Towards an end-to-end data assimilation system for atmospheric composition with JEDI Skylab

Maryam Abdi-Oskouei, UCAR/JCSDA

Joint Effort for Data Assimilation Integration (JEDI) is a unified data assimilation (DA) framework for Earth system prediction and reanalysis. JEDI can be used for both research and operational purposes with the objective of reducing redundant work within the community and increasing the efficiency and flexibility of transition from development to operations. This presentation summarizes the recent progress in data assimilation of aerosol and trace gases using the JEDI framework in the context of the SkyLab model (GOCART and GEOS-CF).

First our generic infrastructure will be presented covering the different building blocks of the JEDI system. The presentation will cover the novel advances and efforts for aerosol DA (using VIIRS and MODIS AOD). We will use an Ensemble of Data Assimilation (EDA), along with other developments incorporated from the JEDI community. Lastly, we will show the recent developments and future directions for trace gas DA using TropOMI CO total column and NO2 tropospheric column retrievals. A novel 4D Ensemble Variational (4DEnVar) approach is being developed to constrain both concentrations and surface fluxes using satellite observations.
 

High-resolution WRF-Chem modeling of June 2022 ozone exceedance events in the Lake Michigan region

Jerrold Acdan, University of Wisconsin-Madison

Many counties along the coastline of Lake Michigan experience surface ozone concentrations exceeding the National Ambient Air Quality Standards set by the U.S. Environmental Protection Agency. This air quality problem affects more than 10 million people living in the states of Illinois, Indiana, Michigan, and Wisconsin and is thus an important topic of research. In this work, we utilize the Weather Research and Forecasting model coupled with Chemistry (WRF-Chem) to study the formation and transport of ozone air pollution in the Lake Michigan region. With June 2022 ozone exceedance events as case studies, we conducted sensitivity tests to investigate the impact of: (1) using the Building Environment Parameterization (BEP) urban physics option and (2) incorporating soil moisture analysis data derived from Soil Moisture Active Passive (SMAP) satellite observations on the simulation of meteorological variables (e.g., wind speed and direction, planetary boundary layer height, etc.) and air pollutant transport within the model.

A Quantile Conserving Ensemble Filtering Framework: Next Generation Nonlinear and Non-Gaussian Data Assimilation for Tracers

Jeffrey Anderson, National Center for Atmospheric Research

Ensemble Kalman filters make implicit assumptions about Gaussianity and linearity. This presentation describes novel efficient algorithms that can use arbitrary continuous priors and likelihoods for an observed variable avoiding the need for assumptions of Gaussianity. The key innovation is to select posterior ensemble members with the same quantiles with respect to the continuous posterior distribution as the prior ensemble had with respect to the prior continuous distribution. While this method leads to improvements in analysis estimates for observed variables, those advantages can be lost when using standard linear regression to update other state variables. However, doing the regression in a transformed bivariate space guarantees that the posterior ensembles for state variables have all the advantages of the observation space quantile conserving posteriors. For example, if state variables are bounded then posterior ensembles will respect those bounds. The posterior ensembles also respect other aspects of the continuous prior distributions. This avoids any implicit assumptions of linearity in the ensembles. Examples are shown for a variety of bivariate prior ensembles including bounded quantities and multimodal distributions. Results from cycling assimilation for an idealized tracer transport model demonstrate significantly improved data assimilation for distinctly non-Gaussian quantities like tracers in Earth system models.

How do we underestimate the impact of dust events on air quality

Karin Ardon-Dryer, Texas Tech University

Dust events are an important and complex constituent of the atmospheric system that can impact Earth’s climate, the environment, and human health. Most of the estimations of the impact of dust events on air quality are based on daily PM10 and PM2.5 concentrations (particulate matter with an aerodynamic diameter <10 and 2.5 μm, respectively). Many dust events (even some extreme dust storms) do not go above the EPA daily thresholds, the reason lies in the time scale of the measurements. Observation of PM during dust events which are based on daily and hourly values showed a low impact of dust events on air quality (even during some extreme dust storms). But these hourly and daily measurements underestimate the true impact of convective dust events. Observations based on a shorter timescale (10 min) reveal the true impact of short-duration (convective) dust events on air quality, leading us to speculate that the impact of these dust events is underestimated with the current (hourly basis) method. It is speculated that dust events contain mainly large particles but measurements during different dust events in Texas showed an increase of concentrations of particles <2.5 μm. Examples from several extreme dust events from Texas and Arizona will be presented.

An evaluation of lidar derived ozone curtain profiles from the TRACER-AQ campaign and WRF-Chem simulation

Claudia Bernier, University of Houston

The Houston-Galveston-Brazoria (HGB) region is susceptible to unique local emissions, transport, and meteorological factors that lead to large tropospheric ozone variations. A consistent and highly detailed variety of measurements is essential in understanding the development and behavior of tropospheric ozone in this region. The 2021 Tracking Aerosol Convection ExpeRiment – Air Quality (TRACER-AQ) campaign addressed this need by utilizing ozone lidars, sondes, off-shore boat, aircraft, and a string of additional measurements to capture a wide range of ozone cases in the HGB region for the month of September. Using the multi-dimensional ozone lidar curtain profiles, we investigate diverse vertical and temporal fine tropospheric ozone cases. The WRF-Chem model is run for the course of the campaign period at nested domains at 12, 4, and 1.33 km. Using TROPOMI satellite retrievals, we infer and test the current emissions used in the simulation. We evaluate the model’s ability in capturing the varying temporal and vertical ozone structure during two distinct ozone episodes.

Influence of meteorological conditions on PM2.5 concentrations in Ho Chi Minh city, Viet Nam: a merging measurements and WRF/CMAQ models approach

Long Bui Ta, Ho chi minh city university of Technology (Bach Khoa University)

The strict clean air action plan in Vietnam launched in 2016 aims to reduce PM2.5 pollution. However, the socio-economic development in HCMC still has to use old technologies, so the plan to reduce pollution is needs to be clarified the dependence of PM2.5 pollution on both emission factors and meteorology. In this study, the coupled WRF/CMAQ models is used to assess the dependence of PM2.5 pollution level on meteorological factors. Three years from 2018 to 2020 were considered. To calibrate and validate the WRF model, a set of measured meteorological data is used. The CMAQ model is used to simulate the distribution of PM2.5 spatiotemporal distribution. Group of six meteorological factors including wind speed (Ws), temperature, precipitation, relative humidity, surface pressure, height of the Planetary Boundary Layer (PBL) are selected for consideration. The results show that, in the period of 2018 - 2020, in the dry season, the concentration of PM2.5 in the previous hour is considered to be the main factor contributing (72%), followed by temperature (8.8%), precipitation (8.3%), surface pressure (6.8%), wind speed (3.5%). In the wet season, the concentration of PM2.5 in the previous hour is still the main factor contributing (43%), however meteorological factors in the wet season have a significant contribution higher than that in the dry season, specifically wind speed (20.3%), temperature (16.6%), precipitation (10.5%), surface pressure (6.1%).
 

Global sectional aerosol microphysics simulations the January 2022 Hunga Tonga Eruption

Parker Case, NASA

We have used a sectional microphysics module within the NASA GEOS Earth system model to simulate the volcanic aerosol plume following the January 2022 Hunga Tonga eruptions. While OMPS nadir mapper observations show the injection of sulfur dioxide was moderate (~0.4 Tg) [Carn et al., 2022], the MLS shows the amount of water vapor injected during the eruptions was immense (>100 Tg) [Millán et al., 2022], with early model simulations projecting the water to impact the composition of the stratosphere for up to a decade. The presence of this amount of water creates a unique natural experiment. Using the NASA GEOS Earth system model coupled to the CARMA sectional aerosol model and the GEOS-Chem tropospheric and stratospheric chemistry mechanism, we have simulated the extent of these effects, focusing on the microphysics and impacts of the sulfate aerosols resulting from the unique conditions of the Hunga Tonga plume. We evaluate our size-resolved model with observations including balloon-born particle size observations from La Réunion (21°S, 55°E) as well as spectral aerosol extinction information from the OMPS-LP satellite instrument. Our simulations are useful in showing how this injection of water and aerosol precursors are impacting the temperature, ozone chemistry, oxidizing power, and aerosol microphysical properties of the stratosphere and complement the story told by satellite, ground-based, and balloon-borne observations of the volcanic plume.

