High-resolution meteorological simulations in Heidelberg using GRAMM/GRAL model

Abstract

<p>More than 60% of global greenhouse gases are produced in urban areas. Urban areas therefore exhibit an immense mitigation potential, which needs to be fully exploited to limit climate change.  In order to monitor the effectiveness of mitigation measures, local decision makers require sub-urban data on the spatio-temporal distribution of greenhouse gases. The GRAMM/GRAL model is capable of calculating high-resolution (10m) wind fields over long time periods by using a “catalogue approach”. The model is therefore well suited to support mitigation efforts in cities.</p><p>GRAMM/GRAL is composed of the mesoscale model GRAMM and a coupled computational fluid dynamics (CFD) model GRAL. GRAMM calculates meteorological wind fields by solving the Reynolds-Averaged-Navier-Stokes (RANS) equation. In the catalogue approach, the model GRAMM calculates about 1000 wind fields with 100 m resolution each with a different set of atmospheric stabilities, wind speeds and directions. The CFD model GRAL uses these wind fields as input at the boundaries and calculates higher resolution (10m) wind fields taking the flow around buildings into account. Passive tracers may be released within the GRAL model to simulate their dispersion using a Lagrangian particle dispersion approach. A time series of wind fields and concentrations can be obtained by matching measured and simulated wind fields. This matching procedure saves computational costs and therefore enables the analysis of longer time periods.</p><p>In this study, we characterize the GRAMM/GRAL model performance in Heidelberg and compare modelled and measured wind fields in an urban setting for a period of three months. In general, we find a good agreement between modelled and simulated wind direction. Wind speeds can be simulated with a root-mean square difference of about 1.0 m/s and a mean bias of about 0.6 m/s. We find that the number of wind stations influences the overall model performance, which is in accordance to Berchet et al. (2017).</p><p>We further present an outlook on possible set-ups of an inversion scheme to estimate greenhouse gas fluxes from a hypothetical measurement network. To this end, we utilize the high-resolution model GRAMM/GRAL to simulate CO<sub>2</sub> concentration in the urban atmosphere and plan to approximate CO<sub>2</sub> fluxes using regularized least-square approaches as well as machine-learning methods. We discuss remaining challenges such as background CO<sub>2</sub> and biogenic CO<sub>2</sub> fluxes.</p>