MERRA-2 Global Precipitation Time Series

Global mean precipitation is an especially sensitive measure of consistency between the forecast model and observations used in a reanalysis. This is because the model precipitation, while not directly constrained by observations, is strongly dependent on physical processes whose parameterizations have known errors and can be highly sensitive to even small changes in large-scale temperature and humidity fields. Observations from new satellite sensors, while providing high-quality information, can be a major source of such changes in the context of reanalysis, resulting in spurious trends and discontinuities in some model output fields.

The figure shows the time series of monthly mean global precipitation for several reanalyses, including MERRA and MERRA-2, as well as for the Global Precipitation Climatology Project (GPCP) data set. All reanalyses exhibit trends and discontinuities not visible in the GPCP record, as well as higher mean values overall (the early years of MERRA being a notable exception). In particular, the MERRA time series has strong discontinuities corresponding to the introduction of several microwave instruments sensitive to temperature and water vapor, including AMSU-A on NOAA-15 in 1998 and NOAA-16 in 2001, and, to a lesser extent, SSM/I on F08 in 1987 and F11 in late 1991. There is also a noticeable decrease in MERRA precipitation with the loss of all SSM/I data in 2009. These sensors have far less disruptive impact in MERRA-2, especially in the case of AMSU-A (although a discernible jump occurs with the introduction of F08 in 1987). The improved results in MERRA-2 are due to several factors including more stringent data selection, a reformulated analysis variable for moisture, updates to the model’s moist physics, and imposed constraints on global imbalances in water vapor and surface pressure.

« GMAO Science Snapshots