Hyperspectral Remote Sensing Applications Training Course

Participants can match hyperspectral platforms and acquisition specifications to project requirements.

• Airborne, spaceborne and UAV hyperspectral platforms matched to project requirements
• Pushbroom, whiskbroom and snapshot acquisition geometries and their artefacts
• Sensor specifications: spectral sampling, bandwidth, swath, GSD and signal quality
• AVIRIS, PRISMA, EnMAP and DESIS mission characteristics and product availability
• Flight planning constraints: solar angle, altitude, overlap and atmospheric conditions
• Acquisition metadata needed for traceable downstream processing decisions

Participants can inspect, organise and prepare hyperspectral datasets for GIS-based analysis.

• Hierarchical data formats, interleave types and hyperspectral cube file structures
• Inspecting wavelengths, bad-band lists, metadata and no-data conventions
• Coordinate reference systems, georeferencing accuracy and orthorectification requirements
• GDAL workflows for conversion, tiling, compression and raster metadata checks
• QGIS visualisation of band composites, spectra and pixel inspection results
• Data management conventions for versioned scenes, derivatives and analytical outputs

Participants can define a defensible hyperspectral project scope, specification and reference-data plan.

• Decision questions translated into target classes, detection limits and map specifications
• Area-of-interest boundaries, sampling units and minimum mapping unit definitions
• Project charter defining users, decisions, accuracy targets and delivery constraints
• Reference-data plans combining field spectra, labels, imagery and laboratory results
• Risk register covering clouds, terrain, mixed pixels and target confusion
• Processing workflow diagram from raw acquisition through validated decision products

Participants can prepare radiometrically reliable image data and document quality-control decisions.

• Dark-current correction, detector striping and dropped-pixel remediation procedures
• Radiometric calibration from digital numbers to at-sensor radiance
• Smile, keystone and band-registration effects on spectral fidelity
• Destriping and denoising choices documented without masking target signatures
• Quality masks for saturation, clouds, shadows and unreliable detector regions
• Preprocessing audit trail recording parameters, software versions and input files

Participants can convert image radiance to fit-for-purpose surface reflectance products.

• Atmospheric absorption windows and aerosol effects on observed spectra
• FLAASH and empirical-line correction assumptions, inputs and output checks
• QUAC correction use cases, limitations and comparative quality assessment
• Reflectance validation against field panels and independent spectrometer measurements
• Topographic correction using illumination geometry, DEMs and surface slope
• Water-vapour bands and noisy spectral regions removed before analysis

Participants can identify spectral features that separate materials and conditions of interest.

• Mean spectra, standard deviation and continuum shape compared by class
• Continuum removal for absorption-feature position, depth and area measurements
• Spectral derivatives highlighting subtle absorption changes and wavelength shifts
• Band ratios and normalized indices designed from target physics
• Spectral libraries curated with provenance, measurement conditions and class labels
• Similarity screening using spectral angle, correlation and Euclidean distance

Participants can reduce hyperspectral dimensionality while retaining meaningful target information.

• Band selection guided by information content, noise and computational cost
• Principal component analysis for decorrelation and exploratory feature reduction
• Minimum Noise Fraction transformation separating coherent signal from noise
• Noise-whitening diagnostics using eigenvalues, images and component spectra
• Feature stacks combining selected bands, indices and spatial texture measures
• Dimensionality choices justified against training-sample volume and target separability

Participants can estimate material fractions and interpret spectral mixture uncertainty.

• Linear spectral mixture analysis assumptions, constraints and endmember selection
• Pixel Purity Index and N-FINDR methods for candidate endmember extraction
• Endmember spectra refined using scatterplots, libraries and field evidence
• Abundance maps interpreted with residual error and fraction-sum diagnostics
• Multiple Endmember Spectral Mixture Analysis for variable background materials
• Subpixel target detection requirements for minerals, contamination and built materials

Participants can detect known materials using spectral matching and rule-based methods.

• Spectral Angle Mapper classification using reference spectra and angular thresholds
• Matched filtering and Adaptive Cosine Estimator for known target detection
• Binary encoding and Spectral Feature Fitting for diagnostic absorption matching
• Rule-based decision trees combining spectral thresholds, masks and ancillary layers
• Class definitions separating materials with similar broadband appearance
• Post-classification filtering controlled to preserve small operational targets

Participants can train and interpret supervised machine-learning models for hyperspectral mapping.

• Training samples stratified by class, location, date and condition
• Random Forest classification tuned through variables, trees and class weights
• Support Vector Machines configured with kernels, C and gamma
• Cross-validation designed to prevent spatial autocorrelation and information leakage
• Feature importance interpreted cautiously alongside physical spectral knowledge
• Probability surfaces and rejection thresholds used for uncertain classifications

Participants can quantify, explain and communicate confidence in hyperspectral outputs.

• Confusion matrices reporting producer, user and overall accuracy
• F1 score, balanced accuracy and Cohen's kappa selected appropriately
• Area-adjusted accuracy estimates derived from probability sampling designs
• Independent validation samples separated from training and model-tuning records
• Uncertainty maps communicate low-confidence pixels, boundaries and unobserved conditions
• Error analysis directs additional labels, revised classes and processing corrections

Participants can apply hyperspectral methods to vegetation condition and crop-management questions.

• Red-edge, water and pigment absorption features linked to vegetation condition
• Continuum-removed metrics for chlorophyll, nitrogen and canopy water assessment
• Field plot data aligned with image footprints and phenological timing
• Regression models estimate biophysical variables with uncertainty intervals
• Stress detection distinguishes disease, nutrient deficiency, drought and soil exposure
• Crop management zones derived from validated spectral and spatial patterns

Participants can identify mineralogical targets and support geological interpretation using spectral evidence.

• Diagnostic mineral absorption features in VNIR and SWIR wavelengths
• USGS spectral libraries used to screen alteration mineral candidates
• Band-depth mapping and feature fitting for mineral abundance proxies
• Geological maps, DEM derivatives and field observations integrated for interpretation
• Mine waste characterisation identifies acid-generating materials and exposed sulphides
• Target maps ranked by confidence, accessibility and verification priority

Participants can design hyperspectral monitoring outputs for water, urban and environmental investigations.

• Water-quality retrieval challenges from adjacency effects, turbidity and bottom reflectance
• Algal pigment and suspended-sediment indicators calibrated against sampled water
• Urban material mapping separates roofs, asphalt, concrete, vegetation and soil
• Contaminant plume screening combines target spectra, masks and wind context
• Change detection compares co-registered reflectance products with seasonal controls
• Operational monitoring thresholds linked to inspection, sampling or remediation actions

Participants can establish a repeatable, reviewable workflow for operational hyperspectral analysis.

• Standard operating procedures defining inputs, parameters, quality gates and outputs
• Metadata records supporting reproducibility, audit review and future scene comparison
• Processing time, storage and computing requirements estimated before project mobilisation
• Supplier statements of work specifying products, validation evidence and acceptance criteria
• Map symbology designed to distinguish detections, probabilities and uncertainty zones
• Technical reports linking spectral evidence to decisions, actions and limitations

Participants can build, review and present a complete hyperspectral analysis project addressing a defined operational question.

• Project brief converted into a defensible hyperspectral analysis and validation plan
• Raw scenes processed through documented correction, masking and feature-generation steps
• Target model selected, trained and applied against stated decision criteria
• Accuracy, uncertainty and limitations reported using independent reference evidence
• Maps, spectra and methods packaged for technical and nontechnical reviewers
• Peer review addresses reproducibility, interpretation risks and operational recommendations
• Final presentation defends findings, assumptions and next-step data requirements