Predicting wildfires: Spatio-temporal Modeling of Wildfires in Maule, Chile, and Analysis of the Risk Communication Tool Botón Rojo
(2025) In Master's Thesis in Mathematical Sciences FMSM01 20251Mathematical Statistics
- Abstract
- As wildfires are becoming more frequent and severe due to the effects of climate change, there is a need for effective risk management and understanding of the drivers behind wildfire ignition and spread. This thesis tackles wildfire risk management from two angles- quantitatively and qualitatively. Primarily it presents spatio-temporal models of the wildfire ignition intensity and the size of burned area by wildfires in the region of Maule, Chile, treating these as a marked point process. The modeling is done using wildfire data from Corporación Nacional Forestal (CONAF). The wildfire ignition points are modeled as a log-Gaussian Cox Process and the burned area is modeled using a Gamma likelihood. Two different temporal resolutions are... (More)
- As wildfires are becoming more frequent and severe due to the effects of climate change, there is a need for effective risk management and understanding of the drivers behind wildfire ignition and spread. This thesis tackles wildfire risk management from two angles- quantitatively and qualitatively. Primarily it presents spatio-temporal models of the wildfire ignition intensity and the size of burned area by wildfires in the region of Maule, Chile, treating these as a marked point process. The modeling is done using wildfire data from Corporación Nacional Forestal (CONAF). The wildfire ignition points are modeled as a log-Gaussian Cox Process and the burned area is modeled using a Gamma likelihood. Two different temporal resolutions are used, one yearly using data from 2002-2021, and one monthly using data from 2020-2021. The model
formulations include sociological, meteorological, and ecological aspects as covariates, and also components to distinguish seasonality effects and long-term trends. The results for the wildfire ignition point pattern show no increased ignition probability for forest plantations compared to natural forest formations, as well as an increasing trend over the years 2002-2021. Furthermore this thesis explores CONAF’s risk communication tool Botón Rojo, which is used in Chilean wildfire risk management, primarily for resource allocation. Through interviews with CONAF employees the tools’ usage and effectiveness is analysed. The analysis concludes that Botón Rojo is an effective and helpful tool for wildfire risk management within CONAF. (Less) - Abstract (Swedish)
- I takt med att skogsbränder blir vanligare och allvarligare till följd av klimatförändringarnas effekter, ökar behovet av effektiv riskhantering och förståelse för de bakomliggande faktorerna till uppkomst och spridning av skogsbränder. Denna rapport angriper riskhantering av skogsbränder ur två perspektiv – kvantitativt och kvalitativt. I huvudsak presenteras spatio-temporala modeller för intensiteten av skogsbränder samt dessa brända yta i regionen Maule i Chile, där dessa tillsammans behandlas som en märkt punktprocess. Modelleringen baseras på skogsbrandsdata från Corporación Nacional Forestal (CONAF). Antändningspunkterna för skogsbränder modelleras som en log-Gaussisk Cox-process och det brända områdets storlek modelleras med hjälp... (More)
- I takt med att skogsbränder blir vanligare och allvarligare till följd av klimatförändringarnas effekter, ökar behovet av effektiv riskhantering och förståelse för de bakomliggande faktorerna till uppkomst och spridning av skogsbränder. Denna rapport angriper riskhantering av skogsbränder ur två perspektiv – kvantitativt och kvalitativt. I huvudsak presenteras spatio-temporala modeller för intensiteten av skogsbränder samt dessa brända yta i regionen Maule i Chile, där dessa tillsammans behandlas som en märkt punktprocess. Modelleringen baseras på skogsbrandsdata från Corporación Nacional Forestal (CONAF). Antändningspunkterna för skogsbränder modelleras som en log-Gaussisk Cox-process och det brända områdets storlek modelleras med hjälp av Gamma-sannolikhetsfördelningar. Två olika tidsupplösningar används: en på årsbasis med data från 2002–2021 och en på månadsbasis med data från 2020–2021. Modellerna inkluderar sociologiska, meteorologiska och ekologiska aspekter som kovariater, samt komponenter som fångar säsongsvariationer och långsiktiga trender. Resultaten för modellen för skogsbrandsuppkomst visar ingen ökad sannolikhet för uppkomst i skogsplantager jämfört med naturlig skog, samt en ökande trend över perioden 2002–2021. Vidare undersöks CONAF:s verktyg för riskkommunikation, Botón Rojo, som används i Chiles riskhantering av skogsbränder, framför allt för resursfördelning. Genom intervjuer med anställda på CONAF analyseras verktygets användning och effektivitet. Analysen visar att Botón Rojo är ett effektivt och användbart verktyg för skogsbrandsriskhantering inom CONAF. (Less)
- Popular Abstract
- Forecasting Flames: Understanding Wildfire Risk in Chile
Wildfires in Chile are becoming more frequent and more severe. We set out to understand what factors increase the risk of wildfires and why these fires spread.
