Update on MISR applications to Shrub Abundance Mapping in Desert Grasslands

Mark Chopping, Lihong Su, Albert Rango, Debra P.C. Peters, John V. Martonchik, and Andrea Laliberte

3 March 2006

Structure: A Strength of Multiangle Remote Sensing

• Multi-angle sampling provides sensitivity to vegetation

canopy structure, even for low density canopies such as

those encountered in the desert grasslands of the

southwest US

• Canopy structure can be inferred through various approaches, including geometric-optical modeling

• A major prerequisite for GO modeling is the ability to estimate the relative contributions of the upper canopy (i.e., shrubs) and the soil-understory component (exposed soil, grasses, sub-shrubs, and forbs)

GOAL: to obtain areally-weighted proportions of background (soil, understory) and upper canopy components. Geometric-Optical (GO) models provide one method of understanding the sensitivities of passive multi-angle sensing to canopy structure parameters of practical significance to ecosystem modeling:

•woody shrub cover•mean shrub crown shape and radius•shrub number density•canopy height•soil/understory brightness and anisotropy

Geometric-Optical Modeling of Vegetation

sunlit crown shadowed crown

sunlit ground shadowed ground

y = 0.9985x + 0.0003

R2 = 1.0000

y = 0.7252x + 0.0306

R2 = 0.7352

0.05

0.10

0.15

0.20

0.05 0.10 0.15 0.20

Measured Fractional Shrub Cover

Mod

ele

d F

ract

ional S

hru

b C

over

Optimal:

Estimated:

The impact of using estimated (∆) over optimal () background BRDFs for 19 cases covering a wide range of shrub cover/size and understory configurations. The background contribution is the most important source of error in modeling canopy-level reflectance at the landscape scale

MISR observationsModeled (simple geometric model)…… optimal background---*--- estimated background- - - without shrubs (G.kG)

A good match is obtained between modeled and observed BRFs from MISR 275 m red band data.GO model: Simple Geometric Model (SGM), using the Walthall model to represent the soil-understory contribution. Shrub crown radii and number density were measured using IKONOS 1 m panchromatic imagery.

Simulations of MISR BRFs

in the red band and at nine viewing angles

0.15

0.20

0.25

0.30

0.35

-80 -60 -40 -20 0 20 40 60 80

0.15

0.20

0.25

0.30

0.35

-80 -60 -40 -20 0 20 40 60 80

0.15

0.20

0.25

0.30

0.35

-80 -60 -40 -20 0 20 40 60 80

0.15

0.20

0.25

0.30

0.35

-80 -60 -40 -20 0 20 40 60 80

View Zenith Angle (˚) View Zenith Angle (˚)

Red Bidirectional Reflectance Factor

(a) (b)

(d)(c)

shrub cover = 0.12(dense understory)

shrub cover = 0.07(sparse understory)

shrub cover = 0.10(sparse understory)

shrub cover = 0.19(sparse understory)

Shrub cover was measured using IKONOS panchromatic imagery for a 5 km2 area in a Chihuahuan Desert grassland with varying degrees of shrub encroachment.

80% of retrieved shrub cover values using MISR data were within 0.05 of the measured value

This is the first time that fractional woody shrub cover has been mapped using medium resolution Earth Observation data

Retrieval of woody shrub cover at landscape scales

250 m

0

10

20

30

40

50

60

70

80

0.00 0.02 0.04 0.06 0.08 0.10

Absolute Difference

Frequency

.0%

20.0%

40.0%

60.0%

80.0%

100.0%

120.0%

Frequency

Cumulative %

Mean (0.026) N = 441

There is broad agreement in the spatial distributions of the retrieved and reference data for fractional woody shrub cover in this Chihuahuan Desert grassland. 80% of retrieved shrub cover values retrieved using MISR data were within 0.05 of the measured value. To our knowledge, this is the first time that woody shrub cover has been mapped over sparse and dense backgrounds using medium resolution Earth Observation. Although there is some divergence between retrieved and measured distributions, there is also error in the reference data; these were obtained via thresholding IKONOS 1 m panchromatic imagery.

Retrieval of woody shrub cover at landscape scales

Measured Retrieved

0.04 0.28 0.07 0.21

250 m2 raster cells

Fractional Woody Shrub CoverWoody shrub cover at landscape scales from MISR

0.00

LTER Vegetation Map (1998)

M.Well, etc indicate Net primary produvity sampling sites: M=mesquite T=tobosa P=playa C=cresosotebush G=grama grass. Boundaries of vegetation mapping units.

0.46

N.B. no corrections were applied for topographic effects on acquisition angles (affecting the E. part of the scene)

mesquite grasses:

snakeweed

creosotebush

tarbushyucca elata

other shrubs

upland

playa