You are here: Home Flexpart Model Information
Document Actions

Model Information

last modified 2011-09-12 14:36

Background information regarding the FLEXPART Atmospheric Transport Model


Provided here is fundamental information for the FLEXPART Backward and Forecast model products. Please scroll down for the FORECAST Model description.


Meteorological Fields Used

Two meteorological input data sets have been used here: 1x1 degree data on 26 pressure levels from the Global Forecast System (GFS) model of the National Center for Environmental Prediction (NCEP), and data from the ECMWF model. The ECMWF data has 91 model levels and was retrieved fully mass-consistently from the T799 spherical harmonics data at ECMWF. The gridded data has 1x1 degree resolution globally.

Model Used

All calculations have been done with the particle dispersion model FLEXPART. For emission input for carbon monoxide, nitrogen oxides and sulfur dioxide, the EDGAR version 3.2 emission inventory for the year 2000 (fast track) on a 1 x 1 degree grid is used outside North America. Over most of North America, the inventory of Frost and McKeen (2004) is used. This inventory has a resolution of 4 km and also includes a list of point sources. Previous experience with the 1995 inventory has shown that Asian emissions of CO are underestimated (probably by as much as a factor of 2 or more) in the EDGAR inventory, while American CO emissions may be overestimated.

Backward simulations

Backward simulations are done from along the flight or ship track, or from the location of a surface station. Whenever an aircraft changes its position by more than 0.15 degrees in either longitude or latitude, a backward simulation is initiated. Also, whenever it changes its altitude by 10 hPa, a new backward simulation is initiated. For the Ron H. Brown, backward simulations were initiated every hour, or whenever the ship changed its position by more than 0.1 degrees in either latitude or longitude. For surface stations, a constant time interval of 3 hours is chosen for model simulations.

Every simulation consists of 40.000 particles released in the volume of air sampled. The backward simulations are done with full turbulence and convection parameterizations. Their theory is described by Seibert and Frank (Source-receptor matrix calculation with a Lagrangian particle dispersion model in backward mode, Atmos. Chem. Phys. 4, 51-63, 2004), and an application to aircraft measurements was presented by Stohl et al. (A backward modeling study of intercontinental pollution transport using aircraft measurements, J. Geophys. Res., 108, 4370, doi:10.1029/2002JD002862, 2003). Output is produced every 24 hours (particle positions plus so-called emission sensitivities accumulated over the 24 hours, see below). The emission sensitivities are stored on a 3-d grid with three levels (0-100 m, 100-3000 m, and above). The horizontal resolution of the output grid is 1 x 1 degree globally, with a 0.5 x 0.5 degree resolution nest over the area of most interest.


Note: for some calculations only passive transport tracers are included

The model calculations were done assuming two ways of transport in the atmosphere: Firstly, passive transport without involving removal processes and, secondly, assuming aerosol-like removal by dry and wet deposition. The aerosol-like transport is treated assuming properties similar to sulfate. It will overestimate the removal of most aerosols because it is applied from the first moment of the emission, whereas in reality sulfate, for instance, is being formed in the atmosphere from sulfur dioxide which is less susceptible to removal. The concentrations of real atmospheric species and the areas influencing them will likely be between the values of the passive and the aerosol-like transport alternatives. Emission sensitivities from both transport alternatives are multiplied with emission fluxes for black carbon, carbon monoxide, nitrogen oxides and sulfur dioxide, giving a total of eight tracers available. Also the emission sensitivity information is displayed for both types of tracer. Comparison will show where dry and, especially, wet deposition has affected the transport.

Emission information

For emission input for black carbon, the emission inventory of Tami Bond has been used. For carbon monoxide, nitrogen oxides and sulfur dioxide, the EDGAR version 3.2 emission inventory for the year 2000 (fast track) on a 1 x 1 degree grid has been used outside North America and Europe. Over Europe, the EMEP emission inventory for the year 2005 with a resolution of 0.5 degree has been used. Over most of North America, the inventory of Frost and McKeen (2004) is used. This inventory has a resolution of 4 km and also includes a list of point sources. Previous experience with the 1995 inventory has shown that Asian emissions of CO are underestimated (probably by as much as a factor of 2 or more) in the EDGAR inventory, while American CO emissions may be overestimated.

Products available from backward simulations

Plots are shown for three plotting regions: Global, regional, and local, with the regional and local plotting regions adjusted to the area of main interest. You can always toggle between plotting regions. Products are organized station-wise on a monthly basis. Once you have selected one of the products (see below), you can enter at a particular day and time of the month and, from there, you can navigate backward and forward in time. You can also change the product displayed for the active time by a simple mouse-click, you can toggle between ECMWF and GFS data displays, or you can go back to the overview page to enter at a different time, or you can go back to the main page.

