FALL 2024
Wednesdays at 3:00 pm, Seminar Room SLAB 103 / Virtual SLAB 103
Aug 21: NO SEMINAR
Aug 28: NO SEMINAR
Sep 04: NO SEMINAR
Sep 11: NO SEMINAR
Sep 18: Peisen Tan
Department of Ocean Sciences, Rosenstiel School
(one-hour OCE student seminar)
The Melodious Encounter of the Atmosphere and the Ocean:
Exploring Air-Sea Momentum Transfer Across a Wide Range of Wind Forcing
Ocean waves profoundly contribute to the upper ocean turbulence and mixing. Most ocean waves are generated by the wind. However, a lack of quantitative measurement of wave growth under wind forcings exists due to instrument limitations. To thoroughly study the wave growth under wind forcing and improve numerical wave models, we conducted laboratory experiments for a wide range of wind forcing in the University of Miami's wind-wave facility SUSTAIN (SUrge STructural Air-sea INteraction Facility). Pure wind seas, irregular JONSWAP spectrum waves, and regular monochromatic waves at different frequencies were generated by the mechanical paddles for testing. We directly measured the static air pressure, wind speeds, and water surface displacement at sampling frequencies up to 1000 Hz. In addition, a wave follower was deployed to sample the static pressure behind the wave crest. Our wave growth rates agree with the revisited Miles' Theory that incorporates airside turbulence. The Jeffreys' Sheltering hypothesis remains valid, although it depends on an empirical constant: the sheltering coefficient A. We found that A follows a "U" shape with wind forcing: The strongest aerodynamic sheltering, corresponding to the smallest A, was identified at U10 around 33 m/s (Category-1 hurricane intensity). Our wind stress partition at the air-sea interface agrees with previous laboratory and model results. Our findings suggest that airflow separation plays a pivotal role in the wave growth and that some tuning of existing wave growth theories paves the way for a novel parameterization of wave growth and wind input in future wave models.
Sep 25: Dr. Myles Allen
Department of Physics, University of Oxford, UK
Guest of the Department of Atmospheric Sciences
Forecast-Based Attribution:
A Probabilistic Approach to Storylines, or a Storyline Approach to Probabilities
Myles Allen, discussing work led by Nicholas Leach, Shirin Ermis & Olivia Vashti Ayim
Recording Available at COMPASS ON DEMAND
In examining the impact of external climate drivers on extreme weather, there has been a lively discussion of the relative merits of probabilistic versus storyline framings. The probabilistic framing uses large ensembles to assess the impact of an external factor on the occurrence-probability of a certain class of extreme event, inspired in some way by the event that actually occurred. Storylines use detailed simulation of the event itself to assess the role of external factors in its evolution or magnitude. This talk argues that we can use operation seasonal or even medium-range ensemble forecasts to develop a new experimental design for the investigation of causes of extreme weather events that combines some of the advantages of both framings, opening the door to applying the best meteorological simulation tools on the planet to address climate questions. I illustrate the point with one of the most exceptional weather events of recent times, the 2021 heatwave in the Pacific Northwest, and the less exceptional but still damaging Storm Eunice of 2022. Whenever you hear the phrase "impossible without climate change" referring to a short-term weather event, there is probably research to be done.
Myles Allen is Head of Atmospheric, Oceanic and Planetary Physics in the Department of Physics, University of Oxford, and Professor of Geosystem Science in the School of Geography and the Environment. His research focuses on how human and natural influences on climate contribute to climate change and risks of extreme weather. In 2003, Allen introduced the concept of probabilistic attribution of extreme weather events, and in 2005, the notion of a finite carbon budget, implying net zero emissions of carbon dioxide are necessary to halt global warming. He has served on the Intergovernmental Panel on Climate Change, most recently as a Coordinating Lead Author on the IPCC Special Report on 1.5°C. He was awarded the Appleton Medal and Prize from the Institute of Physics "for his important contributions to the detection and attribution of human influence on climate and quantifying uncertainty in climate predictions", featured on the BBC's "Life Scientific" as "the physicist behind net zero", was awarded a CBE "for services to climate change attribution, prediction and net zero", and is a Fellow of the Royal Society.
Oct 02: Dr. Lingwei Li
Institute of Arctic and Alpine Research, University of Colorado Boulder
Guest of the Department of Atmospheric Sciences
Understanding Changes of Global Ocean Circulation
Using Geotracers in Earth System Models
Recording Available at COMPASS ON DEMAND
Marine radiocarbon (14C) is widely used to trace deep ocean circulation, providing insight into the atmosphere-ocean exchange of CO2. During the Last Glacial Maximum (LGM), radiocarbon reconstructions suggest that the deep waters were poorly ventilated, trapping CO2 in the deep ocean and thus limiting the air-sea carbon exchange. In the first part of the presentation, I will share results from two transient climate model simulations with radiocarbon and ideal age tracers, and discuss what we can learn from these tracer-enabled simulations about the mechanisms of global ocean ventilation during the last deglaciation. Although simulated radiocarbon shows reduced ventilation at the LGM, the true model ventilation age is modestly younger mainly due to a stronger simulated glacial Antarctic Bottom Water (AABW) transport. Notably, the ocean ventilation age is considerably older around 14-12 ka than at the LGM because of a weakening of AABW transport. Our study suggests that, while 14C is the tracer most relevant to water ventilation ages, the inferred age can be different from the true age of a water mass. That means that, how radiocarbon should be used to infer ocean circulation changes, warrants further exploration. In the second part of the presentation, I will briefly introduce an ongoing project focusing on the sea ice-ocean exchange of microplastics in the Arctic, using a fully coupled Earth system model (CESM) with microplastic tracers.
