The Cooperative Institute for Modeling the Earth System (CIMES) has announced awards to support eight innovative, cross-disciplinary projects aimed at modeling and understanding the Earth system which are aligned closely with the strategic goals of NOAA's Geophysical Fluid Dynamics Laboratory (GFDL). The projects run from 2024 to 2025 and foster research, teaching, and mentorship in Earth system science.
The funded projects include:
Parametrization of atmosphere-surface exchanges in the marginal ice zone
Elie Bou-Zeidm (PU) and Mitch Bushuk (GFDL)
Rapid warming of the Arctic and Antarctic is creating a cascade of environmental changes, the impacts of which will not remain confined to polar regions. One of the most consequential geophysical transformations is the ongoing and substantial reduction in the spatial extent of Arctic Sea ice, its seasonal duration, and thickness. The sensitivity of Arctic sea ice to the polar amplification of global warming is consistently underestimated by a wide suite of climate models. Despite recent advances in CMIP6, many open questions remain concerning the performance of models and the missing processes that need to be adequately captured, particularly in the marginal ice zone (MIZ), or over leads and polynya, where surface heterogeneity is a prominent feature.
The overarching hypothesis of this project is that “novel, scale-adaptive parametrization structures can account for the direct and indirect nonlinear effects of surface heterogeneity in the MIZ, and improve model performance in polar regions”. To test this hypothesis, Drs. Bou-Zeid and Bushuk will answer the following research questions:
Q1: How can the accumulated knowledge on the interaction of a heterogeneous ice pack with the atmosphere be reduced to a succinct dimensionless framework that accurately accounts for all primary driving variables (including temperature contrast, roughness contrast, geostrophic pressure gradient, ice fraction, patch/flow characteristic length scale, ice/water aggregation, among others)?
Q2: How can this framework be applied to develop a scale-adaptive subgrid parametrization that allows climate models to simulate realistic vertical and horizontal fluxes of momentum, mass, and energy through grid cells overlying heterogeneous sea ice, robustly across resolutions?
Development and parameterization of a trait-based model of zooplankton diversity for marine food web and climate feedback studies
Curtis Deutsch (PU), Justin Penn (PU), and Charlie Stock (GFDL)
Zooplankton mediate the flux of carbon from the surface to deep ocean, making them a central part of the “biological pump” that reduces atmospheric CO₂ levels. These fluxes are governed by zooplankton metabolic rates, which vary widely among species, and rise with body size, temperature, and activity level, requiring a higher ambient O₂ supply to sustain them. This trait diversity shapes the biogeography of marine animal species, influencing their ability to compete for resources. Existing Earth system models (ESMs) neglect the interspecific variations in these traits (if they are represented at all) as well as the aerobic habitat constraints imposed by the interactive effects of temperature and O₂, precluding a mechanistic examination of how diverse marine animal metabolisms impact global carbon cycling and its response to climate.
This project seeks to develop a trait-based representation of zooplankton diversity and its role in the carbon and O2 cycles and marine food webs, and to parameterize this diversity for computationally tractable simulations of climate responses and feedbacks. The model builds upon work by the PI and Co-PI to develop an ecophysiological framework that links empirically-calibrated physiological traits for temperature-dependent metabolism and oxygen (O₂) tolerance, to the biogeography and biodiversity of marine species. It will interface with projects underway to add zooplankton behavior to the GFDL ecosystem model, COBALT (Luo, GFDL; Resplandy, Princeton). It also has strong synergies with efforts at GFDL to develop and deploy models of upper trophic levels (e.g. FEISTY) and would help to advance those efforts.
Global storm-resolving modelling studies of climate change
Timothy Merlis, Stephan Fueglistaler, Gabriel Vecchi, Michael Oppenheimer, Jessica Metcalf, Brian Grenfell (all PU), Lucas Harris, and Tom Knutson (both GFDL)
The area of global simulations of atmospheric climate change at convection-resolving scale (order 3km) is an area that has emerged over the last two years as one that is complementary to existing scientific objectives at both Princeton University and GFDL with immense potential for synergies. The project builds on GFDL's world-leading expertise in global storm-resolving modeling and the numerous applications for such simulations for a global warming scenario.
The named co-investigators are particularly interested in simulations that do not require downscaling for assessment of extreme events, high-resolution surface layer climate in the context of diseases spread by aerosols such as COVID-19, and cloud changes and how they affect climate sensitivity. The data produced in the project will be of great interest to a wide range of scientists beyond the named Co-Is, and it is the stated objective to produce a dataset that can be used by anyone within the GFDL/CIMES/PU frame, and potentially beyond.
