坦尚尼亞河口甲烷排放量之遙感技術量化研究

Abstract

本研究以整合性方法量化坦尚尼亞主要河口的甲烷排放，結合遙測技術、次級與衍生的實地測量資料，以及先進的機器學習技術。透過 Google Earth Engine（GEE）與多源衛星資料，本研究建立一套完整的方法論，以估算 Rufiji、Wami、Pangani 和 Ruvuma 四大河口的甲烷通量。研究納入多項環境變數，包括氣溫、降雨量、水體範圍、酸鹼值（pH）、標準化差異水指數（NDWI）及濕度，以探討其對甲烷生成與排放模式的影響。所採用的機器學習模型包括隨機森林（Random Forest）、支持向量迴歸（Support Vector Regression）與梯度提升（Gradient Boosting），成功解析多變量間的複雜關係，預測準確度高達 91%（R² = 0.91）。 研究結果顯示，各河口甲烷排放具有顯著的空間異質性，其中 Rufiji 河口的平均排放量最高（42.3 毫克 CH₄／平方米／日），依序為 Ruvuma（39.7 毫克）、Wami（35.6 毫克）與 Pangani（28.9 毫克）。季節變化亦十分明顯，雨季排放量普遍高於旱季 27–45%。氣溫為甲烷排放最強的預測因子（相關係數：0.69），而酸鹼值則呈顯著負相關（-0.45）。 本研究為坦尚尼亞河口甲烷動態提供首份全面性評估，並建立一套可應用於其他熱帶沿海系統的方法框架。研究成果有助於提升坦尚尼亞溫室氣體清冊的準確性，並為沿海濕地管理的氣候變遷緩解策略提供科學依據。

This study quantifies methane emissions from major estuaries in Tanzania using an integrated approach combining remote sensing, secondary/ derived field measurements, and advanced machine learning techniques. Using Google Earth Engine (GEE) and multi-source satellite data, this research develops a comprehensive methodology to estimate methane fluxes across the Rufiji, Wami, Pangani, and Ruvuma estuaries. The study incorporates multiple environmental variables including temperature, rainfall, water body extent, pH, Normalized Difference Water Index (NDWI), and humidity to understand their influence on methane production and emission patterns. Machine learning models including Random Forest, Support Vector Regression, and Gradient Boosting were employed to analyze complex multivariate relationships, achieving prediction accuracy of up to 91% (R² = 0.91). Results reveal significant spatial heterogeneity in methane emissions across estuaries, with the Rufiji estuary exhibiting the highest mean emissions (42.3 mg CH₄ m⁻² d⁻¹), followed by Ruvuma (39.7 mg CH₄ m⁻² d⁻¹), Wami (35.6 mg CH₄ m⁻² d⁻¹), and Pangani (28.9 mg CH₄ m⁻² d⁻¹). Pronounced seasonal variations were observed, with wet season emissions exceeding dry season values by 27-45% across all sites. Temperature emerged as the strongest predictor of methane emissions (correlation coefficient: 0.69), while pH showed a significant negative relationship (-0.45). The study provides the first comprehensive assessment of methane dynamics in Tanzanian estuaries and establishes a methodological framework applicable to other tropical coastal systems. These findings contribute to improved greenhouse gas inventories for Tanzania and inform climate change mitigation strategies for coastal wetland management.