GRAL System - Mesoscale and Microscale Dispersion Modeling with GRAMM/GRAL
The GRALsystem is a coupled Eulerian (GRAMM and GRAL wind fields) and Lagrangian model (GRAL dispersion). GRAL calculates optionally with its own prognostic or diagnostic wind field model. Depending on the influence of topography and land use, the GRAL calculation area can be approached by the wind with free flow, comparable with MISKAM, or alternatively it can be initialized by the flow properties of GRAMM wind fields. GRAMM regards on a mesoscale area the influence of topography, land use and soil properties. GRAL adds in a smaller, nested area on the micro scale the influence of buildings and highly resolved terrain details. In prognostic mode it works acc. to VDI3783/9.
GRAMM and GRAL calculate as well in neutral atmosphere as in stable or instable atmosphere. As stable and instable atmosphere develop under varying border conditions (feedback), the wind fields can’t always reach steady-state. Therefore, the wind field development is stopped after a user defined, average simulation time and the results are taken as base for steady-state dispersion simulations, each representing average conditions over one hour.
In contrary to the AUSTAL2000/TALdia suite GRAL is validated for urban roads and street canyons and very flexible.
GRAL includes a special approach to regard the effects of tunnel portals and the meandering of wind, what is especially important for low wind speed simulation. This way GRAL is capable to combine the profits of stationary prognostic wind field modeling and the flexibility of Lagrange particle models.
SoundPLAN organizes the parameter settings and program calls of GRAMM Wind field calculation, GRAL Dispersion calculation and several pre processing executables under the proven SoundPLAN user interface. Some pre and post processing executables were replaced by own programming to add functions and to adjust the work flow to the SoundPLAN philosophy.
GRAL fulfills the requirements of the national Austrian regulation for dispersion calculation RVS 04.02.12 - Dispersion of Air Pollutants at transport routes and tunnel portals (original title: Ausbreitung von Luftschadstoffen an Verkehrswegen und Tunnelportalen, April 2014). Everybody who knows the Austrian topography can imagine the worth of such a recommendation.
Not only in Austria, also in Australia exists a recommendation for GRALsystem: GRAL is recommended by the National Health and Medical Research Council, Australian Government, as dispersion model for regulatory purposes for road tunnel portal emissions (NHMRC, 2008).
- The SoundPLAN “Meteorological Station” Library is a mighty tool to analyze, assess and edit meteorologucal data and Background measurement. It helps to calculate stability classes including a special GRAL derivate of the Bowen et al. SRDT method documented by US EPA. You can analyze meteorological data visually and adjust them to your task.
- Time variant emission cycles defined in SoundPLAN enlarge the GRAL standard to an accuracy of 8760 individual hours per year if needed, e.g. for weather dependent emission. However, business as usual it is much easier: You group first the sources to a few different source groups and use exemplary year histograms as dispersion pattern for a whole group. The post processing then refines the dispersion results of classified meteorology hour by hour to the accuracy of a meteorological time row, creating lots of detailed views on the results based on hourly emissions.
- SoundPLAN Geodatabase allows importing, digitizing and editing model data via several interfaces or just to use models prepared for SoundPLAN noise calculations. Background bitmaps can be used, e.g. geo-referenced aerial photos.
- GRAMM and GRAL were formerly offered as 32 bit version and as 64 bit version, but currently they are only maintained as 64 bit version.
- The SoundPLAN calculation kernel guides you through three run types: GRAMM wind field calculations, GRAL wind field and dispersion calculations and post processing calculations.
- Hit-rate maps and steady-state maps help to assess the results of the meso scale GRAMM wind field library, before a match-to-observation function allows to select the best matching GRAMM wind fields out of the pre-calculated meso scale wind field library, in order to initialize the micro scale GRAL calculations. This selection is performed by use of local measurement, documenting the differences between model and measurement in order to assess the quality of the selection. It is planned to optimize the selection by using multiple, synchronized meteo stations in very complex terrain.
- Different dispersion calculations can be performed using the same wind field library and each dispersion result can be used to feed several post processings, e.g. to compare emission scenarios without recalculating the whole GRAL dispersion.
- GRAMM and GRAL can use several processors parallel to calculate a wind field. SoundPLAN cares about the number of processors used. Using the additional module “Distributed Computing”, SoundPLAN also can distribute the calculation of whole wind fields parallel on several computers.
- The SoundPLAN Graphics module assists preparing outstanding presentation graphics. GRAMM and Gral results can be displayed for each single wind flow situation and concentration statistics. The number of available statistical maps supports fully the flexibility of the post processing options. Besides the usual display of mean, max and percentiles there are time slices available, source groups with different emission time histograms, smell assessment and more.
The model system was developed at the University of Graz, Austria, Institute for Internal Combustion Engines and Thermodynamics. It is also supported by the Air Quality Department Styria, Austria. GRAMM was developed at the Technical University of Graz and is highly respected in the scientific world. Both GRAMM and GRAL are validated by many international comparisons and studies and both are recommended by the Austrian authorities. The prognostic wind field calculations fulfill VDI 3783/7 (meso scale) and VDI 3783/9 (micro scale). 30 case studies compare GRAL concentration results with measurement and other models.
