利用衛星影像評估莫蘭蒂颱風對金門本島地景變遷 及森林恢復之影響

The Impact of Typhoon Meranti on Landscape Change and Forest Recovery in Kinmen with Satellite Images

耿翊(Ken I)

指導教授 ： 陳建璋 ； 魏浚紘

國立屏東科技大學/農學院/森林系所/碩士(2019年)

https://doi.org/10.6346/NPUST201900091

摘要

臺灣位於颱風常經路徑上，於夏季時，颱風挾帶之強風及大量降水引起臺灣各類型之災害，金門地區因有臺灣地形屏障雖受颱風之影響較小，但於2016年莫蘭蒂颱風造成金門地區嚴重災情，產生劇烈地景變化，特別是森林生態系深受颱風影響嚴重。隨著遙測(Remote Sensing, RS)技術精進，配合地景生態學(Landscape Ecology)概念的發展，使地景生態監測更加容易執行，且可針對長時間尺度或大範圍之地景進行研究。本研究利用Sentinel 2衛星影像進行影像分類，監測金門地區於莫蘭蒂颱風前後森林及地景變化，並分析森林面積、常態化差異植生指標(Normalized Difference Vegetation Index, NDVI)、地上生物量(Above Ground Biomass, AGB)、土地利用類型及地景指標之變遷與變化，以了解莫蘭蒂颱風對金門地區之衝擊。研究結果顯示，影像分類總體準確度為89%，kappa值為0.87，森林範圍因受颱風影響而受損後減少了約16%，於隔年再減少15%，於颱風後2年部分受損森林逐漸恢復；NDVI於颱風過後年明顯下降，於颱風後1年因林木之重新生長而迅速恢復；AGB估量中以NDVI作為依據，並建立金門地區適用之迴歸式，發現AGB於颱風當年下降約82%，於颱風後兩年即恢復至颱風前之狀態；地景指標中藉由Shannon 地景多樣性(Shannon`s Diversity Index, SHDI)之t-檢定顯示颱風確實造成金門地景多樣性提升(p < 0.01)，因金門本島之森林面積最大，SHDI之上升代表森林面積相對縮減，導致其與其他土地利用面積比例更為接近，因此就森林恢復情形而言，SHDI下降可能代表森林有所恢復。研究結果可供為森林災害監測研究之參考。

關鍵字

地上生物量 ； 森林恢復 ； 地景指標 ； 常態化差異植生指標

並列摘要

Taiwan is located on a frequent pathway of typhoon and is more likely to be affected by strong wind and heavy rainfalls in the summer. Though, typhoon is less likely to have an impact on Kinmen due to the topographic effects of Taiwan; however, in 2016, a catastrophic typhoon Meranti struck Kinmen and has made an major impact to the landscape of Kinmen, and which forests of Kinmen is believed to be affected. With the developing and advancing techniques of remote sensing and its integration with landscape ecology, the execution of landscape monitoring has been enhanced, including monitoring in larger spatial scale or longer time scale available. Sentinel 2 satellite imagery was applied and classified for monitoring Kinmen landscape and land use changes before and after typhoon Mernati. Forest area, NDVI, AGB, landscape metrics and land use transformation matrix were applied in this study. The overall accuracy of classification was 89.89% (kappa = 0.87). Results showed the forest was damaged by typhoon Meranti and leads to a decrease in area by 16% after typhoon Meranti, and 15% more afterwards, but started to recover in 2018. A significant fall in NDVI was discovered during the year of typhoon Meranti occurred, and due to forest regrowth, NDVI returned to pre-typhoon status. AGB estimation was calculated by a regression model developed within this study, and results showed an 82% decrease in 2016, but recovered to pre-typhoon status shortly in 2018. Through SHDI t-test, the landscape diversity of Kinmen raised significantly (p < 0.01). While the forest accounts for the largest land use in Kinmen, a growing SHDI trend indicates that the proportion of forest has become closer to other land use types. In terms of forest recovery, SHDI decreases might indicate that forest is recovering. The result of this study can be used as reference in monitoring studies of forest disasters.

並列關鍵字

Above Ground Biomass ； Forest Recovery ； Landscape Metrics ； Normalized Difference Vegetation Index

參考文獻

中央氣象局 (1974) 民國六十二年颱風調查報告(第一號侵臺颱風魏達)，第1-5頁。民國六十二年颱風調查報告。中央氣象局。39頁。

內政部 (2017) 莫蘭蒂颱風影響臺灣與颱風警報發佈解除時間，第2頁。莫蘭蒂颱風中央災害應變中心總結報告。內政部。175頁。

王素芬、陳永寬、鄭祈全 (1999) 地理資訊系統和碎形維度於森林地景空間變化上之應用。航測及遙測學刊 4(2): 33-53.

吳坤真、何莞薇、陳建璋、陳朝圳 (2013) 墾丁國家公園地景變遷與銀合歡入侵之關係。國家公園學報 23(4): 32-41.

林士強 (2013) 金門島嶼型災害特性及規模設定方法之探討。地理學報 69: 1-24。