Development and Implementation of a Novel Bias Correction Technique for GFSv15-CMAQv5.3.1 Air quality Forecasting System

Xiaoyang Chen, Northeastern University
Will be presented by Yang Zhang on behalf of Xiaoyang Chen

The offline-coupled Global Forecast System with the Community Multiscale Air Quality modeling system (GFS-CMAQ) has been developed by the U.S. National Oceanic and Atmospheric Administration (NOAA) as the intermediate system of the next-generation National Air Quality Forecast Capability (NAQFC). The GFSv15-CMAQv5.3.1 system is employed by NOAA to perform air quality forecast studies for the period of Jul to Dec of 2019 over the contiguous U.S.. Evaluation of the predictions against ground-based observations indicates systematic biases for predicting ozone (O3) and PM2.5 in specific regions and seasons. In this work, a novel bias correction method is developed for the above system to reduce the systematic biases and improve its forecast skills. This method is based on integrating three techniques (Kalman filter, Kolmogorov–Zurbenko filter, and inverse distance weighted interpolation) and is applied to postprocess the predictions of O3, PM2.5, and key precursors such as NOx, SO2, CO, and isoprene. The integrated technique can meaningfully reduce the errors and absolute biases in forecasting these species. The method is then implemented in forecast runs and examined. This study demonstrates that the technique can effectively reduce forecast biases not only for predicted O3 and PM2.5, but also for their key precursor species. The integrated technique can be potentially applied to reducing systematic biases for other air quality forecasting systems.

Turbulence Time Series Analysis for Partitioned Methane Fluxes from Reservoirs

Corrin Clemons, University of California, Davis, Department of of Civil & Environmental Engineering

Global fresh-water methane emissions are historically underestimated due to their difficulty to predict and measure as they are driven by complex environmental, biogeochemical, and physical interactions in the sediment, water column, and local atmosphere. To better understand such systems, we began an ongoing study in May 2023 measuring GHG fluxes to the atmosphere using the eddy-covariance technique at Uvas Reservoir in Morgan Hill, California, managed by Santa Clara Valley Water District. Ebullition (bubbling) fluxes can contribute a significant proportion of methane releases from reservoirs in Mediterranean climates yet is sporadic and challenging to quantify. Our high-frequency measurements allowed us to partition diffusive and ebullitive fluxes through wavelet transformations that shift discrete oscillations, or wavelets, across the data set to generate wavelet coefficients. Coefficients for the scalar values of temperature, water vapor, and carbon dioxide were correlated to those for methane to evaluate the scalar transport similarity. Deviations from a linear correlation signaled heterogenous source distributions caused by ebullition. This partitioning method indicated a significant role of ebullition in total methane emissions and produced diffusive fluxes similar to those measured by point chambers. Future work will incorporate these findings with sediment and water measurements over larger timescales to understand the mechanisms driving these fluxes in the system.

Urban Air Quality Across the Globe with MUSICAv0

Louisa Emmons, NCAR/ACOM

Air quality is primarily driven by local anthropogenic emissions sources, but it can also be strongly influenced by long-range transport of pollutants and regional influences (natural emissions, chemistry, meteorology, climate). In turn, local air quality can have impacts that extend all the way to the global scale.  MUSICAv0, MUlti-Scale Infrastructure for Chemistry and Aerosols Version 0, a configuration of CESM2.2(CAM-chem) with variable resolution, now has the unique capability to simultaneously simulate urban-scale air quality with high horizontal resolution and regional-to-hemispheric-to global influences and impacts of pollutants, while still using an expensive chemistry and aerosol scheme.  Making use of new computer resources at NCAR, a custom variable resolution mesh, with 3 refined regions of special interest targeting the United States, Europe and southern and eastern Asia, is being used to simultaneously simulate air quality at city scales and hemispheric-scale transport of pollutants. This grid will also replicate the coverage of the three geostationary satellites that will soon provide atmospheric composition observations (GEMS, TEMPO, and Sentinel-4).  Prior to that data being available, these model simulations will help prepare for analysis of the constellation observations; and in the near future retrievals from GEMS over East Asia and TEMPO over the U.S. will be used for model evaluation.

On The Role of Simplified Models in Understanding Background Tropospheric Ozone and its Contributions to Maximum Surface Concentrations Across the Continental US

Ian Faloona, UC Davis Air Quality Research Center
Coauthors:  D.D. Parrish, R.G. Derwent, and C.A. Mims

Despite largely successful regulatory control efforts, several areas of California still exceed the ozone National Ambient Air Quality Standard (NAAQS) of 70 ppb, and this threshold is continually lowered as the health impacts of exposure become better understood. As precursor emissions continue to decrease, ozone levels appear to be reaching a plateau, and consequently, the nature and secular trend of hemispheric background ozone is becoming more important as its proportional contribution to exceedances of the NAAQS are on the rise. Moreover, State Implementation Plan (SIP) modeling of the trends of ozone across California have erroneously predicted attainment of the NAAQS which, in fact, have never been observed. While models of reduced complexity are routinely used in geophysical fluid dynamics to study complex phenomena, the use of analogous models in the atmospheric chemistry community is anathema. We present here a conceptual model of the midlatitude background tropospheric ozone field and outline how it impacts surface concentrations across the continental US depending on geography and boundary layer dynamics. We further show that after the mid-2000's background ozone has been gradually decreasing; however, we also call attention to several recent studies indicating a rise in the ozone flux from the stratosphere due to an accelerated Brewer-Dobson circulation and weakening stability caused by rising levels of greenhouse gases.
 

Combining High Spatial Resolution Fire Information with Daily Fire Activity to Improve a Fire Emissions Estimates

Sam D. Faulstich, University of Utah, Department of Chemical Engineering

Modeling human exposure to wildfire smoke is a crucial component of estimating the health impacts of wildfire smoke. In order to estimate smoke exposure, the amount of PM2.5 emitted by each fire is required. Because fire emissions cannot be directly measured, a model called a fire emissions inventory (FEI) is required to estimate the emissions information. FEIs face uncertainties because fire behavior is complex and difficult to model. These uncertainties are further propagated through atmospheric dispersion models and health studies that use FEIs.

Satellite remote sensing is necessary in FEIs due to the large amount of area that needs to be surveilled on a daily basis. However, satellite remote sensing misses fire activity due to cloud cover and sensing limits, leading to underestimated fire activity. This work combines two sources of satellite remote sensing information to provide fire emissions estimates that are high resolution both spatially (30 m resolution) and temporally (daily resolution), reducing errors caused by missing small fires. A linear regression is applied in order to account for missing fire activity likely due to cloud cover. Initial results show that for 2013, approximately 25% more fire activity days are included when using the cloud cover interpolation.

Evaluation of the HYSPLIT-WRF-Chem framework to simulate volatile phenols under wildfire conditions. Case study: two wildfire smoke events at a central Washington State winery.

Ana Carla Fernandez Valdes, Washington State University

The presence of volatile phenol (VP) in the atmosphere can negatively influence human health and the environment. VPs can also accumulate in grapes causing detrimental effects on the wine industry. The primary source of high concentrations of VPs in wine comes from smoke contamination from wildfires, commonly known as smoke taint. These compounds can impart smoky, ashy flavors to wines which affect their quality and marketability. During the past decade, western US wineries have been impacted by this phenomenon, especially during the 2020 season. In this study, we investigate the capabilities of an air quality modeling framework to provide information on two wildfire events in a winery in central WA. Several chemical mechanisms within the WRF-Chem model were used to investigate their capabilities to simulate the behavior of VP in the troposphere, including their sources, chemical reactions, transport, and fate during both smoke-tainted and non-smoke-tainted conditions. Additionally, we performed back trajectories simulations using the HYSPLIT model to identify the sources, pathways, and travel time of the air parcels that arrive at the wineries. Combining the outputs of the chemical transport model and the results from a back trajectory analysis we were able to characterize the photochemical aging of smoke plumes and identify the fate of VP at both the wineries locations and through the trajectories.

Assessment of modern climate change on the territory of Central Asia.

