Origin: First and always First

The need for scientific validation of wildland fire pattern indicators Keith Parker

Abstract

Wildland fire origin investigation relies on accurate analysis of multiple data sets. Accounts from bystander, first-in fire crews and surveillance cams provide helpful information to investigators, but the most important data set is accurate interpretation of directional fire pattern indicators. To accurately interpret fire patterns, we must be able to identify and properly interpret the directional patterns formed when the initial fire front moves past the point where an indicator is located. Fire patterns and their meaning is believed to have been developed through observation of experienced personnel over the years. Searches for scientific studies or research to validate the current understanding of fire indicators have yielded no results. The need to scientifically study the accuracy of fire pattern indicators and their interpretations in varying conditions is critical to ensure accurate origin determination required to accurately determine the fire cause.

Introduction

Wildland fires around the world have become larger and more costly in the last half century.

The process of prosecuting persons intentionally starting fires has been an age-old endeavor while the practice of identifying a responsible party to recover financial damages and suppression costs is relatively new. Regardless of criminal prosecution or financial recovery, locating the correct fire origin is critical to accurate cause determination. NWCG PMS 412, NFPA 921, Paul Kirk’s, Fire Investigation and Richard Ford’s Investigation of Wildfires, and others emphasize the requirement to first accurately identify a fire’s origin before beginning the process of determining the cause. These publications all contain a series of fire pattern indicators but do not identify how, by who or when these patterns were developed. Errors in origin identification will nearly always result in erroneous determination of the fire cause.

Richard Ford wrote in 1987: “When the origin has been pinpointed with known certainty, then begins the crucial task of identifying the fire cause”. Both Ford and Dr. Paul Kirk were ahead of their time in advocating for the use of proven science within a systematic method to determine a fire’s origin; and only after the origin was established to a high level of certainty, to then begin the process to determine the cause. Prior to the creation of NFPA 921 in 1992, fire investigation was considered an “art” or a mix of art and science passed down by experienced fire investigators from generation to generation. Most of the accepted fire pattern interpretations that were handed down have been scientifically proven to be correct. However a significant number have been proven to be folklore without basis in science and were simply wrong. Spalling on concrete, glass crazing, heavy soot deposits on glass and burn holes in floors were believed to only result from the direct pour of flammable liquids used as accelerants. Unfortunately, these folklore patterns wrongly put many people in prison and resulted in one execution for arson that the suspect did not commit, or at least in the manor determined by investigators at trial [14].

The National Wildfire Coordination Group (NWCG) published PMS 412, Guide for Wildland Fire Origin and Cause Determination in 1978, which was updated in 2005 and 2016. This guide is a collection of education, training, and experiences of a group of mostly public agency wildland fire investigators. The 2005 guide embraced the scientific method and methodology used in NFPA 921 directly. The updated 2016 version, some argue, is a departure from NFPA 921 methodologies including alterations of the scientific method. The NWCG group review process of the material is performed within the working group itself; hence no public comment or outside peer review is part of the development or publication process. There is no doubt PMS 412, along with the FI 210 Course exemplify advancement and progress in the specialty training and education of wildland fire investigators. That does not change the fact there is a lack of scientific data supporting the pattern interpretations and significant methodology conflicts between NFPA 921 and PMS 412. These conflicts must be resolved quickly and honestly by all the interested parties using science and scientific studies. The results must come from the science and not from past practices.

Current reference materials and instructional courses on Wildland Fire Investigation provide information and methodology to locate a fire origin using clusters of valid fire spread indicators. The process is designed to locate areas and direction of advancing, lateral and backing fire spread created by the initial fire run. When the area with no specific fire spread direction is identified, (often called the specific origin area) the point of origin will lie within this area. While this methodology is sound and is consistent with the scientific method, determinations are reliant on accurate identification of the initial fire run and interpretation of the wildland fire spread directional pattern indicators. The heart of the issue is, that available materials do not reference how or what data was considered in developing the pattern indicator categories, how to interpret, and nor how to apply the many exceptions identified for each category.

