Advancing 3D Fuel Mapping for Wildfire Behaviour and Risk Mitigation Modelling
Lead Research Organisation:
Swansea University
Department Name: College of Science
Abstract
Wildfires are a natural phenomenon in many regions of the world (e.g. the boreal and temperate North America or the Mediterranean Basin) but, in others (e.g. Atlantic Europe), they are mostly human-caused. Irrespective of their origin, wildfires burn, on average, an area equivalent to about 20 times the size of the UK every year. When they burn through populated areas they can be deadly. For example, in 2018, they resulted in 100 deaths in Greece, 99 in Portugal, and 104 in California alone. In the UK, fires have to date rarely resulted in losses of life but, on average, ~£55M are spent annually in wildfire responses and they have threatened infrastructures and communities (e.g. several wildfires last summer led to evacuations). A combination of climate and land use changes is already increasing wildfire risk in many areas, both inside and outside the UK, and this trend is expected to worsen. In order to develop more effective tools for mitigating and fighting extreme wildfires, we need to advance our ability to understand, predict and, where possible, control fire behaviour.
In this project we aim to improve understanding and mitigation of wildland fire by advancing wildfire behaviour model capabilities through the development of new automated methods (algorithms) to implement, for the first time, ground-breaking real 3D fuel data into physics-based wildfire behaviour models. These models are the most advanced in terms of their ability to forecast fire behaviour, but they remain constrained by the lack of detailed fuel input information to work with (i.e. the amount and structure of live and dead vegetation susceptible to burn).
The advancement we aim to deliver will provide a step-change in physical fire modelling capabilities. The new algorithms will be implemented in the powerful fuel models FUEL3D and STANDFIRE, which provide fuels inputs for the physics-based fire behaviour models FIRETEC and WFDS. We will apply these to forest stands that typify some of the most common flammable conifer forests in the UK, NW Europe and North America. The algorithms produced will be made publicly available and, therefore, can be adapted and applied to many other forest types around the world.
Three-dimensional fuel datasets will be acquired in field campaigns using a range of state-of-the-art laser scanning (terrestrial, wearable and aerial UAV-based laser scanners) and 'Structure from Motion' methods, with traditional fuel inventory measurements being carried out for comparison and model validation. Our case studies will focus on conifer stands in England, Scotland, Wales and the US. In the UK, conifer forests comprise half of the UK's 3.2 Mill. ha of forested land, and they have the greatest potential for crown fires, which spread along treetops and are the most dangerous and challenging to fight. In the US, the work will include real forest fires, carried out for research purposes, which will provide valuable fire behaviour and fuel consumption datasets to validate the improved fuel and fire models.
Fire behaviour depends on weather, topography, and on the type and amount of vegetation fuels, with the latter being the only factor that can be meaningfully influenced through management efforts. By managing fuels, we can reduce the risk of extreme fire behaviour and its impacts. Our project provides a novel approach for designing and testing of 'virtual fuel treatments' aimed at decreasing fuel hazard and, thus, fire risk, under current and predicted future climatic and land use scenarios. The involvement of key UK end-users (Forestry Commission, Met Office, Natural Resources Wales and South Wales Fire & Rescue Service) as partners will maximise the applicability and impact of the project's outputs. The novel 3D fuel data and algorithms will also present a major advance for other forestry applications (e.g. forestry inventory, timber forecasting, forest carbon budgeting, ecosystem services assessment).
In this project we aim to improve understanding and mitigation of wildland fire by advancing wildfire behaviour model capabilities through the development of new automated methods (algorithms) to implement, for the first time, ground-breaking real 3D fuel data into physics-based wildfire behaviour models. These models are the most advanced in terms of their ability to forecast fire behaviour, but they remain constrained by the lack of detailed fuel input information to work with (i.e. the amount and structure of live and dead vegetation susceptible to burn).
