Scientific Research and Essay Vol. 4 (4) pp. 217-225, April, 2009
Available online at http://www.academicjournals.org/SRE
ISSN 1992-2248 © 2009 Academic Journals
Full Length Research Paper
Evaluating spatial variability and scale effects on
hydrologic processes in a midsize river basin
Osman Yıldız
1*
and Ana P. Barros
2
1
Kırıkkale University, Faculty of Engineering, Department of Civil Engineering, 71451 Kırıkkale, Turkey.
2
Duke University, Pratt School of Engineering, Department of Civil and Environmental Engineering, Durham, NC 27708.
Accepted 16 March, 2009
The impact of spatial variability and scale on the dynamics of hydrologic processes in the Monongahela
river basin of USA was investigated using a physically based spatially distributed hydrologic model
developed by Yildiz (2001). The hydrologic model simulations were performed at 1 and 5 km spatial
scales for a 5 month period from April through August of 1993. Effects of spatial variability in
topography, vegetation and hydrogeology and of spatial scale were evaluated through comparisons of
the simulated and observed streamflows for the prescribed resolutions at different locations across the
river basin. The evaluation of observed and simulated streamflows using the statistical measures of
mean, standard deviation, coefficient of variation, root mean square error and bias showed that model
statistics of streamflow followed closely the spatial patterns of those of existing observations, that is,
the model captured the space-time features of the 1993 flood across the basin. The changes in the
nature of the rainfall-runoff response due to changes in the spatial resolution of the model indicated
that there was also a change in governing physical processes at different resolutions. Here, this change
was expressed in terms of the relative contributions of surface and subsurface flows.
Key words: Spatial variability; spatial scale; hydrologic model; streamflow, digital elevation model, stream
network.
INTRODUCTION
The influence of spatial variability and scale on the hydro-
logic response of watersheds and their importance in
hydrologic modeling have been widely studied by various
investigators (e.g., Amorocho 1961; Eagleson, 1970;
Dunne and Leopold, 1978; Wood et al., 1988; Entekhabi
and Eagleson, 1989; Wood et al., 1990; Seyfried and
Wilcox, 1995). In watersheds, spatial variability often
results from interactions between ecosystem character-
ristics such as topography, vegetation, and geology (Sey-
fried and Wilcox, 1995). As the spatial scale of a water-
shed increases, the watershed tends to attenuate the
complex, local patterns of runoff generation and water
fluxes, that is, it functions as a low-pass filter. As pointed
out by Amorocho (1961) the runoff generation at large
scales becomes somewhat insensitive to rainfall intensity
changes recorded at individual gauges and the catch-
*Corresponding author. E-mail: osmanyildiz@kku.edu.tr. or
osmanyildiz2000@hotmail.com. Tel.: +90 318 357 4242. Fax:
+90 318 357 2459.
ment-scale rainfall-runoff appears to be governed by
macroscale catchment characteristics. Therefore, the
transition representing hydrological processes in models
using microscale to macroscale parameterization is a
highly nonlinear process.
As spatial scale increases spatial variability may signi-
ficantly affect hydrological processes in watersheds.
investigating effects of scale on the hydrologic response
of a catchment Wood et al. (1988) proposed the so-called
representative elementary area (REA) concept which is
considered to be the smallest or critical representation of
area at which implicit continuum assumptions can be
applied for the spatially variable controls and parameters
in physical models and therefore, spatial patterns are no
longer needed to be considered explicitly. According to
the authors, a REA can be defined in large-scale hydro-
logic modeling beyond which spatial heterogeneities in
vegetation, topography, and soil can be incorporated into
hydrologic models without considering the detailed spatial
pattern of such heterogeneity within each grid cell.
Conventional lumped rainfall-runoff models generally