Here we focus on variability of ocean evaporation on scales from interannual to decadal by appealing to three sources of data. Two 20th Century reanalyses are used, the NOAA ESRL 20CR V2 and the ECMWF ERA-20C. In contrast to conventional reanalyses climate data reanalyses are run with lesser constraints such as with SSTs, sea-ice (i.e. AMIPs) and additional, minimal observations of surface pressure and marine observations that have longer and less fragmentary observational records. While the 20CR V2 assimilates surface pressure observations, the ERA-20C is distinct in that it also assimilates ICOADS V2 marine wind speeds. A less constrained system we also use consists of an ensemble of 20th Century AMIP-like experiments using the NASA GEOS5 model. These have almost identical external radiative forcing, but assimilate no dynamical or kinematic information. Though limited in temporal extent, state-of-the-art satellite retrievals from the Seaflux and HOAPS (Hamburg Ocean-Atmosphere Parameters and Fluxes from Satellite) projects represent best observationally driven estimates of evaporation. Each of these sources has distinct advantages as well as drawbacks. Physics biases in climate models and the lack of a predicted surface energy balance are a source of concern. Satellite retrievals and comparisons to ship-based measurements offer the most observationally-based estimates, but sensor inter-calibration, algorithm retrieval assumptions, and short records are dominant issues. Our strategy depends on maximizing the advantages of these combined records.
The primary diagnostic tool used here is an analysis of bulk aerodynamic computations produced by these sources and uses a first-order Taylor series analysis of wind speed, SST, near-surface stability and relative humidity variations around climatology to gauge the importance of these components. We find that near-surface relative humidity and stability changes both act to counterbalance the effects of SST alone in driving evaporation changes. This result is consistent with the expectation that the rate of the hydrologic cycle increases is smaller than that of Clausius-Clapeyron scaling. There are however some significant differences in evaporation trends between AMIP experiments and 20th Century reanalyses. The ERA-20C has a significant evaporation trend that appears related to changes in assimilated wind speeds. Significant difficulties persist with satellite-retrieved evaporation due to uncertainties in sensor calibration. Though they can adequately depict climatological states, and to some degree interannual variability, decadal changes or trends are still problematic.