s With the analysis of hierarchies of temporal-spatial data models

it is argued that current temporal-spatial data models lack explicit definition for the geographical phenomena and description for the elementary relations and are unsuitable for data sharing on semantic levels. Traditional theme map layer based spatial data abstraction frameworks and the representation methods taking the elementary units of vector or raster data structures as the elementary units for modeling are deficient in describing complex geographical entities and phenomena and analyzing the geographical procedures. It is set forward that developing feature based temporal-spatial data models is the core for the new generation geographical information systems. The conceptual modeling elements and mutual relations between them are discussed and some key technical problems are identified in the development of feature based temporal- spatial data models.