Atmospheric and Climate Sciences, 2012, 2, 427-440
http://dx.doi.org/10.4236/acs.2012.24037 Published Online October 2012 (http://www.SciRP.org/journal/acs)
Possible Impacts of Climate Change on Daily Streamflow
and Extremes at Local Scale in Ontario, Canada. Part II:
Future Projection
Chad Shouquan Cheng1*, Qian Li1, Guilong Li1, Heather Auld2,3
1Science Section, Operations—Ontario, Meteorological Service of Canada, Environment Canada, Toronto, Canada
2Adaptation and Impacts Research Division, Science and Technology Branch, Environment Canada, Toronto, Canada
3Risk Science International (RSI) Inc., Toronto, Canada
Received May 10, 2012; revised June 14, 2012; accepted June 25, 2012
ABSTRACT
The paper forms the second part of an introduction to possible impacts of climate change on daily streamflow and ex-
tremes in the Province of Ontario, Canada. Daily streamflow simulation models developed in the companion paper (Part
I) were used to project changes in frequency of future daily streamflow events. To achieve this goal, future climate in-
formation (including rainfall) at a local scale is needed. A regression-based downscaling method was employed to
downscale eight global climate model (GCM) simulations (scenarios A2 and B1) to selected weather stations for vari-
ous meteorological variables (except rainfall). Future daily rainfall quantities were projected using daily rainfall simula-
tion models with downscaled future climate information. Following these projections, future daily streamflow volumes
can be projected by applying daily streamflow simulation models.The frequency of future daily high-streamflow events
in the warm season (May-November) was projected to increase by about 45% - 55% late this century from the current
condition, on average of eight-GCM A2 projections and four selected river basins. The corresponding increases for fu-
ture daily low-streamflow events and future daily mean streamflow volume could be about 25% - 90% and 10% - 20%,
respectively. In addition, the return values of annual one-day maximum streamflow volume for various return periods
were projected to increase by 20% - 40%, 20% - 50%, and 30% - 80%, respectively for the periods 2001-50, 2026-75,
and 2051-2100. Inter-GCM and interscenario uncertainties of future streamflow projections were quantitatively as-
sessed. On average, the projected percentage increases in frequency of future daily high-streamflow events are about 1.4 -
2.2 times greater than inter-GCM and interscenario uncertainties.
Keywords: Rainfall-Related Streamflow; Future Projection; Downscaling; Statistic Methods; Ontario; Canada
1. Introduction
It is widely believed that in this century climate change
might result in increased flooding in many regions over
the globe, based on studies conducted in the past decades
(see some of the references listed in Table 1). A number
of studies have specifically focused on projecting chan-
ges in future annual/seasonal average streamflow volum-
es under a changing climate. In most of the studies,
global climate model (GCM) projections are commonly
used as the future climatic conditions. The GCM project-
tions have been used in the studies in different ways: 1)
GCM-implied changes applied to the observed daily cli-
mate; 2) statistically downscaled GCM simulations; and
3) dynamically downscaled GCM data—regional climate
model (RCM) simulations. In addition, hypothesized
scenarios, such as increases in annual mean temperature
of 1˚C, 2˚C, 4˚C and/or changes in annual precipitation
of ±5%, ±10%, ±20%, relative to the baseline climate,
were used in the analysis on hydrological impacts of cli-
mate change.
GCM-implied changes were applied to the observed
rainfall and potential evaporation to generate inputs for
conceptual or semi-distributed rainfall-runoff hydrologi-
cal simulation models (e.g., [1-5]). The change values
that are derived from averaging multi-year monthly
GCM outputs are applied to daily or even hourly ob-
served data [6]. In addition to changes derived from
GCM projections, the hypothesized scenarios were used
as the input to hydrological models to project future wa-
ter balance components (e.g., [7,8]). To project future
climates required by hydrological models, daily and
hourly observed hydrometeorological variables, such as
temperature and rainfall, were modified by adding a sin-
*Corresponding author.
C
opyright © 2012 SciRes. ACS

C. S. CHENG ET AL.
428
gle change value for a month or whole study period. It
seems that this assumption might not be practical since it
does not account for the anticipated changes to the vari-
ability of future climate [9,10].
As was done to project climate change impacts on
streamflow at a local scale, another option is to apply
don scaled simulations of GCMs to hydrological mod- w
els. In addition to dynamic downscaling approach (i.e.,
RCMs) to project future flood frequency (e.g., [6,11]),
another leading technique is statistical (empirical) down-
scaling [12]. The statistical downscaling methods have
been widely used in hydrometeorology to project changes
in frequency of future high-streamflow or rainfall-driven
flood events [13-18].
Table 1. Examples of previous studies on climate change and streamflow.
Scenarios Reference Hydrological Model Key Results Study Area
Bultot et al. [1] Conceptual hydrological
model (IRMB),
Winter streamflow could increase in all of the study basins;
direction of summer flow change depends on basin types under a
2 × CO2 condition
Belgium
Panagoulia &
Dimou [2]
Conceptual hydrological
model
Flood frequency, duration, and volume could increase for all in-
creased precipitation HYPO and GISS scenarios Central Greece
Sefton & Boorman
[32]
Unit hydrograph-based
model
Mean annual flow could increase by up to 60% in most of the region
under a 2 × CO2 condition
England and
Wales, UK
Gellens & Roulin
[33] IRMB Flood frequency could increase, especially in winter season under a 2 × CO2 condition Belgium
Arnell [34] Catchment hydrological
model Winter (DJF)/summer (JJA) streamflow could increase/decrease Six rivers in UK
Eckhardt & Ulbrich
[35]
Conceptual hydrological
model: Soil & Water
Assessment Tool
Winter (DJF)/summer (JJA) streamflow could increase/decrease by
10%/50% by 2070-2099
Rhenish Massif,
Germany
Drogue et al. [3]
HRM (conceptual
Hydrological Recursive
Model)
Daily mean discharge could increase, the magnitude of which de
pends upon the rainfall change scenarios by the 2050s
Grand Duchy of
Luxembourg
Cameron [4] TOPMODEL
(semi-distributed model)Direction of the change by the 2080s varies among the GCMs Northeastern Scotland, UK
GCM-implied
changes
applied to the
observed daily
climate
Forbes et al. [5]
ACRU
agro-hydrological
modeling system
Winter /spring streamflows could increase and summer/fall stream
flows could decrease in the middle and late of this century
Southern Al
b
erta,
Canada
Najjar [13] Statistical and water
balance models
Annual streamflow could increase by 24% ± 13% under a 2 × CO2
condition Pennsylvania, US
Whitfield et al. [14] Hydrograph-based
model
Frequency of rainfall-driven floods could increase in all watersheds
under the projected climate scenarios; rainfall-related low flows
could maintain the same frequency and magnitude
Georgia Basin,
BC, Canada
Dettinger et al. [15]
PPMS (physically based
Precipitation-Runoff
Modeling System)
Winter and summer streamflow could increase and decease in the
21st century
Three rivers in
California, US
Jasper et al. [16]
WaSiM-ETH (Water
Flow and Balance
Simulation Model)
Winter (DJF)/summer (JJA) runoffs could increase/decrease by
14% - 31%/16% - 33% at two basins by 2081-2100
Two Alpine river
basins,
Switzerland
Dibike & Coulibaly
[17]
Distributed hydrological
models: Swedish &
Canadian models
Early spring streamflow could dramatically increase; winter flow
could increase and summer flow could decrease in the 21st century
Northern Quebec,
Canada
Statistically
downscaled
GCM scenarios
Merritt et al. [18] Semi-distributed
conceptual models
Annual, spring, and winter flow volumes could decrease by the
2050s and 2080s
Okanagan Basin,
Canada
Kay et al. [6] Simplified probability
distributed model Flood frequency could increase by the 2080s in most of catchments Fifteen rivers across UK
RCM scenarios
Sushama et al. [11] RCM outputs directly
Annual and seasonal (except summer) streamflow in Canadian river
basins (i.e., Mackenzie, Yukon, Fraser) could increase in the future
(2041-2070)
Six river basins,
North America
Arnell & Reynard
[36]
Conceptual hydrological
models
Annual flow could increase over 20% for the wettest scenarios and
decline over 20% for the driest scenarios by 2050
21 catchments in
UK
Panagoulia &
Dimou [7]
Monthly water balance
model
Summer runoff could decrease by 50% for the driest scenario;
winter runoff could increase by 60% for the wettest scenario Central Greece
Hypothesized
scenarios
Jiang et al. [8] Six monthly water
balance models
The six-model results are similar in reproducing historical water
balance components but different in estimating future changes Southern China

C. S. CHENG ET AL. 429
The statistical downscaling scheme used in this cur-
rent study is built upon the previous studies (i.e., Cheng
et al. [19-21]) for deriving future hydrometeorological
variables that were used in development of daily stream-
flow simulation models (constructed in Part I, [22]),
which is made up of a four-step process. First, daily
rainfall simulation models were developed and validated
using synoptic weather typing with cumulative logit re-
gression and nonlinear regression procedures [20]. The
simulation models consider physical process of rainfall
formation since the theories combining from both con-
ceptual and statistical modeling were applied in the
model development. To more effectively develop daily
rainfall simulation models, the study [20] has used a
number of the atmospheric stability indices in addition to
the standard meteorological variables that were com-
monly used in most of the previous rainfall downscaling
papers. Second, regression-based downscaling transfer
functions developed by Cheng et al. [19] are adapted to
derive station-scale future hourly meteorological vari-
ables (except rainfall) that were used in development of
daily rainfall simulation models. Third, using down-
scaled future hourly climate data, future daily rainfall
quantity can be projected by applying synoptic weather
typing and daily rainfall simulation models [21]. Finally,
it is able to project future daily streamflow volumes by
applying daily streamflow simulation models [22] with
down- scaled future daily rainfall and temperature data.
