| Summary: | Gas hydrates, specifically methane hydrates, are sparsely sampled on a global scale, and their accumulation is difficult to predict geospatially. Several attempts have been made at estimating global inventories, and to some extent geospatial distribution, using geospatial extrapolations guided with geophysical and geochemical methods. The objective is to quantitatively predict seafloor total organic carbon and subsequently the geospatial likelihood of encountering methane hydrates. Predictions of TOC are produced using a sparsely observed dataset (Seiter et al., 2004) through a k-nearest neighbor (KNN) algorithm using 423 predictors and 7 nearest neighbors. KNN is unsupervised and non-parametric, as I do not provide any interpretive influence on prior probability distribution, so results are strictly data-driven. This TOC prediction, along with other global datasets (seafloor temperature, pressure, sediment thickness, and crustal heat flow) are used as parameters to train a KNN algorithm in identifying likely locations of methane and/or methane hydrate accumulation. I have selected as test sites several locations where gas hydrates have been well studied, each with significantly different geologic settings. These are: The Blake Ridge (U.S. East Coast), Hydrate Ridge (U.S. West Coast), and the Gulf of Mexico. I then use KNN to quantify similarities between these sites via the normalized distance in parameter space. Results on identification of likely methane and/or methane hydrate accumulation indicate the use of KNN as an unreliable method of identifying accumulation. However, global seafloor TOC predictions are reasonably accurate and have been incorporated to provide a potential analysis on hydrocarbon accumulation.
|
|---|