Analysis of Seasat laser range data over long arcs has identified geomagnetic modelling within atmospheric density models as a significant source of error. In particular, a major deficiency is the absence or mis-modelling of persistence of a geomagnetic event at heights near 800km. A logical step is to refine persistence and the geomagnetic model parameters within the least-squares differential correction procedure of the orbital refinement. To investigate the feasibility of this approach, persistence has been studied and parameters adjusted within the J71 atmospheric model and its variant due to Slowey, JS84. In J71 two parameters were derived by combination of 7 or 8 Seasat arcs. No advantage was found by refining the 3 extra parameters within JS84. The validity of the solution has been tested on 17 Starlette arcs for which the level of improvement is significant, particularly during levels of high geomagnetic activity. This is the first occasion that atmospheric density parameters have been derived within the orbital analysis data-reduction procedure and paves the way for more extensive analyses in the future.