Scientific research into wildland fire behavior and fire effects on soils, vegetation, hydrology, and other areas of study have provided many scientific based answers and validation of some fire spread indicators under specific conditions. Gustsell and Johnson’s 1996 paper on How fire scars are formed: Coupling a disturbance process to its ecological effect is one example. However, the conclusion of what to my knowledge is the first large scale study of wildfire spread indicators, A preliminary study of wildland fire pattern indicator reliability following an experimental fire, 2017, by Albert Simeoni concluded more research is needed regarding why and how the “indicators are formed” and their proper interpretation. This study placed items in the advancing path of a well-established linear ignition fire under heightened burning conditions. The real time video data showed directly conflicting micro indicators to fire direction at data monitoring locations.

The April 2016 PMS 412, lists eleven Fire Pattern Indicator Categories, down from fourteen in 2005, and mentions multiple exceptions to each indicator category. Categories have between three and six “possible exceptions” regarding the pattern reliability and that is assuming the pattern itself and type of fire spread is correctly interpreted. Ford’s 1995 book includes the paper on Fire Spread Indictors (Florida Forestry) in the appendix that also includes multiple exceptions in pattern interpretation as well. Nearly all exceptions are related to the type of fire spread and or elevated burning conditions. Assessing indicators for reliability is critical to accurate origin identification, however data to reliability perform this task is missing. Reference materials continue to change some pattern interpretation and pattern categories. These changes are not based in scientific study, but rather continuing review and refinement of what has been passed down through the generations of wildland fire investigators. With the ever-changing description and exceptions to fire pattern indicators it is incomprehensible that more effort has not been put into researching these issues to improve reliability.

2. A cause in search of an origin

The lack of research and data specific to fire spread indicators requires Investigators to fill in missing gaps, relying on experience, training, and education that is often not science based. Hence, investigation may become more art, where gaps in the science are filled with unproven past practices. This opens the door to misinterpretations of patterns, inaccurate origin identification and often applying improper methodology, such as determining a cause prior to the origin.

I investigated a number of the October 2017 California North Bay Fires. All were well established with short range spotting zones leading to the advancing fire spread within minutes of ignition. In the months after the fires I reviewed numerous investigation reports, public and some private, and was surprised by the differing interpretations from the same fire spread indicators. Reports and data I developed indicate that in seven of these nine fires, investigations found a possible cause prior to origin determination. Further investigation confirmed that the identified origins of these fires were inaccurate. In one specific instance video taken more than an hour prior to the first report clearly shows the fire over 1/2 of a mile north of the first origin and a 1/3 of a mile north of the second origin determined by the agency investigation. The first origin was dismissed when the suspect cause was ruled out requiring the shift to the second and still incorrect origin. One fire the general origin area was established by initial witnesses: the first in Police Officer who snapped two early photographs and gave specific locations of the fire perimeter. Macro indicators and photos discovered throughout the investigation confirmed the greater origin. The public agency investigation made a specific origin determination using a specific cause. Nearly a year later tests ruled that cause out and the origin changed to fit a slightly different cause in the same greater area. Improperly using the process of elimination, they made a different cause determination without any affirmative evidence.

Shortly after the report was released and recognized for the inaccuracies a new cause was found by private interested parties 300 feet away at a location that was refuted or disproven by numerous data sets including photos and detailed statements of the first arriving police officer. The inaccurate origin determinations in all these cases resulted in equally inaccurate cause determination for at-least seven fires.

Most investigators received their initial training on controlled fires conducted within National Fire Danger Rating System 40th to mid 60th percentile burning conditions. However, most wildfires fires occur in the moderate fire behavior range of the 60th to 80th percentile, and large and devastating fires almost always occur during heightened burning conditions in the 90th plus percentile. Fires burning in these dry windy conditions do not leave fire pattern indicators easily identifiable as commonly found in lessor burning conditions. My experience with the North Bay Fires and others has led me to hypothesize that when clear fire patterns are not present, are conflicting, or difficult to interpret; it has become common practice to locate fire cause first.

