How Accurate is Land / Ocean Moisture Transport

Session 4, topic 2: Advances in analysis of observa8ons, reanalysis and model results to improve energy and water cycle processes 7th Interna*onal Conference on the Global Water and Energy Cycle !"#$%#&'%$"'%(#)*'*%+,%
  Moisture transport crosscuts all four GEWEX Science Ques*ons. But our ability to model this process remains problema*c:   CMIP3 models suggest moisture transport to land increases ~5% in the 21st Century (Liepert and Previdi, 2012), but this value is smaller than ensemble spread.   Ocean  land moisture transport increases with finer resolu8on (Demory et al, 2013). Mean annual fluxes of the global water cycle (1,000 km3/yr), and associated uncertain*es, during the first decade of the millennium. (Rodell et al, 2014). White numbers are based on observa*onal products and data integra*ng models, and blue numbers are es*mates that have been op*mized by forcing water and energy budget closure and taking into account uncertainty in the original es*mates. Yellow numbers are Trenberth (2011)   Reanalyses’ uncertain*es are diminishing but the challenges of model physics, grid resolu*on, observa*onal constraints in assimila*on persist (Trenberth et al, 2011; Robertson et al, 2011; Lorenz and Kunstman, 2012; Robertson et al, 2014). 7th Interna*onal Conference on the Global Water and Energy Cycle To what extent does the evolu8on in (satellite) observing systems affect assimilated VMFC? SSMI AMSU-­‐A Trends (mmd-­‐1 dec-­‐1) Obs Forced LSMs (0.013) JRA55 (-­‐0.031) ERA-­‐I (0.085) MERRA (0.074) CFSR (0.074) 7th Interna*onal Conference on the Global Water and Energy Cycle 1. To what extent are observa*onally-­‐constrained LSMs useful in valida*ng reanalysis moisture transport variability? Reanalyses Observa8onally-­‐constrained LSMs 2. What regional contribu*ons drive reanalysis trends and to what extent can reanalysis ar*facts be isolated removed? 3. Characterize remaining physical variability during the “Satellite-­‐era”. 7th Interna*onal Conference on the Global Water and Energy Cycle Reanalyses (see h8p://www.reanalysis.org/ and h8ps://climatedataguide.ucar.edu/ ) ERA-­‐I Dee et al. 2011; Dee, and Uppala 2008; 2009; Simmons et al, 2010; Trenberth et al, 2011 JRA55 Ebita et al, 2011 MERRA Rienecker et al. 2011;Trenberth et al, 2011;Bosilovich, et al, 2011; Robertson et al, 2011. CFSR Saha et al. 2010; Wang et al, 2010; Trenberth et al, 2011. Land Surface Models Forcing References GLDAS-­‐2 Noah Princeton Chen et al. 1996; Koren et al. 1999; Rodell, 2004 MERRA-­‐Land MERRA, CPCU precip Reichle et al, 2011; 2012 MPI-­‐BGC Machine learning algorithm scales up FLUXNET in satellite-­‐driven SEB model. Combined with GPCC precip to make P-­‐ET. Jung et al., 2009; 2010 CLM4C TRENDY NCEP Rean, CRU Precip, Ts ORCHIDEE EU-­‐WATCH ERA-­‐I, GPCC Precip, Ts (WFDERI) Oleson et al, 2010; Lawrence et al, 2011; Weedon et al. (2011) Kriner et al, 2005; Poulter (201x) Weedon et al. (2012) CMIP-­‐5 AMIP Experiments GFDL-­‐HIRAM-­‐C180, GISS-­‐E2-­‐R, HadGEM2-­‐A, MIROC5, and MRI-­‐CGCM 7th Interna*onal Conference on the Global Water and Energy Cycle ERA-­‐I JRA55 MERRA CFSR LSM 7th Interna*onal Conference on the Global Water and Energy Cycle • S/N is high in populous areas of NH, but drops below 2 in cri*cal areas of Central Africa, South America and other tropical loca*ons • Anomaly Correla*on shows high spa*al coherence among LSMs with occasional problems in MERRA-­‐
