Forest and Wildland Fire Management: a Risk Management Perspective (17w5145)

Arriving in Banff, Alberta Sunday, November 5 and departing Friday November 10, 2017


(University of California, Los Angeles)

David Brillinger (University of California-Berkeley)

Mike Wotton (Canadian Forest Service)

(University of Toronto)

(University of Western Ontario)


The objective of the workshop is to bring together researchers and end-users from a variety of disciplines (e.g., operations research, statistics, actuarial science, mathematics, computer science, forestry and the environmental and health sciences) to provide a forum to discuss fire management needs and to initiate interdisciplinary team-based collaborations aimed at addressing important problems in forest fire management.

Decision Support Tools for Fire Management: In order to make progress on fire management objectives, mathematical and statistical expertise will need to be brought to bear on the issues of uncertainty which complicate operational planning. There is a need to deliver “the right amount of right fire at the right place at the right time at the right cost”. For example, what is the best strategy for an individual airtanker? Is it better to return to its original home base after completing service on a fire, or should it return to the base nearest to that fire? Although progress has been made on this kind of problem using tools from operations research, there is a very urgent need to develop and test better predictive fire occurrence models, especially ones that incorporate the clustering of arrivals which can overwhelm agencies. There is also a need to quantify the initial attack system, including estimating the probabilities for same day initial attack, and of initial attack success since the larger project fires are a large drain on crews and equipment and incur large costs.

A strength of a probability-based model is that the variability in the data will be reflected in the standard errors associated with predictions or fitted values. However, an inadequacy in fire mapping is the difficulty of conveying this uncertainty to users of the modelling product. Two types of uncertainties need to be addressed directly: 1) the standard error of the predicted probability and 2) the fact that predicted probabilities have an inherent uncertainty in the sense that although the model returns a probability, the event of interest is binary. How can these uncertainties be best incorporated into decision support tools aimed at informing fire managers?

Other key risk management issues to be discussed include harvest scheduling optimization, optimal allocation of scarce fire suppression resources, and examining effects of changes in fire detection management and technology.

Fire Regime Modelling: The economic and social impacts of wildfire in Canada raise important statistical questions. Once an area has burned, does it reduce that local area’s risk of burning and if so, for how long and can such effects be replicated by active management of forest fuels? What ecological and forest management strategies are needed to maintain the current landscape mosaic and levels of biodiversity while protecting people, property and other values at risk? There is an interest in developing methods for simulating seasonal boreal forest fire regimes, comprising frequency, size, and severity, noting that there is a strong need to “backcast” characteristics of past fire regimes if there had not been any fire suppression. In this context, the focus involves shifts to modelling aggregate behaviour of fire regimes, say over an entire year and across a large study area (e.g., a province), rather than modelling at fine scale resolutions. Historical records of individual burn scars for many fire seasons are now becoming available in digital format. Fires can be viewed as realizations from a marked spatio-temporal point process (where the points are the ignitions and the marks are the characteristics of each individual fire, such as its duration, intensity, size, etc.). The fire ignition records combined with the burn scar data can be viewed as approximately independent, replicated point pattern data. Important topics for discussion include developing tools for exploring and visualizing these data, for investigating inhibitory impacts of burn scars on the future intensity of points in that area, and for determining whether such marks are separable, and if not, how to jointly model points and marks.

Bridging Gaps Between Researchers and Fire Managers: Many factors, including climate change, changing land use patterns, fuel build-ups, fiscal realities and the fact that fire is a natural ecosystem process contribute to the emergence of more and increasingly complex and challenging fire loads. Hence, there is a growing and urgent need for decision support systems that fire managers can use to enhance their planning and decision-making processes. The development of such decision support systems calls for bringing experts from the mathematical and decision sciences together with ecologists, fire scientists and fire managers to develop collaborative team-based efforts to address these problems and BIRS is an ideal venue to do so.

The researchers on the organizing committee and many on the proposed list of invitees have longstanding, demonstrated success in this area. BIRS has supported this ongoing research program in the past and that support has helped us move well beyond the point of formative collaborative discussions to the strong, interdisciplinary team of researchers with common interests and goals it is today. These past successes have also brought us to the point of considering large-scale (e.g., province and/or national) models and our team has the combined quantitative, applied and subject-matter expertise to be successful at this. But much remains to be done; the emphasis on operations research for this workshop is a new initiative and a natural next step to closing the loop between basic science and management.

Several members of this team are also involved as investigators/collaborators on a CANSSI (Canadian Statistical Sciences Institute) Collaborative Research Team grant. This three year CANSSI project began in April 2015 and runs until March 2018. We envision this proposed BIRS meeting as a mid-project team workshop. It will serve as a forum for a series of in-person meetings to focus concerted efforts on making further headway on these projects, and engage the end-users for feedback at a midway point. Training is a large aspect of this CANSSI team and many HQP arrangements are already in place, including interdisciplinary co-supervision arrangements for MSc students, several PhD students have been identified and are beginning studies, and there are plans for PDFs. Many of these HQP will be invited to attend and participate in the proposed workshop, which will provide an excellent opportunity for networking and interaction with the end-users.