1. NCAR’s CFMIP/COSP contact
Analysis of CAM clouds with COSP (a lot has happened in the last year!)
Jen Kay
NCAR’s CFMIP/COSP contact
Collaborators: Steve Klein, Yuying Zhang, Jim Boyle (LLNL), Ben Hillman, Roj Marchand, Tom Ackerman (UW), Brian Medeiros, Andrew Gettelman, Brian Eaton, Ben Sanderson (NCAR), Robert Pincus (CU)

2. Why are satellite simulators (COSP) useful?
When satellite simulators accurately mimic the observational retrieval process, they enable “apple-to-apple” comparisons between models and observations.
Trusted comparison
Climate Model
Satellite simulator
Satellite retrieval
Observed radiances

3. Is it easy to evaluate CAM clouds with COSP?
Yes!
../models/atm/cam/src/physics/cosp
../models/atm/cam/src/physics/cam/cospsimulator_intr.F90
COSP v1.3 (with local modifications) validated for use with CAM4 and CAM5. Code on CAM trunk and in CESM1 releases. For details, see
Using COSP requires only three simple steps:
1) Configure with cosp. (configure –cosp ….)
2) Set cosp_amwg=.true. in the CAM namelist.
3) Run CAM at least one year, and then use the AMWG diagnostics package to look at COSP outputs.
(Note: Setting cosp_amwg=.true. approximately doubles the CAM run time.)
Andrew says ~4:40 baseline CAM5 on 2 nodes, 2 degree

4. COSP in the AMWG Diagnostics Package (http://www. cgd. ucar
Set now includes COSP plots
Ben Hillman (UW)

5. Cloud Feedbacks Model Intercomparison Project
NCAR run progress (years complete/years planned)
Simulation
CAM4
CAM4(ext)
CAM5
CAM5(ext)
AMIP
30/30
4/4
20/30
0/4
AMIP(4XCO2)
21/30
AMIP(+4K)
18/30
AMIP(patt)
Control SST -
0/30
Control SST (4XCO2)
Control (coupled)
120/120
CO2ramp (coupled)
105/105
Courtesy: Ben Sanderson (NCAR)

6. Paper on COSP-enabled evaluation of CAM clouds for CESM Special Issue
We analyze ten-year 1 degree AMIP (prescribed SST/ice) CAM4 and CAM5 runs. CAM4 and CAM51 degree release code. Tag cam5_0_55.
Kay, J. E., Hillman, B., Klein, S., Zhang, Y., Medeiros, B., Gettelman, G., Pincus, R., Eaton, B., Boyle, J., Marchand, R. and T. Ackerman (2012): Exposing global cloud biases in the Community Atmosphere Model (CAM) using satellite observations and their corresponding instrument simulators, J. Climate, in press
(available at:

7. COSP-enabled total cloud fraction comparisons
Figure 2. Global maps of observed and COSP-simulated annual mean total cloud fraction: a) CALIPSO GOCCP observations (SR > 5), b) ISCCP observations (>0.3), c) MISR observations (>0.3), d) CAM4 CALIPSO, e) CAM4 ISCCP, f) CAM4 MISR, g) CAM5 CALIPSO, h) CAM5 ISCCP, and i) CAM5 MISR. The value after each panel name reports the global annual mean, while the values in parentheses report the global annual mean model bias (CAM - observations) and model RMSE respectively. Because they do not include thin or broken clouds, the global annual mean cloud fractions observed by MODIS (0.49) and CloudSat (0.50) are significantly smaller than those observed by CALIOP, MISR, or ISCCP, and are not included in this plot. The solid gray line in each panel shows the GCSS Pacific Cross Section.
Observational Uncertainty < Model Bias: CAM4 bias > CAM5 bias, Kay et al. (2012)

8. COSP enables evaluation of cloud amount by height and optical depth
credit Swati Gehlot (DWD)

9. COSP-enabled comparisons robustly show that the CAM5 physics has reduced long-standing climate model cloud biases (too many optically thick clouds, too few clouds in CAM4 and many other models, see Zhang et al. 2005).
How do CAM4 and CAM5 produce similar annual mean cloud forcing biases (Figure 1b-f) with such large differences in their annual mean total cloud fractions (Figure 2d-i)? The answer lies in the relationship between cloud fraction and . Figure 3a shows observed and simulator-diagnosed global annual mean  distributions for MISR, MODIS, and ISCCP. This figure shows CAM4 and CAM5 achieve a similar radiative balance because they have compensating differences in the amounts of optically thin and optically thick cloud. Consistent with earlier ISCCP-simulator based evaluations, such as the 10 model inter-comparison study by Z05, CAM4 has too much optically thick cloud and not enough optically thin cloud, resulting in a flatter  distribution than is observed. In contrast, CAM5 has a steeper  distribution that agrees quite well with observations. To the best of our knowledge, CAM5 is the first climate model to match the observed  distribution with this level of fidelity, which is a significant achievement. The reproduction of observed cloud albedos, which are largely determined by , is an important metric of model performance.
Figure 3. Global column-integrated cloud optical depth () distributions: a) MISR, MODIS, and ISCCP from satellite observations, CAM4, and CAM5, b) ISCCP from CAM4 and CAM5 at both 0.9x1.25 and 1.9x2.5 horizontal grid resolutions. Below =3.6, there are large inter-satellite differences in cloud fraction due to difference in the detection and treatment of cloud-edges and sub-pixel clouds. Observational agreement below =1.3 is especially poor (see P11 Figure 6), and is therefore not plotted in a).
Kay et al. (2012)

10. Improved Arctic cloud seasonal cycle in CAM5 (despite known low aerosol issues…)
Kay et al. (2012)

11. CAM5 has improved clouds and increased sensitivity to 2xCO2 forcing… both globally and in the Arctic
Kay et al. (2012): The influence of local feedbacks and northward heat transport on the equilibrium Arctic climate response to increased greenhouse gas forcing in coupled climate models, J. Climate

12. Snow has a large impact on CAM5 COSP diagnostics

13. Important biases remain in both CAM versions (e. g
Important biases remain in both CAM versions (e.g., low cloud deficit in transition from stratocumulus to deep convection)
Figure 5. Global maps of annual mean cloud fraction bias (CAM - CALIPSO GOCCP observations): a) CAM4 CALIPSO low cloud fraction bias, b) CAM5 CALIPSO low cloud fraction bias, c) CAM4 CALIPSO mid cloud fraction bias, d) CAM5 CALIPSO mid cloud fraction bias, e) CAM4 CALIPSO high cloud fraction bias, and f) CAM5 CALIPSO high cloud fraction bias. The value after each panel name reports the global annual mean bias and the value in parentheses reports the RMSE. The solid gray line in each panel shows the GCSS Pacific Cross Section.
Figure 5, Kay et al. (2012)

14. Summary: 1) COSP and CFMIP-requested diagnostics are validated and ready to use within CESM. 2) Analysis using COSP is beginning (and documenting large improvements in clouds from CAM4 to CAM5). COSP/CFMIP help address key climate questions for the AMWG and the larger climate community, such as… How do we know if we have the clouds “right”?

15. EXTRA

16. How does COSP work?
COSP contains satellite simulators for both passive (MODIS, MISR, ISCCP) and active (CloudSat radar and CALIPSO lidar) observations.

17. For COSP metrics, CAM5 > CAM4.
Kay et al. (2012)