Model Type and Field of Application
GRAL dispersion uses a Lagrangian particle model, which is able to cope with vertical inhomogeneous turbulence and inhomogeneous 3D wind-fields. GRAL also calculates diagnostic or prognostic wind fields. In flat terrain it can work independently of meso scale GRAMM wind fields.It has special algorithms for treating dispersion in low wind speed conditions and for treating dispersion from tunnel portals.
The scope of GRAL dispersion calculations reaches from small urban quarters, where buildings must be taken into account, to whole cities or even larger areas, where topography and generalized land use parameters determine the wind flow. Inhomogeneous wind flow caused by topography and land use must be pre-calculated with GRAMM.
Mainly GRAL is used to calculate the impact of traffic exhaust, considering urban pollution background, but it is also used focusing on the whole urban air pollution impact situation e.g. to compare technical regulations for domestic heating.
GRAMM utilizes a prognostic approach to calculate mesoscale wind fields on large areas in complex terrain. The basin of Graz, on the southern side of the Alps, poses a very challenging evaluation area. The system has been used under different climate conditions, so it is a world wide solution. However, the washout by rain is not explicitly parameterized.
With growing complexity of terrain and land use structures, the importance of having a suitable local measurement also grows to fulfill the requirements of a sensible modeling. Transposing measurement from one location to another, what is often done for simple terrain, leads in complex terrain mostly to conflicts and bad results.
In complex terrain, the wind field calculation area must be much bigger than the dispersion calculation area, because the determining topographical input might be located far outside the area of interest - depending also on the place where wind measurement was made. However - once finished, the GRAMM wind field can be used for many more studies within this large area.
The already mentioned match-to-observation reorder function to select the best fitting wind fields according to a local measurement station is helpful, when you calculate for annual periods, because it requires a large pool of calculated wind fields. To get sensible results for only one or few wind flow situations, GRAMM has a reinit function which calculates up to 8 different wind field initializations to get the best fitting solution for an anemometer location within the calculation area. The difference between measured and calculated wind flow at the anemometer location is documented by SoundPLAN.
The GRAL single flow situation results are aggregated and evaluated using a mighty POSTPROCESSING routine. To optimize the relation between calculation time and the level of detail, GRAL operates on a classified meteorological list and annual mean emissions per source. Sources with similar hourly emission variation cycles must be marked by group numbers in the model setup. Those can then be addressed to combine emission variation cycles with an hourly meteorological time series during the post processing. This procedure is much quicker than calculating 8760 dispersion jobs per year with GRAL.
The post processing calculates maps for annual means, hourly or daily percentiles, hourly and daily maximum, user defined seasonal time slices and user defined day time slices. Also you can calculate how many hours, days or % exceed or undercut user defined threshold concentrations.
Additionally SoundPLAN implemented a smell assessment method defined in the German regulation “GIRL” to assess the smell of different source types by their intensity. This method counts first for each smell type the number of hours above a user defined threshold and then it weights the intensity by a user defined factor. For different smells arriving at the same time, the strongest smell determines the weight of this smell hour. The result is given as %.
The SoundPLAN version of this method which we implemented only in the GRAL post processing, is more flexible than the directive, which requires a fixed threshold and a small, limited set of weights. Alternative methods which apply the weight directly on the emissions, underestimate the total number of smell hours.
The NOx conversion is not included here but one step ahead, because all concentration dependent approaches are based on the total load. The total load can be result of an overlay of maps, therefore this conversion is placed in the SoundPLAN Graphics module.
Time Resolution and meteorology
GRAMM and GRAL calculations should use classified wind statistics. For urban projects those have typically 10° wind direction sectors, 7 Pasquill/Gifford classes or, as coarse approximation, 6 Klug/Manier classes. Wind speed classes should consider accordance to the selected stability classification.
Meteorological time series and hourly emission patterns, e.g. daily, weekly and monthly cycles are recommended for the post processing. Emission patterns can easily be defined in the time histogram library module.
The terrain following counting grid, which is used to detect the particle flow, is independent from the calculation grid which is used for the GRAL wind fields. So the wind field calculation can use a finer grid than the concentration maps show, in order to combine high precision of the wind field calculation with optimized significance of the concentration results.
Horizontal calculation grid resolution: typically, 2 x 2 to 5 x 5 m for GRAL when buildings must be resolved, 10 x 10 to 20 x 20 without buildings.
The mesoscale GRAMM wind fields should be rather coarse, typically from 100 x 100 m in extreme topography to 300 x 300 m if topography and land use can be sensibly resolved within this grid or when the calculation area is bigger than 5 x 5 km.
If buildings shall be included in GRAL, it can sometimes be sensible to use a higher grid resolution also for GRAMM.
