WEBVTT
1
00:00:10.439 --> 00:00:14.710 A:middle L:90%
Yeah. Our next speaker is Dr Matthew Sumnall,
2
00:00:14.720 --> 00:00:18.750 A:middle L:90%
who's a postdoctoral associate in Forest Resources and Environmental Conservation.
3
00:00:19.339 --> 00:00:21.809 A:middle L:90%
Matt has a Masters in remote sensing and spatial
4
00:00:21.809 --> 00:00:24.920 A:middle L:90%
analysis from the University of Southampton and a doctorate in
5
00:00:24.920 --> 00:00:28.420 A:middle L:90%
remote sensing from Bournemouth University in the UK. His
6
00:00:28.420 --> 00:00:31.129 A:middle L:90%
research deals with use of lidar to separate over
7
00:00:31.129 --> 00:00:34.590 A:middle L:90%
story trees from understory vegetation, primarily in loblolly pine
8
00:00:34.590 --> 00:00:37.820 A:middle L:90%
plantations. The goal is to develop more accurate methods
9
00:00:37.820 --> 00:00:41.130 A:middle L:90%
to determine leaf area of both the over story ponds
10
00:00:41.130 --> 00:00:46.060 A:middle L:90%
and the understory vegetation. Thanks very much how to
11
00:00:46.060 --> 00:00:53.960 A:middle L:90%
work this thing. Give them a show. Mhm
12
00:00:58.340 --> 00:01:00.969 A:middle L:90%
. Good afternoon everybody. Let's give you a quick
13
00:01:00.979 --> 00:01:02.979 A:middle L:90%
rundown on some of the methods have been working on
14
00:01:02.979 --> 00:01:06.500 A:middle L:90%
over the past few months um sort of been tasked
15
00:01:06.500 --> 00:01:10.629 A:middle L:90%
with developing methods to aid in forest inventory. And
16
00:01:10.640 --> 00:01:12.819 A:middle L:90%
again following on from what Randy said, looking at
17
00:01:12.829 --> 00:01:17.829 A:middle L:90%
sort of identifying features beneath the top of the canopy
18
00:01:17.840 --> 00:01:21.390 A:middle L:90%
, so canopy depth, the presence of understory.
19
00:01:21.390 --> 00:01:23.120 A:middle L:90%
And then there's other conditions find many cannot be in
20
00:01:23.120 --> 00:01:27.120 A:middle L:90%
terms of how much light there is available and you
21
00:01:27.120 --> 00:01:33.090 A:middle L:90%
speak up a tiny bit oh sorry. And it
22
00:01:33.090 --> 00:01:34.859 A:middle L:90%
was sort of a brief overview of what we're doing
23
00:01:34.340 --> 00:01:38.739 A:middle L:90%
. I am I'm pretty sure most of you are
24
00:01:38.739 --> 00:01:41.480 A:middle L:90%
aware of what lighter is. And then we'll go
25
00:01:41.480 --> 00:01:44.659 A:middle L:90%
through sort of creation of how we go about sort
26
00:01:44.659 --> 00:01:47.560 A:middle L:90%
of extracting those metrics beneath the top of the canopy
27
00:01:48.439 --> 00:01:51.519 A:middle L:90%
. And then we'll go into what metrics we actually
28
00:01:51.519 --> 00:01:55.239 A:middle L:90%
have and the sort of prediction of the various other
29
00:01:55.239 --> 00:02:00.530 A:middle L:90%
ones um sort of a sort of brief overview of
30
00:02:00.530 --> 00:02:02.379 A:middle L:90%
what binder is. It's uh what we're talking about
31
00:02:02.379 --> 00:02:07.700 A:middle L:90%
here is the capture of and data from typical airport
32
00:02:07.700 --> 00:02:09.719 A:middle L:90%
platforms that can either be planes or helicopters. And
33
00:02:09.719 --> 00:02:15.199 A:middle L:90%
it generates a vast quantities of three dimensional points from
34
00:02:15.199 --> 00:02:19.629 A:middle L:90%
the ground surface below and in the case of forest
35
00:02:19.639 --> 00:02:22.689 A:middle L:90%
you can see from the pretty picture there and it
36
00:02:22.699 --> 00:02:24.810 A:middle L:90%
penetrates through the canopy. You can get pretty good
37
00:02:24.810 --> 00:02:29.849 A:middle L:90%
measurements of various features which would otherwise be obscured.
