Raster Mosaic and Time Series Support

Earthscale provides native support for tiled, multi-file raster datasets. Some examples of these dataset types:

• Each file has a different projection (e.g., UTM zone)

• Each file represents a separate band

• Each file has a datetime encoded in the filename

By "native support", we mean that we can take a collection of these files and represent them as a single dataset with multiple bands, projections, and time steps. In our web app, you can then configure multi-band visualizations or toggle through time. And you can use our export functionality to download slices of this data in a single projection.

ImageDataset Configuration

For multi-file GeoTIFF datasets (using wildcard patterns like gs://bucket/path/*.tif), the following configuration options are available:

Filename Date Pattern

Extract timestamps from your filenames to enable temporal queries on your dataset. Use Python strptime format codes.

Examples:

Filename Example	Pattern
data_2024-01-15.tif	%Y-%m-%d
tile_20240115_v2.tif	%Y%m%d
2024/01/15/image.tif	%Y/%m/%d/
sentinel_2024-01-15T10:30:00.tif	%Y-%m-%dT%H:%M:%S

Common format codes:

Code	Meaning	Example
%Y	4-digit year	2024
%m	2-digit month	01-12
%d	2-digit day	01-31
%H	Hour (24-hour)	00-23
%M	Minute	00-59
%S	Second	00-59
%j	Day of year	001-366

Filename Band Patterns

Map filename patterns to band names when your bands are stored in separate files.

This is useful when you have files like:

• tile_B02.tif, tile_B03.tif, tile_B04.tif

• scene_red.tif, scene_green.tif, scene_blue.tif

Configuration format:

For each pattern, specify:

1. Pattern: A wildcard pattern including the substring that appears in filenames for that band

2. Band Name: The name to assign to that band

Examples:

Filename Pattern	Band Name
*_B02*	blue
*_B03*	green
*_B04*	red
*_B08*	nir

Or for descriptive filenames:

Filename Pattern	Band Name
*_red*	red
*_green*	green
*_blue*	blue

This can optionally be combined with the filename date pattern argument.