Summer 2003 Banner. Waterfall at Mirror Lake near Arkansas, courtesy of Chuck Haralson, ADPT

National Weather Service
Arkansas-Red Basin
River Forecast Center
10159 East 11th Street
Suite 300
Tulsa, OK 74128

In This Issue

The GAGE Staff

 Diane Innes Cooper, Publisher
 James Paul, Editor

 Download PDF version (369 KB)


Remote Backup Capabilities Are Now Available For RFC’s

By Jeff McMurphy

With the increasing importance and reliability of the National Weather Service (NWS) on computers to obtain data, generate forecasts and transmit products to the public, the NWS has seen the importance for full backup capabilities for an RFC. The most obvious solution would be to install a system similar to what is in place at the NWS Weather Forecast Offices (WFOs). However this option was not feasible because unlike the WFO’s, each RFC has its own unique set of data and forecasting tools. These data and tools, while standardized for basic functions, are configured to the specific needs of each RFC. Each RFC runs their own hydrologic modeling systems for their own specific areas of responsibility. The RFC also has specific data and information that is needed to produce the forecasts they issue for their rivers. In addition, to ensure that the model performs at its best, each RFC continuously maintains the dynamics of their models, especially stage to discharge ratings of river gauges and routings between gauged locations. Thus, the capability for one RFC to backup another RFC in the manner the NWS WFOs do, was not possible because a “backup” RFC would not have the personnel, computing, or storage resources available to maintain two RFC systems.

Previously, the plan for an RFC backup system was for the NWS Office of Hydrology (OH) to have the capability to load each individual RFC’s AWIPS system with the specific programs, tools, and data onto the OH AWIPS system. When backup operations were needed, an RFC would contact OH to initiate the loading of that RFC’s system and begin the data acquisition processes. RFC forecasters would then travel to OH in the Washington D.C. area, where they would use OH’s system to generate and transmit forecasts and products. In theory this seemed like a good idea, yet the logistics of maintaining accurate systems for each of the 13 RFC’s was too complicated. Another problem was the associated expenses of lodging and travel costs for the RFC personnel. Hence, this functionality could only be used in situations when the cost could be justified, typically not for the one to two day events which were most common. Thus, in these situations, the NWS WFO’s were on their own to issue river forecasts without the guidance of a hydrologic model.

Recently, desktop PC processing power, storage capabilities, and software availability began to reach a level in which developing a backup system for local use at an RFC became a true possibility. In 2001, ABRFC began developing a backup Linux desktop PC. Linux was chosen because OH and the Forecast System Laboratory (FSL) had already begun the process of porting NWS applications to Linux. In May of 2002, this local desktop system was tested and proven successful for being able to provide ABRFC forecast and guidance products to customers in a timely and transparent manner. (Please refer to “ABRFC Operational Backup Test Is Successful” article in the Fall 2002 publication of “The Gage,“ for more details.)

Since this first test, ABRFC has taken the backup capability to yet another plateau, this time in the laptop PC arena. The ABRFC has created a Linux laptop with the same functionality, yet with much faster processing speeds, nearly 7 times faster, than the current Operational AWIPS system. This system downloads data from the Internet, runs the hydrologic model as well as supporting programs such as NMAP, XNAV, XDAT, FCSTPROG XSETS, P2 and MPE and issues a full suite of operationally critical products and forecasts including RVFs, FFGs, HMDs, and HCMs. The AWIPS D2D Display is not utilized, yet radar reflectivity and satellite images can be displayed via a variety of web sites. The national Flood Outlook Product cannot be produced as ArcView is unavailable for Linux. Product dissemination is accomplished by using ABRFC developed scripts which drop products on a SRH server where they are picked up by a backup office’s LDAD. Products are then moved through the protective firewall from LDAD and into AWIPS via common LDAD/AWIPS ingest software.

The advantage of the laptop portability was utilized during the test conducted on March 11, 2003. The test mimicked a total office-loss scenario and operations were conducted from a hotel room with a high speed internet connection as well as a second voice line for coordination purposes. The system was cold-started and made ready for use in one hour and ten minutes. Cold-start is defined as the system having no model files, no observed data, no files of estimated radar precipitation, no Quantitative Precipitation Forecasts (QPF) files, and not connected to the Internet. Ready-for-Use is defined as: the hydrologic model run, for the entire river forecast system, is complete along with a current set of data and the system is functional and ready for interactive use by forecasters with data ingest and dissemination processes up and running. This cold start is possible because on a regular basis, ABRFC’s operational hydrologic model information as well as other data is uploaded from a desktop version of the Linux backup system which we locally maintain. (Note: A data conversion problem exists for sending files from HP UNIX to Linux. Hence, once we transition to performing routine operations on a Linux platform, we will then be able to use this data rather than maintaining the desktop backup system.) Not only did this laptop PC perform all operational functionality, but it was also able to perform as a “server” in the sense of allowing multiple laptops to remotely login and allowing these “clients” to be used as additional stations from which to forecast.

