The Eulerian and Lagrangian correlation structures of 13 convective rainstorms were investigated using rainfall data from the 1987 Convective Initiation Downburst Experiment in Denver, Colorado. One minute rainfall data were available for two superposed raingage networks: 46 raingages on the portable automated mesonet (PAM) network at a mean spacing of 10.6 km and 31 additional raingages on the Federal Aviation Administration Lincoln Laboratory Operational Weather Studies (FLOWS) network at a mean spacing of 2.4 km. Eulerian correlation coefficients are low and often negative for high-resolution 1 min rainfall data. Correlation increases with time-averaging, beyond 10--15 min for single rain cells on the FLOWS network and small-mesoscale rainstorms on the PAM network. The average speed of rain cells and small-mesoscale rainstorms was found to be 76.5 and 56.4 km hr-1, respectively. Storm kinematic is identified as the cause of data scatter and low Eulerian correlations. Convective rainstorms are essentially uncorrelated in the Eulerian reference frame at a typical raingage spacing of 2--3 km for rain cells and 10--15 km for small mesoscale rainstorms. A spatial cross-correlation analysis in a Lagrangian reference frame moving with the center of mass of the storm separates the kinematic component from the structural component of a rainfall field. The spatial cross correlations in the Lagrangian rainfall field show considerable improvement over the Eulerian spatial correlation plots, and data scatter is greatly reduced. The increase in correlation coefficients from Eulerian to Lagrangian reference frames typically ranged from 0.5 to 1.1. ¿ 1998 American Geophysical Union