04/09/03 12.540 Lec 15 1
12.540 Principles of the Global
Positioning System
Lecture 15
Prof. Thomas Herring
Propagation Medium: Neutral atmosphere
? Summary
– Basic structure of the atmosphere: Here we exclude
the effects of electrons in the ionosphere (covered
next lecture).
– Refractivity of constituents in the atmosphere
– Separation of atmospheric delay into “hydrostatic”
and wet components.
– Elevation angle dependence:
? Azimuthally symmetric mapping functions
? Gradient formulations
– Effects of atmospheric delays on position estimates
04/09/03 12.540 Lec 15 2
1
04/09/03 12.540 Lec 15 3
Basic
atmospheric
structure
Troposphere is
where the
temperature stops
decreasing in the
atmosphere. (~10
km altitude)
04/09/03 12.540 Lec 15 4
Troposphere
? Lots of examples of web-based documents about the
atmosphere: See for example.
http://www-das.uwyo.edu/~geerts/cwx/notes/chap01/tropo.html
? Tropopause is where temperature stops decreasing.
Generally at pressure levels of about 300 mbar but
can be as low as 500 mbar.
?
only about 70% of delay.
? Generally by height of 50-100km all of atmospheric
delay accounted for.
? Troposphere is where weather systems occur and
aircraft fly on the tropopause.
Sometimes term “tropospheric delay” used but this is
2
Refractivity of air
? Air is made up of specific combination of gases, the
most important ones being oxygen and nitrogen.
? Each gas has its own refractive index that depends on
pressure and temperature.
? For the main air constituents, the mixing ratio of the
constituents is constant and so the refractivity of a
packet of air at a specific pressure and temperature
can be defined.
? The one exception to this is water vapor which has a
very variable mixing ratio.
? Water vapor refractivity also depends on
density/temperature due to dipole component.
04/09/03 12.540 Lec 15 5
04/09/03 12.540 Lec 15 6
Refractivity of air
? The refractivity of moist air is given by:
? For most constituents, refractivity depends on density
(ie., number of air molecules). Water vapor dipole
terms depends on temperature as well as density
N =k
1
P
d
T
Z
d
-1
Density of
dry air
+k
2
P
w
T
Z
d
-1
Density of
water vapor
+k
3
P
w
T
2
Z
d
-1
Dipole compoent
of water vapor r/T
k
1
= 77.60±0.05
k
2
= 70.4±2.2
k
3
=(3.730 ±0.012) ¥10
5
K
2
123 123 123
K/mbar
K/mbar
/mbar
3
04/09/03 12.540 Lec 15 7
Refractivity in terms of density
? We can write the refractivity in terms of density:
? Density r is the density of the air parcel including
water vapor. R is universal gas constant, M
d
and M
w
w
from ideal gas law) See Davis, J. L., T. A. Herring, and I.I.
Shapiro, Effects of atmospheric modeling errors on determinations of
baseline vectors from VLBI,
N =k
1
R
M
d
r+
k'
2
T
+
k
3
T
2
ê
?
á
?
ˉ
?P
w
Z
w
-1
k'
2
=k
2
-k
1
M
w
/M
d
=22.1±2.2 K/mbar
are molecular weights. Z is compressibility (deviation
J. Geophys. Res., 96, 643–650, 1991.
04/09/03 12.540 Lec 15 8
Integration of Refractivity
? To model the atmospheric delay, we express the
atmospheric delay as:
? Where the atm
path; vac is straight vacuum path, z is height for
station height Z and m(e) is a mapping function.
(Extended later for non-azimuthally symmetric
atmosphere)
?
D= n(s)ds- ds
vac
ú
atm
ú
am(e) (n(z)-1)dz=
Z
?
ú
m(e) N(z)¥10
-6
dz
Z
?
ú
path is along the curved propagation
The final integral is referred to as the ”zenith delay”
4
Zenith delay
? The zenith delay is determined by the integration of
refractivity vertically.