Land, GLDAS-­‐2 Noah. • ORCHIDEE and MPI-­‐
BGC share GPCC precipita*on forcing. 7th Interna*onal Conference on the Global Water and Energy Cycle a a.  W. Eq. Africa Differences with LSMs involve trends and non-­‐sta*onary annual cycle with changes rela*ng to the onset of SSMI (1988) and AMSU-­‐A (1998) data. b
b. MERRA VMFC anomaly changes closely related to 1st two PCs of moisture increment AMSU-­‐A SSMI SSMI AMSU-­‐A SSMI c c.  Over Coastal Colombia / Equador ERA-­‐I VMFC jump coincides with addi*onal METAR data ingest in early 2004. Change aoer 1987 SSMI availability begins is also evident. 7th Interna*onal Conference on the Global Water and Energy Cycle Leading Rotated EOF Rotated PCs 1 2 3 • LSM modes are all dominated by ENSO, but the leading reanalysis modes are each trend modes. • 1st and 2nd reanalysis modes have trends and signature of non-­‐
sta*onary annual cycle. • 3rd Reanalysis modes (except JRA55) are ENSO components. 7th Interna*onal Conference on the Global Water and Energy Cycle ERA-­‐I Global Land (60oN/S) VMFC Anomalies (mmd-­‐1) Black: Original Reanalyses JRA55 MERRA Green: Global mean of local correc*ons to be subtracted from reanalyses. Correc*ons remove REOF modes 1, 2 from ERA-­‐I, MERRA, CFSR; modes 1-­‐3 from JRA55 Niño 3.4 SST anomalies (oC /10) (Note inverted scale on r.h.s.) • Global mean correc*on for each is essen*ally a trend with a *me–varying modula*on of the annual cycle. CFSR • Evidence of SSMI-­‐related correc*ons in all reanalyses except CFSR. • MERRA exhibits largest signal of SSMI and AMSU-­‐B influence. 7th Interna*onal Conference on the Global Water and Energy Cycle Trends (mmd-­‐1 dec-­‐1) JRA55(0.008) ERA-­‐I(0.022) MERRA (0.024) CFSR(0.024) /10
Trends (mmd-­‐1 dec-­‐1) LSMens(0.013) Adj Reans (0.016) AMIPens (0.12) /10
LSMs, Reans (0.87) LSMs, AMIPs (0.56) Reans, AMIPs (0.64) LSMs, Nino3.4 (-­‐0.56) Reans, Nino3.4 (-­‐0.63) AMIPs, Nino3.4 (-­‐0.65) 7th Interna*onal Conference on the Global Water and Energy Cycle LSM ensemble P-­‐ET (mmd-­‐1) • Regional parerns of decadal change for Adjusted Reanalysis VMFC agree berer with LSM P-­‐ET (e.g. C. Africa, Amazon, S. Asia). Adjusted Reanalyses VMFC (mmd-­‐1) Raw Reanalyses VMFC (mmd-­‐1) • Less ”ENSO-­‐like” VMFC over C. Pacific and more VMFC over the Warm Pool and Mari*me Con*nent is consistent with post-­‐2000 cooling SSTs in the Eastern Pacific. • Over-­‐ocean correc8ons are enforced by a combina8on of regression using RPCs and enforcement of zero global mean moisture flux divergence. 7th Interna*onal Conference on the Global Water and Energy Cycle 7th Interna*onal Conference on the Global Water and Energy Cycle B A C K U P S 7th Interna*onal Conference on the Global Water and Energy Cycle Local Effects of Reanalysis Adjustments Correla*on Map of Adjusted Reanalysis VMFC and LSM P-­‐ET Correla*on Improvement over Raw Reanalyses (note differing scale) 7th Interna*onal Conference on the Global Water and Energy Cycle 2
2
S/N ! " LSM " LSM '
ACC !
2
! LSM
= variance of the (p
pi, j mi, j
2
i, j
2
i, j
m
)
1/2
ensemble mean monthly P-­‐ET anomaly pi,j are the P-­‐ET anomalies at gridpoint i,j 2
! LSM
' = ensemble mean mi,j is ensemble mean P-­‐ET anomaly of the individual squared departures from the ensemble mean monthly anomaly Brackets denote areal average 7th Interna*onal Conference on the Global Water and Energy Cycle 7th Interna*onal Conference on the Global Water and Energy Cycle