Neural Network Architectures for AirHARP Aerosol Retrieval and Coastal Spatiotemporal Forecasting
Dr. Xiaoguang Xu, Joint Center for Earth Systems Technology.
This project utilizes deep learning and neural network architectures to process high-dimensional environmental data. The first objective focuses on reconstructing multi-angle scattering curves from the NASA AirHARP satellite instrument using PointNet and Transformer architectures to improve microphysical aerosol retrievals. The second objective involves training spatiotemporal regression and Convolutional Neural Networks on NASA and NOAA satellite telemetry to forecast Harmful Algal Blooms (HABs) and water quality shifts in the Chesapeake Bay. Access to the CHIP cluster’s GPU facilities is required to train these complex models on large-scale historical datasets and high-resolution spatial grids without local memory limitations.
HARP 2
Dr. Xiaoguang Xu, Joint Center for Earth Systems Technology.
HARP2 is a contributed instrument from the UMBC Earth and Space Institute to the PACE mission. HARP2 is a copy of the HARP CubeSat polarimeter payload adapted for flying and collecting data from a large spacecraft platform like PACE. HARP2 on PACE will have the capability to collect global data in two days, while the HARP CubeSat payload have important limitations in terms of amount of data that can be collected and downlinked to the ground, The figure bellow compares the main differences between HARP2 and HARP CubeSat.
Facilitating the Next Generation of Air Quality Science: Synergy of NASA’s Ground-Based Observation Networks
Dr. Xiaoguang Xu, Joint Center for Earth Systems Technology.
The proposed work will combine observations of winds, aerosols, and trace-gas pollutants to complement and extract new science from NASA’s existing network of Tropospheric Ozone Lidar (TOLNet) and Pandora spectrometers, focused on regions near complex coastal terrain. The following are two overarching questions from which the tasks of this proposal are derived. Q1) How can observations of ozone from TOLNet and Pandora be combined with regional in-situ measurements to capture and quantify the evolution and impact of ozone episodes? Q2) Do chemical transport models and planned satellite retrievals properly represent observations of vertically resolved and columnated ozone from TOLNet and Pandora.
Determining Atmospheric Aerosol Properties from the UMBC HARP CubeSat in Space
Xiaoguang Xu, Department of Physics
J. Vanderlei Martins, Department of Physics
Rachel Smith, Department of Physics
Designed and built at the UMBC Earth and Space Institute (ESI), the Hyper-Agular Rainbow Polarimeter (HARP) CubeSat was deployed from the International Space Station on the 19th of February 2020 and has been orbiting our planet since. By measuring the intensity and polarization state of Earth-reflected solar radiation from space, HARP provides accurate and comprehensive measurements of atmospheric aerosol and cloud properties. However, HARP CubeSat is funded by NASA as a technology demonstration project and, as such, has no funding for science data analysis, which I intend to advance through this grant. In the proposed project, we aim at an in-depth analysis of the performance of HARP CubeSat imagery, and from which we will determine the distribution of aerosols in terms of their atmospheric loading and microphysical properties. To achieve this goal, we will (1) calibrate the collected HARP imaging data to address potential instrument degradation, (2) develop a retrieval algorithm to extract aerosol information from the calibrated HARP radiance and polarization signals, (3) apply the aerosol retrieval algorithm to the entire collection of HARP CubeSat images and make HARP aerosol products available to the community, and (4) attract external collaboration in major proposals that intend to use HARP-like data for broader applications. A successful outcome on this project will allow us to be more competitive on future application of HARP-like sensors, which are currently being planned by the broad scientific community but require science demonstration and publications.