(Approach)

The main objective of the project was to assess the feasibility of providing, and to provide, a set of data on distribution whose quality was high enough to allow for their use in conservation and management. Behind this ambition was the acknowledgement that geographic analysis is today a powerful tool in decision making, but also the assumption that traditional distribution maps, representing the Extent of Occurrence (EO)of animal species, are of little use in practical conservation and management: their resolution is generally not sufficient to define the species' real Area of Occupancy (AO), and their usefulness in accurate spatial analysis is thus limited. A detailed knowledge of the areas were the species really occur is indeed a fundamental step to address all conservation related issues, and one of the most important aspects to be taken into account in their conservation, in establishing and managing protected areas, and in many programmes of environmental evaluation and management. Of all environmental components, fauna is one of the most difficult to show on a map, because of its mobility and of the difficulties in locating individuals of a species: the production of detailed distribution maps based on true locations can be an enourmous task in terms of effort, especially over large areas. The approach followed in this project was that of maximising the use of the information already available for each species, to produce "true distribution" maps, in which the areas of presence of a species within its general Extent of Occurrence was to be recorded.

As for others biotic components, the distribution of animals species is affected by the distribution of other environmental variables. GIS offer today the opportunity to integrate data on ecology of each species with maps and layers pertaining to a variety of environmental variables: by this process, the environmentally suitable areas can be identified for each species. This sole analysis, however useful, would not give an accurate enough picture of the distribution for many species, as the presence of many species is indeed affected by other ., such as historical (i.e. palaeogeographical) factors, past and present human influence, etc. These factors can be taken into account by integrating the existing distribution data in the process of identifying the areas of expected presence. Only after these areas have been accurately identified, GIS capabilities can again play an important role, by offering a practical tool and effective support to conservation and management actions.

 

Acquiring Extent of Occurrence

The first step in the analysis for each species was thus the acquisition of the existing distribution data, which made up the basis for the subsequent phases. In this phase, distribution maps produced by specialists were considered the best descriptor of each species known Extent of Occurrence. Although in some cases distribution maps can be inaccurate or not updated, they were thus considered the best compromise between the accuracy requirements of the project and the need of coming out in reasonable times with a fully operational tool.

Distribution maps were integrated in the process because of two main reasons: first, as mentioned before they were considered the best descriptor of all the constraints which affect the present distribution of each species anc which cannot be described in terms of distribution of environmental variables. Secondly, they were considered themselves as a source of information on the ecology of each species, from which an estimate of the average ecological conditions selected by the species can be derived.

 


To be used in subsequent analyses, distribution maps had to be converted into GIS layers. For each species, an ARC/INFO polygon coverage defining its distribution range was thus obtained:

1) Bibliographic search

2) Specialists advice

3) The distribution map is referenced to an existing map of Africa at the adequate scale for that species

4) Digitalisation of the map either using a digitiser and/or a scanner

5) Creation of an operational ARC/INFO coverage.

 

Defining Area of occupancy

In the following phase two different, complementary ways of building expected presence models were followed. In fact, two different approaches were followed, a deductive one and an inductive one, resulting in the production for each species of two complementary distribution models.

Both distribution models produced can be used like basis for subsequent analyses, a first set of which, concerning basic parameters and initial assessements of the conservation status, were performed during the project. These included an estimate of the extension of the areas of expected presence, of their degree of fragmentation and on the percentage included in existing protected areas.


  • Categorical Discrete Model

    The first approach consisted basically of the integration of a data set describing the distribution of environmental conditions throughout the continent with the available data on each species' ecology. Based on the information on the species ecological requirements, a suitability rank was assigned to each category present in a series of environmental layers. A predefined, deterministic combination of the ranks allowed for the integration of different layers, to eventually produce a layer describing the different areas in terms of overall suitability for the species.

     

     

    The steps:

    1) Bibliographic search and specialist advice

    2) Definition of the species ecological needs

    3) Overlay of the environmental layes with EO and utilization of the species-habitat relationships within the GIS

     

    Environmental variables:

     

    White's vegetation map of Africa (FAO/UNESCO)
    Global Land Cover Characterisation (GLCC)
    Population densities per district (Africa Data Sampler)
    Hydrological network (DCW)

     

    The result of this process is a map in which a discrete category of suitability is assigned to each portion of the geographic area of interest. The EO has been used as a first means to discriminate between different environmental classes. The ranking of the environmental categories has been assigned only on the basis of and to those categories which falls inside it limits.

    In the final model the boundaries of the Extent of Occurrence as extracted from the distribution maps delimits areas of expected presence/absence, which are thus distinguished from suitable areas falling outside the EO; the latter are considered as areas of "potential" presence, i.e. those areas which appear suitable as the environmental conditions are similar to those present where the species is present, but for which the presence cannot be expected on the basis of the current knowledge of its distribution limits.

     

  • The Probabilistic Continuous Model

    The other, alternative approach sought to define the suitability of each area on the basis of the similarity of its environmental conditions with the conditions found inside the EO; behind this approach is the assumption that the average environmental situation found inside the limits of the EO is, to a certain degree of accuracy, an approximation of the environmental conditions preferred by the species.

    The EO is used to derive the ecological profile, or "signature", of the species, by measuring average value of a series of environmental parameters. These ecological profile is then compared to the conditions found in each portion of the area of interest. With a multivariate statistical approach, namely the Mahalanobis distance, several continuous variables can be taken into account in the model. The result is a continuous surface of values, which give an estimate of the environmental suitability of each point based on the similarity to the conditions found where the species live.

     

    Environmental variables:

    NDVI monthly averages
    DTM
    Population density surfaces (UCSB)
    Distance from water

     

    Validation

    For a subsample of the species considered by the project, further investigations were conducted in order to assess the general reliability of the output of the modelling exercise. To this end, data on presence/absence of the species were directly collected on the field in about 400 plots, randomly scattered over four sample countries.

     

    - Field work carried out in four selected countries in Africa (Botswana, Cameroon, Morocco, Uganda).
    - 427 plots were allocated at random within the four countries
    - The presence (or absence) of each species at each of the predetermined points was verified by:

    direct observation
    in loco collection of publications and scientific reports
    interviews with local authorities and inhabitants

    - In each country a team composed by a researcher from a local Institution and one IEA staff member carried out the field work