Journal of Hydrology 515 (2014) 267280

Contents lists available at ScienceDirect

Journal of Hydrology
journal homepage: www.elsevier.com/locate/jhydrol

Analysis of a changing hydrologic flood regime using the Variable

Infiltration Capacity model
Daeryong Park a, Momcilo Markus b,
a
b

Civil and Environmental System Engineering, Konkuk University, 1 Hwayang-dong, Gwangjin-gu, Seoul 143-701, South Korea
Illinois State Water Survey, Prairie Research Institute, University of Illinois at Urbana-Champaign, 2204 Griffith Dr., Champaign, IL 61820, USA

a r t i c l e

i n f o

Article history:
Received 7 June 2013
Received in revised form 3 April 2014
Accepted 2 May 2014
Available online 14 May 2014
This manuscript was handled by Laurent
Charlet, Editor-in-Chief, with the assistance
of Georgia Destouni, Associate Editor
Keywords:
VIC model
Climate change
Flood frequency
Hydrologic change
Snow hydrology

s u m m a r y
The Pecatonica River and several other streams in the Wisconsin Driftless area show a decreasing trend in
annual peak flows. Previous studies of the Pecatonica River detected a significant decreasing historical
trend in late winter snowmelt-driven floods, while the rainfall-driven spring and summer flood peaks
exhibited no significant trend during the period of record. Unlike several previous studies which attribute
the decline in flood peaks mainly to changes in land management, we hypothesize that climate change
had a significant contribution to the overall decrease in flood peaks. In particular, we hypothesize that
the increase in winter temperatures caused the decrease in snow depth, which in turn resulted in a
decreasing trend in flood peaks. In an attempt to validate this hypothesis, we used long-term daily precipitation, temperature, and river flow data observed in the watershed as inputs to the Variable Infiltration Capacity (VIC) model to generate other non-monitored climatic variables. Trends in these climatic
variables were then related to the trend in flood peaks in the Pecatonica River. Due to the complexity
of the hydrologic system and numerous data and modeling-related uncertainties, the above hypothesis
cannot be validated with certainty. Nonetheless, the results in two different modes (event and continuous
simulation) provide support to the speculation that the decreasing trend in flood peaks was a result of
decreasing snow depth. The model runs resulted in a decrease in snow depths for the period of record
(19152009), increase in sublimation and evaporation, no change in base flow, and mixed results in infiltration. These analyses also suggest that VIC can be used in other similar regions in snowmelt-driven
flood peak studies. It should be recognized, however, that the success of these applications can be
severely constrained by various uncertainties, including but not limited to, the poor quality or absence
of snow depth data.
2014 Elsevier B.V. All rights reserved.

1. Introduction
Many studies on the effects of climate change on water
resources have been published in the past several years. These
studies have investigated historical trends in streamflow, precipitation, and various other variables. Among the studies related to
streamflow trends, most focus on the trends of seasonal and annual
streamflow (Regonda et al., 2005; Hamlet et al., 2005, 2007; Mishra
et al., 2010; Sinha et al., 2010). Relatively fewer studies have
focused on extreme events. Hirsch and Ryberg (2012) studied the
relationships between annual floods at 200 long-term streamgages
in the conterminous United States and the global mean carbon
dioxide concentration. They found no correlation between global
mean carbon dioxide concentration and flooding, except for a
Corresponding author. Tel.: +1 217 333 0237; fax: +1 217 333 2304.
E-mail address: mmarkus@illinois.edu (M. Markus).
http://dx.doi.org/10.1016/j.jhydrol.2014.05.004
0022-1694/ 2014 Elsevier B.V. All rights reserved.

negative relationship between GMCO2 and flood magnitudes in

the southwestern United States. Markus et al. (2007, 2012) and
Hejazi and Markus (2009) studied the frequency of flooding as a
result of the increasing intensity and frequency of heavy storms
and the increase in urbanization. Cunderlik and Ouarda (2009)
detected a significant number of stations with negative trends in
the magnitude of snowmelt-generated floods. Trends in peak
streamflows also have been studied by means of observed data
analysis (Lins and Slack, 1999; Novotny and Stefan, 2007). Lins
and Slack (1999) investigated trends in minimum, median, and
maximum annual flows at 395 daily streamflow stations for
19441993 in the United States. They found that a majority of stations exhibited a decreasing trend in annual maximum flow. However, for annual median and annual minimum flows, a majority of
the stations had an increasing trend. Regonda et al. (2005) analyzed observed weather data in the western United States and
found that peak flow timing had shifted to earlier months, caused

