屬性 j
u u+nσ u-nσ
u u-nσ u+nσ
圖7 資料濾除示意圖(二維為例，實際為多維空間)
5.3 資料濾除後成果
資料濾除後，再度探究崩塌因子的顯著程度。
表5 顯示訓練資料與檢核資料依然相似，且道路距 離也較近似。而預測資料的順序也有所改善，較接 近訓練資料，預測成果應能提高。再者，正規化差 異植生指標、土地利用、斷層距離及水系距離都一 致地位於訓練、檢核與預測資料的前五位，表示這 四個因子與崩塌地呈現高顯著性。
(3)
表 6 (b)為資料濾除前後的檢核與預測成果，
發現濾除後的OA 值都微幅降低，理由是非崩塌地 像元數因濾除機制減少(請見表 4)，增加分類難 度。另外，濾除後非崩塌地的PA 與 UA 值稍微降 低，但崩塌地漏授與誤授明顯減少，尤其是預測結 果，致使kappa 值大為升高，其中 J48 提升約 30%，
貝氏網路改善約20%。值得一提的是，貝氏網路的
預測成果的kappa 值可達 0.8939。針對 2008 年崩 塌較集中的範圍，圖8 與圖 9 展示貝氏網路的預測 及潛勢成果，前者呈現資料濾除後能獲得較佳的偵 測能力，後者顯示濾除機制使得演算法產生的機率 值更符合實況。因此，案例證實濾除機制能保留代 表性資料，提升模型預測和潛勢成果的可靠性。
表4 資料濾除前後像元數
門檻值 崩塌資料 非崩塌資料
訓練及檢核資料 預測資料 訓練及檢核資料 預測資料
原始 37790 1839 329786 19184
3σ 260 0 0 0
4σ 260 117 0 158
5σ 34624 1336 21993 1360
表5 資料濾除後崩塌因子顯著性分析
順序 訓練資料 檢核資料 預測資料
1 正規化差異植生指標 土地利用 正規化差異植生指標
2 土地利用 坡度 斷層距離
3 坡度 正規化差異植生指標 土地利用
4 斷層距離 斷層距離 水系距離
5 水系距離 水系距離 道路距離
6 地質 地質 高程
7 道路距離 土壤 坡度
8 土壤 道路距離 地質
9 高程 高程 土壤
10 坡向 坡向 曲率
11 曲率 曲率 坡向
表6 成果比較((a)資料濾除前，(b)資料濾除後，(c)是對(b)特徵縮減)
非崩塌 崩塌
OA (%) Kappa PA (%) UA (%) PA (%) UA (%)
檢核
J48
(a) 99.1 98.9 88.5 90.2 98.1 0.8829 (b) 95.5 92.8 95.5 97.3 95.5 0.9053 (c) 88.6 88 92.3 92.7 90.9 0.8085 BN
(a) 98.9 98.4 84.4 88.1 97.5 0.8487 (b) 94.4 94.8 96.7 96.4 95.8 0.9115 (c) 91.6 89.2 92.9 94.6 92.4 0.841
預測
J48
(a) 93.1 98.8 65.3 23.9 92.2 0.3182 (b) 75.1 92.4 89.8 68.9 80.7 0.6134 (c) 89.1 88.3 87.9 88.8 88.5 0.7707 BN (a) 96.4 99 86 61.2 95.7 0.6928 (b) 95.1 94.3 94.3 95.1 94.7 0.8939 (c) 90.1 92.5 92.5 91.3 91.3 0.8257
0 3 6 12
Kilometers 0 150 300 600
Meters 0 150 300 600
Meters
位置圖(右側為放大圖) 資料濾除前 資料濾除後
圖8 貝氏網路部分預測成果(黃色：未偵測區域，紅色：偵測區域)
0 3 6 12
Kilometers 0 150 300 600Meters 0 150 300 600
Meters
位置圖(右側為放大圖) 資料濾除前 資料濾除後
圖 9 貝氏網路部分潛勢成果(紅色：較高機率，崩塌機率大於 0.75；黃色：高機率，崩塌機率介於 0.75 與0.51；藍色：低機率，崩塌機率介於 0.5 與 0.26；綠色：較低機率，崩塌機率小於 0.26)
5.4 特徵縮減成果
基於表5，得知正規化差異植生指標、土地利 用、斷層距離及水系距離扮演重要角色。本節據此 四種崩塌因子重新計算，成果如表6 (c)所示。相較 於表6 (b)， J48 決策樹預測成果的 kappa 值大幅 提升，因非崩塌地漏授及崩塌地誤授降低之故。雖 然犧牲檢核和貝氏網路的預測成果(礙於分類策略 不同)，但仍持接受範圍。若為減少資料屬性，以 及節省計算資源，不失為可行辦法。
6. 結論與建議
本研究針對災後以資料導向分析的觀點，探究 資料探勘技術對豪雨促發崩塌事件之驗證與潛勢 評估，為龐大資料量、空間資料來源及格式不一、
崩塌因子與崩塌事件關係未必明確、崩塌因子的獨 立與否、之前未知的崩塌特性等五項議題提供可能 的解決途徑。除此之外，不同時期資料整合、向量 與 網 格 資 料 轉 換 、 套 合 不 同 細 緻 度 或 粒 度 (granularity)圖層、人為數化誤差、複雜的自然現象 等等情形，皆可能造成空間資料的異質性和不確定
性。若直接將崩塌因子代入演算法，勢必面臨模型 第二十二屆 ISPRS 研討會(The XXII Congress of the International Society for Photogrammetry and Remote Sensing)，並發表本研究成果，特此致謝。
參考文獻