The advancement we aim to deliver will provide a step-change in physical fire modelling capabilities. The new algorithms will be implemented in the powerful fuel models FUEL3D and STANDFIRE, which provide fuels inputs for the physics-based fire behaviour models FIRETEC and WFDS. We will apply these to forest stands that typify some of the most common flammable conifer forests in the UK, NW Europe and North America. The algorithms produced will be made publicly available and, therefore, can be adapted and applied to many other forest types around the world.
Three-dimensional fuel datasets will be acquired in field campaigns using a range of state-of-the-art laser scanning (terrestrial, wearable and aerial UAV-based laser scanners) and 'Structure from Motion' methods, with traditional fuel inventory measurements being carried out for comparison and model validation. Our case studies will focus on conifer stands in England, Scotland, Wales and the US. In the UK, conifer forests comprise half of the UK's 3.2 Mill. ha of forested land, and they have the greatest potential for crown fires, which spread along treetops and are the most dangerous and challenging to fight. In the US, the work will include real forest fires, carried out for research purposes, which will provide valuable fire behaviour and fuel consumption datasets to validate the improved fuel and fire models.
Fire behaviour depends on weather, topography, and on the type and amount of vegetation fuels, with the latter being the only factor that can be meaningfully influenced through management efforts. By managing fuels, we can reduce the risk of extreme fire behaviour and its impacts. Our project provides a novel approach for designing and testing of 'virtual fuel treatments' aimed at decreasing fuel hazard and, thus, fire risk, under current and predicted future climatic and land use scenarios. The involvement of key UK end-users (Forestry Commission, Met Office, Natural Resources Wales and South Wales Fire & Rescue Service) as partners will maximise the applicability and impact of the project's outputs. The novel 3D fuel data and algorithms will also present a major advance for other forestry applications (e.g. forestry inventory, timber forecasting, forest carbon budgeting, ecosystem services assessment).
Planned Impact
The ability to predict the behaviour and impacts of wildfires is essential to enable effective wildfire mitigation efforts. The need for further advances in these capabilities is obvious in fire-prone regions where wildfires frequently take a toll on both environmental and societal values, often including the tragic losses of lives. Many countries, such as the UK, where fire impacts are currently less extreme, are expected to see an increase in fire risk due to climate and land changes, yet their wildfire-fighting and mitigation capabilities have remained very limited. This project will therefore fill an important gap. It combines fundamental and applied research and maximises its impact by involving end-users from the outset as project partners, and by including effective pathways for delivering outcomes to the wider academic, end-user and general public communities.
Specifically, the main project beneficiaries are:
SCIENTIFIC AND ACADEMIC BENEFICIARIES: the project outcomes (algorithms, fuel and fire evaluation datasets, and fire risk modelled scenarios) will be freely available; published in golden open access publications and incorporated in the STANDFIRE modelling tool that is freely available at https://standfire.readthedocs.io (website maintained by the USFS). They will provide critical information to researchers in fuel characterization, fire behaviour and fire risk, and, whilst not an essential focus of this project, they will be also relevant to air quality researchers (fire smoke modelling) and to researchers in other areas of forestry and ecology where state-of-the-art 3D fuel information is of great value (e.g. forest carbon budgeting, forestry inventories, ecosystem services assessment).
FOREST MANAGERS, FIRE-FIGHTERS AND POLICY MAKERS: the project will advance tools for fire behaviour and risk modelling and, for our case study forests, will generate fire risk simulations for current and anticipated future climate and forest management scenarios. This is highly valuable for informing policies regarding forest management and fire-fighting in order to promote more effective control of fuel hazards. This is a 'key action' identified in the UK National Adaptation Program that informs the DEFRA's UK Climate Change Risk Assessment (https://assets.publishing.service.gov.uk/government/uploads/system/uploads/attachment_data/file/727252/national-adaptation-programme-2018.pdf).
