Detecting causality signal in instrumental measurements and climate model simulations: Global warming case study

Detecting the direction and strength of the causality signal in observed time series is becoming a popular tool for exploration of distributed systems such as Earth's climate system. Here, we suggest that in addition to reproducing observed time series of climate variables within required accuracy a model should also exhibit the causality relationship between variables found in nature. Specifically, we propose a novel framework for a comprehensive analysis of climate model responses to external natural and anthropogenic forcing based on the method of conditional dispersion. As an illustration, we assess the causal relationship between anthropogenic forcing (i.e., atmospheric carbon dioxide concentration) and surface temperature anomalies. We demonstrate a strong directional causality between global temperatures and carbon dioxide concentrations (meaning that carbon dioxide affects temperature more than temperature affects carbon dioxide) in both the observations and in (Coupled Model Intercomparison Project phase 5; CMIP5) climate model simulated temperatures.