I have a new update to climate model-observation comparisons over at The Climate Brink, covering CMIP3, CMIP5, and CMIP6. Models perform well globally. The latest generation shows too much long-term warming but better reproduces recent trends: www.theclimatebrink....
December 6, 2025 at 8:42 PM
I have a new update to climate model-observation comparisons over at The Climate Brink, covering CMIP3, CMIP5, and CMIP6. Models perform well globally. The latest generation shows too much long-term warming but better reproduces recent trends: www.theclimatebrink....
Model data: CMIP5 multimodel mean, blended tos/tas, RCP8.5
Observations: GISTEMP
Both baselined to a 1951-1980 climatology.
CMIP5 data downloaded from @knmi.nl Climate Explorer climexp.knmi.nl
GISTEMP data from data.giss.nasa.gov/gistemp/
Observations: GISTEMP
Both baselined to a 1951-1980 climatology.
CMIP5 data downloaded from @knmi.nl Climate Explorer climexp.knmi.nl
GISTEMP data from data.giss.nasa.gov/gistemp/
January 24, 2026 at 1:16 AM
Model data: CMIP5 multimodel mean, blended tos/tas, RCP8.5
Observations: GISTEMP
Both baselined to a 1951-1980 climatology.
CMIP5 data downloaded from @knmi.nl Climate Explorer climexp.knmi.nl
GISTEMP data from data.giss.nasa.gov/gistemp/
Observations: GISTEMP
Both baselined to a 1951-1980 climatology.
CMIP5 data downloaded from @knmi.nl Climate Explorer climexp.knmi.nl
GISTEMP data from data.giss.nasa.gov/gistemp/
The CMIP7 scenarios will be finalized in the coming months, and the new high scenario will be notably lower than in CMIP5/CMIP6.
February 12, 2026 at 4:20 AM
The CMIP7 scenarios will be finalized in the coming months, and the new high scenario will be notably lower than in CMIP5/CMIP6.
The aerosol forcing must be highly underestimated too (more desulphurization in reality vs CMIP5 = more real-world warming):
"sulfate aerosols in ERA5 follow CMIP5 historical and scenario forcings, although based on historical emission estimates until 2009 and scenario data thereafter"
"sulfate aerosols in ERA5 follow CMIP5 historical and scenario forcings, although based on historical emission estimates until 2009 and scenario data thereafter"
December 7, 2024 at 3:20 PM
The aerosol forcing must be highly underestimated too (more desulphurization in reality vs CMIP5 = more real-world warming):
"sulfate aerosols in ERA5 follow CMIP5 historical and scenario forcings, although based on historical emission estimates until 2009 and scenario data thereafter"
"sulfate aerosols in ERA5 follow CMIP5 historical and scenario forcings, although based on historical emission estimates until 2009 and scenario data thereafter"
How do recent extreme global temperatures compare to climate models? Lets first look at the last generation of models (CMIP5), which have generally performed well at reproducing observations. Here we see observations well above the 95th percentile of models in September.
October 5, 2023 at 5:38 PM
How do recent extreme global temperatures compare to climate models? Lets first look at the last generation of models (CMIP5), which have generally performed well at reproducing observations. Here we see observations well above the 95th percentile of models in September.
Looking back at the last 56 years (1970-present), the last 25 years, and the last 15 years, we find that recent warming is well outside the range of model projections for CMIP3 and CMIP5, but CMIP6 models tend to overestimate longer term trends.
December 6, 2025 at 8:43 PM
Looking back at the last 56 years (1970-present), the last 25 years, and the last 15 years, we find that recent warming is well outside the range of model projections for CMIP3 and CMIP5, but CMIP6 models tend to overestimate longer term trends.
For reference, here is global land models (CMIP5) vs observations:
August 14, 2025 at 8:05 PM
For reference, here is global land models (CMIP5) vs observations:
RCP8.5 had 1.5C slightly before 2030 in CMIP5. As did all the other RCPs (bar perhaps 2.6) given the small differences in forcing at this point.
March 20, 2026 at 3:44 PM
RCP8.5 had 1.5C slightly before 2030 in CMIP5. As did all the other RCPs (bar perhaps 2.6) given the small differences in forcing at this point.
