Introduction to DEA Water Observations Statistics 0bb04c1842da424aa9256da3d9d564e9

Background

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 1986) gives the information on where water is usually, and where it is rarely.

DEA Water Observations Statistics 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

Water Observations Statistics 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: ga_ls_wo_fq_myear_3: statistics calculated from the full depth of time series (1986 to 2025) unfiltered - DEA WO Calendar Year: ga_ls_wo_fq_cyear_3: statistics calculated from each calendar year (1986 to present) - DEA WO November to March: ga_ls_wo_fq_nov_mar_3: statistics calculated yearly from November to March (1986 to present) - DEA WO April to October: ga_ls_wo_fq_apr_oct_3: statistics calculated yearly from April to October (1986 to present)

Each dataset in this product suite 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)

When loading the data of any Water Observations Statistics product using the datacube Python library or when using DEA Maps, the central date of the observation period is returned. This date typically corresponds to:

  • 15 January for the November to March season

  • 16 July for the April to October season

  • 2 July for calendar year summaries

For example, the November to March 2020–2021 season is reported with a central date of 15 January 2021.

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.

For applications that require water observations from specific satellite images, see instead the Water Observations notebook.

Publications

Description

This notebook will demonstrate how to load and analyse Water Observations Statistics using DEA, including:

  1. Inspecting the products and measurements available in the datacube for Water Observations Statistics.

  2. Load calendar year Water Observations Statistics for an example location.

  3. Load the seasonal and multi-year Water Observations Statistics for the same location and compare them with the calendar year summary.

  4. Example application: Use the calendar year Water Observations Statistics to extract annual shorelines of Lake George, ACT.

Note: Visit the DEA Water Observations product documentation for detailed technical information including methods, quality, and data access.

Getting started

To run this analysis, run all the cells in the notebook, starting with the “Load packages” cell.

Load packages

Import Python packages that are used for the analysis.

[1]:

import datacube
import xarray as xr
import pandas as pd
import seaborn as sb
import contextily as ctx
import matplotlib.pyplot as plt

import sys
sys.path.insert(1, '../Tools/')
from dea_tools.spatial import subpixel_contours

Connect to the datacube

Connect to the datacube so we can access DEA data.

[2]:

dc = datacube.Datacube(app='DEA_Water_Observations_Statistics')

Available products and measurements

List products available in DEA

We can use datacube’s list_products functionality to inspect Water Observations Statistics products that are available in DEA. The table below shows the product name that we will use to load data, and a brief description of the product.

[3]:

# List Water Observations Statistics products available in DEA
dc_products = dc.list_products()
dc_products[dc_products["name"].str.contains("ga_ls_wo_fq")]

[3]:
name description license default_crs default_resolution
name
ga_ls_wo_fq_apr_oct_3 ga_ls_wo_fq_apr_oct_3 Geoscience Australia Landsat Water Observation... CC-BY-4.0 EPSG:3577 (-30, 30)
ga_ls_wo_fq_cyear_3 ga_ls_wo_fq_cyear_3 Geoscience Australia Landsat Water Observation... CC-BY-4.0 EPSG:3577 (-30, 30)
ga_ls_wo_fq_myear_3 ga_ls_wo_fq_myear_3 Geoscience Australia Landsat Water Observation... CC-BY-4.0 EPSG:3577 (-30, 30)
ga_ls_wo_fq_nov_mar_3 ga_ls_wo_fq_nov_mar_3 Geoscience Australia Landsat Water Observation... CC-BY-4.0 EPSG:3577 (-30, 30)

List measurements

We can inspect the contents of any of the listed Water Observations Statistics products using datacube’s list_measurements functionality. The table also provides information about the measurement data types, units, nodata value and other technical information about each measurement.

[4]:

dc_measurements = dc.list_measurements()
dc_measurements.loc[['ga_ls_wo_fq_cyear_3']]

[4]:
name dtype units nodata aliases flags_definition
product measurement
ga_ls_wo_fq_cyear_3 count_wet count_wet int16 1 -999 NaN NaN
count_clear count_clear int16 1 -999 NaN NaN
frequency frequency float32 1 NaN NaN NaN

Loading data

Now that we know what products and measurements are available, we can load Water Observations Statistics data for an example location. We will first load an annual summary, then we’ll load seasonal and all-time summaries to compare with.

Load a calendar year Water Observations Statistics

[5]:

# Set up a region to load data
lat, lon = -24.744, 139.6
buffer = 0.3
time = "2023"

lat_range = (lat - buffer, lat + buffer)
lon_range = (lon - buffer, lon + buffer)

# Load DEA Water Observations Stats
wo_cyear = dc.load(
    product="ga_ls_wo_fq_cyear_3",
    x=lon_range,
    y=lat_range,
    time=time,
    output_crs="EPSG:3577",
    resolution=(-30, 30),
)

We can now view the data that we loaded. The measurements listed under Data variables should match the measurements displayed in the previous List measurements step.

