WATER RESOURCES RESEARCH, VOL. 48, W04509, doi:10.

1029/2011WR011298, 2012

On the effects of improved cross-section representation in

one-dimensional flow routing models applied to ephemeral rivers
Christopher J. Hutton,1,2 Richard E. Brazier,2 Andrew P. Nicholas,2 and Mark Nearing3
Received 18 August 2011; revised 2 March 2012; accepted 3 March 2012; published 10 April 2012.
[1] Flash floods are an important component of the semiarid hydrological cycle, and
provide the potential for groundwater recharge as well as posing a dangerous natural
hazard. A number of catchment models have been applied to flash flood prediction;
however, in general they perform poorly. This study has investigated whether the
incorporation of light detection and ranging (lidar) derived data into the structure of a 1-D
flow routing model can improve the prediction of flash floods in ephemeral channels. Two
versions of this model, one based on an existing trapezoidal representation of cross-section
morphology (K-Tr), and one that uses lidar data (K-Li) were applied to 5 discrete runoff
events measured at two locations on the main channel of The Walnut Gulch Experimental
Watershed, United States. In general, K-Li showed improved performance in comparison to
K-Tr, both when each model was calibrated to individual events and during an evaluation
phase when the models (and parameter sets) were applied across events. Sensitivity analysis
identified that the K-Li model also had more consistency in behavioral parameter sets across
runoff events. In contrast, parameter interaction within K-Tr resulted in poorly constrained
behavioral parameter sets across the multidimensional parameter space. These results,
revealed with a modeling focus on the structure of a particular element of a distributed
catchment model, suggest that lidar derived cross-section morphology can lead to improved,
and more robust flash flood prediction.
Citation: Hutton, C. J., R. E. Brazier, A. P. Nicholas, and M. Nearing (2012), On the effects of improved cross-section representation
in one-dimensional flow routing models applied to ephemeral rivers, Water Resour. Res., 48, W04509, doi:10.1029/2011WR011298.

1. Introduction empirical regression-based models [McIntyre et al., 2007];

[2] Flash floods are defined as runoff events that occur semiempirical models [McIntyre and Al-Qurashi, 2009];
within 6 hours of the causative rainfall event [National spatially lumped models (e.g., Sacramento Soil Moisture
Weather Service, 2002], and are the dominant runoff Accounting Model [Burnash, 1995]) ; and distributed pro-
response in many ephemeral semiarid catchment systems cess based models [El-Hames and Richards, 1998], includ-
[Goodrich et al., 1997; Garcia-Pintado et al., 2009]. Flash ing the Kinematic Runoff and Erosion model (KINEROS)
floods are important elements of the semiarid hydrological [Smith et al., 1995].
cycle that must be understood for two primary reasons. First, [4] The development of distributed process based mod-
these intermittent events provide potential for groundwater els has attempted to overcome some of the empirical limi-
recharge via transmission losses (e.g., infiltration through the tations of simpler model structures. However, predictions
streambed), and are therefore an important water resource in derived from distributed process based models are highly
semiarid environments [Coes and Pool, 2005; Morin et al., uncertain owing to uncertain parameter values and bound-
2006]. Second, flash floods present a dangerous natural haz- ary conditions [Yatheendradas et al., 2008; Garcia-Pin-
ard that can detrimentally impact channel morphology tado et al., 2009], and because of epistemic uncertainty
[Hooke and Mant, 2000], human infrastructure [Foody et al., surrounding the processes themselves. A significant source
2004], and cause a significant number of fatalities [Ashley of model uncertainty results from a paucity of spatial infor-
and Ashley, 2008]. mation, notably information on distributed rainfall and hill-
[3] A number of models have been developed to predict slope infiltration properties [Al-Qurashi et al., 2008;
and understand semiarid catchment hydrology, including: Yatheendradas et al., 2008]. In data poor situations, predic-
tions derived from simpler (semi-) empirical models may
be preferred, and indeed may be less uncertain [McIntyre
1
Centre for Water Systems, College of Engineering, Mathematics and et al., 2007; McIntyre and Al-Qurashi, 2009]. Simpler
Physical Sciences, University of Exeter, Exeter, UK. models, however, may not be appropriate to resolve
2
Geography, College of Life and Environmental Sciences, University of adequately key temporal and spatial processes controlling
Exeter, Exeter, UK. flash flooding in semiarid environments. Simulation of
3
Southwest Watershed Research Center, United States Department of
Agriculture, Tucson, Arizona, USA.
these processes is required to understand and resolve the
complex processes, thresholds and interactions that govern
Copyright 2012 by the American Geophysical Union the rainfall-runoff response in different semiarid catch-
0043-1397/12/2011WR011298 ments [Goodrich et al., 1997].

