Evaluation of Radiation and Clouds From Five Reanalysis Products in the Northeast Pacific Ocean

Abstract

Atmospheric reanalyses are valuable tools for studying the atmosphere, as they provide temporally and spatially complete coverage of atmospheric variables. However, some regions are susceptible to large biases in reanalysis products due to the scarce data available to assimilate into the reanalyses. Consequently, evaluation of reanalyses using available measurements is essential for quantifying regional errors. Here we use NASA's CERES satellite estimates to evaluate surface radiative fluxes and total cloud fraction in the Northeast Pacific from five reanalysis products—ERA‐Interim, MERRA2, JRA‐55, NCEP2, and CFSR—from years 2001 to 2015. Results show that biases of surface incident shortwave radiative flux in reanalyses compared to satellite estimates range from 3.8 (CFSR) to 21.2 Wm‐2 (NCEP2), with significant biases in JRA‐55 and NCEP2. Mean surface downward longwave radiative flux in the reanalysis products is biased by −8.9 (MERRA2) to 3.9 Wm−2 (JRA‐55), with significant biases in MERRA2 and NCEP2. Errors in the surface radiative fluxes are partially linked to differences in total cloud fraction in the satellite estimates and reanalyses, which show significant negative biases ranging from −8% (CFSR) to −21.7% (NCEP2). There is not one reanalysis that outperforms the rest in the NE Pacific. The most appropriate data set depends on the variables of interest, subregion of the NE Pacific being studied, time period of interest, and whether the reanalysis data will be used to study long‐term or short‐term climate processes. Using the errors presented for each reanalysis data set can help guide appropriate use and bound uncertainty for the five reanalysis products analyzed.

Representation of Dropsonde‐Observed Atmospheric River Conditions in Reanalyses

“…The simulation of IVT in reanalyses likely also affects the representation of downstream precipitation. Several studies have compared atmospheric reanalyses to observations, principally relying on those from satellites (e.g., Belmonte Rivas & Stoffelen, 2019;Schmeisser et al, 2018), largely finding differences in skill with location and variable. Jackson et al (2016) found that AR landfall data from reanalyses were generally in good agreement with satellite observations, with reduced bias of AR characteristics such as width and landfall location in those datasets with more data assimilation.…”

Decomposing Shortwave Top-of-Atmosphere and Surface Radiative Flux Variations in Terms of Surface and Atmospheric Contributions

“…They find that the reanalyses tend to underestimate cloud fraction by 8%-21%, resulting in an overestimation of downward SW flux at the surface that ranges from 4 to 21 W m 22 . It is noteworthy that the study of Schmeisser et al (2018) is one of the few that evaluates interannual variations in clouds and radiation. They find that models that adequately reproduce the climatology in downward SW surface flux do not necessarily capture extreme anomalies very well.…”

“…system models (Park et al 2014;Zhao et al 2018), regional models (Lim et al 2014), atmospheric reanalysis products (Zhang et al 2016;Schmeisser et al 2018), and specific model parameterizations, including ice cloud parameterizations (Baran et al 2016;Eidhammer et al 2017;Chern et al 2016), cloud microphysics schemes (Gettelman et al 2015), low cloud parameterizations (Cheng and Xu 2015;Qin et al 2018;Song et al 2018), mixed-phase cloud parameterizations (Furtado et al 2016), parameterizations of deep convection (Boyle et al 2015;Guo et al 2014;Wang and Zhang 2016), and parameterizations of subgrid-scale cloud water content variability (Hill et al 2015).…”

The Role of Clouds and Surface Heat Fluxes in the Maintenance of the 2013–2016 Northeast Pacific Marine Heatwave

“…We use reanalysis due to its complete spatial coverage, long time series, and comprehensive list of variables. CFSR was found to be the most appropriate reanalysis dataset for the NE Pacific region, based on comparisons between six reanalysis datasets and National Aeronautics and Space Administration's CERES satellite estimates of radiative flux and cloud fraction in which CFSR included the smallest errors (Schmeisser et al, ). An analysis using CERES‐EBAF data is included in the supporting information so readers can see the sensitivity of results to dataset choice.…”