This paper is organized as follows: in Section 2, data
sources and their treatments are described. Section 3
summarizes the previous studies (Cheng et al. [19-21])
on future daily rainfall projection on which the current
paper was built. Section 4 describes analysis techniques
as applied to projection of future daily streamflow vol-
umes. Section 5 includes the results and discussion on 1)
changes in frequencies of future daily high- and low-
streamflow events, 2) changes in future return values of
one-day maximum streamflow events, and 3) uncertainty
of the study and limitations of the data. The conclusions
and recommendations from the study are summarized in
Section 6.
2. Data Sources
To project future daily streamflow volumes using stream-
flow simulation models developed in a companyion pa-
per (Part I: Historical simulation, Cheng et al. [22]), the
historical observations and future projections of the me-
teorological/hydrological elements are essential. Histori-
cal observations include daily surface meteorological/
hydrological data (i.e., rainfall, temperature, streamflow)
within the four selected river basins, which were used in
daily streamflow simulation modeling described in Part I
[22]. The future projections include 1) future local-scale
daily rainfall quantities projected by a recent study
(Cheng et al. [21]), using daily rainfall simulation mod-
els developed by Cheng et al. (2010) and 2) station-scale
daily temperature down-scaled by a statistical down-
scaling approach developed by Cheng et al. [19]. To bet-
ter understand the study, the relevant information on pro-
jection of future daily rainfall and extremes as well as
statistical down-scaling will be summarized in Section 3.
In addition to historical observations and projections
of future daily rainfall quantities, daily climate change
simulations from eight GCM models and two emission
scenarios from the Fourth Assessment Report (AR4) of
the Intergovernmental Panel on Climate Change (IPCC)
were used in the study, summarized in Table 2. The
eight GCM models were selected since their simulations
of all weather elements, including surface and upper-air
temperature, dew point, air pressure, total cloud cover,
u-wind and v-wind are available. These climate change
simulations were retrieved from the Web site of the Pro-
gram for Climate Model Diagnosis and Intercomparison
[23]. The PCMDI is archiving the GCM simulations for
two future time periods (2046-2065 and 2081-2100). Fur-
thermore, the historical runs of these GCM simulations
for the period (1961-2000) were used to remove the
GCM model bias from the projection of future daily
streamflow volumes. These three time windows were
used in the analysis because these data are only available
from the PCMDI’s Web site. Furthermore, for project-
tions of future return-period values of annual maximum
high-streamflow events, three CGCM transient model
simulations for a 100-year period (2001-2100) were in-
cluded (Table 2), which were retrieved from Environ-
ment Canada’s Web site [24].
3. Summary of Future Daily Rainfall
Projection
As part of this research, Cheng et al. [21] have projected
future daily rainfall quantities using daily rainfall simula-
tion models (Cheng et al. [20]) with downscaled stan-
dard meteorological variables derived from statistical
downscaling transfer functions (Cheng et al. [19]). Since
the results and methods from these studies were used and
adopted in this current paper to project changes in fre-
quency of future daily streamflow events, it is necessary
to outline these studies focusing on major methods and
results.
As described in a recent study (Cheng et al. [21]), to
project future daily rainfall amounts, the station-scale
future climate data of the meteorological variables are
necessary for the use of within-weather-type daily rain-
fall simulation models developed by Cheng et al. [20].
To derive future hourly station-scale climate information
from GCM-scale simulations, Cheng et al. [19] have
developed a regression-based statistical downscaling

C. S. CHENG ET AL.
430
Table 2. GCM simulations and scenarios used in the study.
GCM IPCC scenario Periods
CGCM3.1/T63
CNRM-CM3
CSIRO-Mk3.0
ECHAM5/MPI
ECHO-G
GFDL-CM2.0
GISS-ER
MIRoc3.2 (medres)
IPCC AR4 SRES
A2/B1
1961-2000
2046-2065
2081-2100
For return period
analysis:
CGCM1
CGCM2
IPCC Third
Assessment Report
(TAR) SRES A2/B2
2001-2100
Note: IPCC AR4—Intergovernmental Panel on Climate Change, Fourth
Assessment Report; SRES—Special Report on Emissions Scenarios;
CGCM3.1/T63—Canadian global climate model (the 3rd generation, T63
version); CNRM-CM3—French global climate model (the 3rd generation)
developed atocenter National Weather Research; CSIRO-Mk3.0—Aus-
tralian global climate model (the 3rd generation) developed by the Com-
monwealth Scientific and Industrial Research Organization; ECHAM5—
Germany global climate model (the 5th generation) developed by the Max
Plank Institute for Meteorology in Hamburg; ECHO-G—Germany and
Korean global climate model consisting of the atmospheric model ECHAM4
and the ocean model HOPE (Hamburg ocean Primitive Equation); GFDL-
CM2—US global climate model (the 2nd generation) developed by Geo-
physical Fluid Dynamics Laboratory; GISS-ER—US global climate model
developed by Goddard Institute for Space Studies, NASA. MIRoc3.2
(medres)—Japanese global climate model (the 3rd generation, med-res.
version), Model for Interdisciplinary Research on Climate.
method to spatially downscale daily GCM simulations to
the selected weather stations in south-central Canada and
then to temporally downscale daily scenarios to hourly
timesteps. The downscaling transfer functions were con-
structed using different regression methods for different
meteorological variables since a regression method is
suitable only for a certain kind of data with a specific
distribution. The downscaled meteorological variables
include surface and upper-air temperature, dew point
temperature, west-east and south-north winds, air pressure,
and total cloud cover. These weather parameters are es-
sential to project future daily rainfall quantities using
rainfall simulation models constructed via combination
of an automated synoptic weather typing and cumulative
logit/nonlinear regression analyses. Performance of the
downscaling transfer functions was evaluated by 1) ana-
lyzing model R2s of downscaling transfer functions; 2)
validating downscaling transfer functions using a leave-
one-year-out cross-validation scheme; and 3) comparing
data distributions, extreme characteristics, and seasonal/
diurnal changes of downscaled GCM historical runs ver-
sus observations over a comparative time period of 1961-
2000. The results showed that regression-based down-
scaling methods performed very well in deriving station-
scale hourly and daily climate information for all
weather variables. For example, the hourly downscaling
transfer functions for surface air temperatures, dewpoint,
and sea level air pressure possess the highest model R2
(>0.95) of the weather elements. The functions for south-
north wind speed (y wind) are the weakest model (model
R2s ranging from 0.69 to 0.92 with half of them greater
than 0.89). Details of the hourly and daily downscaling
methodologies and evaluations of the results are not pre-
sented in this current paper owing to the limitations of
space (refer to Cheng et al. [19] for details).
