Periodic Reporting for period 1 - 3DFIRELAB (3D Fire LABoratory)
Reporting period: 2020-10-01 to 2022-09-30
The project “3D Fire Laboratory” (3DFIRELAB) is a study on field scale forest fire radiative transfer. The project aims at developing monitoring methodology and modelling tools to better understand the limitation of Infra-Red (IR) fire imagery and improve their use in operational product such that fire front monitoring or fire emission estimation.
The project action lies between the two fire science communities of (i) combustion which has interest in upscaling effect of smaller flame scale fire dynamics to field scale front and (ii) atmospheric remote sensing which relies in part on empirical relationship and IR images to compute satellite product and model fire atmospheric disturbance.
One option to monitor fire at global scale is based on the Fire Radiative Power (FRP), a satellite active fire product. It is a measure of the instantaneous total energy released by the fire. It was shown on small scale experimental fire (~1m) to have a linear relationship with vegetation mass consumption. This made it very appealing to estimate global fire emission in real time. Furthermore, one of its formulation is only based on the measure of one single spectral band, making it easy to compute.
The objective of the project is to create a simulation tool that can model radiative transfer in a fire scene at a scale more relevant to satellite scale that is small experimental fire. Targeting field scale fire (~100m), two approaches are considered either only simulating the radiative transfer of an observed experimental fire and assessing the impact of IR camera viewing geometry, or setting up a standalone system than can model both the fire dynamics and IR images. The second approach, much more complex, has the potential to give better understanding of energy transfer in fire scene and implication in the FRP estimation.
Overall conclusions of 3DFIRELAB actions:
- The project was the opportunity to develop a system that can create connection between fire science communities, providing a methodology applicable to field scale simulated fire that can simulate IR images, a direct observable from experimental fire. The different fire (FDS, ForeFire), atmospheric (MesoNH) and radiative transfer (DART) models used in the project were in most of the cases pushed outside their zone of normal utilization. This fueled a lot of theoretical discussions to understand the validity of the result, in particular in the validation of DART. The scientific results of the project were disseminated both among the international and European fire science community, through written articles and oral presentations as well as the organization of a workshop.
- a 3DFireLab creation is also the development of a methodology and image processing algorithms that can extract the essential fire behavior metrics out of imagery from IR airborne handheld camera. Tasks of image orthorectification and fire front segmentation were automatized to process more that 3000 images collected during 4 burns of several hectares conducted in savannah type vegetation. The algorithms and the data are in the process to be published in open source.
- Finally the project provided training to the researcher in various ways (networking capacities, professional and management skills), helping his career prospects in the academia and beyond.
The project action lies between the two fire science communities of (i) combustion which has interest in upscaling effect of smaller flame scale fire dynamics to field scale front and (ii) atmospheric remote sensing which relies in part on empirical relationship and IR images to compute satellite product and model fire atmospheric disturbance.
One option to monitor fire at global scale is based on the Fire Radiative Power (FRP), a satellite active fire product. It is a measure of the instantaneous total energy released by the fire. It was shown on small scale experimental fire (~1m) to have a linear relationship with vegetation mass consumption. This made it very appealing to estimate global fire emission in real time. Furthermore, one of its formulation is only based on the measure of one single spectral band, making it easy to compute.
The objective of the project is to create a simulation tool that can model radiative transfer in a fire scene at a scale more relevant to satellite scale that is small experimental fire. Targeting field scale fire (~100m), two approaches are considered either only simulating the radiative transfer of an observed experimental fire and assessing the impact of IR camera viewing geometry, or setting up a standalone system than can model both the fire dynamics and IR images. The second approach, much more complex, has the potential to give better understanding of energy transfer in fire scene and implication in the FRP estimation.
Overall conclusions of 3DFIRELAB actions:
- The project was the opportunity to develop a system that can create connection between fire science communities, providing a methodology applicable to field scale simulated fire that can simulate IR images, a direct observable from experimental fire. The different fire (FDS, ForeFire), atmospheric (MesoNH) and radiative transfer (DART) models used in the project were in most of the cases pushed outside their zone of normal utilization. This fueled a lot of theoretical discussions to understand the validity of the result, in particular in the validation of DART. The scientific results of the project were disseminated both among the international and European fire science community, through written articles and oral presentations as well as the organization of a workshop.
- a 3DFireLab creation is also the development of a methodology and image processing algorithms that can extract the essential fire behavior metrics out of imagery from IR airborne handheld camera. Tasks of image orthorectification and fire front segmentation were automatized to process more that 3000 images collected during 4 burns of several hectares conducted in savannah type vegetation. The algorithms and the data are in the process to be published in open source.
- Finally the project provided training to the researcher in various ways (networking capacities, professional and management skills), helping his career prospects in the academia and beyond.
Throughout the project, the researcher has achieved:
- Simulation of IR synthetic images from field scale simulated fire. The creation of a standalone system that can simulate convective and radiative energy transfer and link them with direct observable.
- Simulation of a fire plume at 2m resolution forced from IR image observation of a 4ha experimental fire conducted in savanna vegetation.
