DMIP 2 West Final 2013 - Journal of Hydrology

This article appeared in a journal published by Elsevier.
The attached
copy is furnished to the author for internal non-commercial research
and education use, including for instruction at the authors institution
and sharing with colleagues.
Other uses, including reproduction and distribution, or selling or
licensing copies, or posting to personal, institutional or third party
websites are prohibited.
In most cases authors are permitted to post their version of the
article (e.g. in Word or Tex form) to their personal website or
institutional repository. Authors requiring further information
regarding Elsevier’s archiving and manuscript policies are
encouraged to visit:
http://www.elsevier.com/authorsrights
Author's personal copy
Journal of Hydrology 507 (2013) 300–329
Contents lists available at ScienceDirect
Journal of Hydrology
journal homepage: www.elsevier.com/locate/jhydrol
The distributed model intercomparison project – Phase 2: Experiment

design and summary results of the western basin experiments
Michael Smith a,⇑, Victor Koren a, Ziya Zhang a,o, Fekadu Moreda b, Zhengtao Cui a, Brian Cosgrove a,
Naoki Mizukami a,c, David Kitzmiller a, Feng Ding a,d, Seann Reed a, Eric Anderson a, John Schaake a,
Yu Zhang a, Vazken Andréassian e, Charles Perrin e, Laurent Coron e, Audrey Valéry e,1, Behnaz Khakbaz f,
Soroosh Sorooshian g, Ali Behrangi h, Bisher Imam i, Kuo-Lin Hsu g, Ezio Todini j, Gabriele Coccia j,
Cinzia Mazzetti k, Enrique Ortiz Andres l, Félix Francés m, Ismael Orozco m, Robert Hartman n,
Arthur Henkel n, Peter Fickenscher n, Scott Staggs n
a
Office of Hydrologic Development, NOAA National Weather Service, 1325 East West Highway, Silver Spring, MD 20910, USA
b
Water and Ecosystems Management, Research Triangle Institute – International, 3040 Cornwallis Road, Research Triangle Park, NC, USA
c
Research Applications Laboratory, National Center for Atmospheric Research, P.O. Box 3000, Boulder, CO, USA
d
ADNET Systems, Inc., 6720B Rockledge Drive, Suite 504, Bethesda, MD 20817, USA
e
Irstea (former Cemagref) – UR HBAN, 1 Rue Pierre-Gilles de Gennes CS 10030, 92761 Antony Cedex, France
f
URS Corporation, 3500 Porsche Way, Ontario, CA 91764, USA
g
Center for Hydrometeorology and Remote Sensing (CHRS), Department of Civil and Environmental Engineering, University of California, Irvine, E-4130 Engineering Gateway, Irvine,
CA 92697, USA
h
Jet Propulsion Laboratory, California Institute of Technology, 4800 Oak Grove Dr., MS 233-300, Pasadena, CA 91109, USA
i
International Hydrological Programme (IHP), Division of Water Sciences, UNESCO, 1 Rue Miollis, 75732, Paris Cedex 15, France
j
Department of Earth and Geo-Environmental Sciences, University of Bologna, 1 Piazza di Porta S. Donato, 40126 Bologna, Italy
k
PROGEA s.r.l., Via Santo Stefano 6, 40125 Bologna, Italy
l
Idrologia & Ambiente s.r.l, Riviera di Chiaia 72, 80122 Napoli, Italy
m
Research Institute of Water Engineering and Environment, Universitat Politècnica de València, Camino de Vera s/n, 46022 Valencia, Spain
n
Visiting Scientist Programs, University Corporation for Atmospheric Research, P.O. Box 3000, Boulder, CO 80307, USA
a r t i c l e i n f o s u m m a r y
Article history: The Office of Hydrologic Development (OHD) of the U.S. National Oceanic and Atmospheric Administra-
Received 20 November 2012 tion’s (NOAA) National Weather Service (NWS) conducted the two phases of the Distributed Model Inter-
Received in revised form 25 July 2013 comparison Project (DMIP) as cost-effective studies to guide the transition to spatially distributed
Accepted 28 August 2013
hydrologic modeling for operational forecasting at NWS River Forecast Centers (RFCs). Phase 2 of the Dis-
Available online 4 September 2013
This manuscript was handled by
tributed Model Intercomparison Project (DMIP 2) was formulated primarily as a mechanism to help guide
Konstantine P. Georgakakos, Editor-in-Chief, the U.S. NWS as it expands its use of spatially distributed watershed models for operational river, flash
with the assistance of Baxter E. Vieux, flood, and water resources forecasting. The overall purpose of DMIP 2 was to test many distributed mod-
Associate Editor els forced by high quality operational data with a view towards meeting NWS operational forecasting
needs. At the same time, DMIP 2 was formulated as an experiment that could be leveraged by the broader
Keywords: scientific community as a platform for the testing, evaluation, and improvement of distributed models.
Hydrologic model DMIP 2 contained experiments in two regions: in the DMIP 1 Oklahoma basins, and second, in two
Rainfall-runoff basins in the Sierra Nevada Mountains in the western USA. This paper presents the overview and results
Distributed model of the DMIP 2 experiments conducted for the two Sierra Nevada basins. Simulations from five indepen-
Lumped model dent groups from France, Italy, Spain and the USA were analyzed. Experiments included comparison of
Calibration
lumped and distributed model streamflow simulations generated with uncalibrated and calibrated
Simulation
parameters, and simulations of snow water equivalent (SWE) at interior locations. As in other phases
of DMIP, the participant simulations were evaluated against observed hourly streamflow and SWE data
and compared with simulations provided by the NWS operational lumped model. A wide range of statis-
tical measures are used to evaluate model performance on a run-period and event basis. Differences
between uncalibrated and calibrated model simulations are assessed.
⇑ Corresponding author. Tel.: +1 301 713 0640x128; fax: +1 301 713 0963.
E-mail address: [email protected] (M. Smith).
1
Present address: EDF-DTG, 21, Avenue de l’Europe, 38040 Grenoble Cedex 9,
France.
0022-1694/$ - see front matter Published by Elsevier B.V.

http://dx.doi.org/10.1016/j.jhydrol.2013.08.040
M. Smith et al. / Journal of Hydrology 507 (2013) 300–329 301
Results indicate that in the two study basins, no single model performed best in all cases. In addi-
tion, no distributed model was able to consistently outperform the lumped model benchmark. How-
ever, one or more distributed models were able to outperform the lumped model benchmark in many
of the analyses. Several calibrated distributed models achieved higher correlation and lower bias than
the calibrated lumped benchmark in the calibration, validation, and combined periods. Evaluating a
number of specific precipitation-runoff events, one calibrated distributed model was able to perform
at a level equal to or better than the calibrated lumped model benchmark in terms of event-averaged
peak and runoff volume error. However, three distributed models were able to provide improved peak
timing compared to the lumped benchmark. Taken together, calibrated distributed models provided
specific improvements over the lumped benchmark in 24% of the model-basin pairs for peak flow,
12% of the model-basin pairs for event runoff volume, and 41% of the model-basin pairs for peak tim-
ing. Model calibration improved the performance statistics of nearly all models (lumped and distrib-
uted). Analysis of several precipitation/runoff events indicates that distributed models may more
accurately model the dynamics of the rain/snow line (and resulting hydrologic conditions) compared
to the lumped benchmark model. Analysis of SWE simulations shows that better results were
achieved at higher elevation observation sites.
Although the performance of distributed models was mixed compared to the lumped benchmark,
all calibrated models performed well compared to results in the DMIP 2 Oklahoma basins in terms
of run period correlation and %Bias, and event-averaged peak and runoff error. This finding is note-
worthy considering that these Sierra Nevada basins have complications such as orographically-
enhanced precipitation, snow accumulation and melt, rain on snow events, and highly variable topog-
raphy. Looking at these findings and those from the previous DMIP experiments, it is clear that at this
point in their evolution, distributed models have the potential to provide valuable information on spe-
cific flood events that could complement lumped model simulations.
Published by Elsevier B.V.
1. Introduction 1.2. Science questions
1.1. Overview DMIP 2 West was originally formulated to address several ma-
jor science questions (Smith et al., 2006). They are framed for the
The Office of Hydrologic Development (OHD) of the U.S. Na- interest of the broad scientific community with a corollary for
tional Oceanic and Atmospheric Administration’s (NOAA) National the NOAA/NWS. These science questions and issues are highly
Weather Service (NWS) led two phases of the Distributed Model intertwined but are listed separately here for clarity.
Intercomparison Project (DMIP) as cost-effective studies to guide
the transition into spatially distributed hydrologic modeling for 1.2.1. Distributed vs. lumped approaches in mountainous areas
operational forecasting (Smith et al., 2012a; Smith et al., 2004) at Can distributed hydrologic models provide increased stream-
NWS River Forecast Centers (RFCs). DMIP 1 focused on distributed flow simulation accuracy compared to lumped models in moun-
and lumped model intercomparisons in basins of the southern tainous areas? If so, under what conditions? Are improvements
Great Plains (Reed et al., 2004; Smith et al., 2004). DMIP 2 con- constrained by forcing data quality? This was one of the dominant
tained tests in two geographic regions: continued experiments in questions in DMIP 1 and the DMIP 2 experiments in Oklahoma.
the U.S. Southern Great Plains (Smith et al., 2012a,b) and tests in Smith et al. (2012a,b) and Reed et al. (2004) showed improvements
two mountainous basins in the Sierra Nevada Mountains, hereafter of deterministic distributed models compared to lumped models in
called DMIP 2 West. Since the conclusion of DMIP 1, the NWS has non-snow, generally uncomplicated basins. The specific question
used a distributed model for basin outlet forecasts (e.g., Jones et al., for the NOAA/NWS mission is: under what circumstances should
2009) as well as for generating gridded flash flood guidance over NOAA/NWS use distributed hydrologic models in addition to
large geographic domains (Schmidt et al., 2007). The purpose of lumped models to provide hydrologic services in mountainous
this paper is to present the DMIP 2 West experiments and results. areas? While many distributed models have been developed for
Advances in hydrologic modeling and forecasting are needed in
mountainous areas (e.g., Garen and Marks, 2005; Westrick et al.,
complex regions (e.g., Hartman, 2010; Westrick et al., 2002). 2002; Wigmosta et al., 1994), there remains a gap in our under-
Experiments are needed in the western USA and other areas where
standing of how much model complexity is warranted given data
the hydrology is dominated by complexities such as snow accumu- constraints, heterogeneity of physical characteristics, and model-
lation and melt, orographically-enhanced precipitation, steep and
ing goals (e.g., McDonnell et al., 2007). Several major snow model
other complex terrain features, and sparse observational networks. intercomparison efforts have been conducted in recent years such
The need for advanced models in mountainous regions is coupled
as Phases 1 and 2 of the Snow Model Intercomparison Project
with the requirements for more data in these areas. Advanced (SnowMIP; Rutter et al., 2009; Etchevers et al., 2004) and the
models cannot be implemented for operational forecasting without
Project for Intercomparison of Land Surface Process models (PILPS;
commensurate analyses of the data requirements in mountainous Slater et al., 2001). In addition, several comparisons of temperature
regimes.
index and energy budget snow models have been conducted (e.g.,
A major component of the NWS river forecast operations is the Debele et al., 2009; Franz et al., 2008a,b; Lei et al., 2007; Walter
national snow model run (NSM) by the NWS National Operational
et al., 2005; Fierz et al., 2003; Essery et al., 1999; WMO,
Hydrologic Remote Sensing Center (NOHRSC; Rutter et al., 2008; 1986a,b). Comprehensive studies such as the Cold Land Processes
Carroll et al., 2001). For over a decade, NOHRSC has executed the
Experiment (CLPX; Liston et al., 2008) have also been performed.
NSM in real time at an hourly, 1 km scale over the contiguous US However, to the best of our knowledge, there have been few
(CONUS) to produce a large number of gridded snow-related
specific tests of lumped and distributed modeling approaches in
variables. mountainous basins with a focus on improving river simulation
302 M. Smith et al. / Journal of Hydrology 507 (2013) 300–329
and forecasting. One such study was conducted by Braun et al. Bales et al., 2006; Garen and Marks, 2005; Walter et al., 2005;
(1994), who found that finer spatial modeling scales did not lead Simpson et al., 2004; Cline et al., 1998).
to performance gains. Addressing the data limitations in mountainous areas noted by
Garen and Marks (2005) and others (at least in terms of coverage),
1.2.2. Estimation of models inputs and model sensitivity to existing the number of radar, model-based, and satellite-derived products
data is rapidly growing. Efforts are ongoing to improve the ability of
What are the advantages and disadvantages associated with weather radars to observe precipitation in mountainous areas
distributed vs. lumped modeling in hydrologically complex areas (e.g., Kabeche et al., 2010; Gourley et al., 2009; Westrick et al.,
using existing NWS operational precipitation and temperature forc- 1999). Model-based data include the North American Regional
ing data? Current NWS RFC lumped streamflow models in moun- Reanalysis – (NARR, Mesinger et al., 2006), Rapid Update Cycle
tainous areas rely on networks of surface precipitation and (RUC; Benjamin et al., 2004), and Real Time Mesoscale Analysis
temperature gages to derive mean areal averages of point precipi- (RTMA; De Pondeca et al., 2011). Much work has gone into satellite
tation and temperature observations as input forcings for model estimates of precipitation in remote regions (e.g., Kuligowski et al.,
calibration and real time forecasting. The density of these networks 2013; Behrangi et al., 2009; Kuligowski, 2002). As these data sets
varies greatly, with most networks featuring sparse coverage at emerge and become more common, users are cautioned to avoid
high elevations. Even for lumped hydrologic modeling, there are the expectation that increased data resolution in new data sets will
uncertainties in the precipitation and temperature observations translate into increased data realism and accuracy (Guentchev et al.,
used by the NWS RFCs in mountainous areas (Hartman, 2010; Car- 2010; Daly, 2006).
penter and Georgakakos, 2001). Beyond network density issues,
there are problems with observation times, missing data, distribu- 1.2.4. Rain-snow partitioning
tion of multi-day precipitation accumulations, and other difficul- Can improvements to rain-snow partitioning be made? Parti-
ties. It is not known if these data uncertainties preclude the tioning between rainfall and snowfall plays a major role in deter-
application of distributed models, giving rise to the question: can mining both the timing and amount of runoff generation in high
the existing observational networks support operational distrib- altitude basins (Guan et al., 2010; White et al., 2010, 2002; Gourley
uted modeling? Nonetheless, some have attempted to apply dis- et al., 2009; Lundquist et al., 2008; Kienzle, 2008; McCabe et al.,
tributed models using such existing data (e.g., Shamir and 2007; Maurer and Mass, 2006; Westrick and Mass, 2001; Kim
Georgakakos, 2006). The intent in DMIP 2 West was to set up et al., 1998). The question for the NOAA/NWS is: can distributed
and run the models using Quantitative Precipitation Estimates models provide improved representation of the spatial variability
(QPE) derived from the relatively dense existing gage network. Fol- of rain/snow divisions?
low-on experiments would use QPE fields derived from reduced Traditionally, surface temperature observations have been used
networks to investigate the appropriate density for modeling. to determine the form of precipitation, although such data are not
the most reliable indicators of surface precipitation type (Minder
1.2.3. Model complexity and corresponding data requirements and Kingsmill, 2013; Minder et al., 2011; Maurer and Mass,
The NOAA/NWS corollary is: what can be improved over the 2006). Recently, as part of the western Hydrometeorologic Testbed
current lumped model approach used in the NWS for operational (HMT-West; Zamora et al., 2011; Ralph et al., 2005; hmt.noaa.gov),
river forecasting? Is there a dominant constraint that limits the instrumentation such as vertically pointing wind profilers and S-
performance of hydrologic simulation and forecasting in moun- Band radars have been used to detect freezing levels by locating
tainous areas? If so, is the major constraint the quality and/or the bright-band height (BBH, Minder and Kingsmill, 2013; White
amount of forcing data, or is the constraint related to a knowledge et al., 2010, 2002).
gap in our understanding of the hydrologic processes in these
areas? In other words, given the current level of new and emerging 1.2.5. Scale issues
data sets available to drive advanced distributed models, can What are the dominant hydrologic scales (if any) in mountain-
improvements be realized? Or, do we still not have data of suffi- ous area hydrology? Understanding the variations of snowpacks
cient quality in mountainous areas? Additionally, what data and the timing and volume of snowmelt that generate streamflow
requirements can be specified for the NOAA/NWS to realize simu- has grown in recent periods but is complicated by difficult scale is-
lation and forecasting improvements in mountainous areas? sues (e.g., Simpson et al., 2004). Blöschl (1999) describes three
There is a considerable range in the recent literature on the sub- scales related to snow: process, measurement (observational data),
jects of model complexity and corresponding data requirements and modeling scale. Process scale is the variability of a snow re-
for hydrologic modeling in mountainous areas. We provide a sam- lated variable. Measurement scale covers spacing, extent, and ‘sup-
ple here to indicate the range of issues and findings. port’ or area of integration related to an instrument. Modeling scale
For hydrologic models driven solely by precipitation and tem- describes the spatial unit to which the model equations are applied
perature, there are the issues of the gage density and location re- (e.g., grid cell size in a distributed model). Several studies have
quired to achieve a desired simulation accuracy (e.g., Guan et al., investigated the impacts of modeling scale (e.g., Merz et al.,
2010; Tsintikidis et al., 2002; Reynolds and Dennis, 1986). The gage 2009; Leydecker et al., 2001; Cline et al., 1998). However, to the
density issue also affects merged-data precipitation estimates (e.g., best of our knowledge, there is scant literature on modeling scales
satellite–radar–gage) because the gage information is very often that jointly considers snow and runoff processes. One exception is
used to adjust the other (radar or satellite) observations (e.g., Nel- the work of Dornes et al. (2008), who found a spatially distributed
son et al., 2010; Boushaki et al., 2009; Guirguis and Avissar, 2008; approach provided better late season ablation rates and runoff
Young et al., 2000). hydrographs than a spatially aggregated model.
As new precipitation/runoff models with increased complexity For forecasting agencies like NOAA/NWS, the scale question can
are developed, commensurate observations of the forcing variables be restated as: is there an appropriate operational modeling scale in
are also required (e.g., Dong et al., 2005; Fierz et al., 2003; Waichler mountainous areas that captures the essential rain/snow/runoff
and Wigmosta, 2003). Numerous studies have identified the diffi- processes and provides adequate information for forecasting,
culties in collecting the data required to match model complexity water resources management, and decision support? For example,
(Rice and Bales, 2010; Valéry et al., 2009; Feng et al., 2008; Daly can the 4 km grid scale used in the non-mountainous DMIP 1 and 2
et al., 2008; Franz et al., 2008a; Shamir and Georgakakos, 2006; test basins be used instead of the current elevation zones for
operational forecasting? Or is this 4 km scale too coarse to capture parameter estimation, and calibration. It was not the intent of
the large terrain variations and resultant hydrometeorological im- DMIP 2 West to diagnose simulation improvements from specific
pacts on modeling? Shamir and Georgakakos (2006) used a 1 km model structures but rather to examine the performance of the
grid modeling scale in the American basin, but concluded that sig- models as applied by the participants.
nificant improvement in simulation quality would result by better Focusing on precipitation-runoff models, the University of Bolo-
representations of the spatial variability of precipitation and tem- gna (UOB) used the TOPKAPI model (Coccia et al., 2009), which is
perature, especially at the lower elevations of the snowpack. Some based on the idea of combining kinematic routing with a topo-
have commented on the difficulty, or even impossibility, of finding graphic representation of the basin. Three non-linear reservoirs
an optimum element size that effectively comprises measurement, are used to generate subsurface, overland, and channel flow. TOPK-
process, and modeling scales (Dornes et al., 2008; Blöschl, 1999). API includes components that represent infiltration, percolation,
DMIP 2 West intended to examine simulation performance vs. evapotranspiration, and snowmelt. The NWS Office of Hydrologic
modeling scale to infer appropriate model spatial resolution. Development (OHD) used the HL-RDHM model (Koren et al.,
2006; 2004). HL-RDHM uses the Sacramento Soil Moisture
Accounting model (SAC-SMA; Burnash, 1995) applied to grid cells.
1.2.6. Internal consistency of distributed models
Kinematic wave equations are used to route runoff over hillslopes
Another question posed in DMIP 2 West: ‘Can distributed mod-
and through the channel system. The University of California at Ir-
els reproduce processes at interior locations (points upstream of
vine (UCI) also used the SAC-SMA model but applied it to sub-ba-
basin outlet gages) in mountainous areas?’ Inherent in this ques-
sins. Kinematic wave routing was used for channel routing of the
tion is the ability of distributed models to simulate (and therefore
runoff volumes. The Technical University of Valencia (UPV) used
hopefully forecast) hydrologic variables such as SWE, soil moisture,
the TETIS model (Vélez et al., 2009; Francés et al., 2007). TETIS is
and streamflow at points other than those for which observed
a 6-layer conceptual model linked to a kinematic channel routing
streamflow data exist. Successful simulation of such variables at
module. The GR4J model (Perrin et al., 2003) was used by CEMAG-
interior points supports the idea that the models achieve the right
REF (CEM). GR4J is a parsimonious 4-parameter lumped model. For
answer (i.e., basin outlet streamflow simulations) for the right rea-
DMIP 2 West, it was applied to 5 elevation zones.
son, i.e., because they are correctly modeling processes in the basin
interior (Kirchner, 2006).
The remainder of this paper is organized as follows. Section 2 2.2. Benchmarks and performance evaluation
presents the methodology of the DMIP 2 West experiments,
including data derivation and an overview of the modeling instruc- Two benchmarks (e.g., Seibert, 2001; Perrin et al., 2006) were
tions. The results and discussion are presented in Section 3. We used to assess model performance. Observed hourly streamflow
present our conclusions in Section 4, while recommendations for data from the U.S. Geological Survey (USGS) were used as ‘truth.’
future work are offered in Section 5. Simulations from the NWS operational precipitation/runoff model
(hereafter referred to as the lumped (LMP) benchmark) were used
as the second benchmark. This model was selected to address the
2. Methodology science question regarding the improvement of distributed models
compared to lumped models. In addition, the LMP model was cho-
2.1. Participating institutions, models, and submissions sen for consistency with the DMIP 1 and DMIP 2 Oklahoma exper-
iments (Smith et al., 2012b; Reed et al., 2004).
Five groups submitted simulations for analysis. As with other The LMP model actually consists of several NWS components
phases of DMIP, the level of participation varied. Some participants linked together. The NWS Snow-17 model (Anderson, 2006,
submitted all requested simulations, while others submitted only a 1976) is used to model snow accumulation and melt. Rain and melt
subset. Table 1 lists the simulations submitted by each DMIP 2 water from the Snow-17 model is input to the SAC-SMA model.
West participant. Runoff volumes are transformed into discharge using unit hydro-
The models used in DMIP 2 West feature a range of model struc- graphs. This model combination is typically applied over two ele-
tures and approaches to hydrological modeling. Appendix A pre- vations zones above and below the typical rain/snow elevation.
sents information on the models used in DMIP 2 West, as well as Flow from the upper elevation zone is routed through the lower
references that describe the models in more detail. The greatest zone with a lag/k method. Unit hydrographs are used to convert
differences amongst the models seem to be in the precipitation- runoff to discharge in upper and lower zones. Mean areal precipi-
runoff approaches. The snow models all used relationships based tation (MAP) and temperature (MAT) time series for the elevation
on temperature rather than the full energy budget equations. zones were defined from the gridded values on the DMIP 2 ftp site.
Threshold temperatures are used to partition rain and snow. It More information on elevation zone modeling can be found in
should be kept in mind that the results herein reflect the appropri- Anderson (2002) and Smith et al. (2003). In the North Fork Amer-
ateness of model structure and other factors such as user expertise, ican River basin, an elevation of 1,524 m was used to divide the
Table 1
Participants’ submissions. An ‘x’ indicates that a simulation was submitted for analysis.
Participant and DMIP 2 Acronym Modeling Instruction 1 Modeling Instruction 2 Modeling Instruction 3 Markleeville, CA
North Fork American River Outlet at Interior flow point
Gardnerville, NV at Markleeville, CA
Flow SWE (2 sites) Flow Flow Flow SWE (4 sites)
NWS/OHD Distributed (OHD) USA x x x x x x
NWS/OHD Lumped (LMP) USA x x x x x x
U. Bologna (UOB) Italy x x x x x x
CEMAGREF (CEM) France x x x x x x
U. Ca at Irvine (UCI) USA – – – – x x
Technical U. Valencia (UPV) Spain x x x x x x
upper and lower elevation zones for the LMP model. The upper ba- Statistics are computed over two types of time intervals. The
sin comprises 37% of the total basin area, while the lower basin term ‘‘overall’’ refers to multi-year run periods such as the calibra-
comprises 65%. In the East Fork of the Carson River we used an ele- tion, validation, and combined calibration/validation periods.
vation value of 2134 m to separate the upper and lower zones. Event statistics are computed for specific precipitation/runoff
These elevation values agree with the configuration of the NWS events.
California-Nevada River Forecast Center (CNRFC) operational
models. 2.4. Study basins
The LMP model also included the ability to compute the eleva-
tion of the rain/snow line using the MAT data and typical lapse 2.4.1. Description
rates. This rain/snow elevation is then used in conjunction with Two sub-basins in the American and Carson River watersheds
the area-elevation curve for the basin in Snow-17 to determine located near the border of California (CA) and Nevada (NV), USA,
how much of the basin receives rain vs. snow (Anderson, 2006). were selected as test basins (Fig. 1). Although these basins are geo-
A major consideration in any model evaluation or intercompar- graphically close, their hydrologic regimes are quite different due
ison is the question: what constitutes a ‘‘good’’ simulation or an to their mean elevation and location on either side of the Sierra Ne-
acceptable level of simulation accuracy (Seibert, 2001). This is vada divide (Simpson et al., 2004). The Carson River basin is a high-
the subject of much discussion (e.g., Bennett et al., 2013; Ritter altitude basin with a snow dominated regime, while the American
and Muñoz-Carpena, 2013; Puspalatha et al., 2012; Ewen, 2011; River drains an area that is lower in elevation with precipitation
Confalonieri et al., 2010; Andréassian et al., 2009; Clarke, 2008; falling as rain and mixed snow and rain (Jeton et al., 1996). These
Gupta et al., 2008; Moriasi et al., 2007; Schaefli and Gupta, 2007; two basins were selected to represent the general hydrologic re-
Shamir and Georgakakos, 2006; Krause et al., 2005; Seibert, gimes of western mountainous areas with the hope that our mod-
2001). These references and others indicate that there is not yet eling results would be relevant to other mountainous areas. Table 2
an agreed-upon set of goodness-of-fit indicators for hydrologic presents a summary of the characteristics of the American and Car-
model evaluation. Moreover, it has been difficult to specify ranges son River basins.
of values of the goodness-of-fit indicators that determine whether In the American River basin, we selected the North Fork sub-ba-
a model simulation is acceptable, good, or very good, although sug- sin above the USGS gage 11427000 shown in Fig. 2. This gage is at
gested ranges have recently emerged (Ritter and Muñoz-Carpena, the North Fork dam forming Lake Clementine. Hereafter, this basin
2013; Moriasi et al., 2007; Smith et al., 2003). One potential cause is referred to as NFDC1, using the NWS CNRFC basin acronym. This
of these difficulties is that one must consider the quality of the in-
put data when judging simulation results. What is a ‘‘poor’’ simu-
lation for a basin with excellent input data may be considered good
for a basin having poor quality input data (Ritter and Muñoz-Carp-
121 120 119 118
ena, 2013; Moriasi et al., 2007; Seibert, 2001). As a result, the inter-
pretation of goodness-of-fit indices continues to be a subjective
process. California Nevada
40 40
With this in mind, and consistent with DMIP 1 and 2 (Reed Carson
Sink
et al., 2004; Smith et al., 2012b) we use a number of performance
criteria computed over different time periods to evaluate the sim-
ulations compared to the benchmarks. These include measures of
hydrograph shape (modified correlation coefficient rmod; McCuen North Fork Lake Carson River
American River Tahoe
and Snyder, 1975) and volume (%Bias), water balance partitioning, Basin
cumulative runoff error, and specific indices to measure the
improvement compared to the LMP benchmark. Appendix D pre- 39 American 39
sents the statistical measures used herein and target ranges of River
Basin East Fork
these measures are given in the discussion below. In addition, we Carson River
relate our results to those achieved in the relatively simple non-
snow DMIP Oklahoma basins (Smith et al., 2012b; Reed et al.,
2004). 0 50 km
2.3. Definitions
For consistency with the results of the DMIP 1 and 2 experi-

