For the evaluation of spatially explicit land use models, model results are often compared to actual land use maps. However, maps can be compared in many different ways, assessing aspects such as agreement on a pixel level or metrics for landscape patterns. The appropriate method for comparison depends on the model at stake, the interest of the user and the purpose of the exercise. Hence there is not one single assessment that fits all, but instead a number of methods is required for different purposes.
In a series of projects RIKS developed and implemented new methods for map comparisons, and facilitated workshops in the use of these methods. These projects include:
The development of neutral land change models that can serve as a benchmark for the evaluation of model results.
The development of a Fuzzy Kappa statistic. This statistic recognizes that some land use classes are more similar than others. In addition it acknowledges that small displacements of land use changes are more correct than large ones.
An overview and implementation of several methods for continuous valued maps.
The development of a feature that facilitates comparing very large maps, such as the whole European Union at a 1km2 resolution.
Many of these developments are described in more detail in publications that appeared in scientific journals or conferences.
The results of our map comparison projects are implemented in the Map Comparison Kit, a stand alone software tool for map comparisons that is continuously updated based on user requirements. The Map Comparison Kit is owned by the Netherlands Environmental Assessment Agency (Currently PBL) and can be downloaded freely from the dedicated support site. In addition this website gives you more information about the algorithms that are included as well as some exercises and documentation.