Characterising the vertical structure of buildings for use in atmospheric models

Abstract

<p>Urban schemes in weather and climate models often characterise urban structures in a grid cell using the mean building height and street width. This does not capture the sub-grid vertical variability that impacts fluxes. The vertical distribution of wall area and building heights are ideally required but are often unavailable in cities globally. In this work, building footprint and height data from six cities are used to parameterise the geometry with varying levels of detail of input data.</p><p>We conclude the vertical distribution of buildings can be parameterised using a function of mean building height and surface building plan area. Comparisons of the parameterised building plan area fraction with height to ‘true’ data (2 km x 2 km resolution) show 90% of the profiles have bias errors (BE) of < 0.03 (‘true’ values are: 0.05 – 0.55).</p><p>Building horizontal size (or effective building diameter, <em>D</em>) has a six-city mean of ~21 m. As <em>D</em> is impacted by normalised building edge length and building plan area, we use it to parameterise building edge length. The derived <em>D</em> parameterisations have normalised BE (nBE) < 16%, but without total wall area as an input the nBE increases to 26%.</p><p>The combined parameterisations are used with the radiative transfer model SPARTACUS-Urban to simulate total absorption of shortwave (SW) radiation and effective SW albedo. The latter is impacted 2-10% (cf. simulations using ‘true’ data). Larger errors occur when simulating  within-canyon absorption fluxes. Larger errors also occur when fewer morphology inputs are used, with total wall area having the most benefit.</p><p>We conclude urban vertical variability can be acceptably characterised for numerical weather prediction using three parameters: surface building plan area, mean building height, and effective building diameter.</p>