Constraining unforced and forced climate variability impacts interpretations of past climate variations and predictions of future warming. However, comparing general circulation models (GCMs) and last millennium Holocene hydroclimate proxies reveals significant mismatches between simulated and reconstructed low-frequency variability at multidecadal and longer time scales. This mismatch suggests that existing simulations underestimate either external or internal drivers of climate variability. In addition, large differences arise across GCMs in both the magnitude and spatial pattern of low-frequency climate variability. Dynamical understanding of forced and unforced variability is expected to contribute to improved interpretations of paleoclimate variability. To that end, we develop a framework for fingerprinting spatiotemporal patterns of temperature variability in forced and unforced simulations. This framework relies on two frequency-dependent metrics: 1) degrees of freedom (≡
Forced and unforced temperature variability are poorly constrained and understood, particularly that at time scales longer than a decade. Here, we identify key differences in the time scale–dependent behavior of forced and unforced temperature variability using a combination of numerical climate models and principles of downgradient energy transport. This work, and the spatiotemporal characterizations of forced and unforced temperature variability that we generate, will aid in interpretations of proxy-based paleoclimate reconstructions and improve mechanistic understanding of variability.