Sentinel-2, spectral indices and change detection from space.
A free, phone-friendly check for any farm: whether the rains have started, when they usually start, rain and heat for the next 10 days and how green the crops are. Examples from Lilongwe, Malawi, and Gazipur, Bangladesh.
Five new free tools for herders, district offices and cooperatives: a pasture and water bulletin, a water point check, a weekly fire count, flood water near a village today and a rainfall insurance check. Real answers from Kenya and South Sudan.
Two free tools for any district: is this season's rain and crop growth better or worse than usual, and where did it flood and how many people live there? Worked examples from Marsabit, Kenya, in 2022 and Feni, Bangladesh, in 2024.
An introduction to the TechnicalGIS climate time-series tools: five open datasets, how to draw an area on the map, worked examples from London, north Evia, Nairobi and Las Vegas, the processing limits and the terms for reusing every output.
Assess a real land cover map in Python: ESA WorldCover 2021 against 8,597 Eurostat LUCAS 2022 field survey points in Ireland. Build the confusion matrix, compute overall, user's and producer's accuracy, see why kappa is discouraged, and use the Olofsson et al. (2014) stratified estimators to get error-adjusted areas with 95% confidence intervals and to plan a new stratified random sample.
Run a pixel-wise Mann-Kendall trend test and Theil-Sen slope on 24 years of MODIS NDVI in Python. Read MOD13A1 from the Microsoft Planetary Computer, mask it with the QA layer, aggregate to growing-season means, vectorise the tests with xarray.apply_ufunc (checked against pymannkendall), compare the seasonal and modified Mann-Kendall tests, control the false discovery rate across 229,229 pixels, and map greening and browning on China's Loess Plateau.
Detect and quantify methane plumes in Python with NASA's EMIT imaging spectrometer: load open EMIT methane enhancement data from NASA VEDA without a login, separate three Dallas-Fort Worth landfill plumes from the noise, compute the integrated mass enhancement, estimate emission rates with ERA5 winds, and compare them with EPA greenhouse gas inventory figures.
Map earthquake damage from free Sentinel-1 radar in Python. Compute the log ratio and a z-score against a pre-event stack for Antakya after the February 2023 Türkiye earthquakes, aggregate to a 200 m grid, and score the map against 730 destroyed buildings mapped in OpenStreetMap.
Convert raster to vector in Python with rasterio.features.shapes on 10 m ESA WorldCover for Greater Manchester: polygonize water and built-up land, fix invalid polygons, simplify without gaps, rasterize ONS district boundaries for land-cover shares, then write and validate a Cloud Optimized GeoTIFF with rio-cogeo and measure HTTP range reads.
Estimate surface soil moisture from free Sentinel-1 radar in Python with the TU Wien change-detection method. Build dry and wet references per orbit, map relative soil moisture at 100 m, and test it honestly against four COSMOS-UK stations and their rain gauges.
Map blanket bog, heath, bracken, grassland, woodland, water and built-up land in the Peak District from 2023 Sentinel-2 seasonal composites with an eight-rule spectral index classifier, then assess it properly: a confusion matrix, user's and producer's accuracy, kappa and its caveats, block-bootstrap confidence intervals, and why a random train/test split overstates accuracy.
Read the open Meta/WRI global 1 m canopy height map from AWS with no login, compare it with ETH Global Canopy Height 2020, and validate both against a canopy height model made from Environment Agency LiDAR in the Forest of Dean: CRS alignment, bias, RMSE, hexbin plots, height classes and honest limitations.
Build an SBAS InSAR time series in Python from 364 COMET LiCSAR Sentinel-1 interferograms of the Svartsengi volcano in Iceland. Check unwrapping with phase closure, invert pixel by pixel with weighted least squares, project GNSS east, north and up into the line of sight, and validate six inflation–eruption cycles against the SENG station.
Write the classic SAR speckle filters (boxcar, Lee, refined Lee, enhanced Lee, Frost and Gamma-MAP) in NumPy and SciPy, run them on Sentinel-1 over Rutland Water, and compare them by equivalent number of looks, edge preservation and speed.
