The new LiDAR dataset for Glacier Bay includes not only the “bare earth” digital terrain model but also the point cloud which can represent vegetation and other things the airplane-borne LiDAR bounced off first before it bounced off the ground. This “first returns” cloud can show the shape of the upper vegetation canopy and even distinct understory strata. I have been trying to determine if any useful information can be quantified from the point cloud and to use QGIS to make colorful 3D images of the canopy models.
It’s also possible to look at vertical profiles through the point cloud. These often show individual trees especially in the upper canopy. The shape of the tree canopy is revealed, but it is more difficult to distinguish cottonwood from spruce than I would have expected.
The canopy profiles maintain the precise height measurements of the LiDAR data so it is possible to measure tree height. I tried this in multiple profiles at each of my 10 study sites. I systematically selected sampling points along the transects and measured the height of five trees closest to each point.
I measured tree height along three transects at Sites 5-10 where the trees are mostly Sitka spruce and along two transects at Sites 1-4 where the trees are mostly cottonwood. I measured a total of 650 tree heights.
The tallest spruce are at Sandy Cove and York Creek (Sites 7 and 8). There are more trees at York Creek and they have bigger basal diameters, so the basal area (cross-sectional tree trunk area per unit area of forest) is 50% greater at York Creek compared to Sandy Cove. The two older sites have shorter spruce trees probably because the trees are much more densely packed, the soil is less well-drained, and alders never fixed much nitrogen there.
The tree height comparisons are useful because they follow a pattern with site age that is followed by other parameters — something we measure (tree height, biomass, carbon pools, nitrogen) increases along a sequence of progressively older sites until the oldest sites near the mouth of the bay where we measure less of it. This might be the actual time course at a site for some other parameters, but it is less likely that tree height will decrease with time. Average tree height can decline if the tallest trees die, but there is no indication that this has happened at these sites. Spruce bark beetles probably killed some trees at the older sites, but those trees were more likely the slower growing trees that had captured less of a place in the sunny overstory, not the tallest trees. So it makes sense that the decline with tree height along the chronosequence is not something that happened, or will happen, at any single site in its first three centuries. It is an incorrect inference from a flawed chronosequence.
Coloring the canopy models as a function of the intensity of the LiDAR returns allows some of the important plant species to be distinguished. Dryas drummondii returns a very strong signal and is faithfully distinguished at some sites (Figures 1 and 6). However, supratidal meadows also have high intensity LiDAR returns (Figures 7 and 8) and do not include Dryas. Alder tends to return a stronger signal than willow or cottonwood and can sometimes be distinguished from them.
Sites pictured above (Sites 1-4) do not yet have continuous tree canopies. Canopy gaps show remnants of the shrub thicket that is slowly deteriorating as trees overshadow it. Sites pictured below (Sites 5,7,8,10) have almost continuous canopies of Sitka spruce (there are a few cottonwoods at Site 5 and a few hemlocks at Site 10). The four figures below illustrate the variability in spruce density and canopy homogeneity. Site 10 (Figure 12) has more spruce per unit area and the trees appear to be distributed more evenly that at younger sites.
Quantifying tree heights from the LiDAR data seems to have promise. I did not make field measurements of tree height at my study plots very often, so even the simple tree height exercise described here provides useful information about differences among sites. The technique I used does not scale well, but skilled GIS practitioners can produce a continuous map of canopy height values. This can then be used to produce mean canopy heights at any scale for any subset of the landscape.
There are other parameters that could be measured or modeled, e.g., canopy texture, canopy tree density, or tree biomass. I don’t see a pressing need to pursue these metrics at my Glacier Bay sites.
The real power of this type of data might be realized in a decade or two when the data collection is repeated either with LiDAR or some new technology. Then the fate of individual trees could be followed and long-term changes in tree height, canopy width, tree density, or biomass can be documented. The timing might be good if I want to continue monitoring the study sites because by that time they won’t be letting me out of the home.