268

D. Park, M. Markus / Journal of Hydrology 515 (2014) 267280

by increasing spring warm-spell temperatures as well as decreasing precipitation in the Rocky Mountain area. In these studies,
trends in observed and simulated seasonal streamflows were
found to be highly correlated with trends in seasonal precipitation.
The Wisconsin Driftless Area, like surrounding areas in the
Midwestern United States, is a region with several streams showing decreasing annual maximum streamflow. Numerous studies
attempt to explain these trends. Early studies, such as Potter
(1991) and Gebert and Krug (1996), explain the trends by the
changes in land management practices. Potter (1991) argued that
the changes in hydrology are likely due to the adoption of various
measures for soil and water conservation, and are not climate
related. Gebert and Krug (1996) suggested that the decreasing
flood peaks were caused by improving hydraulic structures such
as drainage ways and stream channels, as well as an increase in
wooded areas. Knapp (2005) studied trends in annual minimum,
average, and maximum streamflows in the Midwestern United
States and detected a decrease in annual peak flows in the Pecatonica River watershed, a part of the Wisconsin Driftless area. Knapp
also found no significant correlation between trends in peaks and
trends in average annual flows in this region. Kochendorfer and
Hubbart (2010) confirmed that annual peak flows in the Wisconsin
Driftless area have been decreasing and argued that the decreasing
trend was caused by widespread land-use changes associated with
soil conservation efforts in the 19392008 period.
Some of the more recent studies on the decreasing flood peaks
in this area include climate parameters as explanatory variables.
Juckem et al. (2008) discussed a climate-related step change in
precipitation and base flow around 1970, also globally observed
as a result of the Pacific Decadal Oscillation (Mantua et al., 1997).
Juckem et al. used data from the Driftless area of Wisconsin to
show that both climate and land management changes could
change the hydrologic response of the system. They state, Climatic
change appears to control the timing and direction (increase or
decrease) of the change, while land management changes amplify
the response beyond that which can be explained by climate factors alone. Markus et al. (2013) analyzed the decreasing trend of
annual peak flows in the Pecatonica River for the 19142008
period and found that the annual peak flow time has shifted from
February and March (referred to as the late winter months) to
mostly spring or summer seasons. Markus et al. also suggested that
the overall decreasing annual peak flows could be explained by the
significant decrease in late winter flows, which was possibly
related to the increasing winter temperature. Time series of
minimum, average, and maximum daily temperature in February
(Fig. 1) illustrates this increase. Fig. 2 shows three graphs: the
annual maximum streamflow time series (Fig. 2a), the maximum
FebruaryMarch flood peaks for each water year (Fig. 2b), and
the maximum flood peaks within the remaining months

(OctoberJanuary, and AprilSeptember) for each water year

(Fig. 2c). The discharges in February and March appear to dominate
the annual trend, suggesting that the decrease in annual peaks is
primarily a result of the decrease in the late winter months.
Markus et al. (2013) further speculated that this decrease is primarily a result of the increasing winter season temperatures,
resulting in less snow accumulation and fewer frost days. The
changing hydrologic regime is best illustrated in Fig. 3, showing
the gradual change from predominantly snowmelt-driven annual
maximum flood peaks to predominantly rainfall-driven floods.
Fig. 3 shows times of flood peak in each year, expressed in days
after January 1st. The size of solid circles represents the magnitude
of the peak. The floods between days 0 (January 1st) and approximately 90 (March 31st) are typically snowmelt-driven, while later
peaks are typically caused by large rain events. The figure shows
that the snowmelt-driven floods became less frequent and less
dominant with time compared to the rainfall-driven floods
occurring later in the year. Three weighted trend lines in the figure
also show significant upward trends in average flood timing overall
and for rainfall-driven floods, and a statistically insignificant
downward trend for the snowmelt-driven floods.
We hypothesize that the flood peak decrease in the Pecatonica
River is driven in part by climate change. To test this hypothesis,
this study applied the Variable Infiltration Capacity (VIC) model,
which is a physically based land-surface model capable of simulating energy and water balance. The model simulated a number of
hydrologic and climatic variables, such as frozen soil, snow depth,
snowmelt, soil temperature, and river discharge in the Pecatonica
River watershed. A special focus was given to the late winter flood
events, as the decrease in these events was the primary reason for
the decreasing annual maximum flows in the watershed. The late
winter flood events in this watershed are typically driven by snowmelt, or a combination of snowmelt and rainfall, due to seasonal
increases in temperature to above freezing. Snowmelt simulation
using the VIC model has been analyzed by Sinha and Cherkauer
(2010), Sinha et al. (2010), Tan et al. (2011), Andreadis et al.
(2009), and others. Particularly, Feng et al. (2008) compared the
VIC and the Snow Thermal Model (SNTHERM) by Jordan (1991)
and showed a good agreement between the two models in snow
simulation. In addition, the VIC model can simulate and evaluate
both snow and frozen soil effects unlike the Snowmelt Runoff
Model (SRM; Martinec et al., 1998; Nakayama and Watanabe,
2006). These studies indicated that the VIC model has an ability
to accurately simulate watershed processes, including snow accumulation and ablation. Unlike the published applications, which
used daily step simulations to analyze monthly flows such as Dai
et al. (2004), Andreadis et al. (2005), and Sheffield and Wood
(2008), this study uses VIC to simulate daily flood peaks in two
modes, event and continuous, for late winter events to address
the decreasing flood frequency in a changing hydrologic regime.
The specific objectives of the study are to (i) Assess the applicability of the VIC model to simulate flood peaks in a snowmeltdriven hydrologic regime for a medium-sized agricultural
watershed, and (ii) Attempt to provide a climate change-related
explanation of the trend in the flood peaks in the Pecatonica River
by relating it to trends in climatic variables, both observed and
generated, using VIC for the period 19152009.

2. Study site

Fig. 1. Mean maximum, mean minimum, and mean average daily temperatures
19012008 at Darlington, Wisconsin, in February (Markus et al., 2013).

