The space-time correlation structure of convective rainstorms in the Lagrangian reference frame

This dissertation investigates the statistical nature of subsynoptic, convective rainfall using data from the Convective Initiation Downburst Experiment conducted in Denver, Colorado in 1987. Forty-six PAM network gages, mean spacing of 10.6 km, and 31 FLOWS network gages, mean spacing of 2.4 km were utilized. Both had a measurement time interval of 1 minute.; Thirteen rainfall events are isolated from the raw data and used in this study. The storms included raincells, cell clusters and small mesoscale areas. The average storm speed from the FLOWS and PAM events is 76.5 km/hr and 56.4 km/hr, respectively. The predominant direction of storm movement is at an azimuth between 90{dollar}spcirc{dollar} and 180{dollar}spcirc.{dollar}; The structure of rainstorms relative to the fixed Eulerian reference frame are examined using spatial cross-correlation. Curves of correlation coefficient versus gage separation distance are plotted for all possible pairs of stations. As expected, the correlation coefficient increases as the gage separation distance decreases and as the data time averaging interval increases.; Storm kinematics is the primary cause of data scatter and low levels of correlation for the measured Eulerian rainfall fields. An analysis in the Lagrangian reference frame, moving with the center of mass of the storm, is proposed to minimize data scatter and obtain correlations representative of the internal structure of the rainfall event. The Lagrangian transformation technique consists of three steps: (1) estimation of storm kinematics; (2) spatial interpolation of the Eulerian rainfall field and (3) construction of the Lagrangian rainfall field.; The spatial cross-correlations are then recomputed and compared to the original Eulerian spatial correlation plots. Considerable improvement is seen in the Lagrangian correlation levels with greatly reduced data scatter. For five rainstorms the minimum increase in average correlation coefficient was 0.53 while the largest increase was 1.11.; The final objective of this study was to demonstrate that stochastic generation of the Lagrangian fields was possible with a simple rainfall model. A first-order autoregressive scheme with a random component was used. Spatial correlation curves, computed from numerous realizations of Lagrangian rainfall fields generated using this model, match the correlation of the rainstorms used in model calibration.