In this vignette, we provide an example of using the tools within
`redlistr`

to assess the spatial criteria of the IUCN Red List of Ecosystems. These
criteria assess the change in the extent of an ecosystem over time
(Criterion A) and properties of the geographic distribution size
(Criterion B). Both of these criteria require the use of maps of
ecosystem distributions.

We begin by loading the packages we need for the assessment. Please ensure that these are already installed.

```
library(sf)
library(redlistr)
```

We first import the spatial data we want to analyse. In this
vignette, we will be using two example distributions included with the
package. They are mangrove distributions from the northern regions of
Western Port Bay, Victoria, Australia. The first distribution is from
Giri et al., (2011),
and represents mangrove distribution in 2000. The second distribution is
a classification map generated using the `XGBoost`

package
and data from Landsat 8, and represents mangrove distribution in 2017.
Both of these geographic distributions are stored as rasters, and a
raster value of 1 denotes presence.

```
.2000 <- raster(system.file("extdata", "example_distribution_2000.tif",
mangrovepackage = "redlistr"))
.2017 <- raster(system.file("extdata", "example_distribution_2017.tif",
mangrovepackage = "redlistr"))
```

To import your own raster data, you can use the
`raster::raster`

function, which can handle many different
types of georeferenced spatial data. More information on the
`raster`

package can be found here.

To import data as a shapefile (.shp), we can use the
`sf::st_read`

function.

```
library(sf)
<- st_read('./path/to/folder/', 'shapefile.shp')
my.shapefile <- st_read('./path/to/folder/kmlfile.kml') my.KML.file
```

Once these files are imported, they can then be converted into raster
format via the `raster::raster`

or
`raster::rasterFromXYZ`

functions.

We plot the data to view the two distributions to get a general idea of how the distribution might have changed over time.

```
plot(mangrove.2000, col = "grey30", legend = FALSE, main = "Mangrove Distribution")
plot(mangrove.2017, add = T, col = "darkorange", legend = FALSE)
```

We can see that there has been some change in mangrove cover, although the losses appear to be relatively minor.

At this stage it is important to check that the data are georectified
(in the right location on Earth). This can be acheived by simply
plotting the map and checking the coordinates. Otherwise, check the
distribution maps against satellite images, perhaps by using packages
such as `ggmap`

, `plotgooglemaps`

,
`googleVis`

etc. It is also important to check that the data
are suitable for the task at hand.

We now use some of features of `redlistr`

to get some
basic information on the ecosystem distribution. However, before
proceeding it is necessary to know the data by investigating their
resolution (grain size), extent and spatial properties. Specifically,
the coordinate reference system (CRS) of the datasets must be measured
in metres, and is consistent to compare area.

`isLonLat(mangrove.2000) `

`## [1] FALSE`

`isLonLat(mangrove.2000)`

`## [1] FALSE`

```
# If TRUE, they must be reprojected to a projected coordinate system
crs(mangrove.2000)@projargs == crs(mangrove.2017)@projargs
```

`## [1] TRUE`

Reprojection can be done by using the
`raster::projectRaster`

function.

We start by calculating the area of the dataset. Area is calculated here using the pixel count method, where the number of pixels classified as the ecosystem of interest is multiplied by the area of each pixel, giving the area of the ecosystem distribution. The area calculated is provided in square kilometres.

```
.2000 <- getArea(mangrove.2000)
a.2000 a
```

`## [1] 13.6557`

```
.2017 <- getArea(mangrove.2017)
a.2017 a
```

`## [1] 12.3336`

The results: at time 1 (2000), the ecosystem was 13.6557 km2, and at time 2, (2017) the ecosystem was 12.3336 km2.

An additional parameter in `getArea`

is to specify which
class to count if your raster has more than one value (multiple
ecosystem data stored within a single file). However, for further
functions, it may be wise to convert your raster into a binary format,
containing only information for your target ecosystem.

First, if you do not know the value which represents your target
ecosystem, you can plot the ecosystem out, and use the
`raster::click`

function. Once you know the value which
represents your target ecosystem, you can create a new raster object
including only that value. Here’s an example with a dummy raster.

```
# Create dummy raster for example
<- raster(nrows=10, ncols=10)
r <- r
r.multiple values(r.multiple) <- rep(c(1:10), 10)
# If the target ecosystem is represented by value = 5
<- r.multiple == 5 # Has values of 1 and 0
r.bin # Convert 0s to NAs
values(r.bin)[values(r.bin) != 1] <- NA
```

The IUCN Red List of Ecosystems criterion A requires estimates of the magnitude of change over time. This is typically calculated over a 50 year time period (past, present or future) or against a historical baseline (Bland et al., 2016). The first step towards acheiving this change estimate over a fixed time frame is to assess the amount of change observed in your data.