The Pecatonica River watershed has a drainage area of

3435 km2 and is located in southern Wisconsin and northern Illinois (Fig. 4). The watershed is located approximately 100 km west
of Lake Michigan, near two large coastal metropolitan areas,
Milwaukee and Chicago. This study chose five climate stations

D. Park, M. Markus / Journal of Hydrology 515 (2014) 267280

269

Fig. 2. Observed annual peak flows at Freeport streamgage (USGS No. 0543500) on the Pecatonica River between 1915 and 2009 for (a) complete dataset; (b) February and
March; (c) April through January.

observed snowfall, snow depth, rainfall, and temperature data for

19152009 based on NOAAs National Weather Service Cooperative Observer Program were used (Table 1). Some of these variables, particularly snow depth, have limited data in the earlier
parts of the record, as indicated in Table 1. The daily wind speed
data were available from 1949 to the present, and were obtained
from the National Center for Environmental Prediction National
Center for Atmospheric Research 40-year Reanalysis project,
as described by Kalnay et al. (1996), and are available at
http://www.cdc.noaa.gov/cdc/reanalysis/reanalysis.shtml. The wind
speed prior to 1949 was assumed to match the daily wind climatology based on the period after 1949, similar to the approach of
Hamlet and Lettenmaier (2005). All daily data were gridded for
the Pecatonica River watershed using the method described by
Hamlet and Lettenmaier (2005); spatial resolution of the grids
was 1/8, with grid cells of approximately 10 km  14 km or
140 km2.
Fig. 3. Annual flood peak timing (in days after January 1st) and magnitude
(represented by the size of circles) in the Pecatonica River watershed, showing a
gradual decrease in frequency and magnitude of snowmelt-driven annual flood
peaks (first 34 months in a calendar year).

(Dodgeville, Darlington, Martintown, and Monroe in Wisconsin

and Freeport in Illinois), which will be further described in the next
section. The land use in the watershed has not changed much in
the past; recently Homer et al. (2004) found that it is about 82%
agricultural, 10% forested, and 6% developed.
3. Data
3.1. Climate data
Observed climate forcing data included daily precipitation, daily
maximum and minimum temperatures, and wind speed. Daily

3.2. Soil and land cover data

Soil and land cover parameters were acquired from the Land
Data Assimilation System (LDAS; Maurer et al., 2002). Soil parameters, including infiltration shape parameter, maximum subsurface
flow rate, fraction of the maximum soil moisture, soil layer depths,
and saturated hydrologic conductivity Ksat, commonly used in VIC
(Mishra et al., 2010), were calibrated. Vegetation parameters
(architectural resistance, albedo, minimum stomata resistance, leaf
area index, zero-plane displacement, vegetation roughness length,
fraction of root depth of each soil layer) were represented to subgrid variability with a factional coverage area (Mishra et al., 2010).
They were obtained from the same LDAS dataset. Both soil and vegetation parameters were developed at 1/8 spatial resolution.
Monthly leaf area index (LAI) data for each vegetation type

composed of 1/8 spatial resolution were adapted from the method

developed by Myneni et al. (1997).
3.3. Daily streamflow data
Observed daily streamflow data from 1915 to 2009 were
obtained from the gage operated by the United States Geological
Survey (USGS) at Freeport, Illinois (No. 0543500), with a drainage
area of 3435 km2 (Fig. 4). This gaging station is also operated in
cooperation with the Illinois State Water Survey, University of
Illinois, and maintained by the USGS Illinois Water Science Center.
4. Variable Infiltration Capacity Model
The VIC model was originally developed jointly at the University of Washington and Princeton University. It is a macroscale
hydrology model used to simulate various hydrologic variables as
well as kinetic energy variables such as soil moisture, snowmelt,

274

D. Park, M. Markus / Journal of Hydrology 515 (2014) 267280

data were available only at Freeport, and for 1929 at Darlington.

Data were available at most of the sites for later events. Using
snow depth data with a limited spatial extent made the calibrations more uncertain. Later decades had more data and potentially
more accurate calibrations. Overall, the VIC model produced a reasonable simulation accuracy of snow depths, below 2530 cm, and
underestimated the observed snow depths exceeding that level,
similar to Mishra and Cherkauer (2011). Possible explanations for
these results could be found in the calibration criteria. We calibrated the parameters to minimize the difference between the
observed and model-generated snow and discharge in a spatially
and temporally lumped fashion. Each observation had equal
weight. Extreme values, although critical, might not have carried
sufficient weights due to their infrequent occurrence in a time series of the observed data, suggesting that the calibration was dominated by the medium-range observations. Similarly, the
parameters were spatially averaged, possibly reducing the effects
of the highest observations.
Table 4 shows peak streamflow times of daily observed data,
the VIC model, and the corresponding NSE indexes. The VIC-based
peak times generally match the observed data. One-half of the simulated hydrographs matched the observed peak times, while the

other half were within one (1975, 1985, and 1997) or two days
(1959 and 1963). NSE for calibration ranged between 0.5 and
0.85. Fig. 7 shows the observed and simulated peaks and volumes
(represented as a sum of daily discharges for each event). The figure shows a very good agreement between these variables, particularly between the model and observed peak discharges with a
RMSE of 20.7 m3/s. In terms of volumes, except for three events,
in which the model underestimated the observed volumes by
approximately 2530%, the simulations are reasonably accurate,
within 15% of the observations. The overall accuracy for volume
simulations is RMSE = 658.7 m3/s, which is approximately 20% of
the average event volume.
The snow roughness parameter was also allowed to vary to
achieve best possible simulation accuracy. The time series of these
best-fit snow roughness parameters exhibited an increasing trend
with a 95% statistical significance (Fig. 8).