W04509 1 of 11
W04509 HUTTON ET AL.: FLOW ROUTING IN EPHEMERAL RIVERS W04509

[5] Distributed parameter and initial condition uncertainty account for feedbacks between cross-section shape and
is a significant problem in itself [Yatheendradas et al., 2008; flood-wave propagation, a significant factor controlling
Garcia-Pintado et al., 2009], but also confounds the exercise flood routing [Hassan, 1990], even in the absence of trans-
of identifying structural errors within model components that mission losses. In such cases the applied roughness coeffi-
may contribute to overall predictive uncertainty. A renewed cient will need to account for topographic variability not
focus on reducing model structural uncertainty is evident in represented by a more explicit definition of cross-section
the literature [Refsgaard et al., 2006; Krueger et al., 2009], shape, along with other forms of frictional resistance owing
and will be facilitated by the increased availability of high- to the representation of depth and width using a 1-D approach
quality datasets [Bates et al., 2003; Croft et al., 2009]. One [Lane, 2005]. Assumptions regarding cross-section shape are
uncertain structural component in distributed hydrological also likely to have a negative effect on sediment transport
models is the channel flow routing component. estimates derived from such routing models, given that
[6] In semiarid environments, ephemeral river channels cross-section bed load sediment transport is sensitive to the
have an increasing effect on catchment hydrological lateral distribution of flow [Ferguson, 2003]
response with an increase in catchment size [Goodrich [9] The development and proliferation of topographic
et al., 1997]. Methods that seek to simulate channel hydrol- datasets derived from light detection and ranging (lidar)
ogy include: regression relationships between incoming technology has facilitated the parameterization of numeri-
and outgoing discharge [Walters, 1990]; and empirically cal models at a fine spatial resolution (1 m) over increas-
derived routing methods [Sharma and Murthy, 1995]. ingly large model domains, for both 1-D models [Matgen
Transmission losses, however, are a nonlinear function of et al., 2007; Aggett and Wilson, 2009], and also distributed
discharge and time [Mudd, 2006]. Consequently, explicit (2-D) flow routing models [Cobby et al., 2001; Bates et al.,
routing methods are required to understand how the rela- 2003; French, 2003; Hilldale, 2007]. A lidar Digital Ele-
tionship between inflow discharge and channel characteris- vation Model (DEM) offers the potential to constrain cross-
tics governs infiltration and downstream discharge within section morphology over larger areas than is feasible
ephemeral river reaches [Goodrich et al., 1997]. through ground survey alone, while providing comparable
[7] A number of models have been developed to simu- levels of accuracy [Rayburg et al., 2009]. Lidar is particu-
late ephemeral channel rivers explicitly, based on either larly useful in ephemeral channels as the channel bed may
full [El-Hames and Richards, 1998; Mudd, 2006] or partial be surveyed during no-flow conditions, which is not possi-
solutions [Smith et al., 1995] of the one-dimensional (1-D) ble in perennial rivers.
St. Venant equations. Results from numerical and field [10] High-resolution DEM data available over large
investigations demonstrate the importance of hydrograph areas have the potential to improve the representation of
duration [Parissopoulos and Wheater, 1991], and channel cross-section morphology in 1-D flow routing models
width [Goodrich et al., 1997; Mudd, 2006] in controlling applied to ephemeral rivers. However, given other uncer-
transmission losses in ephemeral channels. The wetted area tainties surrounding flow routing in these environments, it
of the channel bed during flood flows appears to be the pri- is unclear whether such data sources can improve the pre-
mary control on channel transmission losses [Goodrich dictive ability of existing 1-D flow routing models. Although
et al., 1997; Mudd, 2006], and therefore the magnitude and manual surveys of similar accuracy have been conducted
duration of the downstream hydrograph. These results dem- previously, such information has not often been included in
onstrate the importance of accurately parameterizing cross- 1-D models. Climatic scenarios point toward a drier climate
section shape and the processes governing infiltration in for the American Southwest, and more frequent, high-
ephemeral channel flow routing models. Existing methods intensity rainfall events [Seager et al., 2007]. There is there-
applied to simulate channel flow routing in ephemeral rivers fore a need to improve the predictive ability of hydrological
have assumed channel morphology may be approximated by models applied in such regions, to improve understanding
either trapezoidal [Smith et al., 1995] or rectangular (con- and prediction of flash flood hazard and water resources.
stant width) cross-sections [El-Hames and Richards, 1998; [11] This study will investigate whether incorporating
Morin et al., 2009]. Trapezoidal and rectangular channels high-resolution (1 m) topography into the structure of a 1-D
may provide an adequate representation of channel cross- flow routing model can improve flow routing predictions
section morphology in single thread reaches. However, when applied to an ephemeral river, in comparison to a
ephemeral piedmont rivers of the American Southwest alter- model using an existing, simplified representation of cross-
nate between single thread and braided sections [Pelettier section morphology. The study has the following research
and DeLong, 2004]; trapezoidal cross-sections do not aims: (1) Determine whether the integration of distributed
adequately represent multiple thread channels. Differences topographic information can improve 1-D flow routing in
between the simplified cross-section and actual channel mor- ephemeral rivers on an event basis (calibration); (2) Identify
phology will introduce errors into the relationship between how improved topographic representation modifies model
stage and wetted perimeter, which will affect flow convey- structure and affects model parameter uncertainty; (3) Eval-
ance and the bed area available for infiltration. In the case of uate whether modifications to 1-D model structure, and the
KINEROS [Smith et al., 1995], an empirical correction fac- increase in topographic information contained within the
tor is applied that reduces the effective wetted perimeter of model can improve model predictive ability (evaluation).
the cross-section that is available for infiltration at low flows.
However, there is uncertainty regarding the value that this
coefficient should take [Yatheendradas et al., 2008]. 2. Modeling Strategy: Kinematic Wave Model
[8] A parameter applied to correct for the effect on infil- [12] To address the research aims a 1-D kinematic flow
tration of an artificially high wetted perimeter does not routing model was applied to simulate runoff events along

2 of 11
W04509 HUTTON ET AL.: FLOW ROUTING IN EPHEMERAL RIVERS W04509

the main channel of the Walnut Gulch Experimental Water- where BW is the channel bottom width (m), and Wc is the
shed (WGEW) [Renard et al., 2008]. The flow routing empirical Woolhiser coefficient. In K-Tr simulations,
model was applied using two alternative model structures : cross-section cells are ordered from minimum elevation to
First, using a trapezoidal representation of cross-section the maximum elevation, and pe is used to determine the
morphology (K-Tr) ; Second, using laterally distributed fraction of wet cells to calculate infiltration, which are then
cross-section morphology derived from a 1 m lidar derived summed to determine q.
DEM (K-Li).
[13] The 1-D kinematic wave equation, which has been
3. Study Area and Data
widely applied to simulate flow in ephemeral channels
[Garcia-Pintado et al., 2009; Morin et al., 2009; Smith [14] The research aims were addressed using data
et al., 1995; Yatheendradas et al., 2008], is applied here derived from The WGEW, southeast Arizona, USA
and solved at each cross-section in the model domain using (31.45 N, 110.0 W; Figure 1; see WRR special issue)
an explicit scheme: [Moran et al., 2008]. The watershed, with an area of
150 km2, drains west from headwaters in the Dragoon

t t Mountains in the east (1900 m amsl) into the San Pedro
Atþ1
i ¼ Ati  ðQ  Qti1 Þ  qi
t; (1)

