International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B7. Istanbul 2004
Left Floodplain
Channel Right Floodplain
High
Water
Envisat
ASAR {
ERS-2
Figure 5 Dotty plots of parameter distributions for the different
PMs
Obviously, a large number of parameter sets perform almost
equally well. It is shown that the model performance mainly
depends on the channel roughness coefficients whereas the
model only shows limited sensitivity for the two floodplain
friction parameters. Depending on the chofce of the channel
roughness, good fits can be achieved in the whole range of the
sampled parameter values. In particular, a maximum likelihood
value of 1 could be obtained for many diffezent roughnesses in
respect with the measured high water marKs. The dotty plots
also show that a considerable range of pérformance measures
are produced inside the sampled parameter range. Clearly, some
of these parameter sets produce output that has to be considered
to be non-behavioural, i.e. the response deviates so far from the
observations that the model cannot be considered as an
adequate representation of the system. The parameter spaces of
the simulations meeting the behaviourability criteria (Table 2)
have only the channel roughness parameter constrained to its
lower range from the initial sampling limits. The total number
of behavioural simulations for each objective is given in Table
2. The dotty plots associated to each one of these objectives
show the consistencies of the results and the friction parameters
were more or less stationary between the two considered flood
stages. With only 286 behavioural parameter sets remaining, the
most important reduction of parameter sets is achieved with the
HW criteria. As these data are the most reliable, a relatively
high threshold PM value can be chosen. The fuzziness of the
radar data hampers the use of higher threshold PM values.
Behavioural simulations*
Acceptability criteria Total number
HW 286
Envisat 3926
ERS 4542
HW & Envisat 212
HW & ERS | 143
HW & Envisat & ERS 140
* Total number of simulations was 22000
Table 2. Behavioural simulations for indivual and combined
PMs
The effect of combining different PMs is shown in Table 2.
Only parameter sets meeting the behaviourability criterium of
each one of the multiple objectives are retained in the final
sample. Clearly, the number of behavioural simulations is
considerably reduced. More than half of the 286 selected
parameter sets are rejected based on the additional criteria and
finally only 140 among the initial 22000 runs are considered as
behavioural. Most notably, the additional ERS PM constrains
the model response most. This is not surprising as the ERS
picture was taken during the rising limb preceding by several
hours the peak discharge whereas the Envisat picture was taken
close to peak discharge. Hence, the flood data derived from
Envisat is somewhat redundant to the high water marks and the
resulting parameter constrain is not noteworthy.
Envisat 02/01/2003 22h18
TUAE
E .
Observed Sa Simulated Flood \
Eland Envicat 4 Ww =
i. HW
{À
yi
Envisat
02/01/2003
eo
E Ta
"o" A
4 HW à Sf WW HW à 1% HW &
3 ENVISAT ; ERS v ERS &
4 : A ENVISAT
S: t 3
1 X !
x A h
it p
» »
Y. y
Fry t
N & i
ie ^ i
in JA
; *
5 ne
AC Bn... 1
Nat IA
Figure 6. Updated 5% and 95% percentile inundation maps for
behavioural simulations of individual and combined
PMs (Envisat overpass time)
The progressive constraining of the model predictions by
incorporating the additional Envisat radar data is also shown on
Figure 6 with the distance between the boundary limits of the
5% and 95% quantile flood maps gradually narrowing. This
means that simulations conditioned using all the PMs show
smaller ranges of model behaviours than models conditioned
only using the ground data. Only locally, for instance at the
upstream end of the river reach, some major uncertainties
subsist. The uncertainty maps based on the final parameter set
constrained using all 4 PMs of Table 1 give inundation maps at
peak discharge that are very close to what was observed by
Envisat at the same time (Figure 7). Most importantly the range
of simulated flood boundaries generally brackets the observed
extents. It is not surprising that the uncertainty reduction by
incorporating additional distributed radar data becomes most
effective in those areas of the floodplain where surveyed point
data are missing. The resulting uncertainties tend to be higher at
an initial stage of the flood (during ERS-2 overpass). This may
be due to changing roughness values with increasing water
levels that lead to some doubtful parameter values being
included and some good sets being rejected when considering
high water marks only. Moreover, at a preliminary stage of the
flood, small changes of the channel roughness coefficient may
have a large impact on the simulated extent. This is due to the
fact that when the river channel is bankfull, small changes of
the water level tend to induce large changes of the flood extent.
356
Internati
By incor
constrair
The fing
equifinal
multi-res
simulatic
well tha:
main dif
related t«
leads to |
the respc
measurec
constrain
This doe
data sets
test site,
with rad:
occurred.
to existin
in time €
could als
roughnes
time inte
lead to th
be intere
different
paramete
simulatio
Figure 7.
It has alsc
à fuzzy
generalize
procedure
modelling
radar data
way with
ultimately
will have i
