & R. Turcotte (Eds.), Calibration of Watershed Models (pp. 133–152). Washington, DC: American Geophysical Union.Find this resource:
Smith, M. B., Seo, D. J., Koren, V. I., Reed, S. M., Zhang, Z., Duan, Q., . . . Cong, S. (2004). The distributed model intercomparison project (DMIP): Motivation and experiment design. Journal of Hydrology, 298(1), 4–26.Find this resource:
Stallings, E. A., & Wenzel, L. A. (1995). Organization of the river and flood program in the National Weather Service. Weather and Forecasting, 10(3), 457–464.Find this resource:
U.S. Army Corps of Engineers (1955). Snow investigations–Lysimeter studies of snow melt. Snow Investigations Research Note 25, Corps of Engineers, North Pacific Division, Portland, Oregon.Find this resource:
Werner, K., Brandon, D., Clark, M., & Gangopadhyay, S. (2004). Climate index weighting schemes for NWS ESP-based seasonal volume forecasts. Journal of Hydrometeorology, 5(6), 1076–1090.Find this resource:
Werner, K., Brandon, D., Clark, M., & Gangopadhyay, S. (2005). Incorporating medium-range numerical weather model output into the ensemble streamflow prediction system of the National Weather Service. Journal of Hydrometeorology, 6(2), 101–114.Find this resource:
Wu, L., Seo, D. J., Demargne, J., Brown, J. D., Cong, S., & Schaake, J. (2011). Generation of ensemble precipitation forecast from single-valued quantitative precipitation forecast for hydrologic ensemble prediction. Journal of Hydrology, 399(3), 281–298.Find this resource:
Zhang, J., Howard, K., Langston, C., Vasilof, S. D., Karney, B., Arthur, A., . . . Dempsey, C. (2011). National mosaic and multi-sensor QPE (NMQ) system: Description, results, and future plans. Bulletin of the American Meteorological Society, 92(10), 1321–1338.Find this resource: