A Method of Evaluating Atmospheric Models Using Tracer Measurements

Abstract

The authors have developed a method that uses tracer measurements as the basis for comparing and evaluating wind fields. An important advantage of the method is that the wind fields are evaluated from the tracer measurements without introducing dispersion calculations. The method can be applied to wind fields predicted by different atmospheric models or to wind fields obtained from interpolation and extrapolation of measured data. The method uses a cost function to quantify the success of wind fields in representing tracer transport. A cost function, ?tracer potential,? is defined to account for the magnitude of the tracer concentration at the tracer receptors and the separation between each segment of a trajectory representing wind field transport and each of the tracer receptors. The tracer potential resembles a general expression for a physical potential because the success of a wind field trajectory is directly proportional to the magnitude of the tracer concentration and inversely proportional to its distance from this concentration. A reference tracer potential is required to evaluate the relative success of the wind fields and is defined by the initial location of any trajectory at the source. Then the method is used to calculate continuously the tracer potential along each trajectory as determined by the wind fields in time and space. Increased potential relative to the reference potential along the trajectory indicates good performance of the wind fields and vice versa. If there is sufficient spatial coverage of near and far receptors around the source, then the net tracer potential area can be used to infer the overall success of the wind fields. If there are mainly near-source receptors, then the positive tracer potential area should be used. If the vertical velocity of the wind fields is not available, then the success of the wind fields can be estimated from the vertically integrated area under the tracer potential curve. A trajectory with a maximum tracer potential is constructed for each daily tracer measurement, and this tracer potential is used to normalize the relative success of the wind fields in reproducing the transport of tracers. The method is not sensitive to the exact form of the cost function because a test with an inverse square root dependence in the cost function rather than an inverse linear distance dependence ranked the wind fields in the same order. The method requires sufficient spatial coverage of tracer receptors in the vicinity of a source and primarily gives credit to the wind fields that are able to approach areas with high tracer concentrations. The method can quantitatively determine which wind fields are best able to reproduce the main transport of tracers and can be used to determine the most successful wind fields to serve as a solid base for necessary improvement of dispersion models. It can also be used as a screening method prior to using dispersion models. Since the measured tracer concentrations are affected by both transport and dispersion, however, the method does not evaluate the capabilities of successful wind fields, as input to dispersion algorithms, to create tracer concentrations at receptors that are similar to measured ones. The tracer potential method has been applied to data from a comprehensive field program that included tracer measurements and was conducted in the Colorado River Valley area in the southwestern United States in 1992. Wind fields obtained from four atmospheric models as well as those derived from the wind profiler measurements were tested, and the results of their comparison are presented. Since data from the tracer experiment are publicly available, this developed method can be used to test other atmospheric models.