Key terms used throughout the EO Python Foundations course.
Cell A block in a Jupyter notebook. Code cells contain runnable Python; Markdown cells contain formatted text.
Kernel The Python process that executes code in a notebook. Variables persist until you restart the kernel.
f-string
A string prefixed with f that embeds expressions in {curly braces}:
f'Cloud cover: {cloud:.1f}%'
List comprehension
A compact way to build a list from a loop:
[x * 2 for x in range(5)]
Dictionary (dict)
A key → value mapping: {'tile': 'T32UQD', 'cloud_cover': 12}. JSON and STAC metadata use the same structure.
pathlib.Path
A cross-platform way to work with file paths in Python:
Path('data') / 'scene.csv'
pandas Python library for tabular data. A DataFrame is a table with named columns; a Series is a single column.
DataFrame A 2-D table of data with named columns (pandas’ main object).
Series A 1-D labelled array (a single column in pandas).
NumPy Library for fast numerical arrays. Every satellite raster is stored as a NumPy array under the hood.
dtype
“Data type” of an array (e.g., uint16, float32, bool). Affects memory use and numeric behavior.
NaN “Not a Number” — a special floating-point value used to represent missing/invalid data in NumPy/pandas.
Matplotlib
The standard Python plotting library. Key plot types for EO: hist, scatter, imshow.
xarray Multi-dimensional labelled array library. Extends NumPy with named dimensions, coordinates, and metadata — ideal for data cubes.
DataArray xarray object representing a labelled N-D array (data + dims + coords + attributes).
Dataset xarray object holding multiple DataArrays that share coordinates (like a “netCDF in memory”).
Dimension (dim)
A named axis in xarray (e.g., time, y, x, band).
Coordinate (coord)
Label values attached to a dimension (e.g., actual timestamps for time, lat/lon or projected coordinates for y/x).
Dask Parallel computing library that enables out-of-core computation on datasets larger than memory. Integrates with xarray via chunked arrays.
Chunk / chunking Splitting an array into blocks so computations can run lazily and in parallel (common with Dask-backed xarray).
NDVI (Normalised Difference Vegetation Index)
(NIR − Red) / (NIR + Red) — ranges from −1 to 1, with higher values indicating healthier vegetation.
Scene A single satellite acquisition covering a specific area at a specific time.
Tile
A fixed grid cell in a satellite tiling scheme (e.g., Sentinel-2 MGRS tile T32UQD).
Cloud cover The percentage of a scene obscured by clouds. Low cloud cover means clearer imagery.
Cloud mask A boolean array (True = cloud, False = clear) used to exclude cloudy pixels from analysis.
Masking
Setting pixels to NaN or filtering them out based on quality flags (clouds, shadows, snow) before analysis.
NoData
A special value stored in rasters to indicate missing pixels (e.g., 0, -9999). Often needs converting to NaN for analysis.
Raster A grid of pixels representing spatial data — e.g., an NDVI image stored as a 2-D NumPy array.
Band A single channel of a satellite image capturing a specific wavelength range (e.g., Red, NIR, SWIR).
Composite An image formed by stacking or combining multiple bands or scenes — e.g., an RGB composite from Red, Green, Blue bands.
Reflectance The fraction of incoming sunlight reflected by a surface. Satellite sensors measure reflectance at different wavelengths.
Spatial resolution / pixel size The ground size represented by one pixel (e.g., 10 m). Smaller values mean finer spatial detail.
Extent / footprint The spatial coverage of a dataset. A bbox is a rectangle approximation; a footprint can be the true (possibly irregular) outline.
CRS (Coordinate Reference System) Defines how coordinates map to locations on Earth. Common examples: WGS 84 (EPSG:4326), UTM Zone 32N (EPSG:32632).
EPSG code A numeric identifier for a CRS definition (e.g., EPSG:4326 for WGS84 latitude/longitude).
Bounding box (bbox)
A rectangle defined by [min_lon, min_lat, max_lon, max_lat] used to specify a geographic search area.
AOI (Area of Interest) The geographic region you want to analyse.
Reprojection Transforming data from one CRS to another.
Resampling (spatial) Changing pixel size or grid alignment (often happens during reprojection). Common methods: nearest, bilinear, cubic.
CSV (Comma-Separated Values)
A plain-text tabular format. Loaded with pandas.read_csv().
JSON (JavaScript Object Notation) A text format for structured data using key-value pairs and arrays. STAC metadata is JSON.
GeoJSON A JSON format for vector geometries (points/lines/polygons) and their properties.
GeoTIFF A raster image format that includes geographic metadata (CRS, extent).
COG (Cloud Optimised GeoTIFF) A GeoTIFF structured with internal tiling and overviews so that remote clients can read only the part they need via HTTP range requests.
Overviews Lower-resolution copies stored inside a raster (e.g., COG) to enable fast zoomed-out display and quick reads.
NetCDF (Network Common Data Form) A self-describing format for multi-dimensional arrays, common in climate science.
Zarr A chunked, compressed array format designed for cloud storage. Supports parallel access.
NPZ NumPy's compressed archive format for saving multiple arrays to a single file.
STAC An open specification for describing geospatial data using standardised JSON.
STAC API A web API that supports searching STAC catalogs (typically by space, time, and collection).
STAC search parameters
Common fields used when searching: bbox, intersects, datetime, collections, query, limit.
Catalog Top-level container that organises STAC Collections.
Collection A group of related Items sharing common properties — e.g., all Sentinel-2 Level-2A scenes.
Item A single spatiotemporal unit (one scene) with metadata and links to Assets.
Asset An actual data file linked from an Item — e.g., a COG for band B04.
pystac-client Python library for searching STAC APIs.
Planetary Computer Microsoft's platform providing STAC-catalogued EO data with a free Jupyter computing environment.
Signing Planetary Computer often requires “signing” asset URLs (adding temporary auth parameters) before reading remote COGs.
stackstac Python library that loads STAC items directly into an xarray DataArray backed by Dask.
Vectorised operation Applying a function to an entire array at once (no Python loop), leveraging NumPy's C backend for speed.
Boolean indexing
Filtering rows or pixels using a True/False array:
df[df['cloud_cover'] < 20]
GroupBy Splitting data by a categorical column, applying a function to each group, and combining the results.
Lazy evaluation
Deferring computation until results are explicitly requested (e.g., .compute() in Dask).
Resampling (temporal) Changing the temporal resolution of a time series — e.g., daily → monthly.
Last updated: February 2026