Climate change is causing growing concern for extreme weather events and increasing damage to infrastructure and the environment. Trees are a major factor in climate regulation because they absorb CO2 and water and expel oxygen. The loss of trees directly affects the livability quality, economic equity, and physical stability of the environment. Trees are a requirement for a better quality of life for everyone, both economically and environmentally.

Is Tree Cover Changing in King County?

To know where tree cover is changing in the county, an inventory must first be made to determine a baseline. Because there are millions of trees in the county, a manual inventory of every tree would be cost-prohibitive. The KCGIS Center has produced county-wide bare earth Digital Ground Models and Digital Surface Models from LiDAR, and a vector layer of impervious surface. Impervious surfaces include buildings, roads, parking lots, and other paved or compacted ground. Combining these layers in GIS software, a model of tree canopies and crowns can be extracted far more quickly and efficiently than by manual inventory. Although these models are not as accurate as doing a ground inventory, it is a good estimation of what is in the field at a fraction of the cost.

The first step is subtracting the ground-bare earth LiDAR model from the surface model to produce an actual height model of objects on the ground. The impervious layer is used to separate vegetation from manmade objects from the height model to produce two raster layers: Vegetation Height (VHT) and Building Height (BHT).

In 2016, the first tree canopy layer (reg_tree_2016) was produced using 2016 LiDAR data & the 2015 Impervious layer for the western urban area of the county. County experts defined a tree as 10 feet or higher. An extraction of 10’ or higher from the VHT provided the tree cover model. By classifying LiDAR intensity data, tree-type conifer and deciduous could be separated by the level of intensity reflectivity strength. Conifers reflect LiDAR returns stronger than deciduous trees. Filters were used to remove noise and generalize the raster data to better define tree crown boundaries by type.

The 2021 Tree Canopy layer was produced using 2019 impervious to extract vegetation from 2021 Lidar data (reg_tree_2021). The process was essentially the same as in 2016 except the new LiDAR data has an improved resolution from 3 foot to 1½-foot cells or twice the resolution. The results were slightly different from the 2016 tree layer due to the lower resolution of the 3’ 2016 LiDAR and higher noise in the 2021 intensity data. Users who tried to compare the 2016 and 2021 tree canopy layers had difficulty due to the difference in quality and resolution. A major factor was the quality of impervious layers that were used. The 2019 impervious used from the 2021 canopy is a vector layer produced by Ecopia vs the 2015 impervious layer extracted from a 2-foot cell 4-band ortho via Normalized Difference Vegetation Index (NDVI) value for the 2016 canopy.

A New Way to Derive a Tree Canopy Layer

This led to the search for a new way to derive a tree canopy layer that would delineate individual tree crowns. Instead of filtering lidar intensity data for tree canopy, a different method uses watershed tools to delineate tree crowns with the ability to produce polygon (vector data) outlines of individual trees.

Instead of extracting from the VHT and classing trees by LiDAR intensity as was done for 2016, A method for each tree to be inverted by multiplying the actual height surface values by -1 turns tree bumps on the surface into holes in the earth. Using watershed tools, the outline of each tree can be delineated as a crown by processing the holes as mini watersheds.

The draining cell or pour point at the bottom of each hole is found using Flow Direction. An 8-direction value is assigned to each cell as to which direction it flows to. The cell at the bottom of the tree hole that has nowhere to flow receives a null direction value. This is selected and converted to a point layer that will be the pour point for the tree watershed process.  The watershed pour point is also where the trunk would be. The watershed command delineates the area of all cells that flow to each pour point. This is the outline or crown of the tree. The pour point is used to extract height value from the Vegetation Height (VHT) model, Intensity from LiDAR, and NDVI from 4-band orthophotography. Intensity can delineate deciduous from conifer trees by the strength of return and tree health can be determined from NDVI.

The watershed process is more accurate for conifer than deciduous due to tree structure and the limits of the software. Conifers tend to be conical with one point while large deciduous trees can have multiple branches spreading out with multiple point crowns. The software is limited to what is called 2 ½ Dimensions, not true 3D. The software interprets a tree as a bump or hill on flat ground with no undercuts or overhangs. Spatial Analyst cannot tell if multiple crowns are for one tree or many trees close together because it cannot process the underside of branches. Filtering was done to smooth out irregularities while maintaining the fidelity of the data and avoiding loss of detail and smaller trees.

Although the quality of results may vary depending on conditions, it’s a good start for getting a county-wide inventory. It is estimated that there are over 101,001,000 trees in King County as of 2021, and this is probably an underestimate. Many small trees go undetected overshadowed by larger trees or lost in processing.

There will be anomalies due to the different vintages of impervious vs lidar data. Trees are being lost every day and many cases replaced with impervious surfaces. Habitat is destroyed, less carbon is absorbed, surface water does not permeate the soil, runs off and causes floods and the livability of the environment is degraded.

The City of Seattle’s most recent canopy cover study, using data from 2021, found the city lost 255 acres of tree canopy 1.7% relative decline, a .5% absolute decrease. The City of Seattle study can be found here.

How do you get the data?

Go to King County GIS Open Data -> scroll and click “Natural Resources” -> look for a specific township/range (the data is huge so we had to break it up into township/ranges).  There are a total of 79 map services (i.e., “T19R06 Tree Points 2021”)

2021 3-Band Orthophoto, Red, Green, and Blue create a “true color” image.

Building Height (BHT) & Vegetation Height Model (VHT) Hillshades

Oblique view of Vegetation Height 2 ½ Dimension Model.

Reverse Vegetation Height by multiplying elevation by -1, Black = low values, White = high values

Oblique view of Reverse Vegetation Height Tree Holes

                                                                                                                   

Tree Crown Watershed Delineations from Pour Points

Watershed Raster Model built from Tree points Height field in Feet,   Blue = Low values, Orange = High value.    

LiDAR Intensity with Pour Points

Tree Points extract data from LiDAR intensity to symbolize raster tree type. LighterGray = Conifer, Darkergray = Decidious.

Watershed Raster Model built from Tree point LiDAR Intensity data. Values Classed 0-120 = Deciduous, Light Green. 120-200 = Conifer, Dark Green. 200-255 = Other, Red

2021 4-Band Near Infra Red-National Agriculture Inventory Program (NAIP) image.

Tree Crown Watershed Raster to Polygons.

LAS Point Cloud Delineated by Tree Crown polygons and extruded tree points by tree height.