Dr. Jeff Dozier
Computer Systems Laboratory
Center for Remote Sensing and
Environmental Optics
University of California
Santa Barbara, CA 93106
Co-Investigator:
J. Shi, University of Calif., Santa Barbara
SIR-C
Investigations of Snow Properties in Alpine Terrain
OBJECTIVES
Estimate snow-covered area and distribution of snow water equivalence over alpine
drainage basins. Surface material may be trees that are taller than the snow depth,
brush, or grass that will be covered by the snow, soil, scree, talus, or bedrock,
or glacier ice.
Estimate spectral albedo of snow cover.
Model spatial distribution of snow surface energy exchange, melt rate, and snow metamorphism
intensively during two-to-four-week periods around
SIR-C/X-SAR
flights, and less
intensively during the rest of the snow season.
PROGRESS
Our recent studies, using
SIR-C/X-SAR
and
AIRSAR
data, have shown a significant improvement
in understanding and modeling backscattering and polarization properties as a function
of snow parameters. Maps of snow-covered areas derived from
SIR-C/X-SAR
and
AIRSAR
now compare well with those derived from visible images, which require clear
weather and daylight. A better than 80% accuracy for both dry and wet snow covers
can be achieved when compared with
TM
classification results. We have developed
an accurate algorithm to retrieve snow-wetness, which indicates where and at what rate snow
is melting. The tested results using C-band
SIR-C
and
JPLAIRSAR
data indicated
that an accuracy of 2.5% can be achieved at 95% confidence interval. We have developed
an algorithm to estimate snow density and depth. The initial test showed very promising
results. Thus, accurate information about the spatial and temporal distributions
of snow water equivalency and melting status of snow cover can be provided by
SIR-C/X-SAR
for hydrological and climatic investigations and operations.
Field Measurements
During the first and second
SIR-C/X-SAR
missions in April and October 1994, we carried
out near-simultaneous field campaigns at three test sites: Mammoth Mountain in the
Sierra Nevada, California, the Ürümqi River basin in the Tien Shan, China, and the
Echaurren basin in the Andes, Chile.
At the Mammoth site, three corner reflectors were used for calibration. We selected
three test sites near Mammoth Mountain based on
SIR-C/X-SAR
passes. For each data-take
we measured two snow-pits for the vertical profile of snow properties at each site: temperature, snow density, grain size and size distribution, free liquid water content,
and stratification. In addition, we measured snow properties along two transects
that cross each site for information about spatial distribution. Those measurements
included snow depth, density, wetness, and surface roughness.
At the Tien Shan site, five corner reflectors were used for calibration. Snow properties
measurements included snow temperature, depth, and density. At the Echaurren site,
we measured snow temperature, depth, density, particle size, and surface roughness.
Microwave Modeling and Backscattering Response to Snow Parameters
The main problem in estimating snow properties from remote sensing data is understanding
the links between the electromagnetic interactions in different parts of the spectrum
and the physical properties of the snow. Modeling microwave backscattering from
snowpacks requires knowledge of snow characteristics and their dynamics to select
an appropriate model. Both theory and field data show that microwave backscattering
coefficients are sensitive to parameters describing snow microstructure.
Stereological methods, applied to images of sections cut from undisturbed snow, are
used to obtain accurate, unbiased estimates of snow microstructure parameters for
discrete scatterer modeling. Our recent studies have found that the ice particle
size distribution in seasonal snow can be characterized as a log-normal distribution function.
The required parameters (the geometric mean diameter and standard deviation of particle
diameters) for fully describing the particle size variation and distribution can be directly measured by stereological variables. The results show that in addition
to snow density and ice particle size, the particle size variation affects extinction
properties of dry snow. The optically equivalent ice particle size for Rayleigh
scattering in a snowpack with grain size variations can be determined from the stereological
data.
We have developed a polarimetric model that includes both surface and volume scattering
as well as the interaction terms between surface and volume. Radiative transfer
models for dense media and the random media were used for the volume scattering.
The surface scattering models
(IEM,
SPM,
POM,
and
GOM)
were introduced to the radiative
transfer equations in order to evaluate the importance of the interactions between
the surface and volume scattering signals.
