HyspIRI: The potential of Combined

VSWIR-TIR Analysis
Dar A. Roberts
UC Santa Barbara

Acknowledgements: NASA HyspIRI Preparatory Program

Images courtesy of R.O.Green

Outline

The Personal Stuff

What is HyspIRI?
Vegetation and the VSWIR
TIR Remote Sensing
VSWIR-TIR Synergies
The HyspIRI Airborne Campaign
Case Study
HyspIRI: Vegetation Species, Composition and
LST in Santa Barbara

The Personal Stuff: Academic History

Biology Major at UCI
Environmental
Biology/Geology Dbl
Major: UCSB
MA Applied Earth
Sciences: Stanford
PhD Geological Sciences:
University of Washington
Returned to UCSB in
1994 as a Geographer!
Moved slightly
Northwest
Started Here

Got my PhD
Gardened a lot
Wrote music
Came back

Went Here

Research Interests: Wildfire

Photograph of the Station Fire behind the Jet Propulsion Laboratory from Philipp Schneider

Pre-fire Conditions
Live fuel moisture, fuel types
Fire Danger
Weather, live fuel moisture
Active fire
Fire temperature, fire area and perimeters
Post-fire conditions
Burned area, Fire severity, Post-fire recovery

Pre-fire Fuels and Post-fire Severity

dNBR, NPV-GV-Ash mixing model, Lidar Canopy Height

Land-cover Change in Amazonia

Land-cover change in
Amaznia
Land-cover (1985-2012)
Deforestation
Forest degradation
Secondary forest
persistence
Pasture sustainability

Mapping Trace Gasses

VSWIR Mapping of
Methane
Carbon dioxide
Water Vapor

Methane mapped at RMOTC (left) and Coal Oil Point (right)

Urban Remote Sensing

Lidar height map vs MESMA height (left), MESMA cover and classes

VSWIR-Thermal Fusion

Map of cover, species and temperature (left) and scatterplot of LST vs GV Cover (right)

Meet the Viper Lab

What is HyspIRI?
Mission Urgency

Key Science and Science Applications

The HyspIRI science and application objectives are important

today and uniquely addressed by the combined imaging
spectroscopy, thermal infrared measurements, and IPM direct
broadcast.
Ecosystems

Climate: Ecosystem biochemistry, condition & feedback; spectral

albedo; carbon/dust on snow/Ice; biomass burning; evapotranspiration
Ecosystems: Global plant functional-type, physiological condition,
and biochemistry including agricultural lands.
Fires: Fuel status, fire occurrence, severity, emissions, and patterns of
recovery globally.
Coral reef and coastal habitats: Global composition and status.
Volcanoes: Eruptions, emissions, regional and global impact.
Geology and resources: Global distributions of surface mineral
resources and improved understanding of geology and related hazards.

Snow & ice

Fires

Evapotranspiration

Coastal
Volcanoes Habitats

Measurement

Status: Pre Phase A

Preliminary Draft Program Level 1 Requirements: Stable
Payload:
Imaging Spectrometer,
Thermal Infrared Imager
IPM-Direct Broadcast subset
Spacecraft: Small
Launch Vehicle: Taurus-class
Trajectory or Orbit: LEO, Sun sync. 10:30
Concept S/C & Instrument Mass: 561 kg (30% margin)
Concept S/C & Instrument Power: 650W (66% margin)

1
0.9

Relative Spectral Response

Imaging Spectrometer (VSWIR)

- 380 to 2500 nm in 10nm bands
- 30 m spatial sampling
- 16 day revisit
-Global land and shallow water
Thermal Infrared (TIR):
- 8 bands between 4-12 m
- 60 m spatial sampling
- 5 days revisit
- Global land and shallow water
IPM-Direct Broadcast

Mission Concept

H1 (m21)
H2 (m28)
H3 (a10)
H4 (a11)
H5 (a12)
H6
H7
H8 (m32)

0.8
0.7
0.6
0.5
0.4
0.3

The HyspIRI mission concept provides a mature, low risk and

comparatively low cost path to achieve theDecadal Survey Earth
Science and Application objectives.

