Data Test
Resume:
Theor Appl Climatol (****) ***:*** ***
ORIGINAL PAPER
Characterization and estimation of urban heat island
at Toronto: impact of the choice of rural sites
Tanzina Mohsin & William A. Gough
Received: 24 February 2011 / Accepted: 23 August 2011 / Published online: 9 September 2011
# Springer-Verlag 2011
Abstract In this study, the urban heat island of Toronto was The analysis from the current study suggests that the
selection of a unique urban rural pair to estimate UHI
characterized and estimated in order to examine the impact of
the selection of rural sites on the estimation of urban heat intensity for a city like Toronto is a critical task, as it will be
island (UHI) intensity ( Tu-r). Three rural stations, King for any city, and it is imperative to consider some key
Smoke Tree (KST), Albion Hill, and Millgrove, were used features such as the physiography, surface characteristics of
for the analysis of UHI intensity for two urban stations, the urban and rural stations, the climatology such as the
Toronto downtown (Toronto) and Toronto Pearson (Pearson) trends in annual and seasonal variation of UHI with respect
using data from 1970 to 2000. The UHI intensity was to the physical characteristics of the stations, and also more
characterized as winter dominating and summer dominating, importantly the objectives of a particular study in the context
depending on the choice of the rural station. The analyses of of UHI effect.
annual and seasonal trends of Tu-r suggested that urban heat
island clearly appears in winter at both Toronto and Pearson.
1 Introduction
However, due to the mitigating effect on temperature from
Lake Ontario, the estimated trend of UHI intensity was
The Urban Heat Island (UHI) phenomenon has been
found to be less at Toronto compared to that at Pearson
investigated for many cities around the world, which has
which has no direct lake effect. In terms of the impacts of the
triggered a surge in the empirical urban heat island literature
rural stations, for both KST and Millgrove, the trends in UHI
(Oke 1973; Ackerman 1985; B hm 1998; Gough and
intensity were found to be statistically significant and also
Rozanov 2001; i ek and Do a 2006, among others). UHI
were in good agreement with the estimates of UHI intensities
is defined as the relative warmth of a city compared to the
reported for other large cities in the USA. Depending on the
surrounding rural areas (Oke 1976). The primary causes of
choice of the rural station, the estimated trend for the UHI
UHI effect are well described in the urban climate literature.
intensity at Toronto ranges from 0.01 C/decade to 0.02 C/
Some of the most common ones are suggested by Oke
decade, and that at Pearson ranges from 0.03 C/decade to
(1982) as increased absorption of sun s radiation due to
0.035 C/decade during 1970 2000. From the analysis of the
seasonal distribution of Tu-r, the UHI intensity was found to canyon geometry, increased long-wave radiation from the
sky due to city air pollution, decreased long-wave radiation
be higher at Toronto in winter than that at Pearson for all
loss because of the reduction of the sky view factor,
three rural stations. This was likely accounted for by the
anthropogenic heat sources, increases in sensible heat
lower amount of anthropogenic heat flux at Pearson.
storage, decreased evapotranspiration, and decreased total
Considering the results from the statistical analysis with
turbulent heat transport due to wind speed reduction caused
respect to the geographic and surface features for each rural
by canyon geometry. The thermal, moisture, aerodynamic,
station, KST was suggested to be a better choice to estimate
and radiative properties of a city are dramatically different
UHI intensity at Toronto compared to the other rural stations.
from those of rural areas because of the replacement and
T. Mohsin : W. A. Gough vertical screening of natural surfaces with perpendicular
Department of Physical and Environmental Sciences, structures and building materials of high heat capacity and
University of Toronto Scarborough,
low permeability (Oke 1982).
