Australia Saltmarsh Map
A national dataset of saltmarsh distributions to support the Blue Carbon Method
Saltmarshes are one of Australia's most widespread coastal ecosystems, yet their spatial extent around the country remains largely unquantified. Recent advances in cloud-based geospatial platforms have enabled the development of an analysis pipeline to monitor saltmarsh distribution at a continental scale.
Our remote sensing pipeline builds on work to develop the first integrated global maps of tidal flat, saltmarsh and mangrove ecosystems. Read more about that work here.
We are first compiling a large set of occurrence records to train our remote sensing classification models (at least 10,000 point records across Australia's coastline). Our training set includes field data acquired from Australia's saltmarsh research community, published papers and records developed from high-resolution image interpretation conducted at JCU's Global Ecology Lab.
Our remote sensing classification approach uses up to 100 nationwide remote sensing-derived covariate layers to support a suite of machine-learning classification models. The classification models are tasked with estimating the coastal ecosystem type of every 30-m pixel that occurs around Australia's coastal zone. This mapping framework will be implemented as a collaborative effort between JCU, UNSW and Digital Earth Australia.
This project will deliver a wall-to-wall, nationally consistent and updateable map of saltmarsh ecosystems suitable for research and governance of Australia's coastal ecosystems.
Frequently asked questions
What ecosystems will the map depict?
Initially, our map will depict the distribution of vegetated saltmarsh ecosystems. This map is designed to integrate with the Blue Carbon Method, which defines saltmarsh as an ecosystem that: (a) is comprised of salt tolerant plants that are herbaceous as well as some woody shrubs; and (b) occurs on floodplains and in estuaries and can be flushed with water from a combination of water sources.
Additionally, we are collecting training data for sparsely vegetated saltpans for future analysis.
What is the Blue Carbon Method?
This is the first Blue Carbon method under the Emissions Reduction Fund (ERF). Blue Carbon is a term commonly used for coastal wetland and marine ecosystems that can sequester and store high amounts of organic carbon, and release very low amounts of greenhouse gases. This method enables Australian Carbon Credit Units (carbon credits) to be earned by projects that remove or modify tidal restriction mechanisms and allow tidal flow to be introduced to coastal wetland ecosystems (that may be vegetated or unvegetated) through permanent or seasonal inundation with saline or brackish water.
Read more about the Blue Carbon Method here.
How will the saltmarsh map be produced?
We use satellite image composite metrics and stacked machine learning classifiers to assess whether 30-m pixels across Australia's entire coastal zone are likely to be saltmarsh. Pixels with high likelihood of being saltmarsh are classified as saltmarsh pixels and are included in the map.
For more information on our modelling approaches please refer to some of our recent papers:
What is the spatial resolution of the map?
The first map will be delivered at 30-m pixel resolution. However, our training data and models are sufficiently flexible to support predictions using other sensors, such as ESA Sentinel data. Future versions of the map may be produced at resolutions up to 10-m.
Who is behind this project?
This project is funded by the Clean Energy Regulator, with the modelling workflow developed at JCU's Global Ecology Lab (via an ARC Discovery Early Career Fellowship to Murray) in collaboration with UNSW Centre for Ecosystem Science.
The majority of the team is based at JCU Townsville (Nick Murray and Alejandro Navarro) and UNSW (Mitchell Lyons). The team at JCU's focus is to develop the training dataset and the classification models workflows and testing in Google Earth Engine, while supporting the team at UNSW to test the ability for these models to be ported to Digital Earth Australia.
We also have a technical advisory group, which includes scientists from Geoscience Australia, DAWE and the Clean Energy Regulator, and an ecology advisor group, which includes scientists at UQ, JCU and UNSW.
Can I access the data?
This project is running from mid-2021 to mid-2023. During this period we will circulate draft maps to the expert advisory groups and other end users. If you are interested in reviewing draft maps, please contact us directly or join our Google Group.
At the end of the project we expect the map will be made freely available for a range of end-users.
How will you manage data freshness?
Data freshness relates to how up-to-date the dataset is. This project will be implemented on Digital Earth Australia and, at the completion of the project, the remote sensing pipeline and training data will be inherited by Geoscience Australia and an annual update schedule will be established.
Are you planning to map Australia’s other intertidal ecosystem types?
The Global Intertidal Change models have the ability to simultaneously map and monitor other coastal ecosystem types, such as tidal flats, mangroves, and supratidal forests. We are collaborating widely to develop the training sets needed for this type of application in Australia. In addition, we are also developing a seagrass training set to support next-generation seagrass remote sensing models.
The Australia saltmarsh map will consist of several layers:
Raster binary occurrence
Every 30-m pixel in the coastal zone will be flagged as saltmarsh. (0-1)
Quality assurance layer
A layer indicating the likelihood that the classification is correct (0-100)
Example from River Derwent, Tasmania
This project will also deliver a training set of saltmarsh occurrence records which we use to train our classification models. Each row in the training set represents a single occurrence of a saltmarsh ecosystem at the 30-m pixel scale for a known reference date and has the following columns:
class: an integer representing ecosystem type, corresponding to the coastTrain global intertidal training library class values.
latitude: latitude of the training point
longitude: longitude of the training point
start_date: valid start year of training point
end_date: valid end year of training point
geometry: geometry type in WKT format (point)
analyst: the analyst who entered this record
type: the method of collection
source: name of any experts that contributed to this record