Dynamic and Robust Wildfire Risk Prediction System: An Unsupervised Approach
Mahsa Salehi*, IBM Australia; Laura Rusu, IBM Research; Timothy Lynar, IBM Research; Anna Phan, IBM Research
Ability to predict the risk of damaging events (e.g. wildﬁres) is crucial in helping emergency services in their decision-making processes, to mitigate and reduce the impact of such events. Today, wildﬁre rating systems have been in operation extensively in many countries around the world to estimate the danger of wildﬁres. In this paper we propose a data-driven approach to predict wildﬁre risk using weather data. We show how we address the inherent challenge arising due to the temporal dynamicity of weather data. Weather observations naturally change in time, with ﬁner-scale variation (e.g. stationary day or night) or large variations (non-stationary day or night), and this determines a temporal variation of the predicted wildﬁre danger.
We show how our dynamic wildﬁre danger prediction model addresses the aforementioned challenge using context-based anomaly detection techniques. We call our predictive model a Context-Based Fire Risk (CBFR) model. The advantage of our model is that it maintains multiple historical models for diﬀerent temporal variations (e.g. day versus night), and uses ensemble learning techniques to predict wildﬁre risk with high accuracy. In addition, it is completely un-supervised and does not rely on expert knowledge, which makes it ﬂexible and easily applied to any region of interest. Our CBFR model is also scalable and can potentially be parallelised to speed up computation. We have considered multiple wildﬁre locations in the Blue Mountains, Australia as a case study, and compared the results of our system with the existing well-established Australian wildﬁre rating system. The experimental results show that our predictive model has a substantially higher accuracy in predicting wild-ﬁre risk, which makes it an eﬀective model to supplement the operational Australian wildﬁre rating system.
Filed under: Classification