? The atmospheric is very close to hydrostatic
equilibrium meaning that surface pressure is given by
the vertical integration of density. Since the first term
in refractivity depends only on density, its vertical
integration will depend only on surface pressure. This
integral is called the “zenith hydrostatic delay (ZHD)”.
(Often referred to as “dry delay” but this is incorrect
because has water vapor contribution).
04/09/03 12.540 Lec 15 9
Zenith hydrostatic delay
? The Zenith hydrostatic delay is given by:
ZHD=10
-6
k
1
M
R
d
g
m
-1
P a0.00228 m/mbar
s
? Where g
m
is mean value of gravity in column of air
(Davis et al. 1991)
g
m
=9.8062(1-0.00265cos(2f)-3.1x10
-7
(0.9Z+7300)) ms
-2
?P
s
is total surface pressure (again water vapor
contribution included)
? Since P
s
is 1013 mbar at mean sea level; typical ZHD
=2.3 meters
04/09/03 12.540 Lec 15 10
5
Zenith wet delay
? The water vapor delay (second term in refractivity) is
not so easily integrated because of distribution of
water vapor with height.
? Surface measurements of water vapor pressure
(deduced from relative humidity) are not very effective
because it can be dry at surface and moist above and
visa versa.
? Only effective method is to sense the whole column of
water vapor. Can be done with water vapor
radiometer (WVR) which infers water vapor delay from
thermal emission from water vapor molecules and
some laser profiling methods (LIDAR). Both methods
are very expensive (200K$ per site)
04/09/03 12.540 Lec 15 11
Zenith wet delay
? In meteorology, the term “Precipitable water”
(PW) is used. This is the integral of water
vapor density with height and equals the depth
of water if all the water vapor precipitated as
rain (amount measured on rain gauge).
? If the mean temperature of atmosphere is
known, PW can be related to Zenith Wet
Delay (ZWD) (See next page)
04/09/03 12.540 Lec 15 12
6
04/09/03 12.540 Lec 15 13
PW and ZWD
? Relationship:
? The factor for conversion is ~6.7 mm delay/mm PW
? This relationship is the basis of ground based GPS
meteorology where GPS data are used to determine
water vapor content of atmosphere.
? ZWD is usually between 0-30cm.
ZWD=10
-6
R
M
w
(k'
2
+k
3
/T
m
)PW
T
m
=
P
w
/Tdz
ú
P
w
/T
2
dz
ú
Mapping functions
? Zenith delays discussed so far; how to relate to
measurements not at zenith
? Problem has been studied since 1970’s.
? In simplest form, for a plain atmosphere, elevation
angle dependence would behave as 1/sin(elev). (At
the horizon, elev=0 and this form goes to infinity.
? For a spherically symmetric atmosphere, the
1/sin(elev) term is “tempered” by curvature effects.
? Most complete form is “continued fraction
representation” (Davis et al., 1991).
04/09/03 12.540 Lec 15 14
7
04/09/03 12.540 Lec 15 15
Continued fraction mapping function
elevation angles. Form is:
m(e)=
1
sin(e)+
a
sin(e)+
b
sin(e)+
c
sin(e)+L
? Basic form of mapping function was deduced
by Marini (1972) and matches the behavior of
the atmosphere at near-zenith and low
04/09/03 12.540 Lec 15 16
Truncated version
m(e)=
1+
a
1+
b
1+c
e)+
a
e)+
b
e)+c
when e=90; m(e)=1
Davis et al. 1991 solved problem by using tan for second
sin argument.
? When the mapping function is truncated to the
finite number of terms then the form is:
sin(
sin(
sin(
8
04/09/03 12.540 Lec 15 17
Mapping functions
–
height and time of year dependent
– MTT (MIT Temperature) model uses temperature
as proxy for atmospheric conditions.
–
04/09/03 12.540 Lec 15 18
Coefficients in mapping function
– a=1.232e-3, b=3.16e-3; c=71.2e-3
– a =0.583e-3; b=1.402e-3; c=45.85e-3
~36.
