Skip to content

Added information on resampling algorithms for overviews #509

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
wants to merge 12 commits into
base: main
Choose a base branch
from
Open

Added information on resampling algorithms for overviews #509

Show file tree
Hide file tree
Changes from 11 commits
Commits
Show all changes
12 commits
Select commit Hold shift + click to select a range
e708aab
added-resampling-algo-LC-mosaics
GL-S Sep 29, 2025
b2c50d1
added-details-overview-algos
GL-S Sep 29, 2025
c7908f0
small-formatting
GL-S Sep 29, 2025
c7bfa0e
grammar
GL-S Sep 29, 2025
0561e04
note-format
GL-S Sep 29, 2025
75ea35c
fixed-animation-out-dropdown
GL-S Sep 29, 2025
7a9ddc9
Update _access.md
GL-S Sep 29, 2025
6bba2dd
Update _access.md
GL-S Sep 29, 2025
e591123
Update _access.md
GL-S Sep 29, 2025
3ddbc61
edit-after-robbi-suggestion
GL-S Sep 30, 2025
aacb877
removed-unsure-detail
GL-S Sep 30, 2025
9055c54
fixed-indentation
GL-S Oct 1, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
7 changes: 7 additions & 0 deletions docs/data/product/dea-land-cover-landsat/_access.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -33,6 +33,13 @@ For detailed instructions, please visit the [Continental Cloud-Optimised GeoTIFF
<figcaption>The animation above shows the DEA Land Cover COG mosaic moving from the continental-scale down to the 30m x 30m pixel level.</figcaption>
</figure>

:::{admonition} Note
:class: note

The overviews (pyramid layers) embedded in the DEA Land Cover COG mosaics were generated using the **MODE** resampling algorithm. This means that, when viewing the data at coarser zoom levels, each overview pixel is assigned the most common land cover class within its corresponding area.
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@GL-S This text shouldnt be indented. Looks good though.


:::

:::{dropdown} How to integrate DEA Land Cover continental mosaics into your own Python workflow

You can seamlessly open a Land Cover mosaic, such as Level 4 for year 2024, using Python and the `rioxarray` library. For example:
Expand Down
12 changes: 9 additions & 3 deletions docs/guides/continental-cogs-geotiff-mosaics.rst
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -7,11 +7,17 @@ Some Digital Earth Australia (DEA) products are provided as **continental-scale

These datasets are made available as **Cloud-Optimised GeoTIFFs (COGs)**, a format that enables users to efficiently *stream* raster data directly from the cloud without downloading the files. This provides a fast and convenient way to access full-continental coverage from tools like QGIS or ArcGIS Pro, especially when working with large datasets.

COGs include overviews, which are lower-resolution pyramid layers that allow fast rendering at different zoom levels. When zooming out, GIS software displays a coarser overview to improve performance. These layers are generated using a **resampling algorithm**, which determines how pixel values are aggregated when reducing spatial resolution. For DEA products, the choice of resampling algorithm depends on the nature of the data:

- **MODE** – used for **categorical data**; assigns the most frequent value within the extent of an overview pixel.
- **BILINEAR** – used for **continuous data** with minimal missing values; interpolates pixels to produce a smooth gradient.
- **NEAREST** – used for **continuous or categorical data with many missing values**; each overview pixel takes the value of the higher-resolution pixel closest to its centroid. This algorithm appears to maximise the visualisation of products where valid pixels are surrounded by many no-data pixels.
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@GL-S Could "appears to maximise" be replaced by "optimises" ?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I think I need assistance from @robbibt or @vnewey here. Because the source I have found say that NEAREST uses the nearest pixels, valid or invalid. So I am not sure it always optimises/maximises the visualisation, or if just happens to work for those products we applied it to.

Copy link
Member

@robbibt robbibt Oct 1, 2025 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I'll do some quick testing today, but essentially "nearest" was the only method that wouldn't come out empty if an overview pixel overlapped with a high-res single nodata pixel. So using "nearest" was less about maximising data coverage, but more to ensure our overviews contained at least some data.

I think we can probably go with

- **NEAREST** – used for **continuous or categorical data with many missing values**; each overview pixel takes the value of the higher-resolution pixel closest to its centroid. This algorithm enhances the visualisation of products where valid pixels are surrounded by many no-data pixels.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yeah ok, I just wanted to double check this. thanks for confirming!
If you observed it working better to show data, then the statement you suggest is correct. I think it's ok to just describe the effect of applying the algorithm without including too many details on how the algorithm works, considering there is a bit of confusion around it


Not all DEA products provide continental-scale COGs. Currently, this access method is only available for selected products:

- `DEA Land Cover </data/product/dea-land-cover-landsat/>`_
- `DEA Intertidal </data/product/dea-intertidal/>`_
- `DEA Tidal Composites </data/product/dea-tidal-composites/>`_
- `DEA Land Cover </data/product/dea-land-cover-landsat/>`_ - uses `MODE` as the overview resampling algorithm
- `DEA Intertidal </data/product/dea-intertidal/>`_ - uses `NEAREST` as the overview resampling algorithm
- `DEA Tidal Composites </data/product/dea-tidal-composites/>`_ - uses `NEAREST` as the overview resampling algorithm

**VRT (Virtual Raster) files** are provided alongside the ``.tif`` mosaics. These files serve as lightweight wrappers around the main data and can be used to open data in GIS software with visual settings already applied. We use VRTs to provide:

Expand Down