1265 Military Trail,
The analysis of UHI effect involves estimating the
Toronto, ON M1C 1A4, Canada
screen-level air temperature differences between pairs of
e-mail: *******.******@********.**
106 T. Mohsin, W.A. Gough
urban and rural climate stations, or among selected standpoint. A recent study on the impact of different rural
stations to compute urban rural temperature difference
urban or rural measurement sites along a mobile
traverse route. Universally denoted as Tu-r, the analysis suggested that the heat island magnitude can be under-
estimated or overestimated if the selection of rural station is
of UHI intensity has been the backbone of UHI climatology,
not appropriate (Sakakibara and Owa 2005). Only a few
which has largely remained unchanged during the past few
studies on the role of rural stations in assessing urban heat
decades. The UHI intensity is unique to individual cities and
island effects exist in the current literature, and from this
their geographies. Nasrallah et al. (1990) suggested that the
perspective, the objectives of this study will have important
UHI in Phoenix, Arizona, is well developed unlike the UHI
contribution to future research on the estimation of UHI
in Kuwait City, in a similar climate. The study used the
intensity.
airport as the urban station and a desert location as the rural
In this work, we focus on Toronto s UHI with a goal to
station, and the reason for the poorly developed UHI was
identify the appropriate urban and rural station pair in order
attributed to the location of Kuwait City near the Arabian
to quantify the UHI intensity, in particular, the impact of the
Gulf. Ripley et al. (1996) used the mobile traverse method
choice of a rural station to estimate Tu-r. Although a
and found maximum UHI intensities for sunny days in
number of studies exist, which analyzed Toronto s heat
Saskatoon under clear, calm conditions, an observation
consistent with Oke (1982). A more recent study on the island, but these are mainly based on the framework
UHI detection at Ankara, Turkey, looked at the seasonal proposed by Lowry (1977). These studies on Toronto
distribution of UHI using an urban and a typical rural station considered the differences in observations from the urban
( i ek and Do a 2006). The study suggested that the UHI and pre-urban periods (Gough and Rozanov 2001, Mohsin
and Gough 2010) instead of the Tu-r approach. In the
intensities at Ankara have statistically significant increasing
trends for winter and spring after the 1980s. A few empirical current study, three rural stations, situated north, northwest,
analyses of UHI effect have been done on Canadian cities, and southwest of Toronto, Ontario, Canada, are used for the
particularly, for Regina, Montr al, Vancouver, and Toronto. estimation of UHI intensity. No other rural stations were
For Regina, Montr al, and Vancouver, the mobile traverse available with the data for the study period of the current
method was used, which considered a range of distinct land- analysis within the Greater Toronto Area (GTA). Two urban
use types including urban, residential, industrial, urban park, stations are considered, one of which is situated at
and rural to assess the impact of UHI Stewart (2000). These downtown Toronto. Since Toronto downtown station is
analyses suggested that the climatological controls in both situated near Lake Ontario, therefore, to account for the
urban and rural areas play important roles in determining the lake effect, another urban station at Toronto Pearson Airport
intensity of UHI. In a study for Toronto, the downtown core is used to estimate the UHI intensity. A detailed description
was considered as urban and Pearson Airport and Vineland of the stations is given in Section 2, which also includes the
were considered as rural stations (Gough and Rozanov methods and data that are used in this study. The results
2001). The study compared two different periods (1926 from the trend analysis of annual and seasonal time
1936 and 1977 1987) to identify the impact of urbanization series of UHI intensity ( Tu-r) and the seasonal distribu-
tion of Tu-r are presented in Section 3. The UHI at
on temperature change and suggested that UHI is more
pronounced after the post-World War II period when Toronto Toronto and Pearson are characterized and estimated based
became considerably more urbanized. on the results, and suggestions are made on the important
In order to ascertain an unambiguous measure of the criteria to choose a rural station for the estimation of UHI
UHI effect, it is crucial to choose the urban and rural station intensity.
pair using clear, objective, and climatologically significant
criteria and take into account the micro- and local-scale
physical phenomenon. The choice of a rural station is 2 Methodology
particularly crucial. In some UHI studies, the use of airports
represents either urban or rural (Adebayo 1991; Klysik 2.1 Description of the sites
and Fortuniak 1999), which introduces a less coherent and
increasingly worrisome dimension to the representation of The Toronto downtown climate station is considered as an
urban and rural spaces. The concern surrounding the urban urban station since World War II, after which significant
rural dichotomy was raised by Stewart and Oke (2006), urbanization has been taken place in and around the city
who later proposed a Thermal Climate Zones classifica- (Mohsin and Gough 2010). Although in some previous
tion system for classifying UHI measurement sites at the studies Toronto Pearson Airport was considered as a rural
local scale (Stewart and Oke 2009). The choice of station (Gough and Rozanov 2001), during the past 30 years,
unsuitable rural sites to analyze UHI effect may produce massive urbanization has taken place around this station,
results that might be of no importance from a climatological which now exhibits the characteristics of an urban station.
