During the 1992-93 TOGA COARE, sounding systems exhibited a variety of humidity errors. The Vaisala capacitance-type sensors displayed a dry bias in the boundary layer (~1 g/kg for the H-sensors and ~0.25 g/kg for the A-sensors) and the VIZ hygristers a moist bias of ~0.5 g/kg. Owing to the geographical distribution of these sensors, i.e., the VIZ systems situated between 5 and 10 N and the Vaisala systems near and south of the equator, an erroneous north-south distribution of boundary-layer humidity was observed. This, it turn, apparently led to an erroneous north-south distribution of precipitation across the western Pacific warm pool in the ECMWF and NCEP reanalyses.

Correction algorithms have been recently developed for the Vaisala and VIZ sensors by the National Center for Atmospheric Research Atmospheric Technology Division. In this paper we report on several analyses designed to explore the impacts of these corrections. Prior to application of humidity corrections, four-month mean values of Convective Available Potential Energy (CAPE) at stations within the western Pacific warm pool varied by up to a factor of five; after correction they were relatively uniform, as expected over this region. Application of a 1-D buoyancy-sorting cloud model to the uncorrected and corrected data reveals significant differences in detrainment and mass flux profiles for convection over the warm pool. Model results using humidity-corrected data show an increase in upper-tropospheric detrainment and deep convective mass fluxes over Vaisala sounding areas and a decrease over the VIZ areas, consistent with the changes in boundary-layer moisture. There is a corresponding significant change in the north-south distribution of cloud mass fluxes: a shift in the peak deep cloud mass fluxes from 5-10 N prior to correction to a peak centered near the equator after correction. The latter distribution is more consistent with satellite and other independent estimates of precipitation. Finally, it is shown that the corrected humidity data provide a more accurate forcing for cloud-resolving models (CRMs), reducing the enthalpy loss (or climate drift) from that reported in previous CRM simulations using uncorrected data.