x i River (1250 m amsl). The main channel of The WGEW is
where A is the flow cross-section area (m2); t is time (s), a 6.5 km continuous sand bed river, which alternates
and subscript i is cross-section; Q is discharge (m3 s1) cal- between single thread reaches and braided sections with a
wavelength >200 m (Figure 2). The reach may be consid-
culated from A using the Manning equation with Manning’s
ered typical of river morphologies present on Piedmont
coefficient, n; x is distance in the stream-wise direction
slopes of the basin and range province of the semiarid
(m); and q represents transmission losses (m2 s1), which
American Southwest [Pelettier and DeLong, 2004].
are determined at each cross-section by calculating the sum
[15] Lidar data used to determine channel cross sections
of infiltration across all wet cross-section cells. Infiltration
for the 1-D model were acquired from an OPTECH ALTM
rate (I) in each cell is calculated using the Green-Ampt
1233 (Optech Incorporated, Toronto, Canada) laser scanner
equation [Green and Ampt, 1911], capable of simulating
flown over The WGEW in the summers of 2003 and 2004.
run on infiltration:
The Optech ALTM 1233, which has a 1064 nm laser, a
wþzþh pulse rate of 33 kHz, a scanning frequency of 28 Hz, and a
I ¼ Ks ; (2)
z scanning angle of 620 , was flown to obtain a spot size of
approximately 15 cm, and the data processed using Optech
where I is infiltration rate (m s1), Ks is saturated hydraulic REALM proprietary software, alongside a vegetation filter-
conductivity (m s1), w is the wetting front suction (m), z is ing algorithm [Hutton, 2010], to derive a 1 m resolution
accumulated depth of infiltration (m) in the cell, and h is DEM of the watershed. The DEM had a vertical accuracy
the depth of water at the bed surface (m). Infiltration in of 60.15 m derived in comparison to ground based differ-
K-Tr is calculated using the empirical correction factor ential GPS survey points measured in 2003 at Flumes 1 and 2,
applied in KINEROS(2), which uses an effective wetted and also using stable locations at surveyed cross-sections in
perimeter (pe) to correct for the error introduced in the the reach.
actual wetted perimeter (a) when calculating infiltration in [16] Analysis of the 1 m channel DEM was used to deter-
trapezoidal cross-sections [Smith et al., 1995]: mine where to extract representative cross-sections of the
  channel morphology. The 101 cross-sections were extracted
h manually from the DEM at a spacing of 65 m, a spacing
pe ¼ min p ffiffiffiffiffiffiffiffi ; 1 a; (3)
Wc BW sufficient to represent changes between braided and single

Figure 1. Hillshade map of the Walnut Gulch Experimental Watershed, SE Arizona. The main channel
reach under study is enclosed in the black rectangle, as shown in Figure 2.

3 of 11
W04509 HUTTON ET AL.: FLOW ROUTING IN EPHEMERAL RIVERS W04509

Figure 2. Lidar derived DEM hillshade showing downstream changes between braided and single
thread morphology, and two representative cross-sections. Note exaggerated Y-Axis. The upper-left pho-
tograph shows a flash flood event looking upstream from Flume 1.

thread sections in the reach, which alternate downstream [19] Each model structure (K-Tr and K-Li) was applied
with an approximate wavelength of >200 m (Figure 2). Tra- to simulate the events listed in Table 1. For each model
pezoid cross-sections were constructed directly from the application to an event, 40,000 Monte Carlo simulations
lidar data: the channel bottom elevation in each cross were conducted sampling randomly from uniform prior dis-
section was set to the elevation of the thalweg identified tributions for each parameter (Table 2). Prior ranges were
from the lidar data; trapezoidal cross sections were fitted by determined with recourse to studies within similar sand-
varying the channel bottom width to reduce the mean- gravel ephemeral channels [Al-Qurashi et al., 2008; Blasch
square-error between the original cross section data and the et al., 2006; Dahan et al., 2007; El-Hames and Richards,
trapezoidal cross section (Figure 2). 1998; Michaud and Sorooshian, 1994; Morin et al., 2009;
[17] Five discrete flow events recorded at Flume 2 and Yatheendradas et al., 2008], and hydrological properties
Flume 1 upstream and downstream of the study reach, respec- for the bed sediment texture [Rawls et al., 1993]. After
tively, were used to address the research aims (Table 1). The 20,000 simulations the initial parameter ranges were nar-
selected events first activated Flume 2 and then Flume 1 with rowed where no well performing parameter sets were found
little or no rainfall recorded in the tributary catchments that across all events (Table 2). For each event an evaluation of
join the main reach between the flumes (Figure 1). These the convergence of the Cumulative Distribution Functions
events were therefore chosen to ensure mass conservation (CDF) across each parameter range for each measure of
when calculating both infiltration in the reach, and the down- model performance demonstrated 40,000 simulations were
stream hydrograph at Flume 1. sufficient for convergence of the posterior distribution.
[20] In order to identify model structures and parameter
4. Model Evaluation combinations that perform well for the right hydrological
[18] The model evaluation procedure employed to reasons [Brazier et al., 2000], four performance measures
address the research aims follows the concepts outlined (PM) have been chosen to evaluate model performance,
within the Generalized Likelihood Uncertainty Estimation calculated for each event simulation (Table 3). The fourth
(GLUE) methodology [Beven and Binley, 1992; Brazier PM (NT) is derived by multiplying the values of all other per-
et al., 2000], which is applied to understand how different formance measures to define a good model prediction as one
parameter sets (and interactions) within competing models which replicates the magnitude and timing of discharge (NP),
structures affect model performance [e.g., Beven and Freer, and the shape of the hydrograph (NSE), while also maintain-
2001]. ing the correct mass balance (e.g., predict transmission losses

Table 1. Summary Discharge Statistics of Runoff Events Used in Model Evaluationa

Flume 2 Flume 1

Event Number and Date Q Total (m3) Q Peak (m3 s1) Q Total (m3) Q Peak (m3 s1) Q Loss (m3(%))

1. 17 Jul 1999 71,191 12.6 42,572 12.6 28,619 (40)

2. 23 Jul 1999 261,147 61.2 238,838 61.1 22,309 (9)
3. 22 Aug 2005 54,498 12.7 46,587 6.3 7911 (15)
4. 10 Sep 2006 36,202 9 16,227 4.2 19,975 (55)
5. 25 Jul 2007 28,604 7.7 11,169 3.6 17,435 (60)
a
Summary Discharge is Q.

