Modélisation 3D du bilan radiatif des milieux urbains par inversion d'images satellites en cartes de réflectance et de température des matériaux urbains

Abstract : Optical remote-sensing imagery provide a unique and very needed information, but still a partial one, because only in the observation configuration of the satellite sensor (i.e. viewing direction and spectral bands), whereas Q* is an integrated quantity over all the directions and over the whole shortwave (Qsw*) and longwave (Qlw*) spectral domain. These integrations applied to satellite images are very complicated because of the complexity of the urban tri-dimensional (3D) architecture, and because of the urban materials temperature and optical properties spatial heterogeneity. Over the course of this PhD, an innovative approach has been conceived in order to achieve those integrations and thus obtain temporal series of Q* maps at the spatial resolution of the used satellite sensors (i.e. Sentinel-2, Landsat-8, etc.). This approach is using solely a 3D radiative transfer model, satellite images, and a geometrical urban database including the topology, the urban constructions (i.e. buildings, roads, etc.) and the vegetation (i.e. trees, gardens, etc.). Schematically speaking, the radiative transfer model DART (, developed at CESBIO, is used in inverse mode in order to transform satellite images into urban materials optical properties and temperature maps, and then in direct mode in order to compute radiative budget Q*Δλ maps for each spectral band of the used satellite sensor. Then, the spectral integral of those Q*Δλ maps leads to the desired Q* maps. Each temporal series of Qsw* maps is then generated efficiently from direct albedo maps (i.e. black sky albedo) and diffuse (i.e. white sky albedo) pre-computed using DART from the geometrical urban database of the considered city and optical properties derived from the closest satellite image. These maps are complemented by external thermal data for the computation of the temporal series. This method has been conceived and refined using 3 cities with very varying geometries and optical properties: London (United- Kingdom), Basel (Switzerland), and Heraklion (Greece). The H2020 project URBANFLUXES of the European Community used the simulated Q* maps in order to estimate the urban anthropogenic heat fluxes using the derivation of urban energy budget computed from satellite imagery. The precision of the developed method has been estimated using the relative error ER between the radiance images simulated by DART and measured by satellite sensors (ER<2% for any spectral band) and the relative error EQ* between Q* simulated and measured by flux towers. For the year 2016, |EQ*|< 4.5% for 321 Q* maps over Basel, and |EQ*|< 4.4% for 278 London Q* maps. This capacity of deriving from satellite imagery precis Q* maps is really promising in light of the always increasing availability of urban geometrical databases, of high resolution temporal series of satellite images, and of the improvement of 3D radiative transfer modeling.
Document type :
Complete list of metadatas

Cited literature [201 references]  Display  Hide  Download
Contributor : Abes Star <>
Submitted on : Tuesday, June 4, 2019 - 11:13:11 AM
Last modification on : Friday, January 10, 2020 - 9:08:31 PM


Version validated by the jury (STAR)


  • HAL Id : tel-02146724, version 1



Lucas Landier. Modélisation 3D du bilan radiatif des milieux urbains par inversion d'images satellites en cartes de réflectance et de température des matériaux urbains. Climatologie. Université Paul Sabatier - Toulouse III, 2018. Français. ⟨NNT : 2018TOU30139⟩. ⟨tel-02146724⟩



Record views


Files downloads