A microscale resolution of terrain and land use structures with GRAMM is neither necessary nor sensible. Prominent topographic details like causeways and similar embankments are mostly better represented within the GRAL model.
The ratio of investigation area size, grid resolution and the number of source groups which can use different emission time histograms, is limited by RAM, disk space and the time needed to evaluate the produced amount of data. With currently available PCs or PC clusters, small cities can perhaps be evaluated in micro scale. For large city agglomerations instead, micro scale evaluations will be limited to hot spot areas.
Vertical resolution: Lower limit for GRAMM is typically 10 m for the ground layer, spreading in the following layers with a factor 1,2 to 1,35 depending on topography.
GRAL resolution is typically 1 m in presence of buildings, but without buildings it can be coarser. It is possible to apply a stretching factor for a successively increasing layer width, if sensible. The counting grid can be chosen in a different resolution to have on one hand a fine resolved wind field grid and on the other hand a significant amount of particle contacts per cell.
GRAMM and GRAL run as 64 bit programs, so the main limit for area size and resolution will not be work space but calculation time and the number of available processors and computers.
SoundPLAN sources are automatically transformed to GRAL Emission sources. Available are point, line, area and road sources. Additionally there is a tunnel portal source to calculate the tunnel exhaust, including buoyancy and stiffness impulse. All sources can have a vertical expansion, e.g. to create volume sources above roads to simulate initial pollutant dilution by vehicle turbulence. Excluded is the tunnel portal, which has a special geometry. For buoyant plume rise of point sources an effective source height can be defined according to Hurly (Details see GRAL documentation). Road emissions can be lifted above the road surface to simulate noise barrier effects in a coarse approach.
Outlook: The original GRAL kernel can be coupled to the emission model NEMO, developed also at the Technical University of Graz. NEMO uses the same database as HBEFA, but it is quicker up-to-date for all measurement of emission factors which is made at the Graz University. You can import NEMO results in the GEO-Database but a bi-directional communication between the programs is not yet realized.
In SoundPLAN pollutants can be freely defined as neutral gas. PM aerosols are also treated like a neutral gas for the dispersion. That means, deposition speed and sedimentation speed are not immediately regarded during the dispersion calculation. Deposition maps for PM will be derived instead from the ground near pollutant concentration. For roads and diffuse emissions within the urban background this procedure implements a slight overestimation and doesn't matter. Problems can occur for grain sizes > 10 µg. For high sources, especially for sources emitting coarser dust particles, this method might cause an underestimation.
In European law, for some planning, the expansion of areas with critical Nitrogen loads must be outlined in maps. Therefore, there is a discussion, if the deposited mass should be subtracted from concentrations during the transport to avoid overestimations of the critically loaded area size.
This argument sounds logical, but contains a high risk- Deposition speed is a very weak, assumed parameter. In order to avoid an underestimation of deposited mass, the deposition speed must be overestimated to be “on the safe side”. Unfortunately, reducing the concentration of transported air by an overestimated mass deposition will cause an underestimation of the area size which is impacted by a critical load.
However, deposition speed and sedimentation speed, regarded within the dispersion calculation, will be part of the future GRAL development, but not at highest priority.
GRAL can calculate without topography, if all objects are set to a terrain base of 0 m. It must be coupled to GRAMM wind fields if topography and/or land use shall be regarded. In contrary to GRAMM, which uses a terrain following grid, GRAL uses a Cartesian grid to calculate wind fields.
Buildings, Obstacles, Landuse
Buildings can be modeled in GRAL in the resolution of the GRAL calculation grid.
Regard of land use requires GRAMM wind fields. Even when land use can be entered in SoundPLAN as free shaped polygon, the resolution of land use is strictly bound to the GRAMM calculation grid.
GRAMM results can be displayed for each selected height above ground as terrain following maps or terrain following cross section maps.
- Single wind field display shows wind flow parameters like wind speed, wind direction, vertical inclination of wind direction as grid or isolines. Wind vectors can be added as grid symbol.
- Aggregated for the calculated time period, wind field statistic maps show grid or isoline maps of mean, min, max of wind speed, cases above or below a certain wind speed. Threshold exceeding or undercut is counted as hours or as percent.
All GRAMM results can also be displayed as cross section map.
GRAL allows the definition of up to 9 terrain following counting grid layers, which can be displayed as result.
- Single concentration situations: Maps of concentration, the desired height above ground and layer thickness must be defined before GRAL starts.
- Concentration statistics: Map contents must be ordered and calculated by the post processing. Besides the usual display of mean, max and percentiles there are time slices available, source groups with different emission time histograms, smell assessment and more.
If more than 1 counting grid layer was defined, all GRAL results can also be displayed as cross section map.
Grid operations to combine or modify maps by formulas are available for all concentration maps, e.g to derive daily and hourly statistics from annual means by predefined functions or user defined formulas (both using empirical derivations).
Tables for single point results are not yet supported, but all maps can be displayed with grid values to get representative single point results and to check at the same time for which area these points are representative.