ments in Oklahoma, we adopt the definition of Reed et al. (2004)
that a distributed model is one that (1) explicitly accounts for spa-
tial variability of meteorological forcings and basin physical char-
acteristics and (2) has the ability to produce simulations at
interior points without explicit calibration at those points. Inter-
ested readers are referred to Kampf and Burges (2007) for a de-
tailed discussion of definitions and classifications regarding
distributed hydrologic models.
A parent basin is defined as a watershed for which explicit cal-
ibration can be performed using basin outlet observed streamflow
data. In our experiments, these parent basins represent the typical
watershed sizes for which forecasts are generated by the NWS
RFCs. Interior points are locations within the parent basins where
simulations are generated without explicit calibration (hereafter Fig. 1. Location of the American and Carson River Basins in California and Nevada,
also referred to as ‘blind’ simulations). USA (top) and in relation to North America (circle, bottom).
Table 2
Characteristics of the two study basins.
East Fork Carson River above Gardnerville, NV North Fork American River above North Fork Dam
2
Area 922 km 886 km2
Median altitude 2417 m 1270 m
Annual rainfall 560–1244 mm 813–1651 mm
Min, max temperature 0, 14 (°C) 3, 18 (°C)
Precipitation form Mostly snow Snow and rain
Aspect Leeward Windward
Soil Shallow sandy and clay soil Clay loams and coarse sandy loans
Geology Volcanic rock and granodiorite Meta sedimentary rock and granodiorite
Vegetation Rangeland in lower altitude and conifer forests in upper altitude Pine–oak woodlands, shrub rangeland, ponderosa pine forest,
and subalpine forest
USGS gauge 1030900 near Gardnerville, NV 11427000 at North Fork Dam
Ca
C
Neva
Ne
alifforrniia
ad
da
1,524 m
NFDC1 contour
North Fork Dam
GRDN2
Gardnerville
USGS Stream Gauge CMEC1

NCDC T Markleeville
NCDC Hourly P
NCDC
C C Daily
yP 2,134 m
SNOTEL P T
P,T contour
Bureau Reclamation P 0 33 km
Fig. 2. Location of precipitation (P), temperature (T), and streamflow gaging stations. Also shown are the 1524 m and 2134 m elevation contours which separate the upper
and lower elevation zones.
basin is 886 km2 in area and rests on the western, windward side at Markleeville, CA. The Carson River terminates in the Carson Sink.
of the Sierra Nevada crest. Precipitation is dominated by oro- The East Fork Carson River generally flows from south to north so
graphic effects, with mean annual precipitation varying from that its average slope is not as steep as it could be if it were to face
813 mm at Auburn (elev. 393 m above msl) to 1651 mm at Blue directly east–west. GRDN2 is a high altitude basin, with a drainage
Canyon (elev. 1676 m above msl) (Jeton et al., 1996). Precipitation area of 714 km2 above USGS stream gage 10-308200 near Marklee-
occurs as a mixture of rain events and rain–snow events. The basin ville, CA and 922 km2 above USGS stream gage 10–309000 at
mean annual precipitation is 1532 mm and the annual runoff is Gardnerville, NV. Elevations in the GRDN2 basin range from
851 mm (Lettenmaier and Gan, 1990). Streamflow is about two- 1650 m near Markleeville to about 3400 m at the basin divide.
thirds wintertime rainfall and snowmelt runoff and less than Mean annual precipitation varies from 559 mm at Woodfords
one-third springtime snowmelt runoff (Dettinger et al., 2004). (elev. 1722 m) to 1,244 mm near Twin Lakes (elev. 2438 m).
The basin is highly forested and varies from pine-oak woodlands, Fig. 3 shows the rugged, heavily forested terrain in both basins.
to shrub rangeland, to ponderosa pine, and finally to sub-alpine
forest as one moves up in elevation. Much of the forested area is
secondary-growth due to the extensive timber harvesting con- 2.4.2. Rationale for basin selection
ducted to support the mining industry in the late 1800s (Jeton Several factors underscore the selection of the NFDC1 and
et al., 1996). Soils in the basin are predominately clay loams and GRDN2 basins for use in DMIP 2 West. First, both basins are largely
coarse sandy loams. The geology of the basin includes metasedi- unregulated (Georgakakos et al., 2010; Simpson et al., 2004; Dett-
mentary rocks and granodiorite (Jeton et al., 1996). inger et al., 2004; Carpenter and Georgakakos, 2001; Jeton et al.,
In the Carson River basin, the East Fork sub-basin shown in 1996). Second, these basins are geographically close, yet they pres-
Fig. 2 was selected for DMIP 2 West. Hereafter, the CNRFC identifier ent an opportunity to study different hydrologic regimes. More-
GRDN2 is used for the basin above the gage at Gardnerville, NV, over, their proximity allowed for more expedient data processing
and CMEC1 is used for the interior basin above the stream gage by DMIP 2 West organizers and participants.
Fig. 3. Terrain and landcover in the study basins. North Fork of the American River looking downstream from near the Iowa Hill Bridge (upper left). North Fork of the
American River looking southeast from the airport at Blue Canyon (upper right; photo courtesy of Brooks Martner). Forested conditions at the Blue Lakes SNOTEL site in the
East Fork Carson River basin (bottom).
Third, at the time of the design of DMIP 2 West, these basins had 2.5. Calibration
already been studied by numerous researchers. For example, these
basins have been used for parameter estimation experiments Participants were free to calibrate their models using strategies
(Leavesley et al., 2003; Jeton and Smith, 1993), climate change and statistical measures of their choice as this process is usually
studies (Lettenmaier and Gan, 1990; Jeton et al., 1996; Carpenter model-dependent. This provision was also an aspect of the Okla-
and Georgakakos, 2001), and mountainous area hydrometeorology homa experiments in DMIP 1 and 2 (Smith et al., 2012a,b; Reed
investigations (Wayand et al., 2013; Simpson et al., 2004; Lund- et al., 2004) and is similar to the Model Parameter Estimation
quist and Cayan, 2002; Cayan and Riddle, 1993; Kim et al., 1998). Experiment (MOPEX; Duan et al., 2006) and SnowMIP-2 (Rutter
Several authors used these basins to study the dynamics of the pre- et al., 2009). Appendix B presents a brief description of the strate-
cipitation generation process in mountainous areas (Wang and gies followed by the participants to calibrate their models. Three
Georgakakos, 2005; Dettinger et al., 2004; Hay and Clark, 2003; models (OHD, UOB, and UPV) used spatially variable a priori
Tsintikidis et al., 2002; Pandey et al., 1999, 2000; Reynolds and parameters and adjusted the parameter grids uniformly using sca-
Dennis, 1986). Data sets from the HMT-West experiments in lar factors. CEM and UCI maintained the spatially constant param-
NFDC1 have been used to partition rain and snow (e.g., Minder eters in each computational area.
and Kingsmill, 2013; Mizukami et al., 2013; White et al., 2010;
Lundquist et al., 2008), improve radar estimates of QPE in the 2.6. Run periods
mountains (e.g., Gourley et al., 2009; Matrosov et al., 2007), iden-
tify land-falling atmospheric rivers (e.g., Ralph and Dettinger, Specific periods were prescribed for model calibration and
2011; Kingsmill et al., 2006), enhance physics in numerical weath- validation. An initial one-year ‘warm up’ ‘or ‘spin-up’ period was
er models for improved QPF (e.g., Jankov et al., 2007), and under- provided to allow models to equilibrate after a complete annual
stand soil moisture movement in mountainous soils (Zamora wetting/drying cycle. Table 3 presents the computational periods.
et al., 2011). The warm-up and calibration periods for the East Fork of the
Table 3
Time spans of the ‘warm up’, calibration, and validation periods.
Basin ‘Warm Up’ Calibration Validation

North Fork American R. at October 1, 1987 to September 30, 1988 October 1, 1988 to September 30, 1997 October 1, 1997 to September 30, 2006
L. Clementine (NFDC1)
East Fork Carson R. at Markleeville, October 1, 1989 to September 30, 1990 October 1, 1990 to September 30, 1997 October 1, 1997 to September 30, 2006
CA (CMEC1)
East Fork Carson R. at Gardnerville, October 1, 1989 to September 30, 1990 October 1, 1990 to September 30, 1997 October 1, 1997 to September 30, 2006
NV (GRDN2)
Carson and North Fork of the American were slightly different due NRC, 1998) available from the National Climatic Data Center
to the availability of the observed hourly USGS streamflow data. (NCDC). Also used were daily precipitation observations from the
SNOTEL network. Problems with the COOP and SNOTEL data are
2.7. Simulations and modeling instructions well known and include difficulties in distributing multi-day accu-
mulations (Eischeid et al., 2000), data-entry, receiving, and refor-
Participants followed specific modeling instructions to generate matting errors (Reek et al., 1992), observer errors (e.g., NRC,
simulations of streamflow and SWE in order to address the science 1998), and dealing with varying daily station observation times
questions (See http://www.nws.noaa.gov/oh/hrl/dmip/2/docs/ (e.g., Hay et al., 1998).
sn_modeling_instructions.pdf and Table 1). Modeling Instruction Hourly and daily stations from the COOP and SNOTEL networks
1 was for NFDC1. Participants generated hourly simulated stream- were selected inside and near the basins. After initial screening for
flow at the basin outlet gage using calibrated and uncalibrated period-of-record (at least 5–10 years) and windward-leeward ef-
model parameters. There were no interior ‘‘blind’’ streamflow fects, 41 stations for NFDC1 and 69 for GRDN2 were selected for
gages in NFDC1. During the same run to generate the outlet further analysis. The location of the precipitation and temperature
streamflow hydrographs, participants also generated simulations stations is shown in Fig. 2.
of snow water equivalent at two locations where snow pillows
are operated by the U.S. Bureau of Reclamation (USBR): Blue Can- 2.9.1.2. Generation of gridded QPE. Considerable effort was ex-
yon and Huysink. Modeling Instruction 2 focused on the GRDN2 pended to generate a multi-year, hourly, 4 km gridded QPE data
basin. Participants generated uncalibrated and calibrated stream- set (Smith et al., 2010). One goal in the development of the QPE
flow simulations at the GRDN2 outlet. During the same run to gen- data was to generate spatially varying gridded precipitation data
erate GRDN2 outlet simulations, participants generated based on the same types of in situ point measurements currently
streamflow simulations at the interior USGS gage at Markleeville, used by NWS RFCs for lumped model calibration and real time
CA. The gage at Markleeville was considered a blind simulation operational forecasting. In this way we could address the science
point for this test with no explicit calibration. Hereafter, we refer question: can distributed models be operationally implemented
to this test as CMEC1-2. In addition, the basin above the USGS gage with currently available precipitation and temperature data?
at Markleeville, CA was also treated as an independent basin (Mod- An initial QPE data set for 1987–2002 was used to launch the
eling Instruction 3). This test is referred to as CMEC1-3. As part of DMIP 2 West experiments (Moreda et al., 2006), but was found to
Modeling Instruction 3, participants also generated SWE simula- contain a large inconsistency when extended from 2002 to 2006
tions for four U.S. Natural Resources Conservation Service (NRCS) (Mizukami and Smith, 2012; Smith et al., 2009). Developing an
snowpack telemetry (SNOTEL) sites (Serreze et al., 1999). Explicit alternative procedure and QPE data set delayed DMIP 2 West exper-
instructions for modeling scale were not provided. Rather, it was iments by nearly two years. Finally, an approach was developed
hoped that the participants’ models would inherently represent a using modified NWS procedures as shown in Fig. 4. The method
sufficiently broad range of modeling scales from which to make consists of three major steps: (1) data quality control and genera-
inferences on appropriate model scale. tion of hourly point precipitation time series, (2) spatial interpola-
2.8. Data assimilation
Assimilation of observed streamflow or other data to adjust Methodology for Hourly Gauge-Only Gridded QPE
model states was not allowed in DMIP 2 West. The simulations
Data Analysis
were generated by running models continuously over the warm- - Check data consistency –double mass analysis
up, calibration, and validation periods.
NCDC - Generate monthly station means
Hourly & Daily - Estimate missing data using station means
- Disaggregate all daily data to hourly values
NRCS SNOTEL - Non-disaggregated daily obs put into one hour
2.9. Data Daily - Manual QC: Fix ‘non-disaggregated’ values
- Uniformly distribute remaining daily values
Participants were required to use the hourly precipitation and
temperature grids posted on the DMIP 2 West web site. This grid- Generate 4-km QPE Grids
ded data was derived from the same types of in situ measurements Hourly Point - Use NWS Multi-Sensor Precip. Estimator (MPE)
used by RFCs to construct MAP and MAT time series for model
Time Series -‘Gauge-only’ option
-Uses PRISM monthly climatology grids
calibration and operational forecasting. As in the other phases of -Uses single optimal estimation (Seo et al.,
1998, J. Hydrology)
DMIP, basic forms of many other physical basin features and
meteorological variables were provided to promote participation.
2.9.1. Precipitation Water Balance Analysis