Sh. Khabibullaev Forukh Boltabaev, METEOINFOCOM

Modeling of time series of air temperatures and precipitation was carried out for 75 meteorological stations of the Republic of Uzbekistan from 1966 to the present. It was revealed that the greatest manifestation of climate change affects the temperature in July and its stepwise growth has been observed since 1973. At meteorological stations near the Aral Sea, there was a second increase in the average temperature in 2007, associated with the desertification of the territory. The increase in January temperature is insignificant and is observed at several individual stations, as well as the manifestation of modern climate changes in the precipitation series is practically not observed due to their high natural variability.

Predicting major pollutant concentrations and linkages to emissions, meteorology and policy implications in Beijing, China using machine learning methods

Shreya Guha, George Mason University

Air pollution is a major contributor to adverse health outcomes, with particulate matter of aerodynamic diameter less than 2.5um(PM2.5) and tropospheric or surface ozone being the dominant sources for premature mortality globally. The concentrations of these pollutants are highly dependent on both source emissions and meteorological factors. In 2013, China launched the Air Pollution Prevention and Control Action Plan to cut the PM2.5 source emissions, which helped in reducing its concentrations in Beijing. However, the city also experienced an increase in ozone concentration since 2013. This study aims to evaluate and quantify the effects of daily meteorology on these heterogeneous changes in ambient air pollutant concentrations in Beijing from 2011 to 2020. For this, other than relying on a specific dataset, we have assessed and combined several datasets ranging from ground observations to model reanalysis products. This study employs detrending technique in combination with the application of an array of models with increasing levels of intelligence to develop the best possible model for separating out meteorological influences. Thus, we can successfully identify the trend for the anthropological contributions to the changing air quality levels in response to the recent policy implementations in Beijing. Having this information can help in extending the use of this model across various other locations and building more effective environmental policies in the future.

Harnessing our Air Quality Modeling & Observational Capabilities to Establish Key Factors Influencing Ozone Levels in Arizona

Yafang Guo, The University of Arizona

Currently we do not understand how the unique southwest semi-arid environment and potential sources of main O3 precursors in Arizona (NOx, AVOCs, BVOCs) influence O3 abundance across the state so it is not clear which types of controls can be put in place. Because of the unique southwest natural environment including weather, climate, and desert plants in Arizona, it is also important to understand how the extreme heat, low moisture, and all year desert shrubs contribute to the O3 production and hence better control the O3 exceedance and air quality forecast. The objectives of this work are: 1) conduct O3 simulations for the recent decade and evaluate the hourly-to-decadal variations of simulated O3 and associated compounds with existing USEPA AQS datasets; 2) assess how the urban area shifts between NOX-limited and VOC-limited regime over the year through a series of model experiments as weather and desert ecosystem influences O3; 3) quantify the relative contributions of the following: transport of O3 precursors from nearby states (e.g., California, northwest Mexico) and fires. We use a “tagging” approach within the WRF-Chem model that attributes elevated O3 concentration in Arizona to nitrogen oxide (NOx ) emissions from within and outside the state during June month from year 2017 to 2021. Model is configured with two one-way nested domains with 9- and 3-km grid spacing, with the outer domain encompassing the western US and inner domain covering entire state of Arizona.

Investigating the role of nocturnal heterogeneous chemistry on daytime air quality: a comparison of two modeling schemes

Alicia Hoffman, University of Wisconsin - Madison

Nitrogen oxides (NOx ≡ NO + NO2) have a major effect on air quality in the United States because they have direct human health impacts and play a central role in the production of secondary pollutants like ozone and particle pollution. Nighttime heterogeneous chemistry – the chemistry occurring between particles and gases – regulates the reservoirs and sinks of NOx, leading to large impacts on daytime air quality. However, the simplified nocturnal heterogeneous NOx chemistry in air quality models is not consistent with field measurements and results in poor prediction of NOx reservoir species concentrations such as N2O5 and ClNO2. To assess the impact of nocturnal heterogeneous chemistry on daytime NO2 concentrations, we compared two different chemistry schemes in the Community Multiscale Air Quality (CMAQ) model. The two schemes we assessed were the default setting in CMAQ and an updated scheme that had a new N2O5 uptake parameterization and a new ClNO2 yield parameterization. This new chemistry incorporated the effects of organic coatings on particles to reduce N2O5 uptake and the competitive effects of sulfate in particles to reduce ClNO2 yield. These heterogeneous chemistry mechanisms have large impacts on the particle nitrate composition and early-morning NO2 concentrations, which impact human health, environmental quality, and the climate.

Meteorological impacts on the spatial distribution of air pollution in Salt Lake City

Heather Holmes, University of Utah

Meteorology plays a significant role in poor air quality during winter in mountainous regions. In mountain valley basins, local stagnation events lead to pollutants being trapped in cold dense air, a cold air pool (CAP). One location that experiences CAPs is the Wasatch Front in northern Utah, home to more than 2.5 million people. Salt Lake City (SLC) is the largest city this region. There are multiple sources of air pollution emissions in SLC: motor vehicles, aircraft, industrial, and dust from the Great Salt Lake. While the emissions do not change during CAPs, the physics and chemistry of the atmosphere changes, resulting in different PM2.5 composition. The spatial distribution of regulatory air quality monitors is limited, e.g., three speciated PM2.5 monitors in the Wasatch Front with one in SLC. More sensors are available for PM2.5, but not enough to quantify the intra-urban spatial variations. The spatial coverage can be improved by using data form the AQ&U sensor network, with more than 200 PM2.5 sensors in SLC. Another limitation in studying wintertime air pollution is quantifying the meteorological drivers contributing to the poor air quality. This presentation will show results that illustrate the meteorological impacts on the spatial distribution of pollutants and the composition of PM2.5 during CAPs in SLC. We will present a new method to identify CAPs and the CAP influence on the spatial PM2.5 concentrations will be presented using results from AQ&U.

Constraints on anthropogenic NOx, CO, and VOCs emissions over the US by assimilating multi-constituent TROPOMI satellite measurements

Chia-Hua Hsu, Department of Mechanical Engineering, University of Colorado, Boulder

NOx, CO, and VOCs are air pollutants and precursors to PM2.5 and O3 air pollution. The diverse and dynamically evolving natural and anthropogenic sources of these emissions pose a challenge for emissions estimation. This work aims to simultaneously optimize NOx, CO, and VOCs anthropogenic emissions by assimilating multi-constituent (e.g., NO2, CO, CH2O, and O3) TROPOMI observations and studying the impact of emissions inversion on O3 simulation/forecast. In this research, WRF-Chem/DART, a regional ensemble chemical weather forecast and data assimilation system, is used to conduct the emissions inversion on a daily basis. The potential study periods are focusing either on April 2020, which is during the COVID-19 lockdown, or the AEROMMA field campaign (summer 2023). We will first constrain NOx and CO emissions by assimilating TROPOMI NO2 and CO measurements, respectively. For inverse modeling of VOCs emissions, we will jointly assimilate TROPOMI CH2O and O3 data and study the capability of these observations to constrain various VOCs emissions species. We will also investigate the impact of conducting emission inversion in a non-Gaussian framework (e.g., log-normal, and non-Gaussian filter), which may improve the performance of emissions estimates because the distribution of trace gas concentrations and emissions tends to be non-Gaussian.

Towards Improved Understanding of Wildfire Smoke Plume Height Estimation in Western U.S. Using Multisource Satellite Observations

Jingting Huang, University of Utah

As wildfires intensify and fire seasons lengthen in western US regions, accurate models that predict smoke density and location are becoming increasingly critical. Smoke plume height (SPH) is the key to understanding wildfire intensity and aerosol sources in climate and air quality models. Global-scale satellite data with improved spatiotemporal resolution can aid in estimating wildfire emissions and SPH.

This study evaluated different satellite remote sensing techniques used to retrieve wildfire SPH using aircraft data from the Wyoming Cloud Lidar (WCL) during the 2018 BB-FLUX field campaign. Two definitions— “plume top” and “extinction-weighted mean height” —were used to derive representative heights of wildfire smoke plumes, based on aerosol extinction coefficient vertical distribution from the WCL measurements. Smoke plume height products retrieved from multiple satellite sources were intercompared and analyzed with these two definitions for SPH.

Results indicate that plume height products from MISR can capture the topmost aerosol layer, whereas TROPOMI tends to represent the smoke aerosol optical centroid. Besides, the MODIS/MAIAC algorithm-derived injection heights exhibit high uncertainties, and the VIIRS/ASHE algorithm-derived aerosol layer heights are biased high due to their coarse spatial resolution. These findings are expected to inform future satellite remote sensing missions and Earth observation data selection for mapping smoke aerosol vertical distribution.

