Fiji Forestry have a need for accurate maps of land cover fin areas they manage. A pilot project is being done to co-develop a land cover mapping methodology using the geospatial tool, Google Earth Engine, and Collect Earth Online to produce 10m land cover map for a single forestry site (Nalotawa) based on Sentinel-2 satellite data. This workshop focuses on Collect Earth Online (CEO) - a web-based, collaborative tool for collecting land cover data for map generation and accuracy assessment.
There are two components to this training: - Learning CEO - Collecting Land Cover DataAll workshop attendees must register with CEO well before the scheduled day of the workshop.
Before you can start this workshop, you will need to register with Collect Earth Online:
On 14th and 15th July 2020, a field team from Ministry of Forestry and USP labelled land cover plots in the Nalotawa Forestry area. The land cover classes used by the field team are shown below (at Level 2):
|Level 1||Level 2||Code||Definition|
|Forest||Closed Forest||1||Tree > 40% cover|
|Open Forest||2||Tree < 40% cover|
|Grassland||Dry Grassland||4||May include herbs/shrubs. Tree canopy < 10% cover.|
|Wet Grassland||5||May include herbs/shrubs. Tree canopy < 10% cover.|
|Cropland||Cropland||6||Area >= 50% covered by agriculture|
|Wetland||Mangroves||7||Area >= 30% covered in mangroves|
|Settlement||Settlement||8||Area >= 30% developed land including houses, roads|
|Bare Ground||10||Bare soil/rock|
The objective of this exercise is to learn how to label these same plots using Collect Earth Online (CEO). You will be using the provided satellite images in CEO to decide which land cover is present. You will then compare your land cover label with the class assigned to the same plot during the fieldwork. At the end of this exercise, your instructor will lead a short discussion on the results and your experience with CEO.
Ensure you have Unanalyzed plots chosen and click the Go to first plot button.
The instuctor will now demonstrate the various steps to label a plot. Follow along and use Fig. 4 below as a reference.
Your instructor will lead a discussion on some example plots where the particpant’s land cover class did not agree with the land cover class assigned during the fieldwork. Remember, there is not necessarily a correct answer, so it is important to discuss these ambiguities and why they occur. This will help you use your existing knowledge of the area to improve your image interpretation skills.
You will use your CEO skills to label new plots for the Nalotawa forestry site. The labels you assign in this exercise will be used to produce a pilot land cover map for this area. All participants will be labelling unique plots in the same project which demonstrates the collaborative nature of CEO.
One the group has finished the exercise, you instructor will download the labelled data and demonstrate how the land cover map can be produced in Google Earth Engine.