Modelling Catchment and Fluvial Processes and their Interactions - Environmental Modelling: Finding Simplicity in Complexity - page 191

0
20
40
60

Observed
Calibrated
Binary
Area-weighted
Unconditioned
Conditioned 1
Conditioned 2

80

0 20 40 60 80

Time min

(b)

Figure 11.9 (a) Different ways of parameterizing a distributed hydrological model from sparse data. The calibrated values are based
on the optimal fit of the outflow hydrograph, and are thus bound to give good fits to outflow data, at the expense of good spatial
representation of flow variability (see discussion in Section 11.2.5). The area-weighted values are based on measured infiltration
under vegetation and on bare surfaces, apportioned uniformly based on the plot-average vegetation cover; again there is no
representation of spatial variation. The binary approach uses a vegetation map to allocated mean values according to presence or
absence of vegetation and so does provide some simplistic spatial information. Stochastic simulation results are based on spatial
autocorrelation of parameter variability in the field. In the unconditioned case, not further information is used; condition 1 uses
information on vegetation pattern (as for the binary map); and condition 2 uses both vegetation and rill data (see further discussion
in M uller et al ., 2007). (b) The graph shows hydrographs for a representative event in July 2002.

and subsurface runoff. The properties of this connectiv-
ity determine the manner in which patch scale output
aggregate to catchment totals. Total catchment runoff
is not a simple sum of the constituent patch-level runoff
(Bracken and Croke, 2007; Stieglitz et al ., 2003) - see
also the discussion on hillslope complexity versus catch-
ment simplicity in Sivapalan (2003). Recent work by
Mulligan et al . (2010b), using a distributed hydrological
model applied in tropical montane environments,
indicates the importance of hydrological connectivity
for catchment level outputs and the implications for the
study of the hydrological impacts of land-use change.
This modelling study looks at the implications of
progressive deforestation on the catchment scale runoff
and the sensitivity of runoff to forest loss. The study
concludes that in the initial stages of deforestation
(0-75% forest cover lost), the sensitivity of runoff
to forest loss is low compared with the situation
beyond 75% loss. It is concluded that where patches
of

flowlines, they absorb excess water generated on the
poorly infiltrating deforested patches and thus extra
runoff does not cumulate downstream. When so much
forest is lost that flowlines are more or less deforested
along their whole length, runoff generation cumulates
down the flowlines and saturated wedges penetrate
further upslope thus providing a positive feedback for
further runoff generation. In this way the sensitivity of
catchment scale runoff to the deforestation of patches
in this last 25% of deforestation is very high, each patch
contributing greatly to the continuity of deforested
flow paths and thus to enhanced (flashy) streamflow.
Furthermore, the location of the deforestation relative
to the geometry of the flow network, and the properties
of the soil beneath and vegetation above it, determine
the exact outcome of forest loss. This example indicates
the importance of lateral connectivity in driving the
overall response of catchments and in integrating (in
a relatively simple way) the complex spatial hydrology
of them.

forest remain along the surface and subsurface