2

List of Supplementary Figures

FIG. S1: Historical biome maps across CMIP5 models

Historical biome maps across individual CMIP5 models (years 1980-1999). Biomes are defined in Section 2 and indicated by the color bar. MIROC-ESM is not shown due to its similarity to MIROC-ESM-CHEM. Also shown is the multi-model average map of historical biomes (1980-1999) and multi-model map of future biomes (2080-2099).

FIG. S2: Study of similarity across models

A study of similarity between simulations during the historical period (years 1980-2000) for: a) SST (in ºC), b) surface NO3 (mmol/m3), and c) Integrated production Prod (gC/yr). A frequency distribution is calculated for each pair of models from pixel-to-pixel differences of the above variables. Here, we show the absolute mean (x axis) and standard deviation (y axis) of this distribution. Map1 and Map2 refer to all the possible combinations between two models. The cloud of points shows the typical uncertainty among models. We only label the combination of models that are very similar by construction (black cloud), and these lie outside of the main cloud (red cloud). For nitrate, self-similar models are those with dispersion lower than 1. We did not label them because they are all clustered close to the origin in panel (b).

FIG. S3: Multi-model averaged maps

Multi-model average maps for historical (years 1980-1999) and climate change signal (difference between the average for years 2080-2099 and the average for years 1980-1999). Variables shown are a) salinity (averaged over 16 models), b) iron concentration averaged over the top 100m (12 models), c) phytoplankton biomass averaged over the top 100m (16 models), and d) POC export at 100m (12 models). Crossed masked areas show significant trends at 80% to 90% level, and stripped masked areas show a 90% or higher significance.

FIG. S4: Productivity across CMIP5 models

Global maps for primary production (integrated to 100m) during the historical period (years 1980-1999) and changes over 100 years (difference between years 2080-2099 and years 1980-1999) for 16 CMIP5 models. Model MIROC-ESM is not shown due to its similarity to MIROC-ESM-CHEM. Model CMCC-CESM is shown here but is not used in the paper analysis due to the lack of the ice variable in the online CMIP5 output.

FIG. S5: Iron across CMIP5 models

Global maps for iron (averaged over 100m) during the historical period (years 1980-1999) and change over 100 years (difference between years 2080-2099 and years 1980-1999) across CMIP5 models.

FIG. S6: Historical Phytoplankton biomass and POC export across CMIP5 models

Global maps for phytoplankton biomass (averaged over 100m) in top panels and POC export at 100m in the bottom panels during the historical period (years 1980-1999) across CMIP5 models.

Supplementary Table

Temp
thetaosurf / Strat
(calculated) / Zonal wind tauu / MLD
mlotst / Ice coverage
sic / P-E
pr
evpsbl / SSS
salisurf / W velocity
wmo / Light at surface
rsntds / Clouds clt / Nitrate
no3 / Iron
fe / Production
pp or intpp / Export
expc or ecp100 / Phyto
phyc / Diatom
phydiat
CanESM2 / Y / Y / Y / Y / Y / Y / Y / Y / Y / Y / Y / N / Y / Y / Y / N
CESM1-BGC / Y / Y / Y / Y / Y / Y / Y / Y / Y / Y / Y / Y / Y / Y / Y / Y
GFDL-ESM2G / Y / Y / Y / Y / Y / Y / Y / Y / Y / Y / Y / Y / Y / Y / Y / Y
GFDL-ESM2M / Y / Y / Y / Y / Y / Y / Y / Y / Y / Y / Y / Y / Y / Y / Y / Y
GISS-E2-H-CC / Y / Y / Y / Y / Y / Y / Y / Y / N / Y / Y / Y / Y / N / Y / Y
GISS-E2-R-CC / Y / Y / Y / Y / Y / Y / Y / Y / N / Y / Y / Y / Y / N / Y / Y
HadGEM2-CC / Y / Y / Y / Y / Y / Y / Y / Y / N / Y / Y / Y / Y / Y / Y / Y
HadGEM2-ES / Y / Y / Y / Y / Y / Y / Y / Y / N / Y / Y / Y / Y / Y / Y / Y
IPSL-CM5A-LR / Y / Y / Y / Y / Y / Y / Y / Y / Y / Y / Y / Y / Y / Y / Y / Y
IPSL-CM5A-MR / Y / Y / Y / Y / Y / N / Y / Y / Y / Y / Y / Y / Y / Y / Y / Y
MIROC-ESM / Y / Y / Y / Y / Y / Y / Y / Y / Y / Y / Y / N / Y / N / Y / N
MIROC-ESM-CHEM / Y / Y / Y / Y / Y / Y / Y / Y / Y / Y / Y / N / Y / N / Y / N
MPI-ESM-MR / Y / Y / Y / Y / Y / Y / Y / Y / Y / Y / Y / Y / Y / Y / Y / N
MPI-ESM-LR / Y / Y / Y / Y / Y / Y / Y / Y / Y / Y / Y / Y / Y / Y / Y / N
MRI-ESM1 / Y / Y / Y / Y / Y / Y / Y / Y / Y / N / Y / N / Y / Y / Y / N
NorESM1-ME / Y / Y / Y / Y / Y / Y / Y / Y / Y / Y / Y / Y / Y / Y / Y / N
TOTAL NUMBER / 16 / 16 / 16 / 16 / 16 / 15 / 16 / 16 / 13 / 15 / 16 / 12 / 16 / 12 / 16 / 9

Table S1: List of variables used in the paper, downloaded (or calculated) from the CMIP5 database (Y when available, and N when not available). Stratification is calculated as the density difference between 200m and the surface. Mixed layer depth (MLD) is either provided or calculated from the potential density as the depth where the density difference to the surface above is 0.125kg/m3 (Levitus 1982). CESM1-BGC and IPSL-CM5A-LR only provide surface diatom data; all other models provide diatom data at all depths.