Recent Detection and Attribution Work at GFDL: Record Global Temperatures in 2016, Arctic Winter Warmth in 2016, and Regional Precipitation Trends

In this presentation, I will give an update on recent detection and attribution assessment work being done at GFDL. Global annual-mean surface temperature set a record high in 2016 according to three observational datasets. We conclude that, according to the CMIP5 simulations, anthropogenic forcing was a necessary condition for the record global-mean warmth experienced in 2016. Anthropogenic forcing contributed most of (calendar year) 2016’s record annual warmth relative to 1881-1920 conditions, while natural forcings and intrinsic variability (including ENSO) made relatively small contributions to the annual mean. Arctic surface temperatures were anomalously warm during November-December 2016. An Arctic area-averaged temperature index set a new record for warmth in the GISS Surface Temperature Analysis data, and was either a record high or anomalously high (compared to early 20^th century levels) in four other observational products. Using CMIP5 simulations, we assessed the causes of the highly anomalous Arctic warmth. While the century-scale Arctic warming observed since the late 1800s resembles that in the CMIP5 All-Forcing ensemble mean, the CMIP5 model ensemble mean does not show the strong warm phase during 1920-1940, suggesting that the early 20^th century observed Arctic warming may contain a large contribution from internal climate variability. Nonetheless, the model results suggest that the warmth observed during Nov.-Dec. 2016 most likely would not have been possible without a substantial long-term warming contribution from anthropogenic forcing. Precipitation trends for 1901-2010, 1951-2010 and 1981-2010 over relatively well-observed global land regions are assessed for detectable anthropogenic influences and for consistency with CMIP5 historical simulations. Trends in observed data since 1901 tend toward generally wetter conditions in the extratropics, apart from a few regions such as the Mediterranean and Japan. In the CMIP5 historical runs, we find some broad consistency with the observed data trend pattern, but a general tendency for a low bias in modeled trends since 1901, compared to the observations. Despite this bias, we infer that trends in observed data and models are statistically consistent over 59% of the analyzed land area. Trends in observed data were detectable—distinct from natural variability--over 37% of the analyzed area. Over 19% (8%) of the analyzed land area, increased (decreased) precipitation was partly attributable to anthropogenic forcing. These human-induced changes include: increases over regions of the north-central U.S., southern Canada, Europe, and southern South America, and decreases over parts of the Mediterranean region and northern tropical Africa. Trends for the shorter periods (1951-2010 and 1981-2010) do not indicate a prominent negative trend bias in the models, as found for the 1901-2010 trends. An atmosphere-only model, forced with observed SSTs and other climate forcing agents, also underpredicts the precipitation increase in the observed data for the northern hemisphere extratropics since 1901. The model low biases in historical simulated trends since 1901, if borne out in further studies, suggest that future precipitation projections using these regions/models could overestimate future drought risk, and underestimate future flooding risk. Therefore, further study is needed to assess the reliability of the apparent low precipitation trend bias in the CMIP5 models.