|
 Contact: Thomas Albright, email: albright@icess.ucsb.edu, tel: (805)893-8116, fax: (805)893-2578
1 Department of Geography,University ofCalifornia, Santa Barbara, CA 93106,USA
2 Institute for ComputationalEarth System Science, University of California,Santa Barbara,CA,93106, USA
3 Schoolof Environmental Science and Management,University of California, Santa Barbara, CA 93106,USA
|
|
|
KeyWords:snowcoveredarea,classification,synthetic
aperture radar, SIR-C,accuracy
ABSTRACT
Timely and accuratemaps of snow covered area(SCA) are
important toresource managers,planners, and scientists
for applicationsrangingfrom avalanchehazard assessment
to global climate studies. Opticalsensors such as Landsat
Thematic Mapper (TM) have already demonstratedtheir
effectiveness at mapping SCA.Recently,much work has
focusedonthe use ofsyntheticaperture radar(SAR)to
accomplish this task,due to its high resolution,sun inde-
pendent, all-weather capability. Though initial results are
encouraging, an extensiveassessmentofthe accuracy of
these systemsunder a variety of sensorand target condi-
tionsneedstobeperformed.Thisstudyexaminesthe
accuracy of a Spaceborne Imaging Radar - C(SIR-C) algo-
rithmformapping SCA.
We usedawellverifiedLandsatTMfractionalSCA
image to validate SIR-C SCA images of MammothMoun-
tain,CA,USA.We produced images showing the spatial
distribution and magnitude of the errors. Wealso analyzed
what surface conditions correlate with large errors in the
SCAestimation.TheSIR-Calgorithmisaccurateunder
some conditions but needs improvement in other areas. It
does well in pure snow and snow free areas, but overall, it
underestimatessnowrelativetotheTMalgorithm.The
major source for this underestimation in this study is SIR-
C's difficultydetecting snow in moderately vegetated ar-
eas.
INTRODUCTION
The measurement of snow covered area (SCA) in season-
allysnow-coveredalpineregionsisveryimportantfor
investigationsofclimateandhydrology.Snow'shigh
albedoandlowthermalconductivitycanhavegreatef-
fectsonbothlocalandglobalclimate.With respectto
hydrology,snow acts as a waterstorage reservoir,releas-
ing the waterit has accumulated from the winterduring
the spring melt.Many regions throughout the world rely
on this reservoir of water to meet their fresh water resource
needs.
Remote sensing offers excellent opportunities for meas-
uring SCA due to its ability to provide timely information
overlarge areas. Optical sensors have demonstrated their
utilityintheSCAmappingarena (eg.Rango and Itten,
1976, Dozier, 1989, Rosenthal and Dozier, 1996). However,
optical systems are subject to two major disadvantages.The
first of which is that they are dependent on the illumina-
tionofthesunwhich isvariableand,forhigh latitudes
may not exist at some times of the year. In addition to this,
opticalwavelengths are unable topenetratecloud cover
which can be pervasive in some regions.
|
Radar remote sensing is not subject to these disadvantages.
Radar,being an active sensor,provides its own illumina-
tion and is therefore able to operate entirely independently
oftheSun.Furthermore,becauseradaroperatesinthe
microwave portion ofthe electromagnetic spectrum,it is
able to penetrateclouds andallbut the most severeweather
events.
These advantageshave led some researchers to use ra-
dar to map SCA in alpine regions (eg. Rott and Davis, 1993,
Shiand Dozier,inpress).While promisingresultshave
been obtained, an extensive assessment of the accuracy of
these systemsunder a variety of sensorand target condi-
tions needs to be performed This study examines the ac-
curacy ofaSpaceborneImagingRadar-C(SIR-C)algo-
rithm formapping SCA.
BACKGROUND
Imaging radar pulses radiation at a specified frequency in
the microwave portion of the electromagnetic spectrum to
theimagedareaand measuresthecharacteristicsofits
return. Thus, if groundtargetsare to be discriminated, there
mustbedistinctradarreturns,or"backscatter"forthe
targetsrequiringdiscrimination.Inthisregionofthe
spectrum, ice, a major component of snow, is almost trans-
parent and radar penetration depth can reach tens of me-
ters for dry snow.In this case,theground becomesthemajor
scattering source andsnow is difficult to detect. However,
the presenceof liquid water in snow has several effects on
the bacscatterproperties that allow us to detect and map
wet snow. With a small amount of liquid water present in
the snow pack the scattering source shifts from the snow-
ground interface to a mixture ofsnow volumescattering
and surface scattering at theair-snow interface. Additional
water will increase thedominance of the surface scattering
and the air-snow interface. These phenomena, along with
knowledgeoftheroughnessofthesnowsurface,can
distinguishsnowfromothergroundcovertypesinthe
microwaveregion ofthespectrum andallowustomap
snow using radar (Shi and Dozier, 1996).
DATA AND METHODS
An image of the Mammoth Mountain,California area was
acquired fromthe SIR-C/X-SAR sensorflown aboard the
NASA SpaceShuttleonApril 11, 1994. The image,centered
at37.6degreesnor thand119.0degreeswest,is
approximately 11.5 km wide and50 km long and is ori-
ented NW-SEwhich coincides with the orientation of the
Sierra Nevada,the local mountain range. The orbit from
whichtheimage wasacquired wasdescending and the
radar was right looking, thus the sensor was imaging from
the northeast.
The image was radiometrically calibrated, correctedfor
terrain usinga 30m digital elevation model, and processed
|
|
1
2
3
4