38
00:02:29.849 --> 00:02:34.669 A:middle L:90%
And other techniques such as aerial photographs or hyper spectral
39
00:02:34.669 --> 00:02:39.550 A:middle L:90%
imagery and so later is typically processed in the following
40
00:02:39.550 --> 00:02:43.919 A:middle L:90%
way. You use a technique to classify the ground
41
00:02:43.919 --> 00:02:46.849 A:middle L:90%
returns. So features which are the lowest elevation usually
42
00:02:47.750 --> 00:02:51.719 A:middle L:90%
. And then you can trade surface from that and
43
00:02:51.719 --> 00:02:54.169 A:middle L:90%
normalize the whites to remove the effects of terrain and
44
00:02:54.169 --> 00:02:58.669 A:middle L:90%
produce an above ground hide map. Um In the
45
00:02:58.680 --> 00:03:00.669 A:middle L:90%
in the case of some of the examples were used
46
00:03:00.669 --> 00:03:04.349 A:middle L:90%
, we identify features which are sort of plaintive services
47
00:03:04.349 --> 00:03:08.280 A:middle L:90%
which we attributed to being and buildings or my main
48
00:03:08.280 --> 00:03:15.000 A:middle L:90%
structures. And then finally we can extract very statistics
49
00:03:15.729 --> 00:03:21.389 A:middle L:90%
and then perform various analyses on them. And within
50
00:03:21.389 --> 00:03:23.460 A:middle L:90%
the context of this project, we didn't typically use
51
00:03:23.469 --> 00:03:28.310 A:middle L:90%
the standard, basic little metrics such as um sort
52
00:03:28.310 --> 00:03:30.849 A:middle L:90%
of the average height of all returns. So we
53
00:03:30.849 --> 00:03:37.000 A:middle L:90%
produced um a vertical profile and in order to do
54
00:03:37.000 --> 00:03:40.289 A:middle L:90%
that we stratified both, stratified the point cloud both
55
00:03:40.289 --> 00:03:44.610 A:middle L:90%
horizontally and vertically. Um With the example here,
56
00:03:44.620 --> 00:03:49.139 A:middle L:90%
it's it essentially summed up all the leader returns within
57
00:03:49.150 --> 00:03:53.460 A:middle L:90%
each one by one by one m three D objects
58
00:03:53.139 --> 00:03:57.460 A:middle L:90%
um typically typically referred to either as a box along
59
00:03:57.469 --> 00:04:03.639 A:middle L:90%
with volumetric pixel or height bent. Um So so
60
00:04:03.650 --> 00:04:08.620 A:middle L:90%
really it ends up being so each sort of vertical
61
00:04:08.620 --> 00:04:13.129 A:middle L:90%
elevation chunk can be attributed to abandon a hyper spectral
62
00:04:13.139 --> 00:04:20.370 A:middle L:90%
multispectral image. Yeah, for this first example we
63
00:04:20.370 --> 00:04:26.149 A:middle L:90%
used Are kind of 2008 data for three sites in
64
00:04:26.149 --> 00:04:30.829 A:middle L:90%
both Virginia and North Carolina. Um 11 particular example
65
00:04:30.829 --> 00:04:32.759 A:middle L:90%
here which have all run through and we generated a
66
00:04:33.839 --> 00:04:40.170 A:middle L:90%
a vertical profile which is essentially a number of returns
67
00:04:40.170 --> 00:04:44.459 A:middle L:90%
by height to sort of a summary of what the
68
00:04:45.240 --> 00:04:47.329 A:middle L:90%
what the what the well the graphic summary of what
69
00:04:47.329 --> 00:04:50.730 A:middle L:90%
the previous graphic was all about. And as you
70
00:04:50.730 --> 00:04:54.839 A:middle L:90%
can see from this, it's there are two main
71
00:04:54.839 --> 00:04:57.759 A:middle L:90%
features you can see, does this point to.
