To content
Read more about how we use cookies on gu.se
# From model to map

**There are different alternatives for moving from models to maps. In general, the method chosen is context dependent since different software allow for different alternatives.**

There is a wide range of not only operating systems (Windows, Mac OS, Linux), but also statistical (R, S-plus, Statistica, SPSS, Matlab, etc) and GIS-software (ArcGIS, MapInfo, QGIS, GRASS, etc) that need to share common data formats allowing information to be freely exchanged between the statistical software and the GIS in particular.

Regardless of software platform the overarching principle is that the statistical relationship between the response and the predictor variable(s) is applied on “new” spatial environmental predictor(s) for the area that the map should cover (**Göran**: I am unsure what term we should use. This refers to GIS-layers of the predictor variables…** Martynas**:I think its ok.).

**Point files**

The most direct approach for moving from models to maps is to include the spatial predictor data (environmental layers) as point files in the statistical software and use it as basis for model predictions. Subsequently, the predicted point files are re-imported into a GIS platform and converted to a suitable format, often raster grids. These raster grids then show the model response, be it a quantitative/abundance or qualitative/distribution response. However, this approach is computer intensive, especially for large and/or high resolution data and may therefore require specifically dedicated high performance computers.

**Look-up tables**

A generally faster and more efficient way of moving from models to maps is to use look-up tables. Look-up tables are in principle a simplified realisation of the relationship between the response and the predictor layer(s) which is used inside the GIS software to transform the spatial predictor layer(s) into a map of the model response. The look-up technique requires that the statistical software allows for look-up tables to be produced, which most software should be able to do with a bit of help from the modeller. However, constructing the GIS-script for applying the took-up table requires a certain amount of specialist knowledge.

Other** techniques**

A third and straight forward approach for creating map predictions is also possible when using certain statistical techniques. These techniques include for example regression based approaches were parametric coefficients are obtained (e.g. generalized linear models GLM). In these instances the model and coefficients are used in algebraic solutions such as ArcGIS raster calculator.

There are several complete modelling packages that may assist in the process of moving from models to maps (e.g. Maxent, GRASP, Biomapper, DIVA-GIS, MGET??) (*GÖRAN**: Should we mention these also? We are not doing a full review I think, but maybe something on available techniques is appropriate? What do you say? Martynas: "I think it is enough".*

*Discuss briefly issues as:• Probabilities/limitations• Short text and some useful GiS-links*