## Abstract

A comprehensive observational dataset encompassing the entire temporal evolution of horizontal convective rolls was obtained for the first time. Florida, Illinois, and Kansas measurements from preroll conditions through the development of well-defined rolls to their dissipation were utilized to determine the factors influencing roll evolution. When the buoyancy flux reached a critical value of 35–50 W m^{−2}, the first form of boundary layer convection resolved by radar was rolls. It was noted that two-dimensional convective rolls can evolve in a convective boundary layer in the absence of significant wind speed and shear. In fact, the value of wind speed or shear in itself did not seem to determine when or if rolls would form, although it did influence roll evolution. Well-defined, two-dimensional rolls only occurred while −*z*_{i}/*L,* where *z*_{i} is the convective boundary layer depth and *L* is the Monin–Obukhov length, was less than ∼25, which is consistent with previous studies. As −*z*_{i}/*L* increased throughout the day, either open cellular convection or unorganized boundary layer convection was the dominant clear-air convective mode. If the wind speed was low (mean boundary layer winds <3 m s^{−1} or 10-m winds <2 m s^{−1}) during roll occurrences, rolls evolved into open cells. Alternatively, if the wind speed throughout the day was relatively high, rolls broke apart into random, unorganized convective elements. These are unprecedented observations of two-dimensional convection evolving into three-dimensional convection over land, which is analogous to laboratory convection where increased thermal forcing can produce a transition from two-dimensional to three-dimensional structures. Finally, the roll orientation was governed primarily by the mean convective boundary layer wind direction.

*Corresponding author address:* Dr. Tammy M. Weckwerth, NCAR, P.O. Box 3000, Boulder, CO 80307-3000.

Email: tammy@ucar.edu