6046-Fire Notes 57#2

Fire notes ManagementVolume 57 • No. 2 • 1997

Wildland Fire Weather

United States Department of Agriculture Forest Service Fire Management Notes is published by the Forest Service of the U.S. Department of Agriculture, Washington, DC. The Secretary of Agriculture has determined that the publication of this periodical is necessary in the transaction of the public business required by law of this Department.

Subscriptions ($8.50 per year domestic, $10.65 per year foreign) may be obtained from New Orders, Superintendent of Documents, P.O. Box 371954, Pittsburgh, PA 15250-7954. A subscription order form is available on the back cover.

Dan Glickman, Secretary April J. Baily U.S. Department of Agriculture General Manager

Michael P. Dombeck, Chief Donna M. Paananen Forest Service Editor

Mary Jo Lavin, Ph.D., Director Fire and Aviation Management

The United States Department of Agriculture (USDA) prohibits discrimination in its programs on the basis of race, color, national origin, sex, religion, age, disability, political beliefs, and marital or familial status. (Not all prohibited bases apply to all programs.) Persons with disabilities who require alternative means for communication of program information (braille, large print, audiotape, etc.) should contact the USDA Office of Communications at (202) 720-2791.

To file a complaint, write the Secretary of Agriculture, U.S. Department of Agriculture, Washington, DC 20250, or call (800) 245-6340 (voice) or (202) 720-1127 (TDD). USDA is an equal employment opportunity employer.

Disclaimer: The use of trade, firm, or corporation names in this publication is for the information and convenience of the reader. Such use does not constitute an official endorsement of any product or service by the U.S. Department of Agriculture. Individual authors are responsible for the technical accuracy of the material presented in Fire Management Notes. Fire Management notes Volume 57 • No. 2 • 1997

CONTENTS Lightning Detection and Data Use On the Cover: in the United States ...... 4 Brenda L. Graham, Ronald L. Holle, Raúl López

Tracking Thunderbolts: Technology at Work ...... 11 Phil Sielaff

Current Status of the Wildland Fire Assessment System (WFAS) ...... 14 Robert E. Burgan, Patricia L. Andrews, Larry S. Bradshaw, Carolyn H. Chase, Roberta A. Hartford, and Don J. Latham

Lightning strikes such as this one in WFAS Requires a Variety of Weather Information ..... 18 Alabama ignite about 10,000 Robert E. Burgan, Larry S. Bradshaw wildland fires in the United States each year. Lightning detection and other aspects of fire weather High Resolution Fire Weather Models ...... 22 important to the fire community are discussed in this issue. Francis M. Fujioka Photo: Courtesy of Johnny Autery, Dixons Mills, AL, ©1997. Making Sense of Fire Weather...... 26 Brian E. Potter

SODAR and Decisionmaking During the Fork Fire ...... 28 Fred Svetz and Alexander N. Barnett

SHORT FEATURES

Safety from a Lightning Strike ...... 10 Brenda L. Graham, Ronald L. Holle, and Raúl E. López

Guidelines for Contributors...... 31 LIGHTNING DETECTION AND DATA USE IN THE UNITED STATES

Brenda L. Graham, Ronald L. Holle, and Raúl E. López

ach year, lightning ignites about 10,000 wildland fires in Present lightning detection networks provide fire E the United States (DeCoursey managers in the United States with information et al. 1983). Wildland fire manag- about potential lightning-caused wildfires. ers in the 11 Western States and Alaska have routinely used near- Lightning location information can also be used real-time lightning detection net- for other purposes such as planning for work data to determine likely areas prescribed natural fire. for new ignitions. Few of them, however, have had training in the sensor technology, location accu- 1) A thunderstorm cloud (cumu- 4) As the negatively charged step racy, and detection efficiency of lonimbus) becomes predomi- leader approaches the ground, these lightning detection net- nately positively charged at the positively charged “streamers” works. The intent of this article is top and negatively charged in respond by traveling skyward to- to provide background in these the lower part (Uman 1969). ward the step leader. Streamers areas so fire managers can make 2) A typical CG lightning flash be- typically move up from the tall- full use of this technology. gins as a “step leader” that est well-grounded object like a “jumps” about 150 feet (50 m) at tree or building (fig. 1). Lightning Itself a time toward the ground from 5) When the step leader and a There are two types of lightning: the negatively charged region at streamer meet about 150 to 300 the bottom of the cloud (fig. 1). feet (50 to 100 m) above ground, • Cloud-to-Ground (CG) flashes 3) The step leader travels very they form an electrical channel and quickly and takes about 20 mil- that is about as wide as your • Cloud discharges. liseconds to travel about 2 miles thumb. (3 km) (Uman 1969). The following brief description is limited to CG lightning since it is the type that initiates wildland Ð Ð Ð Ð Ð Cloud base fires. Ð Ð Ð Ð Ð Ð Ð Ð Ð Ð Ð Ð A CG lightning flash (commonly Ð called a “strike” in the wildland fire Ð Negatively charged step Ð community) is composed of a se- Ð leader path–average Ð “jumps” are 150 feet (50 m) ries of events that occur very At 150–300 feet quickly. These events are roughly (50–100 m) Ð as follows: above the Positively charged ground, the + streamer reaching skyward step leader to meet step leader and streamer Brenda Graham is a fire weather meteo- will meet to rologist for the U.S. Department of form an Commerce, National Oceanographic and electrical + + + + + + Atmospheric Administration (NOAA), channel. + + Ground National Weather Service, Medford, OR; Ronald Holle and Raúl López are research meteorologists for NOAA’s National Severe Figure 1—A step leader jumps toward the ground and a streamer reaches from Storms Laboratory, Norman, OK. the tree to meet it.

4 Fire Management Notes 6) A series of electrical current surges called return strokes follows. Two to four return strokes, which have the net effect of low- ering negative charge to the ground, comprise a typical lightning flash (Uman 1969). This is the origin of the term “negative lightning,” which is the most common type of CG lightning. “Positive lightning” occurs when a positive charge is lowered to the ground (about 5 to 10 percent of CG flashes are positive). Figure 2 shows a common form of damage when lightning strikes a tree. Figure 3—Three cloud-to-ground (CG) lightning flashes over Norman, OK, photographed in about 1 second by moving the camera from left to right. Photo: W. David Rust, Norman, Figure 3 shows three CG flashes OK, ©1985. photographed in about 1 second by moving the camera from left to Lightning is tremendously ener- waves that can travel great dis- right. The first return strokes of getic. The typical CG lightning tances. Lightning detection net- the two main flashes are to the left; flash carries a peak current of works are designed to use this they have several branches. The left 30,000 amperes. In comparison, information. flash has seven return strokes with most household circuits in the a single-stroke flash just to its United States carry 15 amperes, Lightning detection sensors in the right. The flash on the right has six and a 100-watt light bulb uses United States use “direction-find- return strokes. about 1 amp (Uman 1971). In a ing” (DF) or “time-of-arrival” fraction of a second, the air imme- (TOA) technologies. The perfor- diately around the lightning chan- mance of a lightning detection net- nel is super-heated to 15,000 to work is affected by the character of 60,000 °F (8,000 to 33,000 °C), its sensor technology. which is much hotter than the sun’s surface (Uman 1969). This DF Technology. Based on tradi- rapid heating creates the shock tional radio direction-finding tech- wave we hear as thunder. niques, DF technology sensors have two loop-shaped antennas Most CG lightning flashes in wild- perpendicular to each other. A land areas do not create new fires lightning flash’s magnetic field in- because the fuels are not exposed duces signals in both loops that are to high temperatures long enough determined by factors such as the for combustion to begin. However, flash’s current flow and the orien- when wildland fuels become very tation of the lightning channel to dry due to some environmental the antennas. When the radio sig- condition such as prolonged nal produced by a lightning flash drought, the likelihood of ignition arrives at the antenna, onsite soft- increases. ware immediately compares the ratio of signals in the loops to Technologies Used in identify the flash’s azimuth. All azi- Figure 2—Damage to tree that was struck Lightning Detectors muths from all sensors are sent to by lightning in New Bern, NC. Photo: Kevin Kelleher, National Severe Storms CG lightning flashes produce Laboratory, Norman, OK, ©1995. uniquely shaped electromagnetic Continued on page 6

Volume 57 • No. 2 • 1997 5 a central processor that uses azi- muths from two or more sensor DF1 sites (triangulation) to identify the probable area of flash location (fig. 4) (Holle and López 1993). ± 1° azimuth error DF2 TOA Technology. TOA technology involves an evaluation of when the electric pulse generated by a return stroke arrives at multiple re- Computed flash ceivers. To determine location, the location electric pulse arrival time at each Area of probable receiver site is sent to a central flash location analyzer that computes the difference in arrival times between pairs of receivers. A plot of all the points between a pair of receivers where that time difference is possible de- fines a hyperbola (fig. 5). Compari- sons between four to six sensors provide the optimum location of Figure 4—Flash location by triangulation from two direction-finding (DF) antennas. the flash. Figure 5 shows a probable flash location defined by the intersection of the hyperbolas from two receiver pairs. These receivers must be synchronized to the same Hyperbola branches dependable time source for this defined by time-of-arrival system to work (Holle and López differences 1993). Receiver 1 Receiver 2 Waveform Discrimination. The electromagnetic waves sent out by lightning flashes have particular forms that sensors can resolve (fig. 6). The sensors perform onsite comparisons of detected flash Receiver 3 waveforms with a statistically de- Stroke location termined lightning waveform obtained from previous studies of CG Figure 5—Probable flash location from two pairs of time-of-arrival (TOA) receivers. flashes. If the comparison is a close match, the data are accepted and • Width between peak amplitude Networks in the sent to the network processor. This and when the signal drops to an United States waveform discrimination helps established threshold. keep spurious noise and cloud The BLM Network. From 1976 flash signals from being reported. through 1996, fire managers in the Figure 6b shows some signals from 11 Western States and Alaska re- negative CG discharges. Note these ceived CG lightning data from the Figure 6a shows some of the key waves have the same features as features of CG lightning flashes: U.S. Department of the Interior’s the one in 6a, only they are in- Bureau of Land Management • The (step) leader pulse prior to verted because they are from nega- (BLM) networks. These networks discharge, tive flashes. Figure 6c shows cloud used DF technology. CG lightning • Shape of rise-time portion of the discharges, and they lack the same data were distributed to wildland wave, and defining features. fire managers in near real time via

