-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathdocumentation.html
270 lines (270 loc) · 19.8 KB
/
documentation.html
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
<div class="header">GNN Trend and Accounting Explorer (TrAccEr) Help</div>
<div class="content">
<div class="ui styled fluid accordion">
<div class="active title">
<i class="dropdown icon"></i>
Introduction
</div>
<div class="active content small-panel">
<p>
The gradient nearest neighbor (GNN) imputation mapping approach has been used to develop wall-to-wall,
30-m resolution vegetation maps for forests in western North America, focusing primarily on
California, Oregon, and Washington. However, GNN users must have substantial analytical expertise
and computational resources to effectively leverage these data. This may be a significant barrier for
many interested parties. For example, the most recently available
<a href="http://lemma.forestry.oregonstate.edu/data/structure-maps" target="_blank">GNN structure
maps</a>
are roughly 1 GB in size. The recent development of annual maps stretching back to the early days
of Landsat 5 (1985 to 2012) means that every pixel (900 m<sup>2</sup>) has 28 annual estimates of
each of 333 structure variables associated with it. As a result, users are faced by [28 years] × [333
variables] × [1.8×10<sup>9</sup> pixels] = 1.7×10<sup>13</sup> predictions they might summarize
across space and time. This quantity of data can be truly daunting. To simplify users’ introduction
to the GNN data and facilitate data exploration, we have developed the GNN Trend and Accounting
Explorer (GNN TrAccEr), a web-based platform for visualizing and exporting landscape- and
regional-scale data summaries.
</p>
<p>
While the purpose of the GNN TrAccEr is to simplify the user’s introduction to GNN data, there are
significant challenges to using and interpreting the summarized results presented by the tool. Below,
we provide several links to additional information about the GNN TrAccEr and the GNN method itself.
</p>
</div>
<div class="title">
<i class="dropdown icon"></i>
Description
</div>
<div class="content small-panel">
<p>
The GNN TrAccEr can be thought of as two different components: 1) a background analysis script and 2) a
web-based visualization tool. For the background analysis component, we combine rasters that represent
strata (e.g. ownership, land-use allocation, etc.) together with the full time series of GNN rasters
(1985-2012) so that all unique combinations of strata are calculated for every year. This can be a
time-consuming processing task depending on the number of strata and the geographic extent of the
rasters (up to 12 hours). Once these are calculated, the yearly files are combined into a single
comma-delimited, or CSV, file where every record represents a unique strata combination for a given
year and captures the response variable information for the requested GNN variables, including area.
</p>
<p>
The CSV file is combined with a JavaScript Object Notation (JSON) file (which provides configuration
options and metadata) and placed on the web server. The web-based visualization tool reads the
JSON configuration file, dynamically builds the user interface, and provides logic for summarizing
across unique combinations of strata and providing interactive charting to the user. Note that the
web-based visualization is not specific to GNN data – one should be able to adapt this tool for
summarizing data from any trend-based model or simulation.
</p>
</div>
<div class="title">
<i class="dropdown icon"></i>
Using the GNN TrAccEr
</div>
<div class="content large-panel">
<p>
The front-end of the GNN TrAccEr (i.e., web-based visualization) is quite flexible, allowing for
a wide degree of customization for individual user groups. Examples will be available soon on
an archiving website. Regardless of which configuration of the GNN TrAccEr is used, there are several
steps to using the tool and understanding the data presented. As we describe how to use the GNN
TrAccEr, we will use an example configuration (below) of the tool designed to explore stand biomass
across Oregon through time.
</p>
<img src="images/traccer-overview-800px.png" class="doc-image">
<ol class="ui list">
<li>
<span class="list-item-heading">Tool layout</span>. The GNN TrAccEr is organized into two
sections: the upper portion of the window that allows the user to define how to organize the
data and the lower portion where the trends are summarized and plotted. In the upper portion,
links to the left control how the data are stratified while links on the right control the
sub-setting of the data into different series, altering the range of years, changing the
variable of interest, and exporting the trend information as a comma-delimited file.
</li>
<li>
<span class="list-item-heading">Stratifying data</span>. Initially, all data are combined
together to provide the mean of a variable of interest or the area the data cover. To focus
on only certain strata, the user can click on one of the Focus Areas (e.g., <em>Model Region</em>).
A dialog box will appear and, when the empty field is selected, a dropdown of strata appears. Any
number of these can be selected. Finally, the user needs to select “OK” to use only these strata.
<img src="images/traccer-stratum-dialog-800px.png" class="doc-image"/>
There are two types of strata that have different effects on trends in the data: strata defined
based on the GNN data itself and strata defined by other spatial coverages (i.e., non-GNN strata).
