We present a description of the evolution of the convective boundary layer (CBL) over the boreal forests of Saskatchewan and Manitoba, as observed by the National Center for Atmospheric Research (NCAR) Electra research aircraft during the 1994 Boreal Ecosystem-Atmosphere Study (BOREAS). All observations were made between 1530 and 2230 UT (0930-1630 local solar time (LST)). We show that the CBL flux divergence often led to drying of the CBL over the course of the day, with the greatest drying (approaching 0.5 g kg-1 h-1) observed in the morning, 1000-1200 LST, and decreasing over time to nearly no drying (0--0.1 g kg-1 h-1) by midafternoon (1500--1600 LST). The maximum warming (0.45 K h-1) also occurred in the morning and decreased slightly to about 0.4 K h-1 by midafternoon. The CBL vapor pressure deficit (VPD) increased over the course of the day. A significant portion of this increase can be explained by the vertical flux divergence, though horizontal advection also appears to be important. We suggest a linkage among boundary layer growth, the vertical flux divergences, and boundary layer cloud formation, with cloud activity peaking at midday in response to rapid CBL growth, then decreasing somewhat later in the day in response to CBL warming and decreased growth. We also see evidence of feedback between increasing VPD and stomatal control. We use eddy-covariance flux measurements from the Electra to compute the virtual temperature entrainment ratio Ar. The computed mean value of 0.08¿0.12 is somewhat lower than the commonly assumed value of 0.2, as well as with other estimates from BOREAS. This value is very sensitive to the determination of CBL depth. We find that Ar increases with an increasing jump in mean wind across the CBL top. The entrainment flux of water vapor is found to be most dependent on time of day (negative correlation). The ratio of entrainment to surface flux of water vapor is 1.57¿0.25. Airborne lidar observations of the CBL top reveal a CBL top thickness that is smaller than would be expected from simple theory but consistent with past lidar observations. The normalized thickness is found to have a very consistent value h¿/h0-1=0.116¿0.008, where 12 cases were examined. A new method of computing the variability of the CBL top is illustrated, and we show that this variance in the CBL depth also scales with the depth but that the value of this normalized variance differs substantially from the thickness defined in past literature. ¿ 1997 American Geophysical Union