6 Fire Management Notes the BLM IAMS computer network. poration (Global Atmospherics, short whip antenna, and figure 7 Beginning in 1997, the CG flash Inc. (GAI), of Tucson, AZ*) oper- shows an IMPACT sensor (note the data in the 48 contiguous States ates and maintains this network. loop antennas). Figure 8 shows the will come from the National Light- GAI’s recently updated NLDN uses NLDN sensor network as of May ning Detection Network (NLDN), both DF and TOA technologies. 1996 with the type of sensor tech- but the Alaska network will remain The network is composed of a mix- nology at each site. the same for at least several more ture of sensors, some sites with years. This change in data source TOA only and some with DF and The sensors using only TOA tech- will be transparent to fire manag- TOA together. The sensors with nology have limited CG flash wave- ers in the 11 Western States be- combined DF and TOA technology form discrimination and report cause data distribution will are called IMPACT (Improved Per- only the flash-arrival time. IMPACT continue through the existing formance from Combined Technol- sites use more detailed waveform IAMS system. ogy). TOA sensors are basically a discrimination and report both azimuth and arrival-time data. These *The use of corporation names and/or their products is The NLDN. The current NLDN was for the information and convenience of the reader and data are sent to a single central established a few years ago in the should not be misconstrued as an official endorsement by the U.S. Department of Agriculture or the Forest 48 contiguous States; a private cor- Service. Continued on page 8

a. Waveform discrimination

Rise time Width } Peak amplitude

Threshold Threshold crossing crossing Enable threshold Baseline Leader pulse

b. Cloud-to-ground discharges

-5 -10 0

-5 -10 010203040 Time (µsec)

c. Cloud discharges

0 -5 0

-5 -10 010203040 Time (µsec)

Figure 6—a. Characteristic features of cloud-to-ground (CG) waveform, b. negative CG flashes, c. cloud flashes. (Information provided courtesy of Global Figure 7—IMPACT lightning sensor used in National Lightning Detection Network Atmospherics, Inc., Tucson, AZ.) (NLDN). Photo: Ken Matesich, 1995 (courtesy of Global Atmospherics, Inc., Tucson, AZ).

Volume 57 • No. 2 • 1997 7 processor that resolves all the in- it does not fall within the statisti- Which Flashes formation into individual CG cal CG flash profile. Cause Fires? flashes and locations. GAI gathers, About 30 percent of all lightning processes, and archives all the Location accuracy of networks also flashes are cloud-to-ground data. depends on sensor layout and (Krider 1994), and it would be where the CG flashes occur relative ideal to know which are most Network Performance to the sensors. Location accuracy likely to ignite a fire. Some CG DF sensors are subject to errors with networks based on DF-only lightning flashes have a long pe- caused by topography and human- sensors (like the BLM networks) riod of continuing current (fig. 9), made structures near the site. Lo- has been shown to vary from about meaning that the current is slow cal terrain and buildings can .25 to 2.5 miles (0.4 to 4 km). GAI’s to decay after reaching a peak. The reflect, absorb, and reradiate sig- upgraded NLDN has a location ac- “blur” between some successive nals. However, once each site’s er- curacy of .3 mile (0.5 km) due to strokes shown in figure 3 is due to rors are documented, they are its combined use of DF and TOA continuing current. There is some corrected during data processing technologies that allows the net- evidence that long continuing cur- to minimize potential inaccura- work to determine flash location rent (LCC) flashes are more likely cies. more accurately. to start fires than other CG flashes (Fuquay 1980). An LCC flash Network detection efficiency (per- The BLM network detection effi- would expose fuels to heat longer, centage of actual CG flashes re- ciency has been around 60 to 70 thus increasing the chance of igni- ported) is a function of sensor percent and possibly as low as 40 tion compared to an equal non- layout and where the CG flashes percent in some areas. Networks LCC flash. A limited set of data occur relative to the sensors. In with configurations like GAI’s have from experiments suggests that some cases, CG flashes remain un- a higher network detection effi- positive flashes are quite likely to reported because their signals are ciency on the order of 70 to 90 per- have LCC, while a lower percent- too weak to be received by enough cent (Holle and López 1993). Fire age of negative flashes have LCC sensors for calculating a location. managers in the 11 Western States (table 1). No lightning detection A CG signal that passes the wave- will probably notice about 40 per- network in use today makes any form compatibility test can still be cent more flashes reported by the distinction of current flow dura- discarded if it does not achieve a NLDN compared to the BLM net- tion in CG flashes. GAI is conduct- minimum signal strength thresh- work due to improvements in net- ing research in this area, however, old value. In some cases, the wave- work performance. Reported and expects to add this capability form itself may be rejected because flashes could be even greater for to its network in the future. some areas. Lightning Data Use by Wildland Fire Managers While the most common use of lightning detection network data by wildland fire managers involves near real-time data applied to operations, historical data are also being studied. For example, the Fishlake National Forest in central Utah is using historical lightning data to evaluate the correlation between lightning and ignition oc- currences over a given area. The patterns that emerge from the analysis may be valuable input for deciding which areas of the forest would be good candidates for pre- Figure 8—NLDN sensor types and locations in the Continental United States. scribed natural fire (Chappell, per-

8 Fire Management Notes present network detection efficiencies are very adequate for most wildland fire applications.

Literature Cited Anderson, K. Year unknown. Frequently asked questions (FAQ) about lightning. Canadian Forest Service, Edmonton, Current Alberta. Taken from the following Internet location (web page): http:// www.nofc.forestry.ca/~kanderso Chappell. Unpublished communique available at the office of the first author, Na- tional Weather Service, 4003 Cirrus 0 500 Drive, Medford, OR 97504. Time (µsec) DeCoursey, D.G.; Chameides, W.L.; McQuigg, J.; Frere, M.H.; Nicks, A.D. 1983. Thunderstorms in agriculture and Figure 9—Typical current waveform for a positive cloud-to-ground flash. in forest management. In: Kessler, E., Adapted from Uman (1987). ed. The thunderstorm in human affairs. Norman, OK: University of Oklahoma Press, 67-87. Table 1—Some characteristics of negative and positive CG lightning flashes Fosdick, E.K.; Watson, A.I. 1995. Cloud-to- (Anderson, adapted from Uman 1987). ground lightning patterns in New Mexico during the summer. National Characteristic Negative Positive Weather Digest. 19(4): 17-24. Fuquay, D.M. 1980. Lightning that ignites Percent of occurrence 90 10 forest fires. In: Proceedings, 6th Confer- ence on Fire and Forest Meteorology, Average peak current (kA) 30 35 April 22-24, Seattle, WA. Washington, DC: Society of American Foresters: 109- Average number of strokes per flash 3-4 1 112. Percent with long period of Holle, R.L.; López, R.E. 1993. Overview of real-time lightning detection systems continuing current (LCC) 20-40 50-100 and their meteorological uses. NOAA Technical Memorandum ERL NSSL- 102. Norman, OK: National Severe Storms Laboratory. 68 p. sonal communication). When us- United States and Alaska until Krider, E.P. 1994. Physics of lightning to- ing historical data to develop a recently. Now that the NLDN is in day. Revue générale de l’Electricité. lightning climatology (pattern) for place, CG lightning data are avail- 6: 6 p. López, R.E., and Holle, R.L. 1986. Diurnal an area, it is important to use the able in all wildland areas of the 48 and spatial variability of lightning activ- longest record possible. Lightning contiguous States. The combina- ity in northeastern Colorado and central climatology studies have already tion of DF and TOA technology in Florida during the summer. Monthly Weather Review. 114(7): 1288-1312. been completed for several West- the NLDN has improved detection Uman, M.A. 1971. All about lightning. ern States (López and Holle 1986, efficiency and location accuracy. Mineola, NY: Dover. 167 p. Watson et al. 1994, Fosdick and While some CG lightning flashes in Uman, M.A. 1969. Lightning. Mineola, NY: Watson 1995). your area may not be reported by a Dover. 298 p. Uman, M.A. 1987. The lightning discharge. network, this situation should not Orlando, FL: Academic Press, Inc. 377 p. Conclusions have a significant impact on the Watson, A.I.; López, R.E.; Holle, R.E. 1994. value of the data. Since most thun- Diurnal cloud-to-ground lightning pat- CG lightning detection network terns in Arizona during the Southwest data were available solely to wild- derstorms produce several CG Monsoon. Monthly Weather Review. ■ land fire managers in the Western lightning flashes per minute, 122(8): 1716-1725.

Volume 57 • No. 2 • 1997 9 SAFETY FROM A LIGHTNING STRIKE

Brenda L. Graham, method: Start counting seconds The best option is to go inside a Ronald L. Holle, and (one-thousand-one, one-thousand- public building or a residence. Raúl E. López two, etc.) from the time you see Metal-topped buildings with the flash until you hear the thun- stone or other nonconducting Everyone who works outdoors der. Estimate that lightning is 1 walls are not safe. Don’t touch should be aware of the inherent mile (1.6 km) away for each 5 sec- anything connected to the power threat of being struck by light- onds counted. If lightning is occur- and phone lines or plumbing ning. On average, at least 80 ring closer than 5 miles (8 km) (or pipes. The second best option is people are killed by lightning 25 seconds away), go to a safe shel- to go inside a vehicle with a solid and about five times as many are ter immediately, if possible. Once metal top, but don’t touch the injured each year. These acci- lightning is within 5 miles, the sides of the vehicle. dents may occur due to people’s next flash may occur where you ignorance of how dangerous are, so plan to reach safe shelter Last Minute lightning is, their inability to before lightning occurs that close. If you are caught outside with recognize the potential threat lightning nearby or your hair from lightning, or their lack of Lightning Nearby stands on end and none of the knowledge about lightning safety When outside and far from vehicles above options are available: rules. By heeding the following or buildings, seek a thick grove of advice, you can reduce the • Crouch on the balls of your small trees surrounded by tall chance of death or injury when feet with your head down, and trees. Stay away from individual lightning is nearby. grasp your knees with your trees. hands. This crouching position Before Storms • Don’t stand in an open area minimizes your contact with Develop more than 100 yards (91 m) the ground, which can be across. quite conductive, especially if • Be constantly aware of how • Don’t be the highest object, such wet. long it will take to reach safe as on a ridge, rock outcropping, • Don’t lie flat on the ground or shelter if a thunderstorm or roof. touch the ground with your occurs. • Don’t be in the water or in a hands. • Observe when new thunder- boat. • If in a group, spread out so storms start to develop. • Don’t be near or in contact with that fewer people are likely to • Post a lookout to observe and anything taller than its sur- become victims. communicate the lightning roundings that may be prone to threat. being hit by lightning. When Lightning • Don’t be near or in contact with Strikes When Thunderstorms large metallic objects such as an- Develop Apply CPR to anyone rendered tennas or metal fences. unconscious by lightning. ■ Estimate your distance to lightning using the flash-to-bang

