Rawinsonde observations have long been used to estimate the atmospheric boundary layer (ABL) depth (BLD), which is an important parameter for monitoring air quality, dispersion studies, weather forecast models, and inversion systems for estimating regional surface-atmosphere fluxes of tracers. Although many approaches exist for deriving the BLDs from rawinsonde observations, the bulk Richardson approach has been found to be most appropriate. However, the impact of errors in the measured thermodynamic and kinematic fields on the estimated BLDs remains unexplored. We argue that quantifying BLD error (δBLD) estimates are as equally important as the BLDs themselves. Here we quantified δBLD by applying the bulk Richardson method to 35 years of rawinsonde data obtained from three stations in the US: Sterling, VA; Amarillo, TX; and Salt Lake City, UT. Results revealed similar features in terms of their respective errors; a -2°C (+2°C) bias in temperature yielded a mean δBLD ranging from - 15 m to 200 m (-214 m to +18 m). For a -5 % (+5 %) relative humidity bias, the mean δBLD ranged from -302 m to +7 m (+2 m to +249 m). Differences of ±2 m s-1 in the winds yielded BLD errors of around ±300 m. δBLD increased (decreased) as a function of BLD when introducing errors to the thermodynamic (kinematic) fields. These findings expand upon previous work evaluating rawinsonde-derived δBLD by quantifying δBLD arising from rawinsonde-derived thermodynamic and kinematic measurements. Knowledge of δBLD is critical in, for example intercomparison studies where rawinsonde-derived BLDs are used as references.