The first estimate of the universal form of the atmospheric temperature spectrum for locally stationary conditions was obtained with a statistical accuracy exceeding 5%. The universal form of the temperature spectrum permitted the accurate estimation of the statistics of the locally- stationary inner scale, structure constant, energy dissipation rate, and rate of dissipation of temperature variance. All these parameters have a log-normal distribution. The accuracy of the locally-stationary universal spectrum and estimates of the parameters of the spectrum were determined by a Monte-Carlo simulation. These simulations indicate the feasibility of 4% accuracy for inner scale estimates and 8% accuracy for estimates of the structure constants using 1 second of data. The average temperature spectrum is not a universal function and depends on meteorological conditions. The fluctuations in the locally stationary parameters produce a bias (which depends on experimental conditions) in the estimation of the universal Obukhov-Corrsin constants and the Kolmogorov constant. The variations in inner scale and level of turbulence are large (a factor of 10 for level of turbulence and a factor of 4 for inner scale over a few minutes). These variations may have pronounced effects on imaging systems and boundary layer mechanisms. A rugged laser instrument can be constructed to accurately measure many of the important boundary layer parameters.
