A reliable method to map riparian vegetation by dividing digital imagery into classes based on computer interpretation of training data.

• Difficulty: 2
• Technical level: 2
• Expense: 1
• Scale: Variable
• Accuracy: 2

(Ratings are given on a 1-5 scale. Click on any rating for an explanation)

Supervised classification categorizes an image's pixels into land cover/vegetation classes based on user-provided training data. These training data identify the vegetation or land cover at known locations in an image. The software analyzes the pixel values of the training data and establishes a color profile for each vegetation class. It then classifies, or assigns each pixel in the image to a vegetation class, according to the color profile that best matches that pixel.

The training data are based on manual identification of representative examples from each vegetation class. The identification of representative examples can be derived from field observations at known locations or image interpretation.

Supervised classification has several advantages over simpler methods like unsupervised classification. First, because the classes are user defined, they are ensured to conform to the classification hierarchy of the investigation. Second, the use of training data improves the ability to differentiate between classes with similar color profiles. Finally, the method tends to be more reliable and produce more accurate results.

Supervised 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).

• Image_Interpretation
• Riparian Unsupervised Classification
• Classification and Regression Tree Analysis (CART)
• RSAC Riparian Mapping Tool

Supervised classification can be performed on any digital image. It is frequently applied to satellite or aerial imagery, or to vegetation indexes (e.g., normalized difference vegetation index [NDVI]) derived from such imagery. A high quality training data set is also required.

This method produces a new, simplified image, where each pixel has a vegetation class assignment. The classified image can be used to produce a thematic map showing the distribution of vegetation classes or as an input to more sophisticated processing.

A) B)

Figure 1: A) A false-color (RGB = Bands 7, 3, 2) ASTER image subset around the Soda Butte Creek and Lamar River confluence. B) The image classification result of the ASTER subset (black = unclassified, red = rock/exposed soil, blue = water/shadow, dark green = conifer forest, purple = deciduous, orange = sagebrush, light green = grasslands, maroon = mesic meadow) (From Shive and Crabtree, 2004).

Shive, J.; Crabtree, L. 2004. Mapping Willow Distribution Across the Northern Range of Yellowstone National Park: U.S.A. 44 (3-4), 323-335.

This study mapped willow distribution using ASTER imagery classified using the supervised classification. The classification incorporated the use of RADAR and LiDAR imagery.

Shivem J. 2004, Mapping Amphibian Habitat Distribution in the Frank Church-River of No Return Wilderness, ID Using Multiple Scales of Remotely Sensed Data. Pocatello, ID: Idaho State University. 89 p. Thesis.

Groshong, L. C. 2004. Mapping Riparian Vegetation Change In Yellowstone’s Northern Range Using High Spatial Resolution Imagery. Eugene, OR: University of Oregon. 78 p. Thesis.

Marcus, W. A.; Legleiter, C. J.; Aspinall, R. J.; Boardman, J.W.; Crabtree, R. L. 2003. High spatial resolution hyperspectral mapping of in-stream habitats, depths, and woody debris in mountain streams. U.S.A. Geomorphology 55 (2003), 363-380

• Cingolani, A.M., D. Renison, M.R. Zak, and M.R. Cabido. 2004. Mapping vegetation in a heterogeneous mountain rangeland using landsat data: an alternative method to define and classify land-cover units. Remote Sensing of Environment 92: 84-97.
• Congalton, R.G. 1991. A review of assessing the accuracy of classifications of remotely sensed data. Remote Sensing of Environment 37:35-46.
• Geerken, R., B. Zaitchik, and J.P. Evans. 2005. Classifying rangeland vegetation type and coverage from NDVI time series using Fourier Filtered Cycle Similarity. International Journal of Remote Sensing 26: 5535-5554.
• Ghorbani, A., D. Bruce, and F. Tiver. 2006. Specification: A problem in rangeland monitoring. In: Proceedings of the 1st International Conference on Object-based Image Analysis (OBIA), 4th-5th July 2006, Salzburg, Austria.
• Karl, J. W., and B. A. Maurer. 2009. Multivariate correlations between imagery and field measurements across scales: comparing pixel aggregation and image segmentation. Landscape Ecology. DOI: 10.1007/s10980-009-9439-4

Under ideal conditions supervised classification can produce highly reliable results. However, the method is dependent on:

• The accuracy of the training data.
• How representative the training data are.
• The distinctiveness of the classes.

High quality training data can be time consuming to generate. Please see the Vegetation Mapping Prerequisites and Approaches page for more discussion on training data quality requirements.

Unsupervised classification requires remote sensing or GIS software such as ERDAS Imagine or ArcGIS. This method is processing intensive; processing times will vary by dataset size and computer processing speed.

Advanced Digital Image Processing--Classificaton Techniques (Forest Service Only)

Rangeland Methods Guide: Supervised Classification

_{You must have an account and be logged in to post or reply to the discussion topics below. Click here to login or register for the site.}