Stanley B. Martin

Who cares? The Cold War is over.

Yes, but we are now at war with international terrorists. A mass fire can result from a single nuclear explosion, and one of terrorism’s choices in weapons of mass destruction (WMDs) is nuclear. Nuclear weapons are above all incendiary weapons. The fire that results is often viewed as a major cause of collateral damage because it threatens such large areas, in urban attacks inevitably including large civilian populations. Offensive war planning by the United States tries very hard to avoid (at least minimize) collateral damage, stressing technologic gains in precision, penetration, and reduced energy yields. Exactly the opposite is sought by terrorists, and the extended fire damage offered by a nuclear explosion in/over/near a city, especially an airburst of larger energy yield is ideal. Anyone who has worked the problem, even superficially, knows the potential for mass fires.

My stress on the word superficially is quite intentional, as I hope to make clear in what follows.

It takes time for a technology to mature. Some times money can accelerate the process, but moving too fast too early can result in an unacceptable product.

Fifty years of research into fire effects of nuclear weapons have provided us with detailed insights; however, the whole subject is multifaceted with very different levels of understanding having been attained in each facet. Any comprehensive assessment of urban fire behavior and damage must include: primary fires, those caused by direct exposure to thermal radiation from the nuclear fireball; secondary fires, also known as blast-disruption fires because they are initiated by mechanical damage and displacement due to the air blast wave; and fire spread, both short range (typically within a city block) and long range through effects of fire-induced winds, mass fire phenomena, and firebrand spotting.

During those years, the perceived urgency at times to put the subject to use in national security arenas, prompted short cuts to get answers whether or not they could be justified technically. The facet of mass fires is a case in point. Repeatedly over the years modelers have employed what I call "flat earth models" of the fires on the ground, ignoring the details that are necessary to represent the sources and distribution of heat. These modelers often then add to their presumed uniform disk of heat output an axially symmetrical plume derived with high-powered mathematics and little or no experimental evidence.

In contrast, the facet of fire initiation by the thermal radiation from nuclear-explosion fireballs was meticulously researched for many years with extensive experimental simulation and testing. Its analytical application to whole cities may seem to require a leap of faith, but it was in fact the subject of many years of research and in-the-field surveys of American cities by various independent institutions and contractors. This process culminated in computer implemented analytical codes using geographical information systems (GIS) mapping details to provide the detailed assessments of fire starts and their distribution over any chosen urban area to support further modeling on fire spread and mass fire development. Such a GIS-supported code is documented in Martin, 1987 (See end note.), called the NWFS Code. This code, dealing with both primary and secondary fire causes, was considered state of the art at the time. It was written to facilitate updating as new data and concepts evolved. To date, it remains unmodified

During the 1960s, in the Five-City Study, which was funded and directed by the Office of Civil Defense (OCD), several competitive, yet complementary, efforts were made to develop fire-spread models beyond simply taking credit for within-the-block consolidation. These were of two distinct types: (1) Empirical, derived from historical experience and statistics; (2) Mechanistic, combining experimental data with first-principle analytics. The first type is best represented by the developments of Willoughby and his colleagues. The reader is referred to the DNA Topical Report (Willoughby, 1989). The second type is due to Takata and his colleagues (see Takata, 1990, for an up-to-date summary). A brief overall review of the state of art is given in Martin, 1991. Choice of the type of fire-spread model to employ depends heavily on the application. Concerns about mass fires demand more mechanistic approaches, partly due to the fire-induced winds that seem implicit in definitions of mass fires and "firestorms."

While many definitions of mass fires have been attempted, none is dominant; all considered, they appear to converge on issues of fire-induced winds. The essence is that when a fire is large enough to qualify, it significantly perturbs the natural wind field.

Observations of mass-fire behavior are not limited to the urban bombing raids of the second world war. Mass fires as just defined occur occasionally in forests and woodlands. Experiments intended to represent urban mass fire (but using wildland fuels) were conducted in Operation FLAMBEAU and the subsequent Mass Fires Systems Program between 1964 and 1975. Careful observers of the dynamic events have regularly reported unsteady, pulsating behavior that can lead to "runs" in fire spread behavior. One’s initial intuitive take here could be that the effect is due to gustiness at ground level. Analyses of the data say otherwise, that these pulsations are intrinsic to the burning process, becoming apparent only in areas of fire larger than about 2.5 hectares. Palmer (Palmer and Kreiss, 1985) describes the flow regime near the ground as pulsating with air-layer replacement from above. The picture is one of dome-shaped bubbles of hot combustion products mixed with air periodically breaking away from the ground, forming rising ring vortices while much of the air needed to support the fire is supplied by downdrafts from above (not from the fire’s periphery as many modelers suggest).

This behavior has several practical consequences, and these, in turn, depend on the ambient wind field. In low ambient winds, each vortex packet rises nearly vertically, inducing wind flow around the base of its source-fire or cluster of fires acting as a single source. This may inhibit short-range spread, but it provides a lifting mechanism for long-range spotting by brands. In contrast, ambient wind speeds greater than about 12 mph cause the vortex to tilt and propagate downwind, providing a possible mechanism for conflagration spread.