Characterization and estimation of urban heat island at Toronto 107
According to the Thermal Climate Zones classification natural series . This station is also close to Lake
system (Stewart and Oke 2009), the Toronto downtown site Ontario compared to the other two rural stations, which
can be classified as modern core under the city series with are located further inland. KST is situated to the north of
high surface roughness, high impervious fraction, low sky Toronto (Fig. 1) and is located near the protected area of
view factor, low thermal admittance, low albedo, and high the Oak Ridges Moraine and nearly 70% of the lands of
anthropogenic heat flux. The other urban station, Toronto King Township to this date retain much of its rural
Pearson, cannot be categorized under a single building block characteristics with an assortment of agricultural and farm
of the city series and needs to be considered as a lands (King Township History and Heritage 2006).
Therefore, KST can be classified under the agriculture
combination of old core, compact housing, and blocks.
series, which is surrounded by all four building blocks of
Therefore, Toronto Pearson is an urban site that falls into a
category with moderate surface roughness, high sky view flooded fields, orchards and vineyard, cropped fields, and
factor, and lower anthropogenic heat flux. The rural stations bare fields. All the stations that are used in this study are
used in this study are Albion Hill, Millgrove, and King located within the boundary of the GTA.
Smoke Tree (KST), which are located outside of Toronto at a
distance of approximately 30 50 km (Fig. 1). Albion Hill is 2.2 Data and analysis
located near the Albion Hills conservation area and is
surrounded by the Oak Ridge Moraine, and is situated to The temperature data for all stations are available from
the northwest of Toronto. There are a few farms, golf Environment Canada (EC) as daily temperature data, from
courses, and campgrounds surrounding this rural station, which mean monthly, seasonal, and annual temperature data
which can be classified as a combination of forest and series are produced. Data from all the stations are subjected
wetlands under the natural series of the Thermal Climate to all steps of the quality control process (Vincent and
Zones classification (Stewart and Oke 2009). Millgrove, Gullett 1999).
which is currently a part of the amalgamated city of The typical method to estimate Urban Heat Island (UHI)
Hamilton, is a small town and is located near the Niagara intensity is to use the temperature differences between an
urban station and a surrounding rural station ( Tu-r). In this
Escarpment. It is situated to the southwest of Toronto
and Pearson, and on an average was 0.6 C warmer than study, the time series of temperature differences are obtained
KST and 1 C warmer than Albion Hill over the study using the two urban stations and the three rural stations
period. Millgrove can also be classified under the described above. The analysis of the UHI is carried out for
Fig. 1 Locations of the
meteorological stations used
in the analysis
108 T. Mohsin, W.A. Gough
the interval of 1970 2000 because most urbanization in The MK test determines whether the observations in the
Toronto has taken place during this period (Mohsin and data tend to increase or decrease with time. The MK test is
also referred to as Kendall s tau when x-axis is the time,
Gough 2010). The time series of temperature differences
(urban rural) are tested for homogeneity using the Mann which is the case in our research. The null hypothesis for
Whitney (MW) homogeneity test. The MW test is a non- this test states that all observations are independent; on the
parametric rank-based test for identifying difference between other hand, the alternative hypothesis assumes that a
two samples with respect to their median or means. The two monotonic trend, positive or negative, exists in the time
samples are combined, and all sample observations are series (Helsel and Hirsch 1992). In this analysis, the MK
ranked from smallest to largest. If the two samples have the test is applied to detect if a trend in the temperature time
same distribution, then the sum of the ranks of the first series is statistically significant at 0.05(95%) and 0.01
sample and those in the second sample should be close to the (99%) significant levels (confidence intervals) for a two-
same value (Yue and Wang 2002). The advantage of the MW sided probability. The MK test, however, does not provide
test compared to other parametric tests (e.g., Bartlett test, t an estimate of the magnitude of the trend. For this purpose,
a non-parametric method referred to as the Theil Sen
test) is that the test can be applied to data without the
assumptions of normal distribution and to large data samples approach (TSA) is used to compute the magnitude of UHI
(von Storch and Zwiers 1999). Therefore, it is more robust intensities. This provides a more robust slope estimate than
for temperature measurements, which are affected by the least-squares method because outliers or extreme values
changes in the location of the weather station, changes in in the time series affect it less (Sen 1968). The algorithm for
observing practices, and changes in instrumentation. The TSA is derived by Hirsch et al. (1982) and consists of the
trend in the UHI intensity for Toronto and Pearson are median of all possible pairwise slopes in the dataset.