Following downscaling of the meteorological variables,
future daily rainfall amounts are able to be projected us-
ing within-weather-type daily rainfall simulation models
developed by Cheng et al. [20,21]. As described in the
study [20], 10 synoptic weather types in the study area
were identified over the 45-yr period as primary rainfall-
related weather types. Within-weather-type daily rainfall
simulation models were developed in a two-step process:
1) cumulative logit regression to predict the occurrence
of daily rainfall events; and 2) using probability of the
logit regression, a nonlinear regression procedure to
simulate daily rainfall quantities. To more effectively
distinguish heavy rainfall events, the daily rainfall simu-
lation models were constructed using not only the stan-
dard meteorological variables but also a number of the
atmospheric stability indices (e.g., lifted index [25],
K-index [26], total totals index [27]). The performance of
within-weather-type daily rainfall simulation models was
evaluated, and as described by Cheng et al. [20], the re-
sults showed that the models were successful at verifying
occurrence of daily rainfall events and daily rainfall
quantities. Cheng et al. [20] have found that, across the
four selected river basins, the percentage of excellent and
good daily rainfall-quantity simulations ranged from
62% to 84%, based on absolute difference between ob-
served and simulated daily rainfall amounts. In addition,
it is noteworthy that the rainfall simulation models are
able to capture most of daily heavy rainfall events (i.e.,
≥32.5 mm) with the percentage of excellent and good
simulations: 62%, 68%, 70%, and 81% for Grand, Tha-
mes, Humber, and Rideau River Basins, respectively.
Following development of daily rainfall simulation
models, future daily rainfall and its extremes were pro-
jected by applying within-weather-type rainfall simula-
tion models altogether with downscaled future GCM
climate data. Cheng et al. [21] have used three GCMs
and two emission scenarios (A2 and B2) from the IPCC
Third Assessment Report (TAR) to project future daily
rainfall and its extremes. Since climate simulations from
eight GCMs and two IPCC AR4 emission scenarios A2
and B1 were used in this current study, it is necessary to
update projections of future daily rainfall quantities using
these downscaled updated GCM simulations. The number
of future seasonal rain days and future seasonal rainfall
totals projected by rainfall simulation models versus his-
torical observations are graphically illustrated in Figure
1. The rainfall projections were evaluated by comparing

C. S. CHENG ET AL. 431
differences in the number of seasonal rain days and sea-
sonal rainfall totals derived from down-scaled historical
runs and observations during a comparative time period
1961-2000. As shown in Figure 1, the values derived
from both downscaled historical runs and observations
are very similar. This implies that daily rainfall down-
scaling method used in the study is suitable for projecting
changes in the number of future seasonal rain days and
future seasonal rainfall totals, which is similar to the
conclusion made by Cheng et al. [21].
From Figure 1, it can be seen that the modeled re-
sults found that the frequency of future daily rainfall
events could increase late this century due to the chang-
ing climate projected by GCM scenarios. For example,
on average across the selected GCMs, the frequency of
future rainfall events with daily rainfall ≥15 mm is pro-
jected to increase by about 2 - 7 days in the future across
the selected river basins (from the current four-river-
basin average of 10.5 days for the period April-Novem-
ber 1961-2002). The corresponding increases for rainfall
events with daily rainfall ≥25 mm are projected to be
about 1 - 3 days from the current four-river-basin average
of 3.2 days for the period April-November 1961-2002.
Owing to limitations of the space, refer to Cheng et al.
[20,21] for details on daily-rainfall simulation modeling
and future daily-rainfall projections.
4. Analysis Techniques
Following the projection of future daily rainfall amounts
and downscaling of future daily temperature, future daily
rainfall-induced streamflow volumes can be projected
using daily streamflow simulation models-developed in
Part I (Cheng et al. [22]). Using future downscaled/pro-
jected daily rainfall quantities and temperature data, the
predictors used in the development of daily streamflow
simulation models were derived for the future time peri-
ods according to the same criteria as were the models
developed using historical observations (constructed in
Part I [22]). Following the projection of future daily
streamflow volumes, the changes in frequency of high-/
low-streamflow events and magnitude of daily mean
streamflow volumes from the current condition (1961-
2000) are able to be evaluated. High-streamflow events
were defined as a day with streamflow volume greater
than or equal to the 95th percentile derived from the ob-
servations; low-streamflow events were defined as a day
with streamflow volume less than the 5th percentile.
Although the daily streamflow simulation models
demonstrated significant skill in the prediction of histo-
rical daily streamflow volumes as well as occurrence of
high-/low-streamflow events [22], it is necessary to as-
certain whether the methods are suitable for the future
projection. To achieve this, the data distribution of daily
streamflow volumes was evaluated for both the down-
400
500
600
700
800
900
1000
1100
RainfallTotals(mm)
(c)Sea so nalrainfalltotals
OHA2B1OH A2 B1OHA2B1OHA2B1
0
1
2
3
4
5
6
7
8
Num b erofDays
(b) ≥25mm
CGCM3.1
CNRM‐CM3
CSIRO‐Mk3.0
ECHAM5/MP
ECHO‐G
GFDL‐CM2.0
GISS‐ER
MIROC3.2
OHA2 B1OHA2B1OH A2B1OH A2 B1
0
5
10
15
20
25
Num b erofDays
(a)  ≥15mm
OHA2B1OHA2B1OHA2 B1OH A2 B1
Grand  Humber RideauThames
Figure 1. The number of future seasonal rain days [(a): ≥15
and (b): ≥25 mm) and (c): future seasonal rainfall totals
versus the observed values during the period April-No-
vember 1961-2002 (O-observations and H-downscaled his-
torical runs, following four bars represent eight-GCM-
A2-averaged values and eight-GCM-B1-averaged values,
for two time periods 2046-2065 and 2081-2100).
scaled GCM historical runs and observations over a
comparative time period (1961-2000). Figure 2 shows
quantile-quantile (Q-Q) plots of the sorted daily stream-
flow volumes from both downscaled GCM historical
runs and observations in the selected river basins. The
Q-Q plot is a graphical technique for determining if two
datasets come from populations with a common distribu-
tion, showing that the points should fall approximately
along with the 45-degree reference line. Otherwise, the
greater the departure from this reference line, the greater
the evidence for the conclusion that the two datasets
come from populations with different distributions. From
Figure 2, it is clear that data distributions of both data-

C. S. CHENG ET AL.
Copyrig ACS
432
sets are similar; so that it can conclude that the methods
used in the study are suitable for projecting or downscal-
ing future daily streamflow information on a local scale.
show that for thresholds with daily streamflow of 100, 10,
40, and 40 m3·s–1 or less in Grand, Humber, Rideau, and
Upper Thames river basins, respectively, the differences
between downscaled GCM historical runs and observa-
tions affect the future projections by about 1% - 2%. The
corresponding effects for daily streamflow volumes
greater than the thresholds are about 2% - 5% for Hum-
ber, Rideau, and Upper Thames river basins. However,
for the Grand River Basin, the departure from the
45-degree reference line for a portion of the high stream-
flows is somewhat greater than for the other river basins.