? Basic problem with forming a mapping
function is determining the coefficient a,b, c
etc for specific weather conditions.
? There are different parameterizations:
Niell mapping function uses a, b,c that are latitude,
Recent Niell work uses height of 500mbar surface
(needs to be determined from assimilation models).
? The typical values for the coefficients are
? Hydrostatic:
? Wet delay
? Since coefficients are smaller for wet delay,
this mapping function increases more rapidly
at low elevation angles.
? At 0 degrees, hydrostatic mapping function is
Total delay ~82 meter
9
Mapping functions
? The basic form of the continued fraction fit
raytracing through radiosonde temperature,
pressure and humidity profiles to a few
millimeters at 3deg elevation angle.
? Basic problem is parameterizing a,b,c in terms
of observable meteorological parameters.
? Comparison on mapping functions:
http://gauss.gge.unb.ca/papers.pdf/igs97tropo.pdf
04/09/03 12.540 Lec 15 19
Gradients
? In recent years; more emphasis put on deviation of
atmospheric delays from azimuthal symmetry. See:
Chen, G. and T. A. Herring, Effects of atmospheric azimuthal
asymmetry of the analysis of space geodetic data, J. Geophys.
Res., 102, 20,489–20,502, 1997.
? These effects are much smaller (usually <30mm) but
do effect modern GPS/VLBI measurements.
? There is a mean NS gradient that is latitude dependent
and probably due to equator to pole temperature
gradient.
? Parameterized as cos(azimuth) and sin(azimuth) terms
with a “tilted” atmosphere model (1/(sin(e)+0.032)
04/09/03 12.540 Lec 15 20
10
04/09/03 12.540 Lec 15 21
Effects of atmospheric delays
? Effects of the atmospheric delay can be approximately
assessed using a simple WLS model of the form:
? Simulated data y (e.g. error in mapping function) can
be used to see effects on clock estimate (D
(Dh), and atmospheric delay (Datm)
? If m(e) is removed from partials, then effects in zenith
delay error on height can be estimated.
y=[1 sin(e) m(e)]
Dclk
DH
DZHD
è
?
í
í
í
?
?
˙
˙
˙
clk), Height
Effects of atmospheric delay
? If atmospheric zenith delay not estimated, then when
data is used to 10 degree elevation angle, error in
height is ~2.5 times zenith atmospheric delay error
(see
Herring, T. A., Precision of vertical position estimates from
very–long–baseline interferometry, J. Geophys. Res., 91,
9177–9182, 1986.nobreakspace
? A simple matlab program can reproduce these results
? Herring Kalman filter paper also discusses effects of
process noise value in height estimate uncertainty.
04/09/03 12.540 Lec 15 22
11
04/09/03 12.540 Lec 15 23
Parameterization of atmospheric delay
? Given the sensitivity of GPS position estimates to
atmospheric delay, and that external calibration of the
delay is only good to a few centimeters; atmospheric
precision GPS analyses.
? Parameterization is either Kalman filter or coefficients
of piece-wise linear functions (GAMIT)
? See: http://www-gpsg.mit.edu/~katyq/plots_gpsdelay/
for comparison of GPS estimates with calculations for
global forecast models (NCEP)
04/09/03 12.540 Lec 15 24
Example using NCEP analysis field
Blue is GPS estimates of delay, red is NCEP calculation
zenith delays and often gradients are estimated high-
12
04/09/03 12.540 Lec 15 25
Summary
error sources in GPS
–
mapping functions
–
unsolved problem even with gradient estimates.
– Estimated delays can be used for weather
forecasting if latency <2 hrs.
? Atmospheric delays are one of the limiting
? Delays are nearly always estimated:
At low elevation angles there can be problems with
Spatial inhomogenity of atmospheric delay still
? Class web page has links to some sites that
deal with GPS atmospheric delay estimates
13