Click for next page ( 37


The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 36
Comparison of Weather Services: Pre-NEXRAD and NEXRAD In the preceding sections, and in Appendix A, the pane} examined the technical aspects of the pre-NEXRAD and NEXRAD radars as well as their calculated and "experienced" detection capabilities. This section discusses the panel's comparison of the actual weather services provided by the old and new systems as a final approach to determining whether the implementation of the new system could result in a degradation of service to the nation. It is important to consider the effect of other, nonradar elements of the nation's overall weather infrastructure on the NWS's ability to provide service. It is also important to consider the impact that nearby radars in the network will have on enhancing the service provided to each service area. Another key aspect of the comparison is the performance of both the old and new systems in detecting severe weather events and generating accurate, timely warnings. New technologies in Doppler signal processing, digital information systems, and color displays in the NEXRAD system contribute to the development of new products and services. These technologies are described and discussed in Appendix A. COMPOSITE SYSTEM CONSIDERATIONS The composite system is a combination of scientific, technical, and staff resources that are organized by the NWS to accomplish its mission of providing weather forecast and warning services, primarily for the protection of life and property. These resources include all of the weather data that are collected, processed, and analyzed; specialized equipment; dedicated communication networks and systems; centralized operation of supercomputers and numerical models; and highly trained staff. Detailed information about the composite system can be found in an earlier report by the NRC's NWSMC (NRC, 1994b). Weather semices provided to all users, including the public, are made available through a complex array of observations and models and through the interpretive skills of experienced forecast- ers. In the modernized NWS, there will be a host of new observational systems available to weather forecasters. These will include the NEXRAD network; the ASOS, which continuously provides surface observations of temperature, wind, humidity, pressure, and over atmospheric variables; the next generation of geostationary and polar-orbiting satellites; and a national lightning network. 36

OCR for page 36
Companson of Pre-NEXRAD and NEXRAD - 37 In addition to these systems, there are a number of new observational systems that are currently being demonstrated or developed for future use. These systems include commercial aircraft measurements of wind and temperature, and possibly humidity, both at flight levels and during ascent and descent in the vicinity of larger airports; wind profilers that measure horizontal winds as a function of height above the ground; and the next generation radiosonde system. Operational components of the composite system are shown in Figure 3-~. As shown in the figure, each WFO will have responsibility for what might be viewed as a cylindrical volume of coverage in the United States. These volumes are generally overlapping or contiguous, allowing the WFOs to benefit from one another due to the availability of network data from adjacent regions and nationally. The AWIPS is the critical component for the system because it brings together all of the information for use by the forecasters, as discussed in the following section. The composite system of observation coverage will reside within a substantially restructured weather service that will include the following components: WFOs located near NEXRADs: RFCs collocated at WFOs; WFOs locater! on or near university campuses; thereby enhancing the interactions among forecast office staff, university faculty members, and students; a Science Operations Officer, on each WFO staff, who will be responsible for the development of new forecasting techniques; and WFO staff specially trained in modern forecasting techniques and newly emerging techniques. As in the past, the composite system is also augmented substantially by such things as experienced local spotter and cooperative observer networks, hydrometeorological instrumentation that supports the work of the WFOs and the RFCs; and highly skilled forecasting staff who consider local geographical characteristics in their day-to-day forecasting responsibilities. These local characteristics might include terrain effects that produce elevated heat sources and are, therefore, preferred regions for thunderstorm development. The NWS will use the MARD over the central United States to verify and test the composite system after all components are in place. The results of MARD should not be generalized to other geographical regions of the country due to differences in area weather and other factors. Comparison of the full spectrum of services of the modernized NWS and of the past generation of weather services will be impossible to quantify convincingly until all components of the new system have been installed. NETWORK CONSIDERATIONS In a series of developments extending from the early 1970s into the early 1980s, the NWS provided some NWS offices with a computer-based system for processing radar-generated digital reflectivity data into products that forecasters could use directly (Wilson, 1970; McGrew, 1972; and Devore, 1983~. This system, known as the radar data processor (RADAP Il), ultimately was available at 12 locations. At six of these sites, an interactive color radar display (ICRAD) capability provided