This is likely a result of having limited data; specifically,
hourly meteorological observations within the river
basin are not available. As a result, for the Grand River
Basin, hourly meteorological variables observed at the
London International Airport (located in the Thames
River) were used in the analyses, including synoptic
weather typing, downscaling of meteorological variables,
and daily rainfall historical simulations and future pro-
jections (Cheng et al. [20,21]). In addition, another rea-
son is that a very small data sample for the high-stream-
flow events, for instance, four observations above 150
m3·s–1, contributes a great departure from the 45-degree
reference line.
Any small differences between downscaled GCM his-
torical runs and observations, as shown in Figure 2, were
used to further adjust GCM model biases for projections
of changes in frequency of future daily streamflow
events. To quantitatively assess how much these differ-
ences affect projections of changes in frequency of future
daily streamflow events, we have calculated mean rela-
tive absolute differences (RAD) between observations (Oi)
and downscaled GCM historical runs (Di) by the follow-
ing expression:
1
1nii
ii
OD
NO



RAD  (3)
where N is the number of total pairs of the data sample.
The RAD was calculated for the days when daily stream-
flow volumes greater than each of various thresholds
(e.g., 5, 10, 20, 30, 40, 100 m3·s–1), for each of down-
scaled GCM historical runs and each of four selected
river basins. Then the mean RAD for each of the thresh-
olds was determined by pooling eight downscaled GCM
historical runs for each of four river basins. The results
Figure 2. Quantile-quantile plots of daily streamflow volume derived from downscaled GCM historical runs versus observa-
tions over a comparative time period (May-November 1961-2000) in the selected river basins (A 45-degree reference line
suggests that both datasets come from populations with the same distribution).
ht © 2012 SciRes.

C. S. CHENG ET AL. 433
5. Results and Discussions
0
5
10
15
20
25
30
35
NumberofDay s
(a) High‐streamflow(≥95
th
percentile)
OHA2B1 OHA2B1OHA2 B1OHA2B1
5.1. Changes in Frequency of Future Daily
High- and Low-Streamflow Events
Following the projection of future daily rainfall quanti-
ties, the daily streamflow volumes are able to be pro-
jected by applying streamflow simulation models de-
veloped in Part I (Cheng et al. [22]). Since the antece-
dent precipitation index (API) calculated using rainfall
data from the previous 24 days was used as a predictor in
daily streamflow simulation modeling (refer to [22]), the
time period for the projection of future daily streamflow
volumes is from May, rather than April, to November.
The number of projected future daily high-/low-stream-
flow events and future daily mean streamflow volumes
versus observations are graphically illustrated in Figure
3. The daily high- and low-streamflow events were de-
fined as those days with a streamflow volume greater
than or equal to the 95th percentile and less than the 5th
percentile, respectively derived from the observations.
The frequencies of future high- and low-streamflow
events were determined based upon historical observed
values of the 95th and 5th percentiles listed in Table 3.
As shown in Figure 3, the modeled results from this
study found that the frequency of future daily high-/low-
streamflow events and daily mean streamflow volumes
could increase late this century due to the changing cli-
mate projected by GCM scenarios. In addition, the
streamflow projections were evaluated by comparing
differences in the number of seasonal high-/low-stream-
flow events and daily mean streamflow volumes derived
from downscaled historical runs and observations during
time period 1961-2000. It is noteworthy that as shown in
Figure 3, the values from both datasets are very similar,
which implies that the streamflow downscaling methods
used in the study are suitable to project changes in the
number of future daily high-/low-streamflow events and
daily streamflow volumes.
To more clearly present changes in the frequency of
future daily high-/low-streamflow events and daily mean
streamflow volumes, four-river-basin-average relative in-
creases from the current conditions of May-November
1961-2002 are shown in Table 4. The percentage in-
crease in frequency of the daily high-/low-streamflow
events by 2081-2100 is projected to be greater than that
by the time period 2046-2065. Across the four selected
river basins, for example, on average of eight-GCM
A2projections, the frequency of future high-streamflow
events is projected to increase by 46% and 55%, respec-
tively by the periods 2046-2065 and 2081-2100 (from the
current 11 days for the period May-November). The cor-
responding increases for future low-streamflow events
are projected to be 26% and 89%. Daily mean stream-
flow volume is projected to increase by about 13% - 22%
Gra nd Humber Rideau Thames
0
5
10
15
20
25
30
35
40
NumberofDays
(b ) Low‐streamflow(<5thpercentile)
CGCM3.1
CNRM ‐CM3
CSIRO‐Mk3.0
ECHAM5/MPI
ECHO‐G
GFDL‐CM2.0
GISS‐ER
MIROC3.2
OH A2B1OH A2 B1 OHA2 B1OH A2B1
0
2
4
6
8
10
12
14
16
18
MeanStreamflow(m
3
∙s
‐1
)
(c)Dailymeanstreamflow
OHA2B1OHA2B1OHA2B1OHA2B1
0
5
10
15
20
25
30
35
NumberofDay s
(a) High‐streamflow(≥95
th
percentile)
OHA2B1 OHA2B1OHA2 B1OHA2B1
0
5
10
15
20
25
30
35
40
NumberofDays
(b ) Low‐streamflow(<5thpercentile)
CGCM3.1
CNRM ‐CM3
CSIRO‐Mk3.0
ECHAM5/MPI
ECHO‐G
GFDL‐CM2.0
GISS‐ER
MIROC3.2
OH A2B1OH A2 B1 OHA2 B1OH A2B1
0
2
4
6
8
10
12
14
16
18
MeanStreamflow(m
3
∙s
‐1
)
(c)Dailymeanstreamflow
OHA2B1OHA2B1OHA2B1OHA2B1
Gra nd Humber Rideau Thames
Figure 3. The number of future (a) daily high-streamflow
events (≥95th percentile of the historical observation) and (b)
low-streamflow events (<5th percentile of the historical
observation) as well as (c) future mean daily streamflow
volumes versus the historical observed values (O-observa-
tions and H-downscaled historical runs, following four bars
represent eight-GCM-A2-averaged values and eight-GCM-
B1-averaged values, for two time periods 2046-2065 and
2081-2100).
Table 3. Historical observed streamflow volumes (m3·s−1) of
the 95th and 5th percentiles (May-November 1958-2002 for
Grand and Upper Thames Rivers, May-November 1967-
2002 for Humber River, and May-November 1970-2002 for
Rideau River).
Grand Humber Rideau Upper Thames
95th percentile 18.90 2.91 11.60 6.73
5th percentile 1.93 0.16 0.05 0.25
Daily mean
streamflow 6.61 0.73 2.94 1.93

C. S. CHENG ET AL.
434
late this century. In addition, the projected four-river-
basin-averaged percentage increases derived from eight-
GCM ensemble A2 scenario are greater than those from
B1 scenario for the period 2081-2100; while the in-
creases are very similar between two scenarios for the
period 2046-2065. These projected increases are associ-
ated with GCM simulations: temperature increases be-
tween scenarios A2 and B1 are very similar for the pe-
riod 2046-2065; however, for the period 2081-2100, the
increases simulated from scenario A2 are greater than
those from B1.