Classify nine upland habitats in the Peak District with LightGBM on 77 multi-temporal features: seasonal Sentinel-2 bands and indices, Sentinel-1 VV and VH backscatter and Copernicus DEM terrain. Tune with spatial-fold cross-validation and class weights, run an S2 / S2+S1 / S2+S1+terrain ablation, explain the model with TreeSHAP and permutation importance, and compare it with a random forest and the Part 1 rules on the same test blocks.
Invert Sentinel-1 InSAR for a volcanic source in Python: write Mogi point-source and Okada (1985) sill forward models in NumPy, downsample ascending and descending LOS maps with a quadtree, fit depth and volume change by weighted least squares with scipy.optimize, and estimate uncertainties with a Monte Carlo of correlated noise, compared with published results.
Mask clouds and cloud shadows in Sentinel-2 L2A imagery with Python. Compare the Scene Classification Layer (SCL), s2cloudless, OmniCloudMask and a simple threshold over the Lake District, measure how far they agree, score them against 186 reference points and build a cloud-free median composite.
Map snow cover in Python with the Normalised Difference Snow Index (NDSI) from Sentinel-2 and the daily MODIS snow product on the Microsoft Planetary Computer: choosing the NDSI threshold, separating snow from cloud, snow-covered area through the winter of 2025–26 in the Cairngorms, and the snowline by elevation and aspect with the Copernicus DEM.
Use Google DeepMind's AlphaEarth Foundations satellite embeddings, a geospatial foundation model, in Python without a login. Classify nine Peak District habitats with a linear model, compare a few-shot learning curve against raw Sentinel-2 bands, map uncertainty, check calibration, simulate active learning, and estimate habitat areas with confidence intervals from a stratified random sample (Olofsson et al. 2014).
Map glacier velocity in Python with NASA's ITS_LIVE Zarr datacubes on AWS, with no login: velocity maps, a flowline traced from the velocity field, 40 years of speed profiles, seasonal and multi-year time series from 80,568 filtered image pairs, and strain rates and flow directions on Jakobshavn Isbræ (Sermeq Kujalleq), Greenland's fastest glacier.
Map volcano deformation with Sentinel-1 InSAR in Python: stream free COMET LiCSAR interferograms of the Svartsengi volcanic system in Iceland, read wrapped and unwrapped phase and coherence, count fringes, convert phase to line-of-sight millimetres, and combine ascending and descending tracks into vertical and east–west motion, checked against GNSS.
Segment 2023 Sentinel-2 seasonal composites of the Peak District into image objects with scikit-image SLIC and Felzenszwalb, tune the segment size against Living England boundaries, compute per-object statistics, shape and GLCM texture with NumPy, classify the objects with LightGBM and compare OBIA with a pixel-based classifier on accuracy, salt-and-pepper noise and patch counts. Then export the habitat objects as GeoPackage polygons.
Classify habitats from a full year of cloudy, irregular Sentinel-2 observations with two temporal neural networks in PyTorch: a 1D convolutional TempCNN and a lightweight temporal attention encoder (L-TAE) with day-of-year positional encoding. Compare masking with gap-filling, test against a LightGBM baseline on spatial blocks, read which dates matter from attention and occlusion, and predict a wall-to-wall 10 m map on a CPU.
Measure shoreline change on one of Europe's fastest-eroding coasts with free Sentinel-2 imagery and Python: MNDWI and Otsu thresholding, sub-pixel shorelines with find_contours, shore-normal transects, a tide correction from open tide-gauge data, end-point and linear regression rates, and an honest comparison with East Riding of Yorkshire Council's cliff monitoring.
Map forest loss and deforestation in Python with the Hansen/UMD Global Forest Change v1.13 data: read the open tiles with no login, compute loss by year in hectares with correct pixel areas, compare Landsat and Sentinel-2 images before and after, and check the GFC loss year against Sentinel-2 NBR and MODIS burned area for central Rondônia, Brazil.
Run a trained PyTorch U-Net over a whole Sentinel-2 scene of a city it has never seen: tiled sliding-window inference on a large GeoTIFF with overlapping, blended windows, a georeferenced Cloud Optimised GeoTIFF of classes and confidence, raster-to-vector polygons with rasterio.features.shapes in a GeoPackage, and an honest check against reference labels.
