Systematic Errors and Surface Fluxes

in the NCEP Global Model

 

Glenn White

Environmental Modeling Center

National Centers for Environmental Prediction

NWS/NOAA/DOC

Camp Springs, MD

Glenn.White@noaa.gov

 

††††††††††† This paper reflects the following assumptions:

 

a)     Improving the atmospheric analysis/forecast systems currently used for 0-15 day numerical weather prediction and coupling them to complex land and ocean analysis/forecast systems can significantly improve 15-60 day forecasts.

b)     Improving either the 0-15 or 15-60 day forecasts will usually improve the other.

c)     Investigating the shortcomings of the current forecast system is a key ingredient in improving forecasts.

d)     More realistic physics and tropics are required for successful 15-60 day forecasts than for 0-7 day mid-latitude forecasts.We may well not know the physics in the actual atmosphere as well as we need to produce successful 15-60 day forecasts.

e)     15-60 day and seasonal forecasts need for model verification and diagnosis a reanalysis consistent with the forecast model.

 

The Global Modeling branch plans to implement a global ocean data assimilation system and coupled model for 15-60 day and seasonal forecasts within the next year, replacing the current operational system that is coupled to a tropical Pacific Ocean analysis/forecast system.The atmospheric portion of the system will be upgraded every 1 to 2 years based on changes to the medium-range model.Significant changes to the MRF will be tested for their effects on seasonal forecasts.If improvement is found, the new version will be frozen and tested extensively for seasonal forecast model upgrade.This will mean that the seasonal forecast model will lag the operational medium-range model by 1-2 years.A similar schedule will be applied to improvements to the ocean model.

 

During the late winter of 2001-2002, operational forecasts of 500 hPa height by the MRF reached 60% anomaly correlation (widely accepted as a measure of the limit of useful skill) at day 8 in the extra-tropical Northern Hemisphere, an improvement of 1 day from 10 years ago.Anomaly correlations for operational week-2 (day 8-14 mean) forecasts for the same period were 50% for the MRF and 55% for the MRF ensemble; for extended periods in the late winter the forecast skill for week-2 exceeded 60%.Such levels of skill reflect the efforts of many skilled scientists; however, the results also suggest that new approaches may be needed to extend forecast skill to 15-60 days.†††††††

 

Fig. 1 compares time-mean errors in zonal mean temperature in one and fifteen-day operational MRF forecasts for Dec. 2001-Feb. 2002 with the error in zonal mean temperature from an integration with observed SST of the operational model at T170 resolution from mid-December 2000.Both are verified against operational analyses.The patterns are quite similar, although differences can be seen near the poles in the troposphere.The magnitudes of errors are comparable in the troposphere, but the longer run shows considerable larger errors in the stratosphere.The tropospheric cold bias is saturated by day 15, but the stratospheric bias continues to grow beyond day 15.

 


Fig. 1 Zonal mean error in temperature for (top) 1 day forecasts (middle) 15 day forecasts and (bottom) integration from mid-Dec. 2000 verified against analyses for Dec. 2001-Feb.2002.

Fig. 2 Zonal mean error in transient eddy kinetic energy from (top) integration from mid-Dec. 2000 for all of 2001 and (bottom) 6 hour forecasts for Dec. 2001 verfied against operational analyses.


 

Fig. 2 compares the zonal mean error in transient eddy kinetic energy in the T170m integration, averaged over 2001, and the same field in 6-hour forecasts for December 2001.Both are verified against zonal mean transient eddy kinetic energy in operational analyses. The patterns are quite similar and indicate that the modelís under-forecast of eddy activity reflects problems that begin in the first 6 hours of integration.Figures 1 and 2 show that problems that develop in long integrations of the NCEP global model clearly can be seen in short-term operational medium-range forecasts and suggest that reducing such errors in short-term forecasts will improve 15-60 day and seasonal forecasts.

 

Realistic air-sea fluxes are essential for useful coupled model forecasts. Several groups are currently studying air-sea fluxes.The WGNE/SCOR Working Group on Air-Fluxes

(http: //www.soc.soton.ac.uk/JRD/MET/WGASF) has recently published a report on validation of air-sea fluxes (Taylor, ed., 2000) and held a workshop on air-sea fluxes (White, ed., 2001).The GEWEX SEAFLUX (http://paos.colorado.edu/~curryja/ocean) working group is comparing estimates of air-sea fluxes and emphasizes the use of satellite observations in determining air-sea fluxes.Peter Gleckler and Jan Polcher (Gleckler and Polcher, 2002) are developing SURFA, an ongoing project to collect surface fluxes from operation NWP analysis/forecast systems and verify them against high-quality in-situ observations over both land and ocean.

 

Global Mean Balances

2001

 

CDAS

GDAS

Kiehl and Trenberth

Range

P (mm/d)

2.84

2.97

2.69

2.69-3.1

E

2.86

3.1

2.69

 

P-E

-.02

-.13

 

 

Sens.heat (W/m2)

15.59

9.31

24

16-27

Latent heat

83.06

89.97

78

78-90

Sfc dsw

205.91

201.13

198

 

Usw

45.11

29.01

30

 

NSW

160.8

172.12

168

142-174

Dlw

336.17

333.6

324

 

Ulw

396.81

397.69

390

 

NLW

60.64

64.09

66

40-72

net rad

100.16

108.03

102

99-119

NHF

+1.5

8.76

0

 

Table 1 Global mean surface water and energy balance in the CDAS and GDAS for 2001 and from a climatology by Kiehl and Trenberth (1997).Range displays the range of independent estimates considered by Kiehl and Trenberth (1997).

