PUB in practice: case studies - Runoff Prediction in Ungauged Basins: Synthesis across Processes, Places and Scales - page 274

Figure 11.3. Annual precipitation in
the Krishna Basin, with stream
gauges and catchment boundaries
used in the analysis. Black circles
indicate catchments draining the
Western Ghats, and open circles
indicate catchments draining the
Deccan Plateau. Precipitation is from
TRMM 2B31 ( Bookhagen and
Burbank, 2010 ). Letters A and
B indicate catchments chosen for
the modelling.

B

A

Annual P (mm)
00-200
200-400
400-600
600-800
800-1000
1000-1500
1500-2000
2000-2500
2500-3000
3000-4000
4000-6000

100 km

( Figure 11.2 ). Groundwater is limited to fractured, hard-
rock aquifers, which have limited storage and are quickly
depleted if over-pumped (Dewandel et al., 2006 ).
Irrigation expanded rapidly in the basin from 1960 to
1990. After 1990, the irrigated area stabilised as available
surface water was consumed through crop evapotranspira-
tion (Biggs et al., 2008 ). In 2001, approximately 20% of
the basin was irrigated, half by surface water and half by
groundwater (Biggs et al., 2006 ). The natural vegetation is
dominated by low-lying grasses and scrub, with some
deciduous forest in the Western Ghats ( Figure 11.2 ).

long-term mean evaporation fraction, calculated as evapor-
ation (E

¼ −

Q) divided by precipitation (P), as a
function of the aridity index, calculated as potential evap-
oration (E p ) divided by P (see Chapter 5 , in particular
Sections 5.3.2 and 5.4.1). The observed relationship
between E p /P and E/P was compared with Budyko-type
models calibrated to other catchments (Zhang et al., 2001 ),
which include a single adjustable soil moisture storage
parameter (w). In general, higher soil water storage values
are

P

associated with higher E/P and lower runoff

coefficients.

Method
Budyko framework
A top-down approach was used to investigate the dominant
controls on the long-term average, annual water balance.
Top-down modelling, which is applicable where daily
discharge data are not available, attempts to identify the
key variables necessary for prediction in ungauged basins.
In the Krishna Basin, daily runoff was available for only a
few stations, and was unavailable for any stream draining
the Western Ghats, which supply most of the basin

Water budget model
Two catchments that had different E/P ratios but similar
values of the aridity index were selected for further analy-
sis of the impact of the monthly and spatial patterns of P
and E p on runoff (catchments A and B in Figures 11.3 and
11.4 ). The long-term E/P was calculated using two models:

(1) A lumped, monthly water balance model was con-
structed to test the hypothesis that the temporal distri-
bution of P and E p accounted for the observed
differences in runoff for a given aridity index. For each
month, precipitation in excess of E p was assumed to be
runoff. Given data limitations, the water balance does
not reflect conditions in any particular year, but rather
is the long-term, mean monthly water balance. The
model does not account for soil moisture storage.

s
annual discharge. In addition, the available stream gauge
data did not cover similar time periods, so comparison of
the hydrology of a group of catchments required use of the
long-term, decadal mean discharge.
The Budyko framework (Budyko, 1974 ; Monserud
et al., 1993 ; Zhang et al., 2001 ) was used to estimate the

'