Forest Patterns

remote sensing of the biogeography of forest structure.

[pahbs]

In Tura, our gear is waiting in an old trailer. We have, probably, tons of gear including 1 generator, 7 tents, 3 outboard motors, 2 chainsaws, 4 boats, food for 8 for 2 weeks, beer and vodka. This gear supports the expedition. It is not light.

[pahbs]

On the way to Tura. The best airport food I've ever had is here in Krasnoyarsk. Smoked Arctic Char.

[pahbs]

Toward the Arctic forests Tomorrow we'll head to the upper Kotuy River in the eastern portion of the Putorana Plateau. The monospecific Larch forests we'll be measuring are unique because they grow atop continuous permafrost, their most influential characteristic. This characteristic extends across central and eastern Siberia. This permafrost - forest association is not found in Alaska or Canada, making for an arctic system the functions much differently than its North American counterpart.

[pahbs]

After 24 hrs of flying, which somehow amazingly included 2 red eyes, we’ve arrived safe and sound at the House of Scientists here in Krasnoyarsk. Surprisingly, they have wifi. I think I’ll do a couple posts prior to leaving for the Arctic on Monday morning.

This place is exactly as I remember. Like a bunker, and completely empty of guests other than ourselves. The best adjective I can come up with now is “idiosyncratic”. I have a few initial observations:

The Soviet era TV has a screen the size of my laptop, yet the room features a fridge that is far larger than any guest would likely ever need.

My bed is slightly larger than my toddler’s bed at home (larger not in width, just in length. Maybe.) There are two in my room. Both very comfortable.

The best word to describe the towels is “abrasive”.

First attempts to produce hot water in the shower were unsuccessful.

There seems to be a large outbreak of ants, some of whom have wings, underneath the bathroom sink.

Great view of the Yenisei River nearby.

[pahbs]

The distribution of Larch forests across Siberia. We are headed to the L. gmelinii region.. we’ll be doing field work just under the “g” in gmelinii on the map above, by Abaimov et al. They indicate that “L. gmelinii habitat almost completely coincides with the zone of continuous and interrupted distribution of the permafrost.”

In other news, sadly, updates from the field will not be possible because of Russian government restrictions on communication devices…

[pahbs]

The southern-most bend of the upper Kotuy River in central Siberia in early July 2016. This Landsat-8 image (bands 6-5-4) shows the discontinuous Larch forests (green). Here, near the eastern boundary of the Putorana Plateau, the snow is gone, except for a few patches (cyan) that remain on steep northern slopes and on what appear to be exposed icy patches (permafrost?) in the northwest portion of the image. This particular feature might warrant a closer look. The rivers seem to be running high enough for transit…

[pahbs]

Geo7X GPS, thou shalt be missed in Siberia. Russia does not approve of your presence.

[pahbs]

Popcorn clouds obscure some of the snow cover that remains on the plateaus on either side of the Tembenchi River in central Siberia at the end of May, 2015 in this Landsat-8 scene (bands 6-5-4). Adjacent to the river, the deciduous conifer Larch trees of these sparse forests will soon begin to push out fresh needles for this growing season.

[pahbs]

Permafrost patterns.

February 27 2012, the Worldview-2 satellite captures the geometric features from ice-wedge polygons covered in snow along a portion of the Khatanga River (N72.51280,  E104.16166) in northern Siberia.   

[pahbs]

Forest sample measurements of tree heights provide estimates of forest biomass within a sampling plot.  Corresponding LiDAR measurements, coupled with other remote sensing data, help infer those biomass measurements across countries and biomes.

This map shows, generally, the number and distribution of sample plots within the boreal forest (dark grey area) that have corresponding airborne or spaceborne LiDAR measurements.  Red lines show the distribution of spaceborne LiDAR measurements from the GLAS sensor from 9/25 - 11/19 2003.

Large areas, particularly in Eurasia, that lack forest measurements drive up the uncertainty of estimates of total boreal forest biomass.  

[pahbs]

High-contrast Landsat 8 mosaic (bands: NIR, SWIR, red) of the taiga-tundra ecotone in northern Siberia.  Regional scale vegetation patterns are clear in this late July 2013 scene.

[pahbs]

Larch trees in northern Siberia growing in ice wedges that form the boundaries of the polygonal ground features. 

This oblique view of an early September satellite image from Digital Globe shows good contrast between the senescing Larch trees (yellow), shrubs (dark green), ground (gray-brown) and water (dark). 

[pahbs]

Violin plots showing the distribution of a variable (R2) for 4 different categories (LiDAR height metrics).

[pahbs]

Photon counting LiDAR return simulations show that ~80% of LiDAR shots will only return 1-2 photons (i.e., height measurements).  To characterize vegetation structure, the returns from multiple LiDAR shots will have to be combined. This decreases the effective resolution of vegetation measurements. However, overlapping shots along a LiDAR track can add more photon returns within a very small ground distance from a previous shot (70cm). 

[pahbs]

A side-by-side view of an ALOS PALSAR mosaic (left) and an ASTER GDEM mosaic (right) in northern Siberia.

The ALOS image bands shown are HH, HV, and the normalized difference of HH and HV.

Bright areas generally represent forested areas.  At the taiga-tundra ecotone boundary, where sparse forests give way to tundra, a distinct change in image brightness is visible. The image pixel resolution is 25m.

[pahbs]

A more in-depth analysis of the previous post is presented above.  

The histogram matrix shows the distributions of photon returns per LiDAR shot for each AGB bin (in Mg/ha, top labels) at each link-scale (in meters, right-side labels).

In examining simulations of LiDAR returns across a forest AGB gradient, the distributions of the number of photon returns per LiDAR shot are not all the same (though they appear similar) when compared to distributions from other AGB bins. Is their difference significant?

The heat-map (which is symmetric, about the diagonals) above shows that, at most link-scales, the lowest 2 AGB bins are generally different than all other bins' distributions of returns/shot. Note, the heat-maps are symmetric about the diagonal.

Here, the p-value is used to identify differences in the means of 2 distributions. A KS test (it's non-parametric) was used to derive the p-values.  

Grid values are colored according to the p-value from the comparison of the distributions from the corresponding AGB bins.

White boxes indicate comparisons where p-values were sufficiently high to fail to reject the null hypothesis.

Grey boxes indicate comparisons where p-values = 1, which also fail to reject the null hypothesis.

Both #2 and #3 suggest that the corresponding distributions are not statistically different.  Thus, for the most part, you are getting the same distribution of photon returns per LiDAR shot, regardless of AGB bin, and regardless of link-scale.  The bins for which this similarity does not hold are the two smallest AGB bins.

[pahbs]

These two corresponding images show a small portion of the taiga-tundra ecotone in northern Siberia. The top is a terrain model showing elevation above the ellipsoid(low to high; red - yellow - green).

Apparently, the WorldView-1 satellite provides stereo image pairs detailed enough to resolve elevation differences between the ground surface and the tops of some trees and forest patches. In sparsely forested areas, where the ground surface and the tree tops are both visible, tree heights can be estimated.