Introduction to Satellite Remote Sensing (continued) Miles G. Logsdon, University of Washington, School of Oceanography, mlog@u.washington.edu Miles Logsdon, Univ. of Washington Oceanography SeaWiFS, June 27, 2001

We “approach” RS in two ways To classify or group thematic land surface materials To detect a biophysical process

Cluster and Classify

Spectral Profile

Spatial Profile

Spectral Signatures

1d classifier

Spectral Dimensions

3 band space

Clusters

Dimensionality N = the number of bands = dimensions …. an (n) dimensional data (feature) space Measurement Vector Mean Vector Feature Space - 2dimensions 190 85 Band B Band A

Spectral Distance * a number that allows two measurement vectors to be compared

Classification Approaches Unsupervised: self organizing in multi-dimensions Easy, quick, something you should be able to do Supervised: training with spectral signatures More thought, planning ahead, for target application Hybrid: self organization by categories Just what it implies Spectral Mixture Analysis: sub-pixel variations Seeks the composition of the pixel signatures

Clustering / Classification Clustering or Training Stage: Through actions of either the analyst’s supervision or an unsupervised algorithm, a numeric description of the spectral attribute of each “class” is determined (a multi-spectral cluster mean signature). Classification Stage: By comparing the spectral signature to of a pixel (the measure signature) to the each cluster signature a pixel is assigned to a category or class.

terms Parametric = based upon statistical parameters (mean & standard deviation) Non-Parametric = based upon objects (polygons) in feature space Decision Rules = rules for sorting pixels into classes

Unsupervised Clustering Minimum Spectral Distance ISODATA I - iterative S - self O - organizing D - data A - analysis T - technique A - (application)? Band B Band A Band B Band A 1st iteration cluster mean 2nd iteration cluster mean

ISODATA clusters

Part of the classification process: ISODATA clustering algorithm Band 4 Possible pixel assignments Band 3 Cluster center (mean) Each data pixel is assigned to a cluster based on the distance of a pixel from the center of a cluster (“Euclidean distance”)

Supervised Classification Assigning spectral signatures to clusters by selecting pixel in “geographic space”

Supervised Classification Assigning spectral signatures to clusters by selecting pixel in “Feature space”

Classification Decision Rules If the non-parametric test results in one unique class, the pixel will be assigned to that class. if the non-parametric test results in zero classes (outside the decision boundaries) the the “unclassified rule applies … either left unclassified or classified by the parametric rule if the pixel falls into more than one class the overlap rule applies … left unclassified, use the parametric rule, or processing order Non-Parametric parallelepiped feature space Unclassified Options parametric rule unclassified Overlap Options by order Parametric minimum distance Mahalanobis distance maximum likelihood

Parallelepiped Maximum likelihood (bayesian) Minimum Distance probability Bayesian, a prior (weights) Band B Band A Minimum Distance Band B Band A cluster mean Candidate pixel

Parametric classifiers

Class Names or “Classification Systems” USGS - U.S. Geological Survey Land Cover Classification Scheme for Remote Sensor Data USFW - U.S. Fish & Wildlife Wetland Classification System NOAA CCAP - C-CAP Landcover Classification System, and Definitions NOAA CCAP - C-CAP Wetland Classification Scheme Definitions PRISM - PRISM General Landcover King Co. - King County General Landcover (specific use, by Chris Pyle) Level 1 Urban or Built-Up Land 11 Residential 12 Commercial and Services 13 Industrial 14 Transportation, Communications and Utilities 15 Industrial and Commercial Complexes 16 Mixed Urban or Built-Up 17 Other Urban or Built-up Land 2 Agricultural Land 21 Cropland and Pasture 22 Orchards, Groves, Vineyards, Nurseries and Ornamental Horticultural Areas 23 Confined Feeding Operations 24 Other Agricultural Land

Resolution and Spectral Mixing Thanks to: Robin Weeks

Detecting a Process: Two examples Using “band math”

Laboratory Spectral Signatures II Common Urban Materials Healthy grass Concrete Astroturf wavelength Thanks to Robin Weeks

Vegetation: Pigment in Plant Leaves (Chlorophyll) strongly absorbs visible light (0.4 to 0.7 μm) Cell Structure however strongly reflects Near-IR (0.7 – 1.1 μm) Thanks to Robin Weeks

(courtesy http://earthobservatory.nasa.gov) NDVI When using LANDSAT: Simple Ratio Band 3 Band 4 NDVI Band 4 - Band 3 Band 4 + Band 3 (courtesy http://earthobservatory.nasa.gov)

Ocean Color Let’s begin with phytoplankton Phyton = plant; planktos = wandering. These reproduce asexually, are globally distributed, consist of 10s of thousands of species and make up about 25% of the total planetary veg. These are the grass that the zooplankton graze upon. And, they fix carbon as well.

Chaetoceros species of diatoms: cells are 20-25 mm in diameter. Chloroplasts contain pigments Chaetoceros species of diatoms: cells are 20-25 mm in diameter.

Water provides an internal standard shape for spectral comparison with other variable components Slopes for pigments and CDOM similar from 440 to 600 nm, but are opposite from 400 to 440 nm Note that detritus is include with CDOM since shapes are similar Spectral de-convolution of pigment absorption from CDOM absorption is straight-forward Shapes of phytoplankton or pigment absorption are not constant (next slide) For Case 2 waters, ratio of CDOM to chlorophyll a is not constant Strategy for Spectral Separation of Absorption Components with Semi-Analytic Algorithm Ken Carder: University of South Florida

Colored Dissolved Organic Material (CDOM) Organic Sources Terrestrial CDOM decay vegetation from river and nearshore Ocean CDOM detritus - cell fragments, zooplankton fecal Inorganic Sources Sand & Dust => Errosion rivers, wind, wave or current suspension

What’s the difference between MODIS chlorophylls? “Case 1” waters: Chlor_MODIS (Clark) This is an empirical algorithm based on a statistical regression between chlorophyll and radiance ratios. “Case 2” waters: Chlor_a_3 (Carder) This is a semi-analytic (model-based) inversion algorithm. This approach is required in optically complex “case 2” (coastal) waters and low-light, nutrient-rich regions (hi-lats). A 3rd algorithm was added to provide a more direct linkage to the SeaWiFS chlorophyll: “SeaWiFS-analog” Chlor_a_2 (Campbell) SeaWiFS algorithm OC4.v4 (O’Reilly) Ken Carder: University of South Florida

R(l) Florescence Independent of Chl-a Chl-a increasing

Case 1 Rrs Model with superimposed MODIS bands 8-14: All variables co-vary with chlorophyll a Note that slopes between blue and green wave lengths decrease with increasing chlorophyll, explaining the strategy of empirical algorithms Case 2 waters are more complicated Ken Carder: University of South Florida

SeaWiFS empirical OC4 algorithm for Chl-a; Called a maximum-band ratio alg.

Flying