April 1999 By David L. Martell, Al Tithecott and Paul C. Ward The Ontario Ministry of Natural Resources (OMNR) is responsible for forest fire management on 781,346 square kilometers of public land in the Province of Ontario. From 1987 through 1996, an average of 1,791 fires burned an average of 280,000 hectares per year. Ontario's publicly owned forest land provides many benefits including industrial timber, recreation opportunities, wildlife habitat and wilderness parks in which human activities are curbed to protect natural ecosystems. Fire poses threats to public safety, timber production and other human endeavors. However, because fires are a natural component of Ontario's forest ecosystems, fire managers must balance such threats with the disruption of natural processes. Forest fire managers respond to fires much like their urban counterparts, but their task is complicated by vastly larger protected areas, significant temporal and spatial variability in fire occurrence and behavior processes, and the recognition that fire cannot and should not be completely excluded from all forest areas for social, economic and ecological reasons. Forest Fire Management in Ontario The OMNR's forest fire management program is administered by its Aviation, Flood and Fire Management Branch (AFFMB), which maintains the Provincial Response Centre in Sault Ste. Marie, and two regional fire centers in Dryden and Sudbury. The OMNR's fire control objective is to contain fires while they are small and thereby reduce the number of large and potentially destructive "escaped" fires that can cost hundreds of thousands or even millions of dollars to extinguish. A fire management priority zoning scheme is used to link fire management strategies to land management objectives. The OMNR carries out prevention activities to minimize people-caused fires and administers a detection system designed to find fires soon after they are reported. It depends upon the public to find and report fires near populated areas and uses fixed wing aircraft to fly detection patrols in more remote areas. Ontario is blessed with an abundance of "landable" lakes and "pumpable" water that has contributed to the development of an initial attack system based on fire crews augmented by fixed wing airtankers. Initial attack crews comprised of three or four firefighters travel to each fire by helicopter or truck and use power pumps, hoses and hand tools to establish a control line around the fire. Initial attack crews are often supported by airtankers that usually arrive before them and fight the fire until those crews begin their suppression action on the ground. The OMNR's fleet of amphibious airtankers is based at airports. They skim over the surface of lakes close to the fire, scoop water into their onboard tanks, then drop that water on the fire. The OMNR recently upgraded its fleet of nine piston-engine CL-215 airtankers to more modern turbine-powered CL-415s. The CL-415s are supplemented by a fleet of five smaller Twin Otter aircraft that are used for transportation purposes, but are equipped with special amphibious floats that contain tanks that can be used to pick up water from lakes. Since area burned is a nonlinear increasing function of time and the probability of containment decreases as fire size increases, it is essential to minimize response times. To achieve this, fire managers deploy airtankers and initial attack crews to bases or airports close to areas where fires are expected to occur, from which those suppression resources can be dispatched to fires soon after they are reported. First Steps Forest fire management poses many interesting challenges for operations researchers. Since fire managers tend to be innovative, the two groups have worked very well together in Ontario. In the late 1960s, Glenn Doan, project officer with the Fire Control Unit of what was then known as the Ontario Department of Lands and Forests, learned of some early attempts at the Operations Research Center at the University of California, Berkeley, to apply OR to forest fire management. Doan convinced his colleagues that they should explore the possibility of using OR in Ontario. His foresight created a solid foundation upon which the application of OR to forest fire management has continued to grow in the province. Many of the publications that describe that work are discussed in Martell [1982] and Martell, Gunn and Weintraub [1998]. The OMNR is now comfortable integrating outside OR expertise into its decision processes, and does so on a regular basis. But more importantly, it has moved beyond being an informed consumer; staff members have developed expertise that enables them to develop and implement OR on their own, relying on outside experts for guidance. Many OMNR staffers have come to believe OR is an essential component of sound forest fire management planning. That is evident in their reliance on in-house experts and others to bring OR to bear on important decision-making problems. Initial Attack Study One of the earliest successes in the application of OR to forest fire management in Ontario was a study of initial attack requirements carried out by a four-member working group of two OMNR fire management staff and two OR specialists including the senior author. An initial request for