OCR for page 36
38 Assessment of NEXRAD Coverage A CO/VCEPt OF COVERAGE ., i_ G O E S ~ \ - ~ R H Y D R O __ k ''I--- X ~ H Y Din A S O S - - ~ ~T OP OF TR O POSPHERE A S O S ~SURFACE Figure 3-1 National Weaner Service concept of observation coverage when the modernization is fully implemented. Courtesy of NOAA/NWS.

OCR for page 36
Comparison of Pre-NEXRAD and NEXRAD 39 a color-map presentation of radar reflectivity data and derived products. ICRAD also made it possible to view the RADAP or products at a remote location. Although NWS forecasters experienced occasional hardware and software problems, the RADAP or operations were considered a success. This capability gave both forecasters and researchers valuable pre-NEXRAD experience. Most of the radars at RADAP rr sites were either directly replaced by NEXRADs, representing a clear improvement of both detection and algorithm capabilities, or were decommissioned following the installation of a NEXRAD nearby. The prior existence of RADAP or capability or an TCRAD display of products from a neighboring RADAP I! radar is one of the factors NWS should consider when assessing possible degradation of service at a specific location. While He NEXRAD network is being installed, an internal communications system is being used to distribute the "hourly digital precipitation array" of 2 km and 4 km data. In addition, the "radar coded message" from each NEXRAD site is being sent to the National Severe Storms Forecast Center in Kansas City, Kansas, to generate a national mosaic product. The NEXRAD information dissemina- tion service providers (Beer, 1991) also offer a national mosaic radar reflectivity product that is widely used in the commercial sector and in government, including the National Meteorological Center and the National Severe Storms Forecast Center. The NEXRAD system allows forecasters at each WFO to access some radar products from several neighboring radars. This information will be very important for observing and tracking signifi- cant weather as it approaches the WFO's primary geographic area of responsibility. This wider-area picture of the weather will provide information about the surrounding environment within which the WFO's primary responsibility lies. Some WFOs have multiple, associated principal user processors (PUPs), and others may obtain additional racier information for their area of coverage through the use of nonassociated PUPs. In the four locations with two associated PUPs, high-speed communications are provided to each radar product generator (RPG), and each PUP functions in an icientical manner. The nonassociated PUPs have access to any NEXRAD, but this access is through a low-speed, dial-up communication line, and WFOs can receive products only on a by-request basis. WFOs that require data from DoD NEXRADs for part of their service area have access only through low-speed lines. The NWS has instituted a continuing evaluation and upgrade program for the NEXRAD at the Operational Support Facility in Norman, Oklahoma. This facility is an essential element of the national network that will ensure that the associated services will improve over time. In the modernized NWS, AWIPS will provide interactive analytical capability, of which NEXRAD will be a part. This capability will assimilate data from satellites, surface instruments and human observers, lightning detection networks, radiosondes, and commercial aircraft, along with data from radars and from numerical models run at national centers. Forecasters will use this data base to evaluate the state of the atmosphere and to produce short-term warnings, "nowcasts," and forecasts of weather in support of public safety and other needs. The pane! emphasizes the need and importance for NWS to complete the modernization (including AWIPS) in a timely manner to avoid added risks of degraded coverage and service due to a prolonged transition to new technology. WARNING PERFORMANCE lithe overall performance of a system in producing forecasts, advisories, and warnings is an important criterion for evaluating the adequacy of radar coverage with respect to clegradation of ser- vice. Most of the available data on the performance history of the pre-NEXRAD system concerns the warning function. Although radar is not the only source of data on which storm warnings are based, it tends to play a principal role.

