Many environmental systems are characterized by distinctive spatial features such as rainstorms, ocean currents, algae blooms, wildfires and chemical plumes, among others. Traditional data assimilation methods are not always able to preserve the structure of these features when they combine model forecasts with measurements. One way to address this problem is to reformulate the data assimilation process to more explicitly recognize the role of spatial structure. The resulting feature-based approach works with geometrical objects of uncertain size and shape. Many of the methods required to estimate these shapes from data rely on image-processing methods. However, it is also important to include physical constraints not always considered in image processing applications. This project will focus on developing new feature-based data assimilation methods that are applicable to a range ofapplications.