| 摘要: | 因全球暖化現象日益嚴重，引起全球各國正視二氣化碳的遽增問題，並於1997年在日本京都簽約國會議中達成「京都議定書」簽署共識，以共同推動全球節能減碳任務。協約中明訂造林及森林採伐之二氧化碳吸收或排放之淨值，可併入排放減量值計算，並列為該國之抵減量，所以森林擔任二氧化碳儲存和吸收的重要角色，對減緩溫室效應有相當大的貢獻。據研究顯示，森林調查資料與遙測技術的結合，成為碳儲量、碳庫及碳吸存量的主要方法，尤其依據聯合國跨政府組織間氣候變遷專門委員會良好做法指南（Intergovernmental Panel on Climate Change Good Practice Guidance, IPCC GPG）（2003）指出，遙測技術特別適用於國家土地使用、土地使用變遷的實證及林業碳庫的評估（the national Land Use, Land-Use Change, and Forestry Carbon pool, LULUCF），特別是地上部生物量的估算。因此，IPCC提出建議可整合遙測資源及生態模式（ecological models），用以提供「聯合國氣候變遷綱要」之評估需求。
職是之故，本研究即採用地面調查資料與遙測影像結合之方式，來估測森林的碳儲量之變化情況。實驗中，選用福衛2號及IKONOS-2衛星影像為研究素材，並採用統計「複迴歸（MRA）」與「倒傳遞類神經網路（BPN）」、CBMRA及CBBPN等模式作為分析之技術，主要之探討重點為，不同解析度影像的碳儲存預測能力、MRA 、BPN 、CBMRA及CBBPN的預測精度差異。地面調查之林木資料係由林務局提供，共複查了55個永久樣區，推估之樣區材積總量為868.37 mP3P。另外，從福衛2號及IKONOS-2衛星影像中，則萃取了NDVI、TNDVI、RVI、DVI等植被指標，並與地面調查資料作MRA、BPN、CBMRA及CBBPN分析比較。依實驗RMSE的比較結果顯示，高解析度影像之判釋預測力優於低解析度影像，GBISC法對於資料的預測精度有提升效果，顯示遙測技術對於地方性尺度的森林監測，可以達到有效的監測效益，建議應予充分利用及推廣。
The increasing amount of carbon dioxide has exacerbated global warming and accordingly led to the whole world addressing the dramatic change of weather. Many countries reached a consensus, signing the Kyoto Protocol in 1997, in Kyoto, Japan, with an effort to reduce the mission of carbon. The treaty allows that the carbon dioxide absorption and emission together with afforestation and deforestation, can be incorporated into net value of reduction, and this refers to the amount of reduction for the country. Therefore, forests play an important role in terms of absorbing and storing carbon dioxide; this has enormously contributed to attenuating the greenhouse effect. Studies have shown that the combination of forest inventories and remote sensing has become the main method of assessment for carbon sink, carbon stocks, and carbon sequestration. Particularly, according to the International panel on Climate Change Good Practice Guidance（IPCC GPG）（2003）, remote sensing methods are especially suitable for verifying the national Land Use, Land-Use Change, and Forestry （LULUCF）C pool estimates, specifically the aboveground biomass. Thus, IPCC has proposed integrating remote sensing and ecological models for the requirement of assessment of The United Nations Framework Convention on Climate Change（UNFCCC）.
Therefore, this study aims at estimating the situation of changes in forests carbon stocks with the application of remote sensing, forest investigation data, and the combination of both methods. This experiment adopts FORMOSAT-2 and IKONOS-2 satellite imagery as materials, meanwhile, multiple regression analysis（MRA）, back-propagation network（BPN）, CBMRA and CBBPN analysis technology are also applied. The main ideas are the predicting capability of carbon stocks in different resolution images, and the precision degree of prediction in CBMRA, CBBPN, MRA and BPN. The forest ground survey data is provided by the Forestry Bureau of Taiwan, and this research has reviewed a total of 55 permanent sample plots and estimate total 868.37 mP3P volume values. Furthermore, the vegetation indices of the NDVI, TNDVI, RVI, and DVI, have been extracted from FORMOSAT-2 and IKONOS-2 satellite images, and incorporated with the ancillary data for the analysis of MRA, BPN, CBMRA and CBBPN. The result of the comparison of RMSE indicates that the prediction capability of interpretation of high resolution imagery is superior to the coarse resolution imagery. The GBISC method has promoted the effectiveness of data predicting precision. Thus, overall, it revealed that remote sensing technology could effectively contribute to forest monitoring in terms of local scale. Hence, I propose that remote sensing should be widely applied and promoted. |
