Hybrid Physics ML approaches for Aerosol and Cloud Microphysics
Improving representations of aerosol and cloud microphysics in atmospheric models is key to accurately predicting future changes in climate. Since March 2020, I have been working as a research scientist at Columbia University, where I am investigating new hybrid-physics machine learning approaches for parameterizing aerosol and cloud microphysics for atmospheric models.
Recently I’ve been looking at graph neural networks for zero-shot learning of aerosol optical properties. An Arxiv preprint of this work can be found here:
Zero-Shot Learning of Aerosol Optical Properties with Graph Neural Networks
Global scale observations of aerosols.
Iron oxide combustion aerosols have recently been identified as an anthropogenic source of
atmospheric absorption, and may provide an additional source of particulate iron
to the oceans (where they could play a role in the biogeochemical cycle, ultimately impacting ocean uptake of carbon dioxide from the atmosphere).
However, the sources and atmospheric abundance of these aerosols is still highly uncertain.
To provide the first global scale atmospheric constraints on these aerosols I developed a new methodology to
characterize these aerosols in situ and analyzed data sets from 102 research flights during the
NASA Atmospheric Tomography Mission and the
HIAPER Pole-to-Pole Observations campaigns.
A paper on this project was recently published in npj Climate and Atmospheric Science:
Global-scale constraints on light-absorbing anthropogenic iron oxide aerosols
NASA Frontier Development Lab
Machine learning is a promising tool for providing new insights into important scientific challenges.
I spent 8 weeks in Mountain View, CA, as a researcher during the 2019 NASA Frontier Development Lab, a research sprint to apply AI to earth and space science challenges.
I worked with the Forecasting Geoeffectiveness team to develop a state-of-the-art data-driven machine learning approach for the prediction of space weather phenomena.
Our team received the Unexpected Discovery Award for investigating how the aurora correlates with GNSS phase scintillations at high latitudes. We presented our work during the NASA FDL Show Case at Google Cloud HQ and at the Neurips 2019 Machine Learning and the Physical Sciences Workshop:
Prediction of GNSS Phase Scintillations: A Machine Learning Approach
Correlation of Auroral Dynamics and GNSS Scintillation with an Autoencoder
Example datasets from this project can be accessed at SpaceML
Machine learning for in situ aerosol measurements
Aerosols (small particulate matter) play an important role in the atmosphere, impacting both air quality and climate.
Recently I have been exploring how advanced data analysis techniques such as machine learning can be used to improve the classification and characterization of aerosols measured in situ.
I published work on improving the classification of absorbing aerosols in Atmospheric Measurement Techniques:
Classification of iron oxide aerosols with a single particle soot photometer using supervised machine learning.
I have also been exploring deep learning approaches to classifying black carbon aerosol morphology and presented initial results as a
paper at the Neurips 2019 Tackling Climate Change with AI Workshop:
A deep learning approach for classifying black carbon aerosol morphology.
Forest fires are a major source of both primary and secondary atmospheric aerosols that impact both air quality and climate. Biomass burning aerosols such as
black carbon and brown carbon are important contributors to climate direct radiative forcing.
During the NOAA 2016 Firelab Laboratory Study at the
USDA Fire Sciences Lab,
I spent 2 months in Missoula, MT studying how absorbing aerosols emitted from biomass burning
can be characterized in situ. I collaborated with researchers from NOAA’s Chemical Sciences Division on experiments
focused on a better characterization and understanding of the optical properties of black and brown carbon sourced from biomass burning.
Several papers have been published on these experiments so far:
*Investigating biomass burning aerosol morphology using a laser imaging nephelometer
*Inter-comparison of black carbon measurement methods for simulated open biomass burning emissions.
*Evidence in biomass burning smoke for a light-absorbing aerosol with properties intermediate between brown and black carbon
*Complex refractive indices in the ultraviolet and visible spectral region for highly absorbing non-spherical biomass burning aerosol
Black Carbon sources in East Asia
Since East Asia is the most significant anthropogenic source region for black carbon (BC) aerosol, an important short-lived climate forcer,
my research during the KORUS-AQ campaign focused on providing new observational constraints
on the atmospheric abundance and vertical distribution of BC in East Asia.