analyzed using the Mann Kendall test, which is described In addition to the MK test, the sequential Mann Kendall
test is also applied to the time series of Tu-r to detect any
below.
significant trend. The advantage of sequential MK test over
2.2.1 Mann Kendall test the MK test is that it allows visualizing the year-to-year
changes in the time series for the study period. The
application of sequential Mann Kendall test has the
The trend analysis of each time series requires testing for
serial auto-correlation as a first step before applying the following steps in sequence:
Mann Kendall (MK) test. If the observations in the time
1. The values of the original series xi are replaced by their
series are correlated with preceding or successive observations,
ranks yi, arranged in ascending order
then serial correlation exists in the data. The MK test is
2. The magnitudes of yi (i =1, N) are compared with
applicable only when all the observations in a time series are
yj; that is, the n should be replaced with N (j =1,,
serially independent. If there is a positive serial correlation
i 1). At each comparison, the number of cases yi > yj
(persistence), then it increases the sample variance, and the
needs to be counted and denoted by ni
MK test may falsely detect a significant trend in the time series
3. A statistic ti can, therefore, be defined as follow:
(Helsel and Hirsch 1992). It was suggested by von Storch
(1995) that the time series be pre-whitened to eliminate
t i ni
the effect of serial auto-correlation before applying the MK
test. The method involves removing the lag-1 correlation
coefficient from the time series using autoregressive and
4. The distribution of the test statistic has a mean and a
integrated moving average models. However, it should be
variance as:
noted that the pre-whitening method removes part of the trend
in the data while eliminating the serial auto-correlation
component, which can influence the estimate of the trend
E ti i i 1 =4
(Zhang et al. 2000; Yue et al. 2002). To overcome this
issue, Yue et al. (2002) suggested that the time series be
de-trended before the pre-whitening process is applied. This
and var ti i i 1 2i 5 =72
research incorporates this suggestion and used the following
procedures: (a) compute the lag-1 serial auto-correlation
coefficient (r1); (b) if the calculated r1 is not significant at
5. The sequential values of the statistic u(ti) can then be
the 5% level, then MK test is applied to original values of the
computed as:
time series; and (c) if the calculated r1 is significant, then the
Yue et al. (2002) method is used prior to the application of
u ti ti E ti =SQRT var ti
MK test.