funds to upgrade an aging aircraft fleet was met with a demand from the provincial treasury officials that the OMNR carry out a comprehensive cost-effectiveness analysis of its aircraft needs. It became obvious early in the task that aircraft could not be assessed in isolation from other fire suppression resources. As a result, we developed a relatively simple yet comprehensive deterministic simulation model of the initial attack system in which specified sets of airtankers, transport helicopters and firefighters battle historical fires. The model was used to evaluate initial attack system alternatives and support the OMNR's request to purchase CL-215 airtankers. The model and its implementation are described in Martell et al. [1984]. That project appears to have been the first formally documented implementation of OR in forest fire management. Airtanker Home Basing More recently, OR was brought to bear on the problem of airtanker home basing in Ontario. The OMNR's fire organization has had to deal with significant changes in funding and increased demands on limited resources through a period of reorganization and downsizing. One consequence of these changes was a need to rationalize airtanker home basing in a strategic way. The AFFMB assembled a special working group and invited one of us (Martell) and one of his graduate students (J. MacLellan) to support the deliberations of the working group. We developed a mathematical programming model that was used to help evaluate airtanker home basing strategies in Ontario. The model and its implementation are described in MacLellan and Martell [1996]. Some Open Challenges Some of the problems that initially appeared to be quite simple to solve proved to be much more difficult than we ever anticipated. Daily airtanker deployment is a good case in point. Each day fire managers must decide where to deploy their airtankers to minimize their initial attack response times. The airtanker system can be viewed as a queueing system with airtankers as servers and fires as customers. Two of us developed and field-tested a queueing model (which is described in Martell and Tithecott [1991]). We partitioned the Northwest region of Ontario into several sectors, each of which contained one and only one airport, and modeled airtanker operations in each sector as an independent M(t)/M/S queueing system. We developed the software to solve the time-dependent queueing system equations numerically. We included a user interface for input and a graphical display of the performance of each of the queues, and provided the staff in the regional fire center access to the model. They ran the model throughout the summer and found the results to be interesting, but pointed out that it would not be of any practical use to them unless it was enhanced to account for interaction between bases; that is, that airtankers not be restricted to fires in their "home sector." They encouraged us to return when the model was improved. We learned that daily airtanker deployment support is more important as the fire manager's planning area increases, and that future airtanker deployment models must be designed to deal with geographically large areas where deployment decisions are more important to ultimate success. Airtanker systems pose very difficult challenges to applied queueing theory. Forest fire occurrence can be modeled as a Poisson process with fire arrival rates that vary over both time and space. Some random number of airtankers (sometimes none) is required on each fire, and the service time depends upon waiting time as fires grow while they wait in the initial attack queue. Travel time is a very significant portion of the service time and servers are not always identical. The OMNR can borrow airtankers from other forest fire management agencies during particularly busy periods which fire managers often refer to as "fire flaps." Our return to the field has taken much longer than we anticipated. We were aware of the urban emergency response work carried out by Larson and his colleagues, particularly the hypercube queueing model (HQM) (see for example, Larson and Odoni, [1981]). But in the case of airtankers used for forest fire control, it is necessary to account for diurnal variation in fire arrival rates and the long travel distances involved and their potential impact on dispatch ranges due to fuel tank capacities. Eventually K. Islam, a Ph.D. candidate in Mechanical and Industrial Engineering at the University of Toronto, tackled the problem under the supervision of Martell and M.J.M. Posner. He formulated the airtanker problem with multiple interacting bases and constraints on initial attack dispatch radius as a time-dependent M(t)/Ek /S extension of Larson's HQM. We hope to field test his model during the 1999 fire season. Growing Acceptance of OR Within Ontario's Forest Fire Management Community The initial attack and airtanker home basing projects described above illustrate how operations researchers can work with fire managers to solve important fire management problems. But the most significant contribution of those that have participated in OR fire management initiatives in Ontario has been the development of what might best be characterized as a decision analysis culture. Forest fire managers in Ontario have