Table 4
Peak time and NSE between observed and VIC streamflows.
Year

1916
1929
1937
1948
1959
1963
1975
1985
1997
2005

Peak time

NSE

Observed

VIC

3/28/1916
3/16/1929
3/8/1937
3/1/1948
4/5/1959
3/21/1963
3/25/1975
2/27/1985
2/22/1997
2/16/2005

3/28/1916
3/16/1929
3/8/1937
3/1/1948
4/3/1959
3/19/1963
3/24/1975
2/26/1985
2/23/1997
2/16/2005

0.95
0.85
0.75
0.69
0.88
0.65
0.86
0.65
0.61
0.55

Fig. 8. Calibrated snow roughness for event simulations.

Fig. 7. Event simulation: observed and VIC model simulated peak and cumulative discharges at Freeport on the Pecatonica River for 10 largest late winter events given in
Table 4.

D. Park, M. Markus / Journal of Hydrology 515 (2014) 267280

275

Fig. 9. Continuous simulation: Observed and VIC model simulated snow depth for February and March, for (a) Dodgeville, Wisconsin; (b) Darlington, Wisconsin;
(c) Martintown, Wisconsin; (d) Monroe, Wisconsin; and (e) Freeport, Illinois. Each point represents one day.

Fig. 10. Continuous simulation: observed and VIC model simulated peak and cumulative discharges at Freeport for each late winter event 19152009. Solid points show the
years also used in the event simulation approach.

276

D. Park, M. Markus / Journal of Hydrology 515 (2014) 267280

8.2. Continuous simulation

The scatter plot of observed vs. simulated snow depth (Fig. 9)
shows that the simulation accuracy varies with month (February
and March) and stations. The simulation accuracy appears reasonable on average, with some underestimation of the highest snow
and overestimation of the lowest ones. The simulated snow depths
in March are consistently more accurate than those in February.
Also, the simulations for Darlington and Freeport stations (Fig. 9b
and e) are more accurate with lower RMSE than for the remaining
sites.
The continuous approach produced less accurate flood peak
simulations compared with the event approach. Fig. 10 (unlike
Fig. 7 for event calibration) shows a large variability in predicted
flood peaks and volumes. The solid circles show the years used in
the event calibration. The reason for larger errors is the difference
in calibrations. In the event approach, each event was calibrated
separately, while one set of parameters (except for snow roughness

Fig. 11. Calibrated snow roughness for continuous simulation.

parameter changing every five years) was used for the entire
record in the continuous approach. More importantly, the calibration objective in the event calibration was to match the flood
peaks, while in the continuous approach the goal was to match
snow depths, without regards to the flood peaks.
Similar to the event simulation (Fig. 8), the snow roughness
parameter for the continuous simulation (Fig. 11) exhibited an
increasing temporal trend with a 95% statistical significance based
on the Kendall Tau test. Fig. 11 shows the calibrated values for
snow roughness parameter for each five-year period in the continuous simulation. Despite different datasets, objective functions,
and calibration approaches, both event and continuous calibrations
resulted in similar (increasing) temporal trends for the snow
roughness parameter.

9. Discussion
9.1. Snow monitoring and modeling uncertainty
Accurate monitoring of snow depth is of great importance in
calibration of both the continuous and the event simulations presented in this study. The estimates of snow accumulation and ablation rates, however, can be biased (Varhola et al., 2010). Point
measurements may not be representative of the non-uniform spatial snow coverage, particularly in larger areas. Also, the temporally
discrete, typically daily snow depth measurements do not capture
sub-daily snow variability, caused mainly by temperature and
wind, and manifested through snow melting, compacting, drifting,
or blowing.
These issues could be partially addressed through a process of
data quality control and assurance. For example, to assess the suitability of climate gages for use in trend analysis, Kunkel et al.
(2007) suggest that a careful analysis of station histories and regional comparisons should be required for each gage. Such analyses
are routinely and thoroughly carried out at the Midwestern Regional Climate Center (MRCC) in Champaign, Illinois, resulting in
assessments of accuracy for each station in this study. Despite
the quality assurance and the long-term homogeneity of snow

Fig. 12. Model parameters used in event (represented by circles) and continuous (represented by lines) simulations.

D. Park, M. Markus / Journal of Hydrology 515 (2014) 267280

Fig. 13. Temporal trends in three observed climate variables in the Pecatonica River watershed.

Fig. 14. Temporal trends in eight simulated climate variables by event simulation in the Pecatonica River watershed.

