Quantifying Uncertainties in NCEP Reanalyses Using High-Quality Research Vessel Observations

Abstract

The uncertainties in the NCEP?NCAR reanalysis (NCEPR) products are not well known. Using a newly developed, high-resolution, quality controlled, surface meteorology dataset from research vessels participating in the World Ocean Circulation Experiment (WOCE), regional and global uncertainties are quantified for the NCEPR air?sea fluxes and the component fields used to create those fluxes. For the period 1990?95, WOCE vessel and gridded NCEPR fields are matched in time and space. All in situ data are subject to data quality review to remove suspect data. Adjustment of ship observations to the reference height of the NCEPR variables, and calculation of air?sea fluxes from the in situ data are accomplished using bulk formulas that take atmospheric stability, height of the measurements, and other adjustments into consideration. The advantages of using this new set of WOCE ship observations include the ability to compare 6-h integrated fluxes (much of the ship data originate from automated observing systems recording continual measurements), and the ability to perform more exhaustive quality control on these measurements. Over 4500 6-h component (sea level pressure, air and sea temperature, winds, and specific humidity) and flux (latent, sensible, and momentum) matches are statistically evaluated to quantify uncertainties between the ship observations and the NCEPR. Primary results include a significant underestimation in NCEPR near-surface wind speed at all latitudes. The magnitude of the low bias increases at higher ship wind speeds and may be related to large (rms = 2.7 hPa) errors in sea level atmospheric pressure over the entire globe. The pressure biases show the NCEPR to underestimate the amplitude and/or position of both high and low pressures. The NCEPR slightly underestimates the momentum flux, in part, due to the weaker winds. The NCEPR sensible and latent heat fluxes are largely overestimated when compared to the WOCE ship data. Potential sources of this overestimation (e.g., the NCEPR model flux parameterization) are discussed. Using the NCEPR meteorological variables and an independent flux parameterization, the revised NCEPR sensible heat fluxes are closer to the observations, and the biases of the revised NCEPR latent heat flux change sign. Furthermore, while the revised latent heat flux values reduce the magnitude of the bias at higher wind speeds, they increase the bias at (more frequently occurring) moderate wind speeds and thus may not be suitable for many applications.