UAV Measurements in a Heavily Burdened Air Basin to Understand Meteorological and Emissions Uncertainties in CMAQ

Cesunica Ivey, University of California, Berkeley

Air quality non-attainment is a well-known challenge in the South Coast Air Basin of southern California. The region suffers due to the compounding effects of onshore winds, opposing mountain ranges, abundant anthropogenic emissions, and persistent sunlight to catalyze the formation of secondary pollutants, namely ozone and secondary particulate matter (PM). We explore concentrations of ozone and PM2.5 at the surface and up to 500 meters using an unmanned aerial vehicle in Riverside, CA from August to November 2020. Resulting from this effort were 376 vertical profiles of ozone and PM2.5. The Community Multiscale Air Quality (CMAQ) model was run to generate comparison ozone and PM2.5 concentrations and to better understand model emissions and meteorological discrepancies. Several simulations were run to determine model bias after modifications to eddy diffusivity, planetary boundary layer height, NOX emissions, and VOC emissions. Diurnal dependencies were observed for model biases related to both meteorology and  emissions. Specifically, our evaluation revealed model inconsistencies due to underestimations of modeled NOx and misrepresentation of PBL evolution. We note that these model errors are associated with inland southern California, and we conclude that UAV field experiments are valuable for improvements in region-specific model performance.

Source Apportionment and Integrated Process Analyses to Probe CMAQ Model Biases during Wintertime PCAP Events

Cesunica Ivey, University of California, Berkeley

Mountain cities in the western U.S. are oftentimes impacted by extreme pollution events brought on by stagnant high-pressure systems in the winter, coupled with temperature inversions and limited dilution. Periods of stagnation lasting longer than 36 hours are known as persistent cold air pool (PCAP) events and can be associated with ambient fine particulate matter concentrations well above the 24-hour PM2.5 NAAQS of 35 µg/m3. Here we investigate concentrations, source impacts, and model process contributions within CMAQ for PM2.5 in winter 2016 for three cities: Fresno, CA, Reno, NV, and Salt Lake City, UT. Several PCAPs were identified and were associated with dips in modeled relative humidity. PCAP periods were also characterized by underestimated secondary PM2.5 in all locations. Source-specific underestimations were also explored using the decoupled direct method paired with bias corrections, and sources spanned 20 unique, fuel- and activity-based categories. We also explored the contribution of biases to specific modeled processes using the integrated process analysis tool in CMAQ. Preliminary results suggest that aerosol processes, relative humidity, and the cloud microphysics option are the primary drivers for model bias during PCAP events.

NASA GEOS Composition Forecast System, GEOS-CF: Overview, Applications, and Future Directions

K. Emma Knowland, 1) Morgan State University/GESTAR-II, Baltimore, MD; 2) NASA GSFC, Global Modeling and Assimilation Office, Greenbelt, MD

NASA's Global Modeling and Assimilation Office (GMAO) produces high-resolution analysis and forecasts for weather, aerosols, and air quality. Since 2019, the NASA Global Earth Observing System (GEOS) model provides global near-real-time historical estimates and daily 5-day forecasts of atmospheric composition to the public at unprecedented horizontal resolution of 0.25 degrees (~25 km) from the surface up to 80 km. This composition forecast system (“GEOS-CF”) combines the operational GEOS weather forecasting model with the state-of-the-science GEOS-Chem chemistry module to deliver detailed analysis of a wide range of air pollutants, including the policy-relevant species such as ozone,  nitrogen oxides, and fine particulate matter (PM2.5).

The GEOS-CF is a tool for scientists and the public health community.  This presentation will cover 1) an overview of the GEOS-CF modeling framework and data/visualization access, 2) examples of current and future applications to support NASA missions (e.g., a priori for trace gas retrievals by TEMPO) and stakeholders including US EPA, UNEP and cities in the Global South, and 3) research and development activities as the GEOS-CF system continues to evolve to include multi-constituent data assimilation, down-scaling methods to urban-scale, and data access on Google Earth Engine, and other platforms to integrate our state-of-the-science air quality information onto platforms used by stakeholders, air quality managers, and the public.

Broadening Systematic Reanalysis Intercomparisons in SPARC-Reanalysis Intercomparison Project Phase 2 (S-RIP2): Chemical Reanalyses & Air Quality, Tropospheric Circulation, Extreme Events, and More

K. Emma Knowland, 1) Morgan State University/GESTAR-II, Baltimore, MD; 2) NASA GSFC, Global Modeling and Assimilation Office, Greenbelt, MD

Presenting on behalf of Gloria L. Manney

Katabatic Flow Turbulence Modeling

Yicheng Li, UC Davis, Civil and Environmental Engineering

Katabatic downslope winds are generated by negative buoyancy through the cooling of air near inclined surfaces, especially over mountains and glaciers. This study focuses on RANS (Reynold averaged Navier-Stokes) modeling of these flows. Katabatic winds transport pollutants and CO2 downhill and enhance turbulent mixing. Understanding and predicting their impacts is difficult due to complex nonlinear dynamics, and challenges in measuring over sloping terrain. We use a modified 1-D Prandtl (1942) model to simulate katabatic flow. Our study indicates that a first-order turbulence closure model in the simulations diverges from the observations, while second-order closure models better simulate the katabatic flow structure. However, we found results from second-order closures to be particularly sensitive to the wall model formulation, which provide turbulence fluxes at the grid cell adjacent to the surface in numerical weather predictions and have been shown through observations to require modification for katabatic flows. Therefore, we implement a new data-driven wall model based on Hang (2021), which is based on a 1-D RANS model for an inclined surface based on the buoyancy-modified k-epsilon equations. The capabilities of the model have been assessed against observations. We finally highlight additional modeling challenges and propose strategies for future model improvement.

The GFDL Variable-Resolution Global Chemistry-Climate Model for Research at the Nexus of US Climate and Air Quality Extremes

Meiyun Lin, NOAA Geophysical Fluid Dynamics Laboratory

We present a 33-year AMIP simulation (1988-2020) of a new GFDL variable-resolution global chemistry-climate model (AM4VR) designed for research at the nexus of US climate and air quality extremes.  With a horizontal resolution of 13 km over the continental US, AM4VR shows vast improvements upon  AM4.1 at uniform 100 km resolution in representing: US precipitation seasonal cycles and daily distributions, notably reducing the central US dry and warm bias; pan-US drought, linked to model fidelity of the ENSO teleconnection; and western US snowpack, with implications for wildfire risk. AM4VR also provides improvements over AM4.1 in the process-level representations of BVOC emissions, interactive dust emissions, and interactive tracer dry deposition coupled to hydroclimate and vegetation state. We highlight the skill of AM4VR in simulating summer ozone and winter haze dominated by ammonium aerosols in California’s Central Valley, with implications for food production and public health, and the impact of drought on southwest US dust and eastern US ozone extremes via vegetation feedbacks, with implications for seasonal air quality forecasts. AM4VR captures the observed sensitivities of ozone air pollution extremes to temperature, largest in the Northeast Urban Corridor, coastal counties around Lake Michigan, and other populated areas. AM4VR offers a novel opportunity to study global dimensions to US air quality, especially the role of Earth system feedbacks in a changing climate.

Spatiotemporal Gap-Filling of NASA Satellite-Derived-AOD in North America Using The UNet 3+ Machine Learning Architecture

Marcela Loria Salazar, School of Meteorology, University of Oklahoma

Wildfire emissions forecasting is needed to provide real-time alerts and forecasts for wildfire smoke. Due to the unpredictable and chaotic nature of fire ignition and behavior, it is challenging to forecast real-time wildfire emissions and their associated impacts on air quality. Multiple smoke models have been developed using satellite-derived aerosol optical depth (AOD) because of the improved spatial coverage. However, satellite-retrieved AOD datasets suffer from large portions of missing data, usually due to cloud cover. Consequently, including AOD as a predictor for particulate matter can be challenging for models that cannot inherently deal with missing values. We created a model based on the UNet 3+ architecture to fill in these missing values based on available AOD, meteorological, and land-used datasets. Training, validation, and testing datasets were created using NASA MODIS and VIIRS Deep Blue (DB) AOD retrievals, MERRA-2 AOD, NAM reanalysis model for meteorological land-use variables, NOAA HMS, and NASA FRP for fire emissions. Additional evaluations were performed using ground-based NASA AERONET AOD observations. This study will primarily focus on the fire seasons from (2012-2022) in the continental U.S. Preliminary results show promising results; the model achieved a correlation R~0.72, root mean square error RMSE~0.14, and normalized mean bias NMB~0.13 between collocated estimated AOD values and VIIRS AOD retrievals over non-winter months (April-November).