10 Fire Management Notes TRACKING THUNDERBOLTS: TECHNOLOGY AT WORK

Phil Sielaff

he technology involved in the activity.) As increased experience Automated Lightning Detec- For over 20 years, the and detection efficiencies were at- T tion System (ALDS) has been Automated Lightning tained in Alaska, the NWS began to evolving for over 20 years. The sys- Detection System (ALDS) use the ALDS to supplement tem currently can provide informa- weather radar to detect thunder- tion on lightning “ground strike” has been evolving; as storm activity at long distances activity to land managers and fire new technology is (outside of weather radar range). officials in real time. As a result of developed, the system these two decades of experimenta- will continue to improve Because of their satisfactory expe- tion, others around the world are riences in Alaska, the BLM decided able to use the ALDS to help them in the future. in the late 1970’s to place instru- anticipate natural wildfire occur- ments in the Western United rence. States and establish the first opera- ment at what was then called the tional ALDS network. Background for Boise Interagency Fire Center— Technology developed the prototype of ALDS. The BLM next worked with a new Development The initial operations in Alaska commercial firm—Lightning Location and Protection Inc. In the early 1970’s, fire managers demonstrated that ALDS—when (LLP)*—to place and operate the regularly asked for less expensive used in conjunction with National ALDS, supplying coverage over and safer ways of detecting light- Weather Service (NWS) weather roughly 95 percent of the States ning ground strike activity com- radar—could achieve significant located west of the Rocky Moun- pared with the standard practice of savings in isolating “active” cells. tains. flying aircraft behind storms to vi- (An “active cell” is one that is pro- sually track lightning. The USDI’s ducing lightning ground strike Continued on page 12 Bureau of Land Management (BLM) responded by developing a pilot remote system to track lightning strikes in Alaska. Since then, the BLM has extended the lightning detection network to the Western United States. (See fig. 1 for locations of the BLM’s lightning detection stations in the West.)

Working with the University of Ari- zona, Tucson, AZ, the BLM’s Office of Scientific Systems Development in Denver, CO—jointly with the Di- vision of Communications Manage-

Phil Sielaff is the leader of the Remote Sensing Support Group for the United States Department of the Interior, Bureau of Land Management, National Inter- Figure 1—Map of the Western United States showing locations of Bureau of Land agency Fire Center, Boise, ID. Management (BLM) lightning detection stations in 1996.

Volume 57 • No. 2 • 1997 11 As we operated the early system, many other interested “observers” began to see what this network could do operationally in real time. Thus the NWS, Federal Aviation Ad- ministration, Department of De- fense, Department of Energy, and many private firms began to monitor our efforts. Network Products Having the right information in a form that local managers can readily use for real-time fire management was the BLM’s primary goal in developing ALDS. For example, figure 2 illustrates lightning Figure 2—Locations of 234 lightning strikes in the Western States for 2 hours on strikes for 2 hours on May 14, 1996, May 14, 1996. over the Western United States. The local fire management staffs in these “active” areas have the capabilities at their respective locations to zoom in on this activity. As the local user begins to actively use the ALDS information, additional fire management information is added to the ALDS plots to provide a com- posite of the “big picture” at the user’s particular location. Because it is clear in figure 2 that there is a great deal of thunderstorm activity near the Oregon and Idaho border, local fire managers would actively process all available data to implement action plans (e.g., move to an elevated planning level, pre-position pumper crews, dispatch recon- Figure 3—Positive (red zigzags) and negative (black dashes) lightning strikes in Idaho naissance crews). and Oregon from 12:30 to 8:00 p.m. on May 14, 1996.

Figure 3 shows how the BLM’s These types of graphics are pro- in still water.) Each distinct signal lightning detection system can duced in real time at each local is picked up by the DF’s, which “zoom in” to track 319 strikes in BLM user site that has lightning have specially designed antennas Idaho and Oregon during 7-1/2 ground strike activity. The process and electronics to measure the di- hours on May 14. Note that positive begins when direction finders rection to the ground strike. The lightning strikes are shown with (DF’s) (electronic sensors) detect DF’s determine the direction and red zigzags and negative strikes lightning ground strikes. Every transmit that and other informa- with black dashes. lightning flash generates a distinct tion such as strike polarity and sig- signal that travels outward uni- nal strength to a Position Analyzer formly at nearly the speed of light. (PA). The Position Analyzer (a *The use of corporation names and/or their products is for the information and convenience of the reader and (The movement of this signal is computer-like device) performs should not be misconstrued as an official endorsement similar to the radiation of circles numerous mathematical calcula- by the U.S. Department of Agriculture, the Forest Service, or USDI Bureau of Land Management. around a pebble after it is dropped tions to process incoming DF data,

12 Fire Management Notes triangulates the vectors, validates the information, and then pin- points the exact location (latitude and longitude coordinates) of each lightning ground strike. The final processed PA information is auto- matically sent to any authorized field user who wishes to access it.

Figure 4 illustrates more closely the 319 lightning strikes also shown in figure 3 and shows boundaries of counties. Results of Monitoring Efforts As a result of observing the light- Figure 4—Map of southwest Idaho showing county boundaries and locations of 319 ning locating system and the prod- lightning strikes on May 14, 1996. ucts derived from the system, LLP and other lightning detection com- result, a contract was awarded to will not redistribute lightning data panies began to market and sell Global Atmospherics, Inc. (GAI) of to other Federal or State users. their systems to other government Tucson, AZ, to supply national The only exception to this provi- and nongovernment users. These lightning detection information for sion is if Federal or State fire of- local networks proliferated during the conterminous United States fices are collocated with the BLM. the early and mid-1980’s. The Elec- and some maritime areas. The ser- Those cooperators may continue to trical Power Research Institute, in vices provided under this 5-year use the data at the particular field conjunction with the State Univer- contract will provide increased ca- office but cannot redistribute or sity of New York at Albany, began pabilities for the BLM and many broadcast the information beyond to consolidate all of these indepen- other Federal users. Because Alaska that point. (This provision can be dent detection networks into a “na- has its own unique requirements, modified if the collocated Federal tional” system to supply coverage GAI cannot supply lightning data or State fire office attains rights throughout the conterminous for that State at this time. There- from GAI to distribute the data to United States. As a result of these fore, the BLM will continue to its user community.) and other efforts, for the first time operate its own ALDS in Alaska. at the beginning of 1990, lightning The BLM’s lightning detection ef- data on a quasi-national basis were Currently, the BLM is receiving Na- forts will continue to evolve as available from the private sector. tional Lightning Detection Net- improvements in technology are work (NLDN) data at our National found and implemented. Com- Looking to the Interagency Fire Center in Boise, pared to our standard fire manage- Private Sector ID. We use the existing Initial At- ment practices of just 20 years ago, Because of recurring costs associ- tack Management System/Auto- the efforts have been well worth ated with operating a large ALDS matic Lightning Detection System the investment in time, energy, network and the prospect of having (IAMS/ALDS) telecommunications and money. to replace older ALDS equipment, network to distribute NLDN data to the BLM began looking to the pri- the BLM real-time user community For more information about ALDS vate sector as a possible alternate (up to 70 file server locations). Ne- or NLDN data, please contact Phil source for lightning data. In 1995, gotiations are currently underway Sielaff, USDI, Bureau of Land Man- the NWS, working through the Of- with other USDI fire management agement, 3833 South Development fice of the Federal Coordinator for staffs to see how they will attain Ave., Boise, ID 83705-5354, Meteorology, took the lead to con- data under this new contract ar- telephone 208-387-5363, solidate all Federal users interested rangement. Under current contract fax 208-387-5397, ■ in procuring lightning data. As a provisions, the BLM cannot and e-mail: [email protected]

Volume 57 • No. 2 • 1997 13 CURRENT STATUS OF THE WILDLAND FIRE ASSESSMENT SYSTEM (WFAS)

Robert E. Burgan, Patricia L. Andrews, Larry S. Bradshaw, Carolyn H. Chase, Roberta A. Hartford, and Don J. Latham

he Fire Behavior Research proper application of current frag- Work Unit (RWU) of the Inter- The Wildland Fire mented systems. T mountain Research Station has Assessment System been developing the Wildland Fire (WFAS) will use the The fire behavior component of the Assessment System (WFAS) since WFAS will be based on the FARSITE 1994. The WFAS will eventually best methodology Fire Area Simulator (Finney 1995) combine the functionality of the available to provide fire and will not be discussed here. In- current fire-danger rating system potential assessments stead, we are focusing in this article (Deeming et al. 1977) and the fire for the 21st century. on broad area fire assessment of the behavior system (Andrews 1986). nature typically associated with fire The new system will assess fire po- danger rating. tential across spatial scales ranging Fire behavior must have weather from national to site specific and inputs matched as closely as pos- Development and time scales ranging from near-real- sible to the fire site. Implementation Plan time to 5-day forecasts. The out- We are using a three-phase strategy puts will be useful for fire Fire danger ratings have tradition- to develop the WFAS and have management tasks ranging from ally been relative values or dimen- defined the products and processes strategic planning to current fire sionless indexes, whereas fire that will become available to clients situation analyses (Burgan and behavior assessments are specific, once each stage is finished. This ap- Bradshaw 1997). physical descriptions of expected proach provides a natural transition fire characteristics such as rate of from the existing piecemeal sys- Traditionally, the terms “fire dan- spread, flame length, and fire inten- tems to a new integrated system. ger” and “fire behavior” have de- sity. This development and implementa- scribed assessments of fire tion plan is, of course, dynamic. potential on different time and As we have been developing WFAS, Based on user feedback and on the space scales. Fire danger has been our goal has been to remove the availability of new technology and evaluated routinely, at least once a “seams” that exist between the research products, we expect the day, for broad areas. Fire behavior current systems (Rothermel and WFAS to change during develop- assessments, however, are made as Andrews 1987). For instance, our ment stages. The three phases and needed for specific sites. Both use new “seamless system” will move the planned schedule are as follows: weather that has been forecasted easily from evaluating fire danger • Phase 1—1995 to 1996: Spatial or measured, but fire danger— over a broad area to assessing fire display of weather and current because of its routine and regular behavior at a specific location. computation—requires regularly National Fire-Danger Rating System (NFDRS) values. formatted and routinely available As we develop WFAS, we will con- data for automated processing. • Phase 2—1996 to 1997: Improve- tinue to resolve incompatibilities ments to NFDRS calculations. such as different sets of fuel models • Phase 3—1998: Major fuel and Robert Burgan is a research forester, Pat and equation differences. The fire modeling updates and Andrews is a research physical scientist, WFAS framework will address the Larry Bradshaw is a meteorologist, gridded weather models. Carolyn Chase is a mathematician, issue of weak linkages among cur- Roberta Hartford is a forester, and Don rent systems—fire behavior, fire Phase 1. In the completed phase 1, Latham is a research meteorologist for the danger, fire planning, fire effects, USDA Forest Service, Intermountain we enhanced the current fire-dan- Research Station, Fire Behavior Research and smoke management—and help ger rating system with spatial dis- Work Unit, Missoula, MT. resolve the confusion that exists on