2.9.1.1. Precipitation data sources. DMIP 2 West used precipitation
data collected by the NWS Cooperative Observer Network (COOP; Fig. 4. Methodology for generating gage-only gridded QPE for DMIP 2 West.
tion of the point time series to a 4 km grid, and (3) water balance 0.8 ABRFC basins NFprzPclbPED
CARSnewPprioPED NFnewPclbPED
analyses. Details of these steps can be found on the DMIP 2 West DMIP2 OK basins MAR_newPprioPED
Walker River CARSoldPprioPED
web site: http://www.nws.noaa.gov/oh/hrl/dmip/2/wb_precip.html. MAR_oldPprioPED NFprzmPprioPED
NFnewPprioPED Yuba River
NFnewPRZM MAR_newPRZM
0.6 GARnewPRZM GAR_Jeton_1996
Step 1: Data quality control and generation of point hourly data
time series. East Fork Carson
River North Fork
Q/PET
American
0.4 River
The goal of this step was to generate quality-controlled, serially
complete, hourly precipitation time series at all the gage locations,
including daily stations. The term serially complete means that
precipitation was estimated for each station over the entire analy- 0.2
sis period of 1987–2006 (no missing values; Eischeid et al., 2000).
The hourly and daily data were quality controlled (QC) using stan-
dard NWS procedures described in Smith et al. (2003) and Ander- 0
son (2002). Double mass analysis was used to identify and correct 0.25 0.5 0.75 1 1.25 1.5 1.75
for human influences. Missing precipitation data were estimated P/PET
using weighted observations from the closest station in each of
Fig. 5. Long term mean annual values of Q/PET and P/PET for the western basins.
four quadrants. To account for orographic influences, long-term ra-
Also shown are basins in several climatic regions of the USA. The black diamonds
tios of monthly station means were used to condition estimates of are basins in the U.S. Southern Great Plains, of which the red circles are a subset
missing data by the ratio of long term monthly mean of the estima- used in DMIP 1 and 2 Oklahoma experiments.
tor station to the current station.
Multi-day gage accumulations were distributed over the pre-
ceding flagged days using daily amounts from surrounding sta-
tions. Daily precipitation accumulations were subsequently variables of PET and precipitation for the DMIP 2 West basins agree
distributed to hourly values using the distributions of hourly sta- with the trend established by the other basins, indicating that the
tions surrounding each daily station. A great deal of effort was ex- mean annual precipitation values are reasonable.
pended to correct problems associated with distributing daily
observations (Smith et al., 2010). In particular, two types of prob- 2.9.2. Temperature
lems were addressed: (1) distributing multi-day accumulations 2.9.2.1. Temperature data sources. Hourly 4 km gridded tempera-
to each individual day (e.g., Eischeid et al., 2000) and (2) distribut- ture values were derived using daily maximum and minimum
ing daily totals into hourly values. The precipitation records from (hereafter tmax and tmin) temperature data available from the
1988 to 2006 were examined month by month to identify and cor- NWS COOP and SNOTEL networks shown in Fig. 2. The underlying
rect such errors. interpolation procedure uses an inverse distance weighting algo-
rithm. It also uses PRISM gridded monthly climatological analyses
Step 2. Generation of gridded QPE. of daily maximum and minimum temperature (Daly et al., 1994).
The procedure has the following major steps:
A major task is the estimation of precipitation at ungaged grid-
ded points (Tsintikidis et al., 2002). The multi-sensor precipitation
2.9.2.2. Generation of gridded hourly temperature.
estimation (MPE: Seo, 1998) algorithm was used to spatially dis-
Step 1. Generate gridded daily tmax and tmin.
tribute the point hourly time series onto the DMIP 2 West 4 km
Hydrologic Rainfall Analysis Project (HRAP; Green and Hudlow, Observation times estimated for COOP stations with missing
1982; Reed and Maidment, 1999) grid. Data selected from 27 sta- temperature observation times were assumed to be the same as
tions in and around NFDC1 and 62 stations for GRDN2 were used to for the corresponding daily precipitation observations. The proce-
derive the final gridded QPE data sets. dure used to estimate missing observation times for these stations
MPE uses PRISM data to adjust the interpolation of point precip- is documented in Schaake et al. (2006).
itation to grids. The 800 m resolution PRISM monthly climatologi- A daily temperature processor was used to generate daily tmax
cal precipitation data derived for 1971–2000 (Daly et al., 2008) was and tmin grids for each day of the analysis period. Complex terrain
selected to be consistent with the river forecasting operations at in the DMIP 2 West study area generates spatial variations in tem-
CNRFC (Hartman, 2010). The DMIP 2 West QPE data did not include perature that are comparable to diurnal temperature variations. As
any corrections for gage under-catch (Yang et al., 1998a,b). Treat- a result, simple spatial interpolation of gage observations to grid
ment of gage under-catch is usually model-dependent so partici- locations did not by itself account for the complexity of the actual
pants were free to make adjustments as they chose. spatial variations. Therefore, monthly PRISM climatological grids of
daily mean tmax and tmin were used as part of the interpolation
Step 3. Water balance analysis. process.
The spatial interpolation procedure for daily maximum and
A check of the DMIP 2 West precipitation is shown in Fig. 5. This minimum temperature analysis is as follows. COOP and SNOTEL
figure presents a Budyko-type plot (Budyko, 1974) of water bal- sites to be used for the given day were selected having at least
ance components for a number of basins across the US, represent- 5 years of observations with less than 15% missing data. At each
ing a broad range of basin climatologies. On the abscissa is plotted site, the algorithm computes the difference between the observed
the ratio of observed long-term mean precipitation (Pobs) to poten- gage value for the given day and the monthly PRISM climatological
tial evapotranspiration (PET) while on the ordinate we plot the ra- mean values of tmax and tmin. These differences were interpolated
tio of observed streamflow (Q) to PET. PET was computed from the to the HRAP grid used by DMIP 2 West using an inverse distance
NOAA Evaporation Atlas (Farnsworth et al., 1982). The basins in the weighting interpolation procedure with the distance exponent
domain of the NWS Arkansas-Red Basin RFC (ABRFC) were taken equal to 1.0. Difference values for each of the nearest 2 gages in
from the work of Koren et al. (2006) and range in size from each of 4 quadrants surrounding each HRAP grid point are used.
20 km2 to 15,000 km2. This figure shows that the climatological For each HRAP grid point, the PRISM mean value was added to
the analyzed difference value to get the grid point value of the ana- were also provided at a 400 m resolution. Moreover, we also de-
lyzed maximum or minimum daily air temperature. fined flow directions for coarse resolution model cells that
matched or aligned with the grid of available radar-based forcing
Step 2. Generate hourly temperature grids. data (the HRAP grid) using the algorithm described by Reed (2003).
The hourly temperature processor uses the daily tmax and tmin
2.9.8. Vegetation and land use data
grids to generate hourly temperature grids for each hour of each
DMIP 2 West provided a 1 km gridded vegetation/land use data-
day. This procedure uses the Parton and Logan (1981) algorithm
set covering both basins. These data were originally developed by
to estimate hourly temperatures from daily max–min values. As
Hansen et al. (2000). Thirteen classes of vegetation were defined in
a check of the procedure, gridded climatologies of daily tmax and
these data.
tmin for January for the period 1961–1990, over an area including
the American and Carson River basins were generated (not shown).
The gage analysis and PRISM climatologies were nearly identical. 2.9.9. Hourly observed streamflow data
Instantaneous hourly flow data were acquired from the USGS.
2.9.3. Form of precipitation The data included some corrections and shifts but were defined
Participants were free to determine the form of precipitation as provisional. Unlike the mean daily flow data available from
(rain or snow) for each time step in their modeling. The procedures the USGS National Water Information Service (NWIS) web site,
followed by the participants are shown in Appendix A. the instantaneous observations had not undergone rigorous qual-
ity control. However, OHD performed some rudimentary quality
2.9.4. Potential evaporation control steps. These involved: (1) downloading the approved mean
Participants were allowed to determine the values of potential daily flow data from the NWIS web site for the same time periods,
evaporation data for their models. As in the DMIP 2 experiments (2) computing mean daily flow from the hourly data, (3) visually
in Oklahoma, estimates of climatological monthly PE for both ba- comparing the derived and approved daily flow time series, hourly
sins were provided. Koren et al. (1998) used information from sea- streamflow data, mean areal precipitation data, and basic simula-
sonal and annual Free Water Surface (FWS) evaporation maps in tions for each basin, and (4) setting any suspicious data in the
NOAA Technical Report 33 (Farnsworth et al., 1982) and mean hourly time series to missing values.
monthly station data from NOAA Technical Report 34 (Farnsworth
et al., 1982) to derive parameters for an equation that predicts the
2.9.10. Soils information
seasonal variability of mean daily free water surface (FWS) evapo-
State Soil Geographic (STATSGO) texture data covering the two
ration. These parameters were used to derive the mean monthly
basins were taken from data sets originally derived by Miller and
FWS evaporation estimates for DMIP 2 West basins. The data for
White (1998). These were made available as a grid for each of 11
NFDC1 are in the same range as the estimates derived by Carpenter
soil layers. In addition, a link was provided to finer resolution
and Georgakakos (2001) for the entire American River basin.
county-level soil information called the Soil Survey Geographic
A link to the NARR data (Mesinger et al., 2006) was also pro-
(SSURGO; Zhang et al., 2012) data set. The SSURGO data are typi-
vided, along with guidelines and processing codes for participants
cally available at a scale of at least 1:24,000. They are approxi-
to compute other estimates of PE. The NARR project provides rela-
mately ten times the resolution of STATSGO data in which the
tive humidity, wind speed, air temperature, and radiative flux data.
soil polygons can be on the scale of 100 – 200 km2. SSURGO data
have been recently used to derive a priori estimates of model
2.9.5. Analysis of simulated hydrographs
parameters (e.g., Zhang et al., 2012, 2011). Participants were free
As a final check of the DMIP 2 West forcings, the precipitation,
to use information from either soil data set to derive any necessary
temperature, and climatological PE forcing data were tested in
model-specific parameters.
hourly lumped and distributed simulations over the project period.
Snow correction factors used in Snow-17 for these simulations to
compensate for gage undercatch were derived during the calibra- 2.9.11. Areal extent of snow cover
tion of the benchmark LMP model. All suspect simulated hydro- NWS RFCs use snow covered area (SCA) data in the calibration
graph ‘spikes’ that did not seem to be consistent with the of the operational hydrologic models. For DMIP 2 West, SCA data
observed data were investigated (Smith et al., 2010). Cumulative from the NWS NOHRSC were extracted for the basins. The data
streamflow simulation error plots were also examined. consisted of gridded ‘snap shots’ of SCA on available days, with val-
ues in each cell indicating clouds, snow, or no snow.
2.9.6. Digital elevation data
Participants were not required to use any particular digital ele-
2.9.12. Snow water equivalent
vation model (DEM). 15 arc-s and 1 arc-s DEM data were provided
Observed SWE data were also made available to DMIP 2 West
via the DMIP 2 West website to encourage participation. The
participants. SWE data for two USBR sites in NFDC1 were down-
15 arc-s national DEM was derived by resampling 3 arc-s DEMs
loaded from the California Data Exchange (CDEC). Data from four
(1:250,000 scale) distributed by the U.S. Geological Survey. 1 arc-
SNOTEL sites in the GRDN2 basin were also provided. The sites
s DEM data were available from the USGS National Elevation Data-
are listed in Appendix C. These data spanned the calibration period.
set (NED).
Participants were allowed to use these data in the calibration of
their models.
2.9.7. Flow direction data
Flow direction grid files at a 30 m resolution were provided for
the convenience of any participants who wished to use them. 2.9.13. Cross sections
These 30 m grids were used to define the basin boundaries. The ba- Due to the remote nature of the North Fork basin, we were only
sis for these flow direction grids was the 30 m DEM data from the able to provide cross section data for one location in the North Fork
USGS NED data server. The DEM data were projected and filled, and basin. These data were derived from as-built bridge plans for the
commercial software was used to calculate flow directions using Iowa Hill Bridge near Colfax, CA provided by personnel from Cali-
the D8 algorithms of Jenson and Domingue (1988). Flow directions fornia State Parks.
3. Results and discussion where S is storage, Pobs is observed mean annual basin-average pre-
cipitation in mm, E is evaporation in mm, L represents the inter-
Following the format of the DMIP 2 Oklahoma results (Smith catchment groundwater transfer (losses or gains) and Rmodel is the
et al., 2012b), we present the results of the experiments from gen- depth of model runoff in mm over the basin. We computed these
eral to specific in order to address the science questions in a coher- quantities on an annual basis over a multi-year period and assumed
ent manner. A number of statistical measures were used to assess that the change in storage over that period is equal to zero. Ob-
the simulations. It would be impossible to present and discuss all served mean annual precipitation over the basin and computed
of the analyses that were performed, but the results of the most runoff from each of the models was used to compute a budget-
important and relevant ones are presented below. based estimate of evaporation E and losses L:
3.1. Overall water balance

E þ L ¼ P obs Rmodel ð2Þ
A critical aspect of hydrologic modeling is the partitioning of Fig. 6 can be interpreted as follows. Each diagonal represents
precipitation into runoff and evapotranspiration/losses. This is the partitioning of observed precipitation into computed runoff
especially important in mountainous areas given the large uncer- and evaporation (plus losses) for a basin, with the x and y inter-
tainty in precipitation, temperature, and other meteorological cepts equal to the value of the mean annual areal observed precip-
observations. Following other experiments (e.g., Lundquist and itation. On each diagonal, a model’s plotting symbol can be
Loheide, 2011; Mitchell et al., 2004; Lohmann et al., 2004, 1998; projected to the x and y axes to yield that model’s basin-averaged
Wood et al., 1998; Duan et al., 1996; Timbal and Henderson-Sell- mean annual runoff and evaporation plus losses. All models should
ers, 1998; Shao and Henderson-Sellers, 1996), we investigated plot on a single line with a 1 slope and x and y intercepts equal to
the ability of the participants’ models to partition precipitation the observed mean areal precipitation if they have the correct
into runoff, evaporation, and losses. water budget. All models should plot at the same point if they have
The water balance quantities for each model were computed the same partitioning of water. From Fig. 6, with the exception of
using the general continuity equation: one outlier (CEM for NFDC1), it can be seen that the models parti-
tion precipitation reasonably well. There is slightly more spread in
dS the results for NFDC1 (156 mm) than in the GRDN2 (80 mm) and
¼ Pobs E Rmodel L ð1Þ
dt CMEC1 (105 mm) basins, which may be due to the difficulty in
modeling the rain-snow dominated NFDC1.
820
800 NFDC1
LMP
Evaporation + Losses, mm
780
OHD
760 Obs
UOB
740
CEM
720 UPV
700
680
660
NFDC1
640
North Fork American River
(MAP 1526 mm)
620
600
720 740 760 780 800 820 840 860 880 900 920 940
Observed or Computed Runoff, mm
700
650
CMEC1 GRDN2
East Fork Carson at CMEC1
Evaporation + Losses, mm
600
Markleeville, CA obs
550 (MAP 847 mm) OHD

UOB
500 LMP
CEM
450
UCI
400 UPV
GRDN2
350
East Fork Carson at
300 Gardnerville, NV
250 (MAP 754 mm)
200
200 250 300 350 400 450 500 550 600
Observed or Computed Runoff, mm
Fig. 6. Partitioning of precipitation into runoff and evaporation + losses: NFDC1 (top) and CMEC1 and GRDN2 (bottom).
3.2. Long-term cumulative simulation error provided improved rmod values compared to LMP for the validation
period.
Overall model performance in terms of long-term runoff simu- The bottom panel of Fig. 8 shows the model performance over
lation error as performed by Reed et al. (2004) and Smith et al. the combined calibration and validation periods. The LMP and
(2012b) was also analyzed. This analysis was meant to examine OHD models had the highest rmod for this period in all the basins.
the consistency of the precipitation estimates and subsequent im- The CEM model achieved the next-highest values for three out of
pacts on multi-year hydrologic model simulations. Consistent pre- four basin tests. Only the OHD model provided improved rmod val-
cipitation data have been proven to be necessary for effective ues compared to LMP (cases of GRDN2 and CMEC1–3) in this per-
model calibration (Smith et al., 2012b and references therein). iod. Very similar results to Fig. 8 (not shown) were achieved when
For example, in some cases in the DMIP 2 Oklahoma experiments, using the Nash–Sutcliffe statistic instead of rmod.
the improvements gained by distributed models compared to The %Bias for all the calibrated models is shown in Fig. 9. The
lumped models were negated when the models were calibrated %Bias statistic is used to compare the overall simulated and ob-
using inconsistent precipitation data from DMIP 1 (Smith et al., served runoff volumes. Target values of the %Bias measure vary
2012b). Fig. 7 presents the cumulative runoff error plots for the in the literature from ±5% for NWS model calibration (Smith
NFDC1, GRDN2, and CEMC1-3 basins. In general, the plots are lin- et al., 2003) to ±25% (Moriasi et al., 2007).
ear, indicating that the precipitation, temperature, and evapotrans- A wide range of values can be seen looking collectively at the
piration forcings are temporally consistent. The under-prediction plots. For the calibration period, some models were able to achieve
from CEM shown in Fig. 7 follows from the partitioning shown in a near zero bias for several of the basins (e.g., CEM in GRDN2, LMP
Fig. 6: the CEM model generated less runoff volume over time. in CMEC1-3, OHD in NFDC1). Other models had a consistent posi-
The cumulative error plots in Fig. 7 are considerably improved tive or negative bias for all the basins and all periods (UOB, CEM).
compared to the OHD cumulative error plot using the original As might be expected, the %Bias values were larger for the CMEC1-
(flawed) DMIP 2 gridded QPE (Mizukami and Smith, 2012; Smith 2 test compared to the explicitly calibrated CMEC1-3 test. Recall
et al., 2009; Moreda et al., 2006). Moreover, the results in Fig. 7 that CMEC1-2 simulations were generated as interior points within
span approximately the same error range for two basins in the GRDN2 with no specific calibration (Modeling Instruction 2). In
DMIP 2 Oklahoma experiments (see Fig. 10 in Smith et al., 2012b). terms of %Bias, no model performed best in all cases, nor did any
distributed model consistently outperform the LMP benchmark.
While no one model performed consistently better than the
3.3. Comparison of distributed and lumped model results others in Fig. 8 and Fig. 9, all models achieved relatively high
values of the two statistics. For example, the values of rmod
In this section we begin to address the science question: can (0.63–0.95) and %Bias (20.1 to 5.2) for the combined calibration–
distributed hydrologic models provide increased streamflow simu- validation period in Figs. 8 and 9 fall well within the ranges of rmod
lation accuracy compared to lumped models in mountainous and %Bias in the non-snow DMIP 2 Oklahoma experiments (Figs. 3
areas? Fig. 8 presents the overall performance of the calibrated and 4 of Smith et al., 2012b).
models in terms of rmod (McCuen and Snyder, 1975), computed
hourly for the calibration, validation, and combined periods. The
term ‘overall’ means that the statistic was computed for each hour 3.4. Analysis of precipitation/runoff events
over the entire period specified. The rmod measure was used herein
to provide consistency with the DMIP 2 Oklahoma results (Smith To further investigate the science question of the performance
et al., 2012b) and DMIP 1 results (Reed et al., 2004). The rmod sta- of distributed and lumped models, statistics were computed for
tistic is a goodness-of-fit measure of hydrograph shape. In Fig. 8, 68 events in the NFDC1 and 92 events in the GRDN2 and CMEC1
the results are organized in order of increasing computational ele- basins. These events were selected from the combined calibration
ment size (i.e., ranging from 250 m for UOB to two elevation zones and validation period. Event statistics were computed because
for LMP). Recall that there are two sets of results for the basin our experience has shown that the overall run-period statistics
CMEC1: one is for an independent calibration of the basin can mask the improvement of distributed models over lumped
(CMEC1-3) and the other is for the use of the Markleeville gage models for individual events (Smith et al., 2012b; Reed et al.,
as a ‘blind’ interior simulation point (CMEC1-2). 2004). We use the same two measures as in DMIP 1 and DMIP 2
Looking collectively at the results in the top panel of Fig. 8, no (Reed et al., 2004; Smith et al., 2012b respectively) to evaluate
single model performed best for all the basins in the calibration the model performance for events: the event absolute % runoff er-
period. The models had the most uniform rmod value for the ror and the event absolute % peak error. These measures evaluate
snow-dominated basin GRDN2, and the most spread in rmod for the models’ ability to simulate runoff volumes and peak flow rates.
NFDC1. All models except LMP showed a decrease in rmod for the Eqs. (7a) and (7b) of Appendix D present the formula for % runoff
blind simulation test CMEC1-2. Not surprisingly, all models error and % peak error, respectively. Low values of % runoff error
showed an improvement in rmod when the CMEC1 basin was and % peak error are desired.
explicitly calibrated (compare CMEC1-2 with CEMC1-3). Compared The event statistics for the NFDC1 basin are shown in Fig. 10.
to LMP, the UOB model provided the only improved rmod values for Each plotting symbol represents the average measure of a specific
this calibration period for NFDC1, while the OHD, UOB, and CEM model for the 68 events. As much as possible, the same plotting
models provided equal-to or improved values for GRDN2. No mod- symbols in Smith et al. (2012b) and Reed et al. (2004) are used.
el was able to provide an improvement over the lumped model for The LMP and OHD models have near-identical performance with
the blind test at CMEC1-2. Only the OHD model provided improved the lowest values of the two statistics. The OHD model provides
rmod values for the CMEC1-3 test. a slight improvement in % peak error over the LMP model but at
The rmod values for the validation period are shown in the the expense of a slightly worse runoff volume. Next in order of per-
middle panel of Fig. 8. Models typically perform slightly worse in formance is the UOB model, followed by the UPV and CEM models.
validation periods compared to the calibration period. In our case, The event statistics for GRDN2 are plotted in Fig. 11. All models
mixed results were realized. Some models improved in this period have errors whose magnitudes are similar to Fig. 10. One difference
(e.g., CEM in the CMEC1-2 and CMEC1-3 cases), while others is that there is more spread between the OHD and LMP results here
showed the expected decline in rmod. Only the OHD model compared to the NFDC1 basin. In this case the OHD model provides
1500
NFDC1
1000 (Calibrated)
m
Cumullative errror, mm
500
-500
-1000
1000
CEM
UOB
1500
-1500 UPV
LMP
OHD
-2000
2000
Dec-88
Dec-89
Dec-90
Dec-91
Dec-92
Dec-93
Dec-94
Dec-95
Dec-96
Dec-97
Dec-98
Dec-99
Dec-00
Dec-01
Dec-02
Dec-03
Dec-04
Dec-05
8
5
1000
CMEC1 - 3
Calibrated
500 at Markleeville
Cumulatiive errorr, mm
-500
-1000
1000
C
CEM
UOB
UCI
-1500
UPV
LMP
OHD
-2000
Octt-90
Octt-91
Octt-92
Octt-93
Octt-94
Octt-95
Octt-96
Octt-97
Octt-98
Octt-99
Octt-00
Octt-01
Octt-02
Octt-03
Octt-04
Octt-05
1200
1000
GRDN2
800 (Calibrated)
C lib t d)
m
Cumullative errror, mm
600
400
200
-200 CEM
UOB
UPV
-400
LMP
OHD
-600
Oct-9
Oct-9
Oct-9
Oct-9
Oct-9
Oct-9
Oct-9
Oct-9
Oct-9
Oct-9
Oct-0
Oct-0
Oct-0
Oct-0
Oct-0
Oct-0
90
91
92
93
94
95
96
97
98
99
00
01
02
03
04
05
Fig. 7. Cumulative runoff error (mm) for the combined calibration and validation periods. NFDC1 (top), CMEC1-3 (middle), and GRDN2 (bottom).
improvement compared to the LMP model for both runoff volume GRDN2 basin. The latter test calls for explicit calibration of the
error and peak error. CMEC1 basin as an independent headwater basin (Modeling
Fig. 12 shows the calibrated event results for the CMEC1-2 and Instruction 3). The plotting symbols are shown in different sizes
CMEC1-3 tests. The former calls for the CMEC1 basin to be simu- to distinguish the results for each participant. The arrows show
lated as a blind interior point (Modeling Instruction 2) within the how explicit calibration of the CMEC1 basin impacts the statistics.
rmod - Calibration period

