Eric Maloney -- Assessment of 21st Century CMIP5 Projections of North American Regional Climate -- In Part three of a three-part study on North American climate in Coupled Model Intercomparison project (CMIP5) models, we examine projections of 21st Century climate in the RCP4.5 and RCP8.5 emission experiments. Aspects of North American climate change that are examined include changes in continental-scale temperature and the hydrologic cycle, extremes events, and storm tracks, as well as regional manifestations of these climate variables. We also examine changes in east Pacific and Atlantic tropical cyclone activity and North American intraseasonal to decadal variability, including changes in teleconnections to other regions of the globe. Projected changes are generally consistent with those in CMIP3, although with better model agreement in some areas (e.g. projections of summer time precipitation decreases in the Caribbean and Southern Mexico). Although many projected changes in North American climate are robust across CMIP5 models, substantial disagreement in some areas helps to define priorities for future research. Areas of disagreement include projections of changes in snow water equivalent and diurnal temperature range on a regional basis, precipitation in the Southern U.S., Atlantic and east Pacific tropical cyclone activity, intraseasonal variability, and El Niño teleconnections. Model success in simulating historical climate as shown in Parts 1 and 2 of this study lend confidence to many of the projected results. However, model biases in other areas decrease confidence in projections, including changes in the timing of North American monsoon precipitation, growing season length along the West Coast, and the distribution of persistent drought and wet spells.
Kristopher Karnauskas -- The American Midsummer Drought in CMIP5: Multi-Model Evaluation and Projections -- The rainy season in Central America and southern Mexico spans roughly May through October. For most of the region, the precipitation climatology features maxima in June and September and a period of reduced rainfall during July-August known as the midsummer drought (MSD). The MSD is regular enough to be known colloquially and plays an important role in farming practices. Here we assess CMIP5 model performance at simulating this key climatological feature and evaluate their future projections. Historical, RCP4.5, and RCP8.5 experiments are analyzed from all CMIP5 coupled ocean-atmosphere or Earth system models for which monthly precipitation output is available. The numbers of models included in our analyses are 23, 17, and 20 for Historical, RCP4.5, and RCP8.5, respectively. A simple algorithm for detecting and quantifying the climatological MSD has recently been developed that does not assume a priori which months are maxima and which months constitute the MSD.
Despite biases in overall summertime rainfall amounts, the CMIP5 multi-model mean captures the essence of the MSD over much of the Inter-Americas region. Similar to CMIP3 results, the MSD is not an enigmatic challenge to global models. Out of the 23 individual CMIP5 models analyzed here and included in the multi-model mean, roughly half do a reasonably good job simulating the MSD on an individual basis, with a handful performing very well. Significant differences in the location and strength of the MSD between various observational data sets preclude a definitive evaluation of the CMIP5 multi-model mean, but it is clear that the strength of the MSD is underestimated in some regions, including along the Pacific coast of Central America, the western Caribbean, the major Caribbean islands and Florida.
Consistent with seasonal rainfall projections, the CMIP5 multi-model mean provides a very robust projection of a stronger MSD for most regions that experience an MSD today. The maximum MSD increases from ~2.5 mm/day to ~3 mm/day in the RCP4.5 forcing experiment, and nearly doubles to >4 mm/day in the RCP8.5 forcing experiment. The multi-model mean projection is qualitatively consistent with each of the individual CMIP5 models that best replicate the observed MSD. The stronger MSD is a result of early/mid summertime rainfall being reduced by a greater amount than the late summertime peak. The extension of the MSD northward along the Gulf coast of Mexico and into the U.S. is enhanced in both forcing experiments. A few localities that do not presently exhibit an MSD (e.g., Panama) develop a moderate MSD under either forcing experiments, while others (e.g., northeast Yucatan, western Caribbean) are projected to see a change toward a reduction or even disappearance of the MSD. Overall, however, the CMIP5 projection calls for a robust strengthening of the MSD where it presently exists rather than a broadening of the region that experiences an MSD.
Paul Dirmeyer -- Projected changes in land-atmosphere interactions from CMIP5 simulations. -- We examine results from current (historical) and severe future (RCP8.5) projections from 15 different CMIP5 climate models to assess likely trends in land surface states, fluxes, and atmospheric parameters linked to the land surface. Land-atmosphere coupling defined by several metrics is projected to increase and expand over a larger area and longer fraction of the year in many locations. This implies a greater control by soil moisture variations on surface fluxes and the lower troposphere in the future. There is also a strong consensus for a deepening atmospheric boundary layer (ABL) and diminished gradients across the entrainment zone at the top of the boundary layer, indicating that the land surface feedback on the atmosphere should become stronger both in absolute terms and relative to the influence of the free atmosphere on the ABL. Along with a clear trend toward precipitation extremes, it appears the land surface will play a greater role in amplifying both extremes and trends in climate.
Rym Msadek -- Using CMIP5 retrospective predictions of Atlantic variability to assess the skill of future forecasts -- We analyze retrospective predictions of Atlantic decadal variability with the goal of identifying climate phenomena relevant for society that could be predicted in the future. Because the ability to reproduce past climate variations does not a guaranty the success of predicting future changes, identifying the source of skill in retrospective predictions is key to assess the skill of future forecasts. We focus on the initialized decadal experiments that were conducted as part of CMIP5. Although the dominant source of predictive sill on multi-decadal time scales comes from changing radiative forcing, the Atlantic region stands out in most CMIP5 models as a region where initializing the models with observation estimates can provide additional source of skill, likely related to internal variability. While this talk mainly focuses on results from the GFDL experiments, we relate our findings to published studies using other CMIP5 models and discuss the robustness of our results. We present results from two studies, conducted at GFDL, that show encouraging results for predicting Atlantic variability, in the North and tropical Atlantic, respectively. An abrupt North Atlantic warming associated with a decline of the subpolar oceanic circulation and a shift in ecosystem activity has been observed in the mid-1990s. We show that this climate shift in ocean heat content is rather well predicted by several CMIP5 models and that initializing the ocean appears critical to capture the warming. We then present retrospective predictions of five-year mean and nine-year mean tropical Atlantic hurricane frequency and show significant correlation relative to a null hypothesis of zero correlation. We stress that although encouraging, these results should be interpreted with care. We discuss the source of improved skill in the initialized forecasts and its implication for future multi-year/decadal forecasts. We present the uncertainties of the results arising from the short and non-stationary observational system and conclude by highlighting the remaining challenges of providing successful decadal forecasts.