14 Fire Management Notes plays. Fire danger and weather data Current Status transferred to the National Inter- are obtained from the Weather In- As part of the phased development agency Fire Center (NIFC) in formation Management System Boise, ID, for the 1997 fire season. and implementation, we made pro- (WIMS) and processed to produce totypes of several products and beta graphical displays based on current Colored maps are produced depict- tested them during the summer of NFDRS values. Phase 1 gives users 1995; they are now routinely avail- ing fire weather observations of the opportunity to view spatial dis- able. Broad-area weather and fire temperature, relative humidity, 24- plays and provides us with feedback hour precipitation totals, and danger maps are produced from the for developing future products in fire-weather-station network as part windspeed. Maps of fire danger subsequent phases. components include the 1,000- of phase 1. As a preliminary test of phase 2 and 3 concepts, high-reso- hour timelag fuel moisture, Phase 2. During phase 2, improve- lution fire danger maps are being Keetch-Byram Drought Index ments to NFDRS calculations will (Keetch and Byram 1968), and ad- produced using weather data from be implemented, but that system the Oklahoma Weather Mesonet. jective class (fig. 1). In most cases, will not be completely redesigned. we use a five-color scheme (green Satellite-based .6 mile- (1 km-) Because it is so important to com- resolution fuel and greenness data through yellow to red). pare the current fire danger with are also used. historical values, there will be a na- In addition to collecting the fire- tional reanalysis of historical data Broad Area Maps: National Map weather-network observations, using the Fire Information Re- each morning computers auto- Prototypes. WIMS is queried for trieval and Evaluation System each day’s fire weather observations matically collect data from all the (FIRES) (Andrews and Bradshaw in and NFDRS components for the North America radiosonde stations preparation) to calibrate the in- and compute a Lower Atmosphere primary fuel model for every re- dexes. The main products devel- porting station. Project scientists Stability Index (Haines 1988). oped during phase 2 will include These data are used to create na- have been creating national maps new dead fuel moisture models and on a workstation in the Fire Behav- tional and North American Haines new live fuel moisture models ior RWU in Missoula, MT. The maps Index maps (fig. 2). driven by Normalized Difference are based on a 6 mile (10 km) cell Vegetation Index (NDVI) data re- size using a weighted inverse dis- Twenty-four-hour lightning strike ceived via satellite (Burgan and tance-squared interpolation scheme data are obtained from Global Hartford 1993). Once phase 2 is fin- Atmospherics, Inc.,* a private between fire-weather-station loca- ished, WFAS will use the National tions. The entire process is to be Continued on page 16 Weather Service (NWS) as a primary source of weather data. Inclu- sion of observations from the fire weather network will be optional for broad-scale assessment.

Phase 3. During phase 3, we will use a gridded weather model to calculate fire potential over space and time. In lieu of wind observations or wind models, WFAS will use wind forecasts and estimated wind values. We will implement new fire spread, fire intensity, large fuel burnout, and crown fire models in the new system. In addition, we will use a national fuel map in the calculations. Eventually, WFAS displays will allow dynamic zooming to allow fire behavior assessments Figure 1—Example of a WFAS product—a national fire danger map; squares indicate reporting that are site specific. stations. The Weather Information Management System (WIMS) provides the observed fire danger classifications that are taken directly from levels defined for each station.

Volume 57 • No. 2 • 1997 15 company, and overlaid on either a lightning ignition probability map or a fire danger class map. The lightning ignition probability is calculated as a function of the 100- hour timelag fuel moisture and/or duff depth (Latham and Schlieter 1991).

A four-panel vegetation greenness map derived from data obtained by satellites portrays visual and relative greenness (Burgan and Hart- ford 1993) as well as departure from average greenness (Burgan et al. 1996) and experimental live shrub moisture (fig. 3). These maps have a .6 mile (1 km) spatial resolution and are updated weekly.

The broad area maps are distributed as Graphics Interchange For- mat (GIF) files over three networks. They are uploaded to the WFAS di- rectory in WIMS, mailed as a Data General dumpfile to the NIFC in Boise, ID, and the Northern Rockies Coordinating Center in Figure 2—Another WFAS product—a map of North America using the Haines Lower Missoula, MT, and posted on the Atmospheric Stability Index (LASI) (Haines 1988). Weather data are obtained from all USDA Forest Service’s Internet radiosonde stations. Home Page (www.fs.fed.us/land/ wfas/welcome.html). of the reporting stations have cor- WIMS use the burning index (BI), rectly cataloged annual precipita- and the rest use the energy re- As a result of our current experi- tion values, a requirement for lease component (ERC) to set ence with the maps, we make the correct computation of the KBDI. their adjective class. Groupings following evaluation: The station density for the KBDI for the five classifications on the map is generally very low, there- weather maps were based on our • Station Density and NWS fore the map is quite unreliable perception of reasonable thresh- Weather Streams. The higher on a national basis. This situation olds. Experience has made it station density in the Western could be resolved if fire managers clear that we imparted a western United States facilitates reason- would enter the average annual (arid) bias. “Critical” relative hu- able broad area maps of fire dan- precipitation into their WIMS midities in the Eastern States and ger and fire weather elements. weather station catalogs. Upper Midwest, for example, were The lower station density in the • Coloring the Maps. The daily fire not reflected correctly in the Central and Eastern States leads danger adjective classes are taken color code. More analysis (using to obviously bogus interpolated directly from the WIMS computa- FIRES) is required to delineate fields. This highlights the need to tion for each station, based on the regional threshold values for fire use NWS data in the broad area fire danger index defined by the weather elements. component of WFAS. station, number of staffing • Keetch-Byram Drought Index classes, and staffing level values High Resolution Maps: The Okla- (KBDI). Only about 10 percent cataloged for each station’s prior- homa Prototype. The high resolu- *The use of corporation names and/or their products is ity-one fuel model. About 90 per- tion NFDRS style fire danger map is for the information and convenience of the reader and should not be misconstrued as an official endorsement by cent of the stations cataloged in represented by the Oklahoma fire the U.S. Department of Agriculture or the Forest Service.

16 Fire Management Notes Figure 3—Satellite data is used to produce this WFAS product—a four-panel map that portrays visual and relative greenness (Burgan and Hartford 1993) as well as departures from average greenness and experimental live shrub moisture. Vegetative greenness maps, which are updated weekly, have a .6 mile (1 km) spatial resolution. danger system. This system is de- ellite data. Gen. Tech. Rep. INT-297. Ogden, Finney, Mark A. 1995. A fire area simulator scribed by Crawford (1993) and UT: U.S. Department of Agriculture, Forest for fire managers. In: The Biswell Sympo- Service, Intermountain Research Station. sium: Fire issues and solutions in urban Carlson et al. (1996). Also consult 13 p. interface and wildland ecosystems; 1994; Burgan and Bradshaw (1997) in Burgan, Robert E.; Hartford, Roberta A., Feb. 15-17; Walnut Creek, CA. Gen. Tech. this issue of Fire Management Eidenshink, Jeffery C. 1996. Using NDVI to Rep. PSW-158. Berkeley, CA: U.S. Depart- assess departure from average greenness ment of Agriculture, Forest Service, Pa- Notes. and its relation to fire business. Gen. Tech. cific Southwest Forest and Range Rep. INT-GTR-333. Ogden, UT: U.S. Depart- Experiment Station. Literature Cited ment of Agriculture, Forest Service, Inter- Haines, D.A. 1988. A Lower Atmosphere Se- mountain Research Station. 8 p. verity Index for wildland fire. National Andrews, Patricia L. 1986. BEHAVE: Fire be- Carlson, J.D.; Burgan, Robert E.; Engle, David Weather Digest. 13(2): 23-27. havior prediction and fuel modeling sys- M. 1996. Using the Oklahoma Mesonet in Keetch, John J.; Byram, George M. 1968. A tem—BURN subsystem, part 1. Gen. Tech. developing near-real-time, next-generation drought index for forest fire control. Res. Rep. INT-194. Ogden, UT: U.S. Department fire danger rating system. In: Proceedings Pap. SE-38. Asheville, NC: U.S. Depart- of Agriculture, Forest Service, Intermoun- of the 22nd Conference on Agricultural and ment of Agriculture, Forest Service, tain Research Station. 130 p. Forest Meteorology with Symposium on Southeastern Forest Experiment Station. Andrews, Patricia L.; Bradshaw, Larry S. In Fire and Forest Meteorology; 1996 Jan. 28- 32 p. preparation. FIRES: Fire Information Re- Feb. 2; Atlanta, GA. Boston, MA: American Latham, Don J.; Schlieter, Joyce A. 1991. Ig- trieval and Evaluation System—a program Meteorological Society: 249-252. nition probabilities of wildland fuels based for fire danger rating analysis. Gen. Tech. Crawford, Ken, ed. 1993. The Oklahoma on simulated lightning discharges. Res. Rep. INT. Ogden, UT: U.S. Department of Mesonet. Norman, OK: Oklahoma State Pap. INT-411. Ogden, UT: U.S. Department Agriculture, Forest Service, Intermountain University and the University of Oklahoma. of Agriculture, Forest Service, Intermoun- Research Station. 57(11): 1453-1463. tain Research Station. 16 p. Burgan, Robert E.; Bradshaw, Larry S. 1997. Deeming, J.E., Burgan, R.E., and Cohen, J.D. Rothermel, Richard C.; Andrews, Patricia L. WFAS requires a variety of weather infor- 1977. The National Fire-Danger Rating Sys- 1987. Fire behavior system for the full mation. Fire Management Notes. 57(2): tem—1978. Gen. Tech. Rep. INT-39. Ogden, range of fire management needs. Paper 18-21. UT: U.S. Department of Agriculture, Forest presented at the Symposium on Wildland Burgan, Robert E.; Hartford, Roberta A. 1993. Service, Intermountain Research Station. Fire 2000, April 27-30, 1987. South Lake Monitoring vegetation greenness with sat- 63 p. Tahoe, CA. 145-151. ■