Investigating the Natural Variability of Tropical Cyclone Hazards
Ning Lin (PU) and Hiroyuki Murakami (GFDL)
Tropical cyclone (TC) activity exhibits natural variability and is influenced significantly by climate change, both of which have been explored extensively in previous studies (e.g., Chu and Murakami 2022, Knutson et al. 2020). Recent research has also investigated how long-term hazards (extreme wind, heavy rainfall, and storm surge) induced by landfalling TCs will change due to climate change (e.g., Gori et al. 2022, Xi and Lin 2022a, Xi et al. 2023). However, there is a lack of knowledge of the natural variability of the TC hazards. This insufficient comprehension undermines the reliability of seasonal to decadal forecast of TC hazards, challenging pre-season preparation of TC impact and decadal-scale planning of risk mitigation strategies. In this proposed research, Dr. Lin and team aim to investigate the natural variability of TC hazards and explore applications in seasonal forecasting and risk mitigation.
The Seamless System for Prediction and Earth System Research (SPEAR) developed by GFDL is the next generation climate model for seasonal to multidecadal prediction and projection (Delworth et al. 2020). SPEAR is specially designed to perform large ensemble prediction of climate conditions in seasonal to decadal scales using reasonable computation resources. The Princeton environmental-dependent probabilistic tropical cyclone (PepC) model, developed in Lin’s group, is a statistical TC model that can downscale climate models and generate large numbers of physically plausible TC events (e.g., thousands synthetic TCs used for assessing risk) with low computational costs (Jing and Lin 2020).
In this project, Drs. Lin and Murakami will integrate SPEAR, PepC, and TC hazard models (developed or extensively used in Lin’s group) to investigate the natural variability of TC hazards. To support practical applications, they will examine the predictability of TC hazards in different time scales. They will also explore the role of the natural variability of TC hazards in long-term TC risk, to support risk mitigation through building code revision.
GCM simulations of atmosphere--wildfire coupling in a changing climate
Timothy Merlis (PU), Elena Shevliakova (GFDL), Sergey Malyshev (GFDL), Denise Mauzerall (PU), Wei Peng (PU), and Gabriele Villarini (PU)
Wildfires are incredibly dangerous and expensive natural disasters. There is understanding of the factors that contribute to 'fire weather', such as low relative humidity and high surface windspeeds, and these factors play important roles in determining wildfire spread. There have been a substantial number of observational studies connecting observed fire weather conditions to burned area in wildfires (e.g., Jones et al. 2022). However, the impact of climate change on wildfires in recent decades and in future projections are areas of climate science with substantial uncertainty. This uncertainty is partly from the relative paucity of systematic investigation of climate models with dynamic vegetation under future climate scenarios.
Wildfire models determine burned area from a convolution of the number of fires (scaling ignition by factors determining the probability that a fire starts) by the rate of spread, duration, and burnable vegetation. Ignition can be natural (lightning, in particular) or anthropogenic and the classic fire weather variables like surface relative humidity and windspeed are important to both the number of fires and the rate of spread. Fuel loads depend on a mix of factors that include short and long timescales: drought, past season rainfall, history of past years' wildfire, and the forest growth. This overview makes clear that there are atmospheric factors and land surface factors that are key to determining the future evolution of wildfires. The development of wildfire modelling within the context of the dynamic vegetation land model of GFDL's LM4.1 offers an exciting opportunity to do high-impact research on atmosphere--wildfire coupling in a changing climate. The proposed research here uses GCM simulations designed to tackle this important aspect of modeling the earth system, but the simulations have wider applicability to investigate extreme event and hydrologic changes.
This project will use a combination of atmospheric--land GCM simulations and uncoupled land model simulations to examine the role of atmospheric and large-scale climate uncertainty on wildfire projections within the LM4.1 model. Here, we are focused on the impact of climate model uncertainty on wildfire projections using a land model that can exhibit abrupt changes (Martinez-Cano et al. PNAS, 2022)
Impact of Increases in Climate Change Induced Wildfires on Air Quality and Public Health in North America
Denise L. Mauzerall (PU), Meiyun Lin (GFDL), Larry Horowitz (GFDL), and Yuanyu Xie (PU)
Increases in wildfires in North America over the past decade have increased public concern about the role of climate change in driving wildfires and their detrimental impact on air quality and public health (Burke et al., 2023; McClure et al., 2018). Wildfire smoke can result in substantial exceedances of air quality standards at locations distant from the fire locations. The frequency, intensity and extent of wildfires is likely to increase as climate change accelerates. The previous generation of climate models (e.g., GFDL ESM4 used for CMIP6) included wildfire emissions of CO2 that responded to climate and land use changes, but air pollutant emissions from fires were prescribed on the basis of integrated assessment model projections that did not fully account for the effects of climate change (Feng et al., 2020; Xie et al., 2022). As a result, the full impact of future fires on air pollutant concentrations and public health could not be evaluated.
