Conference Paper

Explainable machine learning approaches for understanding fire outcomes

Authors:
To read the full-text of this research, you can request a copy directly from the authors.

No full-text available

Request Full-text Paper PDF

To read the full-text of this research,
you can request a copy directly from the authors.

ResearchGate has not been able to resolve any citations for this publication.
Article
Full-text available
Whether ignited by lightning or by Native Americans, fire once shaped many North American ecosystems. Euro-American settlement and 20th-century fire suppression practices drastically altered historic fire regimes, leading to excessive fuel accumulation and uncharacteristically severe wildfires in some areas and diminished flammability resulting from shifts to more fire-sensitive forest species in others. Prescribed fire is a valuable tool for fuel management and ecosystem restoration, but the practice is fraught with controversy and uncertainty. Here, we summarize fire use in the forests and woodlands of North America and the current state of the practice, and explore challenges associated with the use of prescribed fire. Although new scientific knowledge has reduced barriers to prescribed burning, societal aversion to risk often trumps known, long-term ecological benefits. Broader implementation of prescribed burning and strategic management of wildfires in fire-dependent ecosystems will require improved integration of science, policy, and management, and greater societal acceptance through education and public involvement in land-management issues.
Article
Coupled fire-atmospheric modeling tools are increasingly used to understand the complex and dynamic behavior of wildland fires. Multiple research tools linking combustion to fluid flow use Navier-Stokes numerical solutions coupled to a thermodynamic model to understand fire-atmospheric feedbacks, but these computational fluid dynamics approaches require high-performance computing resources. We present a new simulation tool called QUIC-Fire to rapidly solve these feedbacks by coupling the mature 3-D rapid wind solver QUIC-URB to a physics-based cellular automata fire spread model Fire-CA. QUIC-Fire uses 3-D fuels inputs similar to the CFD model FIRETEC, allowing this tool to simulate effects of fuel structure on local winds and fire behavior. Results comparing fire behavior metrics to the computational fluid dynamic model FIRETEC show strong agreement. QUIC-Fire is the first tool intended to provide an opportunity for prescribed fire planners to compare, evaluate, and design burn plans, including complex ignition patterns and coupled fire-atmospheric feedbacks.
Conference Paper
In this paper we propose a novel method for learning a Mahalanobis distance measure to be used in the KNN classification algorit hm. The algorithm directly maximizes a stochastic variant of the le ave-one-out KNN score on the training set. It can also learn a low-dimensional lin- ear embedding of labeled data that can be used for data visualization and fast classification. Unlike other methods, our classific ation model is non-parametric, making no assumptions about the shape of the class distributions or the boundaries between them. The performance of the method is demonstrated on several data sets, both for metric learning and linear dimensionality reduction.