本年度為兩年期計畫之第一年，目的為建立有關土地利用之景觀生態與經濟社會因子間之關連，唯研究進行中，遭遇可變動面積單元問題(modifiable aerial unit problem，
MAUP)，亦即在不同觀測尺度下，土地利用、景觀生態與經濟社會等因子間之相關性並不相同，經蒐集資料研究後，發現該問題目前僅於景觀生態論文期刊中受到廣泛討論與關注，然而於其他工程及社會科學領域中，尚未獲得重視，同時有關問題之解決方法亦無定論。為建立相關因子於不同尺度間有意義之關連性，唯有有效解決可變動面積單元問題，本研究於本年度成果中提出整合性架構，結合地質統計、階層分析、景觀尺度及小波分析，嘗試解決可變動面積單元問題，相關論文已於95 年10 月22 日投稿Remote Sensing of Environment，投稿內容詳計畫成果一。此外，探討養殖漁業抽取地下水，除導致地層下陷等土地利用問題，於濁水溪沖積扇及嘉南平原地下水分區，更因該地區地下水中含有高濃度砷，養殖漁業抽取地下水供應養殖用水，將導致漁產品中累積大量有機砷，影響國人食用養殖漁產品之安全。針對該問題，本年度研究成果於94 年12 月1 日投稿，並於95 年8 月9 日刊登於The Science of Total Environment(詳計畫成果二，投稿期刊IF=2.224)，該論文探討水文地質對於地下水中砷累積之影響，計畫主持人並因該論文於95 年8 月20 日獲邀參加國科會與中原大學合辦之「環境污染與人體健康2006 砷、腐植酸與烏腳病國際研討會」，並於會中發表研究成果。 This study establishes a framework for identifying and mapping variations in landscape patterns generated by natural disturbances. The framework is based on a hierarchical perspective of landscape dynamics and takes into account the relationship between landscape patterns and biotic processes in the local ecology. If an organism selects its habitat or food patch via a nested set of hierarchical decisions, the influence of natural disturbances on the nested choices via landscape dynamics should also be assessed hierarchically. In the proposed framework, 2-D Morlet continuous wavelet analysis was adopted to determine the localized variances using anisotropic filtering banks at different scales. The localized variance was used to evaluate the presentation of the patch edge of a low-level entity because high localized variance occurred at the patch edge. From a hierarchical perspective, the significance of variation among low-level entities should be assessed from the viewpoint of a high-level entity. Accordingly, the lacunarity curve was used to assess the gap size between patch edges of low-level entities at different scales. The most significant deviation between lacunarity curves of post-disturbed and pre-disturbed satellite images was recognized as the most significant outcome resulting from the disturbance, while the scale at which the most significant deviation occurred was defined as the characteristic scale at which the disturbance was most relevant to the landscape pattern. A binomial distribution was used to categorize the landscape pattern into binary classes under the assumptions of a random distribution of patch edges and that the highest value in the binary classified map represented the patch edge. Because the binary classified maps delineate the most significant variation in spatial patterns resulting from the disturbance, the maps can be subjected to a pattern–process analysis, e.g., the indirect ordination of bird communities, to determine if the observed variation in patterns also affected related biotic processes. In applying the framework in a case study, the normalized difference vegetation index determined from SPOT HRV images with 12.5 m resolution was used as a measure of changes in land cover within the Chenyulan watershed, Taiwan, resulting from the ChiChi earthquake.