277

278

D. Park, M. Markus / Journal of Hydrology 515 (2014) 267280

depth data, simulation of snow depths remains to be highly uncertain. Fig. 12 shows the ranges of eight parameters, calibrated in
both event (top chart) and continuous simulation (bottom chart).
Most of parameters are consistent between the two approaches,
except for Exp1 and Exp2, which appear more sensitive to the
observed datasets. Nonetheless, the ranges for the correlation
coefficient between the observed and simulated snow depth was
0.40.7 for the continuous simulation and 0.40.9 for the event
simulation, which were comparable to similar studies (e.g.,
Mishra and Cherkauer, 2011).
Additional uncertainties come from model selection and parameter estimation. Essery et al. (2009) described a comparative study
of 33 models, including VIC, for simulations of surface and energy
mass balances. They outlined numerous modeling uncertainties,
including those based on model selection. They also found that
uncertainties in parameter selection overwhelm deficiencies in
model structure when calibration data are not available. This
study, however, had somewhat incomplete datasets in the first
part of the record, but the datasets in the final decades were more
complete.
The VIC model runs in this study allowed the snow roughness
parameter to vary with time periods, every 5 years for continuous
simulation and every 10 years for event simulations. This resulted
in more accurate simulations. It is possible that smaller increments
(e.g. 1 year) for continuous calibration of roughness would have
produced more accurate simulations, but the parameter optimizations and calculations were not performed for each year due large
computational requirements. Regardless, the best-fit roughness
parameter exhibited a significant increasing temporal trend in
both event and continuous simulations. It can be speculated that

the long-term trend in the best-fit roughness parameter is related

to the long-term increase in air temperature. However, more
research, modeling and additional monitoring data would be
needed to determine if and how the increase in this parameter
relates to the changes in other climate variables.
9.2. Temporal trends in hydroclimatic variables
Figs. 1315 show the trends in climate variables in the Pecatonica River watershed, including the observed (Fig. 13), event simulation (Fig. 14), and continuous simulation (Fig. 15) approaches.
Snowfall and precipitations are presented as cumulative values
and all other variables are presented as daily average values for
FebruaryMarch period. Fig. 13 shows three observed variables
(air temperature, snowfall, and total precipitation). Figs. 14 and
15 show eight simulated variables (bare soil temperature, infiltration, evaporation, sublimation, frost days, snow depth, snowmelt,
and SWE). These results show a high degree of consistency for
the two approaches having different data, calibration methods,
and objective functions. Table 5 shows the Kendall-Tau trend test
(Kendall, 1955; Helsel and Hirsch, 1995), the Theil-Sen trend test
(Sen, 1968; Kumar et al., 2009), the slope of the best-fit line, and
the R2 of the best-fit line for each variable in Figs. 14 and 15. Traditionally, the trends would be considered statistically significant,
if the significance levels exceed 95% or 99%. However, some of the
statistically insignificant trends may have serious environmental
consequences. Several studies show that the natural system is very
sensitive to small air temperature increases, such as 1 C per
100 years (National Research Council, 2011; Zhang et al., 2001),
which can cause significant changes in hydrologic and ecologic

Fig. 15. Temporal trends in eight simulated climate variables by continuous simulation approach in the Pecatonica River watershed.

279

D. Park, M. Markus / Journal of Hydrology 515 (2014) 267280

Table 5
Trend analysis results for three observed (in italics) and eight model-generated hydroclimatic variables.
Variable

Event simulation
Kendall
trend
coeff.

Air temperature (C)

+0.511
Snowfall (mm)
0.111
Precipitation (mm)
+0.244
Bare soil temperature (C) +0.333
Infiltration (mm)
+0.244
Evaporation (mm)
+0.156
Sublimation (mm)
+0.289
Frost days (day)
0.311
Snow depth (cm)
0.467
Snowmelt (mm)
0.200
SWE (mm)
0.378

Continuous simulation

Statistical
significance
of trend (%)

Theil-Sen Statistical
estimator significance
of trend (%)

Slope of
the
best-fit
line

R of
the
best-fit
line

Kendall
trend
coeff.

Statistical
significance
of trend (%)

Theil-Sen Statistical
estimator significance
of trend (%)

