Global Precipitation Analyses and Reanalyses

1
Global Precipitation Analyses and Reanalyses

• Phil Arkin, Cooperative Institute for Climate
  Studies
• Earth System Science Interdisciplinary Center,
  University of Maryland
• based on work by Matt Sapiano, Ching-Yee Chang
  and John Janowiak of CICS/ESSIC
• and
• Tom Smith, NOAA/NESDIS/STAR and CICS

2
Issues/Background

• Science questions
• How much precipitation occurs over the globe?
  (i.e. What is the strength of the global
  hydrological cycle?)
• How does global mean precipitation vary with
  time? (i.e. How is the hydrological cycle
  changing?)
• Can coupled climate models reproduce the observed
  changes in global mean precipitation?

3
Integrating/Analyzing Precipitation Observations

• Analysis creating complete (in time and space)
  fields from varying and incomplete observations
• Satellite-derived estimates have complementary
  characteristics (geostationary IR is more
  complete but has poor accuracy, low Earth orbit
  PMW is more accurate but has sparse sampling) so
  combining them makes sense (CMAP, GPCP, CMORPH,
  TMPA, GSMaP)
• Rain gauges are helpful to reduce biases in
  satellite-derived estimates
• Model-generated precipitation may be useful in
  regions where neither gauges nor satellite
  estimates are skillful

• GPCP mean annual cycle (left) and global mean
  precipitation (below)
• Monthly/5-day 2.5 lat/long global
• CMAP has similar characteristics

4
Global Mean Precipitation from Reanalyses

• Modern reanalyses give global means considerably
  greater than GPCP and CMAP
• Reanalyses, which should be most realistic
  model-based product (since they use observations)
  are also higher than most AR4 model simulations
• (figure courtesy Junye Chen, NASA/GMAO-MERRA)

5

• Goal Create global precipitation analysis
  beginning in 1900
• To validate global climate models
• To describe long-term trends in global,
  particularly oceanic, precipitation
• To describe interdecadal variability in phenomena
  such as ENSO, the NAO, the PDO and others
• Approach reconstruct/reanalyze global
  precipitation back to 1900 using 2 methods
• EOF-based reconstruction using GPCP and other
  global precipitation analyses, combined with
  historical coastal and island rain gauge
  observations
• CCA reanalysis using SST and SLP, based on modern
  era analyses
• Compare to GHCN gauge observations, NOAA/ESRL
  20th Century SLP-based reanalysis and IPCC AR4
  C20C products

6
Climate Modes from EOF Reconstructions

• EOF method enables good recovery of climate modes
• Better in NH and tropics SH too poorly sampled
  for good results
• Global time series from EOF reconstructions does
  not have realistic decadal-scale variations
• See Smith et. al. 2008, JGR

7
CCA Reanalyses

• CCA nearly independent of GHCN observations,
  although GPCP uses gauge data to remove bias (CCA
  based on gauge-free version of GPCP gives similar
  results)
• Top panel shows comparison over land areas where
  gauges are found small areal coverage
• Decadal-scale signal looks reasonable
• Ability to resolve finer scale phenomena like
  ENSO is limited yearly, 5, bigger errors on
  short time scales

Fig 1 DJF means.
8

• All values are anomalies relative to CCA mean
  over the period
• /- 1 and 2 SD plotted for AR4 runs
• Compo reanalysis above AR4 range similar to
  modern reanalyses, which are 0.5-0.8 mm/dy gt GPCP
  and CMAP
• GPCP and CCA in lower part of AR4 range

9

• Note scale changed by factor of 10
• Biases removed so means are the same for all time
  series
• Compo interannual variations similar to CCA, but
  long time scales different
• AR4 ensemble mean exhibits much less variability
  since it is an average of many (20 or so) runs

10

• Re-scale AR4 ensemble mean so variance is about
  same as a single realization
• CCA and AR4 ensemble mean show similar
  centennial-scale changes, but interannual
  variations rather different

11
Conclusions/Issues

• EOF-based Reconstruction back to 1900 exhibits
  skill in capturing seasonal-to-decadal variations
• GPCP-based CCA reanalysis matches 20th Century
  variations from IPCC AR4 model simulations
• Both SST and SLP are important, but SST is
  stronger influence over oceans
• Result holds for GPCP_ms, but not for CMAP CMAP
  has very different decadal scale oceanic
  variability
• Best historical analysis may be combination of
  low frequency from CCA and finer scales from
  filtered EOF reconstructions
• CCA is low resolution annual, 5 and so cant
  represent important details
• Compo-Whittaker reanalysis precipitation has
  short time scale variations much like CCA, but is
  very different on longer time scales
• Significant biases still present between models
  and observed datasets