4 of 11
W04509 HUTTON ET AL.: FLOW ROUTING IN EPHEMERAL RIVERS W04509

Table 2. Parameter Ranges Used in Monte Carlo Simulation for top performing 500 parameter sets, based on NT, were also
Each Model Structurea evaluated.
[23] To address the third research objective and identify
Parameter K-Li K-Tr
whether K-Li can outperform K-Tr when a single parame-
Manning (n;) 0.015–0.1 0.015–0.1 ter set is applied across runoff events, the PM scores for
(0.015–0.05) (0.015–0.05) each parameter set are summed across events and renor-
Initial moisture (M;%) 0–1 0–1
Saturated Conductivity 1.8–432 (1.8–144) 1.8–432
malized to 1, allowing a parameter set to perform poorly
(Ks ; mm h1) for an event, yet still score well overall for good perform-
Wetting Front Suction (Wfs ; m) 0.0009–0.1 0.0009–0.1 ance across all other events [Yatheendradas et al., 2008].
Woolhiser Coefficient (Wc) – 0–0.45
a
Ranges in brackets show narrowed ranges sampled during the final
20,000 simulations based on behavioral simulations found in the first 5. Results
20,000 simulations.
5.1. Event Based Model Performance
[24] A comparison of the optimal PM scores derived
correctly) at Flume 1 (NV). A value of 1 for all of the meas- when applying each model structure to each runoff event
ures considered above indicates a perfect fit. All simulations (Table 4), show that for four events K-Li outperforms K-Tr
that produce a value less than zero are considered nonbeha- in terms of total performance (NT). K-Tr produces a better
vioral for that performance measure, and are given a value overall performance for Event 3, however both models out-
of zero. perform the benchmark models (BM) for this event, there-
[21] In order to provide reach specific context to evaluate fore both capture aspects of the reach transfer function that
the quality of model performance, and determine how well is not present in the input hydrograph. For all events both
each model can simulate the reach transfer function models outperform the NP BM, and can produce near opti-
between the upstream and downstream hydrograph, bench- mal peak discharge predictions (NP > 0.96). Neither model
mark values for NSE and NP are derived following Shaefli can outperform the NSE score for Event 2, and K-Tr is also
and Gupta [2007]. However, instead of adjusting input worse than NSE BM for Events 4 and 5. Both models are
rainfall, the input hydrographs for each event are multiplied capable of producing near perfect mass balance for each
by the runoff ratio (Flume 1 volume divided by Flume 2 event when considered only in terms of the NV perform-
volume) and adjusting by an optimum lag which minimizes ance measure. The NT scores for each event are less than
the value of NSE and NP separately for each event com- the product of the optimal scores for NSE, NP and NV ;
pared to the respective hydrograph at Flume 1. therefore the optimal NT score is not produced from a sin-
[22] A global method, Regional Sensitivity Analysis gle parameter set that produces optimal scores for all other
(RSA) [Brazier et al., 2007; Freer et al., 1996; Horn- performance measures.
berger and Spear, 1981], is applied to evaluate model sen-
sitivity. To overcome the problems of specifying a single 5.2. Sensitivity Analysis
restrictive behavioral threshold, and to address the second
research aim, model sensitivity is evaluated by calculating [25] Based on the RSA sensitivity scores for all events
the CDF for each performance measure as a function of and performance measures calculated for the top 10% of
each model parameter for the top 10% and also top 50% of parameter sets (Table 5), K-Li is most sensitive to Ks fol-
model simulations when applied to each event. RSA sensi- lowed by n (except NSE for Event 1), and like the K-Tr
tivity scores are derived by calculating the difference in model, insensitive to the initial moisture (M) and wetting
area between the uniform prior CDF and that of the poste- front suction (wfs). The lowest n sensitivity scores in both
rior CDF for each parameter. The range of each parameter models for NV show the correct runoff volume may be pre-
is normalized when calculating the aerial difference to dicted without necessarily producing the hydrograph shape,
compare sensitivity across parameters, which therefore has as measured by NSE. In K-Tr the single dominant parame-
a maximum value of 0.5. To supplement the RSA sensitiv- ter is n, with the exception of NV for Event 5. The next
ity analysis, which only considers first-order sensitivity, the most sensitive parameters were Ks and Wc. Some of the
strength of linear relationships between parameters for the largest Ks sensitivity scores for the T model occurred in
Event 2, the largest runoff event. The results obtained using
the top 50% of behavioral parameter sets (not shown) iden-
Table 3. Performance Measures Used in Model Evaluation tified the same ordering of the most sensitive parameters.
[26] For K-Li, significant positive linear interactions
Name Equationa were identified between M and Ks, except for Event 3
0 XT 1
ðQ t
 Q t 2
Þ
(Table 6), and between n and M, except for Events 2 and 3.
o s
Nash-Sutcliffe (NSE) max @1  XTi¼1 ; 0A Given RSA identifies optimal Ks and n in a narrow area in
i¼1
ðQto  Qo Þ2 parameter space, such interactions are of secondary impor-
    
Normalized Peak (NP)
jQ Q j jT T j
max 1  psQpo po  1  psTpo po ; 0 : tance. In K-Tr for all events except Event 2, a significant
  positive interaction was identified between Ks and Wc,
Normalized Volume (NV) max 1  jVsVVo
oj
;0 which coupled with low RSA scores shows this interaction
Normalized Total (NT) NSE þ NP þ NV for the optimal performing parameter sets occurs across the
a t
Q ¼ discharge at time t; Q ¼ mean event discharge; Qp ¼ peak dis-
whole range sampled for each parameter. Significant posi-
charge; V ¼ total event volume; Tp ¼ time of peak discharge; subscripts o tive interactions between n and both Ks and Wc were identi-
and s refer to observed and simulated, respectively. fied when K-Tr was applied to Event 3.

5 of 11
W04509 HUTTON ET AL.: FLOW ROUTING IN EPHEMERAL RIVERS W04509

Table 4. Performance Measures for the Optimal Performing Parameter Sets for Each Runoff Event and Model Structurea
NSE NP NV NT

Event Number and Date K-Li K-Tr BM K-Li K-Tr BM K-Li K-Tr K-Li K-Tr

Event 1: 17 Jul 1999 0.965 0.964 0.857 0.984 0.998 0.613 0.999 0.999 0.744 0.651
Event 2: 23 Jul 1999 0.978 0.969 0.983 0.963 0.960 0.913 0.999 0.999 0.911 0.882
Event 3: 22 Aug 2005 0.943 0.962 0.684 0.997 0.994 0.274 0.999 0.999 0.634 0.763
Event 4: 10 Sep 2006 0.986 0.978 0.985 0.991 0.998 0.953 0.999 0.999 0.900 0.811
Event 5: 25 Jul 2007 0.991 0.943 0.964 0.990 0.995 0.845 0.999 0.999 0.955 0.769
Total 0.862 0.845 – 0.868 0.863 – 0.884 0.885 0.662 0.598
a
BM refers to predictions derived from the benchmark model. The bolded scores show the best performing model for each measure of model
performance.

