Welcome to STIMP’s documentation!

The python repository STIMP (GitHub repository) implements the methods in the STIMP paper. STIMP an advanced AI framework to impute and predict Chl_a across a broad spatiotemporal scale in coastal oceans. STIMP’s results can be utilized to diagnose and analyze the ecosystem health of coastal oceans based on the remote sensing measurement.

In this tutorial website, we provide guidelines for using STIMP along with remote sensing Chl_a data analysis examples, including imputation and prediction. The source code for building the website can be found at https://github.com/Ryanfzhang/STIMP-tutorials.

Contents

• Installation
  • Installing STIMP from GitHub
• Usage
  • Pipeline overview
  • Prepare data
  • Train STIMP
    • Step1. train the imputation function
    • Step.2 Generate multiple imputations for Chl_a:
    • Step.3 Train prediction function based on each imputation
• Train baselines
  • Imputation Methods
  • Prediction Metods
• Four representative coastal ocean area analysis
  • Pearl River Estuary
    • Imputation with trained STIMP in the Pearl River Estuary
    • Prediction performance of STIMP in Pearl River Estuary
    • Case study of prediction performance in Pearl River Estuary
  • Northern Gulf of Mexico
    • Imputation with trained STIMP in the Northern Gulf of Mexico
    • Prediction performance of STIMP in the Northern Gulf of Mexico
    • Case study of prediction performance in the Northern Gulf of Mexico
  • Chesapeake Bay
    • Imputation with trained STIMP in the Chesapeake Bay
    • Prediction performance of STIMP in the Chesapeake Bay
    • Case study of prediction performance in the Chesapeake Bay
  • Yangtze River Estuary
    • Imputation with trained STIMP in the Yangtze River Estuary
    • Prediction performance of STIMP in the Yangtze River Estuary
    • Case study of prediction performance in the Yangtze River Estuary
• Supplementary main results
  • Imputation length (including half year and one-and-a-half years)
    • Half-year Imputation
    • One-and-a-half-years Imputation
  • Prediction Length
    • Different Prediction Horizons in PRE
    • Different Prediction Horizons in Chesapeake Bay
    • Different Prediction Horizons in Gulf of the Mexico
    • Different Prediction Horizons in Yangtze River Estuary
  • Recurrent Prediction
    • Recurrent Prediction
  • Case study: predicted Chl_a of STIMP and baselines trained on imputed dataset
    • Predicted Chl_a of STIMP and baselines trained on imputed dataset at three locations
  • STIMP, MaskedAE and Lin-itp
    • STIMP, MaskedAE and Lin-itp
  • Imputation of DINEOF with different t
    • STIMP vs DINEOF with different values of t
  • Predicted Chl_a of STIMP and Climate mean
    • Overall improvement of STIMP compared to two climate mean predictions
    • Predicted Chl_a of STIMP and two climate mean state at four locations

Reference

If you find the STIMP package or any of the source code in this repository useful for your work, please cite:

AI-powered spatiotemporal imputation and prediction of chlorophyll-a concentration in coastal ecosystems.

Fan Zhang, Hiuseut Kung, Fa Zhang, Can Yang and Jianping Gan.

2025.

Development

The python repository STIMP is developed and maintained by Fan Zhang.

Contact

Please feel free to contact Fan Zhang, Prof. Can Yang, or Prof. Jianping Gan if any inquiries.