As shown in Figure 3, the projections of future daily
high-/low-streamflow events and daily mean stream-
flow volumes vary across the selected GCMs. To eva-
luate performance of GCMs’ projections, the individual
GCM’s projections were compared with eight-GCM-
average changes to determine which GCMs’ projections
with the closest to or the most faraway from averaged
future projection. A case with the closest to averaged
future projection was defined as a relative change in fu-
ture projection is within 10% around the average, while a
case with the most faraway from averaged future projec-
tion was defined as a relative change is more than 50%
higher or lower than the average. From Table 5, it can be
seen that the top three best-performed GCMs having the
highest numbers of the cases with the closest to and the
lowest numbers of the cases with the most faraway from
eight-GCM-average relative changes are GFDL-CM2—
US global climate model (the 2nd generation), CNRM-
CM3—French global climate model (the 3rd generation),
and ECHAM5—Germany global climate model (the 5th
generation). While the worst performed GCMs are MI-
Roc3.2 (medres)—Japanese global climate model (the
3rd generation, med-res. version) and GISS-ER—US
global climate model.
The projected increases in the frequency of future
high-streamflow events might be due to the potential
increase in the frequency of future heavy rainfall events
projected by downscaled future GCM scenarios [21].
Possible reasons for an increase in the frequency of fu-
ture low-streamflow events might be an increase in the
frequency and severity of future drought condition and
the magnitude of future evapotranspiration in summer-
time. Although future seasonal rainfall totals in the study
Table 4. Four-river-basin-averaged percentage increases in the frequency of future seasonal high-streamflow/low-streamflow
days and future daily mean streamflow volumes from the current conditions of May-November 1961-2002, presented by
eight-GCM A2 and B1 ensemble.
Eight-GCM A2 Eight-GCM B1
Streamflow events Current conditions
2046-2065 2081-2100 2046-2065 2081-2100
High-streamfow (≥95th percentile) 10.7 days 46 55 44 45
Low-streamflow (<5th percentile) 10.4 days 26 89 25 55
Daily mean streamflow 0.74 - 6.68 m3·s–1 13 22 13 13
Table 5. Evaluation on GCM’s projections of future daily streamflow derived from Figure 3: the number of cases with the
closest to the eight-GCM ensemble future projection (defined as a relative change is within 10% around the average); the
number of cases with the most faraway from the eight-GCM ensemble future projection (defined as a relative change is at
least 50% higher or lower than the average).
Closest to averaged future projections Most faraway from averaged future projections
GCM Grand Humber Rideau Thames Total Grand Humber Rideau Thames Total
CGCM3 5 6 6 3 20 3 1 0 3 7
CNRM3 3 7 6 8 24 1 1 0 2 4
CSIRO3 4 6 2 4 16 5 3 2 4 14
ECHAM5 2 7 5 7 21 2 0 1 0 3
ECHO 3 8 2 6 19 4 2 4 2 12
GFDL-CM2 6 7 5 5 23 0 0 0 2 2
GISS-ER 3 5 0 5 13 4 1 10 1 16
MIRoc3 1 1 1 6 9 0 8 6 2 16
Total 27 47 27 44 145 19 16 23 16 74

C. S. CHENG ET AL. 435
area are projected to increase under a changing climate,
the number of days without rainfall or with a little rain-
fall is also projected to be greater than is currently the
case. For example, the number of future annual average
no-rain days in Upper Thames River Basin is projected to
increase by about 5% from the average condition of 137
days during the period 1961-2002. Furthermore, the fu-
ture warmer temperatures projected by the GCM models
could also enhance the low-streamflow situation due to
increased evapotranspiration capacities.
5.2. Changes in Future Return Values of
One-day Maximum Streamflow Events
The statistical return period analysis was employed to
project the return values of one-day maximum stream-
flow events for a number of return periods. A return pe-
riod also known as a recurrence interval is an estimate of
the likelihood of events like one-day maximum stream-
flow volume of a certain intensity. It is a statistial mea-
surement denoting the average recurrence interval over
an extended period of time. Return values are thresholds
that will be exceeded on average once every return pe-
riod. The design of stormwater infrastructure is cons-
trained by the largest precipitation/streamflow event anti-
cipated during a fixed design period (e.g., 20, 50 or 100
years). Due to climate change, the return values of one-
day maximum streamflow events could increase in the
future.
To project future return values of one-day maximum
streamflow events, an annual series of historical one-day
maximum streamflow events were fitted to the Gumbel
(Extreme Value Type I) distribution for each of the se-
lected river basins. The streamflow data observed for the
entire period used in the study were applied to determine
the return values for the historical period. The projected
future daily streamflow data, using three downscaled
CGCM simulations for three 50-year periods (2001-2050,
2026-2075, and 2051-2100), were used to project future
return values. The results showed that the projected re-
turn values of the one-day maximum streamflow events
for all evaluated return periods (e.g., 20, 25, 30, 50, 100
years) could increase late this century (Figure 4). For
example, in the Grand River Basin, the 20-year return
period values of one-day maximum streamflow events
are 173 m3·s–1 for the past 45 years and potentially 240,
256, and 311 m3·s–1 for the periods 2001-2050, 2026-
2075, and 2051-2100, respectively.
0
50
100
150
200
250
300
350
400
450
500
0 102030405060708090100
One-day Maximum Streamflow (m
3
s
-1
)
Grand River Basin
0
20
40
60
80
100
120
0 102030405060708090100
Rideau River Basin
0
20
40
60
80
100
120
0 102030405060708090100
Return Period (Number of Years)
Upper Thames River Basin
0
5
10
15
20
25
30
35
40
45
0 102030405060708090100
Return period (Number of Years)
One-day Maximum Streamflow (m
3
s
-1
)
Humber River Basin
Observation 2001–2050 2026–2075 2051–2100
Figure 4. Return values of annual one-day maximum streamflow volumes as shown by observation (Grand and Upper
Thames: 1961-2002; Humber: 1967-2002; Rideau: 1970-2002) and downscaled three-CGCM ensemble (2001-2050, 2026-2075,
and 2051-2100).

C. S. CHENG ET AL.
436
In addition to the increase in return values shown in
Figure 4, the relative increases in future return values of
one-day maximum streamflow volume from the current
condition were analyzed. The relative increases are more
similar across the return periods and the river basins than
the absolute increases of the return values, as shown in
Table 6. Across the selected river basins, on average of
the three downscaled CGCM simulations, the return val-
ues of the one-day maximum streamflow volume for
various return periods (i.e., 2, 5, 10, 15, 20, 25, 30, 50,
100 years) are projected to increase by approximately
15% - 40%, 25% - 50%, and 30% - 80% in the future
three 50-year periods 2001-2050, 2026-2075, and 2051-
2100, respectively. More specifically, for example, in the
Grand River Basin, the 20-year return-period values of
one-day maximum streamflow volume for future three
50-year periods are projected to increase by 39%, 48%,
and 80%, respectively from the observed value of 173
m3·s–1 for the past 45 years. Among the three down-
scaled CGCM simulations, the projected percentage
increases in the return values are similar, usually with
slightly greater values derived from downscaled CGCM-
A2 than those from downscaled CGCM-B2. For exam-
ple, across four selected river basins, the difference in
the percentage increases between downscaled CGCM2-
A2 and CGCM2- B2 is usually less than 10% for future
three 50-year periods, with a few exemptions. In addi-
tion, from Table 6, it can be seen that the 95% confi-
dence interval for future projected return-period values
is similar to the observed ones, which implies that the
future projected return-period values are plausibly
reliable.
Table 6. Percentage increases in future annual one-day maximum streamflow volumes for various return periods (down-
scaled three-CGCM ensemble) from current observed values (95% confidence interval in parentheses).