5.3. Predictive Performance parameter sets in the range 7–46 mm h1, and 0.02–0.03,
[27] K-Li outperformed K-Tr when measured by NSE, respectively. All event optimal values of Ks and n lie in these
NP and NT (Table 4) when parameter sets were evaluated ranges, except Event 2, which had a higher roughness coeffi-
across events, and produced a number of better performing cient (0.035). The lack of significant relationships in K-Li
parameter sets (Figure 3). Sensitivity across all events con- across all events for M, alongside the spread in the optimal
firms the sensitivity results identified for each individual initial moisture for each event (Figure 3), suggests M is spe-
event (Table 4; Figure 3): K-Li performance across all cific to each event. The sensitivity to n across events for K-
events is most sensitive to Ks and Manning’s n, with optimal Tr is also shown in Figure 3; the optimal value is smaller
than in K-Li (0.015–0.023), and like K-Li all event optimal
values of n lie in this range, except for Event 2. The strong
Table 5. RSA Sensitivity Scores Derived From the Top 10% of
linear relation between Wc and Ks in K-Tr predictions was
Parameter Sets When Each Model Was Applied to Each Event for
also identified for total model performance across all events.
Each Measure of Model Performancea
In contrast to K-Li predictions, optimal performing parame-
K-Li K-Tr ter sets can be found across the whole range of Ks, which is
also the case for M and Wc. Although, the best performing
PM n Wfs M Ks n Wfs M Ks Wc parameter sets are toward the lower range of Wc.
Event 1 [28] Narrower clustering of the top 500 parameter sets in
NSE 0.436 0.003 0.049 0.405 0.469 0.003 0.023 0.046 0.134 K-Li model (Figure 4) shows the larger number of optimal
NP 0.391 0.006 0.095 0.452 0.425 0.006 0.039 0.254 0.186 performing parameter sets, compared to K-Tr, particularly
NV 0.268 0.004 0.047 0.444 0.180 0.010 0.014 0.024 0.120
NT 0.426 0.003 0.058 0.427 0.465 0.004 0.013 0.012 0.107
for Event 1 and 2. However this is not the same for Event 3,
where K-Tr outperforms K-Li, most notably around the sec-
Event 1 ond hydrograph peak. For Event 1 and Event 2 both models
NSE 0.309 0.005 0.057 0.415 0.371 0.011 0.051 0.203 0.120 struggle to predict peak discharge in evaluation. When both
NP 0.302 0.006 0.129 0.450 0.327 0.008 0.081 0.360 0.117 models were calibrated to NP, peak discharge can be pre-
NV 0.283 0.005 0.069 0.474 0.278 0.007 0.132 0.390 0.117
NT 0.296 0.005 0.087 0.455 0.344 0.011 0.083 0.330 0.129 dicted well, which in the case of Event 1 results in an over
prediction of the receding hydrograph limb. The time to the
Event 3 rising limb is well predicted for all other events, and is mar-
NSE 0.358 0.005 0.023 0.458 0.451 0.004 0.014 0.069 0.165 ginally better in K-Li predictions for Event 4 and Event 5.
NP 0.402 0.004 0.074 0.407 0.443 0.005 0.007 0.022 0.144
NV 0.273 0.006 0.059 0.476 0.312 0.006 0.040 0.305 0.214
For these events K-Li also performs better during the reced-
NT 0.357 0.006 0.025 0.463 0.447 0.003 0.006 0.062 0.176 ing hydrograph limbs, notably in Event 5.

Event 4
NSE 0.373 0.004 0.023 0.413 0.449 0.005 0.006 0.054 0.053 Table 6. Significant Linear Relationships Between Parameters
NP 0.379 0.004 0.056 0.408 0.421 0.004 0.003 0.099 0.073 Determined From the Top 500 Parameter Sets for Each Event and
NV 0.275 0.005 0.055 0.444 0.091 0.004 0.028 0.043 0.069 Model as Measured With NTa
NT 0.369 0.004 0.024 0.425 0.448 0.003 0.006 0.104 0.099
K-Li K-Tr
Event 5
NSE 0.379 0.004 0.026 0.415 0.453 0.005 0.008 0.065 0.056 Event Number and Date Parameters R2
Parameters R2
NP 0.381 0.007 0.061 0.419 0.424 0.007 0.005 0.069 0.088
NV 0.273 0.005 0.050 0.438 0.043 0.006 0.028 0.083 0.071 Event 1: 17 Jul 1999 n-M þ0.17 Wc-Ks þ0.72
NT 0.379 0.004 0.028 0.425 0.460 0.005 0.012 0.105 0.135 Event 1: 17 Jul 1999 M-Ks þ0.41 – –
Event 2: 23 Jul 1999 M-Ks þ0.16 – –
Total Event 3: 22 Aug 2005 n-Ks 0.17 Wc-Ks þ0.47
NSE 0.381 0.003 0.023 0.425 0.452 0.006 0.007 0.020 0.086 Event 3: 22 Aug 2005 – – n-Ks þ0.14
NP 0.379 0.005 0.069 0.424 0.428 0.008 0.007 0.017 0.110 Event 3: 22 Aug 2005 – – Wc-n þ0.63
NV 0.274 0.004 0.047 0.448 0.206 0.003 0.009 0.131 0.158 Event 4: 10 Sep 2006 M-Ks þ0.52 Wc-Ks þ0.75
NT 0.373 0.005 0.039 0.441 0.452 0.005 0.007 0.015 0.083 Event 4: 10 Sep 2006 n-M þ0.15 – –
a
Event 5: 25 Jul 2007 n-M þ0.21 Wc-Ks þ0.7
Total shows values across events. PM is model performance. Bold and Event 5: 25 Jul 2007 M-Ks þ0.55 – –
italic indicates, respectively, the most sensitive parameter (bold), and sec- Total – – Wc-Ks þ0.77
ond most sensitive parameter (italic), for each model and performance
a
measure. For the values, p < 0.01.

6 of 11
W04509 HUTTON ET AL.: FLOW ROUTING IN EPHEMERAL RIVERS W04509

Figure 3. Dotty plots showing total model performance (NT) as a function of the most sensitive pa-
rameters when each parameter set was applied to all runoff events (evaluation), and the optimal perform-
ing parameter when each model was calibrated to each calibrated to each individual event.
7 of 11
W04509 HUTTON ET AL.: FLOW ROUTING IN EPHEMERAL RIVERS W04509

Figure 4. Hydrograph plots showing the optimal performing model predictions of both models (K-Li
and K-Tr) in comparison to 5 runoff events at Flume 1 in both calibration and evaluation, as measured
by NT and NP. The figures also show the top 500 performing parameter sets in evaluation, as measured
by NT when performance was evaluated across events.