To maximize impact for end-users, this project includes several of them as project partners (Forestry Commission, National Resource Wales, Met Office and South Wales Fire & Rescue Service). They will be directly involved in identifying the climate and forest management scenarios used in the 'Fire Risk Assessment Work Package' (WP6 in the Case for Support). They will also be part of the project Advisory group that will meet every 6 months. In addition, the delivered advances in the STANDFIRE and FUEL3D fuel models, and the consequent improvement of the outputs of the FIRETEC and WFDS fire models, will have internationally applicability as these are already used in US and parts of Europe to inform fire mitigation strategies and wider forest management practices (e.g. http://capsis.cirad.fr/capsis/help/fireparadox). STANDFIRE is open access and has a very friendly end-user interface. We have scheduled a workshop to introduce this tool to more end-users, such as the UK project partners.
PUBLIC ENGAGEMENT: The education and involvement of the wider society is critical to reduce fire-induced risks. To that end, we have planned several "Fire Awareness" outreach activities, some for the general public and others targeted specifically to youths. Targeting the young public is especially relevant in South Wales, where a large proportion of wildfires are started by youths. We will also take advantage of wider dissemination opportunities via general media and social media, where our team has an excellent track-record (see Pathways to Impact).
Specifically, the main project beneficiaries are:
SCIENTIFIC AND ACADEMIC BENEFICIARIES: the project outcomes (algorithms, fuel and fire evaluation datasets, and fire risk modelled scenarios) will be freely available; published in golden open access publications and incorporated in the STANDFIRE modelling tool that is freely available at https://standfire.readthedocs.io (website maintained by the USFS). They will provide critical information to researchers in fuel characterization, fire behaviour and fire risk, and, whilst not an essential focus of this project, they will be also relevant to air quality researchers (fire smoke modelling) and to researchers in other areas of forestry and ecology where state-of-the-art 3D fuel information is of great value (e.g. forest carbon budgeting, forestry inventories, ecosystem services assessment).
FOREST MANAGERS, FIRE-FIGHTERS AND POLICY MAKERS: the project will advance tools for fire behaviour and risk modelling and, for our case study forests, will generate fire risk simulations for current and anticipated future climate and forest management scenarios. This is highly valuable for informing policies regarding forest management and fire-fighting in order to promote more effective control of fuel hazards. This is a 'key action' identified in the UK National Adaptation Program that informs the DEFRA's UK Climate Change Risk Assessment (https://assets.publishing.service.gov.uk/government/uploads/system/uploads/attachment_data/file/727252/national-adaptation-programme-2018.pdf).
To maximize impact for end-users, this project includes several of them as project partners (Forestry Commission, National Resource Wales, Met Office and South Wales Fire & Rescue Service). They will be directly involved in identifying the climate and forest management scenarios used in the 'Fire Risk Assessment Work Package' (WP6 in the Case for Support). They will also be part of the project Advisory group that will meet every 6 months. In addition, the delivered advances in the STANDFIRE and FUEL3D fuel models, and the consequent improvement of the outputs of the FIRETEC and WFDS fire models, will have internationally applicability as these are already used in US and parts of Europe to inform fire mitigation strategies and wider forest management practices (e.g. http://capsis.cirad.fr/capsis/help/fireparadox). STANDFIRE is open access and has a very friendly end-user interface. We have scheduled a workshop to introduce this tool to more end-users, such as the UK project partners.
PUBLIC ENGAGEMENT: The education and involvement of the wider society is critical to reduce fire-induced risks. To that end, we have planned several "Fire Awareness" outreach activities, some for the general public and others targeted specifically to youths. Targeting the young public is especially relevant in South Wales, where a large proportion of wildfires are started by youths. We will also take advantage of wider dissemination opportunities via general media and social media, where our team has an excellent track-record (see Pathways to Impact).