1
0.9
0.8
0.7
0.6
0.5
rmod
0.4
0.3
0.2
0.1
0
NFDC1 GRDN2 CMEC1-3 CMEC1-2
Basin
UOB
rmod - Validation period 250 m
1
0.9 UPV
0.8 400 m
0.7
0.6
0.5
OHD
rmod
4 km
0.4
0.3
0.2
UCI
0.1 8 sub basins
0
Basin
CEM
rmod - Combined period 5 zones
1
0.9
0.8
LMP
0.7 2 zones
0.6
0.5
rmod
0.4
0.3
0.2
0.1
0
Basin
Fig. 8. Calibrated rmod for three periods: calibration (top), validation, (middle), and combined periods (bottom). For each basin, models are arranged in order of increasing
model scale. Note: UCI only submitted simulations for CEMC1-3.
Note that UCI only submitted simulations for the CMEC1-3 test, generated a ‘blind’ simulation that was slightly better than the LMP
and the UOB used the same simulations for the CMEC1-2 and model in terms of % peak error (23% vs. 26%). These results may be
CMEC1-3 tests. The CMEC1-2 results (small plotting symbols) span influenced by the large size of the interior basin (714 km2) com-
approximately the same range of error values as in NFDC1 (Fig. 10) pared to the parent basin (922 km2).
and GRDN2 (Fig. 11). Thus, in this case, the distributed models cal- Summarizing the results of Figs. 10–12, only one distributed
ibrated at the basin outlet achieved about the same event simula- model (OHD) was able to perform at a level near or better than
tion performance at this interior location. Explicit calibration of the the LMP benchmark for absolute % runoff error and absolute % peak
CMEC1 basin (i.e., CMEC1-3) improved the event statistics for the error. However, all models achieved levels of performance equiva-
LMP, UPV, and OHD models as one would expect. The OHD model lent to those in one of the non-snow Oklahoma basins in DMIP 2.
%bias - Calibration period

20
15
10
5
%bias
-5
-10 OHD
CEM
-15
-20
Basin
UOB
250 m
% bias - Validation period
25
20
UPV
15
400 m
10 OHD LMP OHD
5
% bias
0 OHD
4 km
-5
-10
-15
-20
UCI
8 sub basins
-25
Basin
CEM
5 zones
% bias - Combined period
25
20
15 LMP
10 2 zones
LMP
5
% bias
-5
-10 OHD
-15
-20
-25
Basin
Fig. 9. Calibrated %Bias for three periods: calibration (top), validation, (middle), and combined periods (bottom). Note: UCI only submitted simulations for CEMC1-3.
For example, the range of values of absolute % runoff error (17.2– ment were computed: % improvement in peak flow, % improve-
47.3) and absolute % peak error (15.5–28.1) agree very well with ment in runoff event volume, and improvement in peak time
the results for the ELDO2 basin in Oklahoma (see Fig. 5 of Smith error in hours (Eqs. (8a)-(8c) in Appendix D). The calibrated simu-
et al., 2012b). lations from the LMP model were used as the benchmark.
Fig. 13 presents the three measures of improvement for the cal-
3.5. Improvement of distributed models over lumped models ibrated models, along with the inter-model average of the values. It
is desirable to achieve values greater than zero in each of the plots.
This analysis specifically addresses the question whether cali- Each data point is the average value of the measure for a specific
brated distributed models can provide improved event simulations model in a specific basin over many events. Looking collectively
compared to lumped models in mountainous regions. Using the at the plots, it can be seen that the OHD model provides improve-
same events from Section 3.4, three specific measures of improvement in peak flow and volume, as seen previously in Figs. 10–12.
50
CMEC1 - 2 and CMEC1 - 3
NFDC1
40
Event Absolute % Runoff Error

CEM
40 CEM
UOB UOB
OHD 30 UCI
LMP UPV
30 UPV OHD
LMP
20
20
10
10
0
0 0 10 20 30 40 50
0 10 20 30 40 50 Event Absolute % Peak Error
Event Absolute % Peak Error
Fig. 12. Average runoff volume errors and peak errors for 92 events in the CMEC1
Fig. 10. Average runoff volume errors and peak errors for 68 events in the NFDC1 basin. Results are from calibrated models from the combined calibration and
basin. Results are from calibrated models from the combined calibration and validation periods. These results are for Modeling Instruction 3 (CMEC1-3) in which
validation periods. CMEC1 was treated as an independent headwater basin calibrated at Markleeville,
NV. These are the larger plotting symbols. Also shown are results from Modeling
Instruction 2 (CMEC1-2), in which CMEC1 is treated as a ‘blind’ interior point within
the calibrated GRDN2 parent basin. These are shown by the smaller symbols for
40 each participant. The arrows show the direction from ‘blind’ to explicitly calibrated
results.
35 GRDN2
30 CEM
UOB experiments showed a greater number of cases of distributed mod-
25 UPV
el improvement than in DMIP 1 (Smith et al., 2012b).
OHD
The convention of Reed et al. (2004) is used herein to identify
20 LMP
the cases in which the ‘improvement’ in Fig. 13 was negative but
near zero. This shows that distributed models can perform nearly
15
as well as a calibrated lumped model. The UCI, UOB, and UPV mod-
10 els show ‘improvement’ values less than 5% for peak volume and
flow, and less than 1 h for peak timing.
5
3.6. Specific examples of distributed model improvement
0
0 10 20 30 40
To complement the event statistics presented in Section 3.5, we
Event Absolute % Peak Error provide an analysis of two events in December, 1995 and January,
Fig. 11. Average runoff volume errors and peak errors for 92 events in the GRDN2
1996 to diagnose the source of distributed model improvement.
basin. Results are from calibrated models from the combined calibration and While other events and models could have been selected, this case
validation periods. is used to intercompare distributed (OHD) and lumped (LMP) mod-
els that share the same precipitation/runoff physics.
Fig. 14 shows the results for the December, 1995 event in
Interestingly three models show improvement in peak timing: NFDC1. At least some of the improvement from the OHD model
OHD, CEM and UPV. (and other distributed models in other cases) may result from im-
Taken as a group, distributed models provided improved peak proved definition of the rain/snow line, and the subsequent im-
flow simulations in 24% of the 17 model-basin pairs, improved run- pacts on runoff generation. As an illustration, Fig. 15 shows the
off volume in 12% of the 17 model-basin pairs, and peak timing time evolution of the hourly rain/snow line as computed by the
improvements in 41% of the 17 model-basin pairs. These values LMP model for the month of December, 1995. The diurnal variation
correspond to 24%, 18%, and 28%, respectively, achieved in the of the rain/snow line is quite evident. The grey zone marks the
DMIP 2 Oklahoma tests (Smith et al., 2012b). However, caution is four-day period from December 11th to December 14th, during
advised as the DMIP 2 Oklahoma results were based on a much lar- which the LMP rain/snow line drops dramatically in elevation.
ger number of model-basin pairs (148 vs. 17). Fig. 16 shows how the LMP model and the OHD model simulate
Visual inspection of the hydrograph simulations revealed that the rain/snow line using the gridded temperature data at 15Z on
the UOB model also provided improved simulations for a few cer- December 12, 1995. The white line denotes the rain/snow line at
tain events, but these improvements were overwhelmed by other 1758 m computed by the LMP model. Areas higher in elevation
events in the average improvements statistics and not visible in (to the right in the figure) than this receive snow, while areas be-
Fig. 13. low this line receive rain. The OHD model receives rain over a lar-
While the results in this section may be discouraging, they are ger area than LMP as denoted by the red grid cells. Fig. 17 shows
entirely consistent with the results from the Oklahoma experi- the precipitation and runoff for this event from the LMP and
ments in DMIP 1 and 2. In DMIP 1, there were more cases when OHD models. The top panel shows that the OHD model generates
a lumped model out-performed a distributed model than vice more runoff than the LMP model. The middle two panels of
versa (Reed et al., 2004). The results of the DMIP 2 Oklahoma Fig. 17 show how the LMP and OHD models partition total
Improvement in Event Peak Flow: Distributed vs Lumped 400

(calibrated, combined period)
10 Observed
350
LMP
5 OHD
Improvement in Peak Error
300
UPV
Discharge m3sec -1
0 UOB
250 Dec 11 Dec 12 Dec 13 CEM
-5
CEM 200
-10 UOB
UCI 150
-15 UPV
OHD 100
-20 Ave
50
-25
-30 0
NFDC1 GRDN2 CMEC1-3 CMEC1-2 1 5 9 13 17 21 25 29 33 37 41 45 49 53 57 61 65 69 73 77 81 85
Basin Time from Dec 11, 1995 @12Z
Improvement in Event Runoff Volume: Distributed vs Lumped Fig. 14. Runoff event of December 12 and 13, 1995 in the North Fork. Simulated and
(calibrated, combined period) observed hourly hydrographs are shown.
10
5
Improvement in volume, %
Rain Snow Elevation Derived from LMP Model

0 4000
1524 m
r/s elev
-5
Rain Snow Elev, m msl
-10 CEM 3000

UOB
-15 UCI
UPV
OHD 2000
-20
Ave.
-25
1000
-30
NFDC1 GRDN2 CMEC1-3 CMEC1-2 Dec 11, hour 0
hour 241
Basin Dec 14, hour 24
0
1 45 89 133 177 221 265 309 353 397 441 485 529 573 617 661 705
Improvement in Peak Time: Distributed versus Lumped
(calibrated, combined period) Hours from December 1, 1995
4
Fig. 15. Time evolution of the elevation of the hourly rain/snow line as computed
3 by the LMP model for December, 1995. The grey shaded area corresponds to the
Average Improvement, hours
period of December 11–14.

2
1
shown in Fig. 18 rather than averaging it over the LMP elevation
0
zones. As a result, fast-responding surface runoff is generated in
-1 CEM
several cells and is routed down the channel network. In the
UOB lumped model, the precipitation is spatially averaged over the ele-
-2 UCI vation zones, resulting in a delayed runoff response. Another cause
UPV
-3 OHD
may be the difference in channel routing schemes between the
Ave. OHD model (kinematic wave) and LMP (unit hydrographs and
-4 lag/k routing).
The contributions of the spatial variability of runoff generation
Basin
and routing were also investigated. Hourly OHD surface and subsur-
Fig. 13. Improvement of distributed models over lumped models: peak flow (top), face routing volumes shown (as accumulations) in Fig. 18 were aver-
event runoff volume (middle), and peak time (bottom). aged over the entire basin, then used as spatially uniform input to
each 4 km grid into the OHD routing network. In this case, the result-
precipitation into different amounts of rain and snow. The bottom ing hydrograph (not shown) was very similar to the OHD hydro-
panel shows the arithmetic difference between the two rainfall graph in Fig. 14 with the result that routing constitutes only 9% of
time series. the difference in root mean square (rms) error statistics between
The OHD model partitions total precipitation into a greater per- the OHD and LMP hydrographs. This small contribution seems rea-
centage of rain than the LMP model, and more runoff is generated sonable given the central location of the runoff volumes. However,
by the OHD model compared to the LMP model. However, our anal- this is only an approximate comparison as the OHD and LMP runoff
ysis did not determine the exact cause of the increased runoff. The volumes prior to routing were not equal (Fig. 17, top panel).
increase could be simply due to the greater amount of total rain in- In another case of distributed model simulation improvement
put into the OHD model. Another cause could be that distributed (January 25, 1996; not shown), liquid precipitation and surface
models (OHD) preserve the precipitation intensity in each grid as runoff were concentrated near the basin outlet. The routing contri-
Dec 12, 1995 at 15Z
= OHD Snow LMP

Snow LMP
= OHD Rain Snow
LMP
Snow
= 1758 m
Contour
Fig. 16. Difference between the LMP model snow areas (to the right of the white 1758 m contour line) and the OHD distributed model snow area (red grid cells) for the hour
ending at 15Z on December 12, 1995. The NFDC1 basin outline is shown in yellow. The grid shows the outline of the 4 km HRAP cells.
bution to the difference in rms error between the OHD and LMP NFDC1, calibration improved the rmod and %Bias measures for sev-
simulations in this case was much larger at 36%. In this case, cap- eral models: UPV, LMP, UOB, and OHD. The results for CEM do not
turing the spatial variability of the liquid precipitation and subse- show any gain in either measure, indicating that the calibration
quent runoff may be more important than differences in process focused on minimizing other error criteria. Indeed, the
precipitation partitioning. CEM model was calibrated using root mean square error (RMSE)
From our limited analysis, it is not clear which factor, or combi- calculated on root square transformed flows, which may explain
nation of factors, was the source of improvement of the OHD mod- why the change in rmod is not so satisfactory.
el. Beyond simply identifying the differences in precipitation For GRDN2, three of the four models achieved improved rmod
partitioning and the spatial variability of precipitation and runoff, values, but at the expense of increasing (positive or negative) the
it was difficult to isolate the source of distributed model improve- overall %Bias (UOB, LMP, and OHD). The most consistent improve-
ment. Further complications arise due to different initial condi- ments from calibration were seen in the CMEC1-3 test in the bot-
tions in the OHD and LMP model for the events studied. In one tom panel of Fig. 19. All models realized gains in the rmod and %Bias
case, the spatial distribution of precipitation appeared to play a statistics with the exception of UPV. The uncalibrated rmod results
major role, while in another case, modeling the rain/snow areas for the OHD and CEM models were better than the calibrated re-
seemed to have a large impact. Dornes et al. (2008) were able to sults for the three remaining distributed models.
attribute the benefits of their distributed model in two events to Although not shown here, parameter calibration resulted in
capturing the impact of topography on shortwave radiation for clear improvements in nearly all cases of other statistics such as
the modeling of snow accumulation and melt. However, their the Nash–Sutcliffe efficiency and root mean square error.
study did not address complexities arising from mixed rain and
snow. 3.8. Analysis of interior processes: streamflow and SWE
3.7. Effect of model parameter calibration This part of the DMIP 2 West experiments was designed to
investigate how distributed models represent basin-interior pro-
As in the DMIP 1 and 2 Oklahoma experiments, participants cesses in mountainous areas. Participants generated simulations
were instructed to submit simulations using calibrated and uncal- of two variables: streamflow and SWE.
ibrated model parameters. This analysis was designed to assess the Only one interior flow point was available (specified as a ‘‘blind’’
efficacy of a priori model parameters as well as schemes to cali- test with no explicit calibration; CMEC1-2). The results for cali-
brate hydrologic models in mountainous areas. The results shown brated simulations are shown in Fig. 8, Fig. 9 and Fig. 10. The mul-
are for the combined calibration and validation periods. The rmod ti-year rmod values for CMEC1-2 in Fig. 8 are only slightly lower
and %Bias measures are presented to provide an overall view of than the CMEC1-3 case in which specific calibration was allowed.
the impacts of parameter calibration. More of a difference between CMEC1-2 and CMEC1-3 is visible in
The calibration results for NFDC1, GRDN2, and CEMC1-3 are Fig. 9, where larger multi-year %Bias values can be seen for
presented in Fig. 19. The rmod and %Bias measures for uncalibrated CMEC1-2. This suggests that specific calibration at the Markleeville
and calibrated results are shown connected by an arrow indicating gage corrects for locally-generated biases that are not readily re-
the directional change in values (e.g., Viney et al., 2009). For moved when calibrating using downstream information at Gardn-
(a) Runoff OHD

