Define the study area Once you know how the map will be used, you need to define the region that will be mapped. This can be a fairly easy task but in some cases it can be difficult to reach consensus among project participants. For example, you might find that adding a small portion to a study area will involve the purchase of additional imagery and therefore increase the required resources to conduct the classification. After discussing this situation with project participants you may decide that the additional cost to include the small portion of the study area cannot be justified and the study area will have to be modified. When delineating the study area it may be important to include an area adjacent to the primary study area. This could be done to better understand the transition to the area outside of the study area. An easy way to do this is to create a buffer around the primary study area so adjacent areas are included (Figure 1). Buffers can be created using common GIS or image processing software.
Which classification scheme do I use? A classification scheme effectively defines the legend that will be used for the final map. For example: will the map show forest and non-forest or will it have several or even dozens of different categories? Should the final map categories represent land cover or something else such as land use, habitat, or conservation importance? The way in which the map will be used, and some practical realities, will dictate the content of the classification scheme. One place to start is to look at some of the common classification schemes. There are a large number of classification schemes used for land use and land cover maps throughout the world. Some of the more common schemes are listed in Table 1.
Classification name
URL
Anderson
http://landcover.usgs.gov/pdf/anderson.pdf
National Land Cover Data
http://landcover.usgs.gov/classes.php
FAO Land Cover Classification System
http://www.africover.org/LCCS.htm
Although this guide focuses on land cover there are other attributes that can be mapped such as:
•Vegetation structure
•Land cover disturbance
•Vegetation age (for example, primary and secondary)
•Distribution of taxa (Chorology)
•Land use
•Ecosystems
•Crown closure
An important point to remember is that no matter what classification scheme is selected each class must be well defined and documented. For example, if you have a class called "Forest" you need to specify what constitutes a forest. Do the trees need to be a certain height? How dense do the trees have to be? Without this sort of information it is impossible for the user of the final map to know what the different classes represent.
When choosing an appropriate classification scheme you should decide if compatibility with existing schemes is necessary or desirable. Some advantages of using an existing system is that the classes are already defined and the map you produce can be easily compared with other maps using the same system.
Defining mapping classes is often an iterative process. A balance must be struck between the classes that are desired, based on the map's purpose, and the classes that can be accurately and economically delimited. In general greater detail translates to greater cost in time and money and lower accuracy per class. A good rule of thumb is to select the minimum number of classes that are practical. When you decide upon a classification system it is important that you document the details for each class and apply these class definitions in an objective manner.
When selecting classes, a hierarchical or non-hierarchical approach can be used. In a hierarchical approach, classes are nested such that major classes are broken into sub-classes and these sub-classes can further be broken into more detail. The advantage of such a system is that it can be easily generalized and it is easy to adapt to various scales (the coarser the scale, the more detailed the hierarchy will be). Hierarchical systems are often used when there is a need to maintain conformity with a national-level classification system. A non-hierarchical approach, however; is designed for a specific purpose with a specific scale in mind. The advantage of a non-hierarchical system is that it can be modified to suit a specific application because it is not restricted by the constraints imposed by a hierarchical system. In other words, it can be easier to customize for specific project goals since it can include a mix of detailed and generalized land cover classes.
When defining classes you must decide how to classify mixed features such as transition and mosaic classes. One approach is to define explicitly these classes as mixed or transition and the other is to ignore the fact that classes are mixed and define classes by the most populous feature on the ground within a delineated unit. A third approach that is gaining popularity is called "continuous fields mapping". With this approach each class is mapped into a separate layer (image) showing the percent coverage of that cover type in each pixel. For example, if "Conifer Forest" was a cover type you would create an image with pixel values representing the percentage of conifer forest in that pixel. Instead of a single classified map you would have several maps (one for each class) detailing the percent coverage for each of the respective cover types.
When deciding which approach to use to represent mixed features, you must take into account the nature of the features being mapped and how important mixed classes are relative to the intended use of the map. You should clearly document whichever method is used.
Another point that has to be considered with the classification scheme is the spatial detail that will be mapped. A minimum mapping unit defines the smallest area that is defined on a map. For instance if the minimum mapping unit is 1 hectare then any feature less than 1 hectare would not be delineated as a unique feature. Instead it would be incorporated into another feature. Minimum mapping units can vary from class to class so more important or rare classes would have a smaller minimum mapping unit to ensure that they are not lost as a result of inclusion in another class. In some cases no minimum mapping unit is used and all recognizable features are identified. No matter what approach is used, it is important that it meets the requirements of the needs assessment and should be well documented.