Multi-decadal surface temperature changes may be forced by natural as well as anthropogenic factors, or arise unforced from the climate system. Distinguishing these factors is essential for estimating sensitivity to multiple climatic forcings and the amplitude of the unforced variability. Here we present 2,000-year-long global mean temperature reconstructions using seven different statistical methods that draw from a global collection of temperature-sensitive paleoclimate records. Our reconstructions display synchronous multi-decadal temperature fluctuations, which are coherent with one another and with fully forced CMIP5 millennial model simulations across the Common Era. The most significant attribution of pre-industrial (1300-1800 CE) variability at multi-decadal timescales is to volcanic aerosol forcing. Reconstructions and simulations qualitatively agree on the amplitude of the unforced global mean multi-decadal temperature variability, thereby increasing

Quantifying past climate variation and attributing its causes improves our understanding of the natural variability of the climate system. Tree-ring-based proxies have provided skillful and highly resolved reconstructions of temperature and hydroclimate of the last millennium. However, like all proxies, they are subject to uncertainties arising from varying data quality, coverage, and reconstruction methodology. Previous studies have suggested that biological-based memory processes could cause spectral biases in climate reconstructions. This study determines the effects of such biases on reconstructed temperature variability and the resultant implications for detection and attribution studies. We find that introducing persistent memory, reflecting the spectral properties of tree-ring data, can change the variability of pseudoproxy reconstructions compared to the surrogate climate and resolve certain model–proxy discrepancies. This is especially the case for proxies based on ring-width data. Such memory inflates the difference between the Medieval Climate Anomaly and the Little Ice Age and suppresses and extends the cooling in response to volcanic eruptions. When accounting for memory effects, climate model data can reproduce long-term cooling after volcanic eruptions, as seen in proxy reconstructions. Results of detection and attribution studies show that signals in reconstructions as well as residual unforced variability are consistent with those in climate models when the model fingerprints are adjusted to reflect autoregressive memory as found in tree rings.

Global warming is expected to not only impact mean temperatures but also temperature variability, substantially altering climate extremes. Here we show that human-caused changes in internal year-to-year temperature variability are expected to emerge from the unforced range by the end of the 21st century across climate model initial-condition large ensembles forced with a strong global warming scenario. Different simulated changes in globally averaged regional temperature variability between models can be explained by a trade-off between strong increases in variability on tropical land and substantial decreases in high latitudes, both shown by most models. This latitudinal pattern of temperature variability change is consistent with loss of sea ice in high latitudes and changes in vegetation cover in the tropics. Instrumental records are broadly in line with this emerging pattern, but have data gaps in key regions. Paleoclimate proxy reconstructions support the simulated magnitude and distribution of temperature variability. Our findings strengthen the need for urgent mitigation to avoid unprecedented changes in temperature variability.

Insolation changes caused by the axial precession induce millennial trends in last millennium temperature, varying with season and latitude. A characteristic seasonal trend pattern can be detected in both insolation and modeled surface temperature response. In the extratropical Northern Hemisphere, the maximum insolation trend occurs around April/May, while the minimum trend occurs between July and September. The temperature trend lags behind insolation trend by around a month. Hence orbital forcing potentially affects long‐term trends in proxy data, which are often sensitive to a distinct seasonal window. We find that tree‐ring reconstructions based on early growing season dominated records show different millennial trends from those for late summer dominated proxies. The differential response is similar to that seen in pseudo proxy reconstructions when considering proxy seasonality. This suggests that orbital forcing has influenced long‐term trends in climate proxies. It is therefore vital to use seasonally homogeneous data for reconstructing multicentennial variability.

No abstract