LMP
(b) Snow
(c) Rain
(d) Rain: OHD - LMP
Fig. 17. Total precipitation and runoff for Dec, 1995 event in NFDC1 by the LMP and OHD models. All quantities are basin-average values. (a) Runoff generated by the LMP and
OHD models. (b) Snowfall registered by both models. (c) Rain for both models. (d) Arithmetic difference in OHD rain and LMP rain shown in panel c.
erville. Fig. 12 shows that specific calibration of the interior point sent the entire snow accumulation and ablation periods (approxi-
CMEC1-3 led to improved values of peak and runoff volume error mately October to June) for all years at each of the six snow gage
statistics for events compared to CMEC1-2 for two models (UPV sites. Table 4 presents the results of this overall analysis for cali-
and OHD). As stated earlier in Section 3.4, the CMEC1-2 results brated models. For the two stations in the NFDC1 basin, %Bias val-
may be influenced by the large size of the CMEC1 basin compared ues were greater in absolute magnitude for the Blue Canyon site
to the parent GRDN2 basin. compared to the Huysink site for all four models. This agrees with
Participants were also requested to generate uncalibrated and Shamir and Georgakakos (2006) who found that the largest uncer-
calibrated hourly SWE simulations at two instrumented points in tainty was for that part of the snow pack located where the surface
the NFDC1 basin and four instrumented points in the CMEC1 basin. air temperature is near the freezing level. With an elevation of
The SWE simulations represent the grid or other computational 1609 m, the Blue Canyon site is near the elevation of 1524 m typ-
area at which the model was run. Given the uncertainties involved ically used by the CNRFC to delineate rain and snow in its river
in comparing point to grid values of SWE, our goal was to under- forecast operations.
stand the general ability of the models to simulate the character The %Bias values at the four SNOTEL sites in or near the higher
of the snow accumulation and ablation processes (e.g., Shamir elevation CMEC1-3 basin are generally less than those of the sta-
and Georgakakos, 2006). Shamir and Georgakakos (2006) defined tions in the lower elevation NFDC1 basin. Here, the UOB and
a ‘‘good’’ SWE simulation as one that fell between simulated uncer- OHD models achieved the lowest values of the %Bias measure.
tainty bounds and also had consistent agreement with sensor Large %Bias values can be seen in Table 4, highlighting the dif-
observations. ficulties of simulating SWE in mountainous areas. For example,
We computed an average value of the %Bias of the simulated the UCI model featured a large over-simulation of SWE for the
SWE compared to the observed SWE. These average values repre- Spratt Creek SNOTEL site. This is due to the fact that the SWE sim-
107 – 123 80
123 – 140 NFDC1
NFDC1
140 – 156 70
156 – 173
173 – 189 60 CEM
189 – 206 UOB
206 – 222 50 UPV
222 – 239 OHD
239 – 255 40 LMP 0.60, 43.73
No Data
Bias, %
30
Cumulative precipitation, mm 0.58, 25.33 0.61, 25.03
20
0 – 14 0.44, 11.54
14 – 25 10 0.63, 6.69 0.74, 5.60
0.85, 4.42
25 – 36
36 – 46 0
46 – 57 0.89, -0.24 0.93, 0.36
57 – 68
-10
68 – 79 0.63, -13.07
79 – 90
90 – 100 -20
0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90 1.00
No Data
rmod
Cumulative surface runoff, mm
20
0 – 14
14 – 25
25 – 36
GRDN2
GRDN2 0.81, 15.60
36 – 46
15
46 – 57
57 – 68 CEM
68 – 79 10 0.61, 8.24
UOB
79 – 90 OHD
0.95, 5.16
90 – 100 LMP
5
Bias, %
No Data UPV
0.36, 0.47
Cumulative sub-surface runoff, mm 0
0.74, -1.17 0.88, -0.83
0.64, -0.79
Fig. 18. Spatial variability of precipitation and OHD runoff components for the 0.87, -5.06
December 1995 event in NFDC1. The top panel shows the cumulative precipitation -5 0.93, -3.87
(rain and snow, mm), the middle panel shows the cumulated fast-responding 0.77, -5.69
surface runoff (mm), and the bottom panel shows the cumulated slow-responding -10
sub-surface runoff (mm).
-15
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
ulation for this site was generated by the UCI sub-basin model in rmod
which the average elevation of the sub-basin containing the site
is over 400 m higher than the site itself—thus leading to overly fre-
quent snowfall events. The large %Bias values may also reflect the 15
scale mismatch between point SWE observations and simulations 10

of SWE generated over a grid, sub-basin, or elevation zone. CMEC1-3
Fig. 20 presents the simulated and observed SWE for a large 5 CMEC1 - 3 0.95, 2.15
0.77, 2.50
calibration at Markleeville 0.70, 0.63
snow year for the Blue Canyon site in the NFDC1 basin and the Blue 0 0.90, -1.61
Lakes site in the CMEC1-3 basin. The dates plotted are October 1,
0.81, -5.70
Bias, %
1992 to July 31, 1993. The Blue Lakes site accumulates about twice -5
0.39, -6.12
as much SWE as does the Blue Canyon site. Up to day 151, all mod- -10 0.73, -9.75
0.86, -10.53
els accumulate snow at a rate similar to that observed for Blue 0.74, -9.46 0.79, -10.06
CEM
Lakes. However, the onset of melt is different amongst models, -15 UOB
UCI
and all models melt off the snow more quickly than is observed. UPV
-20 OHD 0.67, -20.09
Problems with snow accumulation are more evident for the Blue LMP
Canyon site, perhaps as a result of difficulties in tracking the -25
0.50, -25.35
rain/snow line, intermodel differences in treating precipitation
-30
gage undercatch due to wind, and differences in how models deter- 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
mine the form of precipitation. rmod
One issue that arises in modeling studies is whether errors
(uncertainty) in precipitation and temperature forcings mask the Fig. 19. Improvement in rmod and %Bias achieved via model calibration: NFDC1
(top), GRDN2 (middle), and CMEC1-3 (bottom). Arrows indicate the direction from
simulation differences resulting from model physics. Fig. 21 illus-
uncalibrated to calibrated statistics. Results are from the combined calibration and
trates how intermodel spread in SWE simulations compares to dif- validation periods.
ferences in SWE caused by errors in the temperature forcing. We
use the work of Lei et al. (2007), who compared the responses of
the snow model in the Noah LSM (Koren et al., 1999) and the
Snow-17 model to random and systematic errors in temperature, el for the same water year and same levels of random temperature
solar radiation, and other meteorologic forcing variables available error. The spread in the DMIP 2 West SWE simulations at the Blue
in the NARR (Mesinger et al., 2006). SWE simulations for the Lakes SNOTEL site for water year 1999 is shown in the bottom pa-
Snow-17 model with different levels of random error in the tem- nel of Fig. 21. It can be seen from this figure that the spread in mod-
perature data are shown in the top panel for water year 1999. el-generated SWE could be as great as the spread caused by
The middle panel shows the SWE simulations from the Noah mod-
Table 4
Computed vs. observed SWE at six sites. Values shown are the multi-year average %Bias of simulated SWE compared to observed SWE.
Basin Station Elev. m Participant

CEM UOB OHD UCI UPV
NFDC1 Blue Canyon 1609 93.74 84.93 57.96 – 113.49
Huysink 2012 87.30 56.90 34.68 – 13.60
CMEC1-3 Blue Lakes 2456 86.80 7.04 13.85 15.89 11.67
Ebbett’s Pass 2672 69.06 11.39 14.85 31.69 21.72
Poison Flats 2358 79.89 9.66 15.36 21.49 37.55
Spratt Creek 1864 77.49 10.08 27.80 560.34 29.63
random errors in the temperature data, depending on the level of box represents the 25–75% quartile range of SCD differences in
data error. days, while the red line is the median of the values. In Fig. 22, it
The timing of the overall snow accumulation and melt was eval- is desirable to have an SCD difference of zero. Looking collectively
uated by computing the difference between the observed and sim- at the plots, there is a general trend for the models to have an early
ulated SWE centroid dates (SCD; Kapnick and Hall, 2010). The SCD SCD compared to the observed SCD. The values for the Huysink site
is computed using Eq. (3): in the NFDC1 basin have the largest and most consistent departure
P
t SWEi
SCD ¼ Pi ð3Þ
SWEi
1200
Observed
where SWE is the daily observed or simulated SWE in mm, t is the
1000
TSNOTEL Snow-17
number of the day from the beginning of snow accumulation, and i dT(2%*std)
denotes an individual SWE value. dT(10%*std)
800
dT(18%*std)
SWE, mm
Fig. 22 shows the difference SCD for each site for each of the
dT(26%*std)
years of the combined calibration and validation periods. Each 600
dT(34%*std)
400 dT(42%*std)
dT(50%*std)
1400 200
OBS
OHD 0
1200 O N D J F M A M J J
Blue Lakes UOB
CEM Month
1000
UPV
1200
mm.
UCI
Observed Noah Snow Model
WE, m
800
1000 TSNOTEL
dT(2%*std)
SW
600 dT(10%*std)
800
dT(18%*std)
SWE, mm
dT(26%*std)
400 600
dT(34%*std)
200 400
dT(42%*std)
dT(50%*std)
0 200
1 31 61 91 121 151 181 211 241 271 301
0
Days
y from Oct 1, 1992 O N D J F M A M J J
Month
900
1200
800 Blue Canyon Obs
OHD DMIP 2 West
1000
700 UOB
CEM
600 UPV 800
SWE, mm
m
S E, mm
500 600
SWE
400
400
300
200
200
0
100 O N D J F M A M J J
Month
0
1 31 61 91 121 151 181 211 241 271 301 Fig. 21. SWE simulations for the Blue Lakes SNOTEL station for water year 1999.
Days from Oct. 1, 1992 SWE simulations from Snow-17 (top) and Noah snow model (middle) using
different levels of random temperature error are taken from Lei et al. (2007). DMIP 2
Fig. 20. Simulated and observed SWE for water year 1993: Blue Lakes SNOTEL site West simulated SWE is shown in the bottom panel. The observed SWE is the same
(top) and the Blue Canyon site operated by the Bureau of Reclamation (bottom). in all three cases and is shown as a heavy black line.
Blue Canyon
y ((1,609m)) Huysink
y ((2,012m)
, )
s)
CDobbs (days
NFDC1
–SC
North Fork
CDssim –
American
A i River
Ri
SC
CEM UOB UPV OHD CEM UOB UPV OHD
Blue Lakes (2,456m) Ebbet’s

Ebbet s Pass (2,672m)
ays)
(2 456m) (2 672m)
m –S
CDsim
SC Dobs (da
SCD
CMEC1 3
CMEC1-3 CEM UOB UCI UPV OHD CEM UOB UCI UPV OHD
East Fork
Carson River
C Poison Flats (2,358m)
(2 358m) Spratt Creek (1,864m)
(1 864m)
–SC
S Dsim –
SCD s)
days
CDoobs (d
CEM UOB UCI UPV OHD CEM UOB UCI UPV OHD
Fig. 22. Difference in days between observed and simulated SCD each year for SWE measurement sites. Blue boxes represent the 25–75% quartiles, while the red line denotes
the median.
from zero. There is a tendency in the NFDC1 basin for the SCD to be contained in a station’s metadata. Such discrepancies could impact
slightly earlier in time compared to the observed value. The partic- model simulations.
ipants’ SCD values for the Huysink site are consistently about 10–
15 days earlier than the Blue Canyon station. Consistent with Sha- 3.9. Modeling scale
mir and Georgakakos (2006), better results were achieved at higher
elevation snow sites. The best SCD results are for the two highest DMIP 2 West intended to examine the science question of
elevation stations, Blue Lakes and Ebbet’s Pass. At these sites, four appropriate model scale. In the project formulation phase, it was
models achieved the best timing as evidenced by the smallest hoped that there would be a sufficient number of models to make
spread in the 25–75% quartile range and SCD differences less than inferences between simulation results and model scale. Unfortu-
10 days. Larger errors in timing (Fig. 22) and bias (Table 4) could be nately, the number of participants (five plus LMP) did not allow
the result of scale mismatch between the point observations and us to investigate this issue as fully as hoped. Nonetheless, a brief
the models’ computational element size. For example, the larger discussion is provided here given the void in the literature on this
errors seen in the CEM snow simulations may be the result of using aspect of modeling.
a coarse modeling scale (5 elevation zones). The spatial modeling scales ranged from grids of 250 m (UOB),
The statistics presented here should be viewed with caution as to 400 m (UPV) to 4 km (OHD) to elevation zones (five in CEM;
considerable uncertainty exists in the representativeness of point two in LMP) and to sub-basins (8 sub basins with average size
SNOTEL (and other) observations of SWE to the surrounding area 90 km2, UCI). Runoff statistics reflect the integration of many pro-
(e.g., Shamir and Georgakakos, 2006; Dressler et al., 2006; Garen cesses including snow accumulation and melt, rain on snow, rain
and Marks, 2005; Simpson et al., 2004; Pan et al., 2003). Of course, on bare ground, and hillslope and channel routing. Accordingly,
point-to-point comparisons can be made when such data are avail- we acknowledge that the results portray a mix of both model
able (e.g. Rutter et al., 2009; Rutter et al., 2008). However, there is physics and modeling scales and that definitive conclusions
great variability amongst results even when models are run at the regarding modeling scale cannot be made. Nonetheless, it is inter-
point scale and compared to research-quality point observations. esting that no trends between performance and modeling scale
For example, the SnowMIP 2 project examined the performance can be seen in the calibrated rmod and %Bias plots of Fig. 8 and
of 33 snow models of various complexities at four point sites (Rut- Fig. 9, respectively, for the calibration, validation, and total simu-
ter et al., 2009). One of the conclusions from SnowMIP 2 was that it lation periods. Results are plotted in Fig. 8 and Fig. 9 in order of
was more difficult to model SWE in forested sites compared to increasing model resolution. Even where the model physics is
open sites. Moreover, there was no ‘best’ model or subset of mod- similar (i.e., OHD, LMP, and UCI), there is no trend in these run-
els, and models that performed well at forested sites did not nec- period statistics.
essarily perform well (in a relative sense) at open sites. Along However, a different picture emerges when looking at the
these lines, Mizukami and Koren (2008) noted discrepancies be- event statistics for calibrated models. Focusing on the two models
tween satellite-estimated forest cover and the description of cover with common physics which were run in all cases, the average
event statistics in Figs. 10–12 for calibrated models show that for peak flow and runoff volume. However, three models (OHD,
OHD provides better values than LMP. In all three basin cases, CEM, and UPV) achieved improvements in peak event timing com-
the OHD model provides lower peak error values, and in two out pared to LMP, highlighting the potential of distributed models to
of three cases the runoff volume is better. Recall that LMP is actu- capture spatially-variable precipitation and runoff processes. This
ally the case of two elevation zones, above and below 1524 m in evaluation of precipitation/runoff events showed that taken to-
NFDC1 and 1724 m in GRDN2 and CMEC1. The event improvement gether, distributed models were able to provide improved peak
statistics in Fig. 12 further illustrate the improvement of OHD flow values in 24% of the 17 model-basin pairs, improved runoff
compared to LMP. Uncalibrated run period and event statistics volume in 12% of the pairs, and improved peak timing in 41% of
(not shown) also show improvements of OHD compared to LMP. the pairs.
Comparing models with the same physics, the OHD and LMP re- Even though the gains by distributed models over the LMP
sults agree with the scope and results of Dornes et al. (2008), benchmark were modest, all models performed well compared
who modeled the snow and runoff processes of a basin with to those in the less-hydrologically-complex Oklahoma basins in
lumped and distributed applications of the same precipitation- DMIP 2. For example, the rmod and %Bias results of all models in
runoff model. In a limited way, the results herein support the the multi-year run-period tests are commensurate with those in
expectation that higher resolution modeling scale will improve the non-snow-dominated DMIP 2 Oklahoma basins (Smith et al.,
simulation performance. 2012b). Similarly, the event-averaged absolute % runoff error
and absolute % peak error values agree well with the range of
3.10. Rain and snow partitioning values for the DMIP 2 ELDO2 basin (Smith et al., 2012b). These
results are noteworthy in that the DMIP 2 West basins have
DMIP 2 West intended to address the important issue of rain- complexities such as orographic enhancement of precipitation,
snow partitioning. Primarily this was to be addressed via the use snow accumulation and melt, rain-on-snow events, and highly
of radar-detected observations of freezing level from HMT-West varied topography which are not present in the DMIP 2 Oklahoma
(Minder and Kingsmill, 2013; White et al., 2010, 2002) after partic- basins. Looking at the results herein and the DMIP experiments
ipants set-up and ran their models with the baseline DMIP 2 West overall, it is clear that at this point in their evolution, distributed
gridded precipitation and temperature data. Participants would models have the potential to provide valuable information on
then make additional simulations using the radar-based estimates specific flood events that could complement lumped model
of the rain-snow line to note the improvement. Delays in DMIP 2 simulations.
West caused by the need to generate a new QPE data set precluded Based on these mixed results, care must be taken to examine a
the use of HMT-West data in formal experiments with participants. range of statistical measures, simulation periods, and even hydro-
Nonetheless, using the DMIP 2 West modeling framework, Mizuka- graph plots when evaluating the performance of distributed mod-
mi et al. (2013) tested the OHD and LMP models with and without els compared to lumped models. As in the DMIP 2 Oklahoma basins
the HMT-West radar-derived rain-snow data for the 2005–2006 (Smith et al., 2012b), our results actually reflect the model/user
winter period. Mixed simulation results were seen; some runoff combination and not the models themselves.
events were better simulated while other events were worsened. It proved difficult to determine the dominant factors which led
Interested readers are referred to Mizukami et al. (2013) for more to the improvement of the OHD distributed model over the bench-
information. mark LMP model for mixed rain/snow events. Our limited analyses
on one of two study basins identified complex interactions of pre-
cipitation partitioning, spatial variability of liquid precipitation,
4. Conclusions
runoff generation, and channel routing.
We present the major conclusions generally in order of the sci-
ence questions listed in Section 1.2. Interspersed among these are 4.2. Estimation of models inputs and model sensitivity to existing data
additional conclusions and comments.
The distributed models used gridded forms of the precipitation
4.1. Distributed vs. lumped approaches in mountainous areas and temperature data widely used by NWS RFCs in mountainous
areas for hydrologic model parameter calibration. In the study ba-
Overall, no single model performed best in all basins for all sins, the density of precipitation and temperature gauges was suf-
streamflow evaluation statistics. Neither was any distributed mod- ficient to develop useful gridded estimates of these variables over a
el able to consistently outperform the LMP benchmark in all basins 20-year span. A sufficient number of hourly (recording) rain gauges
for all indices. Nonetheless, one or more distributed models were were available to distribute daily precipitation observations. These
able to achieve better performance than the LMP benchmark in a data were able to support effective model calibration and good
number of the evaluations. These results are consistent with the simulations through the validation period, evidenced by %Bias val-
findings of DMIP 1 (Reed et al., 2004) and DMIP 2 West (Smith ues within or near the ±5% criteria for NWS model calibration
et al., 2012b). We highlight several aspects of model performance (Smith et al., 2003), low cumulative runoff errors, and high values
below. of rmod.
Considering the rmod and %Bias measures computed for the mul- For this study, careful quality control of the raw precipitation
ti-year calibration, validation, and combined calibration–validation data was essential. This seemed especially warranted given the
periods, mixed results were achieved. No single model performed sensitivity of the hydrologic models noted in the development of
best in all periods in all basins. In addition, no distributed model the QPE data. Numerous errors, especially in the precipitation
consistently performed better than the benchmark LMP model. observations, were identified and corrected. The OHD and LMP
However, three models (OHD, UOB, and CEM) were able to outper- models were sensitive to these errors for hourly time step simula-
form the LMP model for certain periods in certain basins. tions of mixed rain/snow events. Such errors manifested them-
The models were also inter-compared by evaluating specific selves as anomalous hydrograph peaks. The impact of such
precipitation/runoff events. Here, only one model (OHD) was able precipitation data errors may not be as evident in streamflow
to perform at a level near to or better than the LMP benchmark hydrographs that are dominated by snow melt.
4.3. Internal consistency between SNOTEL SWE observations and the SWE simulations
generated over the computational units within the participants’
The ability of distributed models to simulate snow accumula- models. The amount of forest cover at each SWE site should
tion and melt was investigated at six SWE observation sites. The be derived so that our results can be placed in the context of
best results in terms of timing and volume were seen at the higher the SnowMIP 2 results (Rutter et al., 2009).
elevation stations. Larger errors in simulated SWE were apparent Continued efforts are necessary to diagnose the causes of differ-
at a station near the typical elevation separating rain and snow. ences between distributed and lumped model simulations. These
In addition, larger errors in timing and volume of snow accumula- efforts will require detailed analyses, probably along the lines of
tion and melt were seen in the distributed models featuring larger the hydrologic and meteorological studies of Lundquist et al.
computational element sizes. This result may reflect the scale mis- (2008) and Minder et al. (2011), respectively. While additional
match between the point observations and the computational ele- work can be done with the data on hand, the advanced data avail-
ment size. Our findings should be viewed in light of the able from the HMT-West program will undoubtedly aid in this
considerable uncertainty that exists in the SWE observations and process.
their representativeness of the surrounding area. DMIP 2 West was formulated as a general evaluation of distrib-
A limited test with one interior flow point showed that some uted and lumped models in complex terrain without specific tests
distributed models calibrated at the outlet were able to achieve va- to highlight the benefits of model structure. To address this limita-
lid simulations of streamflow at the interior location. In particular, tion, experiments are recommended to uncover and diagnose the
the overall rmod statistics for the blind interior point CMEC1-2 were impacts of model structure on performance (e.g., Clark et al.,
commensurate with those achieved through explicit calibration at 2011; Butts et al., 2004).
the interior point (CMEC1-3). However, these good results may The benefits of using the HMT-West data sets of additional
hinge on the large size of the interior basin compared to the parent surface temperature (Lundquist et al., 2008) and precipitation,
basin. optical disdrometer, vertically pointing radar-based freezing level
(Mizukami et al., 2013; Minder and Kingsmill, 2013; Lundquist
4.4. Scale issues et al., 2008; White et al., 2010, 2002), soil moisture, and gap-fill-
ing radar-derived QPE (e.g., Gourley et al., 2009) should continue
Scale issues continue to be perplexing in mountainous areas. to be explored. These data sets should also aid in the diagnosis
While our study was limited in scope, the results address a void of modeling improvements. DMIP 2 West was always intended
in the literature regarding modeling scales that consider snow to be a multi-institutional and multi-model evaluation of the
accumulation, melt, and runoff generation. Even in the highly var- QPE, disdrometer, soil moisture, radar-freezing level, and other
iable terrain of NFDC1, a range of modeling scales led to relatively observations afforded by the rich instrumentation deployments
good streamflow simulations. Among the models that shared the in HMT-West. The intent was to first generate streamflow and
same snow/rainfall/runoff schemes, better event statistics were SWE simulations using the ‘basic’ DMIP 2 West gage-only QPE
achieved at higher resolution modeling scales. Considering all the and temperature fields. After calibrating and running their mod-
models, which admittedly represented a mix of physics, user els with the basic data, it was planned to have participants re-
knowledge, and model scales, it was surprising that more apparent run their models using the HMT-West data (radar QPE, snow le-
trends did not appear given the range of modeling resolution from vel, and soil moisture) to note the improvements gained by ad-
250 m to two elevation zones. We were not able to pinpoint an vanced observations. However, both DMIP 2 West and HMT-
optimal modeling scale. West experienced major delays, with the unfortunate result
being that the HMT-West data sets could not be explored in for-
4.5. Parameter calibration mal DMIP 2 West experiments.
Based on our experience with deriving precipitation, tempera-
While not an explicitly identified science question, our results ture, and evaporation forcing data sets for DMIP 2 West, continued
show that parameter calibration led to improved goodness-of-fit work in deriving these forcings in complex terrain is of near-para-
statistics for nearly all model-basin pairs. This suggests that cali- mount importance for model testing, development, and calibra-
bration strategies can be effective in areas with complex hydrol- tion. Continued work is needed to address gage network density
ogy. It also suggests that calibration strategies are needed even issues in mountainous areas. This is true for gage-only QPE and
with advances in model structure and the development of a priori for the use of rain gages to bias-adjust radar estimates of precipi-
parameter estimates. tation. In spite of the enormous effort involved, data sets covering
a large number of basins would support additional experiments
5. Recommendations and lead to broader conclusions (e.g., Andréassian et al., 2009,
2006).
While DMIP 2 West provided interesting and informative re- River Forecast Centers within the NWS should consider the use
sults, much work remains to further address the science questions of the OHD model. Other operational forecasting agencies should
posed in DMIP 2 West and other issues that have bearing on moun- consider the use of distributed models in complex terrain. For
tainous area hydrologic simulation and forecasting. the foreseeable future, such models should be viewed as comple-
Our results should be further examined in the context of ments to existing lumped forecast models rather than outright
uncertainty in forcing data and model parameters. For SWE, replacements.
one simple method would be to use the approach of Shamir
and Georgakakos (2006). In their approach, the uncertainty Acknowledgements
bounds were defined by running the snow model on an adjacent
south-facing grid cell and a nearby north-facing grid cell. The We acknowledge the participants’ institutions for their exten-
resultant SWE simulations formed the lower and upper uncer- sive support of the DMIP 2 West modeling efforts. We thank Gary
tainty bounds, respectively. Another idea is to use the results Estes, organizer of the California Extreme Precipitation Sympo-
from Molotch and Bales (2006) to understand the relationship sium, for his help with finding cross section information. Ken
324
Appendix A. Models used in DMIP 2 West (See Table A1)
Table A1
Participating groups and major model characteristics.
Participant Modeling Primary Primary application Spatial unit Rainfall-runoff/vertical flux Snow Rain/snow partitioning method Channel routing
system reference for rainfall- model model
name runoff
M. Smith et al. / Journal of Hydrology 507 (2013) 300–329