72
00:04:59.540 --> 00:05:01.790 A:middle L:90%
Um Anyway, the main feature here, you can
73
00:05:01.790 --> 00:05:05.589 A:middle L:90%
see that is um well it's obviously that the primary
74
00:05:05.589 --> 00:05:09.709 A:middle L:90%
company will make a canopy. And the the other
75
00:05:09.709 --> 00:05:16.100 A:middle L:90%
feature is the understory and well, there's the main
76
00:05:16.100 --> 00:05:18.829 A:middle L:90%
focus of this particular method was to develop an automated
77
00:05:18.829 --> 00:05:24.560 A:middle L:90%
technique to classify the canopy. Another story components.
78
00:05:25.139 --> 00:05:28.639 A:middle L:90%
We did this first by fitting a curve function to
79
00:05:28.639 --> 00:05:32.310 A:middle L:90%
the data and then identifying maximum minimum rights. And
80
00:05:32.310 --> 00:05:38.000 A:middle L:90%
from that we could estimates from from from the maximum
81
00:05:38.009 --> 00:05:41.220 A:middle L:90%
the presence of a canopy layer for the plot level
82
00:05:41.230 --> 00:05:44.449 A:middle L:90%
or the president to another story away. Mhm.
83
00:05:44.930 --> 00:05:46.889 A:middle L:90%
And this could be replicated for the entire study side
84
00:05:46.889 --> 00:05:49.360 A:middle L:90%
of whatever sort of field plot scale we decided to
85
00:05:49.360 --> 00:05:54.120 A:middle L:90%
implement. Um in terms of a couple of results
86
00:05:54.129 --> 00:05:57.829 A:middle L:90%
for three of our study sites with this Anderson,
87
00:05:57.920 --> 00:06:00.040 A:middle L:90%
which is in North Carolina region wide, 18 Virginia
88
00:06:00.430 --> 00:06:04.420 A:middle L:90%
and um see trees. This is the sort of
89
00:06:04.420 --> 00:06:08.699 A:middle L:90%
validation for the for the primary canopy. So the
90
00:06:08.699 --> 00:06:12.509 A:middle L:90%
top of the canopy height and the height of the
91
00:06:12.509 --> 00:06:15.259 A:middle L:90%
living crown, which is what can be very site
92
00:06:15.240 --> 00:06:17.410 A:middle L:90%
As just as you can see from the vertical profile
93
00:06:17.410 --> 00:06:20.100 A:middle L:90%
, we usually get an accuracy of around one m
94
00:06:20.100 --> 00:06:25.300 A:middle L:90%
for the very top plots height and within about two
95
00:06:25.300 --> 00:06:30.569 A:middle L:90%
m for the base height. Mhm. Um Well
96
00:06:30.569 --> 00:06:32.139 A:middle L:90%
just going a bit more into detail about the detection
97
00:06:32.139 --> 00:06:34.069 A:middle L:90%
of the story. As you can see, we
98
00:06:34.069 --> 00:06:38.310 A:middle L:90%
can detect very much from the profile of the presence
99
00:06:38.310 --> 00:06:41.550 A:middle L:90%
or absence. Um That's where the top of the
100
00:06:41.550 --> 00:06:45.589 A:middle L:90%
canopy can make estimates on the height and depth of
101
00:06:45.600 --> 00:06:49.889 A:middle L:90%
the understory as well. We can also use this
102
00:06:49.889 --> 00:06:54.339 A:middle L:90%
as a sort of an aging management tool. Um
103
00:06:54.350 --> 00:06:58.129 A:middle L:90%
This particular example has four different side treatment techniques.
