Where the shadows fall

A pasture in Marengo County, Alabama. USGS 3DEP lidar flown in 2020, about 5 points per square meter, with a NAIP photo laid over the surface.

Every solar site near woods comes down to one trade: take trees down, or give up the ground their shadows cover. This page measures that trade. The lidar returns become a surface, the surface gets a real sun, and the woods throw their December 21 shadows across the pasture, the longest shadows a fixed layout will ever see. The colored ground is in shadow for a quarter or more of the 9 am to 3 pm window, from tree heights measured, not guessed.

On a project this is the map the layout and the clearing plan get negotiated over. The hard shade says which acres cost production if the trees stay. The crowns by height say which trees, and how many, buy that ground back if they come down. A landowner who wants the tree line kept and a developer who wants the rows kept argue over the same acres and the same trees. Trendline runs this on surveyor lidar for real sites; here it runs on public data over a field nobody is building on.

Trees

Trees--

Canopy--

Mean height--

Tallest--

Field

Measured hard shade--

Measured review--

Trees casting hard shade--

Hard shade is ground in shadow for half the window or more: leave it out of the layout or take down the trees that cast it. Review is a quarter to a half, the band where the trade gets argued row by row. Both come from the December sun run step by step over the woods as the lidar found them, not from a buffer around each tree.

Time of day

Dec 212:30 pm

Show

Measured shade Shade percent Crowns by height Field outline Lidar points

How it was served

The lidar was pulled from the USGS 3DEP Entwine index for the 2020 Alabama 17 County Lidar Project (AL_17Co_2_2020), which is hosted on public S3 as a cloud-optimized octree. That let us query only this window and retrieve about 8.3 million returns without downloading the full collection.

From there we processed the data offline into a set of static files: two 3 ft grids for the canopy surface and the bare ground, a NAIP imagery drape, crown and field polygons, the measured shade grids, and a reduced point cloud of 2.8 million points (68 MB) colored from the imagery.

Rather than relying on a database, tile server or backend to process and serve the data, everything is precomputed and hosted in object storage behind a CDN. The browser retrieves the files directly over HTTP, renders the grids as they arrive, and loads the reduced point cloud into the scene. This keeps the application simple, but it also means the point cloud is prepared in advance rather than selected from the full dataset based on the user's view.

The Flatirons page takes the other approach. Instead of reducing the point cloud to a smaller static file, it streams the full 74 million point COPC directly from object storage with HTTP range requests, keeping a coarse sample of the whole site loaded and retrieving the deeper octree levels only for the current camera view. Both approaches use static object storage with no dedicated point cloud server; the difference is whether the browser loads a smaller preprocessed dataset or requests portions of the original one as the view changes.

Loading the lidar surface

Crown height

10-20 ft20-30 ft30-40 ft40-50 ft50-60 ft60-70 ft70-80 ft80+ ft

Measured hard shade, half the window or more Measured review, a quarter to a half Open field