Retroplume summary

This is perhaps the most complex product and uses a technique described by Stohl et al. (A replacement for simple back trajectory calculations in the interpretation of atmospheric trace substance measurements, Atmos. Environ., 36, 4635-4648, 2002) to display 5-dimensional data. Every 24 hours, particle positions are assigned to one of 5 groups using a clustering algorithm. At the position of every cluster a circle is drawn with the circle's radius scaled with the number of particles the cluster represents (i.e., the fraction of sampled air for which it is representative). The color of the circle indicates the altitude, and the number on top gives the time backward in days. The retroplume's centroid is also displayed by a trajectory, but as plumes get complex back in time, the centroid may not be very representative of the true plume position. It takes some time to get acquainted, but once you know how it can be used, this product tells you where the air sampled was at what time and at what altitude, all in one plot. Also shown are time series of the mean altitude of the retroplume (and the five clusters, red circles in the time series, size again indicating the relative fraction of sampled air it represents), the fraction of particles in the boundary layer, and the fraction of particles in the stratosphere (2 pvu polewards from 30 degree, thermal tropopause in the tropics).

Emission sensitivity integrated over the entire atmospheric column

This product shows the vertically integrated emission sensitivity, which is proportional to the residence time of the particles over a unit area. It is recommended to inspect this product first, because it always shows the entire retroplume and gives the quickest impression where the air did come from (but without altitude information). The emission sensitivity is based on the assumption that transport alone occurs; it does not account for any removal processes, such as wet or dry deposition. The unit shown is nanoseconds times meters divided by kilograms. The numbers superimposed on the shading are the days back in time for the retroplume centroid (see above). They give an approximate indication of where the plume was at what time (but note that the centroids become poorly representative for the plume if the plume shape is too complex. Numbers typically become unrepresentative when they leave the main stream of particles (i.e., a well confined streamer of high values in the column emission sensitivity) or if there are multiple such streams.

You may notice that individual particle trajectories become visible as "lines" of low values of the emission sensitivity. This is due to the logarithmic scale used and typically occurs far backward in time when particle trajectories have already diverged strongly and the 40.000 particles used are not many enough to fully characterize the retroplume's complexity. Also note that low values of the emission sensitivity often can be found appearantly "downwind" of the measurement location. This normally is due to particles having circled the globe.

Footprint emission sensitivity

Same as above, but averaged over the lowest 100 m instead of vertically integrated. As anthropogenic emissions are mostly located at the surface, this gives an indication where emissions were likely taken up. The unit shown is nanoseconds divided by kilograms.

CO, NO2, and SO2 source contributions

This is the product between the emission sensitivity and the anthropogenic emission flux (in kilograms per square meter and second) taken from the inventories. The result is an emission contribution in ppb per square meter. If the emission contribution is integrated over the earth's surface, a "tracer" mixing ratio at the sampling location is obtained. It is also reported on the plot and, furthermore, contributions from different continents are listed separately. These mixing ratios are quantitatively comparable to the measurements under the assumption that the species is conserved (no chemistry, no deposition).

Importantly, source contributions are integrated either for the global output domain at 1 degree resolution, or over the nested regional domain at 0.25 degree resolution. Emissions from outside the domain are not accounted for in the latter. Also, the higher resolution of the nest is not accounted for in the global sum, such that occasionally contributions from the nested region can be higher than for the entire globe.

Emission tracer time series

These plots show time series of the above tracers constructed from the backward simulations for the entire month, displayed seperately for total anthropogenic, Asian, North American, and European pollution.

Biomass burning CO contributions

Hot spot locations are obtained daily from measurements made with MODIS onboard the Aqua and Terra satellites and processed using the MOD14 (MYD14) algorithm described by Louis Giglio (MODIS Collection 4 Active Fire Product User's Guide). The data are downloaded from a server at the University of Maryland in ASCII format and only those hot spot detections with a confidence of 75% or greater are used. Fire hot spots are shown also in the column-integrated and footprint emission sensitivity maps. They are compared to a landuse inventory with 1-km resolution and if they are on forested land, they are marked as red dots (overlaid over larger black dots), for all other landuses as black dots only. The hot spots are shown on the footprint emission sensitivity map only in grid cells where the DAILY footprint emission sensitivity on the very day of the hot spot identification is above 0.005 ps/kg. They are shown on the column-integrated emission sensitivity map only in grid cells where the DAILY column-integrated emission sensitivity on the very day of the hot spot identification is above 8 ns m /kg. The emissions are estimated assuming an area burned of 180 ha/fire detection. They depend on a parameterization based on biomass available to burning, fraction actually burned, and emission factors, all dependent on landuse.

If you have any questions about the Backward modelling products please contact John Burkhart or Andreas Stohl.


General Info

Five-day forecasts are provided to support scientific field campaigns. Normally, they are updated once a day, but during large experiments the system can be run four times a day. If you would like to have the products available more often than once a day during your field mission, please contact Andreas Stohl (Tel: +47 6389 8035) or Sabine Eckhardt (Tel: +47 6389 8187) to discuss possibilities.

Meteorological Fields

The meteorological fields (geopotential height at 500 hPa, surface pressure, equivalent potential temperature, CAPE, and vertical velocity at 500 hPa) are extracted from the Global Forecast System (GFS) data of the National Centre for Environmental Prediction (NCEP)


The trajectory forecasts are calculated with the trajectory model FLEXTRA based on Global Forecast System (GFS) data of the National Centre for Environmental Prediction (NCEP). 8-day backward and 4-day forward trajectory forecasts are provided for two locations. Trajectories are started every 500 m in the vertical and every 3 hours.

