This document proposes a proof-of-concept method to estimate snow water equivalent (SWE) using GPS multipath signals. SWE represents the amount of water available for runoff and is important for hydrological studies. The author developed a theoretical model relating GPS signal properties to snow layer thickness and density. Field measurements in Montana were collected and a nonlinear algorithm was able to estimate SWE based on the model. Future work is suggested to improve the model and test the technique further.
1. Estimating Snow Water Equivalent for a
Snow-Covered Ground Reflector using GPS
Multipath Signals
Dr. Mark Jacobson
Mathematics Department
1500 University Drive, Billings, MT 59101
mjacobson@msubillings.edu
2. I. Introduction
Proof-of-concept method for estimating snow
water equivalent (SWE) at the GPS frequency of
1.5 GHz
SWE is the single most important parameter for
hydrological studies because it represents the
amount of water potentially available for runoff
SWE estimations are used for the management of
water supply and flood control systems
The U.S. D.O.A. operates and manages the
Snowpack Telemetry (SNOTEL) system, 730 sites
GPS signals could provide a new and economical
technique for estimating SWE
4. Relative received power at the antenna is
rh rv exp(i ) 2
4 (h t ) sin
P 1 where,
2 0
h = height of antenna above conducting surface, m
elevation angle, degrees
i 1
c 2.997925 108 m/s, the speed of light in a vacuum
f 1.57542 109 Hz
0 c / f 0.1902937 m, free-space wavelength
t = snow layer thickness, m
snow s' i s" s' 1 2 d
d relative density of dry snow, g cm−3
5. For a single-layer of snow above a conducting
surface with horizontal polarization, x = h and
vertical polarization, x = v
iZ x tan 1 sin
rx Zh
iZ x tan 1 cos 2
2 Zv
cos 2
t cos 2
sin
0
6. III. Measurements and Computation
March 31, 2007 T = -1.7 C, t = 7.6 cm,
h = 45.1 cm, no snow density data
III. Measurements and Computation
March 31, 2007
T = -1.7 C, t = 7.6 cm, h = 45.1 cm
8. In order to utilize a Quasi-Newton Algorithm (QNA)
efficiently in finding estimates of snow depth and density, we
approximate the relative complex permittivity value of dry
snow as
snow i 1 2 d
'
s
"
s
'
s
"
s
'
s
������=������
1 2
SE ������, ������snow = ������������������ ������ ������������ − ������ ������������ , ������, ������snow n = 8,168
������ − 2
������=1
From 45 input pairs, the smallest SE produced the following:
t = 6.8 cm d 0.30 s' 1.60
9.
10.
11.
12.
13. IV. Conclusions
Theoretical results and GPS measurements
are in good agreement using a nonlinear
QNA
Estimating SWE may be possible using a
nonlinear least squares technique
“Inferring Snow Water Equivalent for a
Snow-Covered Ground Reflector Using GPS
Multipath Signals”, Remote Sensing, Vol. 2,
2426-2441, October 2010
14. V. Future Work
Try a QNA for a snow layer above frozen
soil
Collect more in situ measurements of snow
depth, snow density, and frozen soil
permittivity
Try other nonlinear least-squares
algorithms: Levenbeg-Marquardt and
Conjugate Gradient
Incorporate 2 or more snow layers in the
theoretical model
15. V. Future Work (continued)
Incorporate the antenna pattern in the
theoretical model
Incorporate surface roughness of snow and
frozen soil in the theoretical model
Use a horizontally-mounted (zenith-pointing)
GPS antenna
Investigate this technique for GPS antennas
housed on an aircraft or satellite
16. Acknowledgment
Montana State University Billings
– Dr. Tasneem Khaleel, Dean CAS
– Dr. Maggie McBride, Math Dept.
– RACE Grant
Ron and Jeanne Jacobson, my parents
Wade Dotson, Trimble Navigation
C. McFarland and T. McFarland, land
owners
Anonymous reviewers of paper