Thames River Basin Grand River Basin
Obs. 2001-20502026-2075 2051-2100Obs. 2001-2050 2026-2075 2051-2100
Return Period
(Year)
(m3/s) (%) (%) (%) (m3/s) (%) (%) (%)
2 20 (±7.5) 36 (±6.5)44 (±5.6) 56 (±5.2)61 (±5.9) 23 (±9.8) 35 ( ±11) 53 ( ±11)
5 38 (±5.0) 21 (±4.3)27 (±3.7) 36 (±3.4)110 (±4.1) 34 (±6.7) 44 (±7.6) 72 (±7.5)
10 50 (±5.2) 16 (±4.4)22 (±3.8) 31 (±3.5)142 (±4.3) 37 (±7.0) 46 (±7.9) 77 (±7.8)
15 56 (±5.4) 15 (±4.5)21 (±3.9) 29 (±3.6)160 (±4.4) 38 (±7.3) 47 (±8.2) 79 (±8.1)
20 61 (±5.4) 14 (±4.6)20 (±4.0) 28 (±3.7)173 (±4.5) 39 (±7.5) 48 (±8.4) 80 (±8.3)
25 64 (±5.6) 13 (±4.7)19 (±4.0) 27 (±3.8)183 (±4.6) 39 (±7.6) 48 (±8.5) 81 (±8.5)
30 67 (±5.7) 13 (±4.8)19 (±4.1) 26 (±3.8)191 (±4.7) 40 (±7.7) 49 (±8.7) 81 (±8.6)
50 75 (±5.9) 12 (±4.9)17 (±4.2) 25 (±4.0)213 (±4.7) 41 (±8.0) 49 (±9.0) 83 (±8.9)
100 86 (±6.0) 11 (±5.1)16 (±4.4) 24 (±4.1)243 (±4.9) 41 (±8.3) 50 (±9.3) 84 (±9.3)
Mean 17 (±4.9)23 (±4.2) 31 (±3.9) 37 (±7.8) 46 (±8.7) 76 (±8.7)
Humber River Basin Rideau River Basin
Obs. 2001-2050 2026-2075 2051-2100 Obs. 2001-2050 2026-2075 2051-2100
Return Period
(Year)
(m3/s) (%) (%) (%) (m3/s) (%) (%) (%)
2 10 (±3.0) 20 (±6.5) 26 (±8.3) 28 (±7.8) 21 (±4.3) 58 (±5.1) 64 (±6.7) 73 (±6.1)
5 14 (±2.1) 26 (±5.6) 40 (±7.2) 44 (±6.7) 34 (±3.2) 41 (±3.8) 50 (±5.0) 54 (±4.6)
10 16 (±3.1) 29 (±6.3) 46 (±8.1) 51 (±7.6) 43 (±3.7) 35 (±4.1) 45 (±5.4) 48 (±5.0)
15 18 (±2.8) 30 (±6.7) 48 (±8.7) 53 (±8.1) 48 (±3.8) 33 (±4.3) 43 (±5.6) 46 (±5.2)
20 19 (±3.2) 30 (±7.0) 50 (±9.1) 55 (±8.5) 52 (±3.8) 31 (±4.4) 42 (±5.8) 44 (±5.4)
25 20 (±3.0) 31 (±7.2) 51 (±9.3) 56 (±8.7) 55 (±4.0) 30 (±4.5) 41 (±5.9) 43 (±5.5)
30 20 (±3.5) 31 (±7.4) 52 (±9.6) 57 (±8.9) 57 (±4.0) 30 (±4.6) 40 (±6.0) 43 (±5.6)
50 22 (±3.6) 32 (±7.8) 53 ( ±10) 59 (±9.5) 63 (±4.3) 28 (±4.8) 39 (±6.3) 41 (±5.8)
100 25 (±3.6) 33 (±8.4) 56 ( ±11) 61 ( ±10) 71 (±4.5) 26 (±5.0) 38 (±6.6) 39 (±6.1)
Mean 29 (±7.0) 47 (±9.0) 52 (±8.4) 35 (±4.5) 45 (±5.9) 48 (±5.5)
Note: To effectively compare the 95% confidence intervals between observed and future projected return values, as the same as the future projections, the 95%
confidence intervals derived from observations are presented as percentages below or above the return values.

C. S. CHENG ET AL. 437
5.3. Uncertainty and Limitations
The uncertainty of climate change impacts on future
heavy rainfall events described in a recent study (Cheng
et al. [21]) also applies to this paper since the projection
of future daily rainfall quantities was used in the current
study to project future daily streamflow volumes. As
described in the study [21], considerable effort was made
to transfer GCM-scale simulations to station-scale cli-
mate information using statistical downscaling transfer
functions. Through the downscaling process, most of the
GCM model bias was removed, using the 50-year his-
torical relationships between regional-scale predictors
and station-scale weather elements [28]. As a result, the
quality of the GCM climate change projections was
much improved after using the statistical downscaling
(Cheng et al. [19]). However, conclusions made in the
current study about the impacts of climate change on
future high-/low-streamflow events still rely on GCM
scenarios/projections and, consequently, there is corre-
sponding uncertainty about the study findings.
To quantitatively assess inter-GCM and interscenario
uncertainties of future daily streamflow projections, we
have analyzed the four-river-basin-average absolute dif-
ference between pairs of eight selected GCM models
under the SRES B1 scenario as well as absolute differ-
ence between two selected scenarios (A2 versus B1). The
absolute difference used in analysis is to avoid negative
values cancelling out positive values. As shown in Table
7, overall, the inter-GCM uncertainties of percentage
increases in the frequency of future daily high-stream-
flow events are greater than the interscenario uncertain-
ties. For daily mean streamflow projections, both uncer-
tainties are similar. From Tables 5 and 7, it can be seen
that the uncertainties of future daily high-streamflow
events are smaller than the future projections. The overall
mean projected percent- age increases in frequency of
future daily high-stream- flow events are about 1.4 - 2.2
times greater than overall mean inter-GCM and intersce-
nario uncertainties. For projections of future daily low-
streamflow events and daily mean streamflow volumes,
the inter-GCM and interscenario uncertainties are gener-
ally similar to or greater than the projected future per-
centage increases.
Although the models developed from this study can
simulate most high-streamflow events for the selected
river basins, it was found that the models have difficulty
in capturing some of the cases, especially for urban river
basins, such as the Black Creek tributary of the Humber
River (refer to Part I, Cheng et al. [22]). This model
limitation is also reflected by the simulation difficulty of
the localized convective heavy rainfall events. As de-
scribed by Cheng et al. [20], the rainfall simulation mod-
els can simulate most heavy rainfall events, but it was
found that the models have difficulty in capturing some
of localized convective heavy rainfall events. It is likely
that projection of changes in frequency of future rain-
fall-related high-streamflow events offered by this study
will represent the lower bound values for the study area.
As a result, southern Ontario could in the future possibly
receive more rainfall-related high-streamflow events than
is currently projected by the study.
In addition to uncertainty of GCM projections and
limitation of daily streamflow simulation models, the
observed streamflow data used in the study have their
limitation. Daily mean streamflow volumes which are
averaged over a 24-hour period (i.e., 00:00-23:00 LST,
local standard time) are currently used in the analysis.
Daily streamflow data are limited in their usefulness for
studying more detailed information on the simulation of
the high-streamflow events, especially for rapidly rain-
fall-streamflow responding urban watersheds (e.g., the
Black Creek tributary of the Humber River). If the short-
duration (less than one day) streamflow data were avail-
able, the streamflow simulation model for Humber River
Basin could possibly be improved by using streamflow
information at a shorter time step.
Furthermore, the limitation of meteorological data,
described in the study [21] for developing rainfall simu-
lation models, also affect daily streamflow historical
simulation and future projection. The major limitation of
Table 7. Four-river-basin-average inter-GCM and interscenario uncertainties of projected percentage increases in the fre-
quency of future seasonal high-streamflow/low-streamflow days and future daily mean streamflow volumes from the current
conditions of May-November 1961-2002.
