Identify the causes of the signal-to-noise paradox in the North Atlantic Oscillation
PI: Laifang Li (Meteorology and Atmospheric Science)
The tuition for graduate students will be covered by a NASA grant.
Project Description
The North Atlantic Oscillation (NAO) is the most prominent mode of atmospheric variability in the Northern Hemisphere, has broad and distinct climatic impacts, and is thus critical to climate prediction. Attempts to predict the NAO have found that dynamical climate models underestimate its predictability such that the models are significantly better at predicting the observed NAO than they are at predicting individual runs of themselves (Dunston et al. 2016; Smith et al. 2020). This signal-to-noise paradox implies that climate models could be predicting the NAO much better than they currently are if the important processes are correctly modeled (Scaife and Smith 2018).
The proposed work aims to fill this void by identifying those processes and examining their simulation in the current generation of models. The work centers around the publiclyaccessible library of CMIP6 model runs: the Historical experiment and the Atmospheric Modeling Intercomparison Program (AMIP). Each of these experiments varies in their prescription of sea surface temperature (SST) and external forcings, providing us with a laboratory to simultaneously examine the physical drivers of the NAO and model agreement on those drivers. Our hypothesis is that even though the models from various modeling centers tend to su er from the signal-to-noise Paradox and are thus imperfect models of the real-world, the fact that their ensemble means are able to reliably predict the NAO’s temporal variability implies that they must be simulating the correct processes important to the NAO, albeit too weakly.
To quantify these predictable processes (signal) and the unpredictable noise in the climate model simulations, a linear inverse model (LIM) will be applied to sea level pressure over the North Atlantic, a variable defines the NAO. The LIM, as a statistical modeling technique, separates the evolution of a given physical system into its predictable, linear dynamics, a linear approximation of non-linear dynamics ( ) and an unpredictable stochastic forcing (; noise): =+. The LIM will be applied to observed sea level pressure so that the predictable dynamics and the noise of the observed NAO can be quantified. The observations will then be served as a benchmark for model performance. The similar LIM analysis will first be applied to the Historical simulation, and the derived predictable dynamics and unpredictable noise will be compared against observations. This comparison will allow us to quantify the extent to which the signal-to-noise paradox is due to models’ under-representation of predictable dynamics or overestimation of the noise. In addition, the LIM will be applied to the corresponding AMIP-type simulations to explore whether the constraining the SST with observations can improve the model’s performance in simulating the predictable dynamics of the NAO and alleviate the signal-to-noise paradox. Finally, according to the LIM analysis, idealized experiments with SST prescribed (or partially coupled) to certain area of the ocean will be performed to isolate the key physical processes involved, such as wave dynamics or local air-sea positive feedback mechanisms.
Specific Areas of Computational and/or Data Science Expertise the PI is particularly interested in recruiting are: 1) experience handling CMIP6 data archive; 2) knowledge of the linear inverse model; 3) experience performing idealized climate simulations, in particular the partial coupling simulations.
Requirements and Expectations: The project looks for a highly motivated junior researcher with background in atmosphere dynamics. The research should communicate the progress in a timely manner with the PI. The written and oral reports are expected in a regular basis, like the group meeting and relevant joint discussion with collaborating labs.
Specific Objectives: The project outcome will be utilized as extra data for revising an NSF grant proposal, for which the primary comments are around the quantification of predictable and unpredictable dynamics. The LIM analysis aims to address the comments. A peer-reviewed journal articles from the project will be the minimum target, for the PI’s attempt to resubmit the proposal.
Medium to Long-Term Goals: 1) obtain external research grant on NAO predictability; 2) use the project outcome to guide the development of new NAO theories: persistence, periodicity and stochastic forcing; 3) contribute to the improvements of NAO simulations in the NSF NCAR’s Community Earth System Model (CESM).
Connection to ICDS’s mission: The project will utilize advanced data analysis technique to address one of the most challenging Earth system predictability issues in the climate community.
Recent and planned engagement with ICDS: The PI is an ICDS co-hire and has participated in multiple ICDS sponsored activities, like judging the poster presentations for the ICDS symposium. Her graduate students present their research at the past symposiums as well. Through this ICDS support, the PI and the selected junior researcher will plan to: 1) present the research findings at the upcoming ICDS symposium; 2) publish papers acknowledging ICDS as a iliation and funding source (if funded); and 3) plan to co-host workshops on NAO predictability by the Department of Meteorology and Atmospheric Science and ICDS.