calculations
CEMAGREF (CEM) GR4J Perrin et al. Streamflow, Lumped Soil moisture accounting CemaNeige Threshold temperature in each of five Unit hydrograph and
(2003), hydrological with five (degree day elevation zones non-linear channel
Mathevet simulation, and elevation type routing
(2005), Le Moine forecasting zones model)
(2008), Valéry
(2010)
U. Bologna (UOB) TOPKAPI-X Liu and Todini Real time flood 250 m grid Non-linear reservoir with Mass and Percent of precipitation as rain on each cell: Kinematic Wave for
(2001), Todini forecasting cells Dunne infiltration. Percolation energy FðTÞ ¼ 1
TTs
where r = 0.3, T is air slopes greater than 0.1%
1þe r
and Ciarapica by Clapp and Hornberger. ET by balance and Modified
temperature, Ts is derived from historical
(2002) Thornthwaite scaled on an Muskingum–Cunge for
data by plotting frequency of precipitation
hourly basis slopes lower than 0.1%
type vs. air temp.
NWS Office of HL-RDHM Koren et al. Streamflow, 16 km2 grid SAC-SMA modified with heat Snow-17 Threshold temperature Kinematic wave
Hydrologic (2004) water resources cells transfer component for frozen
Development (OHD) forecasting ground effects
NWS Office of NWSRFS Burnash (1995) Streamflow, Elevation SAC-SMA Snow-17 Threshold temperature, lapsed, Unit hydrograph, Lag/K
Hydrologic water resources zones area elevation curve
Development (LMP) forecasting
Technical U. Valencia TETIS v8 Francés et al. Streamflow for flood 400 m grid Japanese Tank model Degree-day Threshold temperature Geomorphologic
(UPV) (2007), Valéry and continuous cells Kinematic Wave
et al. (2009) simulation
U. Ca Irvine Semi- Ajami et al. Streamflow 8 Sub- SAC-SMA Snow-17 Threshold temperature Kinematic wave
(UCI) distributed (2004), Khakbaz simulation basins with
SAC-SMA et al. (2012) the average
and area of
SNOW17 90 km2
Appendix B. Calibration strategies for DMIP 2 West models (See Table B1)
Table B1
Calibration strategies for the DMIP 2 West models.
Group Strategy
LMP Systematic manual adjustment of parameters starting with baseflow and proceeding to fast response flow generation processes (Smith et al., 2003). Several
statistical measures used at different points in the process to evaluate the fit of the simulation
OHD Start with a priori parameters defined from soil texture. Revise a priori parameters using lumped calibrated parameters (derived using procedures in Smith
et al., 2003): scale gridded a priori values by ratio of the SAC-SMA parameter value from the lumped calibration to the average parameter value from the a priori
grid. Evaluate if this initial scaling is appropriate. Then use scalar multipliers to uniformly adjust each parameter field while maintaining spatial variability.
Scalars are calibrated manually and/or automatically. Automatic calibration uses a multi-time scale objective function (Kuzmin et al., 2008)
CEM The parameters were estimated using a steepest-descent type method combined with overall prior screening of the parameter space (see e.g. Mathevet, 2005).
The resulting parameters were applied identically to all sub-basins
UPV Use correction factors to globally modify each parameter map, assuming the prior spatial structure and thus reducing drastically the number of variables to be
calibrated. In the used TETIS configuration, there were a total of nine correction factors: eight affecting the runoff production parameter maps and one for the
stream network velocity. The TETIS model includes an automatic calibration module based on the SCE-UA algorithm (Duan et al., 1994). For this application,
the objective function was the Nash–Sutcliffe efficiency index. The calibration was carried out in three steps: (1) calibration of rainfall-runoff parameters using
the 1990 summer period; (2) calibration of snowmelt parameters using the 1992–93 and 1994–95 winter periods; (3) refinement of rainfall-runoff parameters
using the period 1989–1993
UCI The optimal parameter set was estimated through calibration of the lumped SAC-SMA and SNOW-17 models (over the entire watershed) using SCE-UA
calibration algorithm (Duan et al., 1993) and MACS calibration scheme (Hogue et al., 2000). The resultant parameter set was then applied identically to all sub-
basins in the distributed model configuration in order to generate streamflow at the outlet and interior points
UOB Soils parameters derived from STATSGO texture classes in each grid. Adjusted horizontal and vertical saturated hydraulic conductivity and soil depth.
Calibration process was carried out with a ‘trial and error’ methodology, focusing on the highest flood events in order to obtain the best model performance
according to the Nash and Sutcliffe coefficient (Coccia et al., 2009)
Appendix C. Snow water equivalent stations and observation times (See Table C1)
Table C1
Characteristics of the six snow observation sites.
Basin Station Latitude/longitude Elev.m SNOTEL or CDEC1 ID Owner Observation time

North Fork American Huysink 39.282/120.527 2012 HYS U.S. Bureau Reclamation 0400 Local time; 1200 Z time
Blue Canyon 39.276/120.708 1609 BLC U.S. Bureau Reclamation 0400 Local time; 1200 Z time
East Fork Carson Blue Lakes 38.608/119.924 2456 356 NRCS SNOTEL 2400 Local time; 0800 Z time (next day)
Ebbetts Pass 38.550/119.805 2672 462 NRCS SNOTEL 2400 Local time; 0800 Z time (next day)
Poison Flats 38.506/119.626 7736 697 NRCS SNOTEL 2400 Local time; 0800 Z time (next day)
Spratt Creek 38.666/119.818 6115 778 NRCS SNOTEL 2400 Local time; 0800 Z time (next day)
1
California Data Exchange.
Howard of NSSL assisted with processing the PRISM data. The care- D.2. Simulated or observed mean
ful reviews and comments from the journal reviewers and editors
have contributed to the clarity of this paper.
PN
i¼1 Y i
Y¼
N
Appendix D. Statistical equations used in the analysis of DMIP 2
West results where Y is any type of data value.
D.1. Percent bias, PB (%)

D.3. Standard deviation, r
PB is a measure of total volume difference between two time

series. PB is computed as:
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
PN 2
i¼1 ðY i YÞ
PN rY ¼
i¼1 ðSi Oi Þ
N1
PB ¼ PN ð100Þ
i¼1 Oi
D.4. Correlation coefficient, r