recognized the need to augment their traditional reliance on training, experience and intuition with formal decision analysis. Ontario's forest fire management community now includes many informed OR consumers that use models developed by others, and in some cases, develop and use their own models. They are neither intimidated nor wowed by OR, but they have a healthy respect for many of its strengths and weaknesses. For example, several years ago the OMNR recognized that the process of awarding contracts for seasonal helicopters was cumbersome and limited their ability to consider various contract alternatives. Many helicopter companies are invited to bid on contracts to provide up to a total of 14 aircraft, to be located at many different bases. Contractors may bid on one or many bases with one or many aircraft and, at the time, were invited to provide prices for one, two or three seasons. Of course, only 14 bids would be awarded contracts from the several hundred tendered options; searching for a minimum cost alternative using the qualified bids is a classic mathematical programming problem. Because of the general awareness and training in OR techniques, managers were aware that a more robust and efficient analysis that would minimize overall costs was within reach. With relatively little programming effort, the problem was formulated as a mathematical programming problem by AFFMB staff (Tithecott and Lemon, [1992]), which reduced the analysis time, guaranteed a lowest cost to government, and provided a platform for rapid reanalysis of alternatives. More recently, staff at the AFFMB developed a simple fire fighter and aircraft demand forecasting model that is used during fire flaps when new initial attack fires and ongoing commitments to large escaped fires tax suppression resources to their limits and beyond. Once a fire flap becomes entrenched, the OMNR must look several days in advance to assess firefighter and aircraft needs to cope with existing escaped fires, new escaped fires and new initial attack fires. The consequences of resource shortages can be very significant, but crews can be recycled quickly from one fire to another, and additional resources can be borrowed from other agencies. Two OMNR staff (R. McAlpine and V. Scott) developed a relatively simple spreadsheet forecasting model that is now used to help evaluate manpower planning strategies in anticipation of fire flaps. The most recent example of in-house expertise and the OMNR's acceptance of OR is the development and use of the Leopards level of protection analysis system. In the early 1990s the AFFMB commissioned the development of a more comprehensive level of protection analysis system that would support various strategic decisions of the program. The initial attack system model developed by Martell et al. [1984] had been modified in the intervening years by D. Boychuk to meet the needs of another forest fire management agency. It was later updated by Martell et al. [1995] to produce an application called Lanik for use on desktop computer platforms in Ontario. With the support of AFFMB, R. McAlpine, an OMNR staff member who had not been involved with the earlier initiatives, extended Lanik and embedded it in a geographic information system (GIS) to produce a Windows 95 model called Leopards. Leopards has provided a comprehensive decision-support platform. Transfer of the model to the organization has provided valuable expertise to fire managers who have been struggling with a complex array of issues, including the potential implications of global climate change, initial attack fire crew staffing levels, and a decision to upgrade the airtanker fleet with modern CL-415s. The relationship between the OMNR and OR also extended into other areas of forest management. Forest management planning research carried out by graduate students at the University of Toronto (D. Boychuk and R.G. Davis) was transferred to the OMNR and has created fundamental shifts in forest management modeling. The Strategic Forest Management Model (SFMM) includes average fire losses with harvest and silviculture options to provide for strategic linkages between the level of fire protection and forest sustainability. Ontario's Fire Management Information System The OMNR's ability to use OR for operational planning purposes will ultimately hinge on the availability of a proven management information system. One of us (Ward) served as the project manager for the development, implementation and maintenance of the OMNR's comprehensive Fire Management Information System (FMIS) to support both daily planning and information requirements, as well as strategic planning and policy development. The FMIS currently includes daily operational information systems for tracking fire and weather information; resource management systems for tracking and managing personnel, aircraft and equipment; and cost tracking systems. The AFFMB has been a progressive user of computers to support daily operations since the early 1980s. The Daily Fire Operations Support System (DFOSS) replaced an earlier decision support system in 1993. DFOSS is a client-server system with more than 100 Macintosh workstations located in 30 fire management offices, linked to a central VAX/Alpha-Oracle server. DFOSS provides easy-to-use tools