With the success of this latest test, SRH has provided the funding for the other Southern Region RFC’s, West Gulf River Forecast Center (WGRFC), Southeast River Forecast Center (SERFC), and Lower Mississippi River Forecast Center (LMRFC), to have a laptop Backup PC. Hence, by the time you receive this newsletter, ABRFC will have configured these RFC laptop PC’s for the specialized operations of each RFC and will have trained a member of each of the RFC’s in its use. Final testing and implementation of these backup systems is expected to be accomplished by July 31, 2003. So, in the near future, all Southern Region RFC’s will be able to provide services and products to the WFO’s and public in the events of an AWIPS upgrade, an AWIPS failure, or even in the case of a catastrophic event happening at an RFC’s building facilities.

Back to top

ABRFC Evaluates New Short-Term Probabilistic Forecasts

By Bill Lawrence

Hydrologic forecasts are based on many inputs. Two of the most important inputs include estimates of rain that has already fallen over a basin, and rainfall that is forecast to fall in the future. Using today’s technology of radar-based rainfall estimates along with a relatively dense network of rain gages, we are able to come up with a good estimate of rainfall that has already occurred. Forecasting future rainfall is a totally different situation. There are times when we have high confidence in our Quantitative Precipitation Forecasts (QPF), while other times, especially during summer convective outbreaks, our confidence is very low. Since QPF is one of the most important inputs to a hydrologic model, and its reliability is quite variable, the Office of Hydrologic Development has embarked on a project to help define the variability of hydrologic forecasts resulting from the uncertainty of QPF forecasts. ABRFC has started to participate in this project. The basic premise of the project is to determine exactly what type of confidence to place on QPF using historical QPFs and comparing them to what actually occurred. After this relationship is established, real-time QPF for the first 24 hours of a forecast is adjusted depending on its reliability. Next, historical rainfall from years past is used as QPF for the two to five day window of the hydrologic forecasts. Finally, the hydrologic model is run using an ensemble of QPFs and current conditions to produce a series of river forecasts. Statistical procedures then convert these ensembles of river forecasts into probabilities and a resulting hydrograph can show 2%, 10%, 25%, and etc., chance of exceeding a particular river level for the forecast period. The attached hydrograph plot shows an example of the output from this new experimental procedure. It is hoped that in the future, forecasts like these will give our users more idea of the uncertainty in our forecasts, along with possible outcomes if QPF is much higher or lower than expected.
Figure 1: Example of the Experimental Probabilities Plot of Flow through five days for Carthage, MO. This plot was generated with three inches of QPF forecast to occur in the Second Period, between forecast hour 6 and 12. The plot in purple shows the most likely flow/river stage situation while the plot in aqua indicates the least likely flow/river stage solution.
Figure 1: Example of the Experimental Probabilities Plot of Flow through five days for Carthage, MO. This plot was generated with three inches of QPF forecast to occur in the Second Period, between forecast hour 6 and 12. The plot in purple shows the most likely flow/river stage situation while the plot in aqua indicates the least likely flow/river stage solution.
Back to top


2003 Water Supply Season

By John Schmidt

The Arkansas-Red Basin River Forecast Center (ABRFC), in coordination with the NRCS, provides guidance for the Arkansas River basin in Colorado and the Canadian River headwaters in New Mexico. These data and forecast products are available via the internet at http://www.srh.noaa.gov/abrfc/WaterSupply/index.php. An article on the generation of our forecasts is available in "The Gage," Winter 2001 Edition.

This year, a dry winter gave way to a normal spring snowpack in the Arkansas River basin by the middle of March (See Figure 1). This catch-up was thanks to a record-setting blizzard on March 17-19, 2003. Although the bulk of the snow fell north of the Palmer Divide, the one to four feet that fell in the mountains west of Colorado Springs and Pueblo was welcome moisture.

With most of the snow falling east of Salida, CO, snowmelt forecasts on the Arkansas River at Granite and Salida, CO remained about the same. However, snowmelt forecasts for the tributaries of the Arkansas River east of Salida (Grape and Chalk Creeks and Cucharas, Purgatoire and Huerfano Rivers) jumped from 71 to 94 percent-of-average snowmelt volume. Basin-wide, a slightly below-normal snowmelt season was forecast with a peak flow during the first two weeks of June. This forecast remained about the same from April to May.

