HAL will be down for maintenance from Friday, June 10 at 4pm through Monday, June 13 at 9am. More information
Skip to Main content Skip to Navigation
Poster communications

Urban atmospheric dispersion modeling with artificial neural networks: using the Indianapolis data set

Abstract : Atmospheric dispersion modelling in complex environment is a tricky task because of the number of parameters involved. Indeed, dispersion in the free field is highly influenced by stability of the atmosphere. Moreover, when complex terrain is considered, the process of dispersion may be considered as non-linear due to the directivity of the flow with buildings. In order to evaluate the performance of atmospheric model, different data set are used. One of them is helpful to deal with dispersion around urban area. The Indianapolis dataset correspond to a 170 hours recording of SF6 concentration with 160 ground level monitors. The SF6 source was released from the top of an elevated stack inside the urban area. While Gaussian models are usually applied very fast, they present low level of performance. On the other hand, models from Computation Fluids Dynamics (CFD) are accurate but requires high expertise and important computer resources. In this paper, we investigate the potential of a neural network to predict concentrations in the urban field, using the Indianapolis experiments as a database for the training process.
Document type :
Poster communications
Complete list of metadata

Contributor : Administrateur Imt - Mines Alès Connect in order to contact the contributor
Submitted on : Monday, March 16, 2020 - 2:37:34 PM
Last modification on : Tuesday, April 19, 2022 - 10:12:29 AM


  • HAL Id : hal-02509109, version 1


Pierre Lauret, Frederic Heymes, Laurent Aprin. Urban atmospheric dispersion modeling with artificial neural networks: using the Indianapolis data set. ECCE 12 & ECAB 5 - 12th European Congress of Chemical Engineering & 5th European Congress of Applied Biotechnology, Sep 2019, Florence, Italy. 2019. ⟨hal-02509109⟩



Record views