Awesome

mapme.biodiversity <img src="man/figures/logo.png" align="right" height="110"/>

About

Biodiversity areas, especially primary forests, provide multiple ecosystem services for the local population and the planet as a whole. The rapid expansion of human land use into natural ecosystems and the impacts of the global climate crisis put natural ecosystems and the global biodiversity under threat.

The mapme.biodiversity package helps to analyse a number of biodiversity related indicators and biodiversity threats based on freely available geodata-sources such as the Global Forest Watch. It supports computational efficient routines and heavy parallel computing in cloud-infrastructures such as AWS or Microsoft Azure using in the statistical programming language R. The package allows for the analysis of global biodiversity portfolios with a thousand or millions of AOIs which is normally only possible on dedicated platforms such as the Google Earth Engine. It provides the possibility to e.g. analyse the World Database of Protected Areas (WDPA) for a number of relevant indicators. The primary use case of this package is to support scientific analysis and data science for individuals and organizations who seek to preserve the planet biodiversity. Its development is funded by the German Development Bank KfW.

Installation

The package and its dependencies can be installed from CRAN via:

install.packages("mapme.biodiversity", dependencies = TRUE)

To install the development version, use the following command:

remotes::install_github("https://github.com/mapme-initiative/mapme.biodiversity", dependencies = TRUE)

Available resources and indicators

Below is a list of the resources currently supported by mapme.biodiversity.

Next, is a list of supported indicators.

Usage example

{mapme.biodiversity} works by constructing a portfolio from an sf object. Specific raster and vector resource matching the spatio-temporal extent of the portfolio are made available locally. Once all required resources are available, indicators can be calculated individually for each asset in the portfolio.

library(mapme.biodiversity)
library(sf)

## Linking to GEOS 3.13.0, GDAL 3.9.2, PROJ 9.5.0; sf_use_s2() is TRUE

Once you have decided on an indicator you are interested in, you can start by making the required resource available for your portfolio. Using mapme_options() you can set an output directory, control the maximum size of polygons before they are chunked into smaller parts, and control the verbosity of the package.

A portfolio is represented by an sf-object. It is required for the object to only contain geometries of type POLYGON and MULTIPOLYGON as assets. We can request the download of a resource for the spatial extent of our portfolio by using the get_resources() function. We simply supply our portfolio and one or more resource functions. Once the resources were made available, we can query the calculation of an indicator by using the calc_indicators() function. This function also expects the portfolio as input and one or more indicator functions. Once the indicator has been calculated for all assets in a portfolio, the data is returned as a nested list column to the original portfolio object. The output of each indicator is standardized to common format, consisting of a tibble with columns datetime, variable, unit, and value. We can transform the the data to long format by using portfolio_long().

mapme_options(
  outdir = system.file("res", package = "mapme.biodiversity"),
  chunk_size = 1e6, # in ha
  verbose = FALSE
)

aoi <- system.file("extdata", "sierra_de_neiba_478140_2.gpkg", package = "mapme.biodiversity") %>%
  sf::read_sf() %>%
  get_resources(
    get_gfw_treecover(version = "GFC-2023-v1.11"),
    get_gfw_lossyear(version = "GFC-2023-v1.11"),
    get_gfw_emissions()
  ) %>%
  calc_indicators(calc_treecover_area_and_emissions(years = 2016:2017, min_size = 1, min_cover = 30)) %>%
  portfolio_long()

aoi

## Simple feature collection with 4 features and 8 fields
## Geometry type: POLYGON
## Dimension:     XY
## Bounding box:  xmin: -71.80933 ymin: 18.57668 xmax: -71.33201 ymax: 18.69931
## Geodetic CRS:  WGS 84
## # A tibble: 4 × 9
##   WDPAID ISO3  assetid indicator        datetime            variable unit  value
##    <dbl> <chr>   <int> <chr>            <dttm>              <chr>    <chr> <dbl>
## 1 478140 DOM         1 treecover_area_… 2016-01-01 00:00:00 emissio… Mg    4296.
## 2 478140 DOM         1 treecover_area_… 2016-01-01 00:00:00 treecov… ha    2370.
## 3 478140 DOM         1 treecover_area_… 2017-01-01 00:00:00 emissio… Mg    4970.
## 4 478140 DOM         1 treecover_area_… 2017-01-01 00:00:00 treecov… ha    2358.
## # ℹ 1 more variable: geom <POLYGON [°]>

Using cloud storages

{mapme.biodiversity} leverages GDAL’s capabilities for data I/O. For users of this package, that means that integrating a cloud storage is as easy as setting up a configuration file and changing the outdir argument in mapme_options(). While you could also decide to use environment variables, we recommend to set up a GDAL config file. You can find GDAL’s documentation on this topic here.

Suppose that we want to use an AWS S3 bucket that we control to write resource data to. Let’s assume this bucket is already set up and we wish to refer to it in our R code as mapme-data. The GDAL configuration file should look something like this:

[credentials]

[.mapme-data]
path=/vsis3/mapme-data
AWS_SECRET_ACCESS_KEY=<your-access-key>
AWS_ACCESS_KEY_ID=<your-access-id>

The connection will be handled based on GDAL’s virtual file system. You can find documentation on specific options for your cloud provider here.

Ideally, you would also set the following in the .Renviron file in your user’s home directory to ensure that GDAL is aware of this configuration when an R session is started:

GDAL_CONFIG_FILE = "<path-to-your-config-file>"

Then, in your scripts set the outdir option to the value specified with the path variable in the configuration file:

mapme_options(outdir = "/vsis3/mapme-data")

A note on parallel computing

{mapme.biodiversity} follows the parallel computing paradigm of the {future} package. That means that you as a user are in the control if and how you would like to set up parallel processing. Since {mapme.biodiversity} v0.9, we apply pre-chunking to all assets in the portfolio. That means that assets are split up into components of roughly the size of chunk_size. These components can than be iterated over in parallel to speed up processing. Indicator values will be aggregated automatically.

library(future)
plan(cluster, workers = 6)

As another example, with the code below one would apply parallel processing of 2 assets, with each having 4 workers available to process chunks, thus requiring a total of 8 available cores on the host machine. Be sure to not request more workers than available on your machine.

library(progressr)

plan(cluster, workers = 2)

with_progress({
  aoi <- calc_indicators(
    aoi,
    calc_treecover_area_and_emissions(
      min_size = 1,
      min_cover = 30
    )
  )
})

plan(sequential) # close child processes

Head over to the online documentation find more detailed information about the package.