Water Cycle Variability over the Global Oceans Estimated Using Homogenized Reanalysis Fluxes

Establishing consistent records of the global water cycle fluxes and their variations is particularly difficult over oceans where the density of in situ observations varies enormously with time, satellite retrievals of flux processes are sparse, and reanalyses are uncertain. The latter have the positive attribute of assimilating diverse observations to provide boundary fluxes and transports but are hindered by at least two factors: (1) the physical parameterizations are imperfect and, (2) the forcing data availability and quality vary greatly in time and, thus, can induce time-dependent, false signals of climate variability. Here we examine the prospects for homogenization of reanalysis records, that is, identifying and greatly minimizing non-physical signals. Our analysis focuses on the satellite era, 1980 to near present. The strategy involves three atmospheric reanalysis systems: (1) the NASA MERRA-2, (2) the newest reanalysis produced by the Japanese Meteorological Agency, JRA-55, and (3) the European Centre for Medium Range Weather Forecasts 20th Century reanalysis, ERA-20C. MERRA-2 and ERA-20C are also accompanied by 10-member AMIP integrations, and JRA-55 by a reanalysis using only conventional observations, JRA-55C. Differencing these latter integrations from the more comprehensive reanalyses helps provide a clearer picture of the impact of satellite observations by removing the effects of SST forcing. This facilitates the use of principal component analysis as a tool to identify and remove non-physical signals. We then use these homogenized E, P and moisture transports to examine the consistency of diagnostics of thermodynamic and hydrologic scaling, especially the P-E pattern amplification or the "wet-get-wetter, dry-get-drier" response. Prospects for further validation by new turbulent flux retrievals by satellite are discussed.