COBRA-2000 : Lagrangian Regional-scale Experiment

Regional fluxes of CO2 have been derived from observations on towers and tethered balloons using a boundary-layer budget method. Unfortunately, significant errors that are difficult to detect and evaluate can be introduced into the budget by horizontal advection. Airmass-following (Lagrangian) experiments possess the potential for minimizing the advection term in the budget. We tested this Lagrangian approach in COBRA-2000 by conducting several flights sampling an air mass for 12 - 24 hours.

Trajectories of air parcels arriving at a receptor point have to be forecasted in advance to implement such a Lagrangian experiment. We developed the Stochastic Time-Inverted Lagrangian Transport (STILT) model, built upon the HYSPLIT source code [Draxler and Hess, 1998], to simulate the motions of tracer particles backwards in time, driven by forecasted meteorology at 40-km resolution from the Eta model at NCEP. Mean velocity components for each particle (of several thousand) are specified from the Eta run, and additional stochastic velocity components are computed from turbulence statistics derived by an algorithm based on surface fluxes of momentum and sensible heat from Eta. As particles in STILT are transported back in time from the receptor, they spread due to dispersion and wind separation. Their intersections with the ground define upstream regions that influence the receptor. Particle densities provide an estimate of the footprint (influence strength) for the observations.

Stochastic Time-Inverted Lagrangian Transport (STILT) modelSTILT was used as an operational planning tool in COBRA-2000, as illustrated in the figure to the right for flights over North Dakota. Forecasts for UT2100 on Aug 2nd, obtained from NCEP on Aug 1st, were used to transport particles backward in time 24 hours from the receptor in southern North Dakota. Flights were then planned for UT2100 on Aug 1st to sample the locations marked by the particles, using a sawtooth flight pattern along the principal component of the particle distribution (red line). The region in Canada marked in dark blue was sampled on Aug 1st, and the orange (morning) and light green (afternoon) areas were sampled on Aug 2nd. A biomass burning event upstream on July 30th distinctly labelled the airmass with elevated CO (see below).

CO: morning of 8-2-2000, cross-section in southern ND. CO: afternoon of 8-2-2000, cross-section in southern ND.

The cross-sections for CO from flights on the morning and afternoon of Aug 2nd are shown above. CO can be treated as a quasi-conserved tracer because emissions are small in southern North Dakota, and photochemical loss is negligible for an interval of one day. Both cross-sections exhibit elevated CO in the lower atmosphere, reflecting input from upstream forest fires east of Lake Winnipeg on July 30th. The distinctive CO label imparted by the fires confirmed that the aircraft revisited the same airmass, lending confidence to predictions from the operational flight-planning tool and to the Lagrangian experiment.

CO2: morning of 8-2-2000, cross-section in southern ND. CO2: afternoon of 8-2-2000, cross-section in southern ND.

Cross-sections for CO2 show marked decreases in the lower altitudes. The Citation sampled in the shallow morning mixed-layer on only a few occasions, and since these observations could not be confirmed to be representative, they were excluded from the analysis. Hence tracer concentrations in the morning cross-section reflect values in the residual mixed-layer from the previous afternoon. The CO2 decrease observed between the two cross-sections can thus be attributed to a 24-hr averaged flux.