Once a possible cause is located, the cause itself is often used to determine the point of origin. Not only is the practice violating the critical requirement to identify an origin prior to searching for a cause, it is using a cause as a fire pattern indicator which it is not. Investigators must be aware of the pressure to find a cause and the associated cognitive biases. Not doing so can lead good investigators to make the cardinal error of deciding on a cause without an accurate origin determination.

The extreme burning conditions the week of October 17, 2017, did not leave fire patterns commonly found under moderate burning conditions. The difficulty in finding, interpreting, and validating fire patterns in these conditions, I believe, opened the door to searching for a cause first, thus entering expectation and conformation bias. I call this phenomenon “a cause in search of an origin”. The following is paraphrased from a fire investigation report narrative. Some details were omitted and changed to conceal the specifics of the investigation.

I visually examined the powerline conductors confirming all were intact as I drove down X-Road. I parked near 123 X-Road. The road past 123 X-Rd. was blocked with down tree limbs and debris. I walked up on X-Rd. I observed two powerlines laying on the road between two poles. I walked to the fire’s edge and observed backing on both sides of the road. I did not observe any additional powerline conductors down. I returned to the section of downed powerlines between the poles. I returned to 123 X-Rd. I walked the span of powerlines below. I visualized a pole. After walking the powerlines on the uphill, I walked them in the other direction that took me to a power pole in the field at 123 X-Rd., across the creek. I observed the powerline to the outbuilding resting on the ground. Sometime later (still at nighttime) another Investigator arrived. I briefed Investigator John Doe on the fire and investigation. We flagged off an area with yellow and black flagging above and below 123 X-Rd, designating an area not to be entered.

Based on these nighttime observations in significant smoke conditions, the greater origin was identified. Not surprisingly, marking indicators in a rock/gravel covered field; a series of red, yellow, and blue flags were placed depicting a point of origin where the powerline was down. Had the investigation properly evaluated the indicators and considered all the data they would have realized that the occupants of 123 X-Rd were not the first reporting party as believed, they were not looking at the fire’s first run but rather multiple spot fires that came across the road and creek. The fire had been reported across the road and 1/4 mile or more up the hill 15 minutes prior to the call from 123 X-Rd. Fire units were on scene and radioed that the fire had just jumped X-Road moments prior to the occupants of X-Road calling to report they were trapped by fire. The 911 Operator advised the caller “We are on scene of the fire across the road from your address, are you saying there is fire on your side of the road?”. A clear origin was identified using tight backing, lateral and advancing indicators consistent with the first reported location. The initial run from the correct origin ran down the side of the tight drainage and out onto a flatter slope where a topographic induced wind direction changed the fire spread nearly 90 degrees crossing X-Road to where 123 was located. The correct origin was disclosed to the public investigation however it was dismissed based entirely on observations made that first night. Months later, after all the data had been gathered, the uphill location was conclusively confirmed as the point of origin. This report is one of many that clearly contains an assumed cause as the primary fire indicator for origin determination. Such investigations are purely an exercise in expectation and conformation bias.

3. What can be done

The important outcome of fire investigation to prosecutors, litigators and public, is identification and responsibility of the cause. Accurate origin determination is or should be the critical first step for all involved parties, but commonly vanishes and the cause becomes an argument using the improper uses of the process of elimination. When cause is the focus without an accurate origin determination the argument becomes “A” was all that was there. Investigators must be honest with themselves and do far more to educate our politicians, lawyers, judges, and the public that without an accurate origin determination any cause determination should be highly suspect if not outrightly rejected. Wildland fire investigation must return to the strict methodology of accurate origin identification (without any causal consideration) before beginning the process of identifying all possible causes.

