Weijie Zhang
Currently looking for open positions

Postdoctoral researcher, UC Berkeley

Weijie Zhang

Machine learning & deep learning for ecohydrology

I build machine learning and apply Earth foundation models to understand how land surfaces exchange water, carbon and energy with the atmosphere, working with Prof. Trevor Keenan.

See my research Get in touch
Portrait of Weijie Zhang

Energy imbalance

1:1gapRn − GH + LE

Correcting the energy-balance closure gap in flux data.

Flux upscaling

Satellite + climate dataFLUXNET towersMLGlobal flux maps

From flux towers to global maps with machine learning.

After FLUXCOM

Geo-foundation models

LandsatEmbeddingsFlux map

Satellite embeddings for high-resolution carbon and ET maps.

About

I use machine learning and deep learning to understand the land’s water, carbon and energy cycles.

I combine flux towers, streamflow and satellite data with Earth foundation models, from single sites to the globe. Papers are on Google Scholar and ResearchGate.

Position
Postdoctoral researcher, UC Berkeley
Group
Prof. Trevor Keenan
Ph.D.
Max Planck Institute for Biogeochemistry & Ghent University
Methods
Machine learning, deep learning, Earth foundation models

Research

  • Deep learning and geo-foundation models for carbon and ET mapping

    I apply Earth foundation models such as Prithvi-EO-2 and AlphaEarth, pairing their satellite embeddings with flux-tower observations in deep learning models, to map carbon and evapotranspiration at high resolution from single towers to continents and the globe.

    • Deep learning
    • Earth foundation models
    • Prithvi-EO-2
    • AlphaEarth
    • AmeriFlux & NEON
    Landsat bandsNIR, red, false colourGeo-foundation modelPrithvi-EO-2AlphaEarthembeddingsCarbon & ET maplowhighFlux towers supervise trainingLandsat scene, Berkeley hills
    Animated illustration built on a real Landsat scene of the Berkeley hills (USGS/NASA). The map is a vegetation index from the same scene, shown to illustrate the output.

  • Machine learning for evapotranspiration and hydrology

    How much water land surfaces return to the atmosphere, how evapotranspiration partitions into transpiration and evaporation, and how vegetation water stress shapes both, learned from flux, streamflow and remote-sensing data with interpretable machine learning.

    • ET partitioning
    • Interpretable ML
    • Water stress
    • Streamflow

  • Surface energy balance

    How net radiation (Rn) is partitioned into latent heat (LE), sensible heat (H) and ground heat flux (G), and how that partitioning shapes land-surface warming and drying, including data-driven corrections for energy-balance closure in flux networks.

    • Rn, LE, H, G
    • Data-driven closure correction
    • FLUXNET

  • Wildfire and post-fire carbon recovery

    How burned ecosystems across North America recover their carbon uptake, and what controls the pace of that recovery.

    • Disturbance
    • Carbon recovery
    • North America

News

  • Oral presentations at the AmeriFlux Annual Meeting, Berkeley.
  • Hosting a breakout session on Geo-foundation model embeddings for the flux community at the AmeriFlux Annual Meeting, Berkeley.
  • Primary convener of AGU 2026 Session GC127 on vegetation dynamics and land–atmosphere interactions.
  • New manuscript under review: constrained land–atmosphere energy partitioning points to a hotter, drier future land surface with more limited carbon uptake.
  • Funding awarded for Remote Sensing & Fluxes for Real-World Impact ($2,500) and the HeatHack workshop at NCAR ($2,000).
  • Received the NSF NAIRR AI Education Fellowship ($5,000) and Google Cloud Research Credits ($5,000) to run Earth foundation models for surface carbon and water fluxes.
  • Led the FLUXNET-ECN workshop Data-driven methods to correct the energy balance closure problem in FLUXNET datasets, including a hands-on session.
  • AmeriFlux Annual Meeting, Tucson. Poster with Wantong Li and Trevor Keenan on how representative the AmeriFlux and NEON networks are for carbon and water flux estimates.
  • Chapman Conference on the Energy Balance Closure Problem, Boulder. Presented the correction framework from my PhD and the open problems I plan to address next.
  • Deep Learning for Science School, LBNL Berkeley. Poster on geo-foundation models for carbon and water flux estimates across AmeriFlux and NEON.

