Statistical Analysis of the USU Lidar Data Set with Reference to Mesospheric Solar Response and Cooling Rate Calculation, with Analysis of Statistical Issues Affecting the Regression Coefficients.

摘要

Though the least squares technique has many advantages, its possible limitations as applied in the atmospheric sciences have not yet been fully explored in the literature. The assumption that the atmosphere responds either in phase or out of phase to the solar input is ubiquitous. However, our analysis found this assumption to be incorrect. If not properly addressed, the possible consequences are bias in the linear trend coefficient and attenuation of the solar response coefficient.;Using USU Rayleigh lidar temperature data, we found a significant phase offset to the solar input in the temperatures that varies ±5 years depending on altitude. In addition to introducing a phase offset into the linear regression model, we argue that separating what we identify as the solar-noise is to be preferred because (1) the solar-noise can contain important physical information, (2) its omission could lead to spurious conclusions about the significance of the solar-proxy coefficient, and (3) its omission could also bias the solar proxy coefficient.;We also argue that the Mt. Pinatubo eruption caused a positive temperature perturbation in our early mesopause temperatures, exerting leverage on the linear trend coefficient. In the upper mesosphere, we found a linear cooling trend of greater than −1.5 K/year, which is possibly exaggerated because of leverage from the earlier temperatures and/or collinearity. In the middle mesosphere we found a cooling trend of −1 K/year to near zero.;We use the autocorrelation coefficient of the model residuals as a physical parameter. The autocorrelation can provide information about how strongly current temperatures are affected by prior temperatures or how quickly a physical process is occurring.;The amplitudes and phases of the annual oscillation in our data compare favorably with those from the OHP and CEL French lidars, as well has the HALOE satellite instrument measurements. The semiannual climatology from the USU temperatures is similar to that from the HALOE temperatures. We also found that our semiannual and annual amplitudes and phases compare favorably with those from the HALOE, OHP, and CPC data.;The computer code used to generate the author's figures included in this dissertation is given in Appendix I.