Skip to content

Tracking Cover Crops at Scale: unlocking field-scale cover crop dynamics via AI-driven satellite harmonization

This is a postdoc research project led by Dr. Yu Peng, supported by the Environmental Data Science Innovation & Impact Lab (ESIIL). The aim of this project is to integrate multi-scale field observations, satellite data fusion, and spatial-temporal modeling to map cover cropping across the continental US. and to quantify the ecosystem services that agricultural conservation delivers.

Since, it runs as an open-science system: a GitHub repository where this project is organized, analyzed, and versioned, and a public website where results are explained and shared with collaborators, mentors, and community audiences. The Repo and the website can acceess via following links. Homepage overview image

Open the GitHub repository

Research Abstract

Field-scale monitoring of cover crops is vital for assessing soil organic carbon (SOC) sequestration and greenhouse gas (GHG) dynamics across agroecosystems, yet reliable wall-to-wall information at that scale remains largely unavailable. Existing products are limited by coarse resolution, sparse ground truth, and inconsistent revisit across sensors, so the timing of establishment and termination — the part that matters most for carbon and nitrogen outcomes — is rarely resolved. This project leverages deep learning architectures to harmonize multi-sensor satellite observations (HLS,ECOSTRESS, NISAR), enabling high-resolution tracking of cover crop establishment, biomass accumulation, and termination dates across broad geographic domains. The resulting maps and time series are designed to feed directly into biogeochemical and policy-relevant assessments of agricultural conservation practice.

Research Objectives

  1. Synthesizing ground truths via LLMs. Generate a training and validation database of cover crop presence, timing, and species by combining field records, producer surveys, and LLM-assisted extraction from the published literature. (PD-C)

  2. Sensor harmonization. Fuse multi-source optical, thermal, and earth-embedding observations into a consistent, gap-filled field-scale time series. (Powered by the FireRX ML model, credited to Dr. Cibele Amaral.)

  3. Phenology retrieval. Train and evaluate AI models that retrieve establishment, peak biomass, and termination dates at the field level.

  4. Continental mapping and ecosystem services. Scale the retrieval across the continental US and link the resulting maps to SOC and GHG outcomes for conservation assessment.

research roadmap

Research Resources

The repository has two connected layers. Top-level files configure the project and its automation. The docs/ folder contains the website content. mkdocs.yml tells MkDocs how to turn that content into the public site. Analysis folders hold the working scientific materials that generate the results shown on the website.

Items Types What usually provides there
Data sources and metadata Configure the project and keep shared repository guidance in one place (1) global-scale yield responce to cover crops; (2) Field-scale soil GHG responces to cover crops
Tools and scripts Notebooks and scripts (1) Access cloud computuer-CyVerse ; (2)
mkdocs.yml Workflows and reproducible analysis Navigation, theme settings, plugins, and GitHub edit links
Working folders Figures, tables, manuscripts, and other outputs Data references, notebooks, scripts, workflows, figures, outputs, and reproducibility materials

Research Progress Data & Resources Science Sharing

Use this section to show how the project gets started without manually editing image links one by one.

Placeholder
Placeholder

This gallery displays early setup artifacts for the postdoc project.

Add or replace files in this gallery

To update: upload supported files to this folder and commit. The website sorts files alphabetically. Use this folder for kickoff notes, orientation screenshots, starter diagrams, and early planning visuals.

Delete this note after the site is finalized.

Project Members

Project identity and collaboration image

Name Role Institution Responsibilities
Yu Peng Postdoc Researcher ESIIL Project lead: data fusion, model development, analysis, and public reporting
Cibele Amaral Project Supervisor ESIIL Oversees daily research operations and remote sensing integration
Timothy Bowles Academic Mentor UC Berkeley Guidance on agroecology and academic development
Lixin Wang Advisory Expert IU Indianapolis Long-term research continuity and hydroecology expertise