應用生態氣候指標預測臺灣潛在自然植群之研究

Abstract

本研究之目的在瞭解臺灣氣候與植群類型的空間分布，藉由地理資訊系統、數值高程模型、地理統計等工具之結合，以空間化製圖來呈現此種複雜的變化，基於氣候–植群分類模型將研究過程分為三部分：(1)氣候亞模型—推估氣候網格空間，並設計生態氣候指標；(2)植群亞模型—確立植群類型，並擬定潛在植群形相分類方案；(3)建構氣候–植群分類模型，並預測潛在自然植群。 為建立氣候亞模型，本研究以檢定全島氣象站之詮釋資料與觀測數據為啟端，聯合時空相關之檢定程序，總計濾除18.7%座標與/或海拔錯誤之測站、0.83%不合理的日均溫及日降水量數據，據此再進行氣候空間推估；經比較協同高程、結合非等向性、結合降趨、迴歸及對數常態化等推估方法，結果顯示迴歸修正克利金法係推估月均溫空間之最佳方法，測站之海拔、座標的複線性迴歸約可解釋95%月均溫的變異，套疊迴歸殘差空間推估網格可再修正局部地區氣溫之變化；對數化克利金法係推估月降水空間之最佳方法，可說明臺灣整體概觀之各月降水分布，明顯可見西南部之冬乾現象。以各月均溫及降水空間為材料，可對現有氣候–植群分類模型進行評估與實繪，結果表明Holdridge模型將臺灣劃分為13種生命帶，而Kira模型於劃分山地植群帶時較為合理，但仍須改善其生態氣候指標，為此，本研究以年溫差修正溫量指標創設溫量修正指標(MWI)，並結合冬半年之溫量指標與降水量創設冬乾指標(WDI)，藉以分別代表環境之熱量境制及水分境制兩種生態氣候指標，MWI及WDI之空間分布圖層亦即本研究之氣候亞模型。 為建立植群亞模型，首先依「整合大量樣區資料，由下而上的低階植相聚類」之途徑，設計建檔程序及轉檔程式，成功整併不同來源之585個樣區資料，但因樣區數目不足且分布不均，分析結果未能顯示出臺灣完整的植群類型；故轉循「預先設定植群類型，由上而下的高階形相分類」之途徑，經參考他國植群分類系統及臺灣、東亞植群分類研究報告，訂定分類原則及架構，提出4階層之潛在植群形相分類方案：(1)寒原、森林2類群系綱；(2)亞寒帶、冷溫帶、涼溫帶、暖溫帶、亞熱帶5類群系亞綱；(3) 8類群系組表徵優勢植群之葉片的物候與形態；(4) 16類群系說明其水分境制及優勢分類群；此方案統一處理了過去各種有關植群劃分之紛岐名制，方案之各階植群類型亦即本研究之植群亞模型。 為達繪製潛在植群圖之最終目標，本研究以MWI及WDI之網格圖層代表氣候亞模型，以潛在植群形相分類方案之各階類型代表植群亞模型，藉由布林分離法聯繫此2亞模型之關係，建構氣候–植群分類模型，並以地理資訊系統繪製出包含群系綱、群系亞綱、群系組、群系之潛在植群圖，首次完整呈現臺灣各階植群類型之空間分布，合理說明了山地植群帶分化、北部植群下降及西南部冬季乾旱等現象，亦整合過去各種有關垂直性山地植群帶狀分化與水平性地理氣候區劃分的複雜意見；另外，本研究將585個樣區林木座標標定於MWI及WDI，繪製36種林木於水熱境制及熱量氣候帶之分布範圍。 本研究為瞭解臺灣氣候及植群類型之空間分布提供一種新的研究途徑與預測架構，同時檢討研究限制並提出改善建議，研究過程所得之氣象資料檢定程序、氣候網格圖層、樣區資料整合程序、植群形相分類方案、生態氣候指標、樹種水熱境制、PNV圖層等成果也有助於自然資源之經營管理與科學研究。

Aiming at recognizing the spatial distributions of the climate and vegetation patterns in Taiwan, the climate-vegetation classification model had been implied to illustrate the complex variability through the spatialized cartography using the geographical information system (GIS), digital elevation model, Geostatistics and vegetation classification system as tools. The processes would separate into 3 parts: (1) climate submodel — interpolating the climatic grid space and designing the ecoclimatic indices; (2) vegetation submodel — confirming the vegetation types and establishing the physiognomic classification scheme of potential vegetation; (3) associating climate with vegetation and mapping the potential natural vegetation (PNV). For organizing climate submodel, started with checking the meteorological data by designing a series of examining rules. The procedure filtered out 18.7% stations with irrational coordinate and/or elevation of the metadata, and 0.83% daily mean temperature and daily precipitation data of inaccurate extreme. Then the data could be used to calculate the monthly mean temperature (Tm) and monthly precipitation (Pm) as the samples of interpolating the climatic space. The choice of the method was based on the cross validation. Modified residual Kriging was the best method for interpolating Tm space that the multiple linear regression could predict 95% variability of Tm and then overlapped the local variability from interpolating residual. Logarithmic ordinary Kriging was the best method for interpolating Pm space that performed the whole comprehensive spatial distribution of Pm and discerned the winter drought in the southwest lowland. Then Tm and Pm spaces were used as materials to assess and map the present climate-vegetation classification model. It could be demarcated 13 life zones by Holdridge model in Taiwan, but the result was not suited to local vegetation patterns. Comparatively, Kira's warmth index (WI) was more reasonable to delimit the mountain vegetation zones, but should try to modify the ecoclimatic indices. Therefore, WI had been modified by annual range of monthly mean temperature to originate the modified warmth index (MWI) as thermal regime, and combined WI with precipitation both in the winter half-year to originate the winter drought index (WDI) as moisture regime. For organizing vegetation submodel, started with the bottom-up gateway to agglomerate plot-data. The vegetation types must be confirmed before predictive vegetation mapping, therefore designed a transforming program to combine 585 plot-data from different surveys for the purpose. But the result of the classification and ordination of the combined plot-data couldn't display comprehensive vegetation types of Taiwan, because the plots were insufficient and disproportionate. Consequently, this study adopted the top-down gateway to divisive vegetation. After referring to Taiwan and east-Asia vegetation research papers and American national vegetation classification system etc., the physiognomic classification scheme of Taiwan had been drafted as vegetation submodel. The scheme included 4 levels: 2 classes (tundra, forest), 5subclasses (subarctic, coldtemperate, cooltemperate, warmtemperate, subtropical), 8 groups (based on phenology and types of predominant leaf) and 16 formations (based on moisture regime and predominant taxa). And, the scheme dealt completely with the diverse nomenclatures of vegetation types in the past. Finally, PNV had been mapped according to Boolean discrete methods to connect climate submodel and vegetation submodel by GIS. PNV map first represented the spatial distribution of all vegetation types of 4 physiognomic levels in Taiwan, also integrated the diverse comments of altitudinal vegetation zones and geographical climatic regions in the past. Besides, the scattered ranges of 36 dominant trees in the moisture-thermal regime and climatic zone had been drawn via coordinating 585 plot-data with MWI and WDI. This study provided a new research way and predictive framework for interpreting the complex climate and vegetation of Taiwan, and raised some remedial suggestions. The climatic grid layers, plot-data transforming program, physiognomic classification scheme of vegetation, ecoclimatic indices, moisture-thermal distribution of plants, and PNV map resulted from this study could facilitate further the managements and researches of the natural resources.