Subscribe to our lab email list for the latest updates
Mapping ecosystems has historically been a labour-intensive and complex task that requires advanced geospatial skills, deep ecological knowledge, and the capacity to have and use a range of different data sets (field data, training data, validation data, existing maps). Recently, pre-processed datasets, international standards such as the Global Ecosystem Typology, and user-interfaces to modelling tools are facilitating this process. Further, tools, data, and guidelines developed for the Global Ecosystems Atlas can support countries in national ecosystem mapping.
In this exercise, we will use Google's Satellite Embeddings Classifier (https://tinyurl.com/map-ecosystems), a prototype tool to demonstrate a workflow of how to map ecosystems. This tool, developed by Google Earth Engine team, provides quick and easy access to the new, preprocessed Satellite Embeddings dataset. The satellite embeddings are developed at 10-m spatial resolution and integrate a very broad range of satellite data into a single, analysis-ready dataset. This has greatly reduced the resources required to develop ecosystem maps at any scale.
Today we will train a model to identify ecosystems in the Philippines using the embeddings and yield a map of Ecosystem Functional Groups.
Many thanks to Sean Askay and the team at Google Earth Engine for developing the app that we are learning to use today.
Today, we will use the Satellite Embeddings Classifier to:
Understand the importance of carefully understanding the Global Ecosystem Typology, which forms the basis of the map classification scheme.
Learn the importance of carefully collected training data, which are confirmed occurrences of each ecosystem type in the classification scheme.
Explore how training data is collected and how collecting more points can improve model performance and map accuracy.
Provide the basis of a discussion about how training data could be collected for a national-scale mapping effort in Indonesia.
Global Ecosystems Atlas website (www.globalecosystemsatlas.org)
Global Ecosystems Typology website (www.global-ecosystems.org)
Embedding Fields Classifier App (link)
Introduction to Earth Observation and Remote Sensing (PDF)
A presentation about ecosystem mapping by A/Prof Nicholas Murray on Google Earth's youtube channel: https://www.youtube.com/watch?v=qzN0y884DOs&t=2338s
Step 1. Open the app and find a small area
Navigate to the Satellite Embeddings Classifier application: https://tinyurl.com/map-ecosystems
Select an area of interest or a small geographic region for which you know well and want to develop a map of ecosystems. We suggest a small area to start. Once you're familiar with the tool, you can map a larger area.
Step 2. Decide on your ecosystem classes
Identify the Level 3 Ecosystem Functional Groups present in your area of interest. For example, a coastal area is likely to have some of the following ecosystem functional groups:
MT1.3 Sandy Shorelines
MFT1.2 Intertidal forests and shrublands (mangroves)
MT1.2 Muddy Shorelines
MT1.1 Rocky Shorelines
T7.4 Urban and industrial ecosystems
T7.3 Plantations
M1.7 Subtidal sand beds
M1.1 Seagrass meadows
M1.3 Photic coral reefs
You can see a list of potential Ecosystem Functional Groups by using the 'analyse' tool on the Global Ecosystem Typology website.
To develop a deep understanding of the Ecosystem Functional Groups you are mapping, we suggest you take the time to read the ecosystem descriptions of each of these ecosystem functional groups. Detailed descriptions of each Ecosystem Functional Group are available on the Global Ecosystem Typology website.
Step 3. Specify your map classes in the app
Define the classes you will map by clicking on the + new class button on the left-hand side of the screen. A box will pop up.
Write the name of the ecosystem type in the white box.
Select the corresponding Ecosystem Functional Group for the ecosystem type. You can search for the functional group name or scroll through the list. Click on the colour to select it, and the colour and functional group name will be automatically included.
Click done on the bottom right of the pop-up box.
Repeat these steps for all the ecosystem types. For a demonstration, do at least three classes.
By the end, you should have at least 3-10 classes in your map legend.
Step 4. Develop your training data
Now we must place points on the map for each of these classes.
Click on the name of the class and then on the + button on the right of the name.
Add 5-10 points per class. Use your knowledge of an area and of the ecosystem to find places where you are confident you know that the ecosystem occurs.
Repeat this for each class.
Once you have added some points for at least two classes, the model will automatically run and begin to make a map. You can turn this map on and off by clicking on the 'layers' button on the top left and checking or unchecking 'Classification'.
Step 6. Refine your map through iteration
Now that you have your first draft map, you can begin the process of improvement. Typically, you would continue to work through refining your training data until you cannot make your map any more accurate.
Add more training points, aiming for a well-distributed set of training data that covers all of the ecosystem functional groups in your map.
Add more points in places in your map where your first draft is not correct. See the images below for adding more urban points where the original map did not predict.
As you continue to add new training data, your map will slowly improve until you're unable to make any further improvements.
Finished
With this workflow, you've been able to develop a classified map of ecosystem functional groups for your area of interest!
Notes and advice
Here are a few final tips to advance your map using the Satellite Embeddings Classifier:
The map will only be as good as the training data, so try to be accurate with your training data.
You can try 'lumping' very similar classes. When classes such as two different types of wetland, or even oceanic versus lakes, it's often better to lump them into a single 'water' class. After making your map, we can then do a post-processing analysis to 'split' those classes back into Ecosystem Functional Groups.
Some ecosystems, such as tropical flooded forests, are extremely difficult to map using Earth observation data alone. Therefore, in these situations, it is often necessary to 'lump' similar forest types and then implement a post-processing analysis to split them back into particular ecosystem types.
Ecosystem occurrence points are a valuable resource for training these models. Investment and rigorous methods are important to ensure that the data is the most accurate and can achieve the best map.