Slope of
the
best-fit
line

R2 of
the
best-fit
line

95.0
<70.0
<70.0
80.0
<70.0
<70.0
71.7
75.7
92.6
<70.0
84.8

0.031
0.039
0.407
0.037
0.006
0.001
0.003
0.094
0.105
0.007
0.266

+0.027
0.008
+0.441
+0.033
+0.014
+0.001
+0.002
0.093
0.106
0.003
0.264

0.480
0.013
0.178
0.312
0.219
0.117
0.123
0.148
0.292
0.034
0.242

+0.070
0.061
+0.050
+0.111
0.067
+0.102
+0.233
0.156
0.236
0.240
0.232

<70.0
<70.0
<70.0
88.6
<70.0
85.3
99.9
97.3
99.9
99.9
99.9

0.008
0.057
0.105
0.018
0.002
0.001
0.002
0.079
0.036
0.007
0.069

+0.010
0.006
+0.162
+0.016
0.001
+0.001
+0.004
0.075
0.049
0.007
0.111

0.016
0.009
0.014
0.024
0.002
0.026
0.169
0.052
0.041
0.088
0.028

92.6
<70.0
<70.0
89.3
<70.0
<70.0
71.7
<70.0
92.6
<70.0
89.3

processes. Due to the potential importance of this statistically

insignificant but very important change observed in the Pecatonica
River watershed, the results should be interpreted in a less rigid
fashion. For that reason the table shows trends with significance
levels ranging from less than 70% to 99.9%.
Table 5 shows that the event and continuous simulations for 10
of 11 trends have identical trend signs. The only exception is the
time series for infiltration for the two approaches. The KendallTau and the Theil-Shen tests produced very consistent results.
None of the changes in infiltration exceeded statistical significance
of 70%. While precipitation has been increasing, all four snowrelated variables (snowfall, snow depth, snowmelt, and SWE) have
been decreasing for both event and continuous simulations. Also,
for both the event and continuous simulations, the increasing
trends in sublimation appear to be partly responsible for the
decreasing snow depth, which in turn could result in decreasing
flood peaks. Variability in most of these processes, and thus their
trends, could be attributed to an increasing air temperature. In
conclusion, these results indicate that the VIC model is capable of
simulating snowmelt-runoff hydrologic processes in a changing
hydrologic regime and serving as a tool which can be used to
detect trends in hydroclimatic variables. These trends, such as
the decreasing trends in snow depth, and the increasing trends in
sublimation, for both event or continuous simulations, were consistent with the decreasing trend in flood peaks, potentially suggesting that the contribution of climate change to the decreasing
trend in flood peaks in the Pecatonica River is significant.

10. Conclusions
This research tested the applicability of the Variable Infiltration
Capacity (VIC) model in a flood study in a changing hydrologic
regime. For this task, two modes of model applications, event
and continuous simulations, were designed and implemented.
The event simulation mode tested the models ability to simulate
flood events accurately and also to provide an insight in the
changes in various hydroclimatic variables during the flood event,
while the continuous simulation was designed to provide overall
watershed conditions during the flood season.
The model was applied to the Pecatonica River at Freeport in
Illinois. This river has been known for its decreasing annual flood
peaks based on 95 years of monitoring data. The results indicate
that climate change may have contributed to this trend, and that
the key change in the watershed occurred in the late winter during
the snowmelt. The climate variables produced by both event and
continuous simulations exhibited trends in hydroclimatic variables
from more favorable to snowmelt flooding early in the record, to

<70.0
<70.0
<70.0
89.9
<70.0
84.7
99.9
97.0
99.5
99.9
99.2

less favorable to snowmelt flooding in the past several decades.

In particular, the model results suggest that the observed increasing air temperature resulted in decreasing trends in snow depth
and snowmelt, and also in the increasing trends in sublimation
and evaporation during the late winter months. This shift in flood
peaks from predominantly a snowmelt-driven regime to primarily
a rainfall-driven flood regime could be partly explained by the
temperature increase, and could characterize watersheds in other
geographically similar regions. Nonetheless, to provide a comprehensive analysis of trends in floods, similar studies should be
accompanied by those relating land-use change and other variables to these trends.
Acknowledgments
This research was partially supported by the National Center for
Supercomputing Applications at the University of Illinois in
Urbana-Champaign. The authors would like to acknowledge the
contribution of Dr. Gi-Hyeon Park from the University of Wyoming,
Dr. Tom Over from USGS, and Drs. Laura Bowling and Keith Cherkauer from Purdue University for valuable advice in various stages
of this research. Dr. Edward J. Hopkins, Assistant Wisconsin State
Climatologist, provided valuable information on snow data monitoring, history of the gages, and quality of the observed data. The
authors would also like to acknowledge the help from the following colleagues at ISWS/UIUC: Nancy Westcott and Alena Bartosova
for providing very useful review comments, Lisa Sheppard for editorial review, and Sara Olson for reviewing and modifying the
figures.
References
Anderson, E.A., 1976. A point energy and mass balance model of a snow cover. Tech.
Rep. 19, NOAA, Silver Spring, MD.
Andreadis, K.M., Clark, E.A., Wood, A.W., Hamlet, A.F., Lettenmaier, D.P., 2005.
Twentieth-century drought in the conterminous United States. J.
Hydrometeorol. 6 (6), 9851001.
Andreadis, K.M., Liang, D., Tsang, L., Lettenmaier, D.P., Josberger, E.G., 2008.
Characterization of errors in a coupled snow hydrology-microwave emission
model. J. Hydrometeorol. 9 (1), 149164.
Andreadis, K.M., Storck, P., Lettenmaier, D.P., 2009. Modeling snow accumulation
and ablation processes in forested environments. Water Resour. Res. 45.
Cherkauer, K.A., Bowling, L.C., Lettenmaier, D.P., 2003. Variable infiltration capacity
cold land process model updates. Global Planet. Change 38 (12), 151159.
Cherkauer, K.A., Lettenmaier, D.P., 1999. Hydrologic effects of frozen soils in the
upper Mississippi River basin. J. Geophys. Res.-Atmos. 104 (D16), 1959919610.
Cherkauer, K.A., Lettenmaier, D.P., 2003. Simulation of spatial variability in snow
and frozen soil. J. Geophys. Res.-Atmos. 108 (D22).
Cunderlik, J.M., Ouarda, T.B.M.J., 2009. Trends in the timing and magnitude of floods
in Canada. J. Hydrol. 375, 471480.
Dai, A., Washington, W.M., Meehl, G.A., Bettge, T.W., Strand, W.G., 2004. The ACPI
climate change simulations. Clim. Change 62 (13), 2943.

