應用多時期航拍影像探討1980-2021年間臺灣兩高山地區森林範圍的擴展

Abstract

氣候變遷可能導致物種遷移，進而改變生態系的生物多樣性、結構和功能。高山生態系為對氣候變遷尤為敏感，其中高山森林海拔分布上限的變遷是氣候變遷生態學研究的焦點之一。因為低溫限制減少而使植群往更高海拔移動，但高海拔範圍受限使高山植群分布範圍逐漸縮減，可能使得部分物種滅絕、生態系統服務改變等風險。然而，臺灣針對高山森林海拔變化的研究相對有限，且主要集中於合歡山地區。考量臺灣高山環境的多樣性與區域性條件的差異，需有涵蓋更多地點的研究，才能更全面理解氣候變遷對臺灣高山森林分布的影響。本研究利用1980、2001及2021年的航空照片，結合U-Net深度學習模型進行影像分割，分析玉山與中央山脈地區過去41年間高山森林分布的變化趨勢，並使用空間迴歸分析探討氣候與地形因子對森林擴展的影響。研究發現，使用較小影像尺寸進行模型訓練的影像分割成果最佳。1980至2021年間，玉山與中央山脈的高山森林覆蓋面積淨增約197公頃與158公頃，且樹木界線平均每年向更高海拔推進約1.44至1.55公尺。森林擴展模式呈現出顯著的空間異質性，和過往研究發現的合歡山區的森林擴展以樹木界線推移帶緻密化為主不同，本研究的玉山與中央山脈區的森林擴展以樹木界線推進為主。迴歸分析結果顯示，森林面積擴展與海拔推移的調控因素有所不同，特別是海拔推移上受到更多限制。溫度相關變量及雨量變異為解釋高山森林擴展的主要因子。未來若氣溫持續上升，可能加速森林擴展，而乾旱則可能成為其主要限制條件。地形上，海拔高度為森林擴展與推移的關鍵限制因子，應是因為高海拔環境條件更為嚴苛；此外，坡度與地勢起伏亦為影響海拔推移的重要因子。綜上所述，本研究結合氣候與地形因素，深入分析臺灣高山森林擴展的趨勢與驅動機制，期望能更進一步了解氣候變遷對於臺灣高山森林生態系影響的全貌，以供未來制定高山森林保育與氣候調適策略相關研究資訊參考。Climate change may lead to species migration, subsequently altering biodiversity and ecosystem, structure and function. Alpine ecosystems are particularly sensitive to climate change, and shifts in the upper altitudinal limits of alpine forests have become a focus in climate change ecology. Most alpine forests have expanded as the constraints of low temperatures ease, leading to a gradual contraction of alpine vegetation zones above treelines. This trend poses risks such as potential species extinctions and alterations in ecosystem services. However, studies on altitudinal shifts of alpine forests in Taiwan remain relatively limited and have primarily focused on the Hehuan Mountain area. Given the diversity of Taiwan's alpine environments and the significant regional variations in conditions, research encompassing a broader range of locations is necessary to gain a more comprehensive understanding of the impacts of climate change on the distribution of Taiwan's alpine forests. This study utilizes aerial photographs from 1980, 2001, and 2021, combined with a U-Net deep learning model for image segmentation, to analyze the trends in alpine forest distribution in the Yushan and Central Mountain Range regions over the past 41 years. Spatial regression analysis is used to explore the impact of climatic and topographic factors on forest expansion. The study found that training the model with smaller image sizes yielded the best image segmentation results. Between 1980 and 2021, the alpine forest cover in the Yushan and Central Mountain Range study areas increased by approximately 197 hectares and 158 hectares, respectively. The treeline advanced to higher elevations at an average rate of 1.44 to 1.55 meters per year. The patterns of forest expansion exhibited significant spatial heterogeneity. Unlike in the Hehuan Mountain where previous studies indicated that forest expansion was dominated by tree densification in the ecotone, treeline advancement was the primary characteristic in Yushan and the Central Mountain Range. Regression model indicated that the main factors explaining area expansion and treeline elevation shifts were different, with treeline elevation shifts constrained by more factors. If temperature continues to rise in the future, forest expansion may accelerate, while drought will be a main limiting factor. Topographically, elevation is a critical constraint on both forest expansion and treeline shifts, possibly because environmental conditions are more challenging at higher altitudes. Additionally, slope and terrain variability significantly influenced treeline advancement. In summary, this study integrates climatic and topographical factors to comprehensively analyze the trends and potential mechanisms of alpine forest expansion in Taiwan. The findings help to provide a thorough understanding of the impacts of climate change on Taiwan's alpine forest ecosystems and serve as a reference for future research on conservation and climate adaptation strategies for alpine forests.