Volume 57 • No. 2 • 1997 17 WFAS REQUIRES A VARIETY OF WEATHER INFORMATION*

Robert E. Burgan and Larry S. Bradshaw

s the Fire Behavior Research Weather Data Work Unit (RWU) of the Inter- “The single largest The WFAS is being designed to A mountain Research Station challenge to building provide maximum information to has been developing the Wildland and operating the the user with minimum interfer- Fire Assessment System (WFAS) ence to daily work schedules in (see Burgan et al. 1997 in this is- Wildland Fire Assessment System field offices. We expect that the Na- sue of Fire Management Notes), it tional Weather System’s (NWS) has been abundantly clear that (WFAS) is collecting, weather stations will provide a weather inputs are the most diffi- processing, and stable “minimum set” of weather cult challenge in the system’s stations for broad scale (national) development. In this article, there- archiving weather data.” fire potential assessment, but the fore, we discuss required weather WFAS will also use weather data inputs to the WFAS in terms of collected by various agencies (e.g., weather parameters and temporal USDA Forest Service, USDI Bureau and spatial resolution. We also routinely—at least once a day—for broad areas. Fire behavior assess- of Land Management and National describe briefly a system that pro- Park Service, States) as much as duces fire danger maps four times ments, however, are made for specific sites as needed. possible and include local area re- daily at a .6 mile (1 km) resolution finement of fire assessments. For for the State of Oklahoma and With the attributes already in place broad scale assessments, we hope present it as an early step in the to make use of new technology development of the WFAS. as well as those planned for the WFAS, the system will: available from the NWS modern- ization project—technology that is WFAS Definitions • Require minimal data from user, not likely to be available from and Attributes • Emphasize map and graphical agency-operated units. Thus two Fire potential in WFAS is used in a outputs, major challenges for national level generic sense to refer to assess- • Provide multilevel resolution in- fire danger assessment are 1) how ments (at whatever time and spa- puts and outputs, to make the best use of new NWS tial scale) that fit the context of the • Reflect diurnal weather varia- products and 2) how best to use following discussion. Fire danger tions, varying agency data as a supple- and fire behavior are terms used to • Provide “hooks” to related mod- ment to NWS data. describe assessments of fire poten- els (e.g., smoke dispersion and tial on different time and space fire effects), As the area of interest focuses from scales. Fire danger is evaluated • Be oriented toward workstations national to site specific, we expect and PC’s, that Remote Automated Weather • Simulate fire characteristics, Stations’ (RAWS) data or other • Be operable in the field, Robert Burgan is a research forester and agency-collected weather data will Larry Bradshaw is a meteorologist for the • Relate current to historical con- be used to assess specific fire USDA Forest Service, Intermountain ditions, events. Research Station, Fire Behavior Research • Forecast fire potential and prob- Work Unit, Missoula, MT. abilities on national to local Capability for *This article appeared, in part, as “Weather Needs for scales, and Regular Updates the Wildland Fire Assessment System” in Proceedings • Display inputs, outputs, and in- of the 22nd Conference on Agriculture and Forest Fire danger has traditionally fo- Meteorology with Symposium on Fire and Forest termediate products. Meteorology, published in 1996 in Boston, MA, by the cused on the “worst case” scenario American Meteorological Society, pages 325-328.

18 Fire Management Notes by using 1 or 2 p.m. (local time) Forecasting Fire determine which inputs and out- weather observations. Automated Potential puts the WFAS will display. weather observations and satellite Although fire potential estimates data provide an opportunity to as- will be updated during the day, fire Weather Summary sess fire danger throughout a day. weather forecasts must be made The single largest challenge to Improved temporal resolution of available for as far in the future as building and operating the WFAS is weather data would provide fire they can be reasonably relied upon. collecting, processing, and managers with more timely infor- This is expected to be a maximum archiving weather data. The follow- mation about fire potential, per- of 3 to 5 days. Some measure of ing describes our current assess- haps alerting them to problems forecast reliability will be neces- ment of weather needs, subject to such as severe conditions develop- sary. The specific measure and review and change as development ing late in the day or poor mois- means of producing and displaying continues. Required weather pa- ture recovery at night. it will have to be discussed with rameters for broad area fire poten- the NWS. tial assessments are 1) temper- Fire behavior updates could make ature, relative humidity, solar ra- use of weather data updates every 4 National scale products will have diation, precipitation amount and to 6 hours, but spot weather fore- to rely upon nationally available possibly duration, windspeed, and casts generated “on-demand” for a weather products. We do not ex- wind direction—all near ground fire area will still be required. Fire pect that personal interpretation level, 2) moisture, temperature, researchers are working with the by fire weather forecasters will oc- and stability of the lower atmo- NWS to conduct weather modeling cur at this scale. As the forecast sphere, and 3) cloudiness, lightning tests to determine both if gridded area decreases and one begins to activity, and location of weather data can be refined enough to be look at site- and time-specific situ- fronts. With the possible exception useful for areas as small as fire ations, it is less likely a weather of windspeed and wind direction, sites and how fire weather forecast- station will exist within the area of these inputs should be obtained ers will interface with WFAS. interest. If the forecast area is four times per day to help track di- small enough, there will not be any urnal fire danger changes and pro- Archiving Data weather observations taken within vide input to fuel moisture models. Traditional fire danger systems it unless special provisions are These data should be available on at have long had some capability to made. Therefore forecast weather least a 19-mile (30-km) grid for the relate current fire potential condi- parameters for fire behavior pre- Continental United States, with a 6- tions to historical conditions. Dis- dictions will most likely have to mile (10-km) grid being much pre- playing historical data infers that come from spot forecasts. ferred. These weather parameters major decisions will have to be must apply to the near-surface zone made concerning what weather Data Displays within which fires burn and be ad- data to archive and how to both justed for slope, elevation, and as- Although it is natural to focus on archive and make that data avail- pect. We hope that the NWS will be the final fire potential outputs, it is able to users. It does not seem pru- able to make these adjustments. likely fire managers will want to dent to rely on the NWS to archive see some of the raw inputs and in- weather data for the WFAS because Windspeed and wind direction are termediate products as well. For 1) it would likely have to be pur- very difficult to project realistically example, the NWS should provide chased from them, 2) it would not to a high resolution over large geo- satellite images and graphical be “on line,” 3) it probably would graphic areas, but it would be very products showing weather condi- have to be reformatted before it helpful to obtain this information tions such as cloudiness, the loca- was usable, and 4) it would not be in a relatively broad spatial and tion of fronts, temperature maps, under the control of WFAS manag- temporal context at the least. These and wind graphics. Fire managers ers. Therefore, a procedure must two inputs need to be defined quite will also want to view such graph- be devised to enable WFAS to specifically for fire behavior appli- ics as fuel moisture or drought “grab” data to archive for later cations to be site and time specific. maps, fire potential maps, and computing of current fire danger. Depending on the capabilities of lightning location maps generated mesoscale wind models, it may be by WFAS. Discussions with the NWS and fire managers will help Continued on page 20

Volume 57 • No. 2 • 1997 19 necessary to rely on forecaster or High Resolution Outputs. High county. Figure 1 combines weather user inputs of windspeed and wind resolution outputs require site- stations and fuel models for that direction for local areas. and time-specific assessments of State. Because the topography in fire behavior characteristics such Oklahoma is flat, the entire State For the time of highest fire danger, as flame length, spread rate, and can be included in a single class (0 which varies regionally, 24-hour fire intensity. Such assessments to 20 percent slope). forecast data of fire danger should would benefit greatly from 6- be provided at least once daily, and hourly updates of forecast weather The process of calculating fire dan- 3- to 5-day forecasts should also be conditions at spatial scales of .6 ger maps is done on a pixel-by- updated daily for the Continental mile to 33 yards (1 km to 30 m). pixel basis. First, weather data United States. All weather data Fuels and terrain data should be from the mesonet stations are used should be adjusted for the effects presented in a similar spatial scale. to calculate all the dead fuel mois- of mountainous terrain and/or ma- Because it is unlikely that gridded tures (1, 10, 100, and 1,000 hour) rine influences. weather data can be obtained at associated with each weather sta- these scales, provisions will have to tion. The dead moistures calcu- Outputs be made for direct weather inputs lated from composites for a Low Resolution Outputs. WFAS by fire weather forecasters and weather station are assigned to all products that are user accessible temporary RAWS. The fuel, pixels within the county in which will come in the form of: weather, and terrain data will be the weather station is located. processed in a fire simulation sys- These moistures are saved in a • Primary maps, the lowest reso- tem to model fire behavior and lo- separate data file. lution information displays. The cation of problem fires over the maps will provide the most pro- next 24 hours. Planned outputs in- Live herbaceous and shrub mois- cessed “overview” data. A user clude maps of projected fire loca- tures are calculated for each pixel should be able to click a button tion and intensity, plotted either as a function of the relative green- and view one of these maps. Ex- planimetrically or over terrain. ness index, which is derived from amples include lightning occur- the Normalized Difference Vegeta- rence and energy release Calculating Fire tion Index, which is in turn calcu- component maps. Danger Maps for lated from weekly composites of • Intermediate product maps that Oklahoma Advanced Very High Resolution display underlying information Radiometer Data (Burgan and The primary inputs to a math- are being considered. Examples Hartford 1993). For the fire danger ematical model for calculating fire of such maps are a cloudiness calculations in figure 2, three in- danger are fuels, weather, and map, Haines Lower Atmospheric put map layers are referenced: 1) a topography. See figure 1 for a .6 Stability Index (Haines 1988), map that identifies the county each mile- (1 km-) resolution fire fuels windspeed, wind direction, pixel is in, 2) a Relative Greenness map for Oklahoma. Because Okla- weather fronts, and vegetation map for calculating live fuel mois- homa is relatively flat, elevation greenness. tures, and 3) a fuel model map data are not used for their fire dan- • Non-map graphics that display consisting of a fuel model number ger calculations. However, eleva- information relating to the con- for each pixel. For each pixel, the tion data are available for the ditions of specific points or ar- county map layer is used to deter- conterminous United States at a .6 eas, such as weather station mine the weather station repre- mile (1 km) resolution. A high locations, ranger districts, or na- senting that pixel (to access dead resolution weather network is rare, tional forests. Examples are the fuel moistures); the fuels map but the University of Oklahoma fire characteristics chart and layer is used to identify the fuel (Crawford 1993) is operating one seasonal fire potential graphs. model for that pixel; and the Rela- for Oklahoma. Our RWU is cooper- • Alphanumerics, which are the tive Greenness map layer is used to ating with the university to use actual numbers that underlie the calculate live fuel moistures by weather observations from their various map and graphic outputs pixel. The rest of the calculations mesonet to calculate fire danger (e.g., energy release component, proceed as in the current fire-dan- for the State. One weather station burning index, greenness) at any ger rating system. However, rather in each county has been selected to location for any of the primary than the outputs being a table of or intermediate maps. represent the weather for that

20 Fire Management Notes fire danger indexes printed on paper for individual weather stations, they are multicolored maps of fire danger (spread component) at .6 mile (1 km) resolution for the State (fig. 3). An interesting aspect of the final fire danger map is that it shows even gradation of fire danger across county boundaries, in spite of the fact that all the pixels in each county have the same dead fuel moisture values. This occurs because of the .6 mile (1 km) spatial Figure 1—NFDRS fire fuels map and partial Oklahoma Mesonet weather station network. variability of the fuel model map Crosses indicate each county’s weather station. and the live herbaceous and shrub moisture maps. Such maps can be produced at hourly intervals. Fire Potential Map Calculation While this process works for Okla- County map Relative Fuel (Wx) Greenness map model map homa, extending the process to mountainous terrain will require additional research. Methods are Dead fuel Live fuel Pixel fuel required to adjust the weather in- moisture moisture model puts for the effects of slope, elevation, and aspect.

