DEA Water Observations Statistics (Landsat)

About

Digital Earth Australia (DEA) Water Observations uses an algorithm to classify each pixel from Landsat satellite imagery as ‘wet’, ‘dry’ or ‘invalid’. Combining the classified pixels into summaries, covering a year, season, or all of time (since 1987) gives the information on where water is usually, and where it is rarely.

Background

These are the statistics generated from the DEA Water Observations (Water Observations from Space) suite of products, which gives summaries of how often surface water was observed by the Landsat satellites for various periods (per year, per season and for the period from 1986 to the present).

Water Observations Statistics (WO-STATS) provides information on how many times the Landsat satellites were able to clearly see an area, how many times those observations were wet, and what that means for the percentage of time that water was observed in the landscape.

What this product offers

Each dataset in this product consists of the following datasets:

Clear Count: how many times an area could be clearly seen (i.e. not affected by clouds, shadows or other satellite observation problems)
Wet Count: how many times water was detected in observations that were clear
Water Frequency: what percentage of clear observations were detected as wet (i.e. the ratio of wet to clear as a percentage)

As no confidence filtering is applied to this product, it is affected by noise where misclassifications have occurred in the input water classifications, and can be difficult to interpret on its own.

WO-STATS is available in multiple forms, depending on the length of time over which the statistics are calculated. At present the following are available:

DEA WO Multi-Year: statistics calculated from the full depth of time series (1986 to present) unfiltered
DEA WO Calendar Year: statistics calculated from each calendar year (1986 to present)
DEA WO November to March: statistics calculated yearly from November to March (1986 to present)
DEA WO April to October: statistics calculated yearly from April to October (1986 to present)

In addition a confidence-filtered Multi-Year Summary will become available in 2022 which will contain a confidence layer and subsequent filtered water frequency layer. This provides a noise-reduced view of the unfiltered multi-year summary.

Applications

Helps understand where flooding may have occurred in the past, to inform emergency management and risk assessment.
Provides an indication of the permanence of surface water in the Australian landscape by showing where water is observed rarely in comparison to where it is often observed, informing water management and mapping.
Can assist with wetland analyses, water connectivity and surface-ground water relationships.
The annual product provides information on how surface water changes per year across Australia, and is useful for drought analysis.
The seasonal product is useful for understanding the differences in water availability between the summer and winter periods across Australia.

Publications

Mueller, N., Lewis, A., Roberts, D., Ring, S., Melrose, R., Sixsmith, J., Lymburner, L., McIntyre, A., Tan, P., Curnow, S., & Ip, A. (2016). Water observations from space: Mapping surface water from 25 years of Landsat imagery across Australia. Remote Sensing of Environment, 174, 341–352. https://doi.org/10.1016/j.rse.2015.11.003

Data access

Product ID	ga_ls_wo_fq_apr_oct_3, ga_ls_wo_fq_nov_mar_3, ga_ls_wo_fq_cyear_3, ga_ls_wo_fq_myear_3, ga_ls_wo_filtfq_myear_3,
Link to data	Seasonal Water Summaries Apr-Oct periods since 1986 Seasonal Water Summaries Nov-Mar periods since 1986 Annual Water Summaries per Calendar Year since 1986 All-time Water Summary (Unfiltered) 1986 to present
Link to maps	DEA Maps
Web services	Open Web Services
eCat record	146091
CMI RESTful node ID	686
Access constraints	WO-STATS is available under CC-BY 4.0. The code for DEA-WO and its derivatives is available through GitHub under Apache 2.0 Open licensing.
Security classification	Unclassified
Update frequency	periodic

Access notes

DEA Maps

To view and access the data interactively:

1) Visit DEA Maps.

2) Click 'Explore map data'.

3) Select 'Inland water' > 'DEA Water Observations'.

4) Select which products you would like to display and click 'Add to the map'.

Technical information

As no confidence filtering is applied to this product, it is affected by noise where misclassifications have occurred in the input water classifications, and can be difficult to interpret on its own.

WO-STATS is available in multiple forms, depending on the length of time over which the statistics are calculated. At present the following are available:

DEA WO Multi-Year: statistics calculated from the full depth of time series (1986 to present) unfiltered
DEA WO Calendar Year: statistics calculated from each calendar year (1986 to present)
DEA WO November to March: statistics calculated yearly from November to March (1986 to present)
DEA WO April to October: statistics calculated yearly from April to October (1986 to present)

Accuracy and limitations

Please refer to the Landsat Surface Reflectance Product Description for the accuracy and limitations of the atmospheric, BRDF and topographic shading processing sequence. Please refer to Mueller et al. 2016 for details on the accuracy and limitations of Water Observations from Space (WOfS and WOfS-STATS).

WO-STATS provides a summary of water classification results from the WOfS product for all of Australia. As it cannot perfectly filter out misclassifications due to clouds, cloud shadows and issues to do with satellite sensor problems (such as the Landsat 7 SLC-Off failure), the summary also contains these misclassifications. In general misclassifications occur in the very low frequency observations and so can cause a misrepresentation of flooded areas. Misclassifications can also be caused by the presence of vegetation covering the water or within the water.

Relevant websites

DEA Water Observations in the DEA Website

References

Lineage

This product is created from the WOfS water classification (Water Observations 2 (Landsat)). Every pixel location is analysed statistically to derive the count of clear observations, the count of clear-wet observations and then to calculate the percentage of clear observations that were also wet. This provides a 'normalised' water frequency product for all of Australia.

Each product within the WO-STATS set is derived from the available Landsat observations within the respective period: calendar years; Apr-Oct each year; Nov-Mar each year; all-of-time (first available Landsat observation in the DEA archive to the most recent).

To create the confidence layer required for the Filtered product, a logistic regression is created between the un-filtered product and information about terrain, built-up areas, and coarse national water observations. In this way the confidence reflects the likelihood that the observed water is scientifically feasible at every pixel.

Data sources

Processing steps

Water Observation from Space (WOfS) - statistics

Major algorithms

Water Observations from Space Detection Algorithm

Schema / spatial extent

Geoscience Australia Landsat Collection 3

Update frequency	asNeeded
Temporal extent	1987-01-01 00:00:01
Coordinate reference system	Universal Transverse Mercator (variable)
Cell size X	30.00
Cell size Y	30.00

Example images

The image shows the percentage of time since 1987 that water was observed by the Landsat satellites on the floodplain around Burketown and Normanton in northern Queensland.

Owner

Commonwealth of Australia (Geoscience Australia)

Principal contributors

Norman Mueller, Adam Lewis, Dale Roberts

Subject matter experts

Norman Mueller

License

CC BY Attribution 4.0 International License

Export

Tags

Basics