8 of 11
W04509 HUTTON ET AL.: FLOW ROUTING IN EPHEMERAL RIVERS W04509

6. Discussion activity. However, as the first two events had the largest
[29] The integration of distributed topographic informa- transmission losses, a greater understanding is required of
tion into the K-Li model, in the form of lidar derived eleva- how reduced peak infiltration leads to greater cross-section
tions, generally improves model performance compared to inundation downstream and therefore increasing infiltra-
K-Tr when the models are calibrated to individual events, tion. Infiltration is calculated separately in each cell that
and also when applied across events, producing more be- constitutes the channel cross-section in K-Li, which in con-
havioral simulations. The exception to this is for Event 3, trast to a laterally lumped approach means that each newly
where K-Tr performs better in predicting both hydrograph inundated cross-section cell simulates higher transient infil-
peaks. However, both models significantly outperform the tration rates associated with an initially dry bed [Blasch
benchmark model for this event. et al., 2006]. The model is therefore more representative of
[30] The sensitivity analysis conducted to address the the physical conditions and complex dynamics governing
second research aim supports the conclusion that K-Li is a transmission losses.
better model for future event prediction in comparison to [33] The optimal performing parameter sets across
K-Tr. In the former model, the two most sensitive parame- events in K-Tr were toward the lower range for n. In the
ters Ks and n are similar when calibrated to each event, and braided sections of the main channel a trapezoidal cross-
in evaluation when applied across events. The exception to section provides a poor approximation of the true morphol-
this is the calibrated n for Event 2, which was the largest ogy (Figure 2). The higher wetted perimeter therefore
event considered, and had a peak discharge approximately requires a lower and less physically realistic roughness
five times the size of the next largest event. Event 2 would coefficient to convey the flood discharge. Strong parameter
have inundated a much larger area of the channel and flood- interactions within K-Tr were identified between Ks, and
plains, and also inundated more of the vegetation that has Wc. The compensatory effect of these parameters on the
developed within and alongside the channel over the last rate of infiltration resulted in optimal parameter sets in dif-
40 years [Nichols and Shipek, 2006]. The event, therefore, ferent areas of parameter space for all events considered.
required a higher roughness coefficient to reflect these con- The result is that it is difficult to infer the physical meaning
ditions. In the other events the optimal range for n was of the Wc, and for predictive purposes, optimize or ‘‘fix’’
0.020–0.030, which is consistent with literature values for K-Tr parameters for future application. It is possible to fix
clean, straight channels [Chow, 1959]; i.e., the main inset Wc at the default value (0.15) applied in a number of KINE-
channel in Walnut Gulch. In future application aerial im- ROS(2) applications [Smith et al., 1995], which gives
agery of the channel, alongside the unfiltered lidar DEM, slightly worse optimal predictions than those found in this
may be used to distinguish vegetated from bare channel study (Table 4), and results in optimal infiltration rates in
areas, and allow calibration of roughness coefficients both the range 5–223 mm h1. While the upper end of this range
for the channel and vegetated floodplain. might seem physically more realistic than the optimal val-
[31] In K-Li the optimal infiltration values in both cali- ues found in K-Li (7–46 mm h1), they are obtained by
bration and evaluation are between 7 and 46 mm h1. applying an empirical derived, and poorly justified parame-
Although these infiltration rates are within the wide range ter value. Whereas in K-Li the optimal range of Ks values
of saturated infiltration rates (1.2–254 mm h1) recorded are found with a physically more plausible, and distributed
for comparable rivers of the American Southwest [Hoffman representation of the effects of cross-section morphology
et al., 2002; Constantz et al., 2003; Blasch et al., 2006], they on flow routing.
are lower than that recorded for this channel in previous [34] A number of studies have identified the lack of
experiments [Coes and Pool, 2005]. Dahan et al. [2007] transferability of parameter values between event predic-
found infiltration rates recorded during a natural flow event tions when used in semiarid hydrological models [Al-Qura-
were typically lower than those recorded by ring infiltrome- shi et al., 2008; Yatheendradas et al., 2008], and the need
ters and ponding experiments. Lower infiltration rates during to take an event based approach to understand/predict semi-
natural flow events may result from air escaping at the flood arid hydrological response [Knighton and Nanson, 2001;
bore wave [Hassan, 1990], and the presence of abundant fine Wainwright et al., 2008; Garcia-Pintado et al., 2009]. The
sediment near to the channel bed that may impede infiltration consistency of parameters in K-Li suggests, alongside
[Lange, 2005]. Such processes are not currently represented recourse to aerial imagery and consideration of previous
in existing infiltration models, and might be accounted for events, predictions from this model are more robust for fur-
indirectly by changes in other parameter values (e.g., Ks). ther application than identified in previous studies.
Further work is required to understand the effect of these fac- [35] Data available for this study allowed focus on a par-
tors on infiltration. ticular structural element of distributed catchment models,
[32] The initial moisture content (M) had a secondary which led to development of a better understanding of pa-
influence on model performance on an event basis through rameter interactions, detection of model structural deficien-
interactions with other model parameters in K-Li. Events 1 cies, and identification of where advances in data collection
and 2, which both maintained their peak discharges, may improve model application. Such model limitations
occurred in a period of frequent channel activity. In con- are difficult to infer from whole-of-catchment model appli-
trast, Event 4 and Event 5 were preceded by little channel cations due to the compensatory effects of errors in hill-
activity. These data suggest that the initial moisture content slope model structure, input data and parameters, which
may be important in reducing infiltration rates near the may lead to the identification of erroneous parameter val-
peak, and that model results may be improved by incorpo- ues [Yatheendradas et al., 2008; Bahat et al., 2009]. The
ration of a moisture model dependent on recent channel issue is analogous to the problems of the sediment delivery

