Geography Reference
In-Depth Information
Table 11.11. Limits of acceptability, based on the normal quantile transform
Type of data used
Description
↓
lim
↑
lim
Recession slope (day
−
1
)
Discharge Great East Road Bridge
0.0055
0.014
ρ
1
(Q)(
−
)
0.968
0.994
σ
Q
(m
3
/s)
269
1943
Water balance:
Nov
-
Jan (mm/month)
−
4.24
+10.69
Feb
-
Apr (mm/month)
−
20.7
+12.4
May
-
Jun (mm/month)
−
1.84
+3.3
Jul
-
Oct (mm/month)
−
0.91
+0.92
SEBAL evaporation maps
Evaporation per time step
μ
± 0.3
μ
Total dry-season evaporation
μ
± 0.1
μ
Parameter: S
max
(mm)
Riverine
500
650
Dambos
275
500
Forested
1300
2000
Highlands
1400
2000
Parameter: l
p
(-)
Riverine
0.75
1
Dambos
1
1
Forested
0.25
0.4
Highlands
0.5
0.6
The water balance limits of acceptability are dependent on the output of the monthly HYMOD auxiliary model. Therefore, only the
deviation (±) from the modelled output is given.
parameter inference process, they offer the ability to per-
form a direct comparison between modelled and observed
discharges, given that we have knowledge of the rating
curve and that the SRE rainfall corrected with the
local gauges is accurate enough. One hundred accepted
parameter sets were consequently used to force the hydro-
logical model from the period September 2002 until
August 2008. This ensured a generous spin-up time of 5
years for a proper initialisation of soil moisture storage.
The result of the 100 discharge realisations at Mfuwe
are plotted against the observed Mfuwe water levels in
Figure 11.59
.
As an independent validation, we can conclude that the
simulated discharges, on average, indeed follow a straight
line on the double-log plot, and that the slope of the line is
within what may be expected of a rating curve. Apart from
the hysteretic loops that are normal in the passing of a
flood wave, we can conclude that the timing of the hydro-
graph is well captured. If we plot the observed hydrograph
using this rating curve against the 100 realisations in
Figure 11.60
, then we can see that the timing of the start
of the hydrograph (which is a threshold process) and the
recession are well captured.
It is clearly evident that the validation has been generally
successful. It provides independent credibility that
calibration framework, described in detail in the last two
sections, is valid and useful for predictions in ungauged
basins, where available data may be scarce, intermittent
and non-concomitant, but where we are nonetheless forced
to use them. More importantly, the randomly selected
behavioural parameter sets produce an envelope of realisa-
tions that in most parts of the hydrograph encloses the
observed flows. This means that the model is not over-
conditioned, which is a soft proof that
the framework
indeed ensures a control on subjectivity.
Discussion
In this case study, a real world example of modelling an
ungauged basin has been presented. Crucial components of
our approach are outlined below:
(1) the modeller should get the look and feel of the dom-
inant processes by
and linking
it with catchment characteristics and information
signatures;
(2) one should install water level recorders and rain
gauges immediately during the first visit for validation
purposes;
(3) rather than a classic hydrograph matching calibration,
the modeller
'
reading the landscape
'
the
should focus on small pieces of
Search WWH ::
Custom Search