The process of origin determination must be based on the scientific method using scientific principals and studies. The Daubert ruling, federal rules of evidence and subsequent clarifications, does not recognize “training, education and experience” not based and developed on accepted scientific principals and scientific research, as grounds for developing or proffering expert opinions. While Daubert rules may not apply across every State, NFPA 921, 1033, ASTM Standards on forensic Science do. The challenge for the wildfire investigation community is not only staying true to the scientific method but developing proven scientific principals through valid scientific studies to give guidance on fire pattern interpretation.

I believe one reason studies have not been conducted is the unfounded fear they will invalidate many if not most current practices. This was the concern when NPFA first proposed 921 with the requirement to use the scientific method and principals backed by research. Over the three decades since the first addition of NFPA 921, the Investigation community has made huge strides in knowledge, best practices, and accuracy. Some investigative specialties like Forensic Odontology identifying human bite marks on flesh came into acceptance without any historical empirical evidence, scientific studies, or data to support the claims made by experts. When pressed to validate their findings the response was ipse dixit, it is true because I say it is true and nothing more. After a short run this practice lost acceptance in the forensic community and courts. Nothing like that has happened with the evolution of NFPA 921 and will not happen in the wildfire investigation community either. Unlike human bite marks, wildland fire investigation was developed over decades of examination of empirical data collected by observation and some experimentation by highly experienced practitioners. Now, scientific experimental data is needed to validate and improve most of what is accepted today. Some of our interpretations will probably need to be modified, particularly in the context of the burning conditions and some assumptions may be excluded from future consideration in the same way spalling, glass crazing and holes burned through floors were with the evolution of NFPA 921. To make accurate scientific origin determinations with a very high level of certainty, we must consider only task-relevant valid data. In summary, the scientific data for recognition, identification, interpretation, and validation of wildland fire pattern indicators is incomplete.

4. Studies and data needed

Studies to determine the accuracy of recognized fire spread directional indicators are the logical starting point. Reproducible data must be developed while monitoring fire spread direction and type as the fire perimeter passes control points. Data analysis by experienced investigators using controlled blind assessment to determine the accuracy of investigator observations, analysis of fire direction, and type of fire spread: Advancing, lateral or backing would be one method.

After baseline studies on indicator accuracy are completed, additional test burns under differing weather, fuel, and topography conditions is a logical next step. Data collected and compared to other test burn plots may provide guidance on how differing burning conditions change the nature and reliability of fire pattern indicators.

5. Conclusions

Wildland Fire Investigation origin and cause determination at its core is a forensic fire investigation specialty requiring additional specific training beyond but not in place of standard forensic fire investigation. It has not enjoyed the same attention, research, or funding over the last three decades that compartment or structure fire investigation has. The new standard NFPA 1321 Fire Investigating Units, expected out in 2024 or 2025, revisions to existing standards such as NFPA 1033 Professional Qualifications for Fire Investigators and 921, the increasing spotlight and importance of Wildland Fire Investigation is going to force compliance with the same standards as all other forensic investigation practices.

Wildland fire investigation must be based on accepted scientific principals and research. Training, education, and methodologies must be based on known science, and not exclusively on tradition and experience handed down over the generations. Advanced training in fire investigation methodology, wildland fire behavior, origin determination and cause determination is critical to ensuring accurate findings. The NWCG FI 210 Course is an outstanding basic fire investigation course but there is a need for more advanced origin and cause training. Just as fire behavior has a basic, intermediate, and advanced courses, Wildland Fire and Origin and Cause Determination needs a minimum of three levels of training. Currently the advanced wildland investigation courses such as FI-310 pertain to law enforcement methodologies to determine responsible parties irrelevant to origin and cause determination. The first step for Wildland Fire Investigation’s to embrace a true scientific approach through research validating, revising, adding, and disproving what we currently assume about fire pattern indicators.