0.2
0.1
0
3.00

4.00

5.00

6.00

7.00

8.00

9.00

10.00

11.00

12.00

13.00

VSWIR Spectroscopy: Biochemistry

Mid-IR Plateau (830 nm)
0.7

Anthocyanin
Lignin-Cellulose

0.6

Reflectance

0.5

Water

Red
Edge

0.4

Water

0.3

AcerLf

Carotenoids
0.2

Acerlit
Betula
Fagus

0.1
0
350

850
Chlorophyll

1350

Wavelength (nm)

1850

2350

VSWIR Spectroscopy
Plant Spectra and Structure
0.7
860

LAI1.79

0.6

LAI2.76
LAI3.78

Reflectance

0.5

LAI5.19

0.4

LAI8.75

1650

0.3
970

550

0.2

1240

0.1
0
400
480

900
660

1400

Wavelength (nm)

1900

2400

Species and Plant Functional Types

Provided species or PFTs have unique chemistry, anatomy,

architecture or phenology, they can be discriminated

Mapping Plant Species using MESMA

Complexity: 3,2,1 RGB

Multiple Endmember Spectral Mixture Analysis (MESMA)

Class

Decomposes mixed spectra to estimate type of material present (endmember) and Composition (cover)
Number and types of endmembers varies per pixel

Products

Composition: NPV-GV-Soil
RGB

Model Complexity (Number of endmembers present)

Class (Model selected, 2em case)
Fractional Cover (NPV-GV-Soil-Impervious composition)

Spectral Diversity of Cover Classes

Thermal Remote Sensing

Two key elements
Temperature (kinetic temperature of the object)
Emissivity (material composition)

Temperature Emissivity Separation (TES)

Emitted radiance can be lowered either by a decrease in
surface temperature (T) , or decrease in emissivity (). To
retrieve surface temperature, this ambiguity must be
resolved. This is TES
L =

Atmospheric Correction
Surface temperature is retrieved from at sensor radiance
after accounting for
Atmospheric emitted radiance
Atmospheric transmittance
Reflected downwelling radiance

Urban Heat Islands

1997 Day time TIR image of
Atlanta Georgia
Source: Quattrochi, HyspIRI
Thermal IR (TQ4) Science
Questions .
Image source: Atlas, May 1997
Atlas: 15 channel system
6 TIR, 1 MIR, 8 VNIR-SWIR
http://www.ghcc.msfc.nasa.gov/atlanta/

Urban areas are warmer than natural areas

Low albedo surfaces
Lack of vegetation cover and shading

Urban areas can create their own weather

TIR and Plant Physiology

Temperature is inversely related
to plant cover
Higher temperatures imply lower
rates of ET and photosynthesis

Visible (160x120)

TIR (160x120)

TIR Emissivity Complements VSWIR

AVIRIS reflectance spectra (left) and MASTER emissivity spectra (right).

Biotic materials (green-live plants and senesced materials) are shown above,
abiotic materials (beach sand, roofs, asphalt surfaces) are shown below.

Why Cover and Emissivity Matters

Assumed blackbody underestimates

temperature (a)
Default emissivity of 0.95 over or
underestimates it (b)
The error varies with conditions (c)

TIR Spectroscopy

TIR reflectance varies with biochemistry

Contributed by Meerdink
Live plants are near black bodies,
senesced plants are not

VSWIR TIR Synergies for Vegetation

VSWIR
Species composition
Fractional cover
Canopy structure (LAI)
Canopy chemistry
Photosynthetic function

TIR
Temperature
Stress measure
Evapotranspiration

Emissivity
Species composition
Canopy chemistry

0.7

Water

Anth

Ligno-cellulose

0.6
0.5

Reflectance

0.4

Car
Chl

0.3
0.2

AcerLf
Acerlit

0.1

Betula
Fagus

0
350

850

1350

1850

Wavelength (nm)

HyspIRI will enable those synergies to be fully

explored and utilized

2350

The HyspIRI
Airborne Campaign
Produce HyspIRIlike Products
Sample large areas
Sample strong
environmental gradients
Sample multiple seasons
Sample multiple years

