A data acquisition framework for runoff prediction in ungauged basins - Runoff Prediction in Ungauged Basins: Synthesis across Processes, Places and Scales - page 30

Figure 3.1. (a) Spatial distribution of headwater stream length as a percentage of total stream length in the USA. From Nadeau and Rains ( 2007 ).
(b) Distributions of stream length and stream gauges against stream order in the USA. From Poff et al . (2006) .

the runoff prediction desired. While global or regional data
will be useful for annual runoff prediction, more intensive
local data will be needed for hydrograph prediction.
The objectives of this chapter are to provide an assess-
ment of the data available for PUB, and to provide some
initial guidance on how this data might be acquired. These
data products can often be estimated from global or national
data sets, but their availability at higher resolution or as a
directly measured or observed product from regional to
local scales can enhance data quality and therefore PUB.
Some data sources and observations can be direct (e.g.,
model input parameters or forcing) or indirect (e.g., likely
runoff dynamics interpreted from regionalisation, similar
catchments, or experience). Indirect data or observation
can provide qualitative information and aid in model selec-
tion and evaluation. Auxiliary data types and higher reso-
lution information become increasingly important at short
time scales and smaller spatial extents. The following
sections provide an overview of hydrological landscape
interpretation based on hierarchical data from the global to
regional to local scales that provide a general and practical
pathway to PUB. The introduction of this acquisition frame-
work is followed by a discussion of the main data sources
(separated by hydrological variables) from global to local
scale. In addition, three case studies illustrate the hierarch-
ical data acquisition strategy discussed here.

bias towards larger scales with respect to the US stream
gauge network (see discussion in Wagener and Montanari,
2011 ). At what spatial scale this lack of data becomes a
problem for decision-making varies from country to coun-
try. A general consequence, however, is that data scarcity
is a major issue even for highly developed (and therefore
often highly monitored countries). At the global scale, data
sources are primarily limited to remote sensing and global
climate models, notwithstanding aggregated products such
as global soil maps. In the last few decades new satellite
sensors have also made available useful measurements
across large areas. These global products, together with
regional and local observations and landscape interpret-
ation, can provide data corroboration and validation and a
hierarchy of inputs for hydrological modelling and runoff
estimates. Paradoxically, the data requirements to achieve
accurate simulations increase with decreasing temporal and
spatial scale of prediction. This is because at small spatial
scales runoff tends to be more tightly linked to details of
landscape structure and climate forcing and exhibits
greater space-time variability, thereby hampering param-
eter regionalisation and scaling (Wood et al., 1988 ). At
greater spatial scales, much system heterogeneity is sub-
sumed and averaged, often leading to simpler catchment
response to climate forcing (Sivapalan, 2003a ). Therefore,
data needs and the value or information content of data
products from global data sets to local observations for
runoff prediction are scale dependent. Indeed, the issue of
data adequacy and availability, in the context of natural
variability present across regions of the world, will be a
recurring theme throughout this topic, including in the
assessment of the performance of prediction methods.

3.2 A hierarchy of data acquisition
Most river basins around the world are ungauged. Interest-
ingly, this lack of data often increases with decreasing
catchment sizes. Figure 3.1 shows an example of the data