Train a U-Net to segment Sentinel-2 imagery into buildings, built-up land, trees, grass and water with PyTorch, segmentation_models_pytorch and TorchGeo's Sentinel-2 pretrained ResNet-18 weights, on a CPU in five minutes. Real loss curves, per-class IoU and F1, a confusion matrix and an honest look at what limits accuracy.
Build a labelled deep learning dataset from free data: stream a Sentinel-2 L2A composite from the Planetary Computer, rasterise Overture building footprints over ESA WorldCover, cut georeferenced image chips, split them by spatial block so the test set does not leak, check the class balance and load it all as a TorchGeo dataset.
Map nitrogen dioxide air pollution from Sentinel-5P TROPOMI Level-2 data in Python: stream a full year of orbits from the Planetary Computer, apply the qa_value > 0.75 filter, grid swaths to a regular raster, build monthly and seasonal composites, rank cities and measure the weekend drop, with an honest explanation of what a tropospheric column is.
Map ground subsidence with Sentinel-1 InSAR in Python: stream 170 real interferograms of Mexico City, convert unwrapped phase to millimetres of line-of-sight motion, mask low coherence, invert a small-baseline network into a time series, and cross-check the velocity with MintPy.
Map wildfire burn severity in Python with free Sentinel-2 imagery: NBR calculation, cloud and water masking with SCL, dNBR with an offset, USGS severity classes and burned hectares, a burn perimeter, fire progression, RdNBR and an eight-year recovery record for the 2018 Saddleworth Moor fire.
Monitor crops through cloud with free Sentinel-1 radar. Build VV, VH and VH/VV ratio time series for 247 fields in Cambridgeshire, beat speckle by field-averaging, read the growth signatures of winter wheat and oilseed rape, and compare with the gaps that cloud leaves in Sentinel-2 NDVI.
Run Meta's Segment Anything Model on high-resolution aerial imagery with segment-geospatial (samgeo), on a CPU. Generate masks automatically, vectorise them to a GeoPackage, filter them by area, shape and NDVI to find buildings, and prompt SAM with points and boxes, checked against OpenStreetMap.
Calculate land surface temperature from Landsat 8 and 9 Collection 2 Level-2 data in Python: ST_B10 scale and offset, QA_PIXEL cloud masking, LST by land cover and NDVI, an urban-rural transect, hotspots with OpenStreetMap and an honest comparison with air temperature, for London in the July 2022 heatwave.
Supervised land cover classification in Python: build seasonal Sentinel-2 composites with odc-stac, sample labels from ESA WorldCover, train a scikit-learn random forest, test it with spatial cross-validation and map Sheffield and the Peak District.
Build a five-year Sentinel-2 NDVI time series in Python: load a lazy dask datacube with odc-stac, mask cloud with SCL, make monthly composites, extract field curves and derive crop phenology.
Detect ships in free Sentinel-1 SAR imagery with a cell-averaging CFAR detector written in NumPy and SciPy: mask the land, set a threshold from a speckle model, cluster detections into targets, reject wind turbines, sidelobes and ambiguities, compare VV with VH and export the vessels as GeoJSON.
Track chlorophyll, turbidity and peat colour proxies in four Peak District reservoirs with free Sentinel-2 imagery: cloud and water masking for small, dark upland lakes, NDCI and NDTI time series from 2019 to 2026, anomaly ranking, maps, tidy CSVs and an honest check against Environment Agency samples.
Map the 2022 Sindh floods with free Sentinel-1 radar from the Microsoft Planetary Computer: convert backscatter to decibels, remove speckle with a Lee filter, pick a water threshold with Otsu's method, and measure the flood extent and how it receded over seven months.
Search seven summers of Sentinel-2 imagery with pystac-client, stream only the bands you need from Cloud Optimised GeoTIFFs, mask clouds, compute NDWI and MNDWI and measure how much of Lake Mead dried up between 2019 and 2025.
A look at the dramatic fall in Lake Mead's water level during the 2022 drought.