Rawinsondes invariably sample small-scale spatial and temporal fluctuations of the atmospheric flow. In the course of interpolating fields from networks of sounding stations, these fluctuations are aliased onto larger scales, leading to errors in the analyzed fields. Some of these errors are random (e.g., those due to random turbulence or convection), whereas others are nonrandom (e.g., those due to local effects such as topography, instrument biases, etc.). Assuming nonrandom errors or biases can be effectively removed, the remaining sampling errors should be random and therefore reduced by averaging over successively longer time intervals. To date, quantitative estimates of the impacts of random sampling errors on atmospheric budgets have not been made.

In order to address this problem, sounding data from four ships in the TOGA-COARE Intensive Flux Array (IFA) are used to estimate random sampling errors in the moisture and wind fields. Deviations of individual observations from the four-station means form the basis of a `noise model' that is then applied to the analysis of atmospheric budgets for the four-month Intensive Observing Period (IOP). The impacts of random sampling errors on computations of rainfall rates and heating and moistening profiles for different averaging periods will be reported.