Fire potential The Oklahoma effort shows that calculation calculation of .6 mile- (1 km-) resolution fire danger maps is feasible, Spread Energy release even without weather data at that Burning index component component resolution. The next step is to use NWS gridded data to produce a similar map for the conterminous Figure 2—Flowchart showing how weather, greenness, and fuel model maps are used to United States. calculate fire potential, which is displayed as spread component, energy release component, and burning index maps. Literature Cited Burgan, Robert E.; Hartford, Roberta A. 1993. Monitoring vegetation greenness with satellite data. Gen. Tech. Rep. INT- 297. Ogden, UT: U.S. Department of Ag- riculture, Forest Service, Intermountain Research Station. 13 p. Burgan, Robert E.; Andrews, Patricia L.; Bradshaw, Larry S.; Chase, Carolyn H.; Hartford, Roberta A.; Latham, Don J. 1997. Current Status of the Wildland Fire Assessment System (WFAS). Fire Management Notes 57(2): 14-17. Crawford, Ken, ed. 1993. The Oklahoma Mesonet. Norman, OK: Oklahoma State University and the University of Okla- homa. 57(11): 1453-1463. Figure 3—Fire potential (danger) map for Oklahoma using the spread component. Haines, D.A. 1988. A Lower Atmosphere Se- Note the even gradation of fire danger across county boundaries. verity Index for wildland fire. National Weather Digest. 13(2): 23-27. ■

Volume 57 • No. 2 • 1997 21 HIGH RESOLUTION FIRE WEATHER MODELS

Francis M. Fujioka

he next generation fire danger the near-surface atmospheric con- and fire behavior system envi- “A computerized fire ditions that influence the fire envi- T sioned by the Forest Service weather model coupled ronment. The model output should (see Burgan et al. 1997 and Burgan with a synoptic model is be in a coordinate system and Bradshaw 1997 in this issue of coregistered with the fuels and ter- Fire Management Notes) will as- a powerful means of rain data so that fire behavior cal- sess fire potential across a broad describing the weather culations can be obtained over the spectrum of space and time. At the part of the fire area of interest. A mesoscale model incident level, fire behavior models environment.” based on the physics of atmo- will require detailed descriptions of spheric processes ensures that the the spatial and temporal variations spatial distribution and transfor- of weather in the area of the fire. values of wind direction, wind- mation of fire weather variables are Burgan and Bradshaw (1997) sug- speed, relative humidity, tem- generated methodically and consis- gest that high resolution fire be- perature, solar radiation, and pre- tently. The model should also in- havior simulations “. . . would cipitation amount. All of these can corporate large-scale effects of benefit greatly from 6-hourly up- be simulated in a fire weather synoptic systems that are present, dates of forecast weather condi- model, some more accurately than because generally, weather conditions at spatial scales of .6 mile to others. In meteorology, a fire tions on a local scale are driven by 33 yards (1 km to 30 meters).” By weather model is described as me- large-scale weather. A computer- spatial scale, they imply the dis- soscale, in reference to the par- ized fire weather model coupled tance between the points of a rec- ticular spatial scale of the atmo- with a synoptic model is a powerful tangular grid on which weather, spheric process it simulates (for means of describing the weather fuel, and terrain conditions are example, see Orlanski 1975). part of the fire environment. available for fire behavior simula- Mesoscale models are typically tions. They write: “. . it is unlikely used to simulate convective sys- Recently, three research groups that gridded weather data can be tems (e.g., storm events) and local conducted independent fire obtained at these scales,” but this circulations (e.g., sea breeze, winds weather modeling experiments paper is cautiously optimistic that over complex terrain). They may be that portend the future of fire we are closer than expected to high driven by synoptic scale weather weather analysis and forecasting. resolution fire weather models. models that describe large atmo- The first compared a fire weather spheric circulations up to the plan- model with corresponding data What Is a Fire etary scale. Synoptic scale models from a high density weather sta- Weather Model? provide the national weather maps tion network. The second experi- For the purposes of this article, a seen in newspapers and on televi- ment simulated weather on Storm fire weather model is a scientific, sion. Mesoscale modeling applica- King Mountain during the tragic computerized method of describ- tions are not so routine, but they 1994 fire. The third experiment ing spatial and temporal variations have recently included the analysis coupled a weather model and a fire of weather. The primary weather of fire weather. behavior model so that the fire variables of interest for fire behav- influenced the weather and vice ior modeling are the near-surface The Need for a Fire versa. Weather Model Francis Fujioka is the project leader for The Maui Vortex A fire weather model is needed to Fire Meteorology Research, USDA Forest Wildfires in Hawaii are smaller on Service, Pacific Southwest Station, Forest provide a scientific, comprehensive Fire Laboratory, Riverside, CA. spatial and temporal description of average than those in the other

22 Fire Management Notes Western States but are of no less concern to local fire managers. Fire weather conditions in Hawaii are unique compared to weather in the other States. On the island of Maui, a local circulation feature called the Maui vortex, or Maui eddy, can change the wind direction by 180 degrees within a few miles, significantly affecting the behavior of potential fires. The vortex is also known to trap ash and smoke issuing from pre-harvest burning of sugar cane in the central valley of Maui (Schroeder 1993), thus degrading air quality.

We used a mesoscale model developed over a period of years at Colo- Figure 1—Map showing the high resolution grid used in the Maui fire weather simulations (Ueyoshi et al. 1995). The grid points located at the origins of the wind rado State University to simulate vectors are 1.2 miles (2 km) apart. the weather conditions over Maui in July 1988 (Ueyoshi et al. 1995). The plantation uses weather infor- An intriguing feature of the fire Called RAMS (Regional Atmo- mation to determine when the weather simulations was the be- spheric Modeling System), the sugar cane can be burned without havior of temperature and relative model incorporates the physical significant air quality deterioration humidity in the lee of the ridge laws governing atmospheric pro- in sensitive areas. In July 1988, the line extending northward from the cesses. In this study, we calcu- weather stations lay in that part of summit of Mt. Haleakala, the large lated—at 10-second intervals— the vortex where average wind- shield volcano that dominates winds, temperature, and relative speeds apparently were greatest. Maui’s topography. The mesoscale humidity at points spaced 1.2 model showed that, as the trade miles (2 km) apart in a rectangular We found that the mesoscale winds descended from the ridge grid measuring 67 miles by 55 model frequently overestimated and turned south from their west- miles (112 km by 92 km) (fig. 1). windspeed by as much as 7 mi/h (3 ward trek, the layer of air immedi- The short time step is necessary to m/sec). The model errors were ately above the surface layer preserve the numerical stability of larger at night and early morning became warm and dry, apparently computerized methods used to than they were during the day. The due to the spreading and sinking of solve the model’s underlying equa- model did not reproduce the diur- air as the flow accelerated south- tions. Vertical variations are also nal range in windspeed that actu- ward on the western side of the captured at 23 levels of a terrain- ally occurred, but it seemed to be vortex. The combination of warm- following coordinate system. The in phase with the observed ing, drying, and accelerating air model can thus simulate impor- windspeed oscillation, except near flow can invigorate a fire in this tant fire weather effects such as the center of the vortex. There, the area, if this layer descends to the inversions, precipitation, and peak in the average observed level of the fuels. momentum transfer between windspeed lagged behind the aver- layers of the atmosphere. Model age model windspeed by approxi- The mesoscale model simulations simulations, however, should be mately 4 hours. Consequently, the provided details that we have not checked against the real world model underestimated the average seen before, but the spacing hori- conditions they represent when- windspeed during most of the af- zontally (1.2 miles or 2 km) and ever possible. We compared our ternoon hours. In all cases, the vertically (260 feet or 80 m in the simulations with weather data model wind direction was in fairly lowest layer) is probably still too from remote automatic weather good agreement with the observed coarse. The Maui model also needs stations deployed by a sugar plan- wind direction. tation in the central valley of Maui. Continued on page 24

Volume 57 • No. 2 • 1997 23 a better specification of the surface roughness, vegetation, and soil types. Storm King Mountain Simulation Fire weather models can be used to study historic fires. Scientists at the Los Alamos National Labora- tory applied the RAMS mesoscale model to simulate weather conditions that occurred on Storm King Mountain on the afternoon of July 6, 1994. They used a grid with points spaced 165 feet (50 m) apart in preparation for a subsequent fire behavior simulation (fig. 2). At this grid resolution, the model’s computing requirements are excessive for all but the most powerful ma- chines. At the initial stages of the Figure 2—Simulated wind field in the South Canyon Fire area on the afternoon of Los Alamos simulations, 1 hour of July 6, 1994 (from Bossert 1996). Distance between the bases of adjacent vectors is simulated weather required 35 165 feet (50 m). Vector length is proportional to windspeed. hours of computing time on a relatively high-speed Unix workstation. mospheric Research have coupled shows an example of two fingers Since then, a version of the model a mesoscale weather model and a developing along an initially code has been developed for paral- simple fire behavior model for a straight fire line and the attendant lel processors—computers with dry eucalyptus forest (Clark et al. wind pattern at a height of 363 feet multiple processing units, each 1996). This project started only re- (110 m). In this case, the wind working on a portion of the model cently, but some interesting results ahead of the fire is drawn between calculations. This has brought the have already been obtained. The re- the fingers, where it opposes the computing time for a 1-hour simu- searchers have, for example, iden- direction of the propagating fire. A lation down from 35 hours to 1.5 tified a mechanism that produces critical measure of atmosphere-fire hours and is likely to shrink the ra- parabolic fire fronts. In simula- interaction is the ratio of the ki- tio even further (Bossert 1996). tions of an initially straight fire netic energy of the air flow over The improvement in computing front subjected to a certain range the fire to the buoyant energy of performance brings this mesoscale of light winds, the center of the the heat of the fire. When this ratio modeling code into the range of front accelerates preferentially, be- is small, conditions are enhanced feasibility for fire weather predic- cause the convection column infor a blowup fire. In this case, tion. duced by the fire tilts ahead of the winds are relatively light over the front and draws more air from the fire, while the fire itself is ener- The Interaction middle of the fire line than from getic. A somewhat counterintuitive Between Weather the ends; hence the windspeed and finding is that strong winds (high and Fire fire spread are faster across the kinetic energy to buoyant energy center. As a result, the fire line ratio) stabilize the fire front dy- A significant recent development takes on the shape of a parabola. namics, even as they increase the in fire modeling is the coupled at- Under certain conditions that are spread rate. Apparently, when mosphere-fire model, which simu- still not fully understood, the ac- windspeeds exceed a critical value, lates the feedback between the celerations of the fire front are the convective column is displaced atmosphere and fire. This research rapid and highly localized; the sci- too far from the fire front to pro- is necessary to understand such entists call this event “dynamic fin- duce any significant interaction conditions as blowup fires. Scien- gering.” The simulation in figure 3 with the fire. tists at the National Center for At-