Investigating Impacts of Local Circulation on Coastal Ozone Problem in the New York Metropolitan Area: A modeling and observational study

Sarah Lu, JCSDA & UAlbany

Elevated surface O3 levels are often detected during hot summer days in the New York metropolitan area, due to the rich sources of local anthropogenic and natural ozone precursor emissions. Moreover, surface O3 in this region exhibits extensive horizontal and vertical gradients and distinctive diurnal cycles. This study examines the spatial and temporal O3 characteristics under different cluster-based local circulation scenarios during summertime in the pre-COVID era of 2017-2019, utilizing observations from various surface networks, New York State Mesonet (NYSM) Profilers, and 2018 Long Island Sound Tropospheric Ozone Study (LISTOS), as well as composites from the reanalysis fields and National Emission Inventory. The most polluted days are closely associated with classic sea breeze days with weak large-scale flow. When sea breeze development in the New York Bight is delayed, and its penetration into the NYC and shores of western Long Island Sound is intercepted by the dominant westerlies, daily 8-hr maximum average ozone (DMA8) in the hot spots of NYC and south shore of CT would drop more than 10 ppb under comparable temperature levels. The probability of DMA8 exceeding 70 ppb is also dramatically decreased. Furthermore, meteorological characteristics, such as sea breeze onset time and strength, most critical to ozone exceedances and high peaks are identified and analyzed for future improvement in coastal ozone simulation and prediction.

Data fusion with uncertainty quantification for sub-city-scale air quality assessment and forecasting

Carl Malings, Morgan State University and Global Modeling & Assimilation Office

Many information sources can support air quality assessment and forecasting, including atmospheric chemistry model outputs, satellite retrievals of column chemical and aerosol constituents, and surface-based air quality monitoring data from both regulatory and low-cost instruments. Systematic integration of these data sources provides a major opportunity to improve understanding and management of air quality, but also presents technical barriers, especially in resource- and data-constrained settings in the Global South. This presentation describes a data fusion system, currently under development using the Google Earth Engine platform, which aims to integrate the information sources listed above to support comprehensive sub-city-scale assessment and management of air quality. Furthermore, the data fusion framework includes provisions for the quantification of uncertainties in the resulting fused estimates based on the variability of and among the input data sources. These capabilities will allow air quality managers to better understand their local air quality situation, including relative confidence in the fused estimates for different constituents, locations, and times, leading to better informed air quality management decisions. This presentation covers the underlying methodology of the data fusion and uncertainty quantification approaches, provides an update on the status of its implementation, and presents early qualitative and quantitative results.

Investigating surface ozone sensitivity to HCHO/NO2 ratios over Arizona using the Multi-Scale Infrastructure for Chemistry and Aerosols (MUSICA) model

Seyed Mohammad Amin Mirrezaei, Department of Hydrology and Atmospheric Sciences, University of Arizona

Ozone pollution in semi-arid environments such as Arizona, where the primary precursors including NOx, anthropogenic, and biogenic VOCs varies with time and space, requires determining the relative contributions of these components to ozone production. Due to the lack of spatial distribution of surface measurements and the temporal limitation of satellite retrievals, multiscale air quality modeling can be a great tool for covering the deficiencies of satellite and surface measurements as well as addressing the discrepancies between their measurements. Here, the Multi-Scale Infrastructure for Chemistry and Aerosols Version 0 (MUSICA-v0), a global chemistry-climate model with regional refinement capability that provides down to a few kilometers’ resolution, is evaluated, and used to gain a better understanding of how horizontal resolution and chemical complexity can better represent the diurnal changes of NOx and VOC limited regimes. The MUSICA-v0 simulations with ∼14 km horizontal resolution were evaluated with tropospheric Monitoring Instrument (TROPOMI) HCHO/NO2 column ratios and surface HCHO/NO2 ratios as well as ozone concentrations compared with a Weather Research and Forecasting (WRF) model coupled with Chemistry (WRF-Chem) simulation for July and August 2019-2020. The preliminary results indicate that the performance of MUSICA is comparable to that of WRF-Chem, and it can be used for regional studies while considering all earth system components on a global scale.

WRF-Chem/DART: A Regional Ensemble Atmospheric Composition Forecast/Assimilation/Emissions Estimation System - Recent Developments and Applications

Arthur Mizzi, NASA Ames Research Center, NOAA Chemical Systems Laboratory, University of Colorado at Boulder Mechanical Engineering

WRF-Chem/DART is a state-of-the-science regional atmospheric composition (AC) forecast/assimilation/emissions estimation system based on the ensemble adjustment Kalman filter that uses the state augmentation method for constraining emissions. It has/is being used by researchers throughout North America, Asia, and Western Europe. WRF-Chem/DART’s continued development and application has been supported by NASA, NOAA, NCAR, and the University of Colorado at Boulder. This poster will provide an introduction to WRF-Chem/DART and highlight its recent development/applications which include:

• Assimilation of (i) OMI ozone (O3), nitrogen dioxide (NO2), and sulfur dioxide (SO2); (ii) TROPOMI carbon monoxide (CO), O3, NO2, and SO2; and proxy (iii) TEMPO O3 and NO2 total/partial column and/or profile retrievals (completed).
• Assimilation of (i) OMI formaldehyde (HCHO), (ii) TROPOMI HCHO and methane (CH4); (iii) TES CO, O3, ammonia (NH3), and CH4; (iv) CrIS CO, O3, NH3, CH4, and peroxyacetyl nitrate PAN; (v) GOME2-a NO2; SCIAMACHY NO2; and MLS O3 and nitric acid (HNO3) total/partial column and/or profile retrievals (in progress).
• Spatial and temporal interpolation of global model analyses/forecasts for use as upper boundary conditions for assimilation of retrievals whose vertical weighting functions have an upper boundary that is above the regional model’s upper boundary.
• Applications for COVID OSSE, FIREX-AQ Williams Flats OSSE, TRACER-I, and AC forecasting for central Mexico.

HiRes-X: Forecasting Air Quality and Health Impacts of Prescribed Fires in Southeastern U.S.

M. Talat Odman, Georgia Institute of Technology

Prescribed fire is the leading source of PM2.5 emissions in southeastern United States (US). The use of prescribed burning has economical and ecosystem-related benefits but must be weighed and managed with respect to air quality concerns and potential health and welfare impacts . We have developed the HiRes-X modeling system to provide forecasts of potential prescribed fires and their impact on air quality and human health in southeastern US. Since 2019, we have been disseminating these forecasts daily through the Southern Integrated Prescribed Fire Information System (SIPFIS). The current 48-hour lead time forecasts of PM2.5 and O3 levels and prescribed fire contributions assist the Georgia Department of Natural Resources for official forecasts of air quality for public health protection, and the Georgia Forestry Commission for permitting of prescribed burning. We will describe the components of the HiRes-X modeling system in detail and report on its past performance. We will show examples of how the HiRes-X forecasts and SIPFIS can be used for cohesive prescribed fire, air quality and health management. We will also report our efforts to further improve the HiRes-X’s capability to better predict smoke plumes. We will evaluate model predictions by comparing them to PM2.5 and O3 measurements from prescribed burns conducted at military bases in southeastern US as part of an ongoing Strategic Environmental Research and Development Program (SERDP) fire science initiative.

