Cropland Mapping using Google Earth Engine

Instructor: Ivan Zvonkov (University of Maryland) | Youtube Playlist | Slides Folder | Google Earth Engine Repository

🌾 1. What is cropland and why map it?

Video | Slides

Learning Outcomes

  1. βœ… Understand definitions of cropland
  2. βœ… Understand why cropland maps are needed

Assignment

01_GLAD_cropland

  1. Run script and investigate map
  2. Modify script to display maps from all years in paper
  3. [BONUS CHALLENGE] Modify script to display maps together as in paper Earth Engine App


πŸ‡ΈπŸ‡³ 2. Cropland in a specific region of interest

Video | Slides

Learning Outcomes

  1. βœ… Understand Senegal geography
  2. βœ… Load and investigate available cropland maps

Assignment

02a_GLAD_in_Senegal

  1. Run script and investigate result
  2. Modify the script following the TODO comments to display cropland in Senegal

02b_WorldCover_in_Senegal

  1. Run script and investigate result
  2. Modify the script following the TODO comments to display cropland in Senegal

02c_visualizing_all_maps

  1. Run script and investigate result
  2. Modify the script following the TODO comments to display multiple cropland maps
πŸ›°οΈ 3. Data used to map cropland

Video | Slides

Learning Outcomes

  1. βœ… Use crop calendar for data selection
  2. βœ… Load and process S2 data

Assignment

03a_S2_visualize

  1. Run script and investigate result
  2. Modify the script following the TODO comments to display S2 (Sentinel-2) median composites

03b_S2_cloudscore_visualize

  1. Run script and investigate result
  2. Modify the script following the TODO comments to display S2 cloudfree median composites

03c_S2_cloudscore_gif

  1. Run script and investigate result
  2. Modify the script following the TODO comments to generate S2 cloudfree median composite gif

03d_S2_NDVI

  1. Run script and investigate result
  2. Modify the script following the TODO comments to generate S2 NDVI composite gif
  3. [BONUS CHALLENGE] Modify the script to generate NDVI for the whole country. Tip: Comment out gif generation
πŸ€– 4. Machine learning for mapping cropland

Video | Slides

Learning Outcomes

  1. βœ… Understand rule-based classification
  2. βœ… Collect training data
  3. βœ… Understand ML-based classification

Assignment

04a_rule_based_classification

  1. Run script and investigate result
  2. Modify the script following the TODO comments to classify NDVI pixels into cropland pixels

04b_rule_based_classification_SL

  1. Run script and investigate result
  2. Modify the script following the TODO comments to better classify NDVI pixels into cropland pixels

04c_collecting_training_data

  1. Run script and investigate result
  2. Follow the TODO instructions to collect training data
  3. Follow the TODO comments to save the collected points

04d_ML_classification

  1. Run script and investigate result
  2. Follow the TODO comments to modify the script and classify cropland
πŸ’‘ 5. From map to information

Video | Slides

Learning Outcomes

  1. βœ… Improve cropland map
  2. βœ… Conduct accuracy assessment
  3. βœ… Understand basics of area estimation and uses

Assignment

5a_ML_classification_improve

  1. Run script and investigate result
  2. Follow the TODO comments to modify the script and better classify cropland in Saint Louis, Senegal

5b_assessing_accuracy

  1. Run script and investigate result
  2. Follow the TODO comments to load your map and assess its accuracy

5c_pixel_counting

  1. Run script and investigate result
  2. Follow the TODO comments to print the areas in km2 and the cropland area proportion
6. 🧠 Final Project

Final Project

For the final project, you will be using the knowledge and skills acquired during the course to create a cropland map for a different Senegal admin zone.

Project Deliverables

  1. Admin zone selection (Tip: use this script)
  2. One page document on the admin zone geography and agriculture
  3. GEE Script that processes Sentinel-2 data for cropland investigation
  4. GEE Script for collecting training data points
  5. GEE Script for creating cropland map using collected training points
  6. Cropland map saved as a GEE Asset
  7. GEE Script for conducting accuracy assessment (using provided points)
  8. One page document outlining your process and results

Learning Outcomes

  1. βœ… Analysis of remote sensing data
  2. βœ… Create a quality cropland map from scratch
  3. βœ… Conduct accuracy assessment

Optional Bonus Tasks

  1. Conduct a visual assessment comparing your map to existing cropland maps
  2. Investigate adjustments to Sentinel-2 data time frame
  3. Investigate use of Sentinel-1 data in your admin zone
  4. Investigate other ML classifiers available in GEE

The course was supported by the University of Maryland, NASA SERVIR, and the Xylem Lab.