應用多源遙測影像分析宜蘭地區植被物候之研究

地表植被物候的變化反映了氣候變遷對於生態系統所造成的影響，而透過遙測技術能夠有效率地進行大範圍且長期的植被物候監測。本研究蒐集了2001至2010年間宜蘭地區之MODIS以及Landsat TM影像，先利用MODIS的MOD13Q1植生指標產品提取出一組空間解析度為250公尺之植生指標（NDVI、EVI）時序資料，並透過影像融合模型ESTARFM（Enhanced Spatial and Temporal Adaptive Reflectance Fusion Model）融合MODIS與Landsat TM影像來產生另一組空間解析度為30公尺之植生指標時間序列資料。接著將這兩組不同空間解析度之植生指標時間序列資料分別匯入物候分析軟體TIMESAT來計算宜蘭地區之地表植被物候參數，再配合氣象資料分析氣候因子與宜蘭地區植被及物候參數間之關係，最後比較兩種不同空間解析度之植生指標時序資料所產生的分析結果。研究結果顯示，MODIS-NDVI時序資料與氣候因子之相關性較MODIS-EVI高，並據此將該資料之物候分析成果作為基準，在2001至2010年間宜蘭地區植被生長季起點延後約6至7天；生長季終點約延後2至6天。在ESTARFM影像融合方面，以Landsat-NDVI之成果較Landsat-EVI與氣候因子有較佳的相關性，且其物候分析成果與MODIS-NDVI成果也較接近。

關鍵字

植被物候； MODIS ； Landsat ； ESTARFM ； TIMESAT

並列摘要

Land surface vegetation phenology reflects the responses of a terrestrial ecosystem to climate change. It is effective for monitoring the vegetation phenology of large-scale area using time-series remotely sensed dataset. In this study, we used MOD13Q1 and Landsat TM images for vegetation phenology analysis between 2001 and 2010 in Yilan. First, the monthly maximum value (MVC) images of MODIS were generated from 16-day VI (Vegetation Indices) time-series data. Second, the synthetic Landsat VI data were created using the ESTARFM (Enhanced Spatial and Temporal Adaptive Reflectance Fusion Model) algorithm for all dates when Landsat images were either unavailable or too cloudy. The monthly temperature and precipitation data during 2001 and 2010 were used to generate map layers using ArcGIS 10.1 for vegetation and climate analysis. Then, the MODIS and Landsat images were provided as input for the phenology analysis toolbox TIMESAT, from which phenological metrics including onset (start of growing season), offset (end of growing season) and length of growing season were derived. These phonological parameters of vegetation and climate data were used to investigate the relationship between vegetation phenology and climate change. In the end, both MODIS VI data and synthetic Landsat VI data were analyzed and compared in the study. The results show that MODIS-NDVI time-series data has higher correlation with climate data than MODIS-EVI. Moreover, the phenology analysis shows that in Yilan area the onset of vegetation was 6 days late and the offset was 2-6 days late during the 10 years period. For the ESTARFM model, Landsat-NDVI has higher coorelation with climate data than Landsat-EVI, and the phenology analysis result is more consistent to that of MODIS-NDVI.