The timing of the snowmelt is based on an average distribution of runoff over the past 20-50 years and therefore reflects an average daily temperature. The amount of streamflow forecast through the summer reflects a normal amount rainfall, as most of the snowpack is gone. May proved to be anything but average, especially the last two weeks, as temperatures in the Arkansas River Basin ranked "much above normal" (See Figure 2). Additionally, areas above water supply forecast points received below-normal precipitation amounts over the past 90 days (See Figure 3).

This significantly atypical temperature distribution resulted in higher and earlier than expected crest of streamflows at most water supply forecast points in the Arkansas River basin (See Figure 4). In fact, several flood forecasts and river flood warnings were issued for the Arkansas River near Canon City and Pueblo by the ABRFC and the local NWS forecast office in Pueblo, CO during the week of May 27th.
Figure 1: Accretion and depletion of 2003 Arkansas Basin Snowpack (Courtesy NRCS).  The black line denotes the actual accumulation of snowfall for the 2003 winter season.  The red line indicates the 30 year average snowfall accumulation during a winter season. Hence, one can clearly see that this past winter was much closer to normal in comparison to the 2002 Season (brown line).
Figure 1: Accretion and depletion of 2003 Arkansas Basin Snowpack (Courtesy NRCS). The black line denotes the actual accumulation of snowfall for the 2003 winter season. The red line indicates the 30 year average snowfall accumulation during a winter season. Hence, one can clearly see that this past winter was much closer to normal in comparison to the 2002 Season (brown line).
Figure 2: May 2003 Temperature Departure by Climate Zone. Arkansas Basin in Colorado is outlined in bold black, depicting the much above normal temperatures received in May of 2003. (courtesy NCDC)
Figure 2: May 2003 Temperature Departure by Climate Zone. Arkansas Basin in Colorado is outlined in bold black, depicting the much above normal temperatures received in May of 2003. (courtesy NCDC)
Figure 3: Last 90 days percent-of-normal precipitation, Arkansas River Basin, in Colorado. Normal precipitation in the headwaters of the Arkansas River is typically three to four inches of precipitation. Much of this area has seen only one to two inches of precipitation with isolated areas receiving a half of an inch.
Figure 3: Last 90 days percent-of-normal precipitation, Arkansas River Basin, in Colorado. Normal precipitation in the headwaters of the Arkansas River is typically three to four inches of precipitation. Much of this area has seen only one to two inches of precipitation with isolated areas receiving a half of an inch.
Figure 4: Forecast percent chance of exceedance and observed streamflows. This forecast was issued April 1st with weekly water supply estimates. The various colors  depict the probability of occurrence based on analysis of historical data.  The triangle data points indicate the actual observed volume while the circles data points indicate the volume generated based on native flow calculations. Native Flow is defined as the flow which would have been generated if no human intervention had occurred.
Figure 4: Forecast percent chance of exceedance and observed streamflows. This forecast was issued April 1st with weekly water supply estimates. The various colors depict the probability of occurrence based on analysis of historical data. The triangle data points indicate the actual observed volume while the circles data points indicate the volume generated based on native flow calculations. Native Flow is defined as the flow which would have been generated if no human intervention had occurred.
Back to top


Arconyms in this Edition

ABRFC - Arkansas-Red Basin River Forecast Center
AWIPS- Advanced Weather Interactive Processing System
FFG - Flash Flood Guidance
FCSTPROG - Software Program used to current and previous river forecast to the stage observations
FSL - Forecast Systems Lab
HCM - Hydrologic Coordination Message
HMD - Hydrometeorological Discussion
LDAD - Local Data Acquisition and Dissemination System
LMRFC - Lower Mississippi River Forecast Center
MPE - Multi–Sensor Precipitation Estimator (Used to process observed gridded precipitation)
NMAP - Office of Hydrology
NRCS - Natural Resources Conservation Service
OH - Office of Hydrology
QPF - Quantitative Precipitation Forecast
RFC - River Forecast Center
RVF - River Forecast Guidance from an RFC
SERFC - Southeast River Forecast Center
SRH - Southern Region Headquarters
USDA - United States Department of Agriculture
WFO - National Weather Service Forecast Office
WGRFC - West Gulf River Forecast Center
XDAT - Software program used to query the hydrologic database
XNAV - Multi-Use Software Program used in RFC operations
XSETS - Software program used to generate and transmit river forecasts
Back to top