104
00:06:58.129 --> 00:07:00.670 A:middle L:90%
Well, initial planting techniques where the whole stamp was
105
00:07:00.670 --> 00:07:04.170 A:middle L:90%
removed or the whole tree was removed, including roots
106
00:07:04.519 --> 00:07:08.509 A:middle L:90%
and then, Yeah, so you get sort of
107
00:07:08.509 --> 00:07:14.720 A:middle L:90%
very different sort of initial conditions for these plots and
108
00:07:14.720 --> 00:07:17.490 A:middle L:90%
the vertical profiles are well. Well, for the
109
00:07:17.500 --> 00:07:20.660 A:middle L:90%
service site treatments, where the only story is completely
110
00:07:20.670 --> 00:07:25.689 A:middle L:90%
clear from the start are all very similar. Also
111
00:07:25.689 --> 00:07:29.089 A:middle L:90%
when the one The story is left, you can
112
00:07:29.089 --> 00:07:30.410 A:middle L:90%
see that and some of them is very sort of
113
00:07:30.420 --> 00:07:35.560 A:middle L:90%
minute changes. But for the some removal of aliens
114
00:07:36.139 --> 00:07:40.439 A:middle L:90%
because in but you can see that the primary cannot
115
00:07:40.449 --> 00:07:44.750 A:middle L:90%
be is very much very different because might be able
116
00:07:44.750 --> 00:07:46.620 A:middle L:90%
to infer that because of the presence of the story
117
00:07:46.629 --> 00:07:49.160 A:middle L:90%
. But we don't know ourselves yet. Mhm.
118
00:07:50.540 --> 00:07:55.139 A:middle L:90%
Um Not that potential tool is high scale cannot be
119
00:07:55.139 --> 00:07:58.579 A:middle L:90%
an openness index which is essentially for mapping vertical gaps
120
00:07:58.589 --> 00:08:01.209 A:middle L:90%
. It's another sort of aid in detecting the presence
121
00:08:01.209 --> 00:08:05.089 A:middle L:90%
or absence from the story. This particular technique was
122
00:08:05.089 --> 00:08:09.980 A:middle L:90%
developed in Queensland Australia for um well I think I
123
00:08:09.980 --> 00:08:13.069 A:middle L:90%
guess typically for the detection of suppressed trees or trees
124
00:08:13.069 --> 00:08:15.660 A:middle L:90%
that will not be represented in the cabinet behind model
125
00:08:16.540 --> 00:08:20.540 A:middle L:90%
. Mhm. This is one particular example we produced
126
00:08:20.540 --> 00:08:24.870 A:middle L:90%
where, well each of these figures plots were well
127
00:08:24.879 --> 00:08:26.029 A:middle L:90%
well her but I would use for each of these
128
00:08:26.029 --> 00:08:30.250 A:middle L:90%
field plots to remove any any understory vegetation at all
129
00:08:30.839 --> 00:08:33.620 A:middle L:90%
for the for the plots where you can obviously see
130
00:08:33.620 --> 00:08:35.220 A:middle L:90%
the ground and have been thinned. It's pretty obvious
131
00:08:35.220 --> 00:08:39.200 A:middle L:90%
but closed canopy conditions. Um If you look at
132
00:08:39.200 --> 00:08:43.440 A:middle L:90%
the hcl I you can see that there is very
133
00:08:43.440 --> 00:08:46.039 A:middle L:90%
much in the absence of the story that Yeah,
134
00:08:46.639 --> 00:08:50.440 A:middle L:90%
another technique we've been, I've been trying to develop
135
00:08:50.440 --> 00:08:54.519 A:middle L:90%
is an automated individual treatment technique which will sort of
136
00:08:54.519 --> 00:08:58.570 A:middle L:90%
identify the position and various other attributes of the individual
137
00:08:58.570 --> 00:09:05.559 A:middle L:90%
trees. Um This time it typically starts with utilizing
138
00:09:05.559 --> 00:09:09.769 A:middle L:90%
a high resolution coming behind model and identifying the heights
139
00:09:09.769 --> 00:09:15.110 A:middle L:90%
of maximum elevation using a global search window and each
140
00:09:15.110 --> 00:09:16.049 A:middle L:90%
of the each of the points in the graphic that
141
00:09:16.049 --> 00:09:20.350 A:middle L:90%
represents the the absolute maximum of each of the kind
142
00:09:20.350 --> 00:09:24.080 A:middle L:90%
of people each of the individual trees. Um Once
143
00:09:24.080 --> 00:09:26.299 A:middle L:90%
we know that that location will be used in iterative
144
00:09:26.299 --> 00:09:30.870 A:middle L:90%
technique to grow a region around each point, expanding
145
00:09:30.870 --> 00:09:35.379 A:middle L:90%
into areas of lower elevation from the maximum in order
146
00:09:35.379 --> 00:09:39.559 A:middle L:90%
to generate an estimate of primary radius and area.