Uncertainty (absolute difference)
Mean 8-GCM B1a Mean 8-GCM A2 - B1b
Streamflow events Current conditions
2046-2065 2081-2100 2046-2065 2081-2100
High-streamfow (≥95th percentile)
Low-streamflow (<5th percentile)
Daily mean streamflow
10.7 days
10.4 days
0.74 - 6.68 m3s–1
32
40
19
40
40
21
20
28
11
31
53
21
aMean 8-GCM B1 is average of absolute differences between pairs of eight GCM B1 models; bMean 8-GCM A2 - B1 is average of absolute differences be-
tween scenarios A2 and B1 of eight GCM models.

C. S. CHENG ET AL.
438
meteorological data includes that hourly meteorological
observations are not available in the Grand River Basin,
which are essential to develop synoptic weather typing
and rainfall simulation models. Consequently, hourly
meteorological data gathered at the London International
Airport (located in Thames watershed) were used to clas-
sify synoptic weather types and to derive rainfall predic-
tors (e.g., atmospheric stability indices) for the Grand
River. Therefore, the rainfall simulation results, derived
for the Grand River Basin, were not as accurate as they
might be were hourly meteorological data for the Grand
River Basin available. In turn, these results could affect
projection of frequency of future heavy rainfall events
and ultimately influence on projections of frequency of
future high-streamflow events derived from this study.
6. Conclusions and Recommendation
The overall purpose of this study is to project possible
changes in the frequency of high-/low-streamflow events
late this century for four selected river basins (i.e., Grand,
Humber, Rideau, and Upper Thames) in Ontario, Canada.
To achieve this goal, the streamflow simulation models
developed in Part I (Cheng et al. [22]) were applied col-
lectively with downscaled future GCM simulations. As
described in the studies (Cheng et al. [19,22]), a formal
verification process of model results has been built into
the whole exercise, comprising daily streamflow simula-
tion modeling and the development of downscaling
transfer functions. The study results demonstrate that the
streamflow simulation models are able to reproduce daily
streamflow volumes in the observed period for the se-
lected river basins, through model calibration and verifi-
cation. Furthermore, in this current study, the streamflow
simulation models were evaluated using downscaled
GCM historical runs to ascertain whether the models are
suitable for projecting future daily streamflow volumes.
The results of the verification in terms of data distribu-
tion and frequency of the high-/low-streamflow events,
based on historical observations of the outcome variables
simulated by the models, showed good agreement. As a
result, a general conclusion from this study is that a
combination of the streamflow modeling and statistical
downscaling can be useful to project changes in fre-
quency of future daily high-/low-streamflow events.
The modeled results from this study found that due to
a changing climate, the frequency of future high- and
low-streamflow events for the period 2046-2065 is pro-
jected by averaging eight-GCM A2 projections to in-
crease by about 45% and 25%, respectively from the
current condition, which are similar to the increases by
averaging eight-GCM B1 projections. The corresponding
increases for the period 2081-2100 are much different
between two scenarios: high-streamflow events will in-
crease by 45% and 55% derived from eight-GCM sce-
narios B1 and A2, respectively; for low-streamflow
events, the corresponding increases are 55% and 89%.
These findings are consistent with the results reported in
the IPCC Fourth Assessment Report [29], in terms of the
increase tendency in high- and low-streamflow events.
The implication of these increases should be taken into
consideration when revising engineering infrastructure
design standards (including infrastructure maintenance
and new construction) and developing adaptation strate-
gies and policies. As the IPCC [29] pointed out, “More
extensive adaptation than is currently occurring is re-
quired to reduce vulnerability to future climate change.”
This study aims to provide decision makers with scien-
tific information needed to improve the adaptive capacity
of the infrastructure at risk of being impacted by heavy
rainfall-related flooding in Ontario due to climate change.
The results of the study are intended to contribute to the
Ontario Emergency Management and Civil Protection
Act under Bill 148, which attempts to reduce risks of dis-
asters by requiring that all municipalities, regional gov-
ernments, and provincial ministries develop emergency
and disaster risk management plans [30,31].
7. Acknowledgements
The authors gratefully acknowledge the funding support
of the Government of Canada’s Climate Change Impacts
and Adaptation Program (CCIAP), which made this re-
search project (A901) possible. We also acknowledge the
suggestions made by the Project Advisory Committee,
which greatly improved the study
REFERENCES
[1] F. Bultot, A. Coppens, G. L. Dupriez, D. Gellens and F.
Meulenberghs, “Repercussions of a CO2 Doubling on the
Water Cycle and on the Water Balance—A Case Study
for Belgium,” Journal of Hydrology, Vol. 99, No. 3-4,
1988, pp. 319-347. doi:10.1016/0022-1694(88)90057-1
[2] D. Panagoulia and G. Dimou, “Sensitivity of Flood Events
to Global Climate Change,” Journal of Hydrology, Vol.
191, No. 1-4, 1997, pp. 208-222.
doi:10.1016/S0022-1694(96)03056-9
[3] G. Drogue, L. Pfister, T. Leviandier, A. El Idrissi, J-F.
Iffly, P. Matgen, J. Humbert and L. Hoffmann, “Simulat-
ing the Spatio-Temporal Variability of Streamflow Re-
sponse to Climate Change Scenarios in a Mesoscale Ba-
sin,” Journal of Hydrology, Vol. 293, No. 1-4, 2004, pp.
255-269. doi:10.1016/j.jhydrol.2004.02.009
[4] D. Cameron, “An Application of the UKCIP02 Climate
Change Scenarios to Flood Estimation by Continuous
Simulation for a Gauged Catchment in the Northeast of
Scotland, UK (with Uncertainty),” Journal of Hydrology,
Vol. 328, No. 1-2, 2006, pp. 212-226.
doi:10.1016/j.jhydrol.2005.12.024

C. S. CHENG ET AL. 439
[5] K. A. Forbes, S. W. Kienzle, C. A. Coburn, J. M. Byrne
and J. Rasmussen, “Simulating the Hydrological Re-
sponse to Predicted Climate Change on a Watershed in
Southern Alberta, Canada,” Climatic Change, Vol. 105,
No. 3-4, 2011, pp. 555-576.
doi:10.1007/s10584-010-9890-x
[6] A. L. Kay, R. G. Jones and N. S. Reynard, “RCM Rain-
fall for UK Flood Frequency Estimation. II. Climate
Change Results,” Journal of Hydrology, Vol. 318, No.
1-4, 2006, pp. 163-172.
doi:10.1016/j.jhydrol.2005.06.013
[7] D. Panagoulia and G. Dimou, “Linking Space—Time
Scale in Hydrological Modelling with Respect to Global
Climate Change: Part 2. Hydrological Response for Al-
ternative Climates,” Journal of Hydrology, Vol. 194, No.
1-4, 1997, pp. 38-63.
doi:10.1016/S0022-1694(96)03221-0
[8] T. Jiang, Y. D. Chen, C. Y. Xu, X. H. Chen, X. Chen and
V. P. Singh, “Comparison of Hydrological Impacts of
Climate Change Simulated by Six Hydrological Models
in the Dongjiang Basin, South China,” Journal of Hy-
drology, Vol. 336, 2007, pp. 316-333.
[9] A. W. Wood, D. P. Lettenmaier and R. N. Palmer, “As-
sessing Climate Change Implications for Water Resources
Planning,” Climatic Change, Vol. 37, No. 1, 1997, pp.
203-228. doi:10.1023/A:1005380706253
[10] L. E. Hay, R. L. Wilby and G. H. Leavesley, “A Com-
parison of Delta Change and Downscaled GCM Scenarios
for Three Mountainous Basins in the United States,”
Journal of American Water Resources Association, Vol.
