1. Aaron Kennedy, Xiquan Dong, and Baike Xi University of North Dakota
A Comparison of Atmospheric State, cloud, radiation and precipitation between NARR, MERRA, and ARM Forcing
Aaron Kennedy, Xiquan Dong, and Baike Xi
University of North Dakota
Shaocheng Xie, DOE PNNL
Junye Chen, NASA GSFC

2. Motivation and Goals
NARR and MERRA Reanalysis datasets are widely used for a variety of studies, however, their uncertainties are not quite clear.
In this study, we compare these two popular reanalysis datasets with ARM forcing over SGP from 1999 to 2001
Atmospheric State (RH, vertical motion)
Radiation fluxes (Surface)
Cloud Fraction
Precipitation
ARM continuous forcing developed from RUC-2 model and with ARM observed surface/TOA fluxes and precipitation as constraints, which ensures energy budget in balance.

3. Domain/Methodology
Data averaged for GCM sized box (2°×2.5°) centered on the ARM SGP
NARR
210 km horizontal resolution, 28 vertical levels, and 3h temporal
MERRA
2D 2/3°×1/2°, 1h
Atm. state
3D: 1.25°×1.25°, 6h
Everything else

4. 3-yr averaged Atmospheric State (RH)
ARM NARR MERRA
Similar patterns for mean RHs in troposphere
ARM slightly moister in boundary layer
NARR and MERRA moister in upper troposphere during different months compared to ARM RHs

5. Atmospheric State (ω) ARM NARR MERRA
(-) upward motion, (+) downward motion
Significant differences between all datasets
ARM has large upward maximum in upwelling during late spring
NARR and MERRA have weaker upward motion during late spring and large subsidence during summer.
The model simulations could be significant difference based off the different forcing inputs.

6. Correlations with Cloud Fraction
Omega RH
MAX, MEAN
ARM:
-0.47,-0.37
NARR:
-0.34,-0.27
MERRA:
-0.39,-0.33
MAX, MEAN
ARM:
0.59,0.48
NARR:
0.56,0.44
MERRA:
0.56,0.49
Omega: Highest correlations for ARM forcing, then followed by MERRRA and finally NARR
RH: ARM slightly higher, but virtually the same for NARR/MERRA

7. Radiative Fluxes (Surface SW)
NARR
Both NARR and MERRA have very high correlation ( ) with ARM data.
However, both NARR and MERRA have biases as follow: MERRA vs. NARR SWUP (-4 vs. 37 W m-2)
SWDN (29 vs. 74 W m-2)
Due partially to negative bias for MERRA at low solar zenith angles (see diurnal cycle on next slide)
SWUP
SWDN
MERRA

8. Diurnal cycle of surface SW down
(240)
(192)
(212)

9. Radiative Fluxes (SFC LW)
The correlations for LW fluxes (0.97) are even higher than SW fluxes
MERRA has lower bias for LWUP (1 vs. 11 W m-2) and higher bias for LWDN (-18 vs. -8 W m-2)
MERRA has less tail at high values of LW flux.
NARR
MERRA
LWUP
LWDN

10. Cloud Fraction
Total
Middle
Low (>700mb)
High (<400mb)
ARM observations plotted with min/max overlaps (two blue lines)
Max. overlap is probably closer to reality for low clouds layer
NARR underestimates CF at all levels (red line)
MERRA has reasonable total CF (green line)
High CF in decent agreement except during summer (July-Aug)
Low and mid level CFs lower than observations

11. Precipitation NARR in excellent agreement with ARM
Monthly
Daily
NARR
NARR in excellent agreement with ARM
MERRA underestimates many high precipitation events leading towards negative bias at monthly timescale
MERRA

12. Monthly Precipitation
01/1999
12/2001
Despite negative bias, MERRA captures monthly variability of precipitation well.

13. Summary
Compared to the 3-yr ARM SGP observations, we have the following results for NARR and MERRA
1) Atmospheric State:
RH: Both NARR and MERRA are similar to ARM
ω : Significant differences between the three datasets
2) SW and LW fluxes:
MERRA has smaller bias than NARR
3) Cloud Fraction:
MERRA agrees reasonably well with ARM, but
NARR underestimates CF
4) Precipitation:
NARR has an excellent agreement with ARM, but MERRA underestimates precipitation