Shan Z.

About Shan Z.

I am an interdisciplinary researcher working at the intersection of machine learning, remote sensing—particularly lidar—atmospheric and climate science, computational modeling, and data analysis. My work focuses on developing computational and data-driven methodologies to better understand complex Earth system processes.I am particularly interested in integrating physics-based models and remote sensing observations with modern machine learning techniques to enhance predictive capabilities in atmospheric and climate systems. My research addresses high-dimensional spatiotemporal data, numerical simulation of nonlinear dynamical systems, and uncertainty quantification in environmental modeling.I have applied machine learning methods to semantic segmentation, classification, and regression tasks, as well as to surrogate modeling, denoising, and the development and adaptation of large-scale Earth foundation models. These approaches enable more efficient representation, interpretation, and prediction of complex geophysical phenomena.More broadly, I am motivated by advancing computational frameworks—including emerging paradigms such as quantum computing—to tackle fundamental scientific challenges in Earth and climate science.

Experience

1. Scientific Researcher

   Coherent Applications, Inc

   Jun 2023 — Present · Hampton, VA, US

   Navigation Doppler Lidar Data Analysis and AI Roadmap (Starting October 2025) • Development of an AI research roadmap for navigation Doppler lidar data analysis • Advanced signal denoising and machine learning integration for navigation applicationsMachine Learning Research for Remote Sensing and Climate Modeling(NASA Contractor — through October 2025)Conducted machine learning research applied to remote sensing observations, numerical simulation outputs, and climate model parameterization. Work involved the development and deployment of diverse deep learning architectures, including Multilayer Perceptrons (MLP), LSTM networks, CNNs, Transformers, and Vision Foundation Models. Extensive use of high-performance computing (HPC) environments and CUDA-based GPU acceleration.Key Projects1. ML-Based Cloud Parameterization for Climate Models • Developed neural parameterizations of cloud volumetric fraction • Implemented MLP and LSTM architectures to learn subgrid-scale cloud processes • Improved representation of nonlinear atmospheric dynamics in Earth system models2. Radiative Transfer Surrogate Modeling • Designed Transformer-based surrogate models for radiative transfer simulations • Reduced computational cost while maintaining physical consistency • Enabled faster coupling with climate simulations3. Remote Sensing Classification and Segmentation • Applied CNN, SegNet, and U-Net architectures for image segmentation • Implemented Fuzzy K-Means clustering for unsupervised classification • Addressed noise, spatial heterogeneity, and uncertainty in geophysical data4. Earth Foundation Models • Worked with the Prithvi Foundation Model (based on MAE-ViT architecture) • Fine-tuned large-scale pretrained vision models for Earth observation tasks • Adapted masked autoencoder frameworks for geospatial data analy

Education

• Université Blaise Pascal (Clermont-II) - Clermont-Ferrand

  Master's degree, physics chemistry of the atmosphere and climate

  2006 — 2007

• École des mines d'Albi-Carmaux

  Third year of Engineer, Aeronautic Materials

  2005 — 2006

• University of Lille 1 Sciences and Technology

  Doctor of Philosophy (Ph.D.), Optics, laser, physical Chemistry, Atmosphere

  2007 — 2011

• Nanjing University of Information Science and Technology

  Bachelor's degree, Atmospheric Sciences and Meteorology

  2001 — 2005

Skills

• C
• Fortran90
• Idl
• Linux Operating Systems
• Hdf5
• Matlab
• Windows Operating Systems
• Mac Operating Systems
• Netcdf
• Bash Shell
• Msphinx
• Satellite Remote Sensing