Assemble a science
team to utilize data
Answer diverse questions
from coastal marine to
interior to urban
Strong emphasis on
synergies of VSWIR and TIR

HyspIRI Flight Campaign

Produce HyspIRIlike Products
Sample large areas
Sample large
environmental gradients
Sample multiple seasons
Sample multiple years

Assemble a science team

Answer diverse questions
from coastal marine to
interior to urban
Construct a diverse team

Case Study: The relationship between species

composition, fractional cover and Land Surface
Temperature in a Mediterranean ecosystem
Dar A. Roberts, Phil Dennison, Keely Roth,
Glynn Hulley

Funding:
NASA HyspIRI Preparatory Program
Naval Post Graduate School

Research Objectives
Explore synergies between the VSWIR and TIR
using AVIRIS-MASTER pairs in an ecological
context
Research Questions
What is the relationship between species composition
and land surface temperature (LST)?
What is the relationship between fractional cover and
LST?

Study Site

Processing Scheme

Training/Validation Spectra

Sampled 23 land-cover/species from 307 polygons

Random training/validation (10 max, or 50%)
Three pulls, first pull acceptable (Roth et al., 2012)

Endmember Selection: Forced Iterative

Endmember Selection

Injects EMC optimal ems

into IES for weak or
missing classes
Continues until Kappa
does not improve
Final library can be cut
off at any number
101

0.8
0.7
0.6
Kappa Coefficient

Traditional:
EAR/MASA/COB
Automated: Iterative
Endmember Selection
(IES) (Schaaf/Roth)
Forced IES

0.5
0.4
0.3
0.2

Cut off

0.1

Injection

0
0

100

200
300
Number of Endmembers

Roth et al., 2012

400

Endmember Spectra
2em classification
101 endmembers

67 GV
11 GV-NPV (can be either)
3 Rock
7 Soil
13 urban

GV
No mixtures, reduced
redundancy
25 endmembers
1 to 3 per species

NPV
7 endmembers
3 Arcasale, 2 brni, 2 magf

Endmember Spectra
Soil/Rock
8 endmembers
2 rocks
6 soils

Impervious/urban

10 endmembers
2 roads
3 tile roofs
4 composite roofs
1 painted roof

ClassificationAccuracy

Polygon Accuracy based on most abundant: 84.7%

Pixel-based: Citrus (CISP): 74.35%, Avocado (PEAM): 71.37%

Unclassified pixels largest error source

Poor: Grey Pine (PISA), Bay Laurel (UMCA), Coyote Brush (BAPI overmapped)

Species Mapping

Blue Oak (QUDU) and Grass

ARCA/SALE Stand

Plant Species/Land-cover on the Front Range

23 classes
84.7% polygon-based accuracy
>70% accuracy
ARCA-SALE, CEME, CISP, ERFA, IRGR, MAGF, MARSH, PEAM, SOIL,
AGRES, URBAN

Fractional Cover vs Temperature

Class and Composition: Urban/Natural

Dominant Classes

Grasslands (MAGF)
Black Mustard (BRNI)
Eucalyptus (EUSP)
Avocado (PEAM)/Citrus (CISP)
Coyote Brush (BAPI)
Urban/Soil
Minor Marsh
Black (> 2 ems)

Composition and temperature

Forests: High GV, low LST

Orchards Int. GV, moderate LST
Senesced grass/soil/urban

Low GV, high LST

Temperature Species Relationships

Species composition strongly impacts

temperature, significant clustering
Trees (circles)

Evergreen shrubs (diamonds)

Warmer, high to moderate GV

Deciduous shrubs (triangles)

Coolest, high to moderate GV

Warm, moderate to low GV

Forbs/grasses (squares)

High to low GV, warm to hot

These clusters should move over

a season, with movement varying
by species

Research Data Sets for SARP

Joint AVIRIS/MASTER analysis
2011, 2016
HyspIRI AC analysis
2013, 2014, 2015 (three seasons)
Combined AVIRIS/MASTER Pairs
AVIRIS-NG analysis
June and September 2014
IDEAS Meteorological Data
2007 to present
Field instrumentation (ASD, FLIR)

Questions?