OCR for page 36
40 Assessment of NEXRAD Coverage Evaluation of warning performance requires data on warnings issued and on actual storm events, regardless of warnings issued. A warning is issued by the NWS for a particular county, or set of counties, for a specific period (generally 30-60 minutes). The warning is then logged systematically by the National Severe Storms Forecast Center. Data on storm events are accumulated and published periodically in Storm Data. When the NWS presented its analyses of verification data to this panel, the pane} noted that the historical storm data, in particular, are not as comprehensive as would be desired for rigorous scientific analysis. The pane! is encouraged by NWS efforts to collect more comprehensive storm reports as spotter networks and other networks in county warning areas are expanded and realigned to NEXRAD-equipped weather offices. The NWS customarily uses two measures to evaluate performance: the probability of detection (POD), which in this context refers to the fraction of storm events for which warnings were issued; and the false alarm ratio (FAR), which refers to the fraction of warnings where a storm did not occur. A warning system that was functioning well would have a high POD and a low FAR. ~ Warning perfor- mance exhibits considerable variation from station to station and from year to year for reasons that involve factors other than the radar performance. In addition to ordinary random variations, these reasons include both human factors and the performance of other components of the composite system, the storm climatology of the area, and the time of day of storm occurrence. The POD and FAR statis- tics represent ratios, and, during the panel's consideration of those ratios, concerns were raised about the methodology used to determine whether storms occurred or were reported. Therefore, the pane! found it necessary to examine the number of warnings and reported storm events in addition to the POD/FAR statistics. The pane! used information on weather fatalities as one gauge of storm severity for various phenomena. Data on natural-hazard deaths in the United States show that in 1992 and 1993 the primary causes of severe-weather fatalities include flash floods, lightning, and tornadoes, in that order. (NWS, 1993 and 1994~. These fatality statistics were consistent with similar statistics gathered over 30 years. Early Experience with NEXRAD Several of the NEXRADs have been in service long enough to provide preliminary data on storm-warning performance (Polger et al., 1994; Maddox and Forsyth, 1994~. Figure 3-2 shows severe storm warning statistics for six field offices. The POD and FAR statistics have been plotted versus time in years for each of these sites. The statistics shown for the early years represent the pre-NEXRAD era; those for later years indicate PODs ant! FARs after installation of a NEXRAD at each location. In all cases, POD and FAR performance has improved with the introduction of the new technology. Figure 3-2 also shows that, in general, the number of warnings issued since the advent of NEXRAD has increased. With more warnings being issued and a higher fraction of them being verified, as indicated by the lower FAR, it is clear that better service is being provided to the areas served by the NEXRAD-equipped NWS offices. At the same time, the overall increase in the number of reported storm events makes any attempt at quantitative comparisons difficult. It is unlikely Hat the actual incidence of storms has shifted upward since the advent of the NEXRAD. The increase is probably due to a more vigorous verification program, involving greater ~ The current verification system considers a storm to be covered by a warning, or a warning to be verified, if a reported storm occurs within an area and time period covered by the warning. Some warnings issued after a storm is already in progress are considered as verifications but only if the NWS received another report after the . . ~ warning Is Issued.