In 1987, the Defense Nuclear Agency (DNA) tasked my company, Stan Martin & Associates (SM&A) to "incorporate the results of previous research effort for predicting potential fire effects in actual cities into a computer program that will forecast, at city-block resolution, both blast and fire damage in collateral as well as targeted areas as functions of weapon, urban-makeup, and weather variables." Since the NWFS Code already satisfied the requirements for primary and secondary fire initiation, we viewed the project as one of expanding the existing computer program to include the issues of long-range fire spread, mass-fire development, and fire-induced winds. Thus, the new product would represent a comprehensive state-of-the-art blast/fire code.

To ensure the quality of the final product, SM&A brought together a team of specialists in the appropriate technical disciplines. Hands-on experimental and first-hand observational experiences were stressed. The team was made up of individuals who had directly participated in one or more of the following:

  • The only fire experiments ever conducted during atmospheric testing of nuclear explosives
  • Fire experiments at high-explosive events simulating nuclear explosions
  • Shocktube/shock tunnel experiments of blast-fire interactions
  • Observations of atmospheric dust loading during building demolitions
  • Tests of fire dynamics in real (and model-scale) buildings in various stages of blast damage and collapse
  • Investigations of rates of fire spread in debris
  • Studies of firebrand formation, transport, ignition of host materials
  • Conduct of mass fire experiments and observations of accidental mass fires
  • Observation and measurement of ground-level conditions in experimental mass fire followed by development of fluid-flow modeling

All of these contributed importantly to the total fire picture. The product of the three-year development effort is the comprehensive NWFIRES Code. (Martin, 1991)

A recently published book, written by Lynn Eden, Senior Research Scholar at the Center for International Security and Cooperation, Institute for International Studies at Stanford University, titled: "Whole World on Fire," stresses the importance of fire as an effect of nuclear attack, and reviews the history of fire’s role in U. S. targeting and stockpiling of nuclear weapons (Eden, 2004). To illustrate the potential extent of fire damage from a single nuclear burst, Ms Eden applies to our nation’s capital a 300-kiloton near-surface burst over the Pentagon in conditions of 10-mile visibility. She concludes her summary of the devastation by postulating a mass fire of great extent and impact: "Within tens of minutes, the entire area, approximately 40 to 65 square miles – everything within 3.5 or 4.6 miles of the Pentagon – would be engulfed in a mass fire. The fire would extinguish all life and destroy almost everything else." Consider the homeland security implications of this picture! Then seek the source of her description of this catastrophe. She has chosen to equate the mass fire area with a circle determined by (free-field) exposure to 10 cal/sq cm (or 20 cal/sq cm? You get to choose.). This is what I mean by superficial.

Well then, why not run the same scenario using the comprehensive urban-blast/fires code, NWFIRES? Even though the code has never been validated, it does take into account all of the recognized factors that can affect the outcome, and it is considered to be state of the art. Most importantly, it does not treat the subject superficially, but with as much mechanistic insight as the SM&A team of experts could bring to bear. Unfortunately, the code is not supported by a full-featured GIS fire database. None has yet been developed. Our splendid tool is only as good as we are able to represent the details of the urban target in fire-relevant terms.

Still, similar exercises have been run with an earlier implementation of the fire code, and the Defense Department’s Digital Feature Analysis Data, supplemented with fire-pertinent upgrades. Lynn Eden’s book describes some of these results and illustrates how different from hers they are. The main difference of note here is that our analysis, supported by a GIS showing ground features, while it forecasts large areas of extensive fire damage, forecasts only small pockets of mass-fire activity (called firestorms) in much the same part of the circular urban area she shows, within tens of minutes, to be all mass-fire involved. This illustrates how vitally important inclusion of the details of target makeup are to the predicted outcome.


Martin, S.B., 1987: CODE FOR EARLY TIME FIRE PHENOMENOLGY, Nuclear Weapon Fire Start Code, DNA TR-87-233, Final Report, SRI International.

Willoughby, A.B., 1989: The Use of Historical Data for Predicting the Rates of Fire Spread in Urban Areas Following Nuclear Explosions, DNA Topical Report, SM&A.

Takata, A.N., 1990: Mechanistic Urban Fire Spread Code (FISAT), DNA Topical Report, SM&A.

Martin, S.B., et al., 1991: Model Development for Nuclear Weapon Induced Large Urban Fires (NWFIRES), DNA-TR-90-86, Stan Martin & Associates, May 1991.

Palmer, T., and W.T. Kreiss, 1985: Eddy Coefficients in Large Fire Modeling, Paper presented at Conf. on Large Scale Fire Phenomenology, Sept. 10-13, 1984, Gaithersburg, MD.

Eden, Lynn, 2004: Whole World on Fire: Organizations, Knowledge, and Nuclear Weapons Devastation, Cornell University Press.