UBC Theses and Dissertations
Characterization and prediction of methane spatial heterogeneity within and beyond a flux tower footprint Ng, Darian Chi Yung
Freshwater wetlands are effective nature-based solutions to climate change for their unique ability to sequester atmospheric carbon for extended periods of time. However, they simultaneously emit large amounts of methane gas (CH₄). CH₄ is a highly potent greenhouse gas, and its spatial heterogeneity in wetlands is not yet fully understood. This study is the first of its kind and explores a novel approach to characterizing and predicting the spatial distribution of CH₄ fluxes (FCH₄) within a single flux tower footprint using a combination of remote-sensing via the Landsat 8 satellite and eddy covariance (EC) Footprint-weighted Flux Maps. There were two research objectives: 1) to determine whether there exist significant relationships correlating remote-sensing with within-footprint FCH₄, and 2) to investigate the potential of extrapolating and upscaling these relationships to predict FCH₄ beyond the flux tower. The Normalized Difference Vegetation Index (NDVI), Water Index (NDWI), Moisture Index (NDMI), and Land Surface Temperature (LST) remote-sensing products were collected from Landsat 8, and EC data from four wetland sites, MBPPW1, MBPPW2, US-Myb, and US-WPT were used for the Flux Maps. Hierarchical clustering was used on the Flux Maps to identify distinct FCH₄ features, and then applied to Landsat 8 to process the data for use in two stages of regression analysis. The first stage used simple linear regression to assess the relationships between FCH₄ and the four remote-sensing indices. The second stage used multiple linear regression (MLR) to model the spatial distribution of FCH₄. The simple linear regression analysis demonstrated strong relationships between remotely-sensed surface conditions and FCH₄, and found stronger FCH₄ from open water at MBPPW2 and US-WPT, and stronger FCH₄ from the emergent vegetation at US-Myb. MBPPW1 demonstrated a limitation of this approach requiring sites to have strong FCH₄ signals. The MLR models supported the feasibility of upscaling using remote-sensing by capturing much of the FCH₄ variance, but had reduced accuracy from strong error-biases and low regression slopes. This study demonstrates the applicability of its novel methodology to upscaling FCH₄ using remotely-sensed data and paves the way for future studies aiming to improve FCH₄ estimates from wetlands at a regional level.
Item Citations and Data
Attribution-NonCommercial-NoDerivatives 4.0 International