Constraining Model Predictions of Arctic Sea Ice With Observations. Chris Ander 27 April 2010 Atmos 6030

Transcription

1 Constraining Model Predictions of Arctic Sea Ice With Observations Chris Ander 27 April 2010 Atmos 6030

2 Main Sources Boe et al., 2009: September sea-ice cover in the Arctic Ocean projected to vanish by 2100 Collins, 2009: Insight despite imperfection

3 The Problem? Coupled Atmosphere-Ocean Global Climate Models (AOGCM s) significantly under-estimate the rate of sea ice decline in the 20 th Century and beyond. At issue is the prediction of September Sea Ice Extent (SSIE) and when/if the Arctic will become ice free in the 21 st Century. Still some uncertainties in the models i.e. complex feedbacks, lack sufficient understanding/ability to model certain physical processes.

4 How is Sea-Ice Observed? Satellite passive microwave imagers. Detect surface emissivity, E sea ice > E water. Nimbus-7 SSMR and DMSP SSM/I. Sea Ice Extent= sum of the area of grid cells with at least 15% ice concentration. (1 grid). Monthly mean sea-ice extent in millions of square kilometers. Data generally available from 1979-present. The Satellite Era.

5 What are the models showing? Boe et al., Raw output from 18 different CMIP3 GCM s -Reference sea ice extent is mean SSIE from ( Satellite Era ) -Forced with SRES A1B emissions scenario medium forcing ie CO2 concentrations of 700 ppm by 2100

6 Model Predictions and Current Trends Boe et al., 2009 find that models with the fastest present-day ice loss also predict the fastest decline in future sea ice extent. Consistent with physical reasoning; suggests that the ice-albedo feedback plays an important role.

7 Ice-Albedo Feedback

8 Feedback Evident in AOGCM s -Probabilistic results from 21 CMIP3 models compared to climatology -Values of temperature increase with an 80% chance of occurrence. Meehl, G.A. et al., 2007

9 Relationship between SSIE changes and sea-ice cover properties Solid line=correlation between mean % of remaining ice in September and simulated SSIE trends from Dashed line= % of remaining ice correlated with % of thin ice during the period. -Negative correlation therefore it is multiplied by -1 -Correlations weaken with time -Some of the weakening could be an artifact of the models -Mean of 18 CMIP3 AOGCMs

10 Applying Adjustments to Models Boe et al., 2009 Left: models with strongest declines also have lowest simulated SSIE in the future. Right: Boe et al. approach shows even faster decline compared to the multi-model mean.

11 Predictions based on this approach Estimate to be the first 20-year period with ice-free conditions. Uncertainty increases w/time but remains less than the spread of the multi-model mean. 16% probability of 20% of baseline ( starting ) sea-ice cover remaining in ; same probability that ice disappears completely by

12 Observed trends have a noticeable effect especially important, record minimum in Arctic Sea-Ice observed. Shading=sensitivity to models included (all possible combinations of 15,16,17 of 18 models r,y,b). Sensitivity to Observed Trends Boe et al., 2009

13 2007 Arctic Sea Ice Minimum

14 Improving Feedbacks in AOGCMs Hall and Qu, 2006 found large intermodel variations in snow-albedo feedback strength. Variations in snow-albedo feedback strength with the seasonal cycle strongly correlated with variations under climate change scenarios. Measure observed feedback strength in past seasonal cycles and compare to simulated -> adjust GCM as needed.

15 Snow Albedo Feedback Equation Q/ Ts= change in incoming shortwave radiation with respect to changes in surface temperature. αp/ αs= change in planetary albedo (clouds, atmosphere, etc) WRT changes in surface albedo (snow, ice, trees, etc). Small compared to last term. αs/ Ts= change in surface albedo WRT changes in surface temperature. -> Surface Temperature is well measured. -> Surface Albedo is not->uncertainties! I=incoming solar at top of atmosphere. Hall and Qu, 2006

16 Seasonal Cycle vs Climate Change r2 = 0.92 Calculated αs/ Ts from CMIP3 model solutions climate change. ERA40 and ISCCP satellite data seasonal cycle. Monthly mean differences May-April->feedback strongest in spring. Hall and Qu, 2006

17 Other Factors Cloud cover: is feedback positive or negative in the Arctic? -Increased cloud cover increased downward LW radiation in winter amplified ice-albedo effect. -Increased cloud cover decreased downward SW radiation in summer and at lower latitudes dampened ice-albedo feedback. Ocean heat transport Holland & Bitz, 2003

18 Conclusions AOGCM s generally underestimate the rate of sea ice loss in the Arctic. Physically based relationships can be developed between past observations and future climate variables to help constrain climate model predictions. Results will improve as computing power and parameterizations continue to improve. Observation constrained approaches tend to reduce uncertainty in the model outcomes.

19 References Boe, J., Hall, A., Qu, Z. Nature Geosci 2, (2009). Collins, M., Nature Geosci 2, (2009). Hall, A. & Qu, X. Using the current seasonal cycle to constrain snow albedo feedback in future climate change. Geophys. Res. Lett. 33, L03502 (2006). Holland, M. M., and C. M. Bitz (2003), Polar amplification of climate change in coupled models, Clim. Dyn., 21, Meehl, G. A. et al. The WCRP CMIP3 multimodel dataset: A new era in climate change research. BAMS 88, (2007). Wang, M. & Overland, J. E. Geophys. Res. Lett. 36, L07502 (2009). National Climatic Data Center < National Snow and Ice Data Center <

20 Questions?