Analysis of Sea Ice Fraction in the Community Earth Climate Model

Justin PerketAOSS 586

Motivation

Background in physics & chemistry Want to focus research on improving

microphysics modelling Goal to gain great understanding of how to

handle model/data output

Sea Ice and Observational Data

Dagmar Budikova, Journal of Global & Planetary Change. 68,3. 2009, pg. 149-163.( http://www.sciencedirect.com/science/article/pii/S0921818109000654 )

Understanding the interaction between sea ice conditions and climate requires accurate records of field conditions throughout Arctic with variation documented over multiple decades and seasons.

Reliable records are essential in driving and evaluating the results from atmospheric and coupled General Circulation Models and conducting observational studies.

Satellite technology has enabled conditions to be monitored continuously for since 1978.

Nimbus-7 Scanning Multichannel Microwave Radiometer (SMMR) until 1987

Special Sensor Microwave Imager (SSM/I)

Before the records of sea ice conditions consist primarily of aircraft, ship, and coastal observations

scattered locations, irregular time intervals

Sea Ice and Observational Data

Dagmar Budikova, Journal of Global & Planetary Change. 68,3. 2009, pg. 149-163.( http://www.sciencedirect.com/science/article/pii/S0921818109000654 )

Significant differences among the various sea ice datasets have been documented posing multitude of challenges for effective comparisons of results from studies that utilize them

Continuous records often derived from diverse sources and from different algorithms

Singarayer et al. (2005) compares records of sea ice cover simulated from 3 monthly sea ice climatologies:

NASA's satellite data algorithm

Bootstrap derived PMR datasets,

National Ice Center records

Found the uncertainty in the estimation of sea ice cover to be a combination of random and systematic errors, both temporarily and spatially dependent.

Sea Ice and GCMs

General Circulation Models can predict impact on climate from variations in polar sea ice

Realistic simulations under various polar sea ice conditions require fully coupled ocean–atmosphere models

Captures complex interactions between the atmosphere, oceans,and sea ice conditions.

Simultaneous evolution of atmospheric and oceanic conditions

Challenge of isolating the precise contribution of ocean and atmosphere to the sea/atmosphere interaction.

Determining the nature and strength of the ocean's back interaction on the atmosphere remains a challenge

Example: response to a sea surface temperature or sea ice anomaly can provide a significant signal at the 500 hPa level

But signal usually much smaller than natural variability,making detection difficult in GCM integrations.

Sea Ice and CCSM

Sea ice component in CCSM configuration is the Climatological Data Ice Model

interacts with the CCSM Coupler, but is not an active model Rather, it takes ice fraction data from input data,

infers an ice extent, and sends this data to coupler Ignores any forcing data received from the coupler Useful for seeing how an active atmosphere

component behaves when coupled to climatological ice extent.

http://www.cesm.ucar.edu/models/ccsm3.0/dice6/

Sea Ice Data Used

Ice Frac from CAM2, 1° resolution 20th Century Ensemble Member #6 (MOAR) 1850-2005, monthly averages 1981-2005 with observations

Sanity Check: Global Annual Surf. Temp. Comparison

Source: Goddard Institute for Space Studies Surface Temperature Analysis: http://data.giss.nasa.gov/gistemp/

Checked differences with T-test (α=0.05)

Ave. Ice Fraction

Ice Extent

Ice Extent Diagnostic

http://www.cgd.ucar.edu/cms/rneale/tools/amwg_diagnostics.html

Ice Fraction Diagnostic

Comparison with Historical DataWalsh and Chapman Northern Hemisphere Sea Ice

Data Set http://www.cgd.ucar.edu/cas/guide/Data/walsh.html

Northern Hemisphere, 1870-2008, annual & seasonal averages

Sources of Data:1. Danish Meteorlogical Institute

2. Japan Meteorological Agency

3. Naval Oceanographic Office (NAVOCEANO)

4. Kelly ice extent grids (based upon Danish Ice Charts)

5. Walsh and Johnson/Navy-NOAA Joint Ice Center

6. Navy-NOAA Joint Ice Center Climatology

7. Temporal extension of Kelly data (see note below)

8. Nimbus-7 SMMR Arctic Sea Ice Concentrations or DMSP SSM/I Sea Ice Concentrations using the NASA Team Algorithm

http://arctic.atmos.uiuc.edu/cryosphere/IMAGES/seaice.area.arctic.png

Comparison with Historical Data

Problem: CCSM data independent of Walsh & Chapman data

Dummy Test: Autocorrelation of North Ice Extent

Coherence between North & South Ice Extents

%% Lag Corr.-North & South Ice Frac.clear X; clear b; clear lags; clear xc; b=length(Iexts)-600;for n = 1:b x= Iextn(n:n+600); y =Iexts(n:n+600); [xc,lags] = xcorr(x,y,24,'coeff'); X(:,n) = xc;end[c,h] = contourf(1:b,lags,X); colorbar;xlabel('Month'); ylabel('Lags');

Ice Fraction and Surface Temp.

