Geology Reference
In-Depth Information
.40
.35
.26
OW
YI
FYI
MYI
OW
YI
FYI
MYI
.24
.22
.20
.18
.16
.14
.12
.10
.08
.06
.04
.02
.00
.30
.25
.20
.15
.10
.05
.00
Polarization ratio PR 38
Brightness temperature Tb 38h (K)
.26
.26
OW
YI
FYI
MYI
OW
YI
FYI
MYI
.24
.22
.20
.18
.16
.14
.12
.10
.08
.06
.04
.02
.00
.24
.22
.20
.18
.16
.14
.12
.10
.08
.06
.04
.02
.00
-32 -30-28 -26 -24 -22-20 -18
Backscatter coeff. σ 0 hh (dB)
-16-14 -12-10 -08 -06 -04 -02 00
-32 -30-28 -26 -24 -22-20 -18
Backscatter coeff. σ 0 w (dB)
-16-14 -12-10 -08 -06 -04 -02 00
Figure 10.14 Probability distributions of four radiometric parameters for three ice surfaces (YI, FYI, and MYI)
and OW. Brightness temperatures were obtained from AMSR‐E and backscatter from QuikSCAT. Noise floor of
backscatter is -36 dB [Shokr, unpublished data].
GR 23 V 19 V ≥ 0.03 (the frequencies used in the subscripts
are  approximate values of the AMSR‐E channels).
Figure  10.15 shows the justification for the threshold
selection. It is a scatterplot of OW and sea ice pixels
(regardless of the ice type) produced from using AMRS‐E
measurements in the Arctic region in different days in
September 2004. Data for OW and ice are presented in
different colors for different days. Overlap exists between
OW and ice samples. In selecting the thresholds, it is bet-
ter to misidentify an OW pixel as being a mixture of ice
and water than misidentify an ice (or a mixed pixel of ice
and OW) as being OW. In the first case let the technique
it process the pixel and may produce the correct answer
of 0% ice concentration. In the second case the technique
will exclude the pixel from the calculations and therefore
there will not be a second chance to correct it.
As mentioned in Chapter 7, the brightness temperature
and radar backscatter are triggered by different surface
properties. The former is mainly influenced by physical
temperature and the emissivity and the latter by surface
roughness and dielectric constant. The emissivity and
dielectric constant are related [equation (7.18) and (7.19)]
MAD
MAD max
CL
1
(10.55)
where MAD max is the maximum value of the MAD from
all the trials using the different CRVs. In the case of the
total concentration vector, the average of the Euclidean
distance between the final solution and the concentration
produced by each trial is used. MAD max is 100% for indi-
vidual ice type concentrations and 100%*√2 for the total
concentration. Higher CL values mean less variability
within the trials involving the CRVs (i.e., more robustness
of the ice concentration from the trials). In a perfect situ-
ation when each trial produces the same concentration
(i.e., MAD = 0), the confidence level reaches a maximum
value of 1.
Most of the ice concentration algorithms use a filter to
exclude OW pixels. They are usually based on gradient
ratios from microwave observations. ECICE uses a filter
similar to the one used in NT2 but with different thresh-
olds. The filter identifies the pixel as OW if the follow-
ing two inequalities are satisfied: GR 36 V 19 V > 0.05 and
Search WWH ::




Custom Search