Monthly Teleconferences 2017 December 7, 2017 Dr. Hai Lin of Environment and Climate Change Canada (ECCC) presented an overview of the GEM-NEMO global coupled model. It is anticipated that this model will join the NMME operational system in 2018. The atmospheric model, the Global Environmental Model (GEM), has a horizontal resolution of approximately 1.4° x 1.4°, and 79 vertical levels. During the discussion, a question was raised about the choice to invest in more vertical levels over increasing the horizontal resolution further. The ocean model, NEMO, is coupled with the sea ice model, CICE. Using the standard hindcast period, 1980-2010, and 10 ensemble members, model skill has been assessed, and compared to the CanCM3 and CanCM4 models, both individually and in all combinations of multi-model ensembles. Preliminary results find that the GEM-NEMO is the highest-scoring individual model, and the GEM-NEMO + CanCM4 combination is the most skillful MME. The addition of CanCM3 results in negligible additional skill, or slightly decreased skill. Both the MJO and NAO forecasting skill is improved in GEM-NEMO. (EMILY BECKER) June 1, 2017 Dr. Benjamin Cash of George Mason University presented his research on understanding unexpected rainfall deficiency in Southern California (SOCAL) during the 2015-2016 El Niņo winter. His study revealed NMME multi-model ensemble mean simulation of California rainfall had significant correlation with Niņo 3.4 SST, preferring to produce enhanced rainfall over the area for every El Niņo event. Meanwhile, the model noise component, which had a rainfall pattern similar to ENSO along the west coast but minimal association with SST, was largely in response to variations in strength of the west coast low. Composite and noise analyses of the 2015/16 event delivered consistent messages, that variations in simulated El Niņo event strength did not appear to be the driver of the intra-event variability, neither did the non-ENSO SST anomalies. Atmospheric noise appeared to play a key role. Suggestions for future directions were actively discussed. April 6, 2017 The teleconference presentation by Sarah Strazzo of CPC/INNOVIM explored the idea that, while models may not consistently represent observed teleconnections, e.g. between ENSO and North American temperature and precipitation, they may coherently represent large scale climate variability such as ENSO, AO/NAO, NPO etc. Statistical bridging by Bayesian joint probability model was tested to possibly further improve forecast skill beyond what skill was achieved through calibration. Preliminary results for temperature forecasts showed that the bridging was more skillful than calibration in DJF, particularly over the northern United States, though calibration yielded higher skill relative to bridging overall. At longer leads, skill differences between calibration and bridging varied by model, and the difference generally decreased with lead time. Further testing was planned, such as the performance relative to ensemble regression, merging of bridged and calibrated forecasts, exploring the use of additional bridging predictors etc. February 2, 2017 Dr. Tao Zhang of CIRES-University of Colorado presented a study on the difference of observed Southern California (SCAL) precipitation anomalies between two strong El Niņo years, namely 2015-16 and 1997-98, asking the question: Why are strong El Niņo events not “too big to fail”? The study showed that global SST forcing in 2016 was less effective than 1998 in yielding wet conditions in SCAL, and the 2016 below-average SCAL precipitation was mostly a symptom of internal atmospheric variability. Dr. Zhang concluded that the flavor of “El Niņo” was not the main cause for the weakened SCAL wetness; instead, differences in extratropical SST anomalies were the main driver for a weakened SCAL wet signal. The second presentation, by Dr. Amir Aghakouchak of University of California, Irvine, focused on improving seasonal drought prediction in California using a combined statistical-dynamical model approach. Overall, it demonstrated the hybrid framework performed better in predicting negative precipitation anomalies (10-60% improvement over NMME) than positive precipitation anomalies (5-25% improvement over NMME). |
