PREPARING THE MAPS

This section describes how to prepare your data so that they be used as ecogeographical variables (EGV) in Biomapper. This involves:
  1. Selecting relevant/available data
  2. Importing/converting them into a GIS application
  3. Deriving ecologically-relevant variables from them
  4. Converting them to Biomapper/Idrisi format



Operations

Biomapper requires to kinds of input:
  • Ecogeographical Variables (EGVs): These are the variable defining the species ecological niche, or predictors.
  • Species map: This map holds the locations where the species has been sighted or detected.

Selecting relevant EGVs

First you have to choose which data are relevant (and available: field sampling, government databases, etc.) for your focal species. ENFA is neither sensible to unrelevant data nor to their input order. All useful information will be extracted and summarised into the ecological niche factors. Thus, don’t fear to use too much data. Unrelevant data will increase the computation time and the memory needs, but should not significantly influence the accuracy of the result. Nevertheless, they could decrease the generalisation of your model.

Sources of spatial data

  • (Nothing yet)

Importing to GIS

Some data will be readily available in the correct format. Some will not. There is no way here to describe how importing data to the GIS can be done. Each database will probably require a different kind of treatmen. Typically, 75 to 90% of the time devoted to a project will be devoted to this task. Data verification is also part of this process.
The Biomapper package provides a tool to convert table data into Biomapper maps: the module Convertor.

Deriving ecologically-relevant EGVs

Available data are often collected for other purposes than to model your favourite species. Most often, you need to derive more relevant variables from them. This critical to obtain a good model. By deriving variables more thightly related to the species requirements, you can often improve the model much more than by collecting more observations. So, do not hesitate to devote enough energy to this step.
This is typically performed in a GIS application (Idrisi, ArcGIS), but Biomapper provides several useful modules as well, some of which offering analyses available nowhere else:
CircAn
Performs neighbourhood analyses
DistAn
Computes various distance-related analyses
MapCalculator
Combines several maps and perform mathematical operations on them
Booleanisator
Transforms a continuous map into several binary maps
MapManager
Converts between data types, etc.

BigGroup
Groups identical, adjacent cells together
GroupStat
Performs group analyses

Making EGVs quantitative

The ENFA requires quantitative EGVs. However, data may belong to three types, each needing a different preparation method:

  • Quantitative data

  • Examples: Altitude, slope, mean July temperature, amount of yearly rain, etc..
  • These data are ready to be used by the ENFA. You may want to transform them in order to make them more relevant to the species (smoothing, buffering, averaging, etc.) but they are technically acceptable.

  • Qualitative data

  • Examples: Soil type, land cover, vegetation type, etc..
  • These maps cannot be used in this format. There are two methods to extract quantitative data out of them. Let's imagine we are dealing with a map of vegetation types to illustrate them:
  • The first one is to determine which is the important feature of the represented categories, to reorder them using a semi-quantitative scale determined by this feature and to code it with a integer numerical value. For instance, vegetation could be ordered by height of the canopy as follows: 1.Bare ground, 2.Grassland 3.Bushes 4.Forest
  • The second method implies first to transform the map into several Boolean maps, each describing a relevant category. All available categories can be used, or only a few of them, or it is also possible to pool several categories into one Boolean map. These operations can be easily done by using Idrisi's Image calculator. These maps will then be used as described in the next paragraph. For instance, we could consider that only forests and bushes are relevant for our species and compute two Boolean maps representing presence/absence of these two entities.

  • Boolean data (=binary data)

  • Examples: Presence/absence of a species, cultivated areas, towns, roads, lakes, forests, etc..
  • These maps cannot be used in this format. There are four main methods to extract quantitative data out of them:
  • The first method is preferentially used when the map represents a resource or a shelter for the focal species, when you guess that the species needs a minimal amount of it or cannot live when it is too important in the landscape. The method implies to choose a radius of influence and to compute the frequency of occurrences into a circular area around the focal cell. Generally, the radius is chosen in order to produce a circle area equal to the home range of the species. This analysis can be performed by the BioMapper's module Circan, option Frequency (Maps/Contextual/Circular analysis). For instance, the quantitative map could represent the frequency of forests into a 1km radius buffer around each cell.
  • The second method is typically biological. Many species live at the interface between two habitats (forest edges, lake shores, coasts, etc.). The principle is then to compute the length of the boundary parting them in a circular area defined as in the first method. This can be performed by the BioMapper's module Circan, option Edge length (Maps/Contextual/Circular analysis). A few other landscape-ecology indices are available in Circan.
  • The third method is ideal when the map represent the locations of disturbance sources (towns, roads, etc.). It consists to compute a distance map, attributing to every cell the distance to the nearest occupied cell. This can be done with BioMapper's Distan module (Maps/Contextual/Distance) or Idrisi's Distance module. The real distance can be used, or a cost distance or even a function of the real distance (to lower the influence of the longer distances).
  • The fourth method is more difficult to justify biologically. It consists in smoothing the Boolean map to get values between 0 and 1. This can be done with Idrisi's Filter module or BioMapper's module Circan, option Gaussian or Mean (Maps/Contextual/Circular analysis).

Preparing the species map

  • Prepare a Boolean map (containing 0 and 1 only), the 1 indicating the cells where the species is present (Maps/Convert). Alternatively, you can also attribute an integer weight to the species observations. In this case, replace the 1 by this weight. But you must be sure that your data are the product of a homogeneous collecting effort.
If you have got Idrisi, you can import a point vector of the sightings, and then rasterize it. ESRI Shape files are easy to convert into Idrisi.
If you do not have Idrisi, the easiest way is to somehow get a table (a tab-separated text file, or ASCII), where each row represents a sighting and columns represent X-coordinate, Y-coordinate, Weight. (if no weight, then fill this columns with "1"). You can then make a Biomapper map with the module Convertor.

Converting to Biomapper format

If you did not prepare your maps within Idrisi or Biomapper, you will have to convert them.

Preparing for the ENFA

Now that your EGV maps are at hand, you must prepare them for the ENFA. Basically, it means to make them overlayable and, as far as possible, unimodal and symmetrical. You will also verify that there are no discrepancies between them. You will create a project that will be used for all the subsequent operations.
Here is the step-by-step procedure:
  • Create a list of the ecogeographical maps (Files/Ecogeographic maps/Add maps...)
  • Save your project (Files/Save project as...). It is wise to always work with a named project in Biomapper.
  • You will perhaps have to change the background value for each map; this value will not be used in the ulterior analyses. Furthermore, it is better to assign a "Biomapper extension" to each map, indicating the kind of its data (Boolean, number, frequency, ...). Although it is not strictly required, it will allow you to find more quickly the relevant maps in the jungle of all of them.
  • Normalise the ecogeographical maps (Maps/Formatting/Transformation) using the Box-Cox function. It may happen that the transformation generates a constant map (all cells have the same value) or nearly-boolean maps (almost all cells belong to a small range of values.); as these maps would create problems during the subsequent computations, it is better to revert to the original map (untransformed) or to discard them. If you have checked Replace, the new normalised maps will replace the original ones in the project's EGV list (they will now have a "-box" suffix).
  • Verify the consistency and usability of these maps (File/Ecogeographic maps/Verify maps). This operation verifies that all maps have the same background and non-background cells and that there are neither constant of nearly-boolean maps. If any, it will list the problematic maps. Discrepancies are critical. When you get a "warning", you can still try to continue, and if something goes wrong later, come back to this step to fix the problem.
  • Add the species map among the Work Maps (Files/Work maps/Add map...) and mark it as "Species map" (right click, Mark as species map)
  • Save the project