where Si is the simulated discharge for each time step i, Oi is the ob-
served value, and N is the total number of values within the time While not used explicitly in the DMIP West results analysis, we
period of analysis. present the formula for the correlation coefficient as background
for the discussion on the modified correlation coefficient. The cor- b. Peak flow improvement Ip, %
relation coefficient r is defined as: This statistic quantifies the gain in simulating the peak event
discharge:
P P P
N Ni¼1 Si Oi Ni¼1 Si Ni¼1 Oi PN
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
r ¼ s ffi i¼1 ðjQ p;i Q ps;i j jQ p;i Q pz;i jÞ
P P 2 PN 2 PN 2 Ip ¼ 100
N Ni¼1 S2i N
S N O O N Q p;av g
i¼1 i i¼1 i i¼1 i
c. Peak time improvement It
This statistic measures the improvement in simulated peak
D.5. Modified correlation coefficient, rmod (McCuen and Snyder, 1975) time:
PN
i¼1 ðjT p;i T ps;i j jT p;i T pz;i jÞ
In this statistic, the normal correlation coefficient is reduced by It ¼
the ratio of the standard deviations of the observed and simulated N
hydrographs. The minimum standard deviation (numerator) and where Yi is the observed runoff volume of the ith flood, mm; Ys,i the
maximum standard deviation (denominator) are selected so as to (distributed model) simulated runoff volume of the ith event, mm;
derive an adjustment factor less than unity: Yz,i the (lumped model) simulated runoff of the ith flood to compare
with, mm; Yavg the average observed flood event runoff volume of N
minfrsim ; robs g events, mm; Qp,i the observed peak discharge of the ith event, m3 s1;
r mod ¼ r
maxfrsim ; robs g Qps,i the (distributed model) simulated peak discharge of the ith
event, m3 s1; Qpz,i the (lumped model) simulated peak discharge,
m3 s1; Qp,avg the average observed peak discharge of N events,
D.6. Root mean square error (%)
m3 s1; Tp,i the observed time of the ith peak, h; Tps,i the (distributed
model) simulated time of the ith peak, h; Tpz,i the (lumped model)
simulated time to ith peak, h and N is the number of selected events.
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
PN ffi
2
ðSi Oi Þ
i¼1
N References
%RMS ¼ 100
O
Ajami, N.K., Gupta, H., Wagener, T., Sorooshian, S., 2004. Calibration of a semi-
distributed hydrologic model for streamflow estimation along a river system. J.
Hydrol. 298, 112–135.
D.7. The following aggregate statistics were generated for selected Anderson, E.A., 1976. National Weather Service River Forecast System, Snow
individual events Accumulation and Ablation Model, NOAA Technical Memorandum NWS
HYDRO-17.
Anderson, E.A., 2002. Calibration of Conceptual Hydrologic Models for Use in River
a. Percent absolute event runoff error, Er, %
Forecasting. <http://www.weather.gov/oh/hrl/hsmb/hydrology/calibration/
This is the absolute value of the runoff bias from several events index.html>.
expressed as a percentage: Anderson, E.A., 2006. Snow Accumulation and Ablation Model –SNOW-17. <http://
www.nws.noaa.gov/oh/hrl/nwsrfs/users_manual/part2/_pdf/22snow17.pdf>.
PN Andréassian, V., Hall, A., Chahinian, N., Schaake, J., 2006. Introduction and synthesis:
i¼1 jBi j Why should hydrologists work on a large number of basin data sets?, vol 307.
Er ¼ 100 IAHS Publ., pp. 1–5.
NY av g
Andréassian, V., Perrin, C., Berthet, L., Le Moine, N., Lerat, J., Loumagne, C., Oudin, L.,
Mathevet, T., Ramos, H.M., Valéry, A., 2009. Crash tests for a standardized
b. Percent absolute peak error, Ep, % evaluation of hydrological models. Hydrol. Earth Syst. Sci. 13, 1757–1764.
This is the absolute value of error in peak discharge for several http://dx.doi.org/10.5194/hess-13-1757-2009.
events expressed as a percentage: Bales, R.C., Molotch, N.P., painter, T.H., Dettinger, M.D., Rice, R., Dozier, J., 2006.
Mountain hydrology of the western United States. Water Resour. Res. 42,
PN W08432. http://dx.doi.org/10.1029/2005WR004387.
i¼1 jQ pi Q psi j Behrangi, A., Hsu, K.-L., Imam, B., Sorooshian, S., Huffman, G.J., Kuligowski, R.J., 2009.
Ep ¼ 100 PERSIANN-MSA: a precipitation estimation method from satellite-based
NQ pav g
multispectral analysis. J. Hydrometerol. 10, 1414–1429. http://dx.doi.org/
c. Percent absolute peak time error, Et, h 10.1175/2009JHM1139.1.
Benjamin, S.G., Devenyu, D., Weygandt, S.S., Brundage, K.J., Brown, J.M., Grell, G.A.,
This is the absolute value of the error in peak time for several Kim, D., Schwartz, B.E., Smirnova, T.G., Smith, T.L., Manikin, G.S., 2004. An
events expressed as a percentage: hourly assimilation-forecast cycle: The RUC. Mon. Weather Rev. 132, 495–518.
PN http://dx.doi.org/10.1175/1520-0493(2004)132%3C0495:AHACTR%3E2.0.CO;2.
i¼1 jT pi T psi j Bennett, N.D., Croke, B.F.W., Guariso, G., Guillaume, J.H.A., Hamilton, S.H., Jakeman,
Et ¼ 100 A.J., Marsili-Libelli, S., Newham, L.T.H., Norton, J.P., Perrin, C., Pierce, S.A.,
N Robson, B., Seppelt, R., Voinov, A.A., Fath, B.D., Andréassian, V., 2013.
where Bi is the runoff bias per ith flood event, mm; Yavg the average Characterizing performance of environmental models. Environ. Model. Softw.
40, 1–20. http://dx.doi.org/10.1016/j.envsoft.2012.09.011.
observed flood event runoff, mm; Qp,i the observed peak discharge Blöschl, G., 1999. Scaling issues in snow hydrology. Hydrol. Process. 13, 2149–2175,
of the ith flood event, m3 s1; Qps,i the simulated peak discharge 10.1002/(SICI)1099-1085(199910)13:14/15<2149::AID-HYP847>3.0.CO;2-8.
of the ith flood event, m3 s1; Qp,avg the average observed peak dis- Boushaki, F.I., Hsu, K.-L., Sorooshian, S., Park, G.-H., 2009. Bias adjustment of
satellite precipitation estimation using ground-based measurement: a case
charge, m3 s1; Tp,i the observed time to the ith peak, h; Tps,i the sim- study evaluation over the southwestern United States. J. Hydrometeorol. 10,
ulated time to the ith peak, h, and N the number of selected events. 1231–1242. http://dx.doi.org/10.1175/2009JHM1099.1.
Braun, L.N., Brun, E., Durand, Y., Martin, E., Tourasse, P., 1994. Simulation of
discharge using different methods of meteorological data distribution, basin
D.8. Statistics to measure improvement over the LMP benchmark
discretization, and snow modeling. Nord. Hydrol. 25 (1-2), 129–144. http://
dx.doi.org/10.2166/nh.1994.009.
a. Flood runoff improvement Iy, % Budyko, M.I., 1974. Climate and Life. Academic Press, New York.
Burnash, R.J.C., 1995. The NWS river forecast system – catchment modeling. In:
This statistic measures the improvement in computed runoff
Singh, V.P. (Ed.), Computer Models of Watershed Hydrology. Water Resources
volume: Publications, Littleton, Colorado, pp. 311–366.
Butts, M.B., Payne, J.T., Kristensen, M., Madsen, H., 2004. An evaluation of the impact
PN
Y s;i j jY i Y z;i jÞ of model structure on hydrological modelling uncertainty for streamflow
i¼1 ðjY i
Iy ¼ 100 simulation. J. Hydrol. 298, 242–266. http://dx.doi.org/10.1016/
N Y av g j.jhydrol.2004.03.042.
Carpenter, T.M., Georgakakos, K.P., 2001. Assessment of Folsom lake response to Francés, F., Vélez, J.I., Vélez, J.J., 2007. Split-parameter structure for the automatic
historical and potential future climate scenarios: 1. Forecasting. J. Hydrol. 249, calibration of distributed hydrological models. J. Hydrol. 332 (1), 226–240.
148–175. Franz, K.J., Hogue, T.S., Sorooshian, S., 2008a. Operational snow modelling:
Carroll, T., Cline, D., Fall, G., Nilsson, A., Li, L., Rost, A., 2001. NOHRSC Operations and addressing the challenges of an energy balance model for National Weather
the Simulation of Snow Cover Properties of the Coterminous U.S. Proceedings of Service forecasts. J. Hydrol. 360, 48–66.
the 69th Annual Meeting of the Western Snow Conference, Sun Valley, ID, Franz, K.J., Hogue, T.S., Sorooshian, S., 2008b. Snow model verification using
Western Snow Conference, pp. 1–14. ensemble prediction and operational benchmarks. J. Hydrometeorol. 9, 1402–
Cayan, D.R., Riddle, L.G., 1993. The influence of precipitation and temperature on 1415. http://dx.doi.org/10.1175/2008JHM995.1.
seasonal streamflow in California. Water Resour. Res. 29 (4), 1127–1140. Garen, D.C., Marks, D., 2005. Spatially distributed energy balance snowmelt
Clark, M.P., McMillan, H.K., Collins, D.B.G., Kavetski, D., Woods, R.A., 2011. modelling in a mountainous river basin: estimation of meteorological inputs
Hydrological field data from a modeller’s perspective: Part 2: process-based and verification of model results. J. Hydrol. 315, 126–153.
evaluation of model hypotheses. Hydrol. Process. 25, 523–543. http:// Georgakakos, A.P., Yao, H., Georgakakos, K.P., 2010. Upstream Regulation
dx.doi.org/10.1002/hyp.7902. Adjustments to Ensemble Streamflow Predictions. HRC Technical Report No.
Clarke, R.T., 2008. A critique of present procedures used to compare performance of 7; Sponsored by NOAA/NWS/OHD, Award No. NA08NWS4620023.
rainfall-runoff models. J. Hydrol. 352, 379–387. http://dx.doi.org/10.1016/ Gourley, J.J., Jorgensen, D.P., Matrosov, S.Y., Flamig, Z.L., 2009. Evaluation of
j.jhydrol.2008.01.026. incremental improvements to quantitative precipitation estimates in complex
Cline, D., Elder, K., Bales, R., 1998. Scale effects in a distributed snow water terrain. J. Hydrometeorol. 10, 1507–1520.
equivalence and snowmelt model for mountain basins. Hydrol. Process. 12 (10– Green, D.R., Hudlow, M.D., 1982. Hydrometeorologic grid mapping procedures. In:
11), 1527–1536. AWRA International Symposium on Hydrometeorology, Denver, CO., Available
Coccia, G., Mazzetti, C., Ortiz, E.A., Todini, E., 2009. Application of the Topkapi model on Request from NWS/OHD.
within the DMIP 2 project. In: Paper 4.2, 23rd Conference on Hydrology, 89th Guan, B., Molotch, N.P., Waliser, D.E., Fetzer, E.J., Neiman, P.J., 2010. Extreme
Meeting of the AMS, Phoenix, Arizona, January 10–16. <https://ams.confex.com/ snowfall events linked to atmospheric rivers and surface air temperature via
ams/89annual/techprogram/paper_149632.htm>. satellite measurements. Geophys. Res. Lett. 37, L20401. http://dx.doi.org/
Confalonieri, R., Bregaglio, S., Bocchi, S., Acutis, M., 2010. An integrated procedure to 10.1029/2010GL044696.
evaluate hydrological models. Hydrol. Process. 24, 2762–2770. http:// Guentchev, G., Barsugli, J.J., Eischeid, J., 2010. Homogeneity of gridded precipitation
dx.doi.org/10.1002/hyp.7683. datasets for the Colorado River basin. J. Appl. Meteorol. Climatol. 49, 2404–
Daly, C., 2006. Guidelines for assessing the suitability of spatial climate data sets. 2415. http://dx.doi.org/10.1175/2010JAMC2484.1.
Int. J. Climatol. 26, 707–721. http://dx.doi.org/10.1002/joc.1322. Guirguis, K.J., Avissar, R., 2008. An analysis of precipitation variability, persistence,
Daly, C., Neilson, R.P., Phillips, D.L., 1994. A statistical-topographic model for and observational data uncertainty in the Western United States. J.
mapping climatological precipitation over mountainous terrain. J. Appl. Hydrometeorol. 9, 843–865. http://dx.doi.org/10.1175/2008JHM972.1.
Meteorol. 33, 140–158. Gupta, H.V., Wagener, T., Lui, Y., 2008. Reconciling theory with observations:
Daly, C., Halbleib, M., Smith, J.I., Gibson, W.P., Dogget, M.K., Taylor, G.H., Curtis, J., elements of a diagnostic approach to model evaluation. Hydrol. Process. 22,
Pasteris, P.P., 2008. Physiographically sensitive mapping of climatological 3802–3813. http://dx.doi.org/10.1002/hyp.6989.
temperature and precipitation across the conterminous United States. Int. J. Hansen, M.C., Defries, R.S., Townshend, J.R.G., Sohlbert, R., 2000. Global land cover
Climatol.. http://dx.doi.org/10.1002/joc.1688. classification at 1 km spatial resolution using a classification tree approach. Int.
De Pondeca, M.S.F.V., Manikin, G.S., DiMego, G., Benjamin, S.G., et al., 2011. The real- J. Rem. Sens. 21 (6&7), 1331–1364.
time mesoscale analysis at NOAA’s National Centers for Environmental Hartman, R.K., 2010. Hydrologic forecasting in mountainous terrain. In:
Prediction: current status and development. Weather Forecast. 26, 593–612. Presentation H21H-05, Presented at 2010 Fall Meeting, AGU, San Francisco,
http://dx.doi.org/10.1175/WAF-D-10-05037.1. Calif., 13–17 December.
Debele, B., Srinivasan, R., Gosain, A.K., 2009. Comparison of process-based and Hay, L.E., Clark, M.P., 2003. Use of statistically and dynamically downscaled
temperature-index snowmelt modeling in SWAT. Water Resour. Manage, atmospheric model output for hydrologic simulations in three mountainous
10.1007/s11269-009-9486-2. basins in the western United States. J. Hydrol. 282, 56–75.
Dettinger, M., Redmond, K., Cayan, D., 2004. Winter orographic precipitation ratios Hay, L, Viger, R., McCabe, G., 1998. Precipitation interpolation in mountainous regions
in the Sierra Nevada—large-scale atmospheric circulations and hydrologic using multiple linear regression. In: Preceedings ’98 Conference on Hydrology,
consequences. J. Hydrometeorol. 5, 1102–1116. Water Resources, and Ecology in Headwaters, IAHS Publication 248, 33–38.
Dong, X., Dohmen-Hansssen, M., Booij, M.J., 2005. Appropriate spatial sampling of Hogue, T.S., Sorooshian, S., Gupta, H., Holz, A., Braatz, D., 2000. Multi-step automatic
rainfall for flow simulation. Hydrol. Sci. J. 50 (2), 279–298. calibration scheme for river forecasting models. J. Hydrometerol. 1, 524–
Dornes, P.F., Pomeroy, J.W., Pietroniro, A., Carey, S.K., Qinton, W.L., 2008. Influence 542.
of landscape aggregation in modeling snow-cover ablation and snowmelt Jankov, I., Schultz, P.J., Anderson, C.J., Koch, S.E., 2007. The impact of different
runoff in a sub-arctic mountainous environment. Hydrol. Sci. J. 53 (3), 725–740. physical parameterizations and their interactions on cold season QPF in the
Dressler, K.A., Fassnacht, S.R., Bales, R.C., 2006. A comparison of snow telemetry and American River Basin. J. Hydrometeorol. 8, 1141–1151. http://dx.doi.org/
snow course measurements in the Colorado River Basin. J. Hydrometeorol. 7, 10.1175/JHM630.1.
705–712. Jenson, S.K., Domingue, J.O., 1988. Extracting topographic structure from digital
Duan, Q., Gupta, V.K., Sorooshian, S., 1993. A shuffled complex evolution approach elevation data for geographic information system analysis. Photogramm. Eng.
for effective and efficient global minimization. J. Optim. Theory Appl. 76, 501– Rem. Sens. 54, 1593–1600.
521. Jeton, A.E., Smith, J.L., 1993. Development of watershed models for two Sierra
Duan, Q., Sorooshian, S., Gupta, V.K., 1994. Optimal use of the SCE-UA global Nevada basins using a geographic information system. Water Resour. Bull. 29
optimization method for calibrating watershed models. J. Hydrol. 158, 265–284. (6), 923–932.
Duan, Q.Y., Schaake, J.C., Koren, V.I., 1996. FIFE 1987 water budget analysis. J. Jeton, A.E., Dettinger, M.D., Smith, J.L., 1996. Potential Effects of Climate Change on
Geophys. Res. 101 (D3), 7197–7207. Streamflow, Eastern and Western Slopes of the Sierra Nevada, California and Nevada.
Duan, Q.Y., Schaake, J.S., Andréassian, V., Franks, S., Goteti, G., Gupta, H.V., Gusev, U.S. Geological Survey Water Resources Investigations Report 95-4260, 44pp.
Y.M., Habets, F., Hall, A., Hay, L., Hogue, T., Huang, M., Leavesley, G., Liang, X., Jones, E.T., Roth, K., Costanza, K., 2009. A comparison of the NWS distributed versus
Nasonova, O.N., Noilhan, J., Oudin, L., Sorooshian, S., Wagener, T., Wood, E.F., lumped hydrologic model. In: AMS 23 Conference on Hydrology, Paper 4.3,
2006. Model parameter estimation experiment (MOPEX): an overview of Phoenix Arizona, January 10–15.
science strategy and major results from the second and third workshops. J. Kabeche, F., Figueras, J., Fradon, B., Hogan, R., Boumahmoud, A.-A., Illingworth, A.,
Hydrol. 320, 3–17. http://dx.doi.org/10.1016/j.jhydrol.2005.07.031. Tabary, P., 2010. Towards X-band polarimetric quantitative precipitation
Eischeid, J.K., Pasteris, P.A., Diaz, H.F., Plantico, M.S., Lott, N.J., 2000. Creating a estimation in mountainous regions: the RHYTMME project. In: ERAD 2010 –
serially complete, national daily time series of temperature and precipitation the Sixth European Conference on Radar in Meteorology and Hydrology, Sibiu
for the western United States. J. Appl. Meteorol. 39, 1580–1591. Romania, 6–10 September.
Essery, R., Martin, E., Douville, H., Fernanndez, A., Brun, E., 1999. A comparison of Kampf, S.K., Burges, S.J., 2007. A framework for classifying and comparing
four snow models using observations from an alpine site. Clim. Dyn. 15, 583– distributed hillslope and catchment hydrologic models. Water Resour. Res.
593. 43, W05423. http://dx.doi.org/10.1029/2006WR005370.
Etchevers, P., Martin, E., Brown, R., Fierz, C., et al., 2004. Validation of the energy Kapnick, S., Hall, A., 2010. Observed climate-snowpack relationships in California
budget of an alpine snowpack simulated by several snow models (SnowMIP and their implications for the future. J. Clim. 23, 3446–3456.
project). Ann. Glaciol. 38 (1), 140–158. http://dx.doi.org/10.3189/ Khakbaz, B., Imam, B., Hsu, K., Sorooshian, S., 2012. From lumped to distributed:
172756404781814825. calibration strategies for semi-distributed hydrologic models. J. Hydrol. 418–
Ewen, J., 2011. Hydrograph matching method for measuring model performance. J. 419, 61–77. http://dx.doi.org/10.1016/j.jhydrol.2009.02.021.
Hydrol. 408, 178–187. http://dx.doi.org/10.1016/j.jhydrol.2011.07.038. Kienzle, S., 2008. A new temperature based method to separate rain and snow.
Farnsworth, R.K., Thompson, E.S., Peck, E.L., 1982. Evaporation Atlas for the Hydrol. Process. 22, 5067–5085. http://dx.doi.org/10.1002/hyp.7131.
Contiguous 48 United States. NOAA Technical Report NWS 33, 27pp. Kim, J., Miller, N.L., Guetter, A.K., Georgakakos, K.P., 1998. River flow response to
Feng, X., Sahoo, A., Arsenault, K., Houser, P., Luo, Y., Troy, T.J., 2008. The impact of precipitation and snow budget in California during the 1994/1995 winter. J.
snow model complexity at three CLPX sites. J. Hydrometeorol. 9, 1464–1481. Clim. 11, 2376–2386.
http://dx.doi.org/10.1175/2008JHM860.1. Kingsmill, D.E., White, A.B., Neiman, P.J., Ralph, F.M., 2006. Synoptic and
Fierz, C., Riber, P., Adams, E.E., Curran, A.R., Föhn, P.M.B., Lehning, M., Pluss, C., 2003. topographic variability of Northern California precipitation characteristics in
Evaluation of snow-surface energy balance models in alpine terrain. J. Hydrol. land-falling winter storms observed during CALJET. Mon. Weather Rev. 134,
282, 76–94. http://dx.doi.org/10.1016/S0022-1694(03)00255-5. 2072–2094. http://dx.doi.org/10.1175/MWR3166.1.
Kirchner, J.W., 2006. Getting the right answers for the right reasons: linking McCabe, G.J., Clark, M.P., Hay, L.E., 2007. Rain-on-snow events in the western United
measurements, analyses, and models to advance the science of hydrology. States. Bull. Am. Meteorol. Soc., 319–328. http://dx.doi.org/10.1175/BAMS-88-
Water Resour. Res. 42, W03S04. http://dx.doi.org/10.1029/2005WR004362. 3-319.
Koren, V., Schaake, J., Duan, Q., Smith, M., Cong, S., 1998. PET Upgrades to NWSRFS. McCuen, R.H., Snyder, W.M., 1975. A proposed index for comparing hydrographs.
Internal NWS/OHD report. Available at: <http://amazon.nws.noaa.gov/ohd/ Water Resour. Res. 11 (6), 1021–1024.
publications/publications_google.php>. McDonnell, J.J., Sivapalan, M., Vache, K., Dunn, S., Grant, G., Hatterty, R., Hinz, C.,
Koren, V., Schaake, J., Mitchell, K., Duan, Q.-Y., Chen, F., Baker, J.M., 1999. A Hooper, R., Kirchner, J., Roderick, M.L., Selker, J., Weiler, M., 2007. Moving
parameterization of snowpack and frozen ground intended for NCEP weather beyond heterogeneity and process complexity: a new vision for watershed
and climate models. J. Geophys. Res. 104 (D16), 19569–19585. hydrology. Water Resour. Res. 43, W07301. http://dx.doi.org/10.1029/
Koren, V., Reed, S., Smith, M., Zhang, Z., Seo, D.-J., 2004. Hydrology laboratory 2006WR005467.
research modeling system (HL-RMS) of the U.S. National Weather Service. J. Merz, R., Parajka, J., Blöschl, G., 2009. Scale effects in conceptual hydrological
Hydrol. 291, 297–318. modeling. Water Resour. Res. 45, W09405. http://dx.doi.org/10.1029/
Koren, V., Moreda, F., Reed, S., Smith, M., Zhang, Z., 2006. Evaluation of a grid-based 2009WR007872.
distributed hydrological model over a large area. In: Predictions in Ungaged Mesinger, F., DiMego, G., Kalnay, E., Mitchell, K., Shafran, P.C., Ebisuzaki, W., Jovic, D.,
Basins: Promise and Progress (Proceedings of Symposium S7 Held During the Woollen, J., Rogers, E., Berbery, E.H., Ek, M.B., Fan, Y., Grumbine, R., Higgins, W.,
Seventh IAHS Scientific Assembly at Foz do Iguacu, Brazil, April 2005. IAHS Li, H., Lin, Y., Manikin, G., Parrish, D., Shi, W., 2006. North American Regional
Publication 303, pp. 47–56. Reanalysis (NARR). Bull. Am. Meteorol. Soc. 87 (3), 343–360.
Krause, P., Boyle, D.P., Bäse, F., 2005. Comparison of different efficiency criteria for Miller, D.A., White, R.A., 1998. A conterminous United States multi-layer soil
hydrological model assessment. Adv. Geosci. 5, 89–97. http://dx.doi.org/ characteristics data set for regional climate and hydrology modeling. Earth
10.5194/adgeo-5-89-2005. Interact., 2, Paper No. 2. <http://EarthInteractions.org>.
Kuligowski, R.J., 2002. A self-calibrating real-time GOES rainfall algorithm for short- Minder, J.R., Kingsmill, D.E., 2013. Mesoscale variations of the atmospheric snow-
term rainfall estimates. J. Hydrometeorol. 3, 112–130. http://dx.doi.org/ line over the northern Sierra Nevada: multi-year statistics, case study, and
10.1175/1525-7541(2002) 003<0112:ASCRTG>2.0.CO;2. mechanisms. J. Atmos. Sci. 70 (3), 916–938. http://dx.doi.org/10.1175/JAS-D-
Kuligowski, R.J., Li, Y., Zhang, Y., 2013. Impact of TRMM data on a low-latency, high- 12-0194.1.
resolution precipitation algorithm for flash-flood forecasting. J. Appl. Met. Clim. Minder, J.R., Durran, D.R., Roe, G.H., 2011. Mesoscale controls on the mountainside
52, 1379–1393. http://dx.doi.org/10.1175/JAMC-D-12-0107.1. snow line. J. Atmos. Sci. 68, 2107–2127. http://dx.doi.org/10.1175/JAS-D-10-
Kuzmin, V., Seo, D.-J., Koren, V., 2008. Fast and efficient optimization of hydrologic 05006.1.
model parameters using a priori estimates and stepwise line search. J. Hydrol. Mitchell, K.E. et al., 2004. The multi-institutional North American Land Data
353, 109–128. Assimilation System (NLDAS): utilizing multiple GCIP products and partners in
Le Moine, N., 2008. The Surface Catchment Seen from Underground: A Way to a continental distributed hydrological modeling system. J. Geophys. Res. 109,
Improve the Performance and Realism of Rainfall-runoff Models? PhD Thesis, D07S90. http://dx.doi.org/10.1029/2003JD003823.
University Pierre et Marie Curie (Paris), CEMAGREF (Antony), 324pp. <http:// Mizukami, N., Koren, V., 2008. Methodology and evaluation of melt factor
webgr.irstea.fr/wp-content/uploads/2012/07/2008-LE_MOINE-THESE.pdf>. parameterization for distributed SNOW-17. In: Proceedings of the American
Leavesley, G.H., Hay, L.E., Viger, R.J., Marskstrom, S.L., 2003. Use of a priori Geophysical Union, Fall Meeting 2008, Abstract #H31J-08.
parameter-estimation methods to constrain calibration of distributed- Mizukami, N., Smith, M., 2012. Analysis of inconsistencies in multi-year high space-
parameter models. In: Duan et al. (Eds.), Calibration of Watershed Models. time resolution precipitation grids over complex terrain and impacts on
Water Science and Application. AGU Press, pp. 255–266. hydrologic modeling. J. Hydrol. 428–429, 129–141. http://dx.doi.org/10.1016/
Lei, F., Koren, V., Smith, M., Moreda, F., 2007. A sensitivity study of an energy-budget j.jhydrol.2012.01.030.
snow accumulation and ablation model. In: Proceedings of the Paper J6.4, 87th Mizukami, N., Koren, V., Smith, M., Kingsmill, D., Zhang, Z., Cosgrove, B., Cui, Z.,
Annual Meeting of the AMS, 21st Conference on Hydrology, 13–18 January. 2013. The impact of precipitation type discrimination on hydrologic simulation:
Lettenmaier, D.P., Gan, T.Y., 1990. Hydrologic sensitivities of the Sacramento-San rain–snow partitioning derived from HMT-West radar-retected brightband
Joaquin River basin, California, to global warming. Water Resour. Res. 26 (1), height versus surface temperature data. J. Hydrometeorol 14, 1139–1158.
69–86. http://dx.doi.org/10.1175/JHM-D-12-035.1.
Leydecker, A., Sickman, J.O., Melack, J.M., 2001. Spatial scaling of the hydrological Molotch, N.P., Bales, R.C., 2006. SNOTEL representativeness in the Rio Grande
and biochemical aspects of high-altitude catchments in the Sierra Nevada, headwaters on the basis of physiographics and remotely sensed snow cover
California, USA. Arct. Antarct. Alp. Res. 33 (4), 391–396. persistence. Hydrol. Process. 20, 723–739. http://dx.doi.org/10.1002/hyp.6128.
Liston, G.E., Hiemstra, C.A., Elder, K., Cline, D.W., 2008. Mesocell study area snow Moreda, F., Cong, S., Schaake, J., Smith, M., 2006. Gridded rainfall estimation for
distributions for the cold land processes experiment (CLPX). J. Hydrometeorol. distributed modeling in western mountainous areas. In: Poster H23A, Spring
9, 957–976. http://dx.doi.org/10.1175/2008JHM869.1. Meeting of the AGU, May 23–27, Baltimore, MD.
Liu, Z., Todini, E., 2001. Towards a comprehensive physically-based rainfall-runoff Moriasi, D.N., Arnold, J.G., Van Liew, M.W., Bingner, R.L., Harmel, R.D., Vieth, T.L.,
model. Hydrol. Earth Syst. Sci. 6, 859–881. 2007. Model evaluation guidelines for systematic quantification of accuracy in
Lohmann, D., Lettenmaier, D.P., Liang, X., Wood, E.F., Boone, A., Chang, S., Chen, F., watershed simulations. Trans. ASABE 50 (3), 885–900.
Dai, Y., Desborough, C., Dickinson, R.E., Duan, Q., Ek, M., Gusev, Y.M., Habets, F., Nelson, B.R., Seo, D.–J., Kim, D., 2010. Multisensor precipitation reanalysis. J.
Irannejad, P., Koster, R., Mitchell, K.E., Nasonova, O.N., Noilhan, J., Schaake, J., Hydrometeorol. 11, 666–682. http://dx.doi.org/10.1175/2010JHM1210.1.
Schlosser, A., Shao, Y., Shmakin, A.B., Verseghy, D., Warrach, K., Wetzel, P., Xue, National Research Council (NRC), 1998. Future of the National Weather Service
Y., Yang, Z.-L., Zeng, Q.C., 1998. The project for intercomparison of land-surface Cooperative Observer Network. National Weather Service Modernization
parameterization schemes (PILPS) phase 2(c) Red–Arkansas River basin Committee, Commission on Engineering and Technical Systems, National
experiment: 3. Spatial and temporal analysis of water fluxes. Global Planet. Academy Press, 73pp. <http://www.nap.edu/catalog/6197.html>.
Change 19, 161–179. Pan, M. et al., 2003. Snow process modeling in the North American Land Data
Lohmann, D., Mitchell, K.E., Houser, P.R., Wood, E.F., Schaake, J.C., Robock, A., Assimilation System (NLDAS): 2. Evaluation of model-simulated snow water
Cosgrove, B.A., Sheffield, J., Duan, Q.-Y., Luo, L., Higgins, R.W., Pinker, R.T., equivalent. J. Geophys. Res. 108 (D22), 8850. http://dx.doi.org/10.1029/
Tarpley, J.D., 2004. Streamflow and water balance intercomparisons of four land 2003JD003994.
surface models in the North American Land Data Assimilation System project, J. Pandey, G.R., Cayan, D.R., Georgakakos, K.P., 1999. Precipitation structure in the Sierra
Geophys. Res. 109, D07S91. http://dx.doi.org/10.1029/2003JD003517. Nevada of California during winter. J. Geophys. Res. 104 (D10), 12019–12030.
Lundquist, J.D., Cayan, D.R., 2002. Seasonal and spatial patterns in diurnal cycles in Pandey, G.R., Cayan, D.R., Dettinger, M.D., Georgakakos, K.P., 2000. A hybrid
streamflow in the western United States. J. Hydrometeorol. 3, 591–603. orographic plus statistical model for downscaling daily precipitation in
Lundquist, J.D., Loheide, S.P., 2011. How evaporative water losses vary between wet Northern California. J. Hydrometeorol. 1 (December), 491–506.
and dry water years as a function of elevation in the Sierra Nevada, California, Parton, W.J., Logan, J.A., 1981. A model for diurnal variation in soil and air
and critical factors for modeling. Water Resour. Res. 47, W00H09. http:// temperature. Agric. Meteorol. 23, 205–216.
dx.doi.org/10.1029/2010WR010050. Perrin, C., Michel, C., Andréassian, V., 2003. Improvement of a parsimonious model
Lundquist, J.D., Neiman, P.J., Martner, B., White, A.B., Gottas, D.J., Ralph, F.M., 2008. for streamflow simulation. J. Hydrol. 279, 275–289. http://dx.doi.org/10.1016/
Rain versus snow in the Sierra Nevada, California: Comparing Doppler profiling S0022-1694%2803%2900225-7.
radar and surface observations of melting level. J. Hydrometeorol. 9, 194– Perrin, C., Andréassian, V., Michel, C., 2006. Simple benchmark models as a basis for
211. criteria of model efficiency. Archiv für Hydrobiologie 17 (1–2), 221–244. http://
Mathevet, T., 2005. Which Lumped Rainfall-runoff Models at the Hourly Time Step? dx.doi.org/10.1127/lr/17/2006/221 (Supplement 161/1-2, Large Rivers).
Empirical Development and Intercomparison of Models on a Large Set of Puspalatha, R., Perrin, C., Le Moine, N., Andréassian, V., 2012. A review of efficiency
Catchments. PhD Thesis, ENGREF (Paris), Cemagref (Antony), France, 463pp. criteria suitable for evaluating low-flow simulations. J. Hydrol. 420–421, 171–
<http://webgr.irstea.fr/wp-content/uploads/2012/07/2005-MATHEVET- 182. http://dx.doi.org/10.1016/j.jhydrol.2011.11.055.
THESE.pdf>. Ralph, F.M., Dettinger, M.D., 2011. Storms, floods, and the science of atmospheric
Matrosov, S.Y., Clark, K.A., Kingsmill, D.E., 2007. A polarimetric radar approach to rivers. Eos Trans. Amer. Geophys. Union 92 (32), 265–272.
identify rain, melting-layer, and snow regions for applying corrections to Ralph, F.M. et al., 2005. Improving short-term (0–48 h) cool-season quantitative
vertical profiles of reflectivity. J. Appl. Meteorol. Climatol. 46, 154–166. http:// precipitation forecasting: recommendations from a USWRP workshop. Bull. Am.
dx.doi.org/10.1175/JAM2508.1. Meteorol. Soc. 86, 1619–1632.
Maurer, E.P., Mass, C., 2006. Using radar data to partition precipitation into rain and Reed, S., 2003. Deriving flow directions for coarse resolution (1–4 km) gridded
snow in a hydrologic model. J. Hydrol. Eng. 11 (3), 214–221. hydrologic modeling. Water Resour. Res. 39 (9), 4-1–4-11.
Reed, S.M., Maidment, D.R., 1999. Coordinate transformations for using NEXRAD Tsintikidis, D., Georgakakos, K.P., Sperfslage, J.A., Smith, D.E., Carpenter, T.M., 2002.
data in GIS-based hydrologic modeling. J. Hydrol. Eng. 4 (174). http://dx.doi.org/ Precipitation uncertainty and rain gage network design within the Folsom Lake
10.1061/(ASCE)1084-0699(1999)4:2(174). Watershed. ASCE J. Hydrol. Eng. 7 (2), 175–184.
Reed, S., Koren, V., Smith, M., Zhang, Z., Moreda, F., Seo, D.-J., Participants, D.M.I.P., Valéry, A., 2010. Modélisation précipitations – débit sous influence nivale.
2004. Overall distributed model intercomparison project results. J. Hydrol. 298 Élaboration d’un module neige et évaluation sur 380 bassins versants. Thèse
(1–4), 27–60. de Doctorat, Cemagref (Antony), AgroParisTech (Paris), 405pp. <http://
Reek, T., Doty, S.R., Owen, T.W., 1992. A deterministic approach to the validation of webgr.irstea.fr/wp-content/uploads/2012/07/2010-VALERY-THESE.pdf>.
historical daily temperature and precipitation data from the cooperative Valéry, A., Andréassian, V., Perrin, C., 2009. Inverting the hydrologic cycle: When
network. Bull. Am. Meteorol. Soc. 73 (6), 753–762. http://dx.doi.org/10.1175/ streamflow measurements help assess altitudinal precipitation gradients in
1520-0477(1992)073<0753:ADATTV>2.0.CO;2. mountain areas. IAHS-AISH Publication 333, pp. 281–286.
Reynolds, D.W., Dennis, A.S., 1986. A review of the Sierra Cooperative Pilot Project. Vélez, J.J., Puricelli, M., López-Unzu, F., Francés, F., 2009. Parameter extrapolation to
Bull. Am. Meteorol. Soc. 67 (5), 513–523. ungaged basins with a hydrological distributed model in a regional framework.
Rice, R., Bales, R.C., 2010. Embedded-sensor network design for snow cover Hydrol. Earth Syst. Sci. 13 (2), 229–246.
measurements around snow pillow and snow course sites in the Sierra Viney, N.R., Bormann, H., Breuer, L., Bronstert, A., Croke, B.F.W., Frede, H., et al.,
Nevada of California. Water Resour. Res. 46, W03537. http://dx.doi.org/ 2009. Assessing the impact of land use change on hydrology by ensemble
10.1029/2008WR007318. modeling (LUCHEM) II: Ensemble combinations and predictions. Advances in
Ritter, A., Muñoz-Carpena, R., 2013. Performance evaluation of hydrological models: Water Resources 32 (2), 147–158. http://dx.doi.org/10.1016/
statistical significance for reducing subjectivity in goodness-of-fit assessments. j.advwatres.2008.05.006.
J. Hydrol. 480, 33–34. http://dx.doi.org/10.1016/j.jhydrol.2012.12.004. Waichler, S.R., Wigmosta, M.S., 2003. Development of hourly meteorological values
Rutter, N., Cline, D., Long, L., 2008. Evaluation of the NOHRSC Snow Model (NSM) in from daily data and significance to hydrologic modeling at H.J. Andrews
a one-dimensional mode. J. Hydrometeorol. 9, 695–711. Experimental Forest. J. Hydrometeorol. 4, 251–263.
Rutter, N., Essery, R., Pomeroy, J., Koren, V., et al., 2009. Evaluation of forest snow Walter, M.T., Brooks, E.S., McCool, D.K., King, L.G., Molnau, M., Boll, J., 2005. Process-
processes models (SnowMIP2). J. Geophys. Res. 114, 1–5. http://dx.doi.org/ based snowmelt modeling: does it require more input data than temperature-
10.1029/2008JD011063. index modeling? J. Hydrol. 300, 65–75. http://dx.doi.org/10.1016/
Schaake, J.C., Cong, S., Duan, Q., 2006. The U.S. MOPEX Data Set, vol. 307. IAHS j.jhydrol.2004.05.002.
Publication, pp. 9–28. Wang, J., Georgakakos, K.P., 2005. Validation and sensitivities of dynamic
Schaefli, B., Gupta, H.V., 2007. Do Nash values have value? Hydrol. Process. 21, precipitation simulation of winter events over the Folsom Lake Watershed:
2075–2080. http://dx.doi.org/10.1002/hyp.6825. 1964–1999. Mon. Weather Rev. 133, 3–19.
Schmidt, J.A., Anderson, A.J., Paul, J.H., 2007. Spatially-variable, physically-derived Wayand, N., Hamlet, A., Hughes, M., Feld, S., Lundquist, J., 2013. Intercomparison of
flash flood guidance. In: Paper 6B.2, Proceedings of the 21st Conference on meteorological forcing data from empirical and mesoscale model sources in the
Hydrology, 87th Meeting of the AMS, San Antonio, Texas, January 15–18. N.F. American River Basin in northern Sierra Nevada, California. J.
Seibert, J., 2001. On the need for benchmarks in hydrological modeling. Hydrol. Hydrometeorol. 14, 677–699. http://dx.doi.org/10.1175/JHM-D-12-0102.1.
Process. 15 (6), 1063–1064. Westrick, K.J., Mass, C.J., 2001. An evaluation of a high-resolution
Seo, D.-J., 1998. Real-time estimation of rainfall fields using rain gage data under hydrometeorological modeling system for prediction of a cool-season flood
fractional coverage conditions. J. Hydrol. 208, 25–36. event in a coastal mountainous watershed. J. Hydrometeorol. 2, 161–180.
Serreze, M.C., Clark, M.P., Armstrong, R.L., McGinnis, D.A., Pulwarty, R.S., 1999. Westrick, K.J., Mass, C.F., Colle, B.A., 1999. The limitations of the WSR-88D radar
Characteristics of the western United States snowpack from snowpack network for quantitative precipitation measurement over the coastal Western
telemetry (SNOTEL) data. Water Resour. Res. 35 (7), 2145–2160. United States. Bull. Am. Meteorol. Soc. 80 (11), 2289–2298.
Shamir, E., Georgakakos, K.P., 2006. Distributed snow accumulation and ablation Westrick, K.J., Storck, Pl., Mass, C.J., 2002. Description and evaluation of a
modeling in the American River Basin. Adv. Water Resour. 29, 558–570. hydrometeorological forecast system of mountainous watersheds. Weather
Shao, Y., Henderson-Sellers, A., 1996. Modeling soil moisture: a project for Forecast. 17, 250–262.
intercomparison of land surface parameterization schemes Phase 2(b). J. White, A.B., Gottas, D.J., Strem, E.T., Ralph, F.M., Neiman, P.J., 2002. An automated
Geophys. Res. 101 (D23), 7227–7250. brightband height detection algorithm for use with Doppler radar spectral
Simpson, J.J., Dettinger, M.D., Gehrke, F., McIntire, T.J., Huffurd, G.L., 2004. moments. J. Atmos. Oceanic Technol. 19, 697.
Hydrologic scales, cloud variability, remote sensing, and models: implications White, A.B., Gottas, D.J., Henkel, A.F., Neiman, P.J., Ralph, F.M., Gutman, S.I., 2010.
for forecasting snowmelt and streamflow. Weather Forecast. 19 (April), 251– Developing a performance measure for snow-level forecasts. J. Hydrometeorol.
276. 11, 739–753.
Slater, A.G., Schlosser, C.A., et al., 2001. The representation of snow in land surface Wigmosta, M.S., Vail, L., Lettenmaier, D.P., 1994. A distributed hydrology-vegetation
schemes: results from PILPS 2(d). J. Hydrometeorol. 2, 7–25. model for complex terrain. Water Resour. Res. 30, 1665–1679.
Smith, M.B., Laurine, D.P., Koren, V.I., Reed, S.M, Zhang, Z., 2003. Hydrologic model WMO, 1986a. Intercomparison of Models of Snowmelt Runoff. Operational
calibration in the National Weather Service. In: Duan et al. (Eds.), Calibration of Hydrology Report No. 23. World Meteorological Organisation, Geneva.
Watershed Models: Water Science and Applications. AGU Press, pp. 133–152. WMO, 1986b. Results of an Intercomparison of Models of Snowmelt Runoff, IAHS-
Smith, M.B., Seo, D.-J., Koren, V.I., Reed, S., Zhang, Z., Duan, Q.-Y., Moreda, F., Cong, S., AISH Publication 155, pp. 103–112.
2004. The distributed model intercomparison project (DMIP): motivation and Wood, E., Lettenmaier, D.P., Liang, X., Lohmann, D., Boone, A., Chang, S., Chen, F., Dai,
experiment design. J. Hydrol. 298 (1–4), 4–26. Y., Dickinson, R.E., Duan, Q., Ek, M., Susev, Y.M., Habets, F., Irannejad, P., Koster,
Smith, M.B., Koren,V., Reed, S., Zhang, Z., Seo, D.-J., Moreda, F., Cui, Z., 2006. The R., Mitchel, K.E., Nasonova, O.N., Noilhan, J., Schaake, J., Schlosser, A., Shao, Y.,
Distributed Model Intercomparison Project: Phase 2 Science Plan. <http:// Shmakin, A.B., Verseghy, D., Warrach, K., Wetzel, P., Xue, Y., Yang, Z.-L., Zeng, Q.,
www.nws.noaa.gov/oh/hrl/dmip/2/docs/ 1998. The project for intercomparison of land surface parameterization
dmip_2_plan_march10_06_update.pdf>. schemes (PILPS) Phase 2(c) Red-Arkansas River basin experiment: 1.
Smith, M., Koren, V., Reed, S., Zhang, Z., Cui, Z., Mizukami, N., Moreda, F., Sheldon, S., Experiment description and summary intercomparisons. Global Planet.
2009. Preliminary results of phase 2 of the distributed model intercomparison Change 19, 115–136.
project. In: Presentation 4.1, 23rd Conference on Hydrology, 89th Meeting of Yang, D., Goodison, B.E., Metcalfe, J.R., Golubevy, V.S., Bates, R., Pangburn, T.,
the AMS, Phoenix, Arizona, January 10–15. <https://ams.confex.com/ams/ Hanson, C.L., 1998a. Accuracy of NWS 800 standard non-recording precipitation
89annual/techprogram/paper_148995.htm>. gage: results and application of WMO intercomparisons. J. Atmos. Oceanic
Smith, M., Ding, F., Cui, Z., Koren, V., Mizukami, N., Zhang, Z., Cosgrove, B., Kitzmiller, Technol. 15, 54–68.
D., Schaake, J., 2010. Development of gridded QPE datasets for mountainous Yang, D., Goodison, B.E., Ishida, S., 1998b. Adjustment of daily precipitation data at
area distributed hydrologic modeling. In: Presented at the ASCE EWRI 10 climate stations in Alaska: application of world meteorological organization
Conference, Rhode Island, May 2010. intercomparison results. Water Resour. Res. 34 (2), 241–256.
Smith, M.B., Koren, V., Reed, S., Zhang, Z., Zhang, Yu., Moreda, F., Cui, Z., Mizukami, Young, C.B., Bradley, A.A., Krajewski, W.F., Kruger, A., 2000. Evaluating NEXRAD
N., Anderson, E.A., Cosgrove, B.A., 2012a. The distributed model multi-sensor precipitation estimates for operational hydrologic forecasting. J.
intercomparison project – Phase 2: motivation and design of the Oklahoma Hydrometeorol. 1, 241–254.
experiments. J. Hydrol. 418–419, 3–16. http://dx.doi.org/10.1016/ Zamora, R.J., Ralph, F.M., Clark, E., Schneider, T., 2011. The NOAA hydrometeorology
j.jhydrol.2011.08.055. testbed soil moisture observing networks: design, instrumentation, and
Smith, M., Koren, V., Zhang, Z., Zhang, Y., Reed, S., Cui, Z., Moreda, F., Anderson, E., preliminary results. J. Atmos. Oceanic Technol. 28, 1129–1140. http://
Cosgrove, B., Mizukami, N., DMIP 2 Participants, 2012b. Overall results of the dx.doi.org/10.1175/2010JTECHA1465.1.
DMIP 2 Oklahoma experiments. J. Hydrol. 418–419, 17–48. http://dx.doi.org/ Zhang, Y., Zhang, Z., Reed, S., Koren, V., 2011. An enhanced and automated approach
10.1016/j.jhydrol.2011.08.056. for deriving a priori SAC-SMA parameters from the Soil Survey Geographic
Timbal, B., Henderson-Sellers, A., 1998. Intercomparisons of land-surface Database. Comput. Geosci. 37 (2), 219–231. http://dx.doi.org/10.1016/
parameterizations coupled to a limited area forecast model. Global Planet. j.cageo.2010.05.016.
Change 19, 247–260. Zhang, Z., Koren, V., Reed, S., Smith, M., Zhang, Y., Moreda, F., Cosgrove, B.A., 2012.
Todini, E., Ciarapica L., 2002. The TOPKAPI model. In: Singh, V.P., Frevert, D.K. (Eds.), SAC-SMA a priori parameter differences and their impact on distributed
Mathematical Models of Large Watershed Hydrology. Water Resources (chapter hydrologic model simulations. J. Hydrol. 420–421, 216–227. http://dx.doi.org/
12). 10.1016/j.jhydrol.2011.12.004.