[6]:

wo_cyear

[6]:

<xarray.Dataset> Size: 40MB
Dimensions:      (time: 1, y: 2356, x: 2131)
Coordinates:
  * time         (time) datetime64[ns] 8B 2023-07-02T11:59:59.999999
  * y            (y) float64 19kB -2.661e+06 -2.661e+06 ... -2.732e+06
  * x            (x) float64 17kB 7.278e+05 7.279e+05 ... 7.917e+05 7.917e+05
    spatial_ref  int32 4B 3577
Data variables:
    count_wet    (time, y, x) int16 10MB 0 0 0 0 0 0 0 0 0 ... 0 0 0 0 0 0 0 0 0
    count_clear  (time, y, x) int16 10MB 74 74 74 74 74 75 ... 36 36 36 36 36 36
    frequency    (time, y, x) float32 20MB 0.0 0.0 0.0 0.0 ... 0.0 0.0 0.0 0.0
Attributes:
    crs:           EPSG:3577
    grid_mapping:  spatial_ref

Plotting calendar year Water Observations Statistics frequency and clear count

On the left in the figure below we show the frequency with which water was observed over the calendar year by Landsat for this region in central Australia. Note that frequency = water count / clear count, and in the plot below we have masked out the zero frequency values. On the right of the figure, the number of clear observations is also shown.

On the left we include a basemap image of the region. A basemap is simply background imagery used to provide visual context, it helps orient the viewer but isn’t analytical data itself.

[7]:

fig, ax = plt.subplots(1, 2, figsize=(10, 4), layout="constrained", sharey=True)
wo_cyear["frequency"].where(wo_cyear["frequency"] > 0).plot(
    ax=ax[0],
    cmap=sb.color_palette("mako_r", as_cmap=True),
    vmin=0.1,
    vmax=0.9,
    add_labels=False,
)
ctx.add_basemap(
    ax=ax[0],
    crs=wo_cyear.odc.crs,
    source=ctx.providers.Esri.WorldImagery,
    attribution="",
    attribution_size=1,
)
wo_cyear["count_clear"].plot(ax=ax[1], cmap="magma", add_labels=False, vmin=10, vmax=70)

ax[0].set_xticklabels(labels="")
ax[0].set_yticklabels(labels="")
ax[1].set_xticklabels(labels="")
ax[1].set_yticklabels(labels="")
ax[0].set_title("WOs annual frequency for 2023")
ax[1].set_title("WOs annual clear count for 2023");
../../../_images/notebooks_DEA_products_DEA_Water_Observations_Statistics_18_0.png

Load and plot the other summary periods

Now we will load all the other Water Observations Statistics products, and plot these in the subsequent cell to see how they differ from the 2023 calendar year summary we plotted above.

[8]:

# Load Nov-Mar summaries from 2023
wo_nov_mar = dc.load(product="ga_ls_wo_fq_nov_mar_3", like=wo_cyear, time="2023-01")

# Load apr-oct summaries from 2023
wo_apr_oct = dc.load(product="ga_ls_wo_fq_apr_oct_3", like=wo_cyear, time="2023-07")

# Load the multi-year summary (1986-2025)
wo_myear = dc.load(product="ga_ls_wo_fq_myear_3", like=wo_cyear, time=time)

Plot the seasonal and multi-year summaries

The seasonal frequencies help us interpret what we saw above in the annual summary, while the 1986–2025 summary helps us understand what is typical for this location.

The multi-year Water Observations Statistics frequency (1986–2025; right) is predominantly low (pale cyan), showing that in this arid landscape surface water occurs only intermittently, with moderate water persistence limited to deeper water-holes and main channels. In contrast, 2023 exhibits markedly higher frequencies, especially in Apr–Oct 2023 (centre), indicating sustained inundation driven by higher rainfall and inflows from the upstream catchment following the 2022–23 wet season.

