TRACE A trajectory intercomparison: 1. Effects of different input analyses

Abstract

We address the problem of air mass trajectory uncertainty through an intercomparison of trajectories computed from operational meteorological analyses from the region and time period of the NASA/GTE/TRACE A experiment. This paper examines the trajectory uncertainty that results from the input meteorological analyses. We first compare the National Meteorological Center (NMC) and European Centre for Medium‐Range Forecasts (ECMWF) meteorological analyses to an independent set of observations, the dropsondes released from the NASA DC‐8 over the South Atlantic during TRACE A. We also compare the gridded wind and temperature fields with selected rawinsonde data that entered the analyses. These comparisons show that the ECMWF fields are marginally better than the ones from NMC, particularly in the tropical regions of the southern hemisphere. The NMC analyses are marginally better in the midlatitude westerlies in some cases. In general, slightly more confidence can be placed in trajectories computed with ECMWF data over the TRACE A region, based on our comparisons of the analyses with observations. Second, we compute 5‐day back trajectories with three different models from a grid of points over the South Atlantic and adjacent portions of South America and Africa as well as on the track of TRACE A flight 15 over the South Atlantic. When using the Goddard Space Flight Center isentropic model, horizontal separations of greater than 1000 km occur for about 50% of the points when trajectories run with the ECMWF and NMC analyses are compared. Greater sensitivity to the input analysis differences is noted when trajectories are computed with the FSU kinematic model (separations exceed 1000 km for 75% of the points). The problem of meteorological uncertainty should be addressed with two approaches. There are large differences between both sets of analyses and the TRACE A soundings; this is also likely to be the case in other remote regions. Therefore we recommend that a test set of trajectories be computed with both sets of input data to quantify the uncertainty due to analysis differences. In addition, clusters of trajectories about the points of interest should be run to assess the uncertainty due to wind shear. These recommendations are applicable to any region of the globe with sparse observations. The companion paper [Fuelberg et al., this issue, part 2] addresses uncertainties due to trajectory technique.