to capture and manage daily weather, fire and lightning data, and provides access to decision support systems for daily planning, weather analysis, fire behavior and fire occurrence prediction, and cost tracking, as well as information management tools to generate a wide variety of maps and reports. DFOSS was developed through an extensive user-consultation process designed to create an application that was easy to use, supported the daily operational needs of its users, was sufficiently robust to deal with contingencies such as hardware failures and network outages, and could be operated from remote locations. The OMNR is currently implementing a migration strategy to move DFOSS to a new technology architecture and to expand its capabilities. In phase one, DFOSS data will be linked in real time to its NT server network. Analysis and decision support tools will be migrated or new tools will be developed to access the data through the NT servers; subsequently the transaction processing/data capture applications will be migrated to the new environment. The Importance of Field Experience Most operations researchers are aware of the importance of gaining an understanding of what really happens on the "shop floor," or in this case, the "forest floor," and the OMNR has proven to be an excellent learning environment. All the OMNR staff involved in OR projects have extensive histories of involvement in many aspects of fire fighting and fire management. University researchers have certainly benefited from the OMNR's openness. One of us (Martell) spent several summers working in the provincial fire center in Toronto as a student and then was fortunate enough to work on a fire crew in Thunder Bay District for part of a fire season after he completed his Ph.D. More recently, he had an opportunity to work as a fire intelligence clerk on a project fire on which another one of us (Ward) was the fire intelligence officer. That tradition of cooperative learning and connection with field operations continues as one of Martell's graduate students recently spent two summers working on a fire crew in Ontario. Operations researchers are also welcomed into fire centers to observe and discuss the realities of decision-making under uncertainty. Our success in Ontario is due primarily to the firefighters and managers that have continually demonstrated their commitment to improving Ontario's forest fire management program. They view fire as an interesting challenge and leave no stone unturned in their never-ending quest for better ways to do their jobs. Their tendency is to cooperate fully with researchers who take time to study their operations and identify their true needs, and they are almost always willing to field test new technology. Acknowledgments Much of the research described in this article was supported by grants from the Natural Sciences and Engineering Research Council of Canada, the Ontario Ministry of Natural Resources and the Canadian Forest Service. Many OMNR fire fighters and managers contributed to our successes. At the risk of omitting many important players, we wish to acknowledge the contributions of R.V. Brady, G.E. Doan, R.J. Drysdale, J.F. Goodman, R. Kincaid, P. Parker and H. Redding. D. Boychuk, P. Kourtz and J. Maloney made significant research contributions that are described in Martell [1982] and Martell et al. [1997]. References Larson, R.C, and A.R. Odoni, 1981, "Urban Operations Research," Prentice-Hall, Englewood Cliffs, N.J. Martell, D.L., 1982, "A review of operational research studies in forest fire management," Can. J. Forest Res., Vol. 12, No. 2, pp. 119-140. Martell, D.L., D. Boychuk, J.I. MacLellan, B.M. Sakowicz and R. Saporta, 1995, "Decision analysis of the level of forest fire protection in Ontario," in proceedings: "Symposium on Systems Analysis in Forest Resources: Management Systems for a Global Economy with Global Resource Concerns," Pacific Grove, Calif., Sept. 6-9, 1994. Edited by John Sessions and J. Douglas Brodie. pp. 138-149. Martell, D.L., R.J. Drysdale, G.E. Doan and D. Boychuk, 1984, "An evaluation of forest fire initial attack resources," Interfaces, Vol. 14, No. 5, pp. 20-32. Tithecott, A.G. and C. Lemon, 1993, "The use of management science for the management of helicopter contracts," AFFMB Publication No. 306. Web site: Fire Management Systems Laboratory at the Faculty of Forestry, University of Toronto, http://www.firelab.utoronto.ca. David L. Martell is on the Faculty of Forestry, University of Toronto; e-mail: martell@smokey.forestry.utoronto.ca. Al Tithecott is on staff at the Aviation, Flood and Fire Management Branch, Ontario Ministry of Natural Resources, Sault Ste. Marie, Ontario, Canada; e-mail: titheca@gov.on.ca. Paul C. Ward is with the Science and Information Resources Division, Ontario Ministry of Natural Resources, Sault Ste. Marie, Ontario Canada; e-mail: wardp@gov.on.ca Table of Contents OR/MS Today Home Page OR/MS Today copyright © 1999 by the Institute for Operations Research and the Management Sciences. All rights reserved. Lionheart Publishing, Inc. 506 Roswell Street, Suite 220, Marietta, GA 30060, USA Phone: 770-431-0867 | Fax: 770-432-6969 E-mail: lpi@lionhrtpub.com URL: http://www.lionhrtpub.com Web Site © Copyright 1999 by Lionheart Publishing, Inc. All rights reserved.