24 Fire Management Notes proximately. Fire scientists must describe model uncertainties as de- finitively as possible, and fire prac- titioners must understand the limitations of models. We cannot have one without the other, if we are to apply fire weather and fire behavior modeling successfully in fire management. Literature Cited Bossert, James E. 1996. Unpublished communication on file at the office of the author, USDA Forest Service, Forest Fire Laboratory, 4955 Canyon Crest Drive, Riverside, CA 92507-6099. Burgan, Robert E.; Bradshaw, Larry S. 1997. WFAS requires a variety of weather information. Fire Management Notes. 57(2): 18-21. Burgan, Robert E.; Andrews, Patricia L.; Bradshaw, Larry S.; Chase, Carolyn H.; Hartford, Roberta A.; Latham, Don J. 1997. Current Status of the Wildland Fire Assessment System (WFAS). Fire Management Notes 57(2): 14-17. Clark, Terry L.; Jenkins, Mary Ann; Coen, Janice L.; Packham, David R. 1996. A coupled atmosphere-fire model: Convec- tive feedback on fire-line dynamics. Journal of Applied Meteorology. 35: 875- 901. Orlanski, Isidoro. 1975. A rational subdivi- sion of scales for atmospheric processes. Bulletin of the American Meteorological Society. 56(5): 527-530. Schroeder, T. 1993. Climate controls. In: Figure 3—Simulation from the coupled atmosphere-fire model of Clark et al. (1996). The Sanderson, M., ed. Prevailing trade fire starts along a straight line at the left and develops two dynamic fingers along the 0.4 winds: Weather and climate in Hawaii. km (0.24 mi) and 1.3 km (0.78 mi) marks of the y-axis. The gray scale at the right Honolulu, HI: University of Hawaii indicates vertical velocity of air flow. Press: 12-36. Ueyoshi, Kyozo; Roads, John O.; Fujioka, Francis M.; Stevens, Duane E. 1995 (ac- At this stage, the coupled model is blowups and spotting. All of the cepted). Numerical simulation of the less predictive than it is indicative models described above, complex Maui Vortex in the trade winds. Journal of the complex interactions that as they are, nevertheless are sim- of the Meteorological Society of Japan. 26 p. ■ play out between a fire and the at- plifications of reality. At best, mosphere. It promises significant model simulations can be expected insight into the processes creating to resemble the real world only ap-

Volume 57 • No. 2 • 1997 25 MAKING SENSE OF FIRE WEATHER

Brian E. Potter

district ranger checks the lo- that indicates whether or not the cal weather conditions and When fire managers fire and typical values are signifi- A notes high windspeeds. In an- know which weather- cantly different, to compare the other district, the ranger sees low related variables following six measurements with relative humidity is predicted for “typical” values for the same sta- her area. A third ranger looks at discriminate “large-fire” tion and season: the nearest upper-air report and weather from typical • Surface air temperature, finds strong wind shear and an un- weather conditions, • Relative humidity, stable temperature profile. Which they can prepare for • Dewpoint depression, ranger or rangers should consider • Stability, their observations as indicative of the possibility of a large • Windspeed, and high weather-related fire risk? wildland fire developing. • Wind shear. The general public and fire managers alike commonly consider a va- 1,000 acres (400 ha) or more. I as- Note that dewpoint depression is riety of weather-related variables to sociated each fire with the nearest the difference between actual air be indicators of fire risk. These upper-air weather station and clas- temperature and dewpoint tem- include surface air temperature, sified it according to its season perature. It is not the same as windspeed, and humidity. Fire (spring, summer, autumn, or win- wetbulb depression, which appears managers and foresters may hear ter). I dropped any station and sea- on the psychrometric charts in or believe that wind shear and sta- son with fewer than five associated many belt weather kits. Dewpoint bility also contribute to fire risk. fires from the analysis. Figure 1 depression is always greater than While researchers repeatedly note shows the number of fires for each wetbulb depression. that various conditions precede or station that remained. accompany large wildland fires, Also note both the size of the wild- they have not examined whether For each fire day, I used statistical land fires considered here and the these same conditions are equally analysis of variance, a technique location of the weather observa- common on days when no fires occur or when only small or con- trollable fires occur. When fire managers know which weather- related variables discriminate typical weather conditions from weather associated with large fires, they can prepare for the possibility of a large wildland fire developing. Methods Used I analyzed data for 339 large wildland fires that occurred in the Con- tinental United States from 1971 through 1984. Each fire burned

Brian Potter is a research meteorologist for the USDA Forest Service, North Figure 1—Locations of stations used for this study and the number of fires over Central Forest Experiment Station, East 1,000 acres (400 ha) associated with each station. Three-letter codes are weather Lansing, MI. station identifiers.

26 Fire Management Notes tions. The fires are all large fires— Table 1—Dewpoint Depression Table (in both Fahrenheit and Celsius), which can be the more numerous small fires may photocopied, laminated, and kept in belt weather kits. At any altitude, users can quickly convert relative humidity and temperature to dewpoint depression by consulting the show stronger or weaker correla- applicable version of the Dewpoint Depression Table. tions with some of the weather variables. The observations indicate the Dewpoint Depression Table (Fahrenheit) weather conditions on a horizontal Temperature Relative Humidity (%) scale of about 62 miles (100 km). A (°F) 10 20 30 40 50 60 70 80 fire, once started, begins to alter 110 69 50 38 30 23 17 12 8 some of these conditions (such as 105 68 49 38 29 22 17 12 7 wind), and onsite measurements of 100 66 48 37 29 22 16 12 7 any weather variable may differ 95 65 47 36 28 22 16 11 7 from measurements taken several 90 64 46 36 28 21 16 11 7 miles away from the fire. 85 63 46 35 27 21 15 11 7 80 61 45 34 26 20 15 11 7 Results Found 75 60 44 34 26 20 15 10 7 The results of my analyses show 70 59 43 33 25 19 15 10 6 that surface temperature and 65 58 42 32 25 19 14 10 6 60 57 41 31 24 19 14 10 6 dewpoint depression are substan- tially higher on fire days than they are on a typical day for a given loca- Dewpoint Depression Table (Celsius) tion and season. Relative humidity Temperature Relative Humidity (%) is considerably lower on fire days. (°C) Windspeed, wind shear, and stability 10 20 30 40 50 60 70 80 did not show any significant differ- 44 38 28 21 17 13 10 7 4 ences between fire and nonfire days. 40 37 27 21 16 12 9 7 4 In other words, it is the more 36 36 26 20 16 12 9 6 4 straightforward and commonly 32 35 26 20 15 12 9 6 4 known atmospheric properties— 28 34 25 19 15 11 8 6 4 high temperatures and low air 24 33 24 19 14 11 8 6 4 20 32 24 18 14 11 8 6 4 moisture contents—that most sig- 16 32 23 18 14 10 8 5 3 nificantly contribute to weather-re- 12 31 22 17 13 10 8 5 3 lated fire risk. Thus, the answer to 830221613107 5 3 the question in the first paragraph 4292116129753 is clear: The second ranger, who saw the forecast of low relative humidity, should be particularly alert to was significant were all west of the weather reports. As experienced the possibility that the weather will Mississippi River. fire managers know, even without contribute to the development of a any measurements or weather re- large wildland fire. While they are not a substitute for ports, they can use their own first-hand experience, these find- “senses” to get a good idea what In examining the six weather vari- ings can be helpful to fire manag- the temperature, humidity, and ables at individual stations, I found ers and district rangers. They windspeed are. that neither fire-day stability nor suggest that the conditions that re- fire-day wind shear showed any sig- ally matter are those at the sur- With weather measurements, how- nificant difference from typical val- face—stability and wind shear may ever, fire managers can use the ap- ues at any of the 20 stations tested. not be very good fire-weather indi- plicable Table 1 to quickly convert Windspeed proved significant at just cators. Regardless of what part of relative humidity and temperature two of the stations, relative humid- the country you are in, dewpoint to dewpoint depression. This ity at four, surface temperature at depression is a robust indicator of Dewpoint Depression Table (shown seven, and dewpoint depression at the weather-related fire risk. Rela- both in Fahrenheit and Celsius) ten (or half) of the stations. The sta- tive humidity, which also measures can be photocopied, laminated, tions where surface temperature the moisture in the air, is almost as and kept in belt weather kits for good and more readily available in use at any altitude. ■

