Sensitivity Tests of the Temerin-Li Dst Model

The Dst historically has been used as a measure of the disturbance of the magnetosphere and an indicator of when a geomagnetic storm is occuring. The Temerin-Li [2002] (TL02) model is a semi-empirical Dst model which was trained on five years (1995 - 1999) of data. The TL02 model is the most successful Dst predictor in terms of prediction efficiency, with a reported prediction efficiency of greater then .90. The inputs into this model are the solar wind parameters, solar wind velocity, interplanetary magnetic field, and solar wind density. This model is based on the Burton et al. model, but has many parameters whose physical relavance are not well understood. To better understand the physical relevance of the empirical parameters, we use parameter sensitivity, impulse response, and driver sensitivity analysis. These techniques are used along with seasonal and diurnal analysis of the prediction error to extract a physical understanding of the most influential components of the model. By holding the input parameters constant and comparing this output and the output from sending pulses of delta t = 1 hour at regular intervals over a period of a year to the actual Dst, we can observe the diurnal and seasonal variations due to the individual parameters.