9 of 11
W04509 HUTTON ET AL.: FLOW ROUTING IN EPHEMERAL RIVERS W04509

concept [Parsons et al., 2006] as it highlights the limita- Brazier, R. E., K. J. Beven, J. Freer, and J. S. Rowan (2000), Equifinality
tions of inferring catchment understanding from a single and uncertainty in physically based soil erosion models: Application of
the glue methodology to WEPP-the water erosion prediction project-for
integrated measure of catchment response. Improved data sites in the UK and USA, Earth Surf. Processes Landforms, 25(8),
availability to constrain individual catchment components, 825–845.
both qualitative [e.g., McMillan and Clark, 2009] and Brazier, R. E., A. J. Parsons, J. Wainwright, D. M. Powell, and W. H. Schle-
quantitative (as implemented here), can improve the param- singer (2007), Upscaling understanding of nitrogen dynamics associated
with overland flow in a semi-arid environment, Biogeochemistry, 82(3),
eter inference procedure, leading to more robust models 265–278.
and more robust model predictions. Burnash, R. J. C. (1995), The NWS River Forecast System, in Computer
Models of Watershed Hydrology, edited by V. P. Singh, pp. 311–366,
Water Resour. Publ., Highlands Ranch, Colo.
7. Conclusion Chow, V. T. (1959), Open-Channel Hydraulics, McGraw-Hill, London.
[36] The objective of this paper was to investigate Cobby, D. M., D. C. Mason, and I. J. Davenport (2001), Image processing
whether lidar-derived data could lead to improved predic- of airborne scanning laser altimetry data for improved river flood model-
tion of flow events in ephemeral channels. In general, K-Li ling, ISPRS J. Photogramm. Remote Sensing, 56(2), 121–138.
Coes, A. L., and D. R. Pool (2005), Ephemeral-stream channel and basin-
showed improved performance in comparison to K-Tr, both floor infiltration and recharge in the Sierra Vista subwatershed of the upper
when each model was calibrated to individual events and San Pedro Basin, Southeastern Arizona, USGS Prof. Pap. 1703, 67 pp.,
during an evaluation phase when the models (and parameter U.S. Geological Survey, NW Washington, DC.
sets) were applied across events. Sensitivity analysis identi- Constantz, J., S. W. Tyler, and E. Kwicklis (2003), Temperature-profile
methods for estimating percolation rates in arid environments, Vadose
fied that the K-Li model also had greater consistency in be- Zone J., 2, 12–24.
havioral parameter sets across runoff events, with optimal Croft, H., K. Anderson, and N. Kuhn (2009), Characterizing soil surface
parameter values for the most sensitive parameters (satu- roughness using a combined structural and spectral approach, J. Soil Sci.,
rated infiltration and the roughness coefficient) occurring 60, 431–442.
for all events in a narrower region of parameter space. In Dahan, O., Y. Shani, Y. Enzel, Y. Yechieli, and A. Yakirevich (2007),
Direct measurements of floodwater infiltration into shallow alluvial aqui-
contrast, parameter interaction within K-Tr resulted in fers, J. Hydrol., 344, 157–170.
poorly constrained behavioral parameter sets across parame- El-Hames, A. S., and K. S. Richards (1998), An integrated, physically based
ter space. Interaction between Saturated Infiltration (Ks) and model for arid region flash flood prediction capable of simulating
the Woolhiser Coefficient (Wc), which has little physical dynamic transmission loss, Hydrol. Processes, 12(8), 1219–1232.
Ferguson, R. I. (2003), The missing dimension: Effects of lateral variation on
meaning, had a compensatory effect on model performance. 1-D calculations of fluvial bedload transport, Geomorphology, 56, 1–14.
Data used in this study allowed focus on a particular struc- Foody, G. M., E. M. Ghoneim, and N. W. Arnell (2004), Predicting loca-
tural element common in distributed catchment models. An tions sensitive to flash flooding in an and environment, J. Hydrol., 292
understanding of the channel model component, as devel- (1–4), 48–58.
oped here, has previously been dominated by uncertainty in Freer, J., K. Beven, and B. Ambroise (1996), Bayesian estimation of uncer-
tainty in runoff production and the value of data: An application of the
input conditions and other catchment components. These GLUE approach, Water Resour. Res., 32, 2161–2173.
results suggest that lidar derived cross-section morphology French, J. R. (2003), Airbourne Lidar in support of geomorphological and
can lead to improved, and more robust flash flood predic- hydraulic modelling, Earth Surf. Processes Landforms, 28, 321–335.
tion, particularly in distributed catchment models where the Garcia-Pintado, J., G. G. Barbera, M. Erena, and V. M. Castillo (2009), Cal-
ibration of structure in a distributed forecasting model for semiarid flash
channel component can dominate runoff response. flood: Surface storage and channel roughness, J. Hydrol., 377, 165–184.
Goodrich, D. C., L. J. Lane, R. M. Shillito, S. N. Miller, K. H. Syed, and
[37] Acknowledgments. The research was primarily funded by a Uni- D. A. Woolhiser (1997), Linearity of basin response as a function of scale
versity of Exeter Graduate Fellowship, awarded to the first author. Signifi- in a semiarid watershed, Water Resour. Res., 33(12), 2951–2965.
cant support was also provided by The USDA-ARS Southwest Watershed Green, A., and G. A. Ampt (1911), Studies on soil physics: 1. Flow of air
Research Center. and water through soils, J. Agic. Sci., 4, 1–24.
Hassan, M. A. (1990), Observations of desert flood bores, Earth Surf. Proc-
esses Landforms, 15(5), 481–485.
References Hilldale, R. C. (2007), Using bathymetric LiDAR and a 2-D hydraulic
Aggett, G. R., and J. P. Wilson (2009), Creating and coupling a high-resolu- model to quantify aquatic habitat, paper presented at Proceeding if the
tion DTM with a 1-D hydraulic model in a GIS for scenario-based assess- ASCE World Environmental and Water Resources Congress, Tampa,
ment of avulsion hazard in a gravel-bed river, Geomorphology, 113(1–2), Fla., 15–17 May.
21–34. Hoffman, J. P., M. A. Ripich, and K. M. Ellett (2002), Characteristics of
Al-Qurashi, A., N. McIntyre, H. Wheater, and C. Unkrich (2008), Applica- shallow deposits beneath Rillito Creek, Pima County, Arizona, U.S.
tion of the Kineros2 rainfall-runoff model to an arid catchment in Oman, Geol. Surv. Water Resour. Invest. Rep., 01-4257, 51 pp., U.S. Geological
J. Hydrol., 355, 91–105. Survey, NW Washington, DC.
Ashley, S. T., and W. S. Ashley (2008), Flood fatalities in the United States, Hooke, J. M., and J. M. Mant (2000), Geomorphological impacts of a flood
J. Appl. Meteorol. Climatol., 47, 805–818. event on ephemeral channels in SE Spain, Geomorphology, 34(3–4),
Bahat, Y., T. Grodek, J. Lekach, and E. Morin (2009), Rainfall-runoff mod- 163–180.
eling in a small hyper-arid catchment, J. Hydrol., 373(1–2), 204–217. Hornberger, G. M., and R. Spear (1981), An approach to the preliminary
Bates, P. D., K. J. Marks, and M. S. Horritt (2003), Optimal use of high- analysis of environmental systems, J. Environ. Manage., 12, 7–18.
resolution topographic data in flood inundation models, Hydrol. Proc- Hutton, C. J. (2010), Modelling watershed processes in semi-arid environ-
esses, 17, 537–557. ments, PhD dissertation, Coll. of Life and Environ. Sci., Univ. of Exeter,
Beven, K., and A. M. Binley (1992), The future of distributed models: Model U. K.
calibration and uncertainty estimation, Hydrol. Processes, 6, 279–298. Knighton, A. D., and G. C. Nanson (2001), An event based approach to the
Beven, K., and J. Freer (2001), A dynamic TOPMODEL, Hydrol. Proc- hydrology of arid zone rivers in the Channel Country of Australia,
esses, 15, 1993–2011. J. Hydrol., 254(1–4), 102–123.
Blasch, K. W., T. P. A. Ferre, J. P. Hoffmann, and J. B. Fleming (2006), Krueger, T., J. N. Quinton, J. M. Freer, J. A. Macleod, G. S. Bilotta, R. E.
Relative contributions of transient and steady state infiltration during Brazier, P. Butler, and P. M. Haygarth (2009), Uncertainties in data and
ephemeral streamflow, Water Resour. Res., 42, W08405, doi:10.1029/ models to describe dynamics of agricultural sediment and phosphorus
2005WR004049. transfer, J. Environ. Qual., 38, 1137–1148.