Organisations
- Swansea University (Lead Research Organisation)
- European Cooperation in Science and Technology (COST) (Collaboration)
- Forestry Commission England (Project Partner)
- Natural Resources Wales (Project Partner)
- US Forest Service (Project Partner)
- South Wales Fire & Rescue Service (Project Partner)
- Forestry Commission Research Agency (Project Partner)
- Met Office (Project Partner)
Publications
Carlos Cabo
(2020)
3D forest fuel mapping for wildfire behaviour modelling
Jones M
(2022)
Global and Regional Trends and Drivers of Fire Under Climate Change
in Reviews of Geophysics
Manuel Oviedo De La Fuente
(2020)
Optimum scale selection for 3D point cloud classification through distance correlation function
Oviedo-De la Fuente M
(2021)
A Distance Correlation Approach for Optimum Multiscale Selection in 3D Point Cloud Classification
in Mathematics
Pereira-Obaya D
(2024)
A Handheld Laser-Scanning-Based Methodology for Monitoring Tree Growth in Chestnut Orchards.
in Sensors (Basel, Switzerland)
Prendes C
(2021)
An algorithm for the automatic parametrization of wood volume equations from Terrestrial Laser Scanning point clouds: application in Pinus pinaster
in GIScience & Remote Sensing
Prendes C
(2023)
Optimal bucking of stems from terrestrial laser scanning data to maximize forest value
in Scandinavian Journal of Forest Research
Roces-Díaz J
(2020)
Automatic Delineation of Forest Patches in Highly Fragmented Landscapes Using Coloured Point Clouds
in Forests
Description | We developed and implemented algorithms for point cloud classifications for vegetation 'fuels' (the part of the vegetation that burns during wildfires) using different approaches: heuristic, machine learning and deep learning. We have used so far three datasets from different data and forest types to train, test and compare the different approaches. We have published the basic principles of the machine learning and heuristic methods in 4 different papers (see publication). This is a significant advance over mapping and classifying fuels manually. It is more accurate and far less resource intensive. Mapping of fuels is relevant for assessing and forecasting wildfire risk and planning wildfire mitigation. In addition, the method is valuable for assessing aboveground biomass stock for application such as carbon stock assessment, timber valuation etc. |
Exploitation Route | The methodology and algorithms are available through our publications. |
Sectors | Agriculture Food and Drink Energy Environment Leisure Activities including Sports Recreation and Tourism Other |
Description | A plugin in the widely used open-access software CloudCompare has been developed. This allows 3D laser scanning point clouds taken in forests to be analysed automatically. It has already been applied by numerous users. |
First Year Of Impact | 2023 |
Sector | Digital/Communication/Information Technologies (including Software),Environment |
Impact Types | Societal Economic |
Description | Knowledge Generation Research Grants |
Amount | € 111,078 (EUR) |
Funding ID | PID2021-126790NB-I00 |
Organisation | Ministry of Science and Innovation (MICINN) |
Sector | Public |
Country | Spain |
Start | 01/2022 |
End | 12/2024 |
Description | Maria Zambrano |
Amount | € 190,944 (EUR) |
Funding ID | MU-21-UP2021-030 |
Organisation | University of Oviedo |
Sector | Academic/University |
Country | Spain |
Start | 01/2022 |
End | 12/2024 |
Description | Ramón y Cajal Fellowship |
Amount | € 458,600 (EUR) |
Funding ID | RYC2018- 025797-I |
Organisation | Ministry of Science and Innovation (MICINN) |
Sector | Public |
Country | Spain |
Start | 08/2020 |
End | 08/2025 |
Title | Fuel classification |
Description | Algorithm for the classification of ground-based or ground-based+aerial point clouds according to specific fuel types. Inputs: outputs from the Initial Forest Inventory algorithm, algorithm parameters Outputs: original point cloud with the fuel classification described above as a new attribute. Language: Python |
Type Of Material | Computer model/algorithm |
Year Produced | 2020 |
Provided To Others? | No |
Impact | The outputs of this algorithm are already used by the US partners in this project, and in others, for fuel data implementation in the fire-behaviour model FIRETEC. |
Title | Initial Forest Inventory |
