A long history of human interest in forests includes the estimation and cataloging of tall tree and canopy heights.

Thanks to this longstanding relationship, we are aware of the heights and grandeur of countless giants having the fortune (or misfortune) of living during the Anthropocene.

Superintelligent admirers co-existing with never-logged(?) and protected mature coastal redwoods reaching upwards of 90m (300ft)

The ‘Chicago Stump,’ a remnant of the General Noble Tree, thousands of years in age, 91 m (300ft) tall with a ground circumference of 29m (95ft) before being lopped down for timber in 1892

An ancient coast redwood tree, alive and well at over 1000 years old until toppled by wind in 1905 after its brethren (and shelter) had been logged in 1897. 99m (325ft) tall, diameter 5.8m (19ft) near its base.

Vast collections of writings, resources, records and surveys further detail the living forests we enjoy today, including beautiful web articles showcasing living giants of the Pacific Northwest, a.k.a. ‘Cascadia.’

Now our enjoyment of tall tree lore can benefit widespread aerial LiDAR surveys, which conveniently provide high density, precise height measurements of the ground and canopy.

LiDAR: Public Data

Light Detection and Ranging (LiDAR) is a bright star in the constellation of modern environmental measuring campaigns collectively referred to as remote sensing.

For those of us not professionally engaged in approaching these data with institution-scale talent and computing resources, a bit of home baking can be done to personally explore this laser-based view of the landscape.

The Washington State DNR LiDAR Portal provides impressively comprehensive data downloads from various aerial LiDAR surveys in various forms: DSM for a ‘surface’ model that includes canopy height, DTM for the model of underlying terrain (solid ground), hillshades for convenient pseudo-3D visualizations of these models, and also the multidimensional LiDAR point cloud data itself.

Processing

For personal exploration on a home computer, reasonably-sized areas of special interest can be selected to for processing, working with ~100MB to 1GB blocks at a time (or in parallel).

Height above ground (HAG)

Using the point cloud data directly, it is possible to compute height above ground (HAG), a metric convenient for cataloging and visual rendering of putative tree heights:

#!/bin/sh
pdal translate $1 `basename $1 .laz`-HAG.laz filters.hag_delaunay

A grove of (likely) Douglas-fir trees on a lesser-visited bank of the Nooksack River, with treetop LiDAR points as high as ~90 metres (300ft) above ground. In modern datasets, resolution is sufficient to appear pleasantly as height-shaded trees in 3D (e.g. in QGIS)

Height above ground (HAG) can be rendered alone, subtracting out the ‘ground’ variation for easier tree-to-tree height comparisions on steep slopes–this is the same grove as shown above.

Annotation

A catalog of trees and their heights can also be tabulated directly from these point cloud data, for example using the whitebox_tools function ‘IndividualTreeDetection.’

GeoTIFF difference mapping

A more expedient way to catalog putative tree heights over a larger area and/or smaller data files is to download the DTM and DSM models for an area (these are GeoTIFF files), and perform GeoTIFF subtraction to yield a crude ‘canopy height model.’

gdal_calc -A [DSM GeoTIFF] -B [DTM GeoTIFF] --calc "A-B" --outfile DSM-DTM-diff.tiff

This produces a GeoTIFF-formatted representation of usable canopy heights (when performed on a forested area) that can be loaded and styled in QGIS, and perused for tall trees and their heights (found in the ‘value’ panel).

This DSM-DTM difference map can also be peak-picked for tall trees using home-baked (or ‘AI’-informed) python scripts to generate vector-based features and tables for individual treetops, that be filtered and labeled by height.

Areas of what are likely remarkably tall trees just upriver from Goldmyer Hot Springs, annotated by peak-picking a (DSM)-(DTM) GeoTIFF ‘canopy height’ layer.

Putative trees over 85m (279ft) annotated on a DSM-DTM difference map south of Lake Quinault.

Some of the tallest trees considered to be within the limits of Bellingham are Douglas-firs well above 70 metres according to public LiDAR data

Community projects

I was a bit surprised to find that this kind of tree cataloging isn’t yet included state-wide as a derived product alongside the public data.

There are a few community efforts to bring this new knowledge to the public, such as this LiDAR cataloging effort by the Whatcom Million Trees Project, and some work in Seattle like this Tree Canopy dataset.

While much hype promises ‘AI’-infused approaches for more powerful identification and classification in the near future, it is nice to know that we can access vast datasets that we paid for (as science-appreciating taxpayers) for independent perusals on a single modern laptop.

Now at least, if a tree falls in the woods, the next LiDAR survey might bear witness to it.

CONTENTS

LiDAR literacy: trees

Basic scouting for tall trees in public data