Nansat is a Python toolbox for analysing and processing 2-dimensional geospatial data, such as satellite imagery, output from numerical models, and gridded in-situ data. It is created with strong focus on facilitating research, and development of algorithms and autonomous processing systems. Nansat extends the widely used Geospatial Abstraction Data Library (GDAL) by adding scientific meaning to the datasets through metadata, and by adding common functionality for data analysis and handling (e.g., exporting to various data formats). Nansat uses metadata vocabularies that follow international metadata standards, in particular the Climate and Forecast (CF) conventions, and the NASA Directory Interchange Format and Global Change Master Directory keywords. Functionality that is commonly needed in scientific work, such as seamless access to local or remote geospatial data in various file formats, collocation of datasets from different sources and geometries, and visualization, is also built into Nansat. The paper presents Nansat workflows, its functional structure, and examples of typical applications.