In the current example, <em>Model Region</em>, <em>Ownership</em>, and <em>Land allocation</em>
are non-GNN strata and <em>Old growth – 80</em> and <em>Old growth – 200</em> are GNN variables.
<ol>
<li>
<span class="list-item-heading">GNN strata</span>. Stratification
based on GNN strata are useful for examining changes in the
area of vegetation features through time as the GNN data vary from year to year as a
function of the changes in the satellite imagery. Examination of response
variables relates to the average state on that GNN strata through time. However, the strata
and response variables may not be independent. In the current example, old-growth forests
(i.e., classified as Old growth – 200) tend to have higher biomass than forests without.
</li>
<li>
<span class="list-item-heading">Non-GNN strata</span>. Stratification based on non-GNN
strata are useful for examining changes in response variables (e.g., biomass) through
time. Currently, non-GNN strata are not allowed to vary through time so that area for
each non-GNN strata is constant.
</li>
<li>
<span class="list-item-heading">Combining GNN and non-GNN strata</span>. Using more than
one strata (both GNN and non-GNN strata) can be used for both trends in area or
response variables, though users should take care when interpreting these combinations
and think about how those results relate to the strata being used. For more information,
see the <a href="#" onclick="pitfalls(); return false;">Potential Pitfalls</a> section.
</li>
</ol>
</li>
<li>
<span class="list-item-heading">Sub-setting the data into series.</span> For the purposes of
plotting, series are represented by each individual line. In other words, the user can plot
several strata in the same figure. To the right, the series is defined and can be changed by
clicking on the current series definition. Then, the user can select any of the strata, which
will present results for each strata within the series separately. For example, for <em>Land
allocation</em>, different trends are presented for the mean conditions or area for multi-resource
allocations, Private-OFRA allocations, etc.
<img src="images/traccer-series-dialog-800px.png" class="doc-image"/>
</li>
<li>
<span class="list-item-heading">Altering the range of years.</span> Data are available from
1985 to 2012. Users can define any continuous range within that period.
<img src="images/traccer-year-dialog-800px.png" class="doc-image"/>
</li>
<li>
<span class="list-item-heading">Changing the variable of interest.</span> The response variable
of interest can be altered, usually including area as well as continuous GNN variables.
<img src="images/traccer-variable-dialog-800px.png" class="doc-image"/>
</li>
<li>
<span class="list-item-heading">Altering chart display features.</span> There are several ways
the user can interact directly with the chart itself.
<ol>
<li>
<span class="list-item-heading">Zoom.</span> While holding the cursor over the figure,
using the mouse wheel will alter the scale for the response variable (i.e., y-axis)
<img src="images/traccer-zoom-800px.png" class="doc-image"/>
</li>
<li>
<span class="list-item-heading">Point Data.</span> While hovering the cursor over a
single point in the time-series, GNN TrAccEr will display the name of the strata
represented and the value of the response variable for that year by strata combination.
<img src="images/traccer-hover-800px.png" class="doc-image"/>
</li>
<li>
<span class="list-item-heading">Highlight Strata.</span> While hovering over the name
of the strata in the legend above the chart, GNN TrAccEr will highlight the time-series
for that strata.
<img src="images/traccer-legend-hover-800px.png" class="doc-image"/>
</li>
</ol>
</li>
<li>
<span class="list-item-heading">Exporting the data as a comma delimited file.</span> After
selecting "Export to CSV", the user can specify a file name and the CSV file will be saved to the
user's download location.
</li>
</ol>
</div>
<div class="title">
<i class="dropdown icon"></i>
Potential Pitfalls
</div>
<div class="content large-panel">
<p>
The GNN TrAccEr is flexible and easy to use, but users should take care when interpreting the trends.
In particular, several questions should always be in users’ minds:
</p>
<ul class="ui list">
<li>
<span class="list-item-heading">Should the vegetation trends of interest be captured over broad
geographic areas?</span>
Users should think carefully about whether the change they are attempting to examine is large
enough to be captured at the level of the strata being used. For example, the impacts of
disturbance events on broad-scale temporal trends in forest attributes presented by the GNN
TrAccEr depends on the size of strata and the size of the disturbance. The Biscuit Fire (2002)
covered over 2000 km<sup>2</sup> and altered forest structure across a large portion of southwest
Oregon. As a result, substantial trends in forest structure (e.g., forest area with old-growth
structure index > 200) are clear and large relative to their initial values (approximately
30% reduction) (see below).
