A Method for Monitoring the Heat Flux from an Urban District with a Single Infrared Remote Sensor

Abstract

The proposed methodology relies on the modelling capabilities of the thermo-radiative model Solene to simulate the heat and radiation energy exchanges between an actual urban district and the atmosphere. It is based on the comparison of the simulated upward infrared and sensible heat flux diurnal cycles that may be measured by elevated sensors above the three-dimensional scene, as a function of sensor position: the heat flux is a function of an equivalent surface temperature given by the infrared sensor and an equivalent heat transfer coefficient deduced from Solene simulations with the actual geometry. The method is tested against measurements obtained in the city centre of Toulouse, France during an experimental campaign in 2004–2005. To improve the computation of the heat exchanges between air and building surfaces a new algorithm is first implemented, based on an empirical model of the wind distribution within street canyons. This improvement is assessed by a direct comparison of the simulated brightness surface temperatures of the Toulouse city centre to measurements obtained with an airborne infrared sensor. The optimization of the infrared remote sensor position is finally analyzed as a function of its height above the mean roof level: it allows evaluation of the heat flux from an urban district when the three different classes of surfaces (roofs, walls, grounds) have similar contributions to the infrared flux towards the sensor, and to the heat flux into the atmosphere.

Appendices

Appendix 1: Measurement Description

1.1 CAPITOUL Central Site (Masson et al. 2008; Pigeon et al. 2008)

A 30-m telescopic mast was deployed on the roof of the Monoprix department store, located approximately 20 m above the ground, such that the top of the mast was at 47.5 m above the ground. This provided data continuously from mid-February 2004 to early March 2005. The downward and upward global solar and infrared fluxes were obtained with $180^{\circ }$ aperture radiometric sensors. The turbulent fluxes of sensible heat, latent heat (water vapour) and carbon dioxide were measured using the eddy-covariance method. The wind speed and direction, the air temperature profile along the mast and above the Rue de la Pomme, and the precipitation were also recorded. Some surface radiative temperatures were also monitored by permanent radiometers, and measured by mobile teams with hand-held radiometers during intensive observation periods. To ease the comparison with the simulation results the experimental data, measured every 1 min, are averaged here for a same timestep of 15 min.

1.2 Airborne Infrared Measurements (Lagouarde and Irvine 2008; Lagouarde et al. 2010; Hénon et al. 2012b)

Two TIR cameras were placed aboard a small twin-engine aircraft with backward inclinations of $9.5^{\circ }$ and $50^{\circ }$. Only the images from the $50^{\circ }$ slanted FLIR SC2000 are used here, equipped with a $24 \times 18^{\circ }$ lens and with a band-pass of 7.5–13 $\upmu$m using a filter with a cut-off at 13 ${\upmu }$ m to reduce atmospheric absorption. The aircraft speed was 70 m s$^{-1}$, and the images were acquired at 4.3 Hz. The flight altitude was 650 m above sea level, about 460 m above the ground for the centre of Toulouse. The spatial resolution (pixel size at ground) ranged between 0.8 and 2.3 m for zenith viewing angles between 44 and $62^{\circ }$; the sizes of the images are $240\times 320$ pixels. The LOWTRAN 7 model was used for atmospheric corrections with Météo France radiosoundings performed simultaneously with the flights.

Appendix 2: Simulation Parameters

1.1 District Simulations

The simulated scene is a 135-m square, centred on the position of the meteorological mast. The 3D geometric model was constructed from the GIS data of the city of Toulouse. The mean building height is 20 m and the average built density is 0.68. Most of the buildings are residential and/or commercial. The façades consist mostly of red bricks, with some rare variations including a few concrete façades. Glass surfaces cover between 10 and 30 % of the vertical surfaces. Most of the roofs are made of Roman tiles and have a slope of less than $20^{\circ }$ but some are horizontal, made of concrete and often include metallic appliances.

The geometry is modelled with an unstructured grid composed of 32,669 triangular meshes, each about 2 m$^{2}$ in area. Due to the relative homogeneity of material composition the set of parameters is unique for each of the three surface classes (roof, wall, pavement) and the Solene parameters (Table 3) are given “effective” values either selected from the original analysis of CAPITOUL data by Pigeon et al. (2008) or resulting from the optimization exercises of Hénon (2008) to represent at best the radiative and heat transfers at the district scale (Hénon et al. 2012a). The deep ground layer is set to the mean air temperature during the previous week (20 $^{\circ }$C for July 15, 1.9 $^{\circ }$C for February 25). This condition is also applied to the building internal temperature in summer; in winter all buildings are assumed to be heated to $T_{\mathrm{int}} = 19$ $^{\circ }$C. The sky model of Perez et al. (1993) is used with a clear-sky condition. To take into account the effects of inertia and to overcome the effects of initialization, the simulations are started 24 h in advance, with all initial surface temperatures equal to the measured air temperature. The simulation timestep is 15 min and with a 1.73 GHz mono-processor computer each simulated hour requires 180 min of computation.

Table 3 Thermo-radiative parameters of the materials for Solene simulations ($h_{\mathrm{c\,int}}$ is the heat transfer coefficient for the convection between the building inside air and wall layer 2)

1.2 Street-Canyon Simulations

The street-canyon models are 20 m in height, 70 m in length, 17 m in width for Alsace-Lorraine and 11 m for La Pomme. The computational grid is composed of 1-m$^{2}$ rectangular meshes. The street is long enough for the end-effects to be negligible, both for radiation and convection, on the analyzed surface temperatures at mid-length of the street. Each 2-day simulation requires about 30 min of computation with our 1.73-GHz monoprocessor computer. Here the roofs are assumed flat and horizontal, such that the computations are direct without multiple reflections of radiation.