International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B4. Istanbul 2004 
  
  
The data set we used covers part of the island. The DEMs of 
six consecutive years (from 1989-1995) is displayed in Figure 
2. It is hard to identify the outliers in the images displayed in 
Figure 2. The. purpose of our experiment is to use the 
multiscale approach to detect the outliers in these six year 
DEMs. 
3.2 Implementation details 
We applied the four steps discussed in the previous section. 
First, we classified the DEMs into three landscape classes. 
The classification function was built based upon Dutch 
geomorphologists. For example, the area with height between 
-6 — -1.1 to be the foreshore; the area with height between - 
1.1-2 to be the beach; and the area with height between 2-25 
to be the foredune. The classification results are shown in 
Figure 3. 
Then, we changed the spatial scale of the DEMs by averaging 
the height value by a 3*3 window. We classified them again 
into three landscape classes, according to the class definition 
in the previous step. The aggregated results are shown in 
Figure 4. 
Later, we compared the images in Figure 3 and Figure 4 in the 
same year and found regions that were available in Figure 3 
and disappeared in Figure 4. These regions are potential STOs 
(which are circled in Figure 5). 
For verification, we compared the height values of these 
potential STOs in the consecutive years. If the change of 
height is continuous then the potential STO is not a STO. For 
example, the STO appeared in 1991 (in upper-left corner) 
became part of a big dark area in 1991. It means the change is 
continuous and this is not an outlier in temporal perspective. 
Finally, we identified the STOS, which are circled in 
concreted line in Figure 6. For those circled in dashed lines in 
Figure 6, they are not STOs. 
4. CONCLUSIONS AND FUTURE WORK 
In this paper we discussed spatial-temporal outlier detection. 
We defined a spatial-temporal outlier (STO) to be a spatial- 
temporal referenced object whose thematic attribute values 
are significantly different from those of other spatially and 
temporally referenced objects in its spatial or/and temporal 
neighborhood. We propose a multiscale approach to detect the 
STOs by evaluating the change between consecutive spatial 
and temporal scales. As for further research, the effect of 
granulites of spatial. and temporal scales should be 
investigated. Further, quantitative calibration of the difference 
between two consecutive spatial and temporal scales should 
also be established. 
1010 
ACKNOWLEDGEMENTS 
The first author wishes to thank The Hong Kong Polytechnic 
University for the Postdoctoral Research Fellowship (no. G- 
YWO2). 
REFERENCES 
Adam, *N. R, Janeja, V. P. and ’Atluri, "V. 2004 
Neighbourhood based detection of anomalies in high 
dimension spatio-temporal sensor datasets, 2004 ACM 
symposium on Applied Computing, March 14 — 17, Nicosia, 
Cyprus, P. 576 - 583. 
Barnett, V. and Lewis, T., 1994, Outliers in Statistical Data, 
John Wiley. 
Miller, H., 2003, Geographic data mining and knowledge 
discovery, in J. P. Wilson and A. S. Fotheringham (eds.) 
Handbook of Geographic Information Science, in press. 
Ng, R., 2001, Detecting outliers from large datasets, in H. J. 
Miller and J. Han (eds.) Geographic Data Mining and 
Knowledge Discovery, London: Taylor and Francis, 218-235. 
Johnson, R., 1992, Applied Multilevel Statistical Analysis, 
Prentice Hall. 
Shekhar, S., Lu, C. T. and Zhang, P., 2003, A unified 
approach to detection spatial outliers, Geolnformatica, 7, 139- 
166. 
Yao, X., 2003, Research issues in spatio-temporal data 
mining, a white paper submitted to the University Consortium 
for Geographic Information Science (UCGIS) workshop on 
Geospatial Visualization and Knowledge Discovery, 
Lansdowne, Virginia, Nov. 18-20. 
Intern 
  
  
Fi