Objectives
mapme.biodiversity facilitates statistical data analysis
for protected areas around the globe. It supports a high number of
biodiversity related datasets and associated indicators that can be
utilized to monitor and evaluate the effectiveness of protection
efforts. Several indicators are available at regular intervals for
almost two decades (2000 to 2020). This allows users to analyse spatial
and temporal dynamics of their biodiversity portfolios. The package
abstracts away repetitive tasks, such as temporal and spatial selection
of resources. This allows a seamless approach to quantitative data
analysis even for very large (potentially global) portfolios where users
are enabled to focus on the aims of their analysis. The package has been
tested on Microsoft Azure’s cloud infrastructure as well as on local
machines. The internal framework was designed to allow an easy process
to provide extensions in the form of additional resources and
indicators, unlocking the potential for a future growth in supported
datasets and Pull-Requests are highly appreciated. For geographic data
analysis, the package uses sf for operation of vector
data and terra for raster data.
mapme.biodiversity package
mapme.biodiversity provides a standardized interface to
download and analyse a great variety of biodiversity related spatial
datasets allowing users to focus on the aims of their analysis. The
sometimes cumbersome process of handling different spatial data formats
and the spatial and temporal selection is handled internally. Many
organizations provide value-added datasets related to biodiversity.
These organizations often use different technology stacks to distribute
their data. mapme.biodiversity contains simple routines to
communicate with these different backends to provide a seamless access
to the data. Once the desired resources have been made available
locally, users can decide which indicators they want to calculate and
fine-control of these routines is provided.
Functionalities
Currently, the package offers several functionalities, which should ideally be used in a consecutive manner in order to realize a seamless analysis workflow:
- initiate a portfolio based on an sf object
- get resources for the spatio-temporal extent of the portfolio
- calculate indicators based on the available resources for each asset in the portfolio
- write the results to disk as a GeoPackage and use it with other Geo-Spatial software, or
- conduct your statistical analysis in R
Inputs, Outputs
an sf object containing only geometries of type
'POLYGON'with arbitrary metadata and additional information about the temporal extent and an output folder structureraster and vector resources matching the spatio-temporal extent of the portfolio will be downloaded and made available locally. These are necessary inputs for the subsequent calculation of indicators, but the raw resource of course also can be used, e.g. for custom visualizations or analysis. Importantly, the same resource directory can be used for different portfolios because matching resources are figured out during run time. Thus, there is no need to store multiple copies of the input resources.
the results of the indicator calculation will be added to the portfolio object as nested list columns. This approach makes it feasible to support a variety of indicators with differently shaped outputs (e.g. time variant vs. invariant indicators). When the analysis is done in R, this does not pose serious limitations, because the desired indicator can easily be unnested via
tidyr::unnest(). However, if the data was to be shared to use it in other geospatial software (e.g. QGIS), a routine to write a portfolio object as a GeoPackage to disk is provided. Each indicator will be written to an independent table and a unique identifier allows joining the attributes with the geometries later.
Limitations
a potential limiting factor for now is the processing of single very large polygons. While the terra package provides a memory-save framework to process large raster extents, RAM overflows could occur when several large polygons are processed in parallel. We advise to process very large polygons sequentially or substantially reduce the number of cores.
on Windows, parallel processing currently is not supported. While it has been thoroughly tested that the provided routines also work on Windows machines, the sequential processing of large numbers of polygons can take a while. If you wish to process large portfolios in parallel, please make sure to use a UNIX based operating system (e.g. by using a docker image).
we took a great effort to evaluate the most efficient processing routines for each indicator. If you submit a new indicator where you have a more efficient routine in mind that is currently not supported by the package, please contact the maintainers via e-mail, issue or pull-request and we will happily discuss the options how to integrate your routine into the wider framework
We are planning to add new features and to extend the functionality of mapme.biodiversity, and to address these limitations best possible.