The optical properties of BC evolve in the atmosphere, impacting both its climate effects
and atmospheric lifetime.
I published an overview of these new observations of BC over S. Korea in the Journal of Geophysical Research Atmospheres:
*Estimating Source Region Influences on Black Carbon Abundance, Microphysics, and Radiative Effect Observed Over South Korea
I participated in the field deployment of the NASA-NIER Korean-United States Air Quality Study, NASA KORUS-AQ,
carried out in Korea in Spring of 2016. This aircraft campaign was an international effort to improve understanding of air quality in
A CNN article discussing the campaign is here.
Some of the research papers describing results of aerosol measurements during the campaign are here:
*Investigation of Factors Controlling PM2.5 Variability across the South Korean peninsula during KORUS-AQ.
*Understanding and improving model representation of aerosol optical properties for a Chinese haze event measured during KORUS-AQ.
*Secondary organic aerosol production from local emissions dominates the organic aerosol budget over Seoul, South Korea, during KORUS-AQ
From 2016-2020, I worked as a research scientist at the Cooperative Institute for Research in the Environmental Sciences
at NOAA’s Earth System Research Laboratory in Boulder, CO.
My research focused on the sources, optical properties, and atmospheric lifetime of black carbon (BC).
BC is an aerosol (small particulate matter) emitted during incomplete combustion processes,
from both natural (forest fires) and anthropogenic (vehicle exhaust) sources.
BC is considered the most important short-lived climate forcer, since it strongly absorbs solar radiation, directly heating the atmosphere.
To better characterize this aerosol in the atmosphere, I took part in several laboratory and aircraft
field measurements using single particle soot photometers, the state-of-the-art approach
for in situ characterization of these aerosols.
The isotopic composition of water is an important tracer of geophysical and atmospheric processes,
as the preferential deposition of heavy water as ice can provide information about both the
sources of water and its past history.
My doctoral thesis focused on experimental characterization of isotopic water vapor at low temperatures
as a tracer of cirrus cloud microphysics. This research provided the first experimental verification of the isotopic fractionation factors of water vapor at
low temperatures, in both equilibrium and non-equilibrium conditions characteristic of the upper atmosphere, and was published in PNAS:
Laboratory measurements of HDO/H2O isotopic fractionation during ice deposition in simulated cirrus clouds
Cirrus Cloud Microphysics
Stratospheric water vapor acts as a green house gas, with important implications for climate. Cirrus clouds in the upper tropical tropopause layer (TTL)
regulate the amount of water vapor entering the stratosphere.
Because the TTL is very cold and dry, water vapor in these conditions is very challenging to measure accurately from aircraft and balloon platforms.
During my Ph.D. I worked on the design, construction, and characterization of a mid-infrared spectrometer to measure atmospheric water vapor and its isotopic composition in conditions characteristic of the TTL.
To characterize this instrument in realistic atmospheric conditions, I participated in several research campaigns at the
AIDA Aerosol Interaction and Dynamics in the Atmosphere Chamber at KIT.
Aquavit II systematically tested state-of-the-art aircraft instrumentation for measuring atmospheric water vapor over a range of different conditions.
The IsoCloud campaigns focused on characterizing ice growth in ultra-cold cirrus clouds and the fundamental properties of isotopic water vapor at these temperatures.
A paper discussing the instrument we developed is here:
The Chicago Water Isotope Spectrometer (ChiWIS-lab): A tunable diode laser spectrometer for chamber-based measurements of water vapor isotopic evolution during cirrus formation
Results on characterizing the microphysical properties of ice in cirrus clouds is here:
No anomalous supersaturation in ultracold cirrus laboratory experiments
I studied for both my M.S. and Ph.D. in physics at the University of Chicago, where I was
an NSF graduate fellow and an NDSEG graduate fellow. I graduated with my Ph.D. in the summer of 2015.
The title of my thesis was
In situ isotopic water vapor measurements as a tracer of cold cloud microphysics.
I studied for my B.S. in physics at the University of Illinois in Urbana-Champaign.