Blue carbon remote sensing

Earth observation for assessing the status of coastal ecosystems

Description

Habitat loss, degradation and fragmentation threaten ecosystems worldwide. Satellite remote sensing has been crucial in documenting these threats, and is a powerful tool for identifying areas undergoing losses. However, existing methods for mapping change typically rely on expert implementation, which is a severe constraint in achieving comprehensive information on the status of the world’s ecosystems. Through practical exercises, this short course will train attendees to the use of remote sensing for mapping habitats and quantifying how they change over time. 

 We will learn to use REMAP (https://remap-app.org), a free, open-source, online remote sensing application funded by Google and developed to provide easy access to the immense capabilities of Google Earth Engine. Remap has been used by more 10,000 people around the world to quickly develop highly accurate maps from Landsat data, identify areas undergoing land cover changes and quantify the amount of habitat loss over a two-decade period. 

Today we will use REMAP to make map classifications of coastal ecosystems, collecting information on the amount of coastal wetland lost and gained, and report back to the group.

Workshop materials

Global Intertidal Change website (www.globalintertidalchange.org)

Remap website (www.remap-app.org) 

Google sheet for collecting results (link)

Supporting materials

Introduction to Earth Observation and Remote Sening (PDF)

Introduction to Remap (PDF)

Remap tutorials

Remap paper (Methods in Ecology and Evolution, 2018)

Tidal flats paper (Nature, 2019)

Global intertidal change paper (Science, 2022)

A presentation about Remap by Dr Nick Murray on Google Earth's youtube channel: https://www.youtube.com/watch?v=qzN0y884DOs&t=2338s 

Workshop instructions

1. Select one of the locations above as your area to work on.

   • Identify the main land cover types - for the purpose of this exercise we will classify 3 types (mangrove/marsh, land and water)

   • Use the global intertidal change layers to get an understanding of the extent of expected change

   • Consider what is causing the change, is it natural or anthropogenic?

   • Do a quick search of the internet for any evidence of the main change drivers.

2. Open Remap.

   • Use the search bar to navigate to your area of interest

   • Use the zoom button to situate your area of interest

   • Click Focus Region to choose your study area. You can also move the corner vertices to adapt it.

   • Try to avoid making this too big, small and focused (say 50km wide) will yield better results in just one hour.

3. Observe change

   • Use the Past/Present toggle (above map control) to look at the change, as observed by the Landsat satellite

   • Now, try using the Map Control button to choose different satellite layers - NDVI might be a good choice to identify mangroves (the greenest)

   • Best will be to return to "Natural" for the purposes of your classification

4. Develop a classified map

   • To develop a classified map of the coastal wetlands in your study region, you'll need to "build training set"

   • We will use three classes (land, water, coastal ecosystem). Do note you can increase the number of classes, which will improve your results and allow you to map other things you find interesting (such as the case of Dubai, in which case you only need two classes, land and water).

   • Add at least 50 training points for each of these classes, you can choose colours and play around with how it visualises.

   • When you've built your training set, try to run the classification. Click "Classify!". Here we are packaging up our analysis, sending it off to Google Earth Engine, running it on hundreds of machines, compiling the results, and then returning them to the browser. All in a few seconds! Be patient if you have to be :)

   • Now, if you're happy that your map is accurate and looks good proceed to the next step. If you want to make it better, simply add more training points, perhaps a new class if there is any class confusion, and then rerun your model. You can even try changing predictor layers in the Select Predictors section.

5. Now, collect your results. We will report them back to the group.

   • Click "Results"

   • Here collect your results in our shared Google Sheet. Make sure to include columns for "Year" (which will be 2016), and the area of your classes. Of most interest is the area of coastal wetlands such as mangroves.

6. Now, to understand the extent of loss and gain you need to rerun your model on an earlier set of Landsat images

   • Click the Past/Present toggle to import the year 2000 layer.

   • Make sure your training data is still relevant (ie. mangroves still sit on mangroves, all the data are still correct). If not, move them around until they are correct.

   • Now Click Classify, which will yield a map of these habitats for the year 2000.

   • Collect the data as in the spreadsheet before.

7. Now, compute the % change of the coastal wetlands class

   • This is the area of coastal wetlands in 2016, divided by its original area in 2000, multiplied by 100. 

   • Record your results in this Google Sheet.

8. Save your results by clicking export data, download JSON. This JSON can be re-uploaded at any time via "Build training set"

9. We will reconvene with 20 minutes to go in the lecture to discuss what we found.