SMART Laboratory

SMART Lab Missions:

Science & Mathematics of AI modeling for Research and Technology (SMART) Lab established in 2022. The SMART Lab has three core missions: Education, Research & Technology

Education

With over 20 years of experience in teaching and curriculum development at both national and international levels, we design and deliver courses aimed at educating the next generation of scientists, engineers, and educators. We offer students the opportunity to engage in solo or team research projects, fostering increased interest and achievement in STEM fields.

Research

We design, develop, and implement diverse research projects that heavily utilize advanced coding and computational skills, including AI techniques, machine learning models, big data analysis, physical models, and simulations, spanning a wide range of interdisciplinary research areas.

Technology

We develop innovative software and technologies using cutting-edge techniques in computer science and engineering. Our commitment to fostering collaboration among industry, STEM education, and academia drives continuous progress and innovation.

Lab Director

Current and Previous Research Projects

Sponsor: NASA

Project Summary: This study applied various Machine Learning (ML) algorithms to explore the data science behind fire-induced smoke emissions from California wildfires. We applied an ensemble of NASA’s remotely sensed observations and historical reanalysis data to develop an ML-based predictive to improve the predictability of smoke concentrations over the receptor areas.

Project Title: Development of Smoke Transport Probability and Risk Interactive Map from Trajectories and Climatology Analysis of 2-km CANSAC-Reanalysis Database

Sponsor: CAL FIRE, Forest Health Program

Project Summary: We generated high-resolution smoke transport probability for the 2-km CANSAC reanalysis domain. We used HYSPLIT trajectory modeling with 2001-2020 2-km climatology to characterize smoke sources and transport at each grid cell. We integrated HYSPLIT results with climatological fire weather metrics, utilizing high-resolution historical CANSAC data, to develop an interactive dashboard for wildfire smoke paths. This dashboard was designed to support fire-weather stakeholders and decision-makers.

Project Title: Quantifying variations in atmospheric temperature from light-absorbing aerosols

Sponsor: NASA

Project Summary: In this project we explored impacts of light-absorbing aerosols on atmospheric and surface temperatures broadly across the globe using measurements from multiple sensors on the Aqua satellite spanning the entire Aqua record from mid-2002 to the present. We (a) characterized the dependence of day/night air temperature contrast on aerosol optical thickness globally; (b) explored the modes of coupled variability between aerosols and air temperature at global and hemispheric scales; and (c) quantified the variability in surface temperature related to variations in aerosol amounts. We applied multi-sensor analysis to measurements to the full 20-year record from the Aqua satellite, including from the AIRS, MODIS and AMSR-E instruments.

Project Title: Estimating Background Ozone Using Data Fusion

Sponsor: EPRI

Project Summary: U.S. background ozone is the ground-level ozone present without domestic anthropogenic emissions, estimated through air quality models to determine its regulatory minimum. These estimates carry uncertainty due to the inherent limitations of air quality modeling. To address this, recent advances in data fusion techniques have been employed to improve these estimates by incorporating ground-level ozone observations and satellite data. We applied two adjustment methods to enhance model-derived estimates of background ozone: a multivariate regression model and a machine learning (ML) algorithm, focusing on thirteen urban areas in the U.S. The ML model showed significant improvement in performance when predicted ozone concentrations were compared to observations, and the adjustment with the ML model further increased the background ozone estimates. Read More: https://doi.org/10.1016/j.atmosenv.2023.120145

Project Title: Smoke Concentration Predictions

Project Summary: We conducted a comprehensive climatological analysis using high-resolution 2 km CANSAC historical data, available hourly from 1996 to the present, for California and Nevada. Leveraging the Bluesky framework, we compared smoke concentrations from the CONSUME model with NASA’s MERRA-2 reanalysis as an initial step in predicting smoke behavior. By integrating atmospheric and land cover features, we trained, tested, and validated Machine Learning (ML) models to project future smoke concentrations in receptor areas through 2098.

