A total of 826 temperature and shear microstructure profiles, measured during 12 days in March 1996 in Lake Neuchatel (Switzerland), are used to compare estimates of turbulent kinetic energy dissipation, determined by the Batchelor (mu T; temperature) and the dissipation (mu S; shear) methods. Positioned horizontally 35 m apart, the two different profilers were raised through the nearly-homogeneous surface boundary layer. The collected time series covered different atmospheric forcings, including calm (approximate to 1 m s(-1)) and windy (> 10 m s(-1)) periods, as well as periods of thermally-induced convection (nights) and stratification (sunny days). Consequently, dissipation varied over 6 orders of magnitudes from 10(-11) to 10(-5) W kg(-1). The comparison of the 72'674 pairs of dissipation estimates, determined from 25 cm long vertical profile segments, reveals excellent agreement for the bulk of the estimates, with deviations smaller than a factor of 2. In a relatively small subset, collected under strong heating, deviations of up to a factor of 6 occur in 1-4 m depth. These deviations are, however, just about within the 95% confidence limits of the bootstrap distribution. Under low turbulence the mu T method tends to provide slightly larger dissipation, whereas for high wind-induced turbulence the mu S method gives larger dissipation.