ICARTT Analysis Products: Identification of Lagrangian cases

General
Possible Lagrangian cases were identified based on simulations with the Lagrangian particle dispersion model FLEXPART. During the forward simulation of the North American CO tracer (produce plots of the results with the interactive tool) all particle positions were stored every 2 hours during the period of the Lagrangian experiment, i.e., from 11 July to 4 August. From this particle position output, trajectories were derived that were checked whether they connect two or more ICARTT observations, taking into account the dispersion due to turbulence and convection. The method used was similar to the one described by Stohl et al. (Forecasting for a Lagrangian aircraft campaign. Atmos. Chem. Phys. 4, 11131124, 2004) and which was already used in the field for the flight planning.

Particle initialization and particle tracing
The NOAA P3, NASA DC8, UK Bae146 and DLR Falcon flights were subdivided into small segments (the same as those used for the backward simulations). Then all particles located within 30 kilometers (horizontally) and 200 meters (vertically) of the segment's center at the time of the measurement were identified and traced forward for a maximum of 10.5 days. Every 2 hours, the centroid position of all particles was calculated (representing a particlemean trajectory) along with the standard deviations of the distances both in the horizontal and vertical direction (this represents a measure of the dispersion of the plume). In addition, the mixing ratio of the North American CO tracer was determined at all particle positions and averaged over all particles.

Identification of possible Lagrangian cases
The entire plume trajectory for every upwind measurement was compared with all other later measurement positions (also including the R.H. Brown ship cruise, surface sites, and COBRA King Air flights) by interpolating plume trajectory positions to the measurement time. The following criteria were used for the matching: 1) The time difference between Lagrangian samplings must cover at least 12 hours and a distance of 800 km. 2) The tracer mixing ratio along the plume trajectory must not increase by more than 20%. This criterion shall exclude cases with significant uptake of fresh emissions. 3) The horizontal distance between the plume trajectory and a measurement point PLUS the plume particle standard deviation must be less than 18% of the distance travelled. This criterion limits both the spatial mismatch between the forward projected upwind and the downwind measurement and the horizontal dispersion (i.e., dilution) of the plume. 4) The vertical distance between the plume trajectory and a measurement point PLUS the plume particle standard deviation of the vertical coordinate must be less than 1000 m plus 0.02% of the horizontal distance travelled. This also limits both the spatial mismatch and the dilution.

Ranking of possible Lagrangian cases
Relative values obtained for criteria 3 and 4 are added and used to rank the cases. Only the minimum values for each downwind data set (i.e., each flight, each station, etc.) are used for the ranking and reported. Furthermore, each case is weighted by the number of measurement points from different data sets fulfilling the criteria. This weights Lagrangian chains (i.e., Lagrangian sampling by several platforms) higher than individual Lagrangian cases. Matches with COBRA King Air flights were not used for the weighting because of the limited chemical instrumentation on this platform. While several hundred cases were identified, they are not all truly independent, but often combine data from neighboring sections of the same flights.

Plot description
Access the Lagrangian cases list. Navigate forward and backward along the list using the "PREVIOUS CASE" and "NEXT CASE" buttons on the individual case webpages. The figures show the Lagrangian measurements as dots (color indicating altitude) with a number on top (the number refers to the list of measurements below the figure, giving the platform and time of measurement). The plume centroid trajectory is drawn as a line (inner line indicating altitude, outer line coding for the model mixing ratio scaled to a maximum of 300 ppb). Vertical black lines along the trajectory indicate the standard deviation of particle distances to the centroid location. Orange numbers on the trajectory indicate the day of the months plotted at 1 UTC.

Data file description
Data files are accessible from the individual case webpages. They consist of two parts: The first part lists the measurement times and some distance statistics, the second part lists the entire identified trajectory with positions every 2 hours. In the first part, the first line gives the case number and the number of downwind samplings. The second line gives information on the upwind sampling: Date, time, longitude, latitude, altitude, tracer mixing ratio, two internal numbers (disregard), name of the flight. Subsequent lines give information on downwind flights: Date, time, longitude, latitude, altitude, tracer mixing ratio at closest Lagrangian point, minimum horizontal distance, minimum vertical distance, standard deviation of particle horizontal distance from centroid at closest point, standard deviation of particle vertical distance from centroid at closest point, three internal numbers (disregard), name of the data set. The second part lists the entire plume centroid trajectory. Here, the first line gives the case number and the number of trajectory points. Subsequent lines give date and time, longitude, latitude, altitude, standard deviation of particle horizontal distance, standard deviation of particle vertical distance, tracer mixing ratio. There is also a master file with all Lagrangian cases in one file.

Confirmation of Lagrangian cases
This procedure relies entirely on the accuracy of the FLEXPART simulations. Therefore, the cases identified are certainly not all really Lagrangian (because of model errors and also because of the mismatches tolerated) and their Lagrangian nature must be confirmed using the chemistry measurement data. It is recommended to search data points in the vicinity of the times specified for upwind and downwind measurements to find the best match (or reject the Lagrangian case if no match can be found). Furthermore, in all cases there is significant dilution with surrounding air masses. Therefore, techniques should be employed that are less sensitive against dilution with surrounding air (e.g., correlation analysis of data around Lagrangian points). Furthermore, there was no ranking regarding the tracer mixing ratio, so some cases may be rather clean. Measurement data should be used to select the more polluted ones.

Caveats
The procedure works only if there is at least a minimum amount of North American tracer present. Thus, forest fire plumes without anthropogenic emission input from North America, or European plumes are not covered. For surface stations, positions have been used only every 3 hours. It is unlikely that good cases were missed due to that but the closest distance may actually be closer than reported.