Project Title: Sierra Nevada Extreme Weather: A Novel Investigation of the Energetics of Atmospheric Rivers

Sponsor: NASA

Project Summary: Extreme precipitation in the Sierra Nevada during the cold season is often associated with Atmospheric Rivers (ARs) that bring significant moisture fluxes from tropical regions. We developed and implemented a diagnostic tool to characterize the transient energy of ARs. Additionally, we proposed new metrics, including thresholds for energetic parameters, to better assess these extreme events.

Project Title: Fire Consortia for the Advanced Modeling of Meteorology and Smoke

Sponsor: USFS

Project Summary: We developed simulations using the CONSUME model, integrated with Bluesky, to assess smoke transport. By incorporating land cover and vegetation data as inputs for the smoke model, we enhanced the simulation’s accuracy. We used Artificial Intelligence (AI) / Machine Learning (ML) algorithms to analyze the relationships between land cover and climate variables during fire seasons from a climatological perspective, providing deeper insights into their interactions and impacts.

Project Title: Extreme weather Dynamics and Predictability

Sponsor: Navy

Project Summary: We studied the dynamics of extreme weather events over the Mongolia region using WRF simulations and statistical analyses to estimate model uncertainties. The model was configured and integrated over a large set of historical cases, which were used to calibrate future forecasts for biases and improve predictions.

Project Title: Modeling and Source Attribution of ozone in southeastern New Mexico

Sponsor: NMED

Project Summary: The New Mexico Air Quality Control Act mandates that the New Mexico Environmental Department (NMED) develop and adopt a plan, including regulations, to control emissions of nitrogen oxides and volatile organic compounds. This plan is designed to ensure the attainment and maintenance of the ozone National Ambient Air Quality Standards (NAAQS) when ozone concentrations exceed 95% of the NAAQS. We assisted the NMED in quantifying source attribution for air quality by analyzing the current status and trends using observational data. We employed source apportionment analysis in combination with back trajectories and source apportionment modeling, specifically targeting upwind high-emitting sources, to quantify the contribution of local versus out-of-state sources to New Mexico’s air quality.

Project Title: Bipartisan Infrastructure Law Project- Extending Humidity and Vegetation Condition Datasets Back in Time to Support Climate-scale Wildfire Prediction

Sponsor: NOAA

Project Summary: This work supported preparedness, seasonal assessment, and real-time understanding of fire weather conditions. We used Artificial Intelligence (AI) / Machine Learning (ML) approach to determine techniques to extend humidity and vegetation datasets back in time to support climate-scale wildfire prediction. In addition, we developed AI/ML models for sub-seasonal predictions, forecasting vegetation characteristics such as the Normalized Difference Vegetation Index (NDVI) and Vegetation Health Index (VHI) across the CONUS with a three-week lead time.

Project Title: Disaster Relief Supplemental Appropriations for Fire Weather Dataset Activities

Sponsor: NOAA

Project Summary: This work supported preparedness, seasonal assessment, and real-time understanding of fire weather conditions. The project outcomes helped toward improving wildfire prediction, detection, forecasting, monitoring, and data management. We used Artificial Intelligence (AI) / Machine Learning (ML) approach to determine techniques to extend humidity and vegetation datasets back in time to support climate-scale wildfire prediction.

Project Title: Gridded Wind Climatology Project – Gridded hourly wind climatology products for High Plains Regional Climate Center

Sponsor: NOAA

Project Summary: We developed high-resolution 2.5 km hourly humidity and wind datasets dating back to 1985 for the contiguous United States (CONUS), with extended coverage for Hawaii, Alaska, and Puerto Rico. To support fire management agencies, the dataset also includes historical vegetation conditions, ensuring sufficient data quality based on user engagement and operational needs.

Project Title: Mechanistic Interactions of Dust Radiative Forcing with Energy of Atmospheric Waves

Project Summary: This research investigated the intricate interactions between Saharan dust aerosols and atmospheric dynamics. Utilizing an ensemble of NASA’s historical reanalysis data and satellite remote sensing observations, we analyzed regional climate variability linked to dust fluctuations. Through simulations with the Community Earth System Model (CESM), we assessed how Saharan dust aerosols influence transient eddies and drive regional instability. Our findings offer further insights into the mechanisms driving variations in regional instability caused by dust.