OCR for page 36
Companson of Pre-NEXRAD and NEXRAD Dodge City, KS 1 - enn 0.8 0.6 O 0 04 0.2 O "vv 400 ~ 300 O 0} 100 :q 1989 1990 1991 t992 1993 1994 . _ pr~NEXRAD _ NEXRAD - s POD FAR - ~ -#Warnings t- #Events 41 ~ Norman, OK 0.8 - ~= , , 0.7 - _, \~_._~_, ~ 0.6 - _f r ~ ~+~ ~ o0.5- -~~ ~: ~--r-~ --t-- O0.4 - _~ __ -_ ~ _ ~ ~ + 0.3- ---~ -~ ~ !--~ ~ ~--~ 0.2 0.1 O T~ ;7N ~i100 ~iOOO 900 0 c~ 800 00 ~ 600 80 81 82 '3~4 85 86 87 88 8~9q 91 92 93 94 year pre-NEXRA ~ Research Radar NEXRAD _ . 0 POD ~ FAR --~--#Waminge l" #EYents Houston' TX ' 1 0.8 0.6 o 8 o.4 . c~ 0.2 O 1989 1990 1991 1 1992 1993 1994 ~- pre-NEXRAD NEXRAD POD FAR --~- #Warnings ~- #Events | 500 eq 400 c 300 O . ' t#~, 200 3 I: 100 ~ Sterling, VA , 0.8 ;[ 0.6 o O 0.4 0.2 o 1989 1990 1991 |1992 1993 p94 pre-NEXRAD NEXRAD s POD - - FAR - .--#Wamings -~-#Events 1 0.8 0.64 , O 8 0~4- : Q. 0.2, .. '': O 1989 1990 1991 1 1992 1993 1894 , _ pre-NEXRAD NEXRAD | - POD FAR ~ -#War~ngs l. #Evenis | St. Louis, MO Melbourne, FL . 1 <"o::~4 O02. z: ~. . :" ~ ~.a _ -; ~';' - ';'L'';-; O 1989 1990 1991 11992 1993 ~94 Dr+NEXRAD NEXRAD 1 1 500 i I 4 c ~300 1 m ~1200 5 t '1100 :~- T 1 | ~ POD ~ FAR -* ~Namings .- #Events ~ Figure 3-2 History of severe-storm probability of detection and false-alarm ratio at six field offices before and after installation of NEXRAD. Also shown are the annual number of warnings and reported storm events. The 1994 data are preliminary and still undergoing validation checks by NWS. In the case of Norman, Oklahoma, an experimental Doppler radar was used operationally beginning in 1983. Courtesy of NWS.

OCR for page 36
42 Assessment of NEXRAD Coverage efforts to ascertain the presence of severe storms in the warning areas (e.g., by more phone calls to spotters). However, a more vigorous verification program, with no change in the radar or in the number of warnings, would produce similar changes in the POD/FAR statistics by converting some would-be false alarms into verified warnings. Thus, the trends in the POD/FAR statistics represent some indeterminate combination of improvements in the warning services and enhancements in the verification program. Figure 3-3 shows POD and FAR statistics as a function of distance from NEXRADs to the centers of the served County Warning Areas for the same six sites shown in Figure 3-2. The corre- sponding numbers of warning and reported events per unit area for the 2 years are also shown in the figure. These data, which are intended to illustrate the warning performance as a function of distance from a NEXRAD, reflect some puzzling features. For most of the sites, the PODs and FARs are rela- tively constant with increasing range. Yet, in most cases, the number of warnings per unit area decreases with increasing range. Based on discussion of radar-detection capabilities in Chapter 2, one would expect fewer storm detections at longer ranges. Moreover, if raciars were the major factor in the warning process, there should be fewer warnings at longer ranges, as is usually the case. But with fewer warnings, one would expect more missed detections and, consequently, a lower POD, in contrast to the pattern found in several cases. The reasons for these seemingly contradictory indications, and for the increase in warnings per unit area with distance from the Melbourne NEXRAD (see Figure 3- 2), are not clear. There is no apparent reason why the number of storm events should vary in any systematic way with distance from a radar. Yet, in most cases, the number of reported events decrease with increasing range. This reflects some kind of nonuniformity in the verification data, perhaps due to variations in population density or in the spotter networks. Such nonuniformity makes it difficult to evaluate the effect of range on warning performance wig any confidence. Furthermore, in each case, the number of warnings and reported events per unit area seem to be closely related. A similar correlation is evident in Figure 3-2. This correlation suggests some kind of coupling, either direct or indirect, between the issuance of warnings and the reporting of storm events. The pane! has been unable to account for this relationship and is concerned that it clouds the picture with respect to the effect of distance from a radar on warning performance. The pane! was unable to resolve these complicating factors, which need to be examined further in future studies. Degradation of service evaluations for specific sites will require NWS to (remonstrate that a county that is distant from a NEXRAD and associated WFO will receive, as a minimum, the same quality of warning service that was provided by the existing, nearby office in the past. That is, the new oi~f~ce's performance must be compared against that of the existing office that is equipped with a radar with pre-NEXRAD capabilities and pre-modernization components. Figure 3-4 shows the results of an independent study on tornado warning lead times (Bieringer and Ray, 1995~. As indicated by the figure, there is a clear and substantial improvement after NEXRAD technology is made available to forecasters. Polger et al. (1994) also analyzed the comparative performance of the NEXRAD and pre-NEXRAD eras for flash-flood warnings. The relevant data for flash-flood PODs and FARs for five sites are shown in Figures 3-5a and 3-5b. Substantial improvements are seen at all locations, with the exception of Norman, Oklahoma. There may be several reasons for this anomaly. First, it may be that even in the pre-NEXRAD era Norman and Oklahoma City forecasters had access to data from better experimental radar equipment wig computer processing capabilities (Maddox and Forsyth, 1994~. Second, special flash-flood forecast and warning procedures were in place in the pre-NEXRAD era. In addition, the pre-NEXRAD staff were thoroughly trained, and strong management-support practices were in effect.