<img src="images/traccer-ogsi200-area-800px.png" class="doc-image"/>
</li>
<li>
<span class="list-item-heading">Are the vegetation trends dramatic?</span> For most elements of
vegetation structure and composition, change is usually slow (e.g., biomass increases as trees
grow, species turnover as forest succession progresses, etc.). However, disturbance events, such
as wildfire, timber harvesting, insect and disease outbreak, alter forest conditions rapidly over
the course of days to years. At broad scales, slow, subtle trends may be difficult to capture
and may be hard to differentiate from stable behavior (i.e., no change).
</li>
<li>
<span class="list-item-heading">Are area estimates or continuous attributes most appropriate?</span>
The selection of appropriate response variables to plot trends of is a key decision and requires
some thought. For non-GNN strata, areas do not change through time as we use a single raster to
represent these strata. Because estimates of areas will be static, trends in non-area response
variables are appropriate. For GNN strata, the opposite is true because GNN strata change annually
with imagery, making area estimates interpretable (e.g., how has area of old-growth changed?),
but examination of other response variables difficult. For example, if we plot mean OGSI for
old-growth forests in each physiographic province (see below), we are looking at a chart that
tells us how OGSI changes in forests with OGSI similar to forests with stand age greater than
200. How does one interpret this chart? Is the old-growthiness of old-growth changing over time?
<img src="images/traccer-ogsi200-mean-800px.png" class="doc-image"/>
</li>
<li>
<span class="list-item-heading">Is the trend ecologically meaningful?</span> Users need to be
well-aware of what constitutes an ecologically meaningful change in a response variables. For
example, if we create a chart to look at mean canopy cover on USFS lands in Oregon and Washington
(USFS-R6) for the East Cascades of Oregon (Phys. Province 203) we might see dramatic
fluctuations (see below), but once one looks at the scale of the response variable, it becomes
clear that these fluctuations may not be of much importance (range of canopy covers is 2%).
Therefore, users should carefully examine the scale of the response variables before trying to
interpret a perceived pattern.
<img src="images/traccer-cancov-mean-800px.png" class="doc-image"/>
</li>
<li>
<span class="list-item-heading">Are the attributes of features accurately represented in
vegetation maps?</span>
Since the inception of GNN, there has been a heavy focus on accuracy assessment to inform users
as to the limitations of the data. Those same uncertainties are likely to be translated into
results from the GNN TrAccEr. We suggest that users familiarize themselves with our publications,
especially <a href="http://lemma.forestry.oregonstate.edu/export/pubs/pierce_etal_2009_cjfr.pdf"
target="_blank">Pierce et al. 2009</a> and <a
href="http://lemma.forestry.oregonstate.edu/export/pubs/bell_etal_2015_fem_ImputationUncertainty.pdf"
target="_blank">Bell et al. 2015</a>. Additionally, accuracy assessment reports are
available <a href="http://lemma.forestry.oregonstate.edu/data/structure-maps" target="_blank">here</a>
as part of the GNN download.
</li>
</ul>
</div>
<div class="title">
<i class="dropdown icon"></i>
GNN TrAccEr Development and Archiving
</div>
<div class="content small-panel">
The LEMMA group recognizes that user interest in GNN data arises from diverse questions. Consequently,
there will be a need for the construction of various trend configurations designed to answer different
questions. In addition, some users may wish to develop novel features, such as modifying the code
underlying the tool to change plotting features/options. Such activities can contribute to the development
of the tool. We welcome collaborations with users that increase accessibility to GNN data. To that end,
the GNN TrAccEr code is <a href="https://github.com/lemma-osu/trajectory" target="_blank">open-source
and freely available</a>. In addition, we plan to actively archive different geospatial combines (CSV files)
and JSON files and are willing, to the extent that time allows, to work with others to offer versions of
the GNN TrAccEr for new and interesting questions.
</div>
<div class="title">
<i class="dropdown icon"></i>
Contact Information
</div>
<div class="content small-panel">
For further information, please contact <a href="http://lemma.forestry.oregonstate.edu/about/david-bell"
target="_blank">Dr. David M. Bell</a> (U.S. Forest Service) or
<a href="http://lemma.forestry.oregonstate.edu/about/matthew-gregory" target="_blank">Matt Gregory</a>
(Oregon State University). Both are members of the Landscape Ecology, Modeling, Mapping, and Analysis
(<a href="http://lemma.forestry.oregonstate.edu/about" target="_blank">LEMMA</a>) group, a partnership
between the Department of Forest Ecosystems and Society at Oregon State University and the USDA Forest
Service Pacific Northwest Research Station.
</div>
</div>
</div>
<div class="actions">
<div class="ui positive right labeled icon button">
OK, Got it!
<i class="checkmark icon"></i>
</div>
</div>