The earth's surface is composed of consolidated, unconsolidated, and semi-consolidated materials existing in an infinite variety of mixtures, habits, depths, areas, compositions, heights, expressions, etc. A basis for organization of this apparent diversity can be found in the physical science of geology (the study of the earth's origin, history, and physical structure), its associated fields of physiography (the study of gross aspects of the structure, spatial arrangement, and phenomena associated with the earth's surface), and geomorphology (the study of configuration and evolution of landform surface features of the earth).

A physiographic unit is an area of the earth's surface within which the major topographic features have a single geomorphic history, a definite general structure, certain physical characteristics, and a predictable general pattern of lower order landforms. 

A landform is a terrain unit created by natural processes in such a way that it may be recognized and described in terms of typical attributes where ever it may occur. When identified, a landform provides information concerning its structure and composition. Each physiographic unit has commonly occurring landforms due to its combination of origin, history, and physical structure and these will differ from those in adjacent physiographic units. These differences are discernible and describable. 

Geomorphology is often defined as that part of geologic science concerned with the manner in which landforms are created and destroyed. Physiography is also defined by some to mean ‘regional geomorphology.’ Knowledge of landforms is of great importance to an image interpreter interested in terrain analysis. Once a landform has been identified, many of the dominant physical characteristics of the materials comprising it may be inferred.

Knowledge Based Terrain Analysis

The approach allows prior definition of expected landforms in a physiographic unit in terms of their surface configuration attributes and their associations within the landform set. The approach adds the physiographic context to the analysis process and enables a significant reduction over a super list of all possible landforms. Further, it serves to organize the analysis tasks and significantly reduces the pattern recognition search space.

Knowing the physiography and geomorphology, we can establish an a priori landform model. Model based reasoning then selectively employs various analysis strategies for accomplishing automated terrain delineation, identification, and attribution using the digital elevation data. The classical approach to automated image analysis involves first preprocessing, followed by the segmentation of characteristics, and then the interpretation of these segments as objects. In general this analysis is done with little or no expectation of what may be in the scene. As described above, there is now a significant amount of a priori physiographic information that can be used to apply image understanding algorithms to the identification of landforms in a physiographic area more efficiently and accurately. This a priori information includes the types of landforms that are expected, the relationships that exist between the landforms and their typical characteristics. The knowledge is implemented in a knowledge base consisting of IF-THEN production rules and algorithms. The rules define relationships between landforms, how algorithms should be applied to segment terrain characteristics, and how the results should be interpreted as landforms. This, in essence, is giving the same knowledge to the automated system as that of an analyst trained in geomorphic landform analysis.

The a priori physiographic knowledge base can be constructed for all of the physiographic provinces/sections in the United States. Elements of this knowledge base developed for one physiographic section can be mixed and matched for use in other physiographic sections. A physiographic description can be developed for any geographic area. These modular knowledge bases can then be "plugged" into the system based on the area of interest. This knowledge engineering effort then pays off in the ability of the automated system to perform high resolution, detailed analysis over relatively large geographic areas.