Aleotti, P., and R. Chowdhury, 1999. Landslide hazard assessment: summary review and new perspectives, Bulletin of Engineering Geology and the Environment, Vol. 58, pp. 21-44.
Bai, S., G. Lu, J. Wang, P. Zhou, and L. Ding, 2011.
GIS-based rare events logistic regression for landslide-susceptibility mapping of Lianyungang, China, Environmental Earth Science, Vol. 62, pp. 139-149.
Clerici, A., S. Perego, C. Tellini, and P. Vescovi, 2006.
A GIS-based automated procedure for landslide susceptibility mapping by the conditional analysis method: the Braganza valley case study (Italian Northern Apennines), Environmental Geology, Vol. 50, pp. 941-961.
Cruden, D. M., and D. J. Varnes, 1996. Landslide types and processes In: K.A. Turner and R.L.
Schuster, Editors, Landslides: investigation and mitigation, Transport Research Board Special Report, Vol. 247, pp.36-75.
Dahal, R. K., S. Hasegawa, A. Nonomura, M.
Yamanaka, S. Dhakal, and P. Paudyal, 2008.
Predictive modeling of rainfall-induced landslide hazard in the lesser Himalaya of Nepal based on weights-of-evidence, Geomorphology,
Vol. 102, No. 3-4, pp.496-510.
Dai, F. C., C. F. Lee, and Y. Y. Ngai, 2002. Landslide risk assessment and management: an overview, Engineering Geology, Vol. 64, pp. 65-87.
Dietrich, E. W., R. Reiss, M. L. Hsu, and D. R.
Montgomery, 1995. A process-based model for colluvial soil depth and shallow landsliding using digital elevation data, Hydrological Processes, Vol. 9, pp. 383-400.
Dilley, M., R. S. Chen, U. Deichmann, A. L.
Lerner-Lam, and A. Margaret, 2005. Natural disaster hotspots: a global risk analysis, The World Bank, pp. 145.
Du, Y., P. M. Teillet, and J. Cihlar, 2002. Radiometric normalization of multitemporal high-resolution satellite images with quality control for land cover change detection, Remote Sensing of Environment, Vol. 82, pp. 123-134.
Dunne, T., 1991. Stochastic aspects of the relations between climate, hydrology and landform evolution, Transaction Japanese Geomorphological Union, Vol. 12, pp. 1-24.
Fayyad, U. M., and K. B. Irani, 1993. Multi-interval discretization of continuous-valued attributes for classification learning, International Joint Conference on AI, pp. 1022-1027, Morgan, Kaufmann.
Ercanoglu, M., and C. Gokceoglu, 2004. Use of fuzzy relations to produce landslide susceptibility map of a landslide prone area (West Black Sea Region, Turkey), Engineering Geology, Vol. 75, pp. 229-250.