DMIP 2 West Final 2013 - Journal of Hydrology

Uploaded by

Copyright:

Available Formats

DMIP 2 West Final 2013 - Journal of Hydrology

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

DMIP 2 West Final 2013 - Journal of Hydrology

Uploaded by

Copyright:

Available Formats

This article appeared in a journal published by Elsevier.

Journal of Hydrology 507 (2013) 300–329

Contents lists available at ScienceDirect

The distributed model intercomparison project – Phase 2: Experiment

0022-1694/$ - see front matter Published by Elsevier B.V.

M. Smith et al. / Journal of Hydrology 507 (2013) 300–329 301

1. Introduction 1.2. Science questions

302 M. Smith et al. / Journal of Hydrology 507 (2013) 300–329

M. Smith et al. / Journal of Hydrology 507 (2013) 300–329 303

304 M. Smith et al. / Journal of Hydrology 507 (2013) 300–329

For consistency with the results of the DMIP 1 and 2 experi-

M. Smith et al. / Journal of Hydrology 507 (2013) 300–329 305

USGS Stream Gauge CMEC1

306 M. Smith et al. / Journal of Hydrology 507 (2013) 300–329

M. Smith et al. / Journal of Hydrology 507 (2013) 300–329 307

Basin ‘Warm Up’ Calibration Validation

2.8. Data assimilation

2.9.1. Precipitation Water Balance Analysis

308 M. Smith et al. / Journal of Hydrology 507 (2013) 300–329

M. Smith et al. / Journal of Hydrology 507 (2013) 300–329 309

310 M. Smith et al. / Journal of Hydrology 507 (2013) 300–329

3.1. Overall water balance

550 (MAP 847 mm) OHD

M. Smith et al. / Journal of Hydrology 507 (2013) 300–329 311

312 M. Smith et al. / Journal of Hydrology 507 (2013) 300–329

M. Smith et al. / Journal of Hydrology 507 (2013) 300–329 313

rmod - Calibration period

314 M. Smith et al. / Journal of Hydrology 507 (2013) 300–329

%bias - Calibration period

M. Smith et al. / Journal of Hydrology 507 (2013) 300–329 315

Event Absolute % Runoff Error

316 M. Smith et al. / Journal of Hydrology 507 (2013) 300–329

Improvement in Event Peak Flow: Distributed vs Lumped 400

Rain Snow Elevation Derived from LMP Model

-10 CEM 3000

period of December 11–14.

M. Smith et al. / Journal of Hydrology 507 (2013) 300–329 317

Dec 12, 1995 at 15Z

= OHD Snow LMP

318 M. Smith et al. / Journal of Hydrology 507 (2013) 300–329

(a) Runoff OHD

(d) Rain: OHD - LMP

M. Smith et al. / Journal of Hydrology 507 (2013) 300–329 319

scale mismatch between point SWE observations and simulations 10

320 M. Smith et al. / Journal of Hydrology 507 (2013) 300–329

Basin Station Elev. m Participant

M. Smith et al. / Journal of Hydrology 507 (2013) 300–329 321

Blue Lakes (2,456m) Ebbet’s

322 M. Smith et al. / Journal of Hydrology 507 (2013) 300–329

M. Smith et al. / Journal of Hydrology 507 (2013) 300–329 323

M. Smith et al. / Journal of Hydrology 507 (2013) 300–329

M. Smith et al. / Journal of Hydrology 507 (2013) 300–329 325

Basin Station Latitude/longitude Elev.m SNOTEL or CDEC1 ID Owner Observation time

D.1. Percent bias, PB (%)

PB is a measure of total volume difference between two time

D.4. Correlation coefﬁcient, r

326 M. Smith et al. / Journal of Hydrology 507 (2013) 300–329

M. Smith et al. / Journal of Hydrology 507 (2013) 300–329 327

328 M. Smith et al. / Journal of Hydrology 507 (2013) 300–329