Characterization and estimation of urban heat island at Toronto 109
Here, u(ti) is a standardized variable that has zero mean rural temperature, then it is obvious that the rural temperature
(Tr) should have an impact on Tu-r. However, while
and unit standard deviation. Therefore, its sequential
behavior fluctuates around zero level. Climate change can considering the screen-level temperature of an area, it should
be detected by Kendall coefficient (ti), and when a time be noted that any quantitative changes in meteorological
series shows a significant trend, the period from which the factors such as wind, clouds, and near-surface temperature
trend is demonstrated can be obtained effectively by this lapse rate are associated with the changes in temperature of
that particular area. Therefore, Tu-r has a physical identity
test. More information on this procedure can be obtained
from the WMO paper by Sneyers (1990). because it is a result of a number of physical phenomena and
should not be viewed as simple mathematical differences of
temperatures. The study also suggested that the trend in
temperature for Tr has an impact on the trends of Tu-r, which
3 Results and discussion
could impose a bias on the estimation of the UHI intensity
3.1 Choosing a rural station: correlation analysis based on the choice of a rural station. Therefore, it is possible
that the temperature trends of rural stations may impose bias
The choosing of urban/rural station pairs is a challenging task. on the estimation of UHI intensity. The idea is thus to
The description of urban and rural sites and their screen-level identify a rural station for which the surface characteristics
temperature regimes is crucial, and so far in most of the UHI are unchanging. Here, we suggest that if the temperature
time series between Tu-r and Tr shows no correlation, then
studies, this was based on the criteria of representativeness,
known temperature regimes, local station networks, population the physical interpretation to this statistical observation is
distribution, and access to land cover description and data that such rural station is an appropriate one to measure the
UHI intensity. Since in most urban rural settlements such
availability (Eliasson 1994; Unger 1996; B hm 1998; Stewart
2000; Hansen et al. 2001; Hedquist and Brazel 2006; i ek observation is unlikely, it is thus plausible to choose a rural
and Do a 2006; Stone 2007). A major difference among station that shows the least correlation between the rural
these studies is that some analyses choose the urban rural station and UHI values. The station pair with the least
correlation factor should represent a rural station that have a
pair with respect to human geography (population) and others
considered physical geography (location of station, land similar background and landscape scale climate as that of the
cover, etc.) to construct an urban rural pair. A critical look at urban station.
the current urban heat island literature suggests the need for a In this work, a correlation analysis between each of the
more coherent selection of the urban rural pair for quantifying urban and rural pair is done for the preliminary selection of
the UHI as an urban rural temperature difference ( Tu-r). It a rural station to quantify the UHI intensity. The linear
correlation coefficient, CC, is computed between Tu-r and
was first suggested by Lowry (1977) that while estimating
urban rural difference of any weather element, one should Tr using the following equation:
consider the differences of the background climate, the local
CC s u =s Tu r CCur s r= s u
landscape, and the effect of the urbanization. Lowry also
suggested that there is an influence of the location of the
where u and r are the standard deviations of the yearly mean
boundary between the urban and rural stations on the
estimation of the urban rural difference of an element, urban and rural temperatures, respectively. In (CCur r/ u),
which could be different for different weather elements. He CCur is the correlation between Tr and Tu. The result from the
correlation analysis is presented in Table 1 for both Toronto
thus suggested that urban effects may exist beyond the urban
boundary even in the nearby rural areas and may have impact and Pearson using the three rural stations in the surrounding
on the estimation of urban rural difference of any element areas.
The results from the correlation analysis show that for all
(Lowry 1977).
the rural stations, Tu-r is negatively correlated to Tr but at
In a later study, Camilloni and Barros (1997) suggested that
the UHI intensity depends, among other parameters, on the differing strengths. The rural temperature has the least
correlation with Tu-r in the case of KST compared to
temperature of the rural areas. It can be argued that since
Tu-r is computed from the difference between urban and Millgrove and Albion Hill for both Toronto and Pearson. To
Table 1 Results from the
Stations Toronto (CCur) Toronto (CC) Pearson (CCur) Pearson (CC)
correlation analysis between
Tu-r and Tr
0.60 0.46
Albion Hill 0.38 0.39
0.51 0.20
Millgrove 0.87 0.82
0.44 0.15
KST 0.85 0.86
110 T. Mohsin, W.A. Gough
further clarify the selection on the appropriate rural station, a exception in the case of Millgrove. The year 1998 for Albion
correlation analysis has been done for the trends of Tu-r and Hill is removed from the analysis due to a few months of
Tr which are computed using the sequential Mann Kendall missing data. The heat island intensities are a little higher for
test for both Toronto and Pearson using the three rural Toronto compared to that for Pearson. In terms of each rural
stations. Sequential Mann Kendall is a non-parametric test station, the intensities are higher for Albion Hill and lower for
and can provide time varying statistics of any time series Millgrove compared to that of KST at both Toronto and
indicating the trends, which can be plotted against time Pearson. It is also apparent from the year-to-year variation in
to see the changes in the data. Since the trend of any Fig. 2 that KST and Albion Hill have quite similar patterns
of Tu-r, but Millgrove has a different pattern for both
element represents changes over a period of time, therefore,
examination of the relationship between the trends of the rural Toronto and Pearson. For Pearson using Millgrove as the
rural station, there are negative values of Tu-r during earlier
station and UHI values will take into account any climatic
element inherent to the changing values in temperature. A years of the study period, which means that Millgrove was
positive correlation between the trends of Tu-r and Tr warmer than Pearson until 1988, and after that, due likely to
indicates that the background change in temperature in both the rapid pace of urbanization, Pearson became warmer than
Millgrove which is observed from the positive Tu-r. If the
urban and rural areas is similar, while a negative correlation
between the trends indicates a deviation from the background temperature differences at Toronto and Pearson were caused
climate change. The physical interpretation to this assumption only by the urban effect, then the year-to-year variation
is that over time, the increase in urbanization will cause an pattern would be expected to remain the same, no matter
increase in Tu, which will cause an increase in Tu-r provided which rural station provided the data to compute Tu-r.