Tools

HeatFlux

Interactive app for energy-balance corrections of latent and sensible heat.

Open app
HRHC

Corrects eddy-covariance water-flux underestimation under high relative humidity.

GitHubColabPaper
AEC

Available-energy-based correction for energy-balance closure at flux sites.

GitHubColab

Coming next: global carbon and ET maps from geo-foundation models.

CV

  • 2024–now
    Postdoctoral researcher
    ESPM, University of California, Berkeley, with Prof. Trevor Keenan. DOE project: drought, heat and wildfire impacts on coastal water relations
  • 2020–2024
    Ph.D.
    Max Planck Institute for Biogeochemistry & Ghent University. Advisors: Martin Jung, Jacob Nelson, Markus Reichstein, Diego Miralles
  • 2023–2024
    Research visits
    Northern Arizona University (Andrew Richardson) and Ghent University (Diego Miralles)
  • 2017–2020
    M.S., Tree Ring and Paleoclimate Change
    Nanjing University

Selected publications

  1. Zhang, W., et al. (2024). A new post-hoc method to reduce the energy imbalance in eddy covariance measurements. Geophysical Research Letters, 51, e2023GL107084. DOI
  2. Zhang, W., et al. (2023). The effect of relative humidity on eddy covariance latent heat flux measurements and its implication for partitioning into transpiration and evaporation. Agricultural and Forest Meteorology, 330, 109305. FLUXNET Paper of the Year. DOI
  3. Nelson, J. A., …, Zhang, W., et al. (2024). X-BASE: the first terrestrial carbon and water flux products from an extended data-driven scaling framework, FLUXCOM-X. Biogeosciences, 21, 5079–5115. Highlight paper. DOI
  4. Kraft, B., …, Zhang, W., et al. (2025). On the added value of sequential deep learning for upscaling evapotranspiration. Biogeosciences, 22, 3965–3987. DOI
  5. Li, W., …, Zhang, W., et al. (2023). Widespread and complex drought effects on vegetation physiology inferred from space. Nature Communications, 14, 4640.

Full list on Google Scholar.

Honors & funding

  • 2026Remote Sensing & Fluxes for Real-World Impact ($2,500)
  • 2026HeatHack workshop at NCAR ($2,000)
  • 2026Tree-ring for ecology travel award
  • 2025NSF NAIRR AI Education Fellowship ($5,000)
  • 2025Google Cloud Research Credits ($5,000)
  • 2023Inaugural FLUXNET Paper of the Year

Service

  • 2026Primary convener, AGU 2026 Session GC127
  • 2026Convener, AmeriFlux Annual Meeting breakout on Earth foundation models
  • 2025–FLUXNET2025 Data Support Team
  • 2024Panel reviewer, NASA ROSES Carbon Cycle Science
  • 2022Organizer, FLUXNET Early Career Network; Cross-Scale ET workshop
  • ReviewerScience Advances, GRL, Remote Sensing of Environment, AFM, J. Hydrology, Biogeosciences, and others

Teaching & talks

  • 2026Talks: AmeriFlux Annual Meeting; GEWEX Earth’s Energy Imbalance workshop; GEWEX GLASS Panel conference
  • 2025–Mentoring four graduate and visiting students
  • 2025FLUXNET-ECN workshop on energy balance closure corrections, with hands-on session
  • 2023–24Talks at Columbia, Peking, Ghent and Northern Arizona universities
  • 2021–25Talks at AGU Fall Meeting and EGU General Assembly