OCR for page 36
Companson of Pre-NEXRAD and NEXRAD 43 Dodge City, KS (1992-93) u7s ~. .060 ~i~E .045 ~ .030 c, .015 120 160 200 240 km 1 11 t9 9 5 1 0 #counffes | ~ -- POD ~ FAR - ~ - W/A ~ - EIA I 0 40 8D l 1 0.8 ~0.6 o O 0 4 0.2 O Houston, TX 41992-93) 0 40 80 120 160 200 0 2 2 7 4 2 ,[ ~ POD +FAR .- ~W/& ,. FlA I .VI ;, .060 ~> .', u .045 . E .015 3 ' 240 km n 1$counties Sterling, VA (1992-93) cc 0.8 ~ 0.6 D 0~4 O 0.2 O o 40 80 120 160 2DO 1 11 19 9 5 1 ~ POD +FAR --. -W/A l- E/A | Nonnan, OK (1992-93) 1- ~__ 08 ' . 0.6: 0 c~ 0.4- O 0.2 o --': ~ i ~ t I ru/~ s _ , 4' _ 1- ! --- _~ .045 __ I I - .' ~,,._. _-~.~ !;- ~ `.030 - = ~==; ~_;.015 ~ n. 0 40 80 120 160 200 240 >240 km 1 4 10 9 14 11 6 3#COUllff" POD FAR - *--W/A l- E/A St. Louis, MO (1992-93) 1 _ .075 . oE ~ ~,Off3 O 0.2 .015 240 km O #counff - l ~; ~ ~ , ~ ~ ~ ~_ ~~ _ ,, 0 40 80 120 160 200 1 8 9 14 8 7 Melboume, FL (1992-93) u~ O ~ ~ _ w ~ - . ~ 0 #counties 0.8 O 0.6 O 0'4 ~ 0.2 O ~I ~:~ ,- T 1 t O 1 1.07S ~ ~1.o60- ~ 1 045 o .030 ~ =~.01S 80 120 160 200 240 km 3 t 0 0 0 #count~ I ~ POD ----FAR - ~ -W/A .- E./A | - Figure 3-3 Probability of detection, false alarm ratios, and number of warnings and events per unit area for 1992 and 1993. These data are shown wi~ distance from ~e NI1XRAD for ~e same six field offices depicted in Figure 3-2. Courtesy of NWS.