Observational Assessment of Aerosol Impacts on Updraft Speed in Deep Convection

Hallie Pimperl, UC Davis

Aerosols influence many cloud aspects, such as cloud development and convective dynamics, through their role as cloud condensation nuclei. Previously, modeling studies have shown evidence of higher aerosol concentrations leading to aerosol-induced invigoration of deep convective storms. (In this study, we define invigoration as an increase in cloud updraft speed.) This study uses GOES16 geostationary satellite data to track >100 isolated, deep convection over the Southern Great Plains region in Oklahoma. Updraft speeds are not

measured directly, but they are assumed to be proportional to the cloud top rise rate. The cloud top rise rates have been found to show a positive correlation with some environmental factors, such as the most unstable CAPE, as expected. After controlling for the meteorology, the rise rates are cross-referenced with the daily aerosol loadings from the ARM SGP site to determine whether or not an aerosol effect on the cloud top rise rates can be detected from the environmental controls. While the results are still preliminary, it appears unlikely that we can find a robust relationship between aerosol concentration and updraft speed.

Impacts of climate change on wildfire PM2.5 and the human health burdens in the US

Minghao Qiu, Stanford University

Wildfire and associated smoke exposure have increased dramatically during the last two decades in the US, reversing the multi-decadal improvements in fine particulate matters (PM2.5). Wildfire activity is projected to accelerate under future climate, but our understanding of the magnitude and distribution of the potential resulting impacts on human health remains highly incomplete. Here, we utilize satellite-derived daily smoke PM2.5 at 10km resolution from 2006-2020 over the continental US to project the smoke PM2.5 and associated health burdens under future climate. To do this, we first build a machine learning model to predict monthly wildfire emissions with climatic variables over the entire North America. Then for each 10km location, we establish a causal relationship between the monthly smoke PM2.5 in that location and fire emissions from potential source regions, accounting for variation in the wind directions. Combining with future climate projections, we estimate the monthly smoke PM2.5 under different future climate scenarios for each 10km location over the continental US. Mapping the projected changes to the census tract level and combining these changes with existing smoke-health concentration-response functions, we quantify the health burden due to changes in smoke exposure under future climate, including mortality and hospitalizations. We further quantify the distributions of these health burdens across different social and demographic groups.

Trace gas atmospheric rivers: remote drivers of air pollutants

Mukesh Rai, Jet Propulsion Laboratory, California Institute of Technology

Identifying key atmospheric transport events is essential to providing improved insights into the relative contributions of remote and local sources to local air quality across the globe, which in turn informs environmental policymaking that benefits air quality and climate. The concept of atmospheric rivers (ARs) has been successfully used to demonstrate key transport events for determining global patterns of water vapour and aerosols.  Extending the AR concept, we propose a new AR framework, Trace Gas Atmospheric River (TGAR), to study key long-range transport events by using global trace gas data obtained from JPL chemical data assimilation system that ingests various satellite observations. To that end, we determine the global distribution, seasonal variations, and long-term trends of the TGAR patterns and the relative contribution of remote air pollutants over major megacities in the world. The results obtained for various air pollutants and their precursors with different source patterns and chemical lifetimes, such as ozone, carbon monoxide, and PAN, are evaluated and validated against independent in-situ surface observations and satellite retrievals from NASA’s TES and AIRS/OMI, and NOAA’s CrIS developed by JPL’s TROPESS project. This study will provide insights into the extreme events of pollution transport in terms of TGARs and is expected to provide a useful framework for better characterizing air quality drivers and informing chemical transport model improvements.

An evaluation of Model II Regression techniques for the intercomparison of two instrumental methods for a national air quality monitoring network

Colleen Marciel Rosales, OpenAQ / UC Davis

Most comparisons of two instrumental techniques use ordinary least squares (OLS) regression—simple, but often incorrect. Assumptions of Model I regression techniques, such as OLS, do not completely apply to techniques that both have uncertainties associated with them, since for OLS to be valid, the independent variable “x” should cause or determine the dependent variable “y”. On the other hand, Model II regression techniques, such as major axis (MA) and standard major axis (SMA) regressions, do not require this assumption, such that  “x” and “y” are interchangeable, i.e., one variable does not determine the other. Model II techniques also give the chance to account for errors in both variables, making it a suitable comparison metric for instrumental methods. In this presentation, different Model II regression techniques as well as other statistical comparison metrics are explored and compared side-by-side to OLS as applied to atmospheric instrumental techniques for the elemental characterization of particulate matter collected from an ambient monitoring network.

Development of PM2.5 transport: Modeling the spatial distribution of Camp Fire from California to New York

Xiaorong Shan, George Mason University

The November 2018 Camp Fire in northern California released abundant aerosols into the atmosphere and is one of the costliest disasters in the world for insurers with insured losses totaling $12.5 billion. The event degraded air quality across the United States, with elevated PM2.5 concentrations observed across the country during this period. It is unclear, however, the extent that elevated air pollution concentrations attributable to distant wildfire emissions impacted human health. Modeled wildfire source impacts are highly dependent on the models used and their inputs, including wildfire emissions, which can be highly uncertain. Here we focus on PM2.5 exposures during the Camp Fire period (from Nov. 08.2018 to Dec. 02.2018) in New York state. To estimate exposure, we use the average values of four satellite-based fire emission data sets (FEER, FLAMBE, GBBEPx, and GFAS). Then, we employ HyADS, the HYSPLIT average dispersion model, which combines the HYSPLIT trajectory dispersion with emissions to simulate the PM2.5 daily exposure. We evaluate our results using PM2.5 observations at EPA monitors, which capture total daily variability but are limited in their ability to differentiate PM2.5 from fires. The highest correlation (R) between the HyADS and local monitors is 0.54 using the most detailed emissions data. In continuing work, we plan to employ source apportionment at EPA CSN cites to identify the portion of the observed PM attributable to fires.

Simulating 2020 Creek fire with cloud-resolving E3SM and high-resolution satellite observations

Qi Tang, Lawrence Livermore National Laboratory

The 2020 Creek fire generated pyrocumulonimbus (pyroCb) and injected a large amount of fire smoke into the stratosphere, which contributed to considerable changes in atmospheric radiation and air quality in the western US. The stratospheric aerosols encircled the Northern Hemisphere for months imposing radiative forcing to the Earth system and leading to potential climate impact.  In this study, we will leverage the cloud-resolving Energy Exascale Earth System Model (E3SM) and its regionally refined modeling (RRM) capability to enhance the E3SM grid resolution to ~3 km at the California source region. The rest of the globe remains at the standard 100 km resolution for computational efficiency. This RRM configuration explicitly resolves parts of the critical wildfire processes near source regions and quantifies large-scale climate impact of the wildfire event. The Creek fire emissions will be constrained by high-resolution satellite observations based on VIIRS measurements. We will emphasize on the improved representation of interactive chemistry and aerosols (e.g., brown carbon) and their influence on the stratospheric smoke representations. We will also test the simulation sensitivities due to various nudging strategies and a 1-D plume rise parameterization in the 3-km domain.

Spatial Variability in Formaldehyde and Nitrogen Dioxide Diurnal Cycles in the New York City Area

Madankui Tao, Columbia University in the City of New York/ Lamont-Doherty Earth Observatory

Formaldehyde (HCHO) and nitrogen dioxide (NO2) can be remotely sensed by satellite instruments and are relevant to local ozone formation. We examined spatial variations in the diurnal patterns of HCHO and NO2 concentrations using surface monitors, ground-based Pandora spectrometers, and the WRF-CMAQ model (1.33 km) at seven monitoring locations in the NYC metropolitan area and Connecticut shoreline regions during June-August. We observe evidence of the ozone weekend effect (higher ozone on weekends when NOx is lower) at urban sites, while ozone at Westport (CT) and Flax Pond (NY) was higher on weekdays. HCHO concentrations peak at noon at NY Botanical Garden and Flax Pond but around 2 p.m. at Westport. Surface NO2 concentrations exhibit bimodal peaks at all sites except New Haven (CT), where a third peak was noticeable at noontime due to significant local anthropogenic NOx emissions at this near-roadway site. We find discrepancies in the timing of daily maximum near-surface concentrations and column densities of HCHO and NO2, with simulated daily column densities of HCHO and morning NO2 (excluding the evening peak) being delayed by approximately 4 hours compared to peak surface concentrations. Our ongoing analysis aims to determine how emissions and meteorology contribute to the observed spatial variations in photochemical environments that lead to ozone production. This study has implications for the urban-to-rural applications of the TEMPO instrument.