the background temperature at the rural station is unchanged. However, this is not what is observed in Fig. 2.
This reflects the impact of urbanization on Tu-r. However, the The magnitude of the trends for the temperature time
negative correlation between Tu-r and Tr only demonstrates series of Tu-r is calculated using the TSA approach, and
the increase in Tr but not in Tu. The result is presented in the results are presented in Tables 3 and 4 as degree Celsius
Table 2. per year. According to the analysis, positive trends are
The temperature trend for KST exhibits the highest positive found in all cases for Toronto except for minimum and
mean Tu-r for Albion Hill and for maximum Tu-r for
correlation with both Toronto and Pearson, compared to the
other rural stations, Millgrove and Albion Hill. Thus, we Millgrove. For Pearson, positive trends are observed in all
cases except for maximum Tu-r when Millgrove is
suggest that KST appears to be the appropriate choice as a
rural station for the urban rural stations pair for both Toronto considered as a rural station. The trends for minimum
Tu-r are statistically significant at 99% level for all three
and Pearson since the results from the correlation analysis
suggest that the surface characteristic of KST is more or less rural stations for Pearson. However, for Toronto, the
minimum Tu-r trends are statistically significant at 99%
unchanging relative to the urban stations. The following
sections analyze the annual and seasonal changes of Tu-r level when Millgrove is taken as a rural station. In case of
using the observed temperature data for Toronto and Pearson Toronto with Albion Hill, negative trend for Tu-r for annual
in order to rationalize the choice of the rural station for the mean temperature is observed. When Millgrove is used, the
estimation of Tu-r. trend for annual mean Tu-r is the highest compared to the
other stations. With KST, the increasing annual trend for Tu-r
3.2 Influence of the rural stations on UHI intensity is consistent with the gradual increase in the magnitude of
Tu-r with the ongoing pace of urbanization over the years
3.2.1 Analysis of annual urban rural temperature difference observed in Fig. 2, particularly for Pearson. Although the
choice of KST as the rural station shows urban warming at
The annual urban rural temperature differences calculated both Toronto and Pearson, more analysis is required to make
using the three different rural stations are plotted for a clear decision on the choice of the rural station. However, it
Toronto and Pearson. The plots are shown in Fig. 2. The is evident at this point of the analysis that the choice of a
UHI intensity at both urban stations shows similar year-to- rural station has a significant impact on the quantification of
year variability for all three rural stations with some UHI intensity at both Toronto and Pearson.
Table 2 Correlation between
the trends of Tu-r and Tr Stations Toronto Significant level Pearson Significant level 0.39
Albion Hill 99 0.54 99
0.53 0.81
Millgrove 99 95
KST 0.04 99 0.84 99
Characterization and estimation of urban heat island at Toronto 111
Table 4 Estimated trends of annual UHI intensity for Pearson using
Mann Kendall test for the period of 1970 2000 (degree Celsius per year)
Tu-r_max Tu-r_min Tu-r_mean
Stations
0.025a 0.09a 0.027a
Pear-Alb
0.015 0.03a 0.035a
Pear-Mill
0.012a 0.05a 0.03a
Pear-King
a
Significant at 99%
was suggested that at Toronto downtown the UHI has
reached a saturation point in terms of urban warming, and
thus, the trends for annual mean Tu-r do not show any
significant changes after a certain time.