36, No. 2, 2000, pp. 387-397.
doi:10.1111/j.1752-1688.2000.tb04276.x
[11] L. Sushama, R. Laprise, D. Caya, A. Frigonb and M.
Slivitzkyb, “Canadian RCM Projected Climate-Change
Signal and Its Sensitivity to Model Errors,” International
Journal of Climatology, Vol. 26, No. 15, 2006, pp. 2141-
2159. doi:10.1002/joc.1362
[12] R. L. Wilby and T. M. L. Wigley, “Downscaling General
Circulation Model Output: A Review of Methods and
Limitations,” Progress in Physical Geography, Vol. 21,
No. 4, 1997, pp. 530-548.
doi:10.1177/030913339702100403
[13] R. G. Najjar, “The Water Balance of the Susquehanna
River Basin and Its Response to Climate Change,” Jour-
nal of Hydrology, Vol. 219, No. 1-2, 1999, pp. 7-19.
doi:10.1016/S0022-1694(99)00041-4
[14] P. H. Whitfield, J. Y. Wang and A. J. Cannon, “Model-
ling Future Streamflow Extremes—Floods and Low
Flows in Georgia Basin, British Columbia,” Canadian
Water Resources Journal, Vol. 28, No. 4, 2003, pp. 633-
656. doi:10.4296/cwrj2804633
[15] M. D. Dettinger, D. R. Cayan, M. K. Meyer and A. E.
Jeton, “Simulated Hydrologic Responses to Climate Varia-
tions and Change in the Merced, Carson, and American
River Basins, Sierra Nevada, California, 1900-2099,”
Climatic Change, Vol. 62, No. 1-3, 2004, pp. 283-317.
doi:10.1023/B:CLIM.0000013683.13346.4f
[16] K. Jasper, P. Calanca, D. Gyalistras and J. Fuhrer, “Dif-
ferential Impacts of Climate Change on the Hydrology of
Two Alpine River Basins,” Climatic Research, Vol. 26,
No. 2, 2004, pp. 113-129. doi:10.3354/cr026113
[17] Y. B. Dibike and P. Coulibaly, “Hydrologic Impact of
Climate Change in the Saguenay Watershed: Comparison
of Downscaling Methods and Hydrologic Models,”
Journal of Hydrology, Vol. 307, No. 1-4, 2005, pp.
145-163.
doi:10.1016/j.jhydrol.2004.10.012
[18] W. S. Merritt, Y. Alisa, M. Barton, B. Taylor, S. Cohen
and D. Neilsen, “Hydrologic Response to Scenarios of
Climate Change in Sub-Watersheds of the Okanagan Ba-
sin, British Columbia,” Journal of Hydrology, Vol. 326,
2006, No. 1-4, pp. 79-108.
doi:10.1016/j.jhydrol.2005.10.025
[19] C. S. Cheng, G. Li, Q. Li and H. Auld, “Statistical
Downscaling of Hourly and Daily Climate Scenarios for
Various Meteorological Variables in South-Central Can-
ada,” Theoretical and Applied Climatology, Vol. 91, No.
1-4, 2008, pp. 129-147. doi:10.1007/s00704-007-0302-8
[20] C. S. Cheng, G. Li, Q. Li and H. Auld, “A Synoptic
Weather Typing Approach to Simulate Daily Rainfall and
Extremes in Ontario, Canada: Potential for Climate Change
Projections,” Journal of Applied Meteorology and Cli-
matology, Vol. 49, No. 5, 2010, pp. 845-866.
doi:10.1175/2010JAMC2016.1
[21] C. S. Cheng, G. Li, Q. Li and H. Auld, “A Synoptic
Weather-Typing Approach to Project Future Daily Rain-
fall and Extremes at Local Scale in Ontario, Canada,”
Journal of Climate, Vol. 24, No. 14, 2011, pp. 3667-3685.
doi:10.1175/2011JCLI3764.1
[22] C. S. Cheng, Q. Li, G. Li and H. Auld, “Possible Impacts
of Climate Change on Daily Streamflow and Extremes at
Local Scale in Ontario, Canada,” Part I: historical simula-
tion. Atmospheric and Climate Sciences, Vol. 2, No. 4,
2012, pp. 416-426.
[23] PCMDI, “WCRP CMIP3 Multi-Model Dataset,” 2008.
http://www-pcmdi.llnl.gov/ipcc/about_ipcc.php
[24] Environment Canada, “DATA,” 2006.
http://www.cccma.bc.ec.gc.ca/data/data.shtml
[25] J. G. Galway, “The Lifted Index as a Predictor of Latent
Instability,” Bulletin of the American Meteorological So-
ciety, Vol. 37, 1956, pp. 528-529.
[26] J. J. George, “Weather Forecasting for Aeronautics,”
Academic Press, Waltham, 1960.
[27] R. C. Miller, “Notes on Analysis and Severe Storm Fore-
casting Procedures of the Air Force Global Weather Cen-
tral,” Technical Report 200(R), Headquarters US Air
Weather Service, Omaha, 1972.
[28] R. W. Katz, “Techniques for Estimating Uncertainty in
Climate Change Scenarios and Impact Studies,” Climatic
Research, Vol. 20, No. 2, 2002, pp. 167-185.
doi:10.3354/cr020167
[29] IPCC, “Summary for Policymakers,” In: S. Solomon, D.
Qin, M. Manning, Z. Chen, M. Marquis, K. B. Averyt, M.
Tignor and H. L. Miller, Eds., Climate Change 2007: The
Physical Science Basis. Contribution of Working Group I
to the Fourth Assessment Report of the Intergovernmental
Panel on Climate Change, Cambridge University Press,

C. S. CHENG ET AL.
440
Cambridge and New York, 2007, 18 pp.
[30] Ontario Management Act: Legislative Assembly of On-
tario. “Bill 148 2002: An Act to Provide for Declarations
of Death in Certain Circumstances and to Amend the
Emergency Plans Act,” 2006.
http://www.ontla.on.ca/documents/Bills/37_Parliament/S
ession3/b148ra_e.htm
[31] H. Auld, D. MacIver, J. Klaassen, N. Comer and B. Tug-
wood, “Atmospheric Hazards in Ontario,” ACSD Science
Assessment Series No. 3 Meteorological Service of Can-
ada, Environment Canada, Toronto, 2004.
[32] C. E. M. Sefton and D. B. Boorman, “A Regional Inves-
tigation of Climate Change Impacts on UK Streamflows,”
Journal of Hydrology, Vol. 195, No. 1-4, 1997, pp. 26-44.
doi:10.1016/S0022-1694(96)03257-X
[33] D. Gellens and E. Roulin, “Streamflow Response of Bel-
gian Catchments to IPCC Climate Change Scenarios,”
Journal of Hydrology, Vol. 210, No. 1-4, 1998, pp. 242-
258. doi:10.1016/S0022-1694(98)00192-9
[34] N. W. Arnell, “Relative Effects of Multi-Decadal Cli-
matic Variability and Changes in the Mean and Variabil-
ity of Climate due to Global Warming: Future Stream-
flows in Britain,” Journal of Hydrology, Vol. 270, No.
3-4, 2003, pp. 195-213.
doi:10.1016/S0022-1694(02)00288-3
[35] K. Eckhardt and U. Ulbrich, “Potential Impacts of Cli-
mate Change on Groundwater Recharge and Streamflow
in a Central European Low Mountain Range,” Journal of
Hydrology, Vol. 284, No. 1-4, 2003, pp. 244-252.
doi:10.1016/j.jhydrol.2003.08.005
[36] N. W. Arnell and N. S. Reynard, “The Effects of Climate
Change due to Global Warming on River Flows in Great
Britain,” Journal of Hydrology, Vol. 183, No. 3-4, 1996,
pp. 397-424. doi:10.1016/0022-1694(95)02950-8