Additional Hill-shade Options

[JBLINZ]

User Story

As a map maker/owner, I would like to have multiple options when choosing the hill-shade/relief-shade type that MapLibre renders using my terrain-RGB tiles so that I can deliver high-quality cartographic products.

Rationale

Currently, only a standard hill-shade is available which is a great general hillshade.

By introducing three additional algorithms for rendering hill-shades from terrain RGB tiles, MapLibre will be increasingly useful for advanced cartography. GDAL currently allows 'Combined', 'Multidirectional' and 'Igor', among others. Choosing from these three options when using terrain RGB tiles would allow map makers to select the right theme and feel for their project.

Further explanation of hill-shade types:

• Combined combined shading, a combination of slope and oblique shading.
• Multidirectional multidirectional shading, a combination of hillshading illuminated from 225 deg, 270 deg, 315 deg, and 360 deg azimuth. Multidirectional hill-shading applies the formula of http://pubs.usgs.gov/of/1992/of92-422/of92-422.pdf.
• Igor shading, which tries to minimize effects on other map features beneath. Igor's hill-shading uses a formula from Maperitive http://maperitive.net/docs/Commands/GenerateReliefImageIgor.html.

GDAL’s alternate hillshade algorithms — from top to bottom: multidirectional, combined, and igor. Images derived from the USGS National Map.

[NathanMOlson]

@HarelM please assign this to me. LINZ is funding me to complete this work.

[NathanMOlson]

See my design proposal for the style spec here: maplibre/maplibre-style-spec#1066

[NathanMOlson]

I've got an initial implementation working with all methods except multidirectional. You can check out a demo here: https://nathanmolson.github.io/hillshade and code here: NathanMOlson#4

The existing MapLibre hillshade method is different from GDAL's methods in that MapLibre has both a shadow color and an highlight color which influence the hillshade output, whereas GDAL outputs a monochromatic intensity value from 0-255. MapLibre additionally has an "accent" color, which gets blended in based on the terrain slope, regardless of aspect. (It "accents" steep terrain, on both the sunny side and the shadowed side). I'm not sure how closely to follow GDAL's 1D algorithms in the presence of 3 MapLibre color parameters.

So far, the approach I've take is to use MapLibre's shadow and highlight colors. When GDAL would output 0, we use the shadow color. When GDAL would output 128, we use transparent. When GDAL would output 255, we use the highlight color. So far I've left the accent color out of all the new methods, but it could be easily added in. The benefit of adding it would be more styling flexibility. It could always be removed by setting the accent color transparent. The drawback of using it is that more style modification (from default) is required to replicate GDAL's output closely.

Accent color is not the only parameter being ignored, both standard and igor methods ignore hillshade-illumination-altitude.

I'd love to hear any thoughts on the tradeoff between closely replicating GDAL vs. using the additional styling flexibility that is present in the existing method.

[HarelM]

Is there a hard requirement to replicate gdal output? I would expect the requirement to be just a "it should allow similar behavior/configuration"...