Environmental Data Analytics

Catalog description

With the rapid increase in quality and quantity of environmental data, emerging data-driven analytical methods are becoming important for discovering patterns and making predictions about our Earth and environmental systems. The datasets, such as in-situ measurements, remote sensing observations acquired by satellite, airborne, and UAV systems, Earth system modeling outputs, and energy datasets, are generally characterized by different spatial and temporal scales and coverage, complex data structures, high dimensionality, and large data volume. It is important to understand the principles, strengths, and limitations of data-driven methods for Earth and environmental science applications, and to learn the practical skills for applying them to real-world datasets. This course introduces various data analytical methods focused on machine learning for Earth and environmental sciences. A range of supervised and unsupervised methods for regression, classification, and clustering problems will be discussed with real-world examples, including but not limited to climatological data, biodiversity data, remote sensing imagery classification, and geomorphological analysis. This 3-credit course is designed to include lectures and class projects. The lectures will focus on the basic principles and environmental and earth science applications of machine learning algorithms, including regression (e.g., OLS, Ridge, LASSO, PCR), gradient descent optimization, classification (e.g., logistic regression, naïve Bayes, decision trees, random forest, support vector machine), neural networks and deep learning, and clustering (e.g., k-means, hierarchical, self-organizing map) techniques, etc. The class projects will include practical programming exercises in using these techniques for various real-world datasets. Students are encouraged to bring datasets from their own domain (e.g., geography, geoscience, materials science, meteorology, and energy) for analysis in the final project.