147
00:09:41.539 --> 00:09:43.500 A:middle L:90%
You can see this is a sort of a blow
148
00:09:43.500 --> 00:09:46.340 A:middle L:90%
up of the previous graphic. You can see it
149
00:09:46.350 --> 00:09:50.559 A:middle L:90%
corresponds pretty closely to what you can complete by i
150
00:09:52.139 --> 00:09:54.139 A:middle L:90%
in terms of where the canopy is, it's not
151
00:09:54.139 --> 00:09:58.970 A:middle L:90%
perfect. Unfortunately, He said this one. This
152
00:09:58.970 --> 00:10:01.580 A:middle L:90%
particular one, for example, is bit of an
153
00:10:01.580 --> 00:10:05.980 A:middle L:90%
anomaly. It's still still very much work in progress
154
00:10:07.039 --> 00:10:09.019 A:middle L:90%
. Yes, in terms of accuracy is in terms
155
00:10:09.019 --> 00:10:13.720 A:middle L:90%
of the population of plots here varies by underestimating between
156
00:10:13.720 --> 00:10:16.870 A:middle L:90%
five and centuries, but otherwise is is usually pretty
157
00:10:16.870 --> 00:10:20.840 A:middle L:90%
good. We can also generate vertical profiles for the
158
00:10:20.850 --> 00:10:24.070 A:middle L:90%
spatial extent of each individual tree crown. And obviously
159
00:10:24.070 --> 00:10:28.320 A:middle L:90%
we can we can infer sort of and top flight
160
00:10:28.320 --> 00:10:31.049 A:middle L:90%
and base height metrics as well. You can see
161
00:10:31.049 --> 00:10:37.350 A:middle L:90%
from the graphic vertical profile graphic the The top two
162
00:10:37.600 --> 00:10:41.679 A:middle L:90%
, sorry that the two maximal red fights on the
163
00:10:41.690 --> 00:10:45.590 A:middle L:90%
right hand side of the rough indicative of the canopy
164
00:10:45.590 --> 00:10:48.580 A:middle L:90%
itself. And it corresponds pretty well with field data
165
00:10:48.580 --> 00:10:52.240 A:middle L:90%
at this point. Unfortunately, I'm not sure why
166
00:10:52.240 --> 00:10:54.940 A:middle L:90%
it generates to maximum for each of the three drums
167
00:10:54.940 --> 00:10:56.759 A:middle L:90%
have locked down so far, but it's still like
168
00:10:56.759 --> 00:11:01.820 A:middle L:90%
I said work in progress and the lower points to
169
00:11:01.820 --> 00:11:03.370 A:middle L:90%
the left of the main ones are either indicative of
170
00:11:03.379 --> 00:11:09.799 A:middle L:90%
understory vegetation or potentially that branches. And likewise,
171
00:11:09.809 --> 00:11:13.220 A:middle L:90%
once we have estimates of trial radius and various elevations
172
00:11:13.220 --> 00:11:16.200 A:middle L:90%
, we can make pretty reasonable estimates of mechanical crown
173
00:11:16.200 --> 00:11:22.360 A:middle L:90%
volume. Mm. Yeah. Um The last sort
174
00:11:22.360 --> 00:11:26.240 A:middle L:90%
of lightning tour method I'm going to take you through
175
00:11:26.240 --> 00:11:28.539 A:middle L:90%
is sort of estimation of leaf error index that a
176
00:11:28.539 --> 00:11:33.370 A:middle L:90%
number of vertical levels within the within the within the
177
00:11:33.370 --> 00:11:37.159 A:middle L:90%
field. What extent to do this? We visited
178
00:11:37.159 --> 00:11:39.730 A:middle L:90%
two particular sites in Focus trapped into forest in North
179
00:11:39.730 --> 00:11:45.460 A:middle L:90%
Carolina so separated by probably about 200 miles. Um
180
00:11:45.940 --> 00:11:48.809 A:middle L:90%
Data set was provided very gracefully by Nasa G.
