Alfredo Ruiz-Barradas Sumant Nigam

1. Seasonal Hydroclimate Variability over North America:Global and Regional Reanalyses Faulty Evapotranspiration Department of Atmospheric and Oceanic Science University of Maryland Alfredo Ruiz-Barradas Sumant Nigam CPPA: 2008 PIs Meeting Silver Spring, MD September 29-October 1 September 30, 2008

2. Data Sets • Global Reanalyses: NCEP/NCAR ERA-40 JRA-25 • Regional Reanalysis: NARR • Other data sets: GSWP-2 (multi-model mean, global) VIC (US optimized) UNH/GRDC (Fekete et al. 2002) US-MEX Period: 1979-1999

3. The Beginning: An interannual precipitation variability analysis: NARR J, J, A GPP Index Regressions Precipitation JJA Standard Deviation MFC > ET in reanalyses. High expectations on NARR Box (100º-90ºW,35º-45ºN) defines Great Plains Precipitation (GPP) Index Region is at the center of the Discussion of local vs remote forcing of precipitation variability. CI=0.3 mm/day

4. Atmospheric Water Balance (1979-1999) Area-average over the Great Plains (100ºW-90ºW,35º-45ºN) Is ET large in NARR?

5. Atmospheric Water Balance (1979-1999) Is ET large in NARR? It is comparable to: NCEP/NCAR Larger than: ERA-40 JRA-25 VIC Residual NARR Yes, ET seems to be large From the AWB point of view.

6. Comparison with Rasmusson (1968; MWR, 96, 720-734) over Central Plains and Eastern US regions (110º-80ºW,30º-50ºN) From Rasmusson, E. (1968; MWR, 96, 720-734)

7. Terrestrial Water Balance NARR (1979-1999) Annual mean: Background Annual cycle: Dials P-E ≈ Runoff

8. River Discharge Composite Runoff constrained by spatially distributed river discharge NARR Annual mean runoff in NARR is smaller than the observationally constrained product This implies that annual mean P – E is underestimated or E is overestimated because P is very well assimilated in NARR

9. The Terrestrial Water Balance in Global Reanalyses P-E ≈ Runoff Driest Wettest

10. The Terrestrial Water Balance in Offline Land-Sfc. Models P-E ≈ Runoff Runoff in VIC and UNH/GRDC products have a similar structure in the annual mean This supports the idea that Runoff (E) is underestimated (overestimated) in NARR

11. P-E Winter Runoff In NARR: P-E ≈ ∂ws/∂t P-E Summer Runoff

12. Area-average over the Great Plains (100ºW-90ºW,35º-45ºN) • Seasonal imbalances in NARR from spring to fall are due to the highlighted errors in runoff and evaporation. • JRA-25 has the largest seasonal imbalances, although its annual • imbalance is comparable to that • in ERA-40 and ~ 1/2 of that in NCEP/NCAR. • Seasonal imbalances in JRA-25 • are due to overestimation in • P - E, the change in water storage, and runoff as well as the almost nonexistent annual cycle in P - E. • ERA-40 has larger imbalances • from spring to fall due to the • small and almost uniform change • in water storage during those • seasons.

13. It is known that: Noah, the LSM in NARR, had a large positive bias in summer evaporation over regions of non-sparse vegetation cover, such as the eastern US, and that the bias was related to canopy resistance parameters in the model (Mitchell et al 2005; http://www.emc.ncep.noaa.gov/gc_wmb/Documentation/TPBoct05/T382.TPB.FINAL.htm). Upgrades to the Noah model, including the correction of the evaporation bias, were implemented in middle 2005, almost two years after the completion of NARR at the end of 2003. Conclusions: • It is clear that reanalyses have some problems in reproducing the terrestrial water cycle over North America, particularly the global products. The most recent global reanalysis, the Japanese reanalysis, does not improve the representation of the climatological features of the terrestrial water cycle over North America. • The regional reanalysis NARR severely overestimates evaporation (and the change in water storage) that leads to the underestimation of runoff. Is the problem fix in its LSM? • In any case, these results make clear the need for a correction in the assimilation process in NARR in which some observational constraints on the land-surface part are needed to generate realistic hydroclimate fields.