Forecasting daily fire radiative energy using scaled persistence and machine learning for air quality applications

Laura Thapa, University of California, Los Angeles, Atmospheric and Oceanic Sciences, Los Angeles, CA, USA

This work uses random forests and develops new fire indices to forecast daily and sub-daily fire radiative power (FRP) for previously ignited fires in the Western US. FRP forecasts can be used to predict fire emissions, diurnal cycles, and plume heights, all of which are important but still relatively uncertain inputs to the state-of-the art air quality models that help society cope with the negative impacts of smoke. This work seeks to improve upon the persistence assumption used in most air quality models to predict FRP.

The variables which govern wildfire intensity, such as weather, plant and land characteristics, and fire suppression efforts, are monitored via satellite, aircraft, and ground based sensors and are incorporated into reanalysis datasets. We develop a training dataset that consists of fire weather indices, containment percentages, fuel loading and soil moisture estimates, ecosystem-water interaction metrics, and atmospheric stability indices. We also develop an algorithm that aggregates VIIRS fire detections into polygon objects and use these polygons to select portions of the gridded datasets most relevant for each day of the fire. Additionally, we explore relationships between variables in the input space and identify an optimal subset of the input variables for forecasting FRP. We show that random forests methods and scaling previous FRP estimates produce 1-day forecasts of FRP that generally improve upon 1-day persistence forecasts.

Configuration and evaluation of the WRF-Chem air quality simulations over Thailand

Worapop Thongsame, University of Colorado Boulder

Air pollution, especially PM2.5, is a major issue in Thailand. To address this, we aim to configure and evaluate the WRF-Chem to simulate PM2.5 concentrations in Thailand and understand the contributing factors. We assess the model's performance using ground-based observations and satellite data (MODIS AOD and MOPITT CO). The study utilizes the MOZART-MOSAIC scheme with 9 km resolution. Various anthropogenic emission inventories (CAMS, ECLIPSE, REAS, and EDGAR) were evaluated during the off-haze season, revealing that CAMS and ECLIPSE provided comparable PM2.5 concentrations, while REAS and EDGAR overestimated levels.

During the haze season, three biomass-burning inventories (QFED, FINN1.5, and FINN2.5) were assessed alongside the CAMS inventory for anthropogenic emission. FINN1.5 demonstrated the best performance, showing a correlation coefficient of 0.75 when compared to air quality station data and 0.63 for MODIS AOD. FINN2.5 overestimated PM2.5 with a 30% mean fractional error compared with air quality stations and overestimated MODIS AOD in the Northern part of the domain. QFED, on the other hand, underestimated AOD. Overall, FINN1.5 exhibited the best performance, with a correlation coefficient of 0.63 between modeled and MODIS AOD. However, when compared with MOPITT CO, FINN2.5 outperformed FINN1.5. Based on our study, the combination of CAMS and FINN1.5 is the most suitable setup for the WRF-Chem model in Thailand.

Multi-Model Ensemble Forecasts of Hazardous Air Quality Events: Comparisons of Weighted and Unweighted Approaches

Yunyao Li, George Mason University
Will be presented by Daniel Tong on behalf of Yunyao Li

Wildfires are a major source of atmospheric aerosols and trace gases, which can cause hazardous air quality conditions and associated health problems. Accurately predicting wildfire smoke dispersion is challenging due to uncertainties in fire emissions, plume rise, etc. Ensemble forecasting techniques have been increasingly used to improve the predictability. We developed novel ensemble creation methods to improve the ensemble forecast compared to the traditional ensemble mean. 5 operational models, NASA GEOS, NRL NAAPS, NOAA GEFS, HRRR, and NACC-CMAQ model, were used to create the ensemble. Results show that the ensemble mean outperforms each individual model. To further improve the ensemble forecast, we employ statistical methods such as ridge regression, weighted regression, and quantile regression to develop a weighted ensemble. The weighted ensemble reduces the model system error compared to the ensemble mean. Furthermore, the weighted ensemble has a higher hit ratio, weighted success index, and lower false alarm ratio compared to that of individual models and the ensemble mean. This demonstrates that the weighted ensemble is an effective method for reducing forecast uncertainty and improving the accuracy of air quality forecasting during wildfire events. Our findings provide insights into the development of advanced ensemble forecast methods for extreme air quality events, which can aid in decision-making and mitigation efforts for protecting public health.

Extending AIRPACT Simulations to a Third Day

Mohammadamin Vahidi Ghazvini, Washington State University

AIRPACT (Air Indicator Report for Public Awareness and Community Tracking) is a complex air quality forecasting system that is used to predict air pollutants concentrations and depositions, developed at Washington State University over many years. The simulation domain of this model is the northwestern United States, includes the entirety of the states of Washington, Oregon, and Idaho, as well as the northern parts of California, Nevada, and Utah, the western parts of Wyoming and Montana, and the southern parts of the Canadian provinces of British Columbia and Alberta. AIRPACT contains 73530 cells (including 285 horizontal and 258 vertical cells) the size of each grid cell is 4 km by 4 km. Also, AIRPACT has 37 vertical layers.

Currently, AIRPACT forecasts two days (48 hours), and the goal of this research is to increase the forecasting time to 3 days (72 hours). AIRPACT includes the 9 main scripts for simulating the second day, and in this research, we used all of them for simulating the third day. AIRPACT uses 11 main scripts for simulating the first day. Each stage has a main script that contains other sub-scripts.

Advancing Wildfire Research using Large Eddy Simulation (LES) and Machine Learning

Siyuan Wang, CIRES/NOAA/CSL

Wildfires affect weather and climate, with costly impacts on human health and properties. Despite decades of research, wildfires are still challenging to represent in air quality models. One fundamental challenge is that the spatial resolution at which the models are operated is much coarser than the spatial extent of most wildfires. As a result, several key processes in wildfires cannot be explicitly resolved. One of such poorly resolved processes is plume rise. It has been well documented that state-of-the-art plume rise models are subject to large uncertainties, with major impacts on air quality downwind.

In this work, I present the ongoing development of a machine learning emulator for wildfire plume rise. This new framework, trained using a high resolution, turbulence-resolving Large Eddy Simulation (LES) model, emulates the plume rise process considering the impacts of the fire-induced buoyancy, entrainment, and moisture processes. Rigorous measures are taken to mitigate overtraining, and the outcomes are physically sound and interpretable. Preliminary results show that this LES-trained machine learning emulator outperforms a widely used physics-based plume rise model in terms of mean smoke injection height and smoke profile shape. The machine learning emulator is also considerably faster. In summary, this machine learning emulator for plume rise provides an accurate and computationally efficient solution for air quality models and chemistry-climate models.

Slope Angle Impacts on the Turbulence Structure of Daytime Anabatic Winds and Modeling Implications

Ting (Diane) Wang, UC Davis

Buoyantly-driven anabatic winds are induced by radiative surface heating over slopes. They transport air pollutants and moisture, affecting mountain weather via the formation of clouds and orographic precipitation. This study utilizes turbulence observations at multiple levels from two datasets (over a steep Alpine slope in Val Ferret, Switzerland and over a gentle slope of at Granite Mountain in northern Utah) to characterize the mean and turbulence structure of daytime anabatic winds. Observed surface-normal profiles of velocities and temperatures show steep near-surface gradients from both datasets. A jet-shaped profile for streamwise velocity is identified at the steep slope site but not at the gentle slope site likely due to the depth of the anabatic layer. The estimated jet-peak location is ~2.7 m for the steep slope and ~20 m for the gentle slope. We also observe significant momentum and heat flux divergence for anabatic winds at both sites. This is significant because it violates the underlying constant-flux layer assumption for Monin-Obukhov similarity theory, which provides a basis for parameterizing near-surface turbulent flux transport in many numerical weather prediction models. Quantifying the daytime near-surface flux transport and turbulent motion structures with observations over slopes aids in understanding the fundamental physics and improve the existing surface-layer parameterizations in turbulence modeling.

Connecting Aerosol Modeling and Numerical Weather Prediction from Data Assimilation

Shih-Wei Wei, Joint Center for Satellite Data Assimilation and University at Albany

Aerosols affect the radiation budget of the Earth system through the attenuation and the interactions with clouds. This aerosol-cloud-radiation interaction has been studied in climate projections and weather prediction for decades. Its impacts on the context of satellite data assimilation (DA), however, are often neglected despite the considerable reduction in the upwelling radiance at the top of the atmosphere in the thermal infrared (IR) window. As developing the all-sky IR DA, it is worthwhile to understand the advantages and challenges of an aerosol-aware DA framework. This framework also adds another layer of interaction between composition modeling and weather prediction. It further provides insights for coupling Earth DA system.