OCR for page 36
- ~ A en ~ ~ o o o to ~ - ~ - - ~ 1 1 1 1 , 1 1 1 in B B et: ~. an c A o o ~o US _. paupers saopell~o`L JO a8~U9339d 44 ~o ._ - u~ ~ 1 - is" ._ Ed In -4 en In 8 c) ~ .= ~ ~q ~ o ~ o ~ a, ~ A> ~4 - ~ - ~ o os ~ Q ~ a, TIC ~ To ID ~ a: o ~ 4,,) ID of .0 3 ~ .~ ._ oo a, ~ .c o {Q ~ ~~ as c: :- ~D 00 3 =: ~ ~ ~a' ~= E 3 o ~ o _ ~4 ~ o U] .~; o 4,,) o es ~ ~ ~q o ~ - , o ~ ~ 04 e . . 8 V' z

OCR for page 36
o ~ rat - ~ ~To ~ to pa 6D to He to oh :b p oh Q PA - ~D p p o hi. ~ oh S ~ C t P ~ rat _ _ ~ p be. of g O O . ~X \ ~ O O V: (D - . O :D ~ :5' W_ , o ~ ~ Vie rat in; an ~ ~ ~OO OO ~ P logO ~I ' \ . . ? i ~ ~ ~ i. i.: ~ i. ? ? ~- t~ - ~ ~ ~E ~ ~ . -. .~ it i~E - ~ -- . o p ~-.~.--~ ~. ~-.~.~.~.-.~.- ~.~.~-~ ~.~ ~. ~.- ~.~; ~-.~-,.~.~,~ ~ -~.~- ~-~-~- ~.~-~ ~ ~- ~ ~ _ L:~':'~'~'.~:51~.~,~,,,~,~.-.~ ~,~''~ ~: ~'~ ~'~ ~ ~,~ ~ ~ ~'~ ~ ~ ~ ". ~,.~.~'~.~i,:~.~,:. ~.j~,.~,,~"':~',~,~,~2~'~.:2'~.~:~2~',~"2.'~'~.~'~ ~'~'~'~':~: .,.~.i .~;'~'.~'~,~ ~' ~ ~,~'~. ~"~'~: ~.~'~.~'~' ,~."~.~.~',~'~'~'~.~"~ ~ O ~'~ ~ ~$~ ~'~g'~.~.~ ~'~ ~'~ ~ ~.~ ~'~.~ ~.~'~.~.~''~'~.'.~ ~.~'~ ~ ~',.''~ ~ ~ ~ ~ ~. . . . ~,:~,~.'.~,~,. ~ ~,:~,~, ~,~ ~.~.~. '~.~,~':.i'~:.',.~'~'~'~"'~''~'~i~" ~''~'~' m0 ~ ~ ~ ~ ~ 1 ~ O (Q o ~r _. Ot p 6g ~D _. p D U) o o o '5 _. _. o - (D ~D ~. o 1 ~

OCR for page 36
46 Assessment of NEXRAD Coverage Comments and Conclusions on Warning Performance It is apparent that warning performance at all locations analyzed by the pane! has improved substantially with the introduction of NEXRAD. PODs, FARs, and lead times for tornadoes and flash floods have also improved significantly. It is reasonable to expect that performance will improve further as forecasters gain additional experience with the system, forecasting methods are streamlined, and algorithms are refined. This general result is consistent with what would be expected from the information found in the tables and coverage maps in Chapter 2. Even though many of the early NEXRAD sites had specially qualified staff, the pane! is confident that overall increases in warning performance are likely at all sites. This improvement will stem from the use of the new technology, the extensive training provided to all the NEXRAD operators, the sharing of lessons learned within the NEXRAD community, the continuing development efforts of the Science Operations Officers and the Operational Support Facility, and similar efforts O The data regarding warning performance as a function of distance from a NEXRAD site present a more mixed picture. In all likelihood, the warnings and storm reports for the more distant locations concern mostly larger and more intense storms. If further study shows that smaller and less intense storms are less likely to result in warnings and storm reports from a NEXRAD that is more distant from them than a nearby, older radar would be, then this effect of distance may result in a degradation of these particular services at some locations. In the panel's judgment, insufficient data currently exist to permit definitive conclusions to be made. Although the early results are very promising, further testing in the MARD and at diverse geographic locations will be necessary to draw more general conclusions regarding the quality of weather warnings as a function of distance from the radar. The pane! concludes with confidence that the introduction of NEXRAD will significantly improve overall storm detection and warning performance.