In this project, Dr. Mauzerall and team aim to develop an integrated modeling framework that establishes critical links between climate change, wildfires, air pollution and public health. We plan to utilize the GFDL chemistry – climate model AM4.1 (Horowitz et al., 2020) with updated wildfire smoke emissions from the United Nations Task Force on Hemispheric Transboundary Air Pollution (TF HTAP). They will conduct a suite of AM4.1 simulations with prescribed sea surface temperature (SST) and sea ice distributions to examine how increasing wildfire smoke in various Shared Socioeconomic Pathways - Representative Concentration Pathways (SSP-RCP) scenarios affect the ability of the US to meet national PM2.5 air quality standards. These simulations, integrated with a health impact assessment model, will allow the team to build on earlier CMIP6 simulations that only examined CO2 emissions from fires, to make preliminary estimates of the effect of increasing wildfire smoke on air quality and health.
Expanding the capabilities of the SHiELD modeling framework to high-resolution flood inundation modeling
Gabriele Villarini (PU), Lucas Harris (GFDL), Jan-Huey Chen (GFDL), and Gabriel Vecchi (PU)
GFDL SHiELD (System for High-resolution prediction on Earth-to-Local Domains) is a Unified Forecast System (UFS) prototype atmosphere model showing the power of a unified prediction system across a variety of time and space scales designed for a wide array of applications. It provides critical information for both weather and climate time scales from local to global domains.
The research team at Princeton University has been working on the application of flood modeling across the contiguous United States using the Hillslope-Link Model (HLM). This model has been performing very well and is able to provide hydrologic predictions at the community level anywhere along the river reaches, providing basic information to enable decision-making at the local scale and improve the preparedness against this natural hazard.
This collaborative project will build on the high-resolution outputs from SHiELD and use it as forcing for the HLM. This information will be then translated into flood inundation mapping using the Height Above Nearest Drainage (HAND) technique, providing a short-term forecasting of flood inundation at the community level.
This project will be the first of its kind to provide short-term forecasting (i.e., lead times up to few days) of flood inundation at the community level: it will synergistically combine tools and methodologies developed at GFDL and Princeton University to provide forecasts of flood impacts and fill a gap related to the availability of this information across broad regions.
Therefore, the goals of this project are to: 1) combine the System for High-resolution prediction on Earth-to-Local Domains (SHiELD) by GFDL with a physically based hydrologic model (Hillslope-Link Model or HLM); and 2) provide flood inundation forecasts with a lead time up to 5 days. Dr. Villarini and team will start with the eastern United States, with a focus on the Delaware River, and provide information along all the river reaches
Understanding the predictability and climate context of extreme hourly precipitation in extratropical-transition hurricanes
Gabriel Vecchi (PU), Sofia Menemenlis (PU), Jan-Huey Chen (GFDL), Kun Gao (PU), Mingjing Tong (GFDL), Ryan Eusebi (Caltech), Lucas Harris (GFDL)
This work is part of an ongoing collaboration between GFDL and Princeton and aims to advance the understanding of hurricane-induced extreme rainfall and climate across scales. Model development, forecasting expertise, and TC analysis at GFDL (Chen, Gao, Tong) reinforce research on climate-weather connections and hurricane impacts at Princeton (Vecchi and Menemenlis). Using T-SHiELD allows us to perform simulations with clear operational relevance. In turn, the ensemble forecasts and analysis performed at Princeton offer insight into models under active development at GFDL.
Previous work provides a framework for probabilistic analysis of an extreme event, and a case study of extreme rainfall predictability using hindcasts and observations of Hurricane Ida. We will use this framework to address additional questions related to hurricane forecasting and climate.
Building on Menemenlis et al. (2023, in review), Vecchi and team will investigate the effect of changes in the large-scale environment on hindcasts of Ida. They will use a conditional / pseudo-global warming approach (e.g., Liu et al. 2020; Jung and Lackmann 2019) in which they modify the thermodynamic environment to test the effects of a warmer & higher-moisture environment under fixed dynamical conditions. They may also perform attribution experiments to more directly address questions about how historical changes in climate may have impacted Ida’s aftermath, with perturbations taken from historical GCM simulations.
Vecchi and team also plan to use this framework to investigate the effects of urbanization on extreme rainfall. Urbanization is known to influence the magnitude and pattern of precipitation. Enhancement of precipitation downwind of urban areas is observed on climatological timescales (Liu & Niyogi, 2019), and urbanization was shown to exacerbate rainfall and flooding in hindcasts of Hurricane Harvey (Zhang et al., 2018). Targeted experiments altering surface land characteristics could test the extent to which urbanization played a role in the spatial patterns of likelihood of extreme rainfall.
An additional project step would build on Eusebi et al. (2023, in review), which demonstrated that a physics-informed neural network (PINN) may be used to reconstruct, with observations as an input, realistic hurricane wind fields. This project will investigate whether initializing a TC forecast with realistic wind fields generated by a PINN can improve track and intensity predictions.