Comprehensive analysis of high-resolution dispersion simulations in urban area using the GRAMM/GRAL model

During the last decade, atmospheric measurement networks in urban areas have become important drivers for global pollution and greenhouse gas (GHG) mitigation. For city stakeholders to effectively plan GHG emissions mitigation measures and to monitor changes in emissions, GHG concentration data both, from measurements and simulations on high resolution, are required, but still lacking in most cities. The accurate simulation of high-resolution dispersion in urban areas enables the interpretation of concentration measurements as well as quantitative estimates of local emissions in an inverse modelling framework.

High-resolution dispersion simulations on neighborhood scale are generally computationally costly, preventing the analysis of long time periods. To overcome this limitation, we use the coupled GRAMM/GRAL model which is computationally efficient due to a ‘catalogue approach’. GRAMM/GRAL is composed of the mesoscale model GRAMM, solving the Reynolds Averaged Navier Stokes equations for an outer domain (resolution 100 m), and the computational fluid dynamics model GRAL for an inner domain (resolution 10 m). We run GRAMM for 1008 different wind situations differing in synoptic wind forcing and stability class. GRAL is initialized by the GRAMM fields and calculates wind fields taking the flow around buildings into account.  A time series of hourly, 10 m resolved wind fields and concentrations can be obtained by matching catalogued, simulated wind fields with measurement of a local wind measurement network (‘catalogue approach’).

Here, we evaluate the GRAMM/GRAL model in the urban area of Heidelberg for 12 months. 14 wind measurement stations within the inner GRAL domain enable a thorough evaluation of GRAL for yearly time periods and for a 12x12 km² area under challenging topography. Our evaluation also includes wind profile measurements (up to 200 m) from a LIDAR. We find good agreement between modelled and simulated wind directions. Wind speeds can be simulated with an overall root-mean square difference of about 1 ms^-1 and a mean bias of about -0.3 ms^-1. Measurement sites with poorer model representation are located in the forest or the outer domain with coarser resolution. We conclude that GRAMM/GRAL is capable of simulating high-resolution wind fields in urban areas of complex topography. We showcase first dispersion simulations for carbon dioxide in Heidelberg.