One notable finding is that the trends over the past 15 years are more on the high-end of CMIP5 model expectations than trends were below expectations during the hiatus years. Though the end of the hiatus is a cautionary tale to not over-interpret short-term trends.
December 6, 2025 at 8:42 PM
One notable finding is that the trends over the past 15 years are more on the high-end of CMIP5 model expectations than trends were below expectations during the hiatus years. Though the end of the hiatus is a cautionary tale to not over-interpret short-term trends.
The regional climate sensitivity of ESMs (dXreg/dTglob) is found to be mostly independent from the ESMs' global climate sensitivity (dTglob/dCO2). Hence "hot models" can be used to derive regional climate changes as function of global warming
See also:
agupubs.onlinelibrary.wiley.com/doi/full/10....
See also:
agupubs.onlinelibrary.wiley.com/doi/full/10....
December 8, 2024 at 8:32 AM
The regional climate sensitivity of ESMs (dXreg/dTglob) is found to be mostly independent from the ESMs' global climate sensitivity (dTglob/dCO2). Hence "hot models" can be used to derive regional climate changes as function of global warming
See also:
agupubs.onlinelibrary.wiley.com/doi/full/10....
See also:
agupubs.onlinelibrary.wiley.com/doi/full/10....
A study in Communications Earth & Environment suggests that the North Atlantic Warming Hole is tied to a weakened Atlantic Meridional Overturning Circulation during the past century. go.nature.com/3HkLwJr 🌊 🧪
June 3, 2025 at 4:38 PM
A study in Communications Earth & Environment suggests that the North Atlantic Warming Hole is tied to a weakened Atlantic Meridional Overturning Circulation during the past century. go.nature.com/3HkLwJr 🌊 🧪
My favorite ECS-is-hard example is when I found that a synthesis of all the proposed emergent constraint ECS estimates based on CMIP5 models ended up giving the exact same ECS range as the unconstrained CMIP5 models 😛
November 2, 2023 at 5:22 PM
My favorite ECS-is-hard example is when I found that a synthesis of all the proposed emergent constraint ECS estimates based on CMIP5 models ended up giving the exact same ECS range as the unconstrained CMIP5 models 😛
And a similar figure for CMIP5 models:
October 31, 2023 at 4:20 PM
And a similar figure for CMIP5 models:
This plot also highlights the point I’ve made about a poorly designed RCP 6 meaning, effectively, than CMIP5 didn’t have a usable scenario in between RCP 8.5 and RCP 4.5.
Back in 2012 we did a figure comparing historical emissions to previous generations of emission scenarios. rdcu.be/brDGx
Surprisingly, the figure is still doing the rounds. We update it every year, with many interesting changes primarily due to revisions in land-use change emissions.
1/
Surprisingly, the figure is still doing the rounds. We update it every year, with many interesting changes primarily due to revisions in land-use change emissions.
1/
May 18, 2026 at 9:53 PM
This plot also highlights the point I’ve made about a poorly designed RCP 6 meaning, effectively, than CMIP5 didn’t have a usable scenario in between RCP 8.5 and RCP 4.5.
In Colorado, we looked into the performances of CMIP3 and CMIP5 with their future predictions overlaid on observations made since their releases. TLDR - Trends followed along pretty closely.
See climatechange.colostate.edu for more!
See climatechange.colostate.edu for more!
February 7, 2025 at 5:57 PM
In Colorado, we looked into the performances of CMIP3 and CMIP5 with their future predictions overlaid on observations made since their releases. TLDR - Trends followed along pretty closely.
See climatechange.colostate.edu for more!
See climatechange.colostate.edu for more!
התחזיות לגבי התחממות גלובלית היו די spot on, זה המודל המשולב CMIP5 שפותח ב2011 פורסם בועידת האקלים של 2014 והיה חלק מהגל שהביא לבסוף להסכם פריז ב2015. האפור זה טווח ההתחממות שחזה המודל, האדום זה ההתחממות הנצפית (מזכיר לי את הקיץ של 2024 וכמה פאקין חם היה באירופה)
June 12, 2026 at 7:59 AM
התחזיות לגבי התחממות גלובלית היו די spot on, זה המודל המשולב CMIP5 שפותח ב2011 פורסם בועידת האקלים של 2014 והיה חלק מהגל שהביא לבסוף להסכם פריז ב2015. האפור זה טווח ההתחממות שחזה המודל, האדום זה ההתחממות הנצפית (מזכיר לי את הקיץ של 2024 וכמה פאקין חם היה באירופה)
but it’s not like nothing happened between 2007 and 2024 to explain that difference.
The real pblm with RCP8.5 from my perspective is
RCP6.0 wasn’t run enough and was constructed weirdly so that its forcing started out lower than RCP4.5, meaning CMIP5 analyses were stuck w/on much btwn 4.5 & 8.5
The real pblm with RCP8.5 from my perspective is
RCP6.0 wasn’t run enough and was constructed weirdly so that its forcing started out lower than RCP4.5, meaning CMIP5 analyses were stuck w/on much btwn 4.5 & 8.5
May 17, 2026 at 8:06 PM
but it’s not like nothing happened between 2007 and 2024 to explain that difference.
The real pblm with RCP8.5 from my perspective is
RCP6.0 wasn’t run enough and was constructed weirdly so that its forcing started out lower than RCP4.5, meaning CMIP5 analyses were stuck w/on much btwn 4.5 & 8.5
The real pblm with RCP8.5 from my perspective is
RCP6.0 wasn’t run enough and was constructed weirdly so that its forcing started out lower than RCP4.5, meaning CMIP5 analyses were stuck w/on much btwn 4.5 & 8.5
'Emergent constraints provided by observed internal variability are weak and largely unchanged from CMIP5 to CMIP6 due to the shortness of the observational record' agupubs.onlinelibrary.wiley.com/doi/full/10....
Links Between Internal Variability and Forced Climate Feedbacks: The Importance of Patterns of Temperature Variability and Change
Internal variability feedbacks better predict forced feedbacks in CMIP6 relative to CMIP5 by over 50%
The improved prediction derives in part from the greater similarity between the patterns of i...
agupubs.onlinelibrary.wiley.com
December 20, 2024 at 4:47 PM
'Emergent constraints provided by observed internal variability are weak and largely unchanged from CMIP5 to CMIP6 due to the shortness of the observational record' agupubs.onlinelibrary.wiley.com/doi/full/10....
My stab at today's radiative forcing vs. CMIP5/CMIP6 scenarios. "Observed" (at least as of 2024) doesn't quite exceed the highest scenarios that I've seen in other versions, partly due to anchoring scenarios to "obs" at 2005 (for RCPs) or 2014 (for SSPs). Not sure if that's the most sensible thing.
May 20, 2026 at 9:49 PM
My stab at today's radiative forcing vs. CMIP5/CMIP6 scenarios. "Observed" (at least as of 2024) doesn't quite exceed the highest scenarios that I've seen in other versions, partly due to anchoring scenarios to "obs" at 2005 (for RCPs) or 2014 (for SSPs). Not sure if that's the most sensible thing.
For the dummies here (ie me!) could you please explain what “CMIP5 historic forcing” means and what exactly the RCP8.5 scenario represents?
March 26, 2025 at 12:51 PM
For the dummies here (ie me!) could you please explain what “CMIP5 historic forcing” means and what exactly the RCP8.5 scenario represents?
I had read that study. It is curious. Although am jealous of their ability to use 239,616 CMIP5 ensemble members... I have only used order of 100s ;-)
March 3, 2026 at 10:54 AM
I had read that study. It is curious. Although am jealous of their ability to use 239,616 CMIP5 ensemble members... I have only used order of 100s ;-)
Does anyone know where I can download pre-computed annual global mean temperatures from CMIP5/6 model runs for a period that covers at least 1960-2100 and a range of scenarios?
January 10, 2024 at 10:33 AM
Does anyone know where I can download pre-computed annual global mean temperatures from CMIP5/6 model runs for a period that covers at least 1960-2100 and a range of scenarios?
Here’s the link to the paper.
BG - Multiple stressors of ocean ecosystems in the 21st century: projections with CMIP5 models
bg.copernicus.org
July 13, 2023 at 10:56 AM
Here’s the link to the paper.
