Science Serving Data Science

Current Projects

• 

  Enabling Operational Computed Cloud Tomography by Combining Physical and Data Science

  Atmospheric tomography is an emerging remote sensing technique that enables passive imagers to “slice” the atmosphere through multi-angle observations, recovering the 3D structure of clouds and other atmospheric components. Current approaches are computationally expensive and are not used beyond small scale case studies of individual clouds. To reduce the computational burden and demonstrate mission operational readiness, we combine physics and data science to speed up the tomographic reconstruction from hours to seconds by decomposing the problem into an inverse radiative transfer and a reconstruction part. This technology is a key building block for future mission concepts that seek to perform cloud tomography globally.
• 

  Methane Plume Detection with Imaging Spectroscopy

  This project leverages nearly a decade of experience in developing automated plume detection systems for methane remote sensing campaigns. Using imaging spectrometer observations from airborne platforms like AVIRIS-NG and GAO, as well as spaceborne platforms such as EMIT and Carbon Mapper/Tanager, the team has demonstrated robust capabilities for detecting methane plumes across diverse geographical locations. The project employs advanced deep learning techniques, including Convolutional Neural Networks (CNNs), Generative Adversarial Networks (GANs), and Transformers, to capture complex geospatial structures in remote sensing data. Through rigorous statistical validation and in-depth analysis of spatiotemporally diverse retrieval products, the research aims to address key challenges in operational plume detection and monitoring for both airborne and spaceborne platforms.
• 

  Bias Correction and Data Quality Filtering for OCO-2/3

  The Orbiting Carbon Observatory-2 makes space-based radiance measurements. Using a physics-based retrieval algorithm, these measurements are inverted to atmospheric CO2. However, biases are present in the retrieved CO2 due to calibration errors and discrepancies between the physics-based retrieval and nature. We employ a Random Forest, a non-linear, interpretable machine learning technique, to correct these biases and filter out data that cannot be corrected. We perform rigorous validation, quantify uncertainties, and apply explainability methods to build trust in the approach with the science community.
• 

  SUDS: Discovering the Drivers of Systematic Errors in Global Air Quality Estimates Using Explainable Machine Learning

  Accurate global estimates of air pollutants are essential to evaluate the global public health burden of disease associated with air quality exposure, which in turn will help environmental policy making to reduce risks from associated disease and to save human lives. The experimental machine learning framework uses NASA JPL's Multi-mOdel Multi-cOnstituent Chemical data assimilation (MOMO-Chem) estimates of ozone and the respective observed values of surface ozone from Tropospheric Ozone Assessment Report (TOAR) network to model the differences in their values.
• 

  Scientific Understanding from Data Science (SUDS) Council

  We aim to measurably improve JPL’s scientific understanding and productivity within Astrophysics, Earth and Planetary sciences by infusing data science techniques, such as uncertainty quantification, model-based inference, machine learning, and artificial intelligence, that can directly support science and its applications. The council shall develop a strategy that advances JPL leadership, in collaboration with Campus, in scientific understanding based on state-of-the-art data science methodologies. This council will first evaluate the current state of practice at JPL for scientific understanding from data science methods.

Past Projects

• 

  Adaptive Data Processing for Next Generation Radio Arrays

  We utilize machine learning methods to enable large-scale radio astronomy data analyses in real-time. Specifically, we focus on the detection and characterization of time-varying sources, one of the major scientific goals of radio arrays.
• 

  Agricultural Research Via Interactive, Scalable Technology

  We developed and demonstrated a machine learning analysis toolkit that uses Support Vector Machines, clustering, and regression models to identify the connections between weather and agriculture.
• 

  MISR Automated Cloud Classification

  By comparing images of the same area of the earth from different angles, scientists are able to identify thin clouds and determine approximate cloud heights with unprecedented accuracy.
• 

  Mars Target Encyclopedia (MTE)

  The Mars Target Encyclopedia (MTE) aims to collect in one place information about observational targets (e.g., rocks, soils) identified and named during Mars surface exploration. The MTE links targets to publications that provide information (e.g., composition, provenance, interpretation) about each target. Mission science team members, planetary science researchers, educators, students, and the general public can find relevant information about targets of interest or compare targets found by different missions. The MTE database will be released through the PDS and searchable through the Analyst's Notebook.