280

D. Park, M. Markus / Journal of Hydrology 515 (2014) 267280

Essery, R.L.H., Rutter, N., Pomeroy, J., Baxter, R., Sthli, M., Gustafsson, D., Barr, A.,
Bartlett, P., Elder, K., 2009. SNOWMIP2: an evaluation of forest snow process
simulations. Bull. Am. Meteorol. Soc. 90, 11201135. http://dx.doi.org/10.1175/
2009BAMS2629.1.
Feng, X., Sahoo, A., Arsenault, K., Houser, P., Luo, Y., Troy, T.J., 2008. The impact of
snow model complexity at three CLPX sites. J. Hydrometeorol. 9, 14641481.
http://dx.doi.org/10.1175/2008JHM860.1.
Gebert, W.A., Krug, W.R., 1996. Streamflow trends in Wisconsins driftless area.
Water Resour. Bull. 32 (4), 733744.
Hamlet, A.F., Lettenmaier, D.P., 2005. Production of temporally consistent gridded
precipitation and temperature fields for the continental United States. J.
Hydrometeorol. 6 (3), 330336.
Hamlet, A.F., Mote, P.W., Clark, M.P., Lettenmaier, D.P., 2005. Effects of temperature
and precipitation variability on snowpack trends in the western United States. J.
Clim. 18 (21), 45454561.
Hamlet, A.F., Mote, P.W., Clark, M.P., Lettenmaier, D.P., 2007. Twentieth-century
trends in runoff, evapotranspiration, and soil moisture in the western United
States. J. Clim. 20 (8), 14681486.
Hejazi, M., Markus, M., 2009. Impacts of urbanization and climate variability on
floods in northeastern Illinois. J. Hydrol. Eng.. http://dx.doi.org/10.1061/
(ASCE)HE.1943-5584.0000020.
Helsel, D.R., Hirsh, R.M., 1995. Statistical Methods in Water Resources. Elsevier
Science Publishing Company Inc., New York, 529 p.
Hirsch, R.M., Ryberg, K.R., 2012. Has the magnitude of floods across the USA
changed with global CO2 levels? Hydrol. Sci. J.. http://dx.doi.org/10.1080/
02626667.2011.621895.
Homer, C., Huang, C.Q., Yang, L.M., Wylie, B., Coan, M., 2004. Development of a 2001
national land-cover database for the United States. Photogramm. Eng. Rem. S 70
(7), 829840.
Jordan, R.E., 1991. A one-dimensional temperature model for a snow cover:
Technical documentation for SNTHERM.89. CRREL Special Rep. 9116, Cold
Regions Research and Engineering Laboratory, 49 pp.
Juckem, P.F., Hunt, R.J., Anderson, M.P., Robertson, D.M., 2008. Effects of climate and
land management change on streamflow in the driftless area of Wisconsin. J.
Hydrol. 355 (14), 123130.
Kalnay, E. et al., 1996. The NCEP/NCAR 40-year reanalysis project. B Am. Meteorol.
Soc. 77 (3), 437471.
Kendall, M.G., 1955. Rank Correlation Methods. Hafner Publishing Co., New York.
Knapp, H.V., 2005. Analysis of streamflow trends in the Upper Midwest using longterm flow records, EWRI World Water and Environmental Resources Congress.
ASCE, Anchorage.
Kochendorfer, J.P., Hubbart, J.A., 2010. The roles of precipitation increases and rural
land-use changes in streamflow trends in the Upper Mississippi River basin.
Earth Interact, 14.
Kumar, S., Merwade, V., Kam, J., Thurner, K., 2009. Streamflow trends in Indiana:
effects of long term persistence, precipitation and subsurface drains. J. Hydrol.
374 (12), 171183.
Kunkel, K.E., Palecki, M.A., Hubbard, K.G., Robinson, D.A., Redmond, K.T., Easterling,
D.R., 2007. Trend identification in twentieth-century U.S. snowfall: the
challenges. J. Atmos. Oceanic Technol. 24, 6473.
Liang, X., Lettenmaier, D.P., Wood, E.F., Burges, S.J., 1994. A simple hydrologically
based model of land-surface water and energy fluxes for general-circulation
models. J. Geophys. Res.-Atmos. 99 (D7), 1441514428.
Liang, X., Wood, E.F., Lettenmaier, D.P., 1996. Surface soil moisture parameterization
of the VIC-2L model: evaluation and modification. Global Planet. Change 13 (1
4), 195206.
Lins, H.F., Slack, J.R., 1999. Streamflow trends in the United States. Geophys. Res.
Lett. 26 (2), 227230.
Lohmann, D., Nolte Holube, R., Raschke, E., 1996. A large-scale horizontal routing model
to be coupled to land surface parametrization schemes. Tellus A 48 (5), 708721.
Mantua, N.J., Hare, S.R., Zhang, Y., Wallace, J.M., Francis, R.C., 1997. A Pacific
interdecadal climate oscillation with impacts on salmon production. Bull. Amer.
Meteor. Soc. 78, 10691079.
Markus, M., Angel, J.R., Yang, L., Hejazi, M.I., 2007. Changing estimates of design
precipitation in northeastern Illinois: comparison between different sources
and sensitivity analysis. J. Hydrol. 347 (12), 211222.

