Download the entire proceedings as an Adobe PDF - Eastern Snow ...
Download the entire proceedings as an Adobe PDF - Eastern Snow ...
Download the entire proceedings as an Adobe PDF - Eastern Snow ...
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
105<br />
63 rd EASTERN SNOW CONFERENCE<br />
Newark, Delaware USA 2006<br />
Time Series Analysis <strong>an</strong>d Algorithm Development<br />
for Estimating SWE in Great Lakes Area Using Microwave Data<br />
ABSTRACT<br />
AMIR E AZAR, 1 HOSNI GHEDIRA 1 , PETER ROMANOV 2 ,<br />
SHAYESTEH MAHANI 1 , AND REZA KHANBILVARDI 1<br />
The goal of this study is to develop <strong>an</strong> algorithm to estimate <strong>Snow</strong> Water Equivalent (SWE) in<br />
Great Lakes area b<strong>as</strong>ed on a three-year of SSM/I dat<strong>as</strong>et along with corresponding ground truth<br />
data. The study area is located between latitudes 41N <strong>an</strong>d 49N <strong>an</strong>d longitudes 87W <strong>an</strong>d 98W. The<br />
area is covered by 28*35 SSM/I EASE-Grid pixels with spatial resolution of 25km. Nineteen test<br />
sites were selected b<strong>as</strong>ed on se<strong>as</strong>onal average snow depth, l<strong>an</strong>d cover type. Each of <strong>the</strong> sites<br />
covers <strong>an</strong> area of 25km*25km with minimum of one snow reporting station inside. Two types of<br />
ground truth data were used: 1) point-b<strong>as</strong>ed snow depth observations from NCDC; 2) grid b<strong>as</strong>ed<br />
SNODAS-SWE dat<strong>as</strong>et, produced by NOHRSC. To account for l<strong>an</strong>d cover variation in a<br />
qu<strong>an</strong>titative way a NDVI w<strong>as</strong> used. To do <strong>the</strong> <strong>an</strong>alysis, three scattering signatures of GTVN<br />
(19V–37V), GTH (19H–37H), <strong>an</strong>d SSI (22V–85V) were derived. The <strong>an</strong>alysis shows that at lower<br />
latitudes of <strong>the</strong> study area <strong>the</strong>re is no correlation between GTH <strong>an</strong>d GTVN versus snow depth. On<br />
<strong>the</strong> o<strong>the</strong>r h<strong>an</strong>d SSI shows <strong>an</strong> average correlation of 75 percent with snow depth in lower latitudes<br />
which makes it suitable for shallow snow identification. In <strong>the</strong> model development a non-linear<br />
algorithm w<strong>as</strong> defined to estimate SWE using SSM/I signatures along with <strong>the</strong> NDVI values of <strong>the</strong><br />
pixels. The results show up to 60 percent correlation between <strong>the</strong> estimated SWE <strong>an</strong>d ground truth<br />
SWE. The results showed that <strong>the</strong> new algorithm improved <strong>the</strong> SWE estimation by more th<strong>an</strong> 20<br />
percent for specific test sites.<br />
Keywords: Microwave SSM/I, NDVI, SWE.<br />
INTRODUCTION<br />
Knowing <strong>the</strong> se<strong>as</strong>onal variation of snowcover <strong>an</strong>d snowpack properties is of critical import<strong>an</strong>ce<br />
for <strong>an</strong> effective m<strong>an</strong>agement of water resources. Satellites operating in <strong>the</strong> optical wavelength<br />
have monitored snowcover throughout <strong>the</strong> Nor<strong>the</strong>rn Hemisphere for more th<strong>an</strong> thirty years. These<br />
sensors c<strong>an</strong> detect snowcover during daylight <strong>an</strong>d cloud-free conditions. In contr<strong>as</strong>t to visible<br />
b<strong>an</strong>ds, remote me<strong>as</strong>urements operation in microwave region offers <strong>the</strong> potential of monitoring <strong>the</strong><br />
snowpack water equivalent <strong>an</strong>d wetness due to penetrating capability of <strong>the</strong> radiation at <strong>the</strong>se<br />
frequencies. Hallikainen et al. (1984) introduced <strong>an</strong> algorithm for estimating SWE using p<strong>as</strong>sive<br />
microwave Sc<strong>an</strong>ning Multi-ch<strong>an</strong>nel Microwave Radiometer (SMMR) data. The process involved<br />
<strong>the</strong> subtraction of a fall image from a winter image in vertical polarization of 18 <strong>an</strong>d 37 GHz<br />
frequencies. The difference, ΔT, w<strong>as</strong> used to define linear relationships between ΔT <strong>an</strong>d SWE.<br />
Aschbacher (1989) proposed <strong>an</strong> SPT algorithm for estimating snow depth <strong>an</strong>d snow water<br />
equivalent that w<strong>as</strong> b<strong>as</strong>ed on a combination of SSM/I ch<strong>an</strong>nels. Fur<strong>the</strong>r studies revealed that since<br />
1 NOAA-CREST, City University of NY, 137 th St & Convent Ave. New York, NY.<br />
2 NOAA World Wea<strong>the</strong>r Building, 5200 Auth Rd, Camp Springs, MD.