Oct 09: NO SEMINAR
Oct 16: Junfei Xia
Department of Ocean Sciences, Rosenstiel School
(one-hour MPO student seminar)
Attention Tells the Story – Vision Transformer Applications in SAR Imagery and
Dynamic System Decoding From Drifter Data
Recent advancements in machine learning, particularly Vision Transformers (ViT), have opened new opportunities for analyzing complex datasets in various scientific fields. In this presentation, we explore two distinct applications of ViT within the realm of oceanographic data analysis. The first application focuses on Synthetic Aperture Radar (SAR) imagery, where ViT's attention mechanisms are leveraged to identify both structural and textural features, surpassing traditional Convolutional Neural Networks (CNNs) in classification accuracy. Using a training dataset of VV polarized SAR images, ViT achieved high precision, recall, and F1 scores above 0.97 across geophysical categories when tested on a sub-dataset of 5000 images. An additional test on the AI4Arctic Sea Ice dataset, which includes two polarized images (HH and HV) with different spatial resolutions and preprocessing steps, demonstrated an accuracy of around 90%. Despite the differences in polarization and spatial characteristics, ViT proved to be adaptable and effective in handling these variations.
In the second application, we introduce a novel approach to decoding dynamic systems from drifter data. A multilabel ViT model was first trained on drifter trajectories derived from simple simulated velocity fields. This trained model was then evaluated using real-world observational drifter data from the SPLASH and LASER experiments. The results demonstrate that ViT, after proper training, can effectively identify underlying dynamic systems from drifter trajectories, revealing phenomena such as shear flows, converging / diverging nodes, and various foci and saddle points. Central to both applications is the attention mechanism – particularly multihead attention and overall attention derived from the attention rollout method – which provides insights into the interpretability of the ViT model. These findings not only highlight the versatility of Vision Transformers in oceanographic data analysis but also propose new directions for improving the accuracy and interpretability of machine learning models in Earth science.
Oct 23: Dr. Mathieu Ratynski
Department of Atmospheric Sciences, Rosenstiel School
Oct 30: Dr. Alina Nathanaël Dossa
Department of Ocean Sciences, Rosenstiel School
Nov 06: Samantha Nebylitsa
Department of Atmospheric Sciences, Rosenstiel School
(one-hour ATM student seminar)
Nov 13: Dr. Yasushi Fujiwara
Kobe University, Japan
Nov 20: NO SEMINAR (RoRo out of town)
Nov 27: NO SEMINAR (Thanksgiving Recess)
Dec 04: Hanna Miličević
University of Zagreb, Croatia
Guest of Milan Curcic, Department of Ocean Sciences
The Effects of Submerged Sill Geometry
on the Morphodynamics of Artificial Beaches
The Croatian coastline, which stretches over 6,000 km along the eastern Adriatic, is under increasing pressure from tourism, which has led to the construction of artificial beaches, especially gravel pocket beaches. However, insufficient knowledge of appropriate design parameters and hydraulic engineering structures has led to significant material losses and costly beach maintenance. This study investigates the profitability of constructing submerged sills and their optimal geometry to mitigate erosion, stabilize artificial gravel beaches, and reduce material loss and the need for costly nourishment. The study also aims to improve beach protection and increase coastal biodiversity. Natural forces such as waves, sea currents, and freshwater springs, alongside human interventions such as beach nourishment, contribute to morphological changes that require frequent and expensive maintenance. From 2015 to 2019, Croatia spent around 6 million euros on beach restoration, highlighting the economic burden of erosion during harsh winters.
While natural gravel beaches can recover from storm events, artificial beaches are at risk of permanent material loss. Using coastal monitoring systems, including unmanned aerial vehicles and the XBeach-Gravel numerical model, this research will analyze the morphodynamics of artificial gravel beaches and the effects of submerged sills on wave energy dissipation and material transport. The results will serve as a guide for the optimal design of artificial gravel beaches in Croatia and similar low-energy coastal areas to promote sustainable coastal management, reduce long-term maintenance costs, and protect marine ecosystems.
SPRING 2025 PREVIEW
Feb 05: Yixin "Berry" Wen
Department of Geography, University of Florida, Gainesville