The 2024 wildfires in Chile resulted in the deaths of 137 and extensive losses of property. Megafires like this will become increasingly frequent and increasingly severe due to climate change and land use practices, and the wildfire season in Chile has been observed to get longer every year.
Quenching a raging wildfire is resource-demanding, difficult and dangerous. This is especially true in Chile where resource allocation often takes a long time due to the country´s elongated shape and mountainous lands. In dry... (More) - Forecasting Flames: Understanding Wildfire Risk in Chile
Wildfires in Chile are becoming more frequent and more severe. We set out to understand what factors increase the risk of wildfires and why these fires spread.
The 2024 wildfires in Chile resulted in the deaths of 137 and extensive losses of property. Megafires like this will become increasingly frequent and increasingly severe due to climate change and land use practices, and the wildfire season in Chile has been observed to get longer every year.
Quenching a raging wildfire is resource-demanding, difficult and dangerous. This is especially true in Chile where resource allocation often takes a long time due to the country´s elongated shape and mountainous lands. In dry conditions, the time it takes to move firefighters is highly important as fires can spread very quickly. However, nipping a fire in the bud is easier. For this reason, a great deal of effort is being put into making reliable predictions of the wildfire risk in Chile.
In our thesis, we modeled wildfires in the Chilean region of Maule using data from the Chilean National Forest agency (CONAF). The methodology builds on an advanced statistical modeling package, and we built models with some chosen components we thought could explain the wildfire data well. We found that wildfire occurrence increased between the years 2002 and 2021, and that wildfires became larger on average during that same period. One investigated factor to affect wildfire risk was land cover type, where agricultural land was found to have the highest wildfire risk. Surprisingly, forest plantations did not differ in risk from natural forests, despite expectations.
We also saw that the wildfire ignition probability has a very clear season that follows a yearly cycle. Wildfire size was less predictable and not strongly linked to the factors we tested. We could see that the partly meteorological Forest Fire Weather Index (FWI), that we used as a factor, affected the wildfire ignition probability and together with the season captures the high intensity of the wildfire season. The relationship between proximity to urban areas and higher probability of ignition was not well explained by our models, and would be interesting to investigate further.
The results in this study can be used in order to understand what has an impact on wildfire occurrence and wildfire spread. The methodology of the study has not been used very widely before, and can also be used for further research on the matter. This thesis contributes to the understanding of wildfire risk and supports future prevention strategies. (Less)
Please use this url to cite or link to this publication:
http://lup.lub.lu.se/student-papers/record/9198781
- author
- Duvsten Östin, Hanna LU and Gasslander, Tilda LU
- supervisor
- organization
- course
- FMSM01 20251
- year
- 2025
- type
- H2 - Master's Degree (Two Years)
- subject
- keywords
- Wildfires, Spatio-temporal, Log-Gaussian Cox process, Marked point process, Risk communication, inlabru
- publication/series
- Master's Thesis in Mathematical Sciences
- report number
- LUTFMS--3534-2025
- ISSN
- 1404-6342
- other publication id
- 2025:E86
- language
- English
- id
- 9198781
- date added to LUP
- 2025-06-24 10:22:19
- date last changed
- 2025-06-24 10:22:19
@misc{9198781,
abstract = {{As wildfires are becoming more frequent and severe due to the effects of climate change, there is a need for effective risk management and understanding of the drivers behind wildfire ignition and spread. This thesis tackles wildfire risk management from two angles- quantitatively and qualitatively. Primarily it presents spatio-temporal models of the wildfire ignition intensity and the size of burned area by wildfires in the region of Maule, Chile, treating these as a marked point process. The modeling is done using wildfire data from Corporación Nacional Forestal (CONAF). The wildfire ignition points are modeled as a log-Gaussian Cox Process and the burned area is modeled using a Gamma likelihood. Two different temporal resolutions are used, one yearly using data from 2002-2021, and one monthly using data from 2020-2021. The model
formulations include sociological, meteorological, and ecological aspects as covariates, and also components to distinguish seasonality effects and long-term trends. The results for the wildfire ignition point pattern show no increased ignition probability for forest plantations compared to natural forest formations, as well as an increasing trend over the years 2002-2021. Furthermore this thesis explores CONAF’s risk communication tool Botón Rojo, which is used in Chilean wildfire risk management, primarily for resource allocation. Through interviews with CONAF employees the tools’ usage and effectiveness is analysed. The analysis concludes that Botón Rojo is an effective and helpful tool for wildfire risk management within CONAF.}},
author = {{Duvsten Östin, Hanna and Gasslander, Tilda}},
issn = {{1404-6342}},
language = {{eng}},
note = {{Student Paper}},
series = {{Master's Thesis in Mathematical Sciences}},
title = {{Predicting wildfires: Spatio-temporal Modeling of Wildfires in Maule, Chile, and Analysis of the Risk Communication Tool Botón Rojo}},
year = {{2025}},
}