1. Arctic Sea Ice, Clouds and Atmosphere Interactions from Satellite Observations
Yinghui Liu1, Jeff Key2, and Xuanji Wang1
1CIMSS, University of Wisconsin-Madison
2NOAA/NESDIS, Madison, Wisconsin USA
Earth Observation and Cryosphere Science, November 2012, Frascati, Italy

2. Arctic surface temperature has been changing
Arctic-wide annual averaged surface air temperature anomalies (60°–90°N) based on land stations north of 60°N relative to the 1961–90 mean. Data are from the CRUTEM 3v dataset (Overland et al., 2011 Arctic Report Card)
IABP/POLES ( )
Rigor et al. (2000)

3. Arctic cloud cover has been changing
Spring
Winter
Time series of seasonal cloud amount from APP-x and TOVS (left). Trends in cloud amount in spring (right, upper), and winter (right, bottom) from TOVS, (Schweiger 2004).

4. Sea ice has been changing
Monthly October ice extent for 1979 to 2012 shows a decline of -7.1% per decade (NSIDC).
The image above shows the different distribution of ice extent at the time of the September 2012 minimum, compared to the September 2007 minimum. Dark gray indicates where ice extent was present only in 2007; white indicates where ice extent was present only in 2012; and light gray shows where ice extent was present in both 2007 and 2012.

5. The Arctic climate system is complex
Wiring diagram for the atmospheric component of the Arctic hydrologic system. Blue (yellow) hubs are drivers (recipients), red (blue, black) arrows denote interactions of the same (opposite, competing) sign (Francis et al. 2009).
What are the relationships between changes in Arctic surface temperature, sea ice, and cloud cover?

6. Data: part 1 Dataset Extended AVHRR Polar Pathfinder (APP-x)
Cloud mask
Surface temperature under all-sky conditions
Other parameter, not used here: surface albedo, cloud properties, radiative fluxes, ice thickness
Sea ice Concentration from SMMR and SSM/I with NASA team algorithm
Sea ice concentration
Derived parameters
Seasonal means and trends of surface temperature under clear, cloudy, and all-sky conditions
Seasonal means and trends of cloud amount
Seasonal means of surface temperature over water and ice
Seasonal trends of sea ice concentration

7. Winter Spring Summer Autumn Annual
APP-x: Trends in surface temperature
Winter
Spring
Winter
Spring
Summer
Autumn
Annual
-0.36oC/dec
0.68oC/dec
0.70oC/dec
0.45oC/dec
0.34oC/dec
Summer
Autumn
APP-x all-sky surface temperature trends,
Surface temperatures from satellites show cooling trends in winter and warming trends in other seasons.

8. APP-x: Trends in cloud cover
APP-x cloud cover trends,
Spring
Winter
Summer
Autumn
-3.43%/dec
2.30%/dec
0.50%/dec
0.02%/dec
-0.24%/dec
Cloud amount from satellites show; decreasing clouds in winter, increasing in spring.

9. SSM/I: Sea ice concentration trend
Winter
Spring
Summer
Autumn
Seasonal mean sea ice concentration trends from 1982 to 2004.
Sea ice concentration from satellites show decreasing sea ice concentration in summer and autumn.

10. Trend Partitioning
Quantitative determination of the effect of trends in cloud cover and sea ice on the trend in surface temperature:
A: cloud effect
B: sea ice effect
Average surface temperature is a function of clear and cloudy temperatures and the area fractions of cloud and ice
C: residual trend
The fine print: Changes in cloud cover are assumed to be independent of changes in sea ice. We’ll address that later.

11. Results: Cloud effect on surface temperature trend
Winter
Spring
Summer
Autumn
Surface temperature trends introduced by cloud amount changes (K/decade)
In winter, the cloud cover trend explains out of -1.2 K decade-1 of the cooling. In spring, it is 0.55 of the total 1.0 K decade-1 warming.

12. Results: Sea ice effect on surface temperature trend
Winter
Spring
Summer
Autumn
Surface temperature trends introduced by sea ice concentration changes (K/decade).
In the Chukchi and Beaufort Seas in autumn, surface warming due to changes in sea ice accounts for 0.9 K decade-1 of the total 1.1 K decade-1 warming trend over the Chukchi and Beaufort Seas.

13. Challenges and Opportunities
What other parameters contribute to the residual trend?
Are changes in cloud independent of changes in sea ice?
Residual trends as the difference of the total trend and trends introduced by changes in cloud amount and sea ice concentration.

14. Are changes in cloud independent of changes in sea ice?
Qualitatively, an increase in cloud amount tends to correspond to a decrease in SIC and ice extent.
(a) Difference of Sept-Oct 2007 MODIS cloud amount and the Sept-Oct mean, (b) AMSR-E SIC anomalies for 2007 based on the Sept-Oct mean.

15. Sea ice and Cloud: Data and Methods
Datasets
MODIS (Moderate-resolution Imaging Spectroradiometer) Level-3 atmospheric daily global product, including daily mean cloud amounts for each 1 degree by 1 degree cell,
Sea ice concentration (SIC) with spatial resolution of 25 km from the Special Sensor Microwave/Imager (SSM/I),
Methods
The Equilibrium Feedback Analysis (EFA) is used to assess the local “feedback” of sea ice concentration on cloud amount. The Monte Carlo method is used to calculate the confidence level.
Cov is covariance. Tau must be larger than dt. The first equation is not used explicitly in the rest of this analysis. The decorrelation time (tau), or “memory”, of cloud over the Arctic Ocean is around two weeks. A two-week memory for clouds might be due to persistent Arctic stratus. The decorrelation time is (1+a)/(1-a), where “a” is the autocorrelation of the weekly cloud amount anomalies with a one-week lag.
EFA has been extensively used to assess the local feedback of SST on the overlying atmospheric fields, and the feedback of vegetation on climate (Frankignoul et al. 1998, Liu et al. 2006, 2008, and etc.).

16. Sea ice and Cloud: Feedback Coefficient
Feedback coefficient of sea ice concentration on cloud cover (a) from July to November, and (b) from January to December. Only those significant at the 90% confidence level are shown.

17. Area averaged feedback of sea ice on cloud amount
Feedback parameter
Autumn average
Annual average
Beaufort Sea
-0.47
-0.33
Chukchi Sea
-0.39
-0.37
Central Arctic
-0.45
Laptev Sea
-0.36
Barents Sea
-0.27
Pacific Section
-0.41
Overall
Over Beaufort and Chukchi Seas, a 1% decrease in SIC leads to % increase in cloud amount in different seasons.

18. Summary The trend partitioning scheme shows that
In winter, the cloud cover trend explains out of -1.2 K decade-1 of the cooling. In spring, it is 0.55 of the total 1.0 K decade-1 warming.
In the western Arctic Ocean in autumn, warming due to changes in sea ice accounts for 0.9 K decade-1 of the total 1.1 K decade-1 warming.
The feedback (response) parameter showed that a 1% decrease in sea ice concentration corresponds to a % increase in cloud amount in Beaufort, Chuckchi, and Laptev Seas.
Changes in cloud cover, sea ice, and surface temperature are closely related. Their internal feedback and interactions with other climate processes need further investigation.