Imagery on the map

After an orthomosaic is uploaded we check it and, unless it already is one, convert it to a cloud-optimized GeoTIFF, so the map can read just the tiles it needs. Big files take a while to convert; the file's row shows the stage. Then the map shows it in true colour, above the basemap and under the agent's highlights.

A transparency band is recognised even when the export didn't label it (its values are only fully clear or fully opaque), so it hides the empty margins instead of being mistaken for a colour band.

How bands are recognised

To draw an index we need to know which band is red, green, blue, near-infrared or red edge. We look, in order, at:

the file's own band descriptions, colour tags and wavelengths;
the camera named in its tags (DJI Mavic 3 Multispectral and P4 Multispectral, MicaSense Altum, RedEdge and RedEdge-P);
three data bands: read as red, green and blue;
otherwise, a comparison of the bands' brightness, with thermal and empty bands left out.

When the guess is wrong, tell the agent ("band 4 is NIR") and it uses your bands for that layer. Band numbers are always counted as in the file you uploaded.

Indices

Index	Name	Needs
VARI	Visible atmospherically resistant index	Red, green, blue
ExG	Excess green	Red, green, blue
GLI	Green leaf index	Red, green, blue
NGRDI	Normalized green-red difference	Red, green
NDVI	Normalized difference vegetation index	Red, NIR
SAVI	Soil-adjusted vegetation index	Red, NIR
EVI	Enhanced vegetation index	Red, NIR, blue
NDRE	Normalized difference red edge	NIR, red edge

The index picker shows every index; the ones your file can't support are dimmed, with the band they're missing. An RGB orthomosaic has the first four. Ask for an index by name ("show NDVI") or pick it.

Raw values

SAVI and EVI have constants that assume surface reflectance (values from 0 to 1). On imagery in raw digital numbers they still show where vegetation is stronger or weaker, but the values only compare within that file.

Reading an index layer

Colours run from blue for the lowest values to green for the highest. The stretch spans the 2nd to the 98th percentile of the file, so a few extreme pixels don't wash out the rest; the legend shows the values at each end.
Ask for a measurement over an area ("how much of the north block has NDVI below 0.2?") and the agent computes it from the pixels: the mean, spread and percentiles, plus the share of the area that is low, moderate or high.
The low, moderate and high bands use general thresholds (for NDVI, below 0.2 and above 0.5). They're a quick read, not an agronomic diagnosis.

Downloading

The download button on a file's row gives you the file exactly as uploaded, through a link that works for 15 minutes.