 

Table 1 compares global mean surface fluxes averaged over 2001 from CDAS (done with the 1995 NCEP operational global data assimiliation system) and the currently operational GDAS to climatological estimates by Kiehl and Trenberth (1997) and the range of estimates they considered. The large range of estimates is one measure of the uncertainty in our knowledge of air-sea fluxes.†† The hydrological cycle is more intense in the NCEP analyses than in the climatological estimate, but its magnitude lies within the range of independent estimates.Sensible heat flux in the current system is lower than other estimates and may reflect the introduction of a new boundary layer parameterization in the late 1990s.CDAS had too high an oceanic surface albedo; short wave radiation in the current GDAS is closer to the estimates of Kiehl and Trenberth.The current GDAS however has a larger imbalance at the surface than CDAS.

 

††††††††††† Table 2 compares the global mean water and surface energy balance over the oceans from CDAS and GDAS for 2001 to climatological estimates from da Silva et al. (1994), based on COADS observations for 1981-1992, and from the Surface Radiation Budget (Darnell et al., 1992; Gupta et al., 1992), based on retrievals of surface radiation fields from top of the atmosphere satellite observations and ISCCP clouds.The COADS-based estimate produces a large imbalance, indicating a large uncertainty in our knowledge of air-sea fluxes.The SRB net long wave is lower than other estimates (Kiehl and Trenberth, 1997).

 

Global Mean Balances

2001 Ocean

 

CDAS

GDAS

COADS

SRB

P mm/day

3.08

3.24

 

 

E

3.35

3.64

 

 

P-E

-.27

-.40

 

 

Sensible heat W/m2

11.41

5.65

10.1

 

Latent heat

97.01

105.47

88

 

Sfc dsw

198.66

201.56

 

 

Usw

34.85

18.47

 

 

Nsw

163.81

183.09

170.4

173.4

Dlw

353.21

347.35

 

 

Ulw

408.4

409.05

 

 

Nlw

55.19

61.7

49.2

41.9

net rad

108.62

121.39

121.1

131.5

NHF

0.2

10.27

23.3

 

Table 2 Global mean surface water and energy balance over the ocean in the CDAS and GDAS for 2001 and from climatologies by da Silva et al. (1994) and from the Surface Radiation Budget.

 

†††††††††††

Fig. 3(left)compares surface stress along the equator from 12-year climatologies from the NCEP/NCAR reanalysis and from COADS.The reanalysis had too weak zonal surface stress.Fig. 3 (right) shows zonal surface stress along the equator for 2001 from CDAS (the NCEP/NCAR reanalysis), GDAS (the operational analysis), and 3 integrations from mid-Dec. 2000 with observed SSTs. GDAS and the parallel X (PRX) give reasonable zonal surface stress in the east Pacific; two other versions of the model, including the operational seasonal forecast model (SFM)and a different convection (RAS) have too strong stress in the East Pacific.

 



Fig. 3 Annual mean zonal surface stress over the ocean averaged from 5S-5N for (left) the NCEP/NCAR reanalysis and COADS (da Silva et al., 1994) for 1981-92 and (right) GDAS, CDAS and 3 integrations from mid-Dec. 2000 for 2001.

 

††††††††††† Considerable work is needed to improve both the forecast systemís air-sea fluxes and our knowledge of air-sea fluxes.Considerable differences exist between different estimates of air-sea fluxes; our current knowledge of air-sea fluxes is insufficient to close the surface energy budget.Current global forecast systems have problems with cloudiness that produce inaccurate short wave fluxes and problems with moisture that affect long wave fluxes.Low-level oceanic stratus clouds are difficult for current systems to model correctly.

 

References

 

da Silva, A., C.C. Young, and S. Levitus, 1994:Atlas of Surface Marine Data 1994.Vol. 1: Algorithms and Procedures.NOAA Atlas NESDIS 6, U.S. Dept. of Commerce, Washington, D.C., 83 pp.

Darnell, W.L., W.F. Staylor, S.K. Gupta, N.A. Richey, and A.C. Wilber, 1992:Seasonal variation of surface radiation budget derived from International Satellite Cloud Climatology Project C1 Data.J. Geophys. Res., 97, 15741-15760.

Gleckler, P., and J. Polcher, 2002:Project SURFA: A WGNE Pilot Study.WCRP/SCOR Workshop on Inter-comparison and validation of ocean-atmosphere flux field..Bolger Center, Potomac, MD, USA, 21-24 May 2001, WCRP-115, WMO/TD-No. 1083, 6-7.

Gupta, S., W. Darnell, and A. Wilber, 1992:A parameterization of long wave surface radiation from satellite data:Recent improvements.J. Appl. Meteor., 31, 1361-1367.

Kiehl, J. T., and K.E. Trenberth, 1997:Earthís annual global mean energy budget.Bull. Amer. Meteor. Soc., 78, 197-208.

Taylor, P.K., Ed., 2001:Inter-comparison and validation of ocean-atmosphere energy flux fields.Joint WCRP/SCOR Working Group on Air-Sea Fluxes Final Rep., WCRP-112, WMO/TD-No. 1036, 306 pp.

White, G., Ed., 2002:WCRP/SCOR Workshop on Inter-comparison and validation of ocean-atmosphere flux field..Bolger Center, Potomac, MD, USA, 21-24 May 2001, WCRP-115, WMO/TD-No. 1083, 362 pp.