Gemitzi, A., G. Falalakis, P. Eskioglou, and C. Petalas.
2011. Evaluating landslide susceptibility using environmental factors, fuzzy membership functions and GIS, Global NEST (Network for Environmental Science and Technology) Journal, Vol. 13, No. 1, pp. 28-40.
Guzzetti, F., A. Carrara, M, Cardinali, and P.
Reichenbach, 1999. Landslide hazard assessment in the Saffora basin, northern Italian Apennines, Geomorphology, Vol. 72, pp.
272-299.
Mettemicht, G., L. Hurni, and R. Gogu, 2005. Remote sensing of landslides: An analysis of the potential contribution to geo-spatial systems for hazard assessment in mountainous environments, Remote Sensing of Environment, Vol. 98, pp.
284-303.
Miller, J., and J. Han, 2001. Geographic data mining and knowledge discovery, in: Geographic data mining and knowledge discovery (ed. by H.J.
Miller and J. Han), Taylor and Francis, London.
Montgomery, D. R., and W. E. Dietrich, 1994, Landscape dissection and drainage area-slope
thresholds. Process Models and Theoretical Geomorphology, British Geomorphological Research Group Symposia Series, John Wiley &
Sons Ltd.,Chichester, pp. 221-246.
Nichol, J., and M. S. Wong, 2005. Satellite remote sensing for detailed landslide inventories using change detection and image fusion, International Journal of Remote Sensing, Vol. 26, No. 9, pp.
1913-1926.
Okimura T., and T. Kawatani, 1987. Mapping of the potential surface- failure sites on granite slopes, International Geomorphology 1986 Part I., Wiley, Chichester, pp. 121-138.
Openshaw, S., and C. Openshaw, 1997. Artificial intelligence in geography, John Wiley and Sons., U.S.A.
Peduzzi, P., 2010. Landslide and vegetation cover in the 2005 Northern Pakistan earthquake: a GIS and statistical quantitative approach, Natural Hazards and Earth System Sciences, Vol. 10, pp.
623-640.
Sakar, S., and D. P. Kanungo, 2004. An integrated approach for landslide susceptibility mapping using remote sensing and GIS, Photogrammetric Engineering & Remote Sensing, Vol. 70, pp.
614-625.
Schott, J. R., C. Salvaggio, and W. J. Volchok, 1988.
Radiometric scene normalization using pseudoinvariant features, Remote Sensing of Environment, Vol. 26, pp. 1-16.
Sidle, R. C., A. J. Pearce, and C. L. O'Loughlin, 1985.
Hillslope stability and land use, Water Resources Monograph, Vol. 11, pp.140-141.
Tan, P.-N., M. Steinbach, and V. Kumar, 2006.
Introduction to data mining, Addison Wesley, U.S.A.
Tsai, F., and L. C. Chen, 2007. Long-term landcover monitoring and disaster assessment in the Shiman Reservoir Watershed using satellite images, in: Proc. 13th CeRES International Symposium on Remote Sensing, Chiba, Japan.
Wang, X., and R. Niu, 2010. Landslide intelligent prediction using object-oriented method, Soil Dynamics and Earthquake Engineering, Vol. 30, pp. 1478-1486.
Wu, W., and R. C. Sidle, 1995. A distributed slope stability model for steep forested basins. Water Resource Research, Vol. 31, pp. 2097-2110.
Zhou, C. H., C. F. Lee, J. Li, and Z. W. Xu, 2002. On the spatial relationship between landslides and causative factors on Lantau Island, Hong Kong.
Geomorphology, Vol. 43, No. 3-4, pp. 197-207.
1 PhD Candidate, Department of Civil Engineering, National Central University Received Date: Feb. 25, 2013 2 Associate Professor, Center for Space and Remote Sensing, National Central University Revised Date: Aug. 07, 2013 3 Master, Department of Civil Engineering, National Central University Accepted Date: Aug.16, 2013 4 Associate Professor, Department of Civil Engineering, National Taipei University of Technology
*.Corresponding Author, Phone: 886-3-4227151 ext. 57619, E-mail：[email protected]