Abstract: Uncertainty of spatial data in GIS can be in the aspect of position,attribution,temporary,logical relation and completeness.Among them,positional uncertainty of linear segments is an important aspect.There are uncertainty models of linear segments,such as"E-band","g-band"and so on,in some confidence bands,and they are relevant to different confident levels,but choice of confident levels exists a certain extent subjectivity and therefore cannot be completely determined.In this paper,on the basis of union entropy and maximum entropy theorem in information theory,an error entropy band is put forward and the comparison between error entropy band and"E-band" is given,and then the probability of falling into it is calculated.Finally,some conclusions are drawn as follows:1) Error entropy band for linear segment is an average uncertainty measure of linear segments,and can be solely determined by union entropy and is independent of the choice of confident level.Therefore,it is a comparatively impersonal index.2) The ratio of band width between error entropy band and E-band is √e times,and the ratio of area is about times,and the probability of falling into them is 99.87%.3) Information theory can provide a new approach for solving uncertainty problems in GIS.