Through
SAR
data analyses and model simulations, we found that backscattering measurements
from dry snow are affected by three sets of parameters: (1) sensor parameters, (2)
snowpack parameters, and (3) ground parameters. The relationships between backscattering signals and snow water equivalence can be either positive or negative, depending
on the snow physical parameters, ground surface parameters, and incidence angle.
In addition to snow density and ice particle size, size variation, snowpack stratification, and underlying ground conditions affect the interpretation of the observed
backscattering signals. When the scattering signal from the snowpack is greater
than the signal from the ground (attenuated by the overlying snow), a positive correlation
is expected. Otherwise, the correlation is negative.
We also found that the relationship between the co-polarization signals at C-band
and snow wetness is controlled by the scattering mechanisms. In addition to the
snow properties, the surface roughness and local incidence angle play an important
role in the relationships between the co-polarization signals and snow wetness. This relationship
can be either positive or negative, depending on the snow properties, surface roughness,
and incidence angle.
Snow Mapping
Previously, snow in alpine regions has been mapped with conventional
SAR
imagery by
comparing a geocoded
SAR
image with a simulated image. This method requires a compatible
digital elevation model (DEM) for generation of the simulated image, and that the
DEM
and
SAR
data be coregistered. The capability of a single-polarization
SAR
in mapping
snow-cover is limited to wet snow condition. Single polarization SARs operating
at 5.3 GHz (C-band) can map wet snow and ice-free surfaces, but they poorly separate
glacier ice from snow and rock.
Through evaluation of the characteristics of the backscattering and polarization response
of snow, we found that mapping snow in remote alpine regions by using intensity measurements
requires topographic information to obtain correct radiometric measurements and to reduce angular dependence for discrimination. Because the measurements of
polarization properties and of the ratios between different frequencies provided
by
SIR-C/X-SAR
do not require topographic correction, an appropriable selection of
these measurements through evaluating the incidence angle dependence and separability between
different targets within the scene makes it possible to map snow covers without requiring
topographic information in remote alpine regions. Using this technique, we compared the coregistered
TM
classification results with
SIR-C/X-SAR
classification results.
Accuracies better than 80% for wet snow-covered from pixel-by-pixel comparison can
be obtained for binary classification. The major difference between
TM
and
SIR-C/X-SAR
binary classification results are due to the mixed pixel problem. In addition,
we have found that
SIR-C/X-SAR
can map dry snow cover when multifrequency and multipolarization
data with
DEM
are used. This is in contrast with single-polarization SARs which are limited only to mapping wet snow cover. A similar accuracy can also be obtained.
Currently, we are evaluating the possibility of mapping dry snow cover without requiring
topographic information. Thus,
SIR-C/X-SAR
has comparable capability with the visible and near-infrared sensors to map both dry and wet snow covers.
Snow Wetness
From a verified backscattering model, we have developed an algorithm for retrieving
snow wetness using C-band polarimetric
SAR
imagery. Our algorithm is based on a first-order
scattering model, with consideration of both surface and volume scattering. At the regional scale, the inferred spatial distribution of snow wetness from two
SIR-C
data-takes showed clear characteristics. Higher snow wetness was estimated on the
south slope than on the north slope, and the elevation dependence of snow wetness
was inferred. This agrees well with our knowledge of the characteristics of spatial distribution
of free liquid water content in snow. The comparison of SAR-derived snow wetness
with ground measurements indicated that the absolute error at 95% confidence interval
was 2.5%. Both regional and point measurement indicate that the inversion algorithm
performs well using C-band
SIR-C/X-SAR
data and will prove useful for routine and
large-area snow wetness measurements if a multi-parameter spaceborne
SAR
becomes
available.
Snow Water Equivalency
Based on the numerical simulations of our polarimetric model for dry snow covered
conditions, we have developed a physical-based algorithm to estimate snow water equivalency
and the snowpack equivalent particle size. This algorithm is based on the first-order backscattering model so that it should have less effect on the site-specific
problems. It uses L-band co-polarization measurements to estimate snow density and
the underlaying surface dielectric and roughness properties. Then, the backscattering
signals from the snow-ground interface at C- and X-band can be decomposed from the total
backscattering signals by estimation of the surface backscattering components from
the underlaying surface dielectric and roughness properties. Finally, the snow depths
and snowpack equivalent particle sizes can be estimated from C- and X-band measurements.
