Prediction of annual runoff in ungauged basins - Runoff Prediction in Ungauged Basins: Synthesis across Processes, Places and Scales - page 93

Figure 5.21. (a) Rainfall surplus (gross rainfall minus actual evaporation) between June 1999 and April 2000 across Sri Lanka, with the river
basin boundaries superimposed; (b) observed and remote sensing predicted runoff for Kelani Ganga and Gin Ganga, two major rivers in Sri
Lanka between June 1999 and April 2000; (c) comparison of measured and predicted runoff volumes for a majority of river basins in Sri Lanka.
From Bastiaanssen and Chandrapala ( 2003 ).

Many of these studies are based on large data sets provid-
ing a broad range of results from catchments in different
climates. The number of catchments the studies evaluate
ranges from 1 in local studies to more than 1000 in
global studies. There are several studies that compare
different regionalisation approaches, which results in a
total of 41 and 19 results for predictive performance of
annual runoff and inter-annual variability, respectively.
The regionalisation methods used are regression, index
methods, spatial proximity, process-based methods and
methods using proxy data. The performance measure
used in the assessment is the squared correlation coeffi-
cient, r², between predicted and observed mean annual
runoff and between predicted and observed standard
deviation of annual runoff. It is important to note that

performance values presented in Tables A5.1 and A5.2
differ greatly depending on whether they are reported as
specific mean annual runoff (in mm) or as volumetric
runoff (in m 3 ). Performance values tend to be much
higher when annual runoff is expressed in volumetric
units (i.e., in m 3 ) since they include the effect of catch-
ment area, which is always a strong predictor of runoff
volume. In the figures that follow, this distinction is
highlighted by displaying the performance values based
on volumetric units as crosses andthosebasedonspe-
cific runoff as circles. For comparison with the other
runoff signatures in Chapter 12 ,ther² of annual runoff
were calculated for all methods in all studies with the
exception of tree rings. The 25% and 75% quantiles of
these r² are 0.65 and 0.91, respectively.