top of page

% Reference: "False alarms: How early warning signals falsely predict abrupt sea

% ice loss",

% T.J.W. Wagner & I. Eisenman (2015), Geophys. Res. Lett., 42, 066297 

%

% This script numerically solves the system described by eqn (1) in the 

% article above.

%

% The full model description is given in "How Model Complexity Influences

% Sea Ice Stability", 

% T.J.W. Wagner & I. Eisenman, J Clim (2015), henceforth WE15.

%

% The present model is equivalent to that of WE15, but with the added

% feature of (reddened) stochastic noise forcing.

%

% Till Wagner and Ian Eisenman, November 2015

%

% Minor bug fix, Jan 2022: in eq.A1 S(:,i)->S(:,i+1)

% (and accordingly repeated 1st column at end of S array)

% ----------------------------------------------

% ----------------------------------------------

n   = 200;   %grid resolution

dur = 350;   %duration of simulation

sig = 0.5;   %noise amplitude

Fdef= 0;     %initial Forcing level 

             %(F = 0 corresponds roughly to pre-industrial levels)

spinup = 50; %start ramping after 'spinup' years

%%Model parameters (WE15, Table 1 and Section 2d) -------------------------

D  = 0.6;     %diffusivity for heat transport (W m^-2 K^-1)

S1 = 338;     %insolation seasonal dependence (W m^-2)

A  = 193;     %OLR when T = T_m (W m^-2)

B  = 2.1;     %OLR temperature dependence (W m^-2 K^-1)

cw = 9.8;     %ocean mixed layer heat capacity (W yr m^-2 K^-1)

S0 = 420;     %insolation at equator  (W m^-2)

S2 = 240;     %insolation spatial dependence (W m^-2)

a0 = 0.7;     %ice-free co-albedo at equator

a2 = 0.1;     %ice=free co-albedo spatial dependence

ai = 0.4;     %co-albedo where there is sea ice

Fb = 4;       %heat flux from ocean below (W m^-2)

k  = 2;       %sea ice thermal conductivity (W m^-2 K^-1)

Lf = 9.5;     %sea ice latent heat of fusion (W yr m^-3)

cg = 1e-3;    %ghost layer heat capacity(W yr m^-2 K^-1)

tau = 3e-6;   %ghost layer coupling timescale (yr)

%%time stepping

nt = 1e3;

dt = 1/nt;

%%Spatial Grid ------------------------------------------------------------

dx = 1/n;  x = (dx/2:dx:1-dx/2)';  %native grid

%%Diffusion Operator (WE15, Appendix A) -----------------------------------

xb = (dx:dx:1.0-dx)';  lambda=D/dx/dx*(1-xb.*xb);

a=[0; -lambda]; c=[-lambda; 0]; b=-a-c;

diffop = - diag(b) - diag(c(1:n-1),1) - diag(a(2:n),-1);

%%Definitions for implicit scheme on Tg

cg_tau = cg/tau; dt_tau = dt/tau; dc = dt_tau*cg_tau;

kappa = (1+dt_tau)*eye(n)-dt*diffop/cg;

%%Seasonal forcing (WE15 eq.3)

ty = dt/2:dt:1-dt/2;

S=repmat(S0-S2*x.^2,[1,nt])-repmat(S1*cos(2*pi*ty),[n,1]).*repmat(x,[1,nt]);

S=[S S(:,1)];

%%Further definitions

M = B+cg_tau; 

aw= a0-a2*x.^2; 

kLf = k*Lf;

%%Set up output arrays, saving 100 timesteps/year

E100 = zeros(n,dur*100); 

T100 = zeros(n,dur*100); 

%%ramping rate

dF = 1/(20*nt);

%%Initial conditions ------------------------------------------------------

T = 10*ones(n,1);

Tg = T; E = cw*T;

%%noise timeseries

sig_noise = sig/sqrt(dt);

noise = sig_noise*randn(1,dur*nt);

lp = 1/52;  %1 week 'decorrelation' time

alpha = exp(-dt/lp); nalpha = sqrt(1-alpha^2);

N_red = noise*0;

N_red(1) = noise(1);

for i=2:length(noise)

    N_red(i)=alpha*N_red(i-1)+nalpha*noise(i);

end

%%run the model -----------------------------------------------------------

p = 0; m = 0; N = 0; F = Fdef;

%%Integration (see WE15_NumericIntegration.pdf)----------------------------

%%Loop over Years ---------------------------------------------------------

tic

for years = 1:dur

    %Loop within One Year-------------------------------------------------

    for i = 1:nt

        m = m+1;

        if years > spinup  %start ramping and noise after spin up

            F = F+dF;

            N = N_red(m);

        end

        %store 100 timesteps per year

        if (p+1)*nt/100 == m

            p = p+1;

            E100(:,p) = E;

            T100(:,p) = T;

        end

        % forcing

        alpha = aw.*(E>0) + ai*(E<0);        %WE15, eq.4

        C =alpha.*S(:,i)+cg_tau*Tg-A+F+N;

        % surface temperature

        T0 =  C./(M-kLf./E);                 %WE15, eq.A3

        T = E/cw.*(E>=0)+T0.*(E<0).*(T0<0);  %WE15, eq.9

        % Forward Euler on E

        E = E+dt*(C-M*T+Fb);                 %WE15, eq.A2

        % Implicit Euler on Tg

        Tg = (kappa-diag(dc./(M-kLf./E).*(T0<0).*(E<0)))\ ...

            (Tg + (dt_tau*(E/cw.*(E>=0)+(ai*S(:,i+1) ...

            -A+F+N)./(M-kLf./E).*(T0<0).*(E<0))));        %WE15, eq.A1

    end

    yrs = sprintf('year %d complete',years); disp(yrs)

end

toc

%%--------------------------------------------------------------------------

%compute ice edge

xi = ones(1,p);

for i = 1:p

    if any(E100(:,i)<0)==1

    xi(i) = x(find(E100(:,i)<0,1,'first'));

    end

end

%--------------------------------------------------------------------------

%compute yearly summer and winter ice area and temperature at pole

SIA = zeros(1,dur-spinup);

Tpole = SIA;

        sept= (76+spinup*100):100:p; %summer dates excluding spinup

        mar = (26+spinup*100):100:p; %winter dates excluding spinup 

        SIA_sept= 255*(1-xi(sept));  %summer ice area in million km^2

        SIA_mar = 255*(1-xi(mar));   %winter ice area in million km^2

        Tpole_sept= T100(end,sept);  %summer temperature at pole

        Tpole_mar = T100(end,mar);   %winter temperature at pole

 

Fv = linspace(Fdef,F,dur-spinup);   %yearly forcing without spinup

%--------------------------------------------------------------------------

%plot summer/winter ice areas and temperatures at the pole

figure(1); clf

subplot(2,1,1)

plot(Fv, SIA_sept,'r',Fv,SIA_mar,'b')

xlabel('F (W m^{-2)}')

ylabel('Sea Ice Area (10^6 km^2)')

legend('winter','summer')

 

subplot(2,1,2)

plot(Fv, Tpole_sept,'r',Fv,Tpole_mar,'b')

xlabel('F (W m^{-2)}')

ylabel('Temperature at Pole (^oC)')

 

bottom of page