A Robust Exploratory Method for Qualitative Trend Curve Analysis against Poor Quality Data

This paper proposes a theoretical method for qualitative trend curve analysis, supposing the situation that data quality is poor and the amount of data is very large. Here the ‘trend curve’ means, for example, a sequence of monthly temperature over twelve months at a location. The paper first defines ‘qualitative’ characteristics of a trend curve in terms of isomorphic relations. Second, these concepts are extended to macroscopic qualitative similarity. Third, using the macroscopic qualitative similarity, the paper proposes a method for categorical clustering and a method for measuring the magnitude of qualitative change in trend curves. Fifth, the proposed theoretical method is implemented in a GIS environment and an exploratory tool (a computer program called QuaT) is developed. Sixth, this tool is applied to the analysis of land cover change in the Persian Gulf Area between 1982 and 1993. Last, the paper discusses the limitations of the proposed method.