[9]:

fig, axes = plt.subplots(1, 3, figsize=(16, 5), layout="constrained", sharey=True)

dss = [wo_nov_mar, wo_apr_oct, wo_myear]
titles = ["Nov-Mar 2022-23", "Apr-Oct 2023", "All-time (1986-2025)"]

for ax, ds, t in zip(axes.ravel(), dss, titles):
    im = (
        ds["frequency"]
        .where(ds["frequency"] > 0.01)  # help filter out noise
        .plot(
            ax=ax,
            cmap=sb.color_palette("mako_r", as_cmap=True),
            vmin=0.1,
            vmax=0.9,
            add_labels=False,
            add_colorbar=False,
        )
    )
    ctx.add_basemap(
        ax=ax,
        crs=wo_cyear.odc.crs,
        source=ctx.providers.Esri.WorldImagery,
        attribution="",
        attribution_size=1,
    )
    ax.set_xticklabels(labels="")
    ax.set_yticklabels(labels="")
    ax.set_title(f"WOs frequency: {t}")

# Add colorbar
cbar = plt.colorbar(im, ax=axes[2]);
../../../_images/notebooks_DEA_products_DEA_Water_Observations_Statistics_22_0.png

Example Application: Delineate a time series of lake shorelines

In this example, we will use the calendar year Water Observations Statistics product, ga_ls_wo_fq_cyear_3, to extract annual shorelines of Lake George, ACT, from 2013 to 2024. This will show us how the lake’s shorelines have changed over this period in response to water availability (predominantly rainfall).

[10]:

# Extent covers Lake George, ACT
lat, lon = -35.0989, 149.4252
buffer = 0.095
time = ("2013", "2024")

lat_range = (lat - buffer, lat + buffer)
lon_range = (lon - buffer, lon + buffer)

Load calendar year Water Observations Statistics and geomedian data

Note we are loading a single year of the DEA Geometric Median (Landsat) product so we can plot our lake extents over a nice image of the region.

[11]:

# Load calendar year Water Observations Statistics
wo_cyears = dc.load(
    product="ga_ls_wo_fq_cyear_3",
    x=lon_range,
    y=lat_range,
    time=time,
    output_crs="EPSG:3577",
    resolution=(-30, 30),
)

# Load Landsat 8/9 geomedian product for context
ds = dc.load(
    product="ga_ls8cls9c_gm_cyear_3",
    measurements=["nbart_red", "nbart_green", "nbart_blue"],
    like=wo_cyears,
    time="2024",
).squeeze()

Find the outlines of the permanent water bodies

Here we set a threshold of 0.90, meaning if the Water Observations Statistics product observed water in a given pixel for more than 90 % of the year, then this pixel will be included in the annual shoreline extent of Lake George.

We will use the dea_tools.spatial.subpixel_contours function to extract the boundaries of Lake George from the image data. You can find out more about this function in the Extracting contours notebook.

[12]:

# Convert to a binary water/not-water image
threshold = 0.90
annual_water = xr.where(wo_cyears["frequency"] >= threshold, 1, 0)

# Extract contours into a geodataframe
water_bodies = subpixel_contours(annual_water, min_vertices=25, verbose=False)

# Add a simple 'year' column to help with the plotting
water_bodies["year"] = [i[0:4] for i in water_bodies.time]

Plot the annual series of shorelines for Lake George

We also plot the 2024 Landsat geomedian beneath the shoreline contours to provide contextual background.

From the image below, it’s clear that the extent of Lake George has changed considerably over the past decade. In the most recent years (2022–2024), the lake has remained relatively full, whereas during the earlier part of the time series (2013–2019), it held significantly less water. The shifting shoreline positions reflect these variations in water level.

[13]:

fig, ax = plt.subplots(1, 1, figsize=(8, 7), layout="constrained")
ds[["nbart_red", "nbart_green", "nbart_blue"]].to_array().plot.imshow(
    robust=True, ax=ax, add_labels=False
)
water_bodies.plot(column="year", legend=True, cmap="viridis", ax=ax)
ax.set_xticklabels(labels="")
ax.set_yticklabels(labels="")
ax.set_title(f"Annual shorelines of Lake George");
../../../_images/notebooks_DEA_products_DEA_Water_Observations_Statistics_30_0.png

Additional information

License: The code in this notebook is licensed under the Apache License, Version 2.0. Digital Earth Australia data is licensed under the Creative Commons by Attribution 4.0 license.

Contact: If you need assistance, please post a question on the Open Data Cube Discord chat or on the GIS Stack Exchange using the open-data-cube tag (you can view previously asked questions here). If you would like to report an issue with this notebook, you can file one on GitHub.

Last modified: February 2026

Compatible datacube version:

[14]:

print(datacube.__version__)

1.8.19

Tags

Tags: NCI compatible, sandbox compatible, DEA products, ga_ls_wo_fq_cyear_3, water observations,:index:water observations, :index:subpixel contours`, water observations statistics, ga_ls_wo_fq_myear_3, ga_ls_wo_fq_apr_oct_3,:index:ga_ls_wo_fq_nov_mar_3,