The Morning Inversion Near the Ground and Its Daytime Transition at Two Rural Sites in the Carolinas

Abstract

Six years of tower data from two dissimilar sites in the eastern piedmont of the Carolinas are analyzed to yield a selective climatology of the lower portion of the morning inversion. Its transition to daytime conditions is then described and statistically modeled. Both sites are in clearings surrounded by forest, but one site is in a valley by a lake, while the other, 175 km to the north, is on a low hilltop. Measurements of wind speed and direction, the standard deviation of wind direction, dew point, and temperature at 11 m, temperature difference (?T) between 11 and 60 m, plus solar radiation, were analyzed for an 8-h period starting from three hours before local sunrise each day for both locations. Results show that predawn inversions characterize over 70% of the data and strong inversions of over 5°C per 100 m in the tower layer characterize 30% of the mornings at the hilltop site. At the valley site, strong inversions are less common, probably because of the proximity of the lake. There is a correlation of 0.71 in daily site-to-site ?T at dawn. This suggests strong overall synoptic control of the local inversion frequency. The transition to well-mixed conditions after sunrise depends chiefly on ?T prior to sunrise. Analysis of mean trends in variables during the transition shows it is a remarkably well-ordered process. The time from sunrise to a mean isothermal state (between 11 and 60 m only) takes about 1 to 2 h. Daily transition is predicted by a linear regression scheme based on predawn conditions and developed and tested separately at each site. Chief predictors are inversion intensity, dew point and 60 m wind speed. For cloudy mornings the rms error for the prediction time from sunrise to mean isothermal conditions is 0.3 h. For days with variable cloudiness, a rather unspectacular R2 value of 0.3 to 0.4 is, nevertheless, statistically significant. A similarity in models at both sites is noted. In cloudless conditions the models are, in fact, nearly interchangeable.