This paper attempts to use spatial wavelet transformation derived multi-scale module maximum as a geographical locator to objectively identify eco-geographical transition zone and its boundary. Employing the spatial patterns of NDVI and precipitation, we decomposed 49 transects with wavelet transformation of db3 kernel function. The geographical coordinates of all singularity points were located and marked in GIS. The results show that optimal spatial scale for identifying eco-geographical boundary of semiarid-subhumid transition zone is about 20-40 km. When the scale is less than this threshold value, the locating process is likely influenced by local scale ground structure such as built-up areas or fine scale landforms. On contrary, when the scale is larger than 20-40 km, the accuracy of locating points will drop sharply owing to over- smoothing treatment. In spite of the higher concentration degree of NDVI derived locating points than that of precipitation derived, the comprehensive locating accuracy for precipitation transition zone is superior to that of NDVI spatial pattern in terms of long axis direction coherence, mean minimum distance and point balance degree indicates. Our results suggest that the combination of spatial wavelet transformation and GIS is a powerful complement approach to expert based eco-geographical transition zone identification and locating.