181
00:11:48.809 --> 00:11:52.669 A:middle L:90%
Like team. Um In order to estimate, oh
182
00:11:52.669 --> 00:11:54.409 A:middle L:90%
, I should probably mention a bit, but leaf
183
00:11:54.409 --> 00:11:56.629 A:middle L:90%
area index says, I'm sure most of you are
184
00:11:56.629 --> 00:12:01.529 A:middle L:90%
aware but it's essentially can be defined as the amount
185
00:12:01.529 --> 00:12:05.570 A:middle L:90%
of leaf matter above ground. It's going to have
186
00:12:05.570 --> 00:12:11.250 A:middle L:90%
various various influences on growth models and the things that
187
00:12:11.240 --> 00:12:13.309 A:middle L:90%
may be quite useful in a forest in for treatment
188
00:12:13.320 --> 00:12:18.419 A:middle L:90%
sense. In order to measure leaf area index at
189
00:12:18.429 --> 00:12:22.870 A:middle L:90%
this scale, we used The Guy 20 200 which
190
00:12:22.870 --> 00:12:26.360 A:middle L:90%
is this Magic one here. Yeah. And then
191
00:12:28.840 --> 00:12:31.789 A:middle L:90%
We used to 50 m. Transact with one with
192
00:12:31.789 --> 00:12:35.279 A:middle L:90%
one measurement every meter. Um three vertical levels of
193
00:12:35.289 --> 00:12:41.850 A:middle L:90%
0m ground level 1.5 m on 2.5 m. So
194
00:12:41.240 --> 00:12:45.379 A:middle L:90%
this is this is order to test the hypothesis that
195
00:12:46.049 --> 00:12:50.929 A:middle L:90%
lie measurement under on the story of, Sorry and
196
00:12:50.929 --> 00:12:54.000 A:middle L:90%
I measurement of understory 0m may not be indicative of
197
00:12:54.009 --> 00:12:56.850 A:middle L:90%
the the leaf area conditions of the main canopy.
198
00:12:58.399 --> 00:13:01.360 A:middle L:90%
And obviously this is very important for growth models.
199
00:13:03.740 --> 00:13:07.210 A:middle L:90%
Um for this, for this particular project we had
200
00:13:07.220 --> 00:13:09.950 A:middle L:90%
to to field sites were only able to do 18
201
00:13:09.950 --> 00:13:13.659 A:middle L:90%
parkas tracks due to relatively small size of the acquisition
202
00:13:13.240 --> 00:13:18.600 A:middle L:90%
and well relatively hi difficulty actually getting around the site
203
00:13:18.610 --> 00:13:20.580 A:middle L:90%
because of high levels of understory and a lot of
204
00:13:20.590 --> 00:13:26.720 A:middle L:90%
pain. And with the Duke forrest We were able
205
00:13:26.720 --> 00:13:31.049 A:middle L:90%
to install 20 plots at a far greater spatial extent
206
00:13:31.639 --> 00:13:35.639 A:middle L:90%
and we're able to example various um stand ages and
207
00:13:35.639 --> 00:13:39.860 A:middle L:90%
cannot be structural types. Um In terms of wider
208
00:13:39.860 --> 00:13:43.929 A:middle L:90%
processing, it continue very much as it has been
209
00:13:43.929 --> 00:13:48.009 A:middle L:90%
the previous steps And we extracted metrics and produce vertical
210
00:13:48.009 --> 00:13:52.539 A:middle L:90%
profiles for 20 by 20 metre areas. Um We're
211
00:13:52.539 --> 00:13:58.750 A:middle L:90%
also able to estimate so various various metrics based on
212
00:13:58.750 --> 00:14:01.549 A:middle L:90%
the stratification of the point cloud based on those layers
213
00:14:03.039 --> 00:14:05.750 A:middle L:90%
. Um So we're able to for example, estimate
214
00:14:05.759 --> 00:14:09.250 A:middle L:90%
the total proportion of points within the main canopy itself
215
00:14:11.240 --> 00:14:16.809 A:middle L:90%
. We're also able to modify an existing technique first
216
00:14:16.820 --> 00:14:20.049 A:middle L:90%
employed by uh Alicia producing a couple of years ago
217
00:14:20.649 --> 00:14:26.470 A:middle L:90%
, developing gave certainly called canopy Yeah, kind of
218
00:14:26.500 --> 00:14:33.259 A:middle L:90%
authenticity slices. So it's essentially focuses on finding the
219
00:14:33.840 --> 00:14:35.460 A:middle L:90%
The point within three canopy way, you'll find the
220
00:14:35.460 --> 00:14:39.220 A:middle L:90%
most light or returns, identifying that point and developing
221
00:14:39.220 --> 00:14:43.389 A:middle L:90%
metrics, sort of both above and below that five
222
00:14:43.389 --> 00:14:48.720 A:middle L:90%
m meter increments. Uh For each of these these
223
00:14:48.730 --> 00:14:52.029 A:middle L:90%
slices you can have estimate metrics such as the mean
224
00:14:52.029 --> 00:14:56.549 A:middle L:90%
standard deviation various. And I'm coefficient two variants.
