A Comparison of Atmospheric Reanalysis Surface Products over the Ocean and Implications for Uncertainties in Air–Sea Boundary Forcing

Abstract

This paper investigates the uncertainties related to atmospheric fields from reanalysis products used in forcing ocean models. Four reanalysis products, namely from 1) the interim ECMWF Re-Analysis (ERA-Interim), 2) version 2 of the Common Reference Ocean–Ice Experiments (CORE2), 3) the 25-Year Japanese Reanalysis Project (JRA-25), and 4) NCEP–NCAR, are evaluated against satellite-derived observations for eight different fields (zonal and meridional winds, precipitation, specific humidity, continental discharge, surface air temperature, and downwelling longwave and shortwave radiation fluxes). No single product is found to agree better in all fields with satellite-derived observations. Reanalysis products are mostly comparable to each other because of their similar physical assumptions and assimilation of common observations. Adjusted atmospheric fields from the Estimating the Circulation and Climate of the Ocean (ECCO) optimizations are also in agreement with other reanalysis products. Time-mean and time-variable errors are estimated separately and mapped globally in space, based on 14-day average fields to focus on monthly to interannual periods. Time-variable errors are larger in comparison to the signal than time-mean errors for most fields, thus justifying the need to separate them for studying uncertainties as well as formulating optimization procedures. Precipitation and wind stress fields show significant time-mean and time-variable errors whereas downwelling radiation, air temperature, and humidity fields show small time-mean errors but large time-variable errors, particularly in the tropics. Uncertainties based on evaluating multiple products presented here are considerably larger than uncertainties based on single product pairs.

Biogeochemical modelling of dissolved oxygen in a changing ocean

“…Particularly, in agreement with forced ocean biogeochemistry model results of [86] for the North Atlantic, simulated interannual variability in [O2] is shown here to be primarily associated with heat fluxes (σθ) in extratropical (mainly subpolar) regions and wind stress (σwspd) in the tropics. Large uncertainties between prescribed meteorological datasets in these regions propagate into simulated thermocline [O2], consistent with a recent inter-comparison of surface reanalysis data, which attributes elevated multiproduct inconsistency in the tropics and extratropics chiefly to wind stress and heat flux uncertainties, respectively [99]. The important role of tropical zonal wind str ess in controlling variations in model simulated low-O2 water bodies has been demonstrated in recent work [100,24] providing further motivation for the provision of appropriate wind forcing to ocean-only models investigating O2 dynamics.…”

“…Further work is required to better understand the biases in atmospheric forcing datasets, both in terms of comparing meteorological fields (e.g. [99]) and assessing how uncertainties in these prescribed forcings (and bulk formulae) impact upon the evolution of hindcast variables. Towards this objective, the Coordinated Ocean-ice Reference Experiments (COREs) project proposes a standard protocol for running hindcast ocean-ice models, emphasising the need for models to be integrated using different atmospheric forcings in order to "assess implications on the ocean and sea ice climate of various atmospheric reanalysis or observational products" [101].…”

Interannual sea level variability in the tropical Pacific Ocean from 1993 to 2006

“…The different magnitudes of trend shown in Fig. 5a and b may be attributed to the individual biases in respective reanalysis fields (Chaudhuri et al, 2013). However, the positive trend along the western equatorial Pacific is not reproduced in several contemporary studies (e.g., Timmermann et al, 2010, Fig.…”

“…To be consistent with the net surface heat flux products and surface wind stress products used in later sections, the altimeter data are interpolated to a 1°× 1°s patial resolution over the time span from January 1993 to December 2006. Considering that different atmospheric reanalyses have individual biases in their reanalysis fields, and no single reanalysis product is found to agree better in all fields with satellite-derived observations (Chaudhuri et al, 2013), three net surface heat flux products (CORE.2, ECMWF ORA-S3 and OAFlux) and three wind stress products (CORE.2, ECMWF ORA-S3 and SODA) are used to evaluate the contributions of different physical processes to sea level variability. The version 2 of Common Ocean Reference Experiment (CORE.2) Global Air-Sea Flux Dataset (Large and Yeager, 2004) Yu et al, 2008) products can be obtained from the CISL RDA: a daily version (Jan 1985 to Dec 2011) and a monthly version (Jan 1958to Dec 2011.…”

Assimilation of GRACE-derived oceanic mass distributions with a global ocean circulation model

“…The spread of the ensemble itself is a consequence of the different initial state of each ensemble member and the therefore different reactions to the atmospheric forcing. Since the forcing itself has substantial errors (e.g., Trenberth et al, 2011;Chaudhuri et al, 2013), the forcing fields are included in the assimilation process (see Sect. 2.3).…”