Volume 57 • No. 2 • 1997 27 SODAR AND DECISIONMAKING DURING THE FORK FIRE

Fred Svetz and Alexander N. Barnett

hat does an Incident Meteo- rologist (IMET) do when On the Fork Fire, a SODAR unit assisted personnel W anticipating a frontal pas- with forecasting and decisionmaking. During a sage likely to create strong winds weather-driven firestorm that consumed 40,000 and erratic fire behavior during a wildland fire? At the Fork Fire on acres (16,200 ha) in 6 hours, there were no the Mendocino National Forest in burnovers, no equipment loses, and no injuries August 1996, the IMET requested attributed to fire behavior. an acoustic wind profiler or SODAR (SOund Detection And Ranging) unit. He felt that a SODAR’s capability to measure low-level winds continuously would help him monitor expected environmental changes and improve his forecasts. What Is SODAR? The SODAR is a remote-sensing instrument that uses sound to measure windspeed, wind direction, and atmospheric stability up to 2,000 to 3,200 feet (750 to 1,050 m) above the ground at 100 feet (30 m) intervals. It can provide averaged wind Figure 1.—The 8-foot- (2.4-m-) tall, foam- and lead-lined wooden enclosures that data every 5 minutes, every hour, or surround SODAR’s antennas and drivers mounted on the trailer that transports them to the field. The enclosures dampen SODAR’s sound pulses and reduce the impact of at other intervals in between. background noises to increase the sensitivity of its measurements. Photo: Ken Underwood, AeroVironment Inc., Monrovia, CA, 1996. While other sizes of wind profilers with varying operating or output • A controller unit that houses SODAR Set Up frequencies are available, the control electronics, a power am- The SODAR unit was deployed in SODAR unit used on the Fork Fire plifier, and a power supply as- the extreme northwestern corner was an AeroVironment Inc. Model sembly. of the fire zone, just below the 2000 (AV2000),* consisting of • A 486 IBM-compatible computer ridge defined in figure 3. The site to run the interface software that • Three parabolic dish antennas was 3,200 feet (1,050 m) above sea logs and processes data and pro- and acoustic drivers mounted on level in a picnic area on a hillside vides the user with both tables a 30-foot (10 m) trailer (fig. 1). overlooking the fire zone. This site and graphs of wind and atmo- was selected because spheric data (fig. 2). • Background noise levels were Fred Svetz is a fire weather forecaster and low (below 50 decibels). incident meteorologist for the National Weather Service, Redding, CA, and Alex • There were no physical obstruc- Barnett is a certified consulting meteo- *The use of corporation names and/or their products is tions in the SODAR beam paths. rologist and senior project manager, for the information and convenience of the reader and should not be misconstrued as an official endorsement • The elevation was adequate to AeroVironment Environmental Services by the U.S. Department of Agriculture or the Forest Inc., Monrovia, CA. Service. ensure that beam coverage

28 Fire Management Notes verbal windspeed and wind-direction information onto data transfer forms to interpret and use later. As a backup, the SODAR operator also had a command radio, but he used it only when the SODAR noted changes in wind patterns or trends between the hourly cellular con- tacts or when cellular communications in the remote location were not available. It should be noted that future technological improvements could include the use of radio or satellite phone communications to allow the SODAR’s computer to be located where the IMET and Fire Behavior Analysts (FBA’s) are preparing their Figure 2.—Data displayed by a SODAR unit’s computer. This graphic shows wind forecasts. direction and windspeed profiles as displayed on the SODAR computer screen. Each vertical column of wind barbs represents 10-minute averages of the windspeed and wind direction versus height. Windspeed is indicated by the wind barb feathers. Wind direction The IMET used the data to develop is indicated by the orientation of the wind barbs. The colors of the wind barbs are an a “blueprint” of wind patterns and indication of atmospheric stability. Photo: Ken Underwood, AeroVironment Inc., stable layer development and dissi- Monrovia, CA, 1994. pation occurring during a normal would be representative of the operator called the IMET and read daily cycle in the fire zone. The low-level wind flow pattern him the windspeed and wind-direc- IMET found that weather observa- above the fire. tion data in miles per hour and de- tions from the fire crew showed ex- • It was upwind of the fire zone to grees for the 10-minute interval cellent agreement with the SODAR help safeguard the operator and occurring just previous to tele- data. This confidence in the to detect changes before they phoning. The IMET transcribed the Continued on page 30 reached the fire zone, thus mini- mizing burnovers. • It contained a road adequate for maneuvering a 30 foot (9 m) trailer. SODAR in Operation Because the IMET wanted the SODAR to collect data for 24 to 48 hours prior to the arrival of the weather front, SODAR began its measurements at 0900 P.d.t. on August 16. Deploying the SODAR unit well before the arrival of the front meant that the IMET could easily compare baseline weather conditions with changes brought about by the frontal passage.

The SODAR operator and the IMET communicated primarily by cellu- Figure 3.—Map showing the location of the SODAR unit during the Fork Fire on the lar phone. Each hour, the SODAR Mendocino National Forest in California on August 18, 1996.

Volume 57 • No. 2 • 1997 29 SODAR prompted the IMET to use the observations in the preparation of twice-daily weather forecasts and encouraged the FBA’s to use the data in their forecasts as well. SODAR’s Success When the Incident Command (IC) staff determined that weather and fire behavior forecasts continued to correlate well with actual conditions in the fire zone, their confidence in the IMET’s ability increased. On August 18, the day the front was expected, the IMET’s forecast was: Morning winds south to west, 4 to 8 mph (6 to 12 kph) with gusts to 14 mph (23 kph), changing in the afternoon to west to northwest winds, 10 to 20 mph A view of the SODAR model being used 3,200 feet (1,050 m) above sea level in a picnic area on a hillside overlooking the Fork Fire in August 1996. Photo: Alex Barnett, (16 to 32 kph) with gusts to 30 AeroVironment Environmental Services Inc., Monrovia, CA, 1996. mph (48 kph). The confidence of the IC staff in the weather and fire crews in the expected path of the fluenced fire behavior on the Fork behavior forecasts weighed heavily fire. They held back all fire crews Fire. It provided the IMET with a in their deciding not to assign and equipment, except for a lim- rare opportunity to gather quanti- ited number of crews assigned to tative low-level wind data on a near hold the flanks and bulldozer op- real-time basis. The consistency erators constructing fire breaks to and reliability of these data from save endangered communities. an area immediately adjacent to the wildfire (compared to occa- One hour before the weather sional surface observations from front’s arrival, the SODAR detected spotters in the field) made it pos- increasing windspeeds at the level sible to verify wind forecasts for of the ridge tops (approximately the fire area, which in turn in- 1,000 feet (330 m) above the aver- creased the IMET’s confidence in age elevation of the fire). This the accuracy of his forecasts and 1-hour warning gave the IC staff prevented misanalysis of the low- time to move remaining fire crews level wind field. and equipment out of danger from the expected runs. Within the next The IC staff also gained confidence 6 hours, an explosively developing in the forecasts. As a result, they firestorm consumed 40,000 acres placed greater emphasis on them (16,200 ha). Yet during this period, when planning their fire suppres- there were no burnovers, no equip- sion strategies. SODAR data helped Each SODAR antenna consists of a ment losses, and no injuries attrib- to provide them with an objective parabolic dish that faces upward with an uted to fire behavior. basis for their decisionmaking con- acoustic driver mounted above each dish cerning the expected erratic fire pointing downward. By emitting a sound pulse at a known frequency and monitor- Results of SODAR Use behavior during the frontal pas- ing the Doppler shift of the return echoes, The SODAR unit proved to be a sage and helped to give an addi- the instrument calculates the windspeed valuable tool for characterizing tional margin to ensure firefighter and wind direction at various altitudes. safety. Photo: Jeff Bradley, AeroVironment Inc., wind and stability patterns that in- Monrovia, CA, 1996.

30 Fire Management Notes The Future considered for temporary (24 to 48 providing wind and stability data in We hope that the success of the hour) deployments to characterize the future. SODAR at the Fork Fire will open daily wind and stability patterns in discussions among IC staffs na- the fire area both during wildland For more information about how tionwide concerning how this fires and prescribed burns. SODAR units measure windspeed, technology can regularly be incor- wind direction, and atmospheric porated into wildland fire suppres- We anticipate that the incorpora- stability and their possible future sion strategies. When erratic fire tion of radio- or satellite-data uses during other wildland fires or behavior is occurring or expected, transfer technology and the use of during prescribed fires, readers a SODAR can be a resource to as- existing single-antenna SODAR should contact Alex Barnett; tele- sist the IMET with forecasting and units will improve the rapid de- phone 818-357-9983, IC staffs with their decision- ployment and portability of the fax 818-359-9628, or e-mail ■ making. When erratic fire behavior units to and in the field. This will [email protected] is not expected, a SODAR can be make SODAR more effective in

GUIDELINES FOR CONTRIBUTORS

Editorial Policy Donna Paananen, Editor Authors are asked to use the English unit Fire Management Notes (FMN) is an inter- Fire Management Notes system of weight and measure, with national quarterly magazine for the wild- USDA Forest Service—NCFES equivalent values in the metric system. land fire community. FMN welcomes 1407 South Harrison Rd.—Rm. 220 Tables should be typed, with titles and col- unsolicited manuscripts from readers on East Lansing, MI 48823-5290. umn headings capitalized, as shown in re- any subject related to fire management. Telephone 517-355-7740, fax 517-355-5121 cent issues; tables should be under- (See the subject index of the first issue of e-mail: [email protected] standable without reading the text. Include each volume for a list of topics covered in DG: S23L03a tables at the end of the manuscript. the past.) Include with the paper copy the complete Figures, illustrations, slides (original Because space is a consideration, long name(s) and address(es) of authors as well transparency preferable), and clear photo- manuscripts are subject to publication de- as telephone and fax numbers and e-mail graphs (preferably glossy prints) are often lay and editorial cutting; FMN does print information. If the same or a similar ar- essential to the understanding of articles. short pieces of interest to readers. ticle is being submitted elsewhere, include On the back, please label carefully (Figure that information also. 1, Figure 2; photograph A, B, C, etc.); in- Submission Guidelines clude your complete name and address if Disks should be submitted with the paper you wish your material returned, and indi- Authors are asked to type or word-process copy. FMN prefers WordPerfect 5.1 or an cate the “top.” Clear, thorough captions their articles on white paper (double- ASCII text file on 3-1/2 inch, IBM/Dos- (see recent issues) should be labeled to spaced) on one side. Try to keep titles con- compatible disks. Please label the disk correspond with these designations. In- cise and descriptive; subheadings and carefully with system being used and name clude the name of the photographer and bulleted material are useful and help read- of file. When possible, submit illustrations year when photo was taken. ability. As a general rule of clear writing, on disk as well and include instructions for use the active voice (e.g., Fire managers use on the label. All photos and illustrations require a writ- know . . . not, It is known . . .). ten release, whether taken or drawn by Consult recent issues for placement of the government or nongovernment photogra- Submit articles to either the general man- author’s name, title, agency affiliation, and phers and designers. Authors not in the ager or the editor. Complete details to location as well as style for paragraph Federal government sign a release ac- reach them follow: headings and references. FMN uses the knowledging that they are aware that their spelling, capitalization, hyphenation, and work will be in the public domain and on USDA Forest Service other styles as recommended by the U.S. the World Wide Web. The photo, illustra- Attn: April J. Baily, F&AM Staff Government Printing Office “Style tion, and author release forms are available P.O. Box 96090; Manual.” Inhouse editing can be expedited from General Manager April Baily. ■ Washington, DC 20090-6090. if authors have their article reviewed by Telephone 202-205-0891, fax 202-205-1272 peers and by someone with editing skills. e-mail:/s=a.baily/ou1=wO1c@mhs- Please list the editor and/or reviewer/s fswa.attmail.com when submitting articles. Data General (DG): wO1c.

Volume 57 • No. 2 • 1997 31 5614

subscription(s) to Fire Management Notes (FMN) for $ 8.50 each per year ($ 10.65 foreign).

4/95