Markus, M., Knapp, H.V., Flegel, A., McConkey, S., Thomas, W.O., 2013. Episodic
change analysis of the annual flood peak time series for a flood insurance study.
J. Hydrol. Eng. 18 (1), 8591.
Markus, M., Wuebbles, D.J., Liang, X.-Z., Hayhoe, C., Kristovich, D.A.R., 2012.
Diagnostic analysis of future climate scenarios applied to urban flooding in the
Chicago metropolitan area. Climatic Change 111, 879902.
Martinec, J., Rango, A., Roberts, R., 1998. Snowmelt Runoff Model (SRM) Users
Manual. In: Baumgartner, M.F., Apfl, G.M. (Eds.), Geographica Bernensia Ser. P,
no. 35. Berne: University of Berne.
Maurer, E.P., Wood, A.W., Adam, J.C., Lettenmaier, D.P., Nijssen, B., 2002. A longterm hydrologically based dataset of land surface fluxes and states for the
conterminous United States. J. Climate 15 (22), 32373251.
Mishra, V., Cherkauer, K.A., 2011. Influence of cold season climate variability on
lakes and wetlands in the Great Lakes region. J. Geophys. Res.-Atmos. 116,
D12111.
Mishra, V., Cherkauer, K.A., Shukla, S., 2010. Assessment of drought due to historic
climate variability and projected future climate change in the Midwestern
United States. J. Hydrometeorol. 11 (1), 4668.
Myneni, R.B., Nemani, R.R., Running, S.W., 1997. Estimation of global leaf area index
and absorbed par using radiative transfer models. IEEE T Geosci. Remote 35 (6),
13801393.
Nakayama, T., Watanabe, M., 2006. Simulation of spring snowmelt runoff by
considering micro-topography and phase changes in soil layer. Hydrol. Earth
Syst. Sci. Discuss. 3 (4), 21012144.
Nash, J.E., Sutcliffe, J., 1970. River flow forecasting through conceptual models part
IA discussion of principles. J. Hydrol. 10 (3), 282290.
National Resources Council, 2011. Climate Stabilization Targets: Emissions,
Concentrations and Impacts over Decades to Millennia. National Resources
Council, Washington, DC.
Novotny, E.V., Stefan, H.G., 2007. Stream flow in Minnesota: indicator of climate
change. J. Hydrol. 334 (34), 319333.
Potter, K.W., 1991. Hydrological impacts of changing land management
practices in a moderate-sized agricultural catchment. Water Resour. Res. 27,
845855.
Regonda, S.K., Rajagopalan, B., Clark, M., Pitlick, J., 2005. Seasonal cycle shifts in
hydroclimatology over the western United States. J. Climate 18 (2), 372384.
Sen, P.K., 1968. Estimates of the regression coefficient based on Kendalls tau. J. Am.
Stat. Assoc. 63 (324), 13791389.
Sheffield, J., Wood, E.F., 2008. Global trends and variability in soil moisture and
drought characteristics, 19502000, from observation-driven simulations of the
terrestrial hydrologic cycle. J. Climate 21 (3), 432458.
Sinha, T., Cherkauer, K.A., 2010. Impacts of future climate change on soil frost in the
Midwestern United States. J. Geophys. Res.-Atmos., 115.
Sinha, T., Cherkauer, K.A., Mishra, V., 2010. Impacts of historic climate variability on
seasonal soil frost in the Midwestern United States. J. Hydrometeorol. 11 (2),
229252.
Tan, A., Adam, J.C., Lettenmaier, D.P., 2011. Change in spring snowmelt timing in
Eurasian Arctic rivers. J. Geophys. Res.-Atmos., 116.
Tarboton, D.G., Chowdhury, T.G., Jackson, T.H., 1995. A spatially distributed
energy balance snowmelt model. In: Tonneson, K.A., Williams, W., Tranter, M.
(Eds.), Biogeochemistry of Seasonally Snow Covered Catchments, vol. 228. Int.
Assoc. of Hydrol. Sci., Wallingford, UK, pp. 141155.
Varhola, A., Wawerla, J., Weiler, M., Coops, N.C., Bewley, D., Alila, Y., 2010. A new
low-cost, stand-alone sensor system for snow monitoring. J. Atmos. Oceanic
Technol. 27, 19731978.
Wigmosta, M.S., Vail, L.W., Lettenmaier, D.P., 1994. A distributed hydrologyvegetation model for complex terrain. Water Resour. Res. 30, 16651679.
Wood, E.F., Pan, M., Sheffield, J., Vinukolu, R., Ferguson, C.R., Gao, H., Tang, Q., Shi, X.,
Zhu, C., Bohn, T., Su, F., Lettenmaier, D.P., Sahoo, A.K., De Lannoy, G.J.M., Houser,
P.R., Pinker, R.T., Ma, Y., Li, C., Kummerow, C.D., 2009. Algorithm theoretical
basis document for terrestrial water cycle data records. Princeton Univ.,
Princeton, NJ.
Zhang, T., Barry, R., Gilichinsky, D., Bykhovets, S.S., Sorokovikov, V.A., Ye, J., 2001. An
amplified signal of climatic change in soil temperatures during the last century
at Irkutsk, Russia. Climatic Change 49, 4176.