HIND_CTRL(uten 2005-2007) T2m anomaly Rank histogram

Tsfc anomaly Rank histogram at global, NAtl(20-70N, 110W-40E) and WNPac(20-70N, 110E-180)
GISS as observation data.
Rank histogram is ranking frequency of observation data with ensemble members. 
Histogram all flat -> good ensemble distribution.
Large at most right bin -> model members are overestimate.
Large at most left bin  -> model members are underestimate.
Bell shape -> model simulation is too diverse.
Inverted bell shape -> model simulation is too concentrated.

HIND_CTRL(uten 2005-2007) SST anomaly Rank histogram

SST anomaly Rank histogram at global, NAtl(20-70N, 110W-40E) and WNPac(20-70N, 110E-180)
HadISST as observation data.
Rank histogram is ranking frequency of observation data with ensemble members. 
Histogram all flat -> good ensemble distribution.
Large at most right bin -> model members are overestimate.
Large at most left bin  -> model members are underestimate.
Bell shape -> model simulation is too diverse.
Inverted bell shape -> model simulation is too concentrated.

HIND_CTRL(uten 2005-2007) Z700 heterogeneous Anomaly Correlation(AC), 1-4 lead season

Following Becker et al. 2014 approach ( 10.1175/JCLI-D-13-00597.1 ) 1-4 lead seasonal forecast correlation. 

The correlation is between ERA5 reanalysis and ensemble mean of simulation members.

HIND_CTRL(uten 2005-2007) Z700 anomaly Rank histogram

Z700 anomaly Rank histogram at global, NAtl(20-70N, 110W-40E) and WNPac(20-70N, 110E-180)
ERA5 as observation data.
Rank histogram is ranking frequency of observation data with ensemble members. 
Histogram all flat -> good ensemble distribution.
Large at most right bin -> model members are overestimate.
Large at most left bin  -> model members are underestimate.
Bell shape -> model simulation is too diverse.
Inverted bell shape -> model simulation is too concentrated.

HIND_CTRL(uten 2005-2007) SST regional Rank histogram

SST Rank histogram at global, NAtl(20-70N, 110W-40E) and WNPac(20-70N, 110E-180)
HadISST as observation data.
Rank histogram is ranking frequency of observation data with ensemble members. ie. 
Histogram all flat -> good ensemble
Large at most right bin -> model members are overestimate.
Large at most left bin  -> model members are underestimate.
Bell shape -> model simulation is too diverse.
Inverted bell shape -> model simulation is too concentrated.

HIND_CTRL(uten 2005-2007) Atlantic-3 predictionskill

Atlantic-3 (20W-0, 3S-3N) prediction skill, compare with persistence prediction.

HIND_CTRL(uten 2005-2007) Tsfc anomaly Rank histogram

Tsfc anomaly Rank histogram at global, NAtl(20-70N, 110W-40E) and WNPac(20-70N, 110E-180)
GISS as observation data.
Rank histogram is ranking frequency of observation data with ensemble members. 
Histogram all flat -> good ensemble distribution.
Large at most right bin -> model members are overestimate.
Large at most left bin  -> model members are underestimate.
Bell shape -> model simulation is too diverse.
Inverted bell shape -> model simulation is too concentrated.

HIND_CTRL(uten 2005-2007) ENSO predictionskill

nino3.4 prediction skill, compare with persistence prediction.

HIND_CTRL(uten 2005-2007) Tsfc heterogeneous Anomaly Correlation(AC), 1-4 lead season

Following Becker et al. 2014 approach ( 10.1175/JCLI-D-13-00597.1 ) 1-4 lead seasonal forecast correlation. 

The correlation is between GISS and ensemble mean of simulation members.

HIND_CTRL(uten 2005-2007) Ocean Heat Content heterogeneous Anomaly Correlation(AC), 1-4 lead season

Following Becker et al. 2014 approach ( 10.1175/JCLI-D-13-00597.1 ) 1-4 lead seasonal forecast correlation. 

The correlation is 100m OHC between EN4 4.2.1 analysis and ensemble mean of simulation members.

HIND_CTRL(uten 2005-2007) SST homogeneous Anomaly Correlation(AC), 1-4 lead season

Following Becker et al. 2014 approach ( 10.1175/JCLI-D-13-00597.1 ) 1-4 lead seasonal forecast correlation. 

The correlation is between 1st member and ensemble mean of rest of members.

HIND_CTRL(uten 2005-2007) SST heterogeneous Anomaly Correlation(AC), 1-4 lead season

Following Becker et al. 2014 approach ( 10.1175/JCLI-D-13-00597.1 ) 1-4 lead seasonal forecast correlation. 

The correlation is between HadISST and ensemble mean of simulation members.