Article Forest Fire Forecasting Using Fuzzy Logic Models Àngela Nebot * and Francisco Mugica   Citation: Nebot, À.; Mugica, F. Forest Fire Forecasting Using Fuzzy Logic Models. Forests 2021, 12, 1005. https://doi.org/10.3390/f12081005 Academic Editors: Michele Salis, Grazia Pellizzaro, Bachisio Arca, Pierpaolo Duce, Donatella Spano, Costantino Sirca and Valentina Bacciu Received: 7 June 2021 Accepted: 27 July 2021 Published: 29 July 2021 Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affil- iations. Copyright: © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/). Soft Computing Research Group at Intelligent Data Science and Artificial Intelligence Research Center, Universitat Politènica de Catalunya, Jordi Girona Salgado 1-3, 08034 Barcelona, Spain; fmugica@cs.upc.edu * Correspondence: angela@cs.upc.edu; Tel.: +34-934137783 Abstract: In this study, we explored hybrid fuzzy logic modelling techniques to predict the burned area of forest fires. Fast detection is crucial for successful firefighting, and a model with an accurate prediction ability is extremely useful for optimizing fire management. Fuzzy Inductive Reasoning (FIR) and the Adaptive Neuro-Fuzzy Inference System (ANFIS) are two powerful fuzzy techniques for modelling burned areas of forests in Portugal. The results obtained from them were compared with those of other artificial intelligence techniques applied to the same datasets found in the literature. Keywords: hybrid fuzzy techniques; FIR; ANFIS; forest fire; burned areas prediction 1. Introduction As discussed in [1], weather and climate are the most important factors influencing fire activity, and they are changing due to human industry. In the near future with a warmer climate, we expect more severe fire weather, more burned area, more ignitions and a longer fire season. It is possible to see these effects right now, for example: Australian fires since September 2019 that have burned at least 17.9 million acres, destroyed more than 3000 homes, and killed at least 28 people. The Climate Atlas of Canada [2], clearly explains this vicious cycle connecting forest fires and climate change. All efforts to reduce global warming help prevent forest fires, which, in turn, mitigates climate change. In this context, any effort to improve forest fire management is of special relevance and utility. In this regard, accurate predictions of burned areas offer useful knowledge for management decisions and resource planning. It is possible to find in the literature a considerable number of research papers that deal with predicting burned areas using different artificial intelligence and machine-learning models. Several of these, mainly based on neural networks, were developed in [3–5], and their usefulness for predicting burned areas were studied and compared. Random forest, radial basis functions, genetic algorithms and support vector machines were also used by several authors for the task at hand [6–9]. In [9], a wildfire probability prediction method employing Bayesian networks and fuzzy logic was studied and evaluated through a case study in Australia. Other machine-learning methodologies, such as ensemble learning, were also addressed, of which random forest approximations delivered the best results [10]. Another interesting ensemble approach for improving the predictive accuracy of forest fire can be found in [11]. The authors combined a locally weighted learning algorithm with the Cascade Generalization, Bagging, Decorate, and Dagging ensemble learning techniques and applied it to a region of Vietnam. An investigation focused on the Monte Carlo heuristic search algorithm was conducted to predict the spread of forest fires [12]. All of these papers used real data from different parts of the world to train and validate their models. In this paper we addressed the challenge of modelling forest fires by means of hybrid machine-learning techniques based on fuzzy logic to predict the areas forest fires will burn. We studied two fuzzy-based learning approaches: Fuzzy Inductive Reasoning (FIR) and the Adaptive Neuro Fuzzy Inference System (ANFIS) because they have been shown Forests 2021, 12, 1005. https://doi.org/10.3390/f12081005 https://www.mdpi.com/journal/forests