The difference in the trends between Toronto and
Pearson could be attributed from the relationships among
UHI intensity, near-surface phenomenon, and the localized
topographic effect on the temperature such as the lake
breeze arising from Lake Ontario. The formation of UHI is
also largely influenced by meteorological conditions (e.g.,
mixing heights, inversions) and the physical characteristics
of a surface (e.g., thermal capacity).
Toronto experiences a direct lake effect particularly
Fig. 2 Year-to-year variation of UHI intensity at Toronto and Pearson noticeable in the summer during the day (Gough and
with three rural stations for the period of 1970 2000
Rozanov 2001), whereas Pearson experiences no direct
lake effect. Therefore, due to the presence of the lake, the
heat island effect will have a mitigating influence on
An important observation that can be made from the
temperature trends for Toronto that is absent from Pearson.
above analysis is that the urban heat islands that exist
It is thus suggested that Toronto will experience a strong
at Toronto and at Pearson are different in magnitude
heat island effect on minimum temperature that is in
(Fig. 2). While it is apparent from Fig. 2 that the
winter, but a mitigating effect from the lake on the
magnitude of Tu-r is higher for Toronto than that for
maximum temperature, particularly in summer as
Pearson, however, the trend analysis suggests that the
explained above and also observed by Gough and
trends are higher for Pearson compared to that of Toronto
Rozanov (2001). The question that arises at this point is:
in all cases of Tu-r. The trends for UHI intensities are 4%
is the UHI intensity with or without the lake effect the
and 16% higher for the minimum and maximum Tu-r, same with three different rural stations? Essentially, the
respectively, at Pearson compared to that at Toronto when
UHI intensity, computed using the three rural stations, is
using KST as the rural station. In addition, the annual
considerably different at both Toronto and Pearson, which
mean trends of Tu-r are statistically significant for
can be related to the distinctive physical and surface
Pearson but not for Toronto in case of all the rural characteristics of individual rural station, an issue which is
stations. This observation could be attributed to the explored further in the following sections.
proposal put forth in Mohsin and Gough (2010), where it
3.2.2 Analysis of seasonal urban rural temperature
difference
Table 3 Estimated trends of annual UHI intensity for Toronto using
Mann Kendall test for the period of 1970 2000 (degree Celsius per year)
The variation of monthly mean temperature differences
Tu-r_max Tu-r_min Tu-r_mean
Stations
between the two urban stations and the three rural
stations, Albion Hill, Millgrove, and KST, are shown in
0.015 0.005
Tor-Alb 0
Fig. 3.
0.03b 0.04a
Tor-Mill 0.002
The urban heat island intensity at Toronto seems to be
Tor-King 0.002 0.002 0.001
higher during December, January, and February, however,
a
with differing strengths for the three rural stations. The heat
Significant at 99%
b
island intensity at Pearson shows a similar pattern for
Significant at 95%
112 T. Mohsin, W.A. Gough
Fig. 3 Monthly mean changes of urban rural temperature difference
at Toronto and Pearson with three different rural stations for the period
of 1970 2000 Fig. 4 Seasonal changes in Tu-r at Toronto and at Pearson for three
different rural stations for the period of 1970 2000
Albion Hill and Millgrove but a different one for KST. To
study the features of the seasonal variation of urban rural
temperature difference, winter is classified as December, a rural station shows the lowest magnitude of UHI intensities
January, and February; spring is classified as March, April, (Fig. 4). In general, it is known that urban heat island is
and May; summer is classified as June, July, and August; more pronounced on the minimum temperature than on the
and, finally, autumn is classified as September, October, and maximum temperature, and this was observed to be the case
November. The seasonal changes in urban rural tempera- for Toronto as well (Mohsin and Gough 2010). In the current
ture difference computed using the three rural stations for study, the former is observed for Albion Hill and KST and
the period of 1970 2000 are shown in Fig. 4 for Toronto the latter is observed for Millgrove. Therefore, the UHI
and Pearson. intensity can be characterized as winter dominating at
In order to analyze the seasonal variation of Tu-r, two Toronto using both Albion Hill and KST as a rural station,
types of patterns are considered to characterize the changes while with Millgrove the UHI intensity shows a consistent
