Attribution of recent temperature behaviour reassessed by a neural-network method

Area of Science:

• Climate Science
• Environmental Science
• Data Science

Background:

• Global Climate Models (GCMs) suggest anthropogenic forcings drive recent global warming.
• GCMs share a common dynamical approach, raising questions about the robustness of their attribution results.
• A need exists for independent methods to validate climate change attribution.

Purpose of the Study:

• To independently attribute the drivers of global temperature changes over the last 160 years.
• To assess the contributions of anthropogenic forcings, solar activity, and internal climate variability.
• To validate and expand upon existing climate model findings using a novel approach.

Main Methods:

• Employed a data-driven, non-dynamical, and fully nonlinear approach using neural network (NN) modeling.
• Analyzed global temperature data spanning the last 160 years.
• Conducted attribution experiments and sensitivity analyses to various forcing changes.

Main Results:

• Confirmed anthropogenic forcings as the primary drivers of global temperature increase in the last half-century.
• Identified significant solar influence on global temperatures between 1910 and 1975.
• Clarified the roles of sulphate aerosols and Atlantic Multidecadal Oscillation in decadal temperature variability.

Conclusions:

• Neural network analysis corroborates GCM findings on anthropogenic warming.
• Provides new insights into the relative contributions of external forcings and internal climate variability.
• Highlights the value of diverse, data-driven approaches in climate change attribution studies.