This study aims to exploit aerosol-affected IR observations in DA. To do this, we identify aerosol-affected observations based on brightness temperature (BT) differences of multiple IR channels. Modeling aerosol mass mixing ratios are then incorporated into the radiance observation operator to calculate the aerosol-affected BTs and Jacobians. The observation errors are dynamically determined by adapting the algorithm from all-sky microwave DA. Using this framework, we conducted an aerosol-aware experiment and compared it with a baseline experiment. Both experiments are based on the National Centers of Environmental Prediction (NCEP) Global Data Assimilation System (GDAS). The impacts on analyses and forecasts will be investigated.

Analyzing Trends in Air Quality During a Drought: A Case Study to Improve Public Health Response to Drought Threats (Withdrew Abstract)

Taylor West, NASA DEVELOP

Air quality has been shown to be significantly correlated with drought severity in the United States, and secondary impacts of droughts such as wildfires and dust storms are known to increase concentrations of airborne particulate matter with adverse effects on human health. This project represents the initial phase of measuring trends in air quality indicators, including Aqua and Terra Moderate Resolution Imaging Spectroradiometer (MODIS) aerosol optical depth (AOD), Suomi NPP Visible Infrared Imaging Radiometer Suite (VIIRS) aerosol optical thickness (AOT), and Environmental Protection Agency’s Air Quality System (AQS) PM2.5 during recent drought conditions in the Pacific Northwest. Using data from the U.S. Drought Monitor (USDM) and Standardized Precipitation Evapotranspiration Index (SPEI), we evaluated these trends across the evolution of drought conditions and constructed maps depicting areas in Oregon and Washington that were vulnerable to changes in air quality during the drought event. By partnering with the University of Nebraska Medical Center’s Water, Climate, and Health Program, NOAA National Integrated Drought Information System, Oregon Health Authority, and Washington State Department of Health, these trend analyses can inform public health departments’ efforts to prepare for and mitigate the effects of drought on human health.
Withdrew Abstract

High Spatiotemporal Resolution Modeling of PM2.5 in West Africa Using Satellite Data and Machine Learning

Benjamin Yang, Columbia University

Exposure to ambient fine particulate matter (PM2.5) is a leading environmental risk factor for premature death. In Africa, surface PM2.5 data is sparse, hindering pollution mitigation plans and human health improvement. NASA satellite observations provide near-complete spatial coverage, but their columnar nature is imperfect representations of surface pollution. To estimate PM2.5 concentrations at high spatiotemporal resolution (1 km^2, daily) across West Africa over the past two decades, we trained, tested, and fine-tuned a machine learning (XGBoost) model with the following data: PM2.5 from about 100 reference-grade and calibrated low-cost monitors, aerosol optical depth from MODIS MAIAC satellite retrievals, five meteorological features from ERA5, and seven tropospheric trace gas column or aerosol property features from TROPOMI/OMI satellite retrievals. Preliminary results show that the model performs reasonably well (r^2 = 0.73, mean absolute error = 11 µg m^-3) in predicting daily PM2.5 compared to observations in six cities across West Africa. Likewise, we will develop a machine learning model focused on Ghana for more localized epidemiological and environmental justice studies. These novel PM2.5 datasets will enable us to identify and explain long-term trends, cycles (seasonal, weekly, and diurnal), and significant sources of PM2.5 in both urban and rural areas. As air quality monitoring networks expand across Africa, the spatial predictions are expected to improve.

CANCELLED: Impact of the Urban Heat Island Effect on Simulating Air Quality in Seoul, South Korea

Katherine (Katie) Travis, NASA Langley Research Center

In Seoul, it has been shown that the urban heat island effect deepens the nighttime boundary layer, reducing ozone titration at night, and strengthens the urban breeze, bringing in more ozone from the surrounding areas. Here, we apply the urban heat island effect to simulating the 2016 NIER/NASA Korea-United States Air Quality (KORUS-AQ) field campaign using the newly developed WRF-GC model that incorporates the GEOS-Chem chemical mechanism and Harmonized Emissions Component (HEMCO) with the Weather Research and Forecasting (WRF) model and its urban physics schemes. We assess how the inclusion of the urban heat island effect impacts simulations of the diurnal cycle of surface and column pollutants and whether including urban physics improves model comparisons against the detailed aircraft chemical observations. We also analyze whether the urban heat island effect can improve long-standing model overestimates of nitrate due to incorrect nighttime chemistry driven by insufficient nighttime mixing.

CANCELLED: Forecasting hourly wildfire emissions using a modulated persistence approach based on an  hourly wildfire potential index

Johana Romero-Alvarez, Cooperative Institute for Research in Environmental Sciences, University of Colorado at Boulder/NOAA Global Systems Laboratory

As wildfires increase in frequency and severity, smoke forecasting becomes fundamental to help minimize exposure, improve weather forecasting, and guide wildfire-fighting operations. Most smoke and air quality forecast models assume a persistence approach in which the latest satellite-based biomass burning emissions estimates are modulated by a parameterized climatological diurnal cycle during the forecast duration. The only variable to account for weather impact on smoke emissions is the effect of precipitation, which is simplistically included in some of the models. NOAA's Global Systems Laboratory is developing a new experimental weather forecast model, the Rapid-Refresh Forecasting System (RRFS), that covers the CONUS domain at 3 km resolution. The model also simulates 3D transport and mixing of smoke emissions. Here, an hourly wildfire potential (HWP) index computed using predicted 10-m wind gust, surface dewpoint depression, and soil moisture availability is used to modulate persistence-based forecasted wildfire emissions estimates within the RRFS smoke and dust (RRFS-SD) model. The method is compared against the persistence approach for two fire seasons, 2019, which coincides with the FIREX-AQ campaign and represents a low-intensity fire season, and 2020, characterized by severe fires in the western US. The model performance is evaluated using ground-based PM2.5 and aerosol optical depth observations and in-situ measurements acquired by an aircraft during FIREX-AQ.

CANCELLED: Evaluation of the Air Quality in the Vicinity of Quarrying at Ebonyi State, Nigeria

SAMUEL AKPAN, Federal College of Fisheries and Marine Technology, Victoria Island, Lagos, Nigeria

In Ebonyi State, Nigeria, numerous quarries contribute considerably to the growth of the economy. However, these sectors have had a significant effect on the environment, and one of those impacts is air pollution. Extech Model VPC300 and Aeroqual 500 series were used to measure the state of the air to determine the consequences of quarry dust on the environment. The quarry regions and the area around them were found to be much more polluted than indicated by international standards for PM2.5, PM10, and NO2. There is a strong correlation matrix between the detected quarry contaminants, the weather, and distant measurements. To reduce the amount of dust released into the atmosphere, it is advised to use modern, environmentally friendly crushing equipment with built-in dust collectors. It is also advised to create a green belt using trees that can withstand pollution around the quarrying locations to lessen the impact of air pollution.

CANCELLED: Quantification of crop residue burning using WRF-Chem model over North Indian region

Ummed Singh Saharan, National Physical Laboratory, New Delhi, India

Intensive Crop Residue Burning (CRB) in north India causes severe air pollution episodes over the Indo-Gangetic Plain (IGP), India, every year during October-November. Present study has quantified the role of the CRB using the Weather Research and Forecasting model coupled with Chemistry (WRF-Chem) using recent anthropogenic (EDGARv5) and biomass burning emissions (FINNv2.4) inventories and the MOZART-MOSAIC chemical scheme. First, we have investigated long term CRB data (2012 to 2019) to find out the dominant CRB areas, period of burning, and hotspot districts in Punjab and Haryana. Then, we have utilized WRF-Chem with different sensitivity runs to quantify the role of the CRB. Hotspot districts contributed ~80% and ~50% of total fire counts in Haryana and Punjab, respectively. PM₂.₅ was well captured with NMB  < 0.2 and R > 0.6 over the majority of sites across the domain. The hotspot districts contributed about 70% of total CRB-induced PM₂.₅ enhancement at the western, central, and eastern sites, and approximately 50% at the  northern sites. We could save ~18k lives by eliminating CRB from the entire domain and policymakers can address this issue by focusing on hotspot districts first.

Learn more about the presenters here