CRS4

An integrated system for the forecasting of wildfire behavior based on Cloud and Virtualization technologies

Antioco Vargiu, Marino Marrocu, Luca Massidda, Muriel Cabianca, Carlo Impagliazzo
Proceedings, Number 20150526 - may 2015
Télécharger la publication : Vargiu_An_integrated system_Alghero 20150526.ppt [8.7Mo]  
In this work we illustrate some results in the field of wildfire modeling, achieved in the project "Cloud for Remote Visualization", financed by the Autonomous Region of Sardinia with the call PIA 2010 and developed in collaboration with Nice srl. The project consisted in the development of a system based on Cloud and Virtualization technologies for the remote visualization and handling of large 3D datasets. The architecture has been tested by means of an integrated system for the forecasting of wildfire behavior, producing large datasets to be examined remotely. A web portal has been therefore developed. It offers a user friendly interface that allows an easy interaction with the complex HPC computing operations required for wildfire simulation as well as for the analysis and evaluation of the results. The service allows the remote 3D visualization of the results through an application running on specific virtual machines, taking advantage of the hardware acceleration. The simulation system implements a fire propagation algorithm based on the Fast Marching Method specifically developed for the project, with the Rate of Spread calculated on the basis of the classic Rothermel models. The weather conditions necessary to drive the wildfire model are produced daily by means of a high resolution weather forecast chain. The assessment of the types of fuel is obtained by recasting the land cover Globalcover-2009 in standard NFFL fuel models and a specific fuel model is used for the Mediterranean brush. The accuracy of the model is assured by the high resolution of the terrain and vegetation maps and by the high performance numerical solver adopted for the fire spread. A higher accuracy of the simulation is further obtained with a dedicated fluid dynamic solver that calculates a high resolution wind pattern. This module is based on the mass-consistent approximation, allowing for a quick but efficient downscaling of the wind, from typical weather scales to grid resolutions compatible with the fire forecast requirements. The solver, based on the CFD library OpenFOAM, runs on demand over a virtual machine. In order to test the forecast system, some relevant wildfires happened in Sardinia were simulated. The simulations show reasonable evolution of typical fires in the Mediterranean area. The system is open to many improvements both in terms of computing efficiency and as forecast quality. The first depends on the available computing power as the system scales well in a parallel environment. For the latter we can have improvements stepping up, for instance, the accuracy of the vegetation and the fuel modeling for the Rate Of Spread computations. On the other hand, is important to remark that the mass-consistent approximation we used for high resolution wind simulation is a CFD method can not handle nonlinear phenomena such as the turbulence and the interaction between atmospheric flow and fire.

Références BibTex

@Proceedings{VMMCI15,
  editor       = {Vargiu, A. and Marrocu, M. and Massidda, L. and Cabianca, M. and Impagliazzo, C.},
  title        = {An integrated system for the forecasting of wildfire behavior based on Cloud and Virtualization technologies},
  number       = {20150526},
  series       = {II International Conference on Fire Behaviour and Risk, Alghero},
  month        = {may},
  year         = {2015},
  editor       = {Luca Carongiu, Enrico Usai},
  organization = {CNR, Istituto di Biometeorologia, Università di Sassari, Dipartimento di Scienze della Natura e del Territorio,  Università di Sassari, Dipartimento di Scienze Economiche e Aziendali,  Centro Euro-Mediterraneo sui Cambiamenti Climatici},
  address      = {Sassari},
  keywords     = {wildfire behavior modeling, Rothermel's fire spread model, mass-consistent downscaling, cloud computing, remote visualization},
  url          = {http://publications.crs4.it/pubdocs/2015/VMMCI15},
}

Autres publications dans la base

» Antioco Vargiu
» Marino Marrocu
» Luca Massidda