Importance-driven approach for reducing urban radiative exchange computations

In the context of large scale urban heat transfer simulation, the prediction of radiative flux at short and long wave spectra is a step necessary to obtain accurate results. From a computational perspective, this task is expensive because realistic conditions require calculations in many sensors, considering multiple radiation bounces, and evaluating many hundred daylighting conditions. Radiosity-based approaches are adequate methods for processing the large number of diffuse surfaces that are usually present in city models. However, the high memory consumption of these algorithms turns them inefficient for handling big geometries, and therefore ray tracing techniques are commonly used. In this article we present a study on using the importance concept to improve the performance of radiosity calculations at the urban scale. The algorithm is able to consider diffuse and specular materials, and it proves to be a viable alternative to ray tracing. Since most of the information contained in big city models is not needed for simulating a selected zone of interest, the computational requirements can be reduced drastically. Several experiments are conducted to test the approach, and promising results are reported.