A quick and easy method to map riparian vegetation that involves dividing digital imagery into discrete classes using a combination of computer categorization and human interpretation.
(Ratings are given on a 1-5 scale. Click on any rating for an explanation)
Unsupervised classification is the quickest and easiest way to divide digital imagery into discrete classes that represent potential map units, such as vegetation types. In this semi-automated process, the user specifies the number of classes and then, based on natural groupings or clusters of the pixel values, the software assigns each pixel in the image to one of the classes. The software requires no user guidance in this process. After assigning each pixel to a class, the user must identify what vegetation type that class represents.
Typically, the user specifies more classes for unsupervised classification than the actual number of vegetation classes. This allows vegetation with variable appearance to be divided into more than one unsupervised class. During the identification–or assignment phase–the user can merge classes representing the same type of vegetation, as necessary. In some cases, a single class may represent more than one type of land cover, suggesting that a larger number of unsupervised classes may be required to adequately distinguish similar types of vegetation.
Although unsupervised classification is quick and easy, it is not as robust as other types of classification that require user-provided training data, and it may not be able to distinguish similar types of vegetation that another classifier could distinguish. Nevertheless, for relatively simple mapping efforts, such as distinguishing between vegetated and non-vegetated areas, it may produce acceptable results.
Unsupervised classification can also be applied to groups of pixels or “objects” that are derived from segmentation. This is useful when using high-resolution imagery where features on the ground can be larger than a pixel. For more information on objects and segmentation, please refer the Classification and Regression Tree Analysis (CART) method.
Other Classification Methods
- Riparian Supervised Classification
- Object-based Classification: Classification and Regression Tree (CART)
- RSAC Riparian Mapping Tool
Unsupervised classification can be performed on any digital image. It is frequently applied to satellite or aerial imagery, or to indexes (e.g., normalized difference vegetation index [NDVI]) derived from such imagery.
This method produces a new, simplified image where each pixel has a class assignment. That classified image can be used to produce a thematic map showing the distribution of vegetation classes or as an input to more sophisticated processing.
Riparian Application Example
Evans, D.; Vanderzanden, D.; Lachowski, H. 2002.
This study mapped stream geomorphology and riparian vegetation on the Upper Middle Fork of the John Day River in Oregon. The study combined photo interpretation and a smart buffering technique based on channel gradient and sinuosity to establish the width of the riparian area along the stream. Vegetation within the riparian zone was mapped using unsupervised classification and manual photo interpretation.
- Congalton, R.G. 1991. A review of assessing the accuracy of classifications of remotely sensed data. Remote Sensing of Environment 37:35-46.
- Jensen, J. R. 1996. Introductory digital image processing. Prentice-Hall, Inc., Upper Saddle River, NJ.
- Karl, J. W., and B. A. Maurer. 2010. Multivariate correlations between imagery and field measurements across scales: comparing pixel aggregation and image segmentation. Landscape Ecology 25:591–605.
- Lillesand, T. M., and R. W. Kiefer 1994. Remote sensing and image interpretation. John Wiley & Sons, Inc., New York, NY.
The classes produced by unsupervised classification are based on natural breaks in the distribution of pixel color in the image. As a result, the established classes may not distinguish between the features that the user needs to resolve. For example, the color of two riparian vegetation types maybe too similar for the software to separate into distinct classes. In such a situation, another classification method such–as Riparian supervised classification or object-based classification–may be more appropriate.
Unsupervised classification requires remote sensing or GIS software such as ERDAS Imagine or ESRI’s ArcGIS. This method is moderately processing intensive; processing times will vary by dataset size.