- Development of a system to operate an Optris PI640 IR camera from a drone. The camera is controlled from a raspberry pluged on its own battery and can record IR images at 30Hz.
- Collaboration with Tjarda Roberts from CNRS on the use of drone in small scale outdoor fire to simultaneously monitor fire behavior and fire emission.
- Organization of an international workshop to emphasize the added value of high-resolution active fire remote sensing to support the current development of the community in fire modeling.
- A review article from the organized workshop is in preparation with the participation of the 17 attendees.
- Publication of an article on the orthorectification of IR airborne imagery.
- Submission of an article on Fire Front Segmentation and Rate of Spread Calculation. This manuscript is currently split in two for resubmission.
- Participation in several seminars and workshops.
- Dissemination of the results through invited academic talks conference and workshops (France, Portugal, On-line) and other outreach activities targeting students, fire practitioners and citizens.
- The creation of a website that report the advancement of the project.
- Release of the orthorectification algorithm OrthoIRCam under GPL License on github.
- Release on the Catalan dataverse repository of an IR image data set to support OrthoIRCam application.
- Simulation of IR synthetic images from field scale simulated fire. The creation of a standalone system that can simulate convective and radiative energy transfer and link them with direct observable.
- Simulation of a fire plume at 2m resolution forced from IR image observation of a 4ha experimental fire conducted in savanna vegetation.
- Development of a system to operate an Optris PI640 IR camera from a drone. The camera is controlled from a raspberry pluged on its own battery and can record IR images at 30Hz.
- Collaboration with Tjarda Roberts from CNRS on the use of drone in small scale outdoor fire to simultaneously monitor fire behavior and fire emission.
- Organization of an international workshop to emphasize the added value of high-resolution active fire remote sensing to support the current development of the community in fire modeling.
- A review article from the organized workshop is in preparation with the participation of the 17 attendees.
- Publication of an article on the orthorectification of IR airborne imagery.
- Submission of an article on Fire Front Segmentation and Rate of Spread Calculation. This manuscript is currently split in two for resubmission.
- Participation in several seminars and workshops.
- Dissemination of the results through invited academic talks conference and workshops (France, Portugal, On-line) and other outreach activities targeting students, fire practitioners and citizens.
- The creation of a website that report the advancement of the project.
- Release of the orthorectification algorithm OrthoIRCam under GPL License on github.
- Release on the Catalan dataverse repository of an IR image data set to support OrthoIRCam application.
The work performed on fire simulation within the 3DFIRELAB project emphasize on two aspects of fire mechanisms that were not very much tackled in the literature yet. First, the simulation of a 4ha savanna experimental burn forced from sensible heat flux computed from IR airborne observation shows that the cooling area can have impact on the fire front dynamics. Recent developments in fire-atmosphere coupled models mainly focused on effect of moisture or compressibility in the flame, but nothing was never considered on the effect of the cooling area. The inhomogeneity of the sensible heat fluxes due to the patchiness of the vegetation forms a lot of vortices otherwise not present in the traditional setup of constant cooling with set with e-folding time. In the back of the fire front, the forced sensible heat fluxes help creating stream-wise pair of vortices that were recently documented in the literature, but also formed much more complex structure ahead of the fire front, pushing, in the example presented in 3DFIRELAB, to the formation of a tornado. So far, I only looked at one fire, 3 others will be soon analyzed with the same approach.
The second achievement of 3DFIRELAB is the development of a stand-alone simulation tools capable to simulate Infra Red imagery from simulated fire. It is based on the coupling of the FDS physical-based fire model and the radiative transfer model DART. It provides a methodology to the large FDS community to improve model validation. And more importantly it creates a system that simulates Fire Radiative Product (FRP) from 3D realistic fire scene. FRP is an active fire satellite product which is routinely used in global fire monitoring system like Copernicus and listed as a fire disturbance Essential Climate Variable. Despite its massive use in operational product, the question of the impact of non-lambertian flame present in large outdoor fire on the FRP measure was never tackled. The system developed within 3DFireLab is now capable to simulate FRP from small field scale fire front (>30m) showing up to 50% increase in FRP for varying viewing geometry. Future work will consider fire size and fire behavior impacts on the FRP calculation.
The second achievement of 3DFIRELAB is the development of a stand-alone simulation tools capable to simulate Infra Red imagery from simulated fire. It is based on the coupling of the FDS physical-based fire model and the radiative transfer model DART. It provides a methodology to the large FDS community to improve model validation. And more importantly it creates a system that simulates Fire Radiative Product (FRP) from 3D realistic fire scene. FRP is an active fire satellite product which is routinely used in global fire monitoring system like Copernicus and listed as a fire disturbance Essential Climate Variable. Despite its massive use in operational product, the question of the impact of non-lambertian flame present in large outdoor fire on the FRP measure was never tackled. The system developed within 3DFireLab is now capable to simulate FRP from small field scale fire front (>30m) showing up to 50% increase in FRP for varying viewing geometry. Future work will consider fire size and fire behavior impacts on the FRP calculation.