225
00:14:56.440 --> 00:15:01.730 A:middle L:90%
Yeah. And we're able to produce a couple of
226
00:15:01.730 --> 00:15:03.700 A:middle L:90%
days ago actually a number of models for the estimation
227
00:15:03.700 --> 00:15:09.429 A:middle L:90%
of a at these three particular levels. Um Used
228
00:15:09.440 --> 00:15:13.460 A:middle L:90%
standard linear regression technique, story stepwise linear progression technique
229
00:15:13.840 --> 00:15:18.190 A:middle L:90%
, Our party's 100 30 metrics. And eventually we
230
00:15:18.190 --> 00:15:22.000 A:middle L:90%
came to these three models. You got like that's
231
00:15:22.840 --> 00:15:26.570 A:middle L:90%
the vast majority of the main predictors of these models
232
00:15:26.570 --> 00:15:28.409 A:middle L:90%
where the kind of identity slices, the proportion of
233
00:15:28.409 --> 00:15:33.009 A:middle L:90%
ground returns and a number of metrics related to the
234
00:15:33.009 --> 00:15:37.240 A:middle L:90%
stratification of the point. Cloud yourself. Yeah.
235
00:15:39.139 --> 00:15:43.740 A:middle L:90%
As Vincent greek. Um So the vertical profiles can
236
00:15:43.740 --> 00:15:48.220 A:middle L:90%
be a very powerful tool to estimate directly and statistically
237
00:15:48.409 --> 00:15:52.350 A:middle L:90%
elements within the within the forest canopy and underneath itself
238
00:15:52.139 --> 00:15:58.159 A:middle L:90%
. And yeah, so automated techniques for developing individual
239
00:15:58.159 --> 00:16:00.129 A:middle L:90%
tree methods, R. And D. Possible and
240
00:16:00.139 --> 00:16:04.450 A:middle L:90%
fairly accurate and Yeah, that's pretty much it does
241
00:16:04.460 --> 00:16:17.179 A:middle L:90%
. Does anybody have any questions? Yes. 1st
242
00:16:17.190 --> 00:16:21.210 A:middle L:90%
, Great Presentation. Thank you. two I guess
243
00:16:21.210 --> 00:16:25.179 A:middle L:90%
to me some idea of the process and kind of
244
00:16:25.190 --> 00:16:27.879 A:middle L:90%
uh kind of machines and people used to use in
245
00:16:27.879 --> 00:16:30.870 A:middle L:90%
the process. The vast majority of processing was done
246
00:16:30.870 --> 00:16:34.429 A:middle L:90%
on my my work laptop, which is probably about
247
00:16:34.429 --> 00:16:37.789 A:middle L:90%
five years old and pretty rubbish. Um But in
248
00:16:37.789 --> 00:16:41.779 A:middle L:90%
terms of processing time it probably took probably about the
249
00:16:41.789 --> 00:16:45.830 A:middle L:90%
day her study site. So what about 24 hours
250
00:16:45.830 --> 00:16:51.059 A:middle L:90%
per study site? So on a more modern machine
251
00:16:51.059 --> 00:16:59.850 A:middle L:90%
of people fossil period? Yeah. Mhm. Mhm
252
00:17:03.840 --> 00:17:08.859 A:middle L:90%
. Well, thank you one.