2D correlation between Ice Fraction & Surf Temp Grids

Normalized

Deasonalized

Coherence of Desaonalized Data

Code Used%% midterm scripts

%% Global Surf Temp.

l=length(TSglob);monind=reshape( (1:l)',12,l/12)'; % ea. col. has indices for a month% subtract monthly variability:for i=1:12 TSmonmean(:,i)=mean(TSglob(monind(:,i))); %mean for each month TSdese(monind(:,i))=TSglob(monind(:,i))-TSmonmean(:,i);end

plot(giss(:,1),giss(:,2),giss(:,1),giss(:,3),time2,TSdese)

%% Global Average Icefracsubplot(2,1,1);plot(time2,Igs*100,1850:2005,Ibs*100);title('Northern Hemisphere Average Ice Fraction'); ylabel('% Ice');subplot(2,1,2); plot(time2,Ign*100,1850:2005,Ibn*100);title('Southern Hemisphere Average Ice Fraction'); ylabel('% Ice');

%% Ice Extent%[Iextn1,Iexts1]=hemisum(ICEFRAC,lat,lon); % With all Icefrac[Iextn,Iexts]=hemisum(ICEFRAC,lat,lon); % With ICEFRAC > 0.15for k=1:(length(Iexts)/12) %get annual means a=12*k-11; b=12*k; IextsY(k) = mean(Iexts(a:b) ); IextnY(k) = mean(Iextn(a:b) );% Iexts1Y(k) = mean(Iexts1(a:b) );% Iextn1Y(k) = mean(Iextn1(a:b) );end

% plot( ... % 1850:2005,IextsY, ... % 1850:2005,IextnY, ...% seaiceextent(1:end-2,1),seaiceextent(1:end-2,2)*10^6, ...% seaiceextent(1:end-2,1),seaiceextent(1:end-2,3)*10^6, ...% seaiceextent(1:end-2,1),seaiceextent(1:end-2,4)*10^6, ...% seaiceextent(1:end-2,1),seaiceextent(1:end-2,5)*10^6, ...% seaiceextent(1:end-2,1),seaiceextent(1:end-2,6)*10^6 ...% )% %1850:2005,Iexts1Y, ...% %1850:2005,Iextn1Y, ...% %)% legend('Ice Extent North','Ice Extent Sorth','annual mean','winter','spring','summer','autumn');

subplot(2,1,1); plot(time2,Iexts*1e-6,1850:2005,IextsY*1e-6); ylabel('millions km^2');title('Northern Hemisphere Ice Extent'); legend('monthly','yearly');subplot(2,1,2); plot(time2,Iextn*1e-6,1850:2005,IextnY*1e-6); title('Sorthern Hemisphere Ice Extent'); legend('monthly','yearly'); ylabel('millions km^2');

%%plot(1850:2005,IextsY,seaiceextent(1:end-2,1),seaiceextent(1:end-2,2)*10^6);h=ttest2(seaiceextent(1:136,2),IextsY(21:156))

%% Ice Frac. Annual Cycle

for i=1:12 Iextsmm(:,i)=mean(Iexts(monind(:,i))); %mean for each month Iextnmm(:,i)=mean(Iextn(monind(:,i))); %mean for each monthendplot(1:12,Iextsmm*1e-6,1:12,Iextnmm*1e-6);title('Annual Cycle Hemispehre Sea Ice Extent'); legend('North','South');

Code Used%% Autocorrelationplot(...xcov(... sin(1:100)+2*sin(1:100)+0.6*randn(1,100) ... ) ... )

%% Cross corr.% [xc,lags]=xcorr(Igs,Ign,120);% plot(xc);

% compute autocorr:r=xcorr(Igs, 240, 'coeff');r=r(26:27); % keep r for 1-2 month time laga=( r(1)+sqrt(r(2)) )/2; % est. the AR-1 coeff.W=240;[pw,f]=pwelch(Igs,48,24,length(Igs)) %,W,[],W/2,[]); % format is pwelch(data, window length,... % noverlap, # points in FFT, [] ) pest = (1-a^2)./(1-2*a*cos(2*f*pi)+a^2); %compute the AR-1 power spectrumDOF= 1.2*length(Igs)/length(f); %degree of freedeom, f_e=1.2;

% alternative:% ACF=autocorr(SOInorm,24);% acoef=aryule(SOInorm,1);

% use f statistic% H_0 = spectral peak > AR-1 spectral val. <- 1-side est., not 2 side.fp=icdf('f', 0.95, floor(DOF), floor(DOF));p95 = pest*fp;figure;plot(f,pw, 'o-', f,pest,'-',f,p95,'r-');fval=pw./pest;pval=1-fcdf(fval,floor(DOF),floor(DOF));

%% Lag Corr.-North & South Ice Frac.clear X; clear b; clear lags; clear xc;%X = zeros(81,2172);b=length(Ibn)-120;for n = 1:b x= Ign(n:n+400); %running window of 20 years y =Igs(n:n+400); [xc,lags] = xcorr(x,y,24,'coeff'); X(:,n) = xc;end[c,h] = contourf(1:b,lags,X); colorbar;xlabel('Year'); ylabel('Lags');figure;plot(time2(1:48),Ign(1:48),time2(1:48),Igs(1:48));plot(time2,Ign,time2,Igs)%%[y,lags]=xcorr(Ign,Igs);R=fliplr(y(1:24));stem(y,lags)