CCI blog: Protected Area connectivity: an assessment for Aichi target 11

8 Apr 2016

Aichi Target 11 commits the Convention on Biological Diversity (CBD) Parties to conserving 17% of the terrestrial surface of the earth, especially “areas of particular importance for biodiversity” through “well-connected” systems of protected areas. This project aims to assess the degree of connectivity of protected and unprotected habitat patches for forest dependent birds in Africa, highlighting species and locations for which connectivity is a priority for their conservation. For further information, please visit the project page on the CCI website.

The following blog is written by Andy Arnell, Programme Officer, UNEP-WCMC (Project team member)

Determining a strategy for moving forward

Our project kicked off with an inception meeting in December 2015, held at UNEP-WCMC. We started with overview presentations (including other ongoing connectivity projects) and then had discussion sessions on how to: a) create extent of suitable habitat (ESH) maps for each species and b) calculate the connectivity metric. We discussed available forest and land cover datasets that could be used to refine IUCN range maps and how to use habitat preference data in this process.

We decided that Hansen et al. (2013) forest cover would be most appropriate, based on considerations of spatial resolution, possibility of analysing change over time and the continuous values for cover, all of which avoid reliance upon categorical land cover products. No one dataset could provide all of the attributes needed; negative aspects include: choosing cover thresholds, distinguishing between forest types, commission errors from plantations and habitat models only including forest areas. The connectivity metric methodology discussions focused on improving dispersal distance estimates, testing processing on the University of Cambridge’s High Performance Cluster (HPC), calculating inter-patch distances and simplifying processing for species with a prohibitively large number of habitat patches.

Our second meeting at the end of March 2016 discussed progress to date and has agreed a number of approaches and actions for taking the project forward.

In terms of key achievements so far, the team has:

  • Reviewed available land cover and tree cover and ancillary datasets that are available
  • Created >1200 altitude adjusted range maps for forest-dependent African bird species
  • Refined these range maps using two thresholds (15% and 40%) of tree cover using Hansen data
  • Investigated further options to link tree cover thresholds to species’ habitat affiliations in an attempt to limit potential errors (of both commission and omissions) for extent of suitable habitat (ESH) maps. This has included using supplementary data from GBIF species records. So far there appears little clear link between forest affiliations and percentage tree cover. 
  • Adapted Conefor software (for calculating connectivity metrics), to allow focus on specific patches (nodes) of interest, thus saving processing time.
  • Tested Conefor with a small subset of species using the Linux-based High Performance Cluster (HPC) at Cambridge University. This needs to be trialled with a larger subset and then with the final ESH maps when they are produced.
  • Run the proposed connectivity metrics on the ESH maps of a subset of 20 species
  • Created various python scripts (ArcPy) for creating ESH maps and running connectivity metrics, including trial code for a novel “nested connectivity” approach to make calculating large networks (graphs) feasible in the time frame. The results from this novel nested approach still needs further testing so that the results align with those from the normal application of the metric.

Lessons learnt so far

Establishing the best structure for collaboration

There are quite a large number of participants for the size of the project, so organising meetings can be a challenge, but the structure of an afternoon follow-on technical group meeting with a small subset of partners allowed for efficient use of everyone’s time.

We split up the work into two work packages based on some of the main outputs: a) habitat suitability models and b) connectivity maps. This allowed us to cut down on unnecessary communication time.

Overcoming technical challenges

We have had technical challenges from software choice and the large number of species in the analysis. Some tools to run processing steps will have to be converted from ArcPy into R, for compatibility with the Cambridge University computing cluster. This is because even light processing can be time consuming when working with a ~1000 species on a single computer.

Similar challenges arose from the number of nodes for some species but are being mitigated by aggregating habitat patches when very near to each other and by nesting connectivity metric calculations (attempting to process connectivity on a grid square basis, similar to a ‘moving window’ approach).

Preparing for outputs

In due course we plan to produce at least one paper, a draft idea for a follow on project, species maps with connectivity values for each habitat patch, combined connectivity values for protected areas and areas where aggregate connectivity has been most impacted by forest loss since 2000, thus showing applicability as a change metric for connectivity of protected areas (as well as and unprotected forest habitats).

All of these outputs will be available from the project page on the CCI website. In the meantime, if you have any questions about our project, please drop the project lead, Neil Burgess, a line on