This algorithm does not require a background image (without snow cover) so that the
problems due to mis-coregistration and the surface roughness properties change can
be avoided. The current test results from two
SIR-C/X-SAR
data takes over our Mammoth
test site showed the accuracies of 62 kg/m3, 38.1 cm and 19.8 cm for snow density,
depth, and water equivalence, respectively. Currently, we are testing this algorithm
over all data takes for dry snow cover at the Mammoth site.
PUBLICATIONS
O'Neill, P. E., A. Hsu, and J. Shi, Soil Moisture Estimation Using Time-Series Radar
Measurements of Bare and Vegetated Fields in Washita '92, Proceedings
IGARSS
'95
, volume I, pp. 498-500,
IEEE
No. 95CH35770, 1995.
Shi, J. and J. Dozier, Inferring Snow Wetness Using
SIR-C
C-Band Polarimetric Synthetic
Aperture Radar, IEEE Transactions on Geoscience and Remote Sensing
, July, 1995.
Shi, J. and J. Dozier,
SIR-C/X-SAR
Investigations of Snow Properties in Alpine Regions,
Proceedings
IGARSS
'95
, volume II, pp. 1582-1584,
IEEE
No. 95CH35770, 1995.
Shi, J. and J. Dozier,
SIR-C/X-SAR
Mapping Snow in Alpine Regions, Proceedings
IGARSS
'95
, volume II, pp. 1508-1510,
IEEE
No. 95CH35770, 1995.
Shi, J., J. Wang, A. Hsu, P. O'Neill and E. T. Engman, Estimation of Soil Moisture
and Surface Roughness Parameters Using L-band
SAR
Measurements, Proceedings
IGARSS
'95
, volume I, pp. 507-509,
IEEE
No. 95CH35770, 1995.
Shi, J. and J. Dozier, Measurement of C-band
SAR
Response to Wet Snow: From Modeling
to Observation, Proceedings
PIERS
'95
, pp. 218-1041, 1995.
Shi, J. and J. Dozier, Estimation of Snow Water Equivalence Using
SIR-C/X-SAR
Image
Data, Proceedings
PIERS
'95
, p. 1041, 1995.
Shi, J., J. Wang, A. Hsu, P. O'Neill and E. T. Engman, Estimation of Soil Moisture
and Surface Roughness Parameters Using SIR-C's and AIRSAR's L-band Measurements,
to be submitted to the special issue of IGARSS '95
, in preparation.
Shi, J. and J. Dozier, Estimation of Snow Water Equivalence Using
SIR-C/X-SAR
Image
Data, Journal to be submitted to IEEE Transactions on Geoscience and Remote Sensing
, in preparation.
Shi, J. and J. Dozier,
SIR-C/X-SAR
Mapping Snow in Alpine Regions, to be submitted
to Remote Sensing of Environment,
in preparation.
Shi, J., J. Dozier and H. Rott, Snow and glacier mapping in alpine regions using polarimetric
SAR,
IEEE Transactions on Geoscience and Remote Sensing
, 32, pp. 152-158, 1994.
Shi, J., J. Dozier and H. Rott, Active microwave measurements of snow cover-Progress
in polarimetric
SAR,
Proceedings
IGARSS
'94
, pp. 1922-1924,
IEEE
No. 94CH3378-7, 1994.
Shi, J. and J. Dozier, Estimating snow particle size using
TM
band-4, Proceedings
IGARSS
'94
, pp. 1747-1749,
IEEE
No. 94CH3378-7, 1994.
Shi, J., P. O'Neill, A. Hsu, J. van Zyl and M. Seifried, Estimating soil moisture
and surface roughness parameters using L-band
SAR
measurements, Proceedings SPIE '94
, 1994.
Shi, J., J. Wang, P. O'Neill, A. Hsu and T. Engman, Application of
IEM
model on Estimations
soil moisture and surface roughness, Fifth Annual
JPL
Airborne Earth Science Workshop
, 1994.
Wang, J., P. E. O'Neill, E. T. Engman, R. Pardipuram, J. Shi and A. Hsu, Estimating
Surface Soil Moisture from
SIR-C
Measurements over the Little Washita Watershed,
Proceedings
IGARSS
'95
, volume III, pp. 1982-1984,
IEEE
No. 95CH35770, 1995.