10 of 11
W04509 HUTTON ET AL.: FLOW ROUTING IN EPHEMERAL RIVERS W04509

Lane, S. N. (2005), Roughness—Time for re-evaluation?, Earth Surf. Proc- Parsons, A. J., J. Wainwright, R. E. Brazier, and D. M. Powell (2006), Is
esses Landforms, 30, 251–253. sediment delivery a fallacy?, Earth Surf. Processes Landforms, 31(10),
Lange, J. (2005), Dynamics of transmission losses in a large arid stream 1325–1328.
channel, J. Hydrol., 306, 112–126. Pelettier, J. D., and S. DeLong (2004), Oscillations in arid alluvial-channel
Matgen, P., G. Schumann, J. B. Henry, L. Hoffmann, and L. Pfister (2007), geometry, Geology, 32, 713–716.
Integration of SAR-derived river inundation areas, high-precision topo- Rawls, W. J., L. R. Ahuja, D. L. Brakensiek, and A. Shirmohammadi
graphic data and a river flow model toward near real-time flood manage- (1993), Infiltration and soil water movement, in Handbook of Hydrology,
ment, Int. J. Appl. Earth Obs., 9(3), 247–263. edited by D. R. Maidment, pp. 5.1–5.51, McGraw-Hill, New York.
McIntyre, N., and A. Al-Qurashi (2009), Performance of ten rainfall-runoff Rayburg, S., M. Thoms, and M. Neave (2009), A comparison of digital ele-
models applied to an arid catchment in Oman, Environ. Modell. Soft- vation models generated from different data sources, Geomorphology,
ware, 24, 726–738. 106, 261–270.
McIntyre, N., A. Al-Qurashi, and H. Wheater (2007), Regression analysis Refsgaard, J. C., J. P. van der Sluijs, J. Brown, and P. van der Keur (2006),
of rainfall-runoff data from an arid catchment in Oman, Hydrol. Sci. J., A framework for dealing with uncertainty due to model structure error,
52(6), 1103–1118. Adv. Water Resour., 29, 1586–1597.
McMillan, H., and M. Clark (2009), Rainfall-runoff model calibration using Renard, K. G., M. H. Nichols, D. A. Woolhiser, and H. B. Osborn (2008), A
informal likelihood measures within a Markov chain Monte Carlo sam- brief background on the U. S. Department of Agriculture Agricultural
pling scheme, Water Resour. Res., 45, W04418, doi:10.1029/2008 Research Service Walnut Gulch Experimental Watershed, Water Resour.
WR007288. Res., 44, W05S02, doi:10.1029/2006WR005691.
Michaud, J., and S. Sorooshian (1994), Comparison of simple versus com- Seager, R., et al. (2007), Model projections of an imminent transition to a
plex runoff models on a midsized semiarid watershed, Water Resour. more arid climate in southwestern North America, Science, 316(5828),
Res., 30(3), 593–605. 1181–1184.
Moran, M. S., et al. (2008), Preface to special section on fifty years of Sharma, K. D., and J. S. R. Murthy (1995), Hydrologic routing of flow in
research and data collection: U.S. Department of Agriculture Walnut arid ephemeral channels, J. Hydraul. Eng., 121(6), 466–471.
Gulch Experimental Watershed, Water Resour. Res., 44, W05S01, Smith, B. J., D. C. Goodrich, D. A. Woolhiser, and C. L. Unkrich (1995),
doi:10.1029/2007WR006083. KINEROS—A Kinematic runoff and erosion model, in Computer Mod-
Morin, E., D. C. Goodrich, R. A. Maddox, X. G. Gao, H. V. Gupta, and S. els of Watershed Hydrology, edited by V. P. Singh, pp. 697–732, Water
Sorooshian (2006), Spatial patterns in thunderstorm rainfall events and Resour. Publ., Highlands Ranch, Colo.
their coupling with watershed hydrological response, Adv. Water Wainwright, J., A. J. Parsons, E. N. Muller, R. E. Brazier, D. M. Powell,
Resour., 29(6), 843–860. and B. Fenti (2008), A transport-distance approach to scaling erosion
Morin, E., T. Grodek, O. Dahan, G. Benito, C. Kulls, Y. Jacoby, G. Van rates: 2. Sensitivity and evaluation of MAHLERAN, Earth Surf. Proc-
Langenhove, M. Seely, and Y. Enzel (2009), Flood routing and alluvial esses Landforms, 33(6), 962–984.
aquifer recharge along the ephemeral arid Kuiseb River, Namibia, J. Walters, M. O. (1990), Transmission losses in arid region, J. Hydraul. Eng.
Hydrol., 368, 262–275. ASCE, 116(1), 129–138.
Mudd, S. M. (2006), Investigation of the hydrodynamics of flash floods in Yatheendradas, S., T. Wagener, H. Gupta, C. Unkrich, D. Goodrich, M.
ephemeral channels: Scaling analysis and simulation using a shock- Schaffner, and A. Stewart (2008), Understanding uncertainty in distrib-
capturing flow model incorporating the effects of transmission losses, uted flash flood forecasting for semiarid regions, Water Resour. Res., 44,
J. Hydrol., 324(1–4), 65–79. W05S19, doi:10.1029/2007WR005940.
Nichols, M. H., and C. Shipek (2006), Channel response to changes in run-
off regime along the Walnut Gulch Channel, paper presented at Research
Insights into Semi-arid Ecosystems Third Annual Symposium, Univ. of R. E. Brazier and A. P. Nicholas, Geography, College of Life and Envi-
Ariz., Tucson, 7 Oct. ronmental Sciences, University of Exeter, Rennes Dr., EX4 4RJ Exeter, UK.
National Weather Service (2002), Advanced hydrologic prediction services C. J. Hutton, Centre for Water Systems, College of Engineering, Mathe-
concept of services and operations, report, NOAA, Nat. Weather Serv., matics and Physical Sciences, University of Exeter, Harrison Building,
Silver Spring, Md. North Park Road, EX4 4QF, Exeter, U.K. ([email protected])
Parissopoulos, G. A., and H. S. Wheater (1991), Effects of Wadi flood M. Nearing, Southwest Watershed Research Center, United States
hydrograph characteristics on infiltration, J. Hydrol., 126(3–4), 247–263. Department of Agriculture, 2000 East Allen Rd., Tucson, AZ 85719, USA.

11 of 11