Project Overview
This project provides an introduction to machine learning using Earth observation and hydrologic data. The first component focuses on land cover classification using Landsat imagery, where a supervised Random Forest model is trained to distinguish water from non-water pixels and estimate changes in lake surface area over time. This workflow includes image preprocessing, manual labeling, feature engineering, model training, and evaluation.
The second component applies machine learning to time series prediction by modeling streamflow using a Long Short-Term Memory (LSTM) neural network. Using historical hydrologic data, the LSTM captures temporal dependencies to forecast streamflow dynamics.
Together, these examples demonstrate how machine learning can be applied to both spatial (remote sensing) and temporal (hydrologic) datasets, providing a foundation for environmental data science and predictive modeling.