-
Notifications
You must be signed in to change notification settings - Fork 3
/
Help.Rmd
103 lines (88 loc) · 7.95 KB
/
Help.Rmd
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
---
output:
html_document: default
---
# Extra help for running Goose-GMSE simulations
## Methodology
This model is a user interface to the model presented in [SNH Research Report 1039](https://www.nature.scot/snh-research-report-1039-development-population-model-tool-predict-shooting-levels-greenland).
Please refer to this report for full details on the background, model and the methods used. The information provided here
focuses only on the use of the interface itself and details relevant to that.
## Input file requirements
Overall, the input file represents the latest observed data available, i.e. all goose counts, climate variables up to a
given September, with the model projecting a required number of years forward from the last year listed.
<span style="color: red;">
A "year" as referred to in the input data (i.e. a line in the input file) represents a winter, as opposed to a calendar
year. For example, __a line for year 2015 reflects winter 2015-16, i.e. November 2015 to March 2016 in terms of goose
counts on Islay__. Any reference to "the current year", "the current winter", or "given winter" in the following should
be interpreted as such.
Please also note the point about relevant Greenland climate values per line below.
Thus, for example, in September 2016 all values for the line 2015 (i.e. winter 2015-16) should be available, and model
runs project goose counts for winter 2016-17 onwards.
</span>
- A base data (input) file __must__ be loaded and formatted correctly.
- The input file must be in .csv, .xls or .xlsx format only.
- The input file content must be layed out as the _example_data.csv_ file, and must be in order of year (first column).
Years have to be sequential with no missing year values.
- Note that whole blank lines are not allowed in the input file, and that any missing values in
.xls or .xlsx input files have to be kept blank, and should not contain 'NA' text. The latter point does not apply to
.csv input files; in this case, NA text values will be correctly interpreted as missing.
- Projections for future years are taken from the last year (last line) in the data file, and the user is expected to add
new observed data as this becomes available.
- Although it is expected that all data for all past/present years is available (i.e. all rows are complete), missing
values are allowed for some columns. Please refer to the [report](https://www.nature.scot/snh-research-report-1039-development-population-model-tool-predict-shooting-levels-greenland)
for further details on how missing values are interpreted and handled (e.g. missing data are assumed to be equal to prior
means, etc.).
- <span style="color: red;">__NOTE: Greenland climate data, Iceland cull and Greenland cull numbers__. Because (as set
out above) a line in the input file refers to a given winter (Nov-Mar on Islay), and the model predicts the _following_
winters, the values needed for both Greenland climate ("AugRain" and "AugTemp"), and those for numbers culled in both
Iceland and Greenland, are effectively those for the "following year", i.e. the summer months following the given year.
For example, for a line of observed data for __2013__, Greenland weather values should be those for __August 2014__,
and Iceland and Greenland cull numbers should be those for summer/autumn 2014.</span>
The columns in the input file should be titled (first row only, with the name as specified under "Header" below) and
arranged as follows (note that any deviations from the above format may cause the model not to run properly and/or
projections to be unreliable):
| Column | Header | Required | Description |
| :--- | :--------: | :-------: | :-----------|
| 1 | Year | __YES__| Year in four-digit format. Has to be sequential over successive rows (no gaps). Represents a winter (Nov-Mar in terms of geese on Islay), and is referred to below as the "given winter" (e.g. line marked 2015 refers to winter 2015-16).|
| 2 | November | No | Monthly average goose count on Islay in given winter
| 3 | December | No | Monthly average goose count on Islay in given winter
| 4 | January | No | Monthly average goose count on Islay in given winter
| 5 | February | No | Monthly average goose count on Islay in given winter
| 6 | March | No | Monthly average goose count on Islay in given winter
| 7 | Count | __YES__| The average over-winter count of geese on Islay (i.e. November-March). If monthly counts are included (columns 2-6), this figure needs to be equal to the average of those columns.
| 8 | IcelandCull| No | The number of geese culled on Iceland __in the summer following the given winter__.
| 9 | IslayCull | __YES__| The number of geese culled on Islay in given winter.
| 10 | GreenlandCull | No | The number of geese culled on Greenland __in the summer following the given winter__.
| 11 | AIG | No | The area of improved grassland in hectares available on Islay in given winter.
| 12 | IslayTemp | No | The average temperature (degrees C) on Islay in given winter, calculated as the average over October to March inclusive, in given year.
| 13 | AugRain | No | The average rainfall (mm) in Greenland __during the August following given winter__.
| 14 | AugTemp | No | The temperature (degrees C) in Greenland __during the August following given winter__.
## Interpretation of output
### Uncertainty
The output graph and summarised output are the result of simulations including observation uncertainty and some process
uncertainty regarding future environmental conditions. Thus, it should be noted that the output and any inferences from
them are inherently variable (i.e. contain a degree of uncertainty). Thus, any interpretation and recommendations
derived from this output should take account of this uncertainty: neither the projected population sizes nor the numbers
culled should be taken as an absolute recommendation for numbers to be culled.
### Number of simulations
In particular, the interpretation of the model output should take account of the number of simulations that was used to
produce the output: a lower number of simulations will limit the ability of the user to make fair judgements of the
likely variability in projected population sizes. To minimise processing time, we suggest that trial runs of the model
could be made with a small number of simulations, before a final run with a larger number of simulations (e.g 100), on
which inferences could be based.
## Extra information on simulation parameters
### Number of years projected
Because of the various sources of uncertainty included in the model (see above and the report), it is important to note
that it should be assumed that uncertainty is going to be greater further in the future. For example, the model
currently makes the very simplistic assumption that climatic conditions will change according to linear trends in the
future, with trend parameters and error estimated from input data. Further details of this can be found in the
accompanying [report](https://www.nature.scot/snh-research-report-1039-development-population-model-tool-predict-shooting-levels-greenland"),
Thus, it is best to concentrate interpretation on 1-2 years in the future. Moreover, it is very important that
simulations are re-run as more data becomes available).
### Maximum cull per year
It is very important to note that this parameter represent the maximum number culled only and is not a specific
recommendation (i.e. it is set by the user of the model). The actual number of individuals culled in each year
projected may be lower than this and the mean, variation and range of this across simulations is provided in the output
table. In some cases (e.g. when the population size exceeds the population target by a large amount), it is likely that
the number culled simply equals the maximum. __Note that this reflects a situation when there is no error in the number
actually culled (e.g. marksmen are able to hit their target exactly__.