For the four-day warm conveyor belt trajectories, trajectories are started at 500 m in the domain 110 deg W to 50 deg W and 20 deg N to 60 deg N. Only those trajectories are displayed that ascent more than 5000 m and travel north-eastwards within the four days.

Emission Tracer Forecasts

Emission tracer forecasts are calculated with the particle dispersion model FLEXPART based on Global Forecast System (GFS) data of the National Centre for Environmental Prediction (NCEP). Tracer masses are carried by particles following trajectories calculated using the GFS winds and stochastic components for turbulence and convection. Tracer forecasts are run separately for anthropogenic emissions from Asia, Europe and North America. For each of these forecasts, three tracers are available: carbon monoxide, nitrogen oxides (expressed as NO2), and sulfur dioxide (including direct emissions of sulfate). These species are run as passive tracers for a duration of 20 days, after which tracer particles are dropped from the simulation. In addition, a biomass burning CO tracer is available; the emission algorithm ingests actual information on MODIS fire detections, landuse information, and emission factors.

As the emission basis, the EDGAR version 3.2 fast track inventory for the year 2000 with a resolution of 1 degree is used, excecpt for most of North America where the inventory of Frost and McKeen (2004) is used. This inventory is based on the U.S. EPA NEI-99 inventory (National Emissions Inventory, base year 1999, version 3) and has an original resolution of 4 km, plus point sources. For initializing FLEXPART, the high resolution has been kept for high-emission grid cells and strong point sources. Weaker sources were aggregated to coarser resolution. For Mexico City, an inventory provided by Jerome Fast was used, which has a resolution of 4.5 km. Again, the high resolution was kept for high-emission grid cells.

The different regional tracers are available for the following computational domains:
70 W - 74 E, 30 N - 90 N at a resolution of 0.6 deg.
0 E - 30 E, 65 N - 90 N at a resolution of 0.25 deg.

FLEXPART forecast system essentials

Special Volcano Tracer

Two emission tracers carried for 20 days:
VO-TR – purely passive (#1)
VO-AER – susceptible to dry and wet deposition (#2)

Standard Tracers

Four anthropogenic emission tracers carried for 20 days:
CO – purely passive
NOx – purely passive
SO4 – susceptible to dry and wet deposition
BC (Black carbon) – susceptible to dry and wet deposition
Two biomass burning emission tracers:
CO – purely passive
BC (Black carbon) – susceptible to dry and wet deposition
The removal of BC and oxidized sulfur is likely to be overestimated, since BC is assumed to be hydrophilic from right after the emission, sulfur is assumed to be emitted as sulfate.

The anthropogenic tracers also carry regional information: European, North American, Asian origin.

Input data

0.5°x0.5° resolution GFS data for the real-time analysis and up to 3 hour forecast, 1°x1° resolution GFS data for the forecast; the forecasts strongly benefit from the high-resolution 0.5°x0.5° analyses!

26 vertical levels, every 3 hours

Output resolution

1°x1° globally, 0.25°x0.25° over Europe (10°W-40°E, 40°N-90°N)

16 vertical levels, every 3 hours

Update frequency and timing

4 times daily; forecasts will be started ca. 4 hours after analysis time and will be available ca. 8 hours after analysis time, i.e., forecasts based on 0-UTC meteorological analysis will be available at about 8 UTC (±1 hour, depending on the tracer).

A note about the pole

Since grid cells become very small near the pole, the particle counting statistics (which depend on the number of particles residing in a grid cell) become worse towards the pole, especially when concentrations are low. This leads to somewhat noisy concentration fields just around the pole. However, the transport of the particles itself is not affected by this – FLEXPART does not have any numerical problems near the poles. Just disregard that the output is somewhat more noisy around the pole, which is just due to the extremely high resolution of the output grid around the pole.

Quasi-Lagrangian flight opportunities

IMPORTANT NOTE: this product is not active at this time

Quasi-Lagrangian flight opportunities into the Mexico City plume are calculated for the aircraft involved in the MIRAGE and IMPACT campaigns. Currently, these are the NCAR C-130, the NASA DC-8, and the DLR Falcon. A paper describing the forecasts for such a Lagrangian campaign in detail has recently been published in ACP.

The orange dots in the figures show the assumed bases of three aircraft and orange circles indicate their ranges of operation. The centroid locations of air parcels selected as Lagrangian opportunities are presented as dots superimposed on maps of the total Mexico City CO tracer columns. The dots are colored according to the actual CO tracer mixing ratio in the air parcel using the same color scale as used for the total columns, but scaled to a maximum value of 249 ppb. Note, therefore, that the colors of the dots do not match with the colors of the background contours, which show total columns rather than mixing ratios. The number drawn on top of each dot indicates the air parcel's centroid altitude in kilometers. The five best Lagrangian opportunities are drawn as bigger circles surrounded by a white ring (sometimes only few are visible which cover the others). The trajectories of the 50 best cases are also shown as black lines.

If you have any questions about the GFS modelling products please contact Sabine Eckhardt or Andreas Stohl.

Personal tools