Area change can be calculated using the `getAreaLoss`

function. It simply calculates the difference in area between two
datasets. It can accept data from numbers, `RasterLayers`

or
`SpatialPolygons`

. Make sure to use area measure in km2 if
you are using numbers as the input.

```
<- getAreaLoss(a.2000, a.2017)
area.lost # getAreaLoss(mangrove.2000, mangrove.2017) generates identical results
```

In the Red List of Ecosystems, two methods are suggested to determine the rate of decline of an ecosystem, each of which assumes a different functional form of the decline (Bland et al., 2016). The proportional rate of decline (PRD) is a fraction of the previous year’s remaining area, while the absolute rate of decline (ARD) is a constant fraction of the area of the ecosystem at the beginning of the decline (Bland et al., 2016). These rates of decline allow the use of two or more data points to extrapolate to the full 50 year timeframe required in an assessment.

The annual rate of change (ARC) uses a compound interest law to determine the instantaneous rate of change (Puyravaud 2003).

To estimate the rate of changing using these methods, we use the
`getDeclineStats`

function.

```
<- getDeclineStats(a.2000, a.2017, 2000, 2017,
decline.stats methods = c('ARD', 'PRD', 'ARC'))
decline.stats
```

```
## absolute.loss ARD PRD ARC
## 1 1.3221 0.07777059 0.5972082 -0.5989987
```

Each method represent a different shape of decline. For further information about the choice of each of these methods to extrapolate refer to the IUCN Red List of Ecosystems guidelines (Bland et al., 2016).

Now, it is possible to extrapolate, using only two estimates of an ecosystems’ area, to the full 50 year period required for a Red List of Ecosystems assessment.

```
<- futureAreaEstimate(a.2000, year.t1 = 2000,
extrapolated.area ARD = decline.stats$ARD,
PRD = decline.stats$PRD,
ARC = decline.stats$ARC,
nYears = 50)
```

```
## Warning: 'futureAreaEstimate' is deprecated.
## Use 'extrapolateEstimate' instead.
## See help("Deprecated") and help("redlistr-deprecated").
```

` extrapolated.area`

```
## forecast.year A.ARD.t3 A.PRD.t3 A.ARC.t3
## 1 2050 9.767171 10.12146 10.12146
```

50 years from our first estimate is the year 2050.

As we included all three methods of calculating rate of decline currently included in the package, the results produced shows three estimated areas under the various decline scenarios. It is important to note that this relatively simple exercise is founded on assumptions that should be fully understood before submitting your ecosystem assessment to the IUCN Red List of Ecosystems Committee for Scientific Standards. Furthermore, the guidelines suggest using area estimates from more than two time points to estimate change, and providing a measure of uncertainty if possible. Please see the guidelines (Bland et al., 2016) for more information.

If we were to use the Proportional Rate of Decline (PRD) for our example assessment, our results here will be suitiable for criterion A2b (Any 50 year period), and the percent loss of area is:

```
<- (extrapolated.area$A.PRD.t3 - a.2000)/a.2000 * 100
predicted.percent.loss predicted.percent.loss
```

`## [1] -25.88108`

Criterion B utilizes measures of the geographic distribution of an ecosystem type to identify ecosystems that are at risk from catastrophic disturbances. This is done using two standardized metrics: the extent of occurrence (EOO) and the area of occupancy (AOO) (Gaston and Fuller, (2009), Keith et al., (2013)). It must be emphasised that EOO and AOO are not used to estimate the mapped area of an ecosystem like the methods we used in Criterion A; they are simply spatial metrics that allow us to standardise an estimate of risk due to spatially explicit catastrophes (Murray et al., (2017). Thus, it is critical that these measures are used consistently across all assessments, and the use of non-standard measures invalidates comparison against the thresholds. Please refer to the guidelines (Bland et al., 2016) for more information on AOO and EOO.

