What are the levels of uncertainty associated with back trajectory calculations in HYSPLIT? Some of the FAQ’s suggest that running the forward trajectory and backward trajectory simulators produce answers that will vary by approximately 1/2 the distance between the start and end points on average. Could this number be interpreted as the uncertainty in the numerical calculation? On top of that error, do you have an estimate for how far away from reality you expect the simulator to be on average?
The difference between the start position and end position of a forward-backward trajectory combination indeed represents the numerical error. However, more significant error is introduced because the gridded meteorological data file used for the computation is a discrete representation of a continuous function. Therefore, depending upon the the spatial and temporal complexity of the meterological fields, additional error will be introduced if the grid resolution is not sufficient. Typically a grid resolution of “x” can only represent wavelengths of “4x”. This error will be a function of the meteorological conditions. Then there is the implicit error introduced by the assumption that the trajectory (the integration of a single point in space and time) represents something more than its explcit definition. For instance, assuming that a single trajectory represents the pollutant flow from a source to a receptor, when in fact that situation must be represented by many trajectories to represent the effects of turbulence and temporal variations in the flow over the duration of emissions and sampling.