DETERIORATING AIR OR IMPROVING MEASUREMENTS - ON INTERPRETING CONCATENATE TIME-SERIES

Measurements reported by Eldred and Cahill [1994] provide an extended record of fine-particle sulfur concentrations at Shenandoah National P ark. Begun in 1982, the measurements were upgraded in 1986 to improve time resolution and again in 1988 to improve sensitivity and particle size discrimination. Standard hypothesis rejection tests indicate a st atistically significant upward trend in measured concentrations from 1 982 to 1992. However, similar tests also indicate statistically signif icant differences between measurements made by the different methods, even when the apparent trend is accounted for. Does the trend in measu red concentrations reflect increasing atmospheric concentrations, or i s it an artifact of evolving measurement methods? This paper frames th e question in rudimentary Bayesian terms and shows that one's interpre tation of the trend is sensitive to one's prior confidence in the meas urements.