in UHI intensity: UHI intensity that is dominating in the pattern in both winter and summer. For Pearson, UHI
cooler part of the year or winter dominating and UHI intensity is clearly summer dominating when Millgrove is
intensity that is dominating in the warmer part of the year selected as a rural station and winter dominating when KST
is used to compute Tu-r. When Albion Hill is used as a rural
or summer dominating. The factors that are considered to
identify what type of UHI intensity exists at Toronto and station for Pearson, the UHI intensity is uniformly observed
Pearson are as follows: (a) the location of the rural station in both winter and summer time. Inland areas are generally
in relation to the urban station, (b) the influence of the hotter in summer and colder in winter than the areas near a
physical geography and environment surrounding the rural water body. Therefore, the observed UHI at Toronto with
station on the seasonal temperature time series (Tr), and (c) Millgrove (both are influenced by the lake) has a physical
the distance of both the urban and rural stations from a basis to be consistent in winter and summer. However, this is
water body. not the case for Pearson for which the UHI intensity shows
In terms of the observed seasonal temperatures, Millgrove summer dominating characteristic with Millgrove as the rural
experienced the highest average seasonal temperatures station. To summarize, at both Toronto and Pearson, the UHI
(Tr_seas) among the three rural stations during 1970 2000. intensities can be characterized as winter dominating
Consequently, the seasonal Tu-r estimated with Millgrove as irrespective of the choice of different rural stations to
Characterization and estimation of urban heat island at Toronto 113
compute the Tu-r. An exception is observed when Millgrove urban development (see site description at Section 2.1).
Toronto is categorized as modern core under the city
is used as a rural station for Pearson, which is probably due
series, and thus, the density of development is higher at
to different geographic settings of the two stations. A recent
study of Tokyo s UHI suggested that it is necessary to select a Toronto than that at Pearson which is categorized as
compact housing and blocks under the city series. Thus,
rural station that has the same distance from a water body as the
urban station to calculate the heat island magnitude accurately the less winter forcing of the Pearson heat island is related
(Sakakibara and Owa 2005). In this perspective, for Toronto, to its surface disturbances and possibly a result of the much
Millgrove may be a better choice to compute Tu-r since both reduced anthropogenic heat flux.
To summarize, at both Toronto and Pearson, the frequency
experiences the lake effect (Fig. 1).
of UHI intensities are dominant in winter for the rural stations,
Another important point to notice from Fig. 4 is that the
seasonal pattern of Tu-r is remarkably similar for both Millgrove and KST, with a much higher magnitude compared
to that for Albion Hill, which shows higher frequencies in
Toronto and Pearson when KST is used as the rural station.
summer. Considering the surface characteristics of each
The surface characteristic of the rural stations has a role to
station, we can only speculate on the reasons for such
play to this observation. As described in Section 2.1, both
observation. One of the major controlling factors to the
Albion Hill and Millgrove were categorized under the
urban rural differences in seasonal temperature is the urban
natural series of the Thermal Zone Classification system,
whereas KST was categorized under the agriculture series, rural differences in surface wetness, which is suggested to play
a role in the variability of UHI intensities in terms of heat gain
depending on the land uses at each of these rural stations.
or loss (Runnalls and Oke 2000). In cities, in summer when
Although the micro-scale climates of these two series may
roads are wet during or after rain, the extra moisture diverts
overlap, the surface properties differ fundamentally on the
more of the net radiation into latent heat and thus reduces the
principles of landscape disturbances (Stewart and Oke 2009).
The observed seasonal changes of Tu-r with KST from sensible heat storage. Toronto experiences frequent convec-
tive precipitations events in summer due to the presence of
Fig. 4 reveal that the surface characteristics of this rural
the lake nearby, which also has moderating effect on
station are more or less constant relative to the urban stations
temperature change. Therefore, in support of the above