For subcriterion B1, we will need to calculate the extent of occurrence (EOO) of our data. We begin by creating the minimum convex polygon enclosing all occurrences of our focal ecosystem.

```
<- makeEOO(mangrove.2017)
EOO.polygon plot(EOO.polygon)
plot(mangrove.2017, add = T, col = "green", legend = FALSE)
```

We can then calculate the area of the EOO polygon.

```
<- getAreaEOO(EOO.polygon)
EOO.area EOO.area
```

`## [1] 610.4198`

The calculated EOO area for this subset of mangroves in Victoria is 610.4198076 km2. This result can then be combined with additional information required under B1(a-c) to assess the ecosystem under subcriteria B1.

For subcriterion B2, we will need to calculate the number of 10x10 km grid cells occupied by our distribution. We begin by creating the appopriate grid cells.

```
<- makeAOOGrid(mangrove.2017, grid.size = 10000,
AOO.grid min.percent.rule = F)
plot(AOO.grid)
plot(mangrove.2017, add = T, col = "green", legend = FALSE)
```

Finally, we can use the created grid to calculate the AOO for our mangroves.

```
<- length(AOO.grid)
n.AOO
# the getAOO function can also be used to directly get the AOO
# n.AOO <- getAOO(mangrove.2017, grid.size = 10000,
# min.percent.rule = T, percent = 0.1)
n.AOO
```

`## [1] 11`

Although we have a number for AOO 11, this may not be the actual
minimum AOO for this ecosystem, because the placement of the AOO grid
will influence this number. To take this into account, we have included
a few additional functions. We will be using
`gridUncertainty`

in this vignette.

`gridUncertainty`

simply moves the AOO grid systematically
(with a small random movement around fixed points), and continues
searching for a minimum AOO until additional shifts no longer produce
improved results.

```
<- gridUncertainty(mangrove.2017, 10000,
gU.results n.AOO.improvement = 5,
min.percent.rule = F)
# If it takes your computer very long to run this command, consider reducing
# n.AOO.improvement
$min.AOO.df gU.results
```

```
## n.splits min.AOO
## 1 1 10
## 2 2 10
## 3 3 10
## 4 4 10
## 5 5 10
## 6 6 10
```

The result here shows that the minimum AOO does indeed change in response to where the AOO grid is placed, and that our original AOO value of 11 is not the true minimum AOO. The true minimum AOO (and the one which should be used for the assessment) is 10.

It is also possible to plot out the grid which was used to produce the minimum AOO result, which can be used to report our final results.

```
plot(gU.results$min.AOO.grid$out.grid)
plot(mangrove.2017, add = T, col = "green", legend = FALSE)
```

The results of the above plot from `gridUncertainty`

is a
shapefile, which can be exported with the `raster::shapefile`

function.

In addition to the size of the grids used (which will be different
for species assessments), there is also an option in the
`makeAOOGrid`

function to specify whether a minimum of
percent of the grid cell area must be occupied before they are included
as an AOO grid. Typically, this is only used in very special cases for
ecosystems with highly skewed distribution of patch sizes (Bland
et al., 2016).

Here, we demonstrate the differences between including, or not including the one percent rule:

```
<- makeAOOGrid(mangrove.2017, grid.size = 10000,
AOO.grid.one.percent min.percent.rule = T, percent = 1)
plot(AOO.grid.one.percent)
plot(mangrove.2017, add = T, col = "green", legend = FALSE)
```

There is an additional parameter - `percent`

- which
adjusts the threshold for the AOO grids. Here, we set it to 0.1% to
demonstrate its functionalities.

```
<- makeAOOGrid(mangrove.2017, grid.size = 10000,
AOO.grid.min.percent min.percent.rule = T, percent = 0.1)
par(mfrow = c(2,2))
plot(AOO.grid, main = 'AOO grid without one percent rule')
plot(mangrove.2017, add = T, col = "green", legend = FALSE)
plot(AOO.grid.one.percent, main = 'AOO grid with one percent rule')
plot(mangrove.2017, add = T, col = "green", legend = FALSE)
plot(AOO.grid.min.percent, main = 'AOO grid with one percent rule at 0.1%')
plot(mangrove.2017, add = T, col = "green", legend = FALSE)
```

We have demonstrated in this vignette a typical workflow for
assessing an example ecosystem under the RLE. The methods outlined here
can easily be adapted for the Red List of Threatened Species as well,
with slight adjustments in the parameters (specifically the grid.size
parameter for `getAOO`

or `gridUncertainty`

). We
hope that with this package, we help ensure a consistent implementation
for both red lists criteria, minimising any misclassification as a
result of using different software and methods.