Description | Algorithm for the initial processes for forest inventory: individual tree detection, initial stem model, point clustering, assignment of all points to the tree they belong to and computation of the distance of all points to their tree axis. Inputs: original point cloud (height normalized), algorithm parameters Outputs: original point cloud with the attributes described above Language: Python |
Type Of Material | Computer model/algorithm |
Year Produced | 2020 |
Provided To Others? | No |
Impact | The outputs of this algorithm are already used by the US partners in this project, and in others, for fuel data implementation in the fire-behaviour model FIRETEC. |
Title | Machine Learning classification model for fuel types |
Description | Machine Learning classification model for fuel types from ground-based point clouds. This model/algorithm computes the classification based on trainning data, and chooses the most suitable method and parameters (Support Vector Machines, Random Forest or Generalized linear models) |
Type Of Material | Computer model/algorithm |
Year Produced | 2021 |
Provided To Others? | No |
Impact | The use of ML techniques for point cloud classifications brings to our project the use of a different classification paradigm. We aim now at developing hybrid ML-heuristic models to overcome some of the drawbacks of each methodology. |
Title | Merging terrestrial/aerial point clouds |
Description | Algorithm that merges ground-based and aerial point clouds based on the position of targets/key points in forest plots. Inputs: original point clouds (aerial and ground-based), targets/key points position, algorithm parameters Outputs: original point clouds in the same reference system, and merged point cloud Language: Python |
Type Of Material | Computer model/algorithm |
Year Produced | 2020 |
Provided To Others? | No |
Impact | The outputs of this algorithm are already being used by the Forest Research partners for forest inventory. |
Description | Cost Action 3DForEcoTech CA-20118 |
Organisation | European Cooperation in Science and Technology (COST) |
Country | Belgium |
Sector | Public |
PI Contribution | Carlos Cabo is one of the proposers of the Cost Action. He is a Management Commitee member and he leads the Working Group 3 (WG3: 'Laser- and image-based point cloud processing'). The aim of WG3 is to gather compile, test and compare all relevant public algortithms that allow automatic characterisation of terrestrial point clouds in forest plots. The Cost Action started 6 months ago. So far, we have launched an initiative to obtain a preliminar list of target algorithms (https://forms.gle/t3AARdd4cwMzd6oq9), and we are in the process of selecting those that are relevant to our goal. |
Collaborator Contribution | Our partners have given us information about their algorithms, and will process some datasets that will provide a valid comparison with our implementations. |
Impact | List of relevant target algorithms. |
Start Year | 2021 |
Title | 3DFIN |
Description | Software for automatic forest mensuration and monitoring at plot/stand level from ground-based dense point clouds (terrestrial laser scanning, handheld laser scanning and terrestrial photogrammetry) |
Type Of Technology | Software |
Year Produced | 2023 |
Open Source License? | Yes |
Impact | This software is just being released now so no impacts yet. |
Description | Contribution to an outreach article by the Natural History Museum |
Form Of Engagement Activity | A press release, press conference or response to a media enquiry/interview |
Part Of Official Scheme? | No |
Geographic Reach | International |
Primary Audience | Media (as a channel to the public) |
Results and Impact | Contribution to an outreach article by the Natural History Museum explaining the causes of wildfires and links to climate change |
Year(s) Of Engagement Activity | 2020 |
URL | https://www.nhm.ac.uk/discover/does-climate-change-make-wildfires-worse.html |
Description | School visit (Spain) |
Form Of Engagement Activity | A talk or presentation |
Part Of Official Scheme? | No |
Geographic Reach | Local |
Primary Audience | Schools |
Results and Impact | Talk given by the PI, Cristina Santin, for the international day of Women in Science (11 February) in her former secondary school (Spain) via zoom. |
Year(s) Of Engagement Activity | 2021 |