Satellite Remote Sensing Data Bootcamp With Opensource Tools

Explore the theory and principles of collecting optical remote sensing data, focusing on passive sensing, multispectral bands, and spectral reflectance to map vegetation, forests, and land use with Landsat.

Explore optical remote sensing data types, from Modis and Landsat at regional and global scales to Sentinel-2’s 10-meter bands, accessible via Earth Explorer and Snap for vegetation and landscape monitoring.

Learn to download Landsat data from the USGS Earth Explorer by creating an account and selecting coordinates or map areas, choosing Landsat 8 collection one data and level one data.

Learn to download Landsat optical data from Earth Explorer using the semi-automatic classification plugin in QGIS, then view, download, and pre-process bands.

Master optical remote sensing principles and Landsat, MODIS, and Sentinel data. Download Landsat data from USGS, grasp band designations, and pre-process for land use maps using open-source tools.

Explore the theory of pre-processing Landsat optical data, address atmospheric effects, and perform atmospheric correction to obtain top of the atmosphere reflectance with open-source tools.

Implement atmospheric correction on Landsat data in R using the artist toolbox, converting to top-of-atmosphere radiance and apparent reflectance. Read metadata, stack bands, and apply dark object subtraction.

Learn to perform Landsat 8 pre-processing in QGIS using the semi-automatic classification plugin, including radiometric and atmospheric corrections to derive surface reflectance.

Conclude section three by outlining atmospheric corrections and dark object subtraction for Landsat data, demonstrate pre-processing steps, and preview the data products derived from Landsat data.

Learn to stack and unstack raster bands in QGIS, including 10-band stacks, using the pre-processing plugin to split stacked rasters into individual bands and save to a folder.

Compute band ratios and vegetation indices like ndvi in R and QGIS using raster calculator, then read data, set the working directory, and generate rasters to reveal landscape patterns.

Explore texture metrics theory, from first-order statistics to second-order co-occurrence (GLCM), including texture bands, to describe spatial texture in raster imagery and improve classification and forest carbon modeling.

Compute texture metrics with the ESA SNAP raster image analysis using grade level co-occurrence metrics on the H.H. data, and save results or export as texture tiff.

Explore the theory and application of tasseled cap transformations on multi-band Landsat data. This linear dimensionality reduction yields brightness, greenness, and moisture insights for forestry, soil science, and geology.

Compute tasseled cap transformations for Landsat 8 data in GRASS GIS using the Itasca module, reading bands b2 to b7, to derive four caps: brightness, greenness, wetness, and atmospherically corrected.

Learn the theory and applications of dimension reduction, using principal component analysis to reduce hyperspectral and multispectral bands while preserving maximum variance and improving classification accuracy.

Learn to perform dimensionality reduction on a stack of raster bands using principal component analysis in QGIS with the BCA plugin, producing uncorrelated principal components and revealing data patterns.

Revisit band ratios, vegetation indices, texture matrix theory, Bethel cap transformations, and dimension reduction in GRASS GIS and easy snap, with a preview of unsupervised and supervised classification.

Learn the theory of supervised classification in satellite remote sensing using training sites to define spectral signatures. Compare algorithms such as minimum distance, maximum likelihood, and spectral angle mapper.

Learn the preliminary steps for supervised land-use classification in QGIS, including stacking Landsat 8 bands, creating training signatures, and digitizing macro-class signatures for vegetation, bare earth, and degraded vegetation.

Perform supervised classification in QGIS using training data and maximum likelihood, then assess the output with a validation shapefile to report overall accuracy.

Explore supervised classification of Landsat data for forests, degraded, and bare areas across three classes using a random forest, with reflectance conversion, band selection, and model validation on unseen data.

Conclude section five by summarizing unsupervised and supervised classification theory and applying training data polygons with machine learning, using train-test splits for validation and accuracy metrics.

Discover why active remote sensing with synthetic aperture radar offers cloud-free, all-weather insights into forest canopy structure and height, using dual-polarization data such as HH and HP.

Learn how to obtain freely available ALOS PALSAR and PALSAR-2 data from JAXA, register to access, and download forest and non-forest datasets for 2010.

Learn pre-processing of lost pulsar alos palsar data, applying calibration, speckle reduction with leaf filters, and terrain correction when needed, then convert digital numbers to backscatter to obtain sigma naught.

Learn to obtain backscatter values from ALOS PALSAR data using R, compute sigma in linear form, and derive the radar forest degradation index (RFI) from HH and HB polarization data.

Learn to assign legends in QGIS by using right-click properties and unique values to classify a 12-category map, customize color ramps, and label categories like dry and wet mountains.

Learn how distributed computing frameworks coordinate parallel tasks across multiple nodes, enabling fault tolerance and scalable performance. Hadoop and Apache Spark illustrate these ideas in big data processing.