The discovery of moderate non-linear surprises in CMIP6 models
Abstract
This study presents an automated algorithm for detecting Moderate Non-linear Surprises (MNS) in the outputs of climate models participating in the Coupled Model Intercomparison Project phase 6 (CMIP6), building upon and relaxing existing methodologies for detection of critical transitions in climate models. The approach combines four statistical criteria — standard deviation, maximum 10-year jump, a multimodality test, and deviation over the last 30 years relative to the preindustrial distribution — with a density-based spatial clustering step to objectively identify regions undergoing non-linear transitions, without requiring expert judgment. The algorithm is applied across sea-ice, ocean, land, and atmospheric variables from up to 37 CMIP6 models under four Shared Socioeconomic Pathways (ssp) scenarios. Results highlight multiple MNS in high-latitude ocean and ice regions, and a recurrent signal of hydrological decline over tropical South America, including the Amazon Basin, detected in large fraction of the models with land-surface output. Some models also exhibit significant shifts in the extratropical atmospheric circulation, most notably in the blocking frequency, particularly under the strong ssp585 forcing. These examples demonstrate that the tool offers a systematic, reproducible framework for screening emerging non-linear climate responses in large model ensembles that may not qualify as tipping points, but still challenge the adaptation of human societies and ecosystems
Related
- Catalogue of strong nonlinear surprises in ocean, sea-ice, and atmospheric variables in CMIP6
- Comprehensive assessment of historical surface air temperature simulations over iran using 51 CMIP6 models
- A machine learning model for early season wildfire risk in Ghana’s Guinea savannah assessing discrimination, operational prevalence, spatial transferability, and CMIP6 climate scenarios
- Emulating absent land processes reduces carbon budgets in CMIP6 models
- CMIP6‐Driven CaMa‐Flood Simulations Reveal Anthropogenic Control on Global Flood Extremes via Precipitation and Snowmelt Drivers
BibTeX
@misc{lembo2026discovery,
title = {{The discovery of moderate non-linear surprises in CMIP6 models}},
author = {Valerio Lembo and Marco Possega and Susanna Corti and Annalisa Cherchi and Andrea Alessandri},
howpublished = {EarthArXiv},
year = {2026},
month = oct,
doi = {10.31223/x5hn6w},
url = {https://eartharxiv.org/repository/view/15658/},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - The discovery of moderate non-linear surprises in CMIP6 models
AU - Valerio Lembo
AU - Marco Possega
AU - Susanna Corti
AU - Annalisa Cherchi
AU - Andrea Alessandri
PY - 2026
DA - 2026/10/21
PB - EarthArXiv
DO - 10.31223/x5hn6w
UR - https://eartharxiv.org/repository/view/15658/
AB - This study presents an automated algorithm for detecting Moderate Non-linear Surprises (MNS) in the outputs of climate models participating in the Coupled Model Intercomparison Project phase 6 (CMIP6), building upon and relaxing existing methodologies for detection of critical transitions in climate models. The approach combines four statistical criteria — standard deviation, maximum 10-year jump, a multimodality test, and deviation over the last 30 years relative to the preindustrial distribution — with a density-based spatial clustering step to objectively identify regions undergoing non-linear transitions, without requiring expert judgment. The algorithm is applied across sea-ice, ocean, land, and atmospheric variables from up to 37 CMIP6 models under four Shared Socioeconomic Pathways (ssp) scenarios. Results highlight multiple MNS in high-latitude ocean and ice regions, and a recurrent signal of hydrological decline over tropical South America, including the Amazon Basin, detected in large fraction of the models with land-surface output. Some models also exhibit significant shifts in the extratropical atmospheric circulation, most notably in the blocking frequency, particularly under the strong ssp585 forcing. These examples demonstrate that the tool offers a systematic, reproducible framework for screening emerging non-linear climate responses in large model ensembles that may not qualify as tipping points, but still challenge the adaptation of human societies and ecosystems
ER -