The United States spends over $17 billion per year on maintenance and rehabilitation of pavements [FHWA 87]. The substantial resource costs involved have motivated the development of automated methods for acquisition of pavement surface condition information and the use of this information for pavement management While new data acquisition technologies have proliferated, the manipulation of surface observations and the representation of surfaces has typically been ad hoc and narrowly focused. This thesis addresses these problems. Its main contribution is to demonstrate the critical importance of spatial relationships among surface characteristics for describing pavement conditions. It also contributes a formal spatial model of pavement surfaces as well as the tools to use it The model incorporates multiple types of surface data, maintains spatial relationships among the data types, provides different levels of surface data abstraction, represents data at different levels of spatial aggregation, and facilitates flexible access of surface information. The model has been implemented as a software library. Its application to perception for automated pavement crack filling, which is reported in this thesis, serves to illustrate its potential for combining sensor information. The model’s potential is further illustrated by its application to a class of pavement management activities involving automated pavement surface characterization. Its use as a standard representation is demonstrated, and its capabilities for advanced characterization are i’lustrated. For instance, the thesis reports on the incorporation of three different commercial systems’ condition assessment data into the standard model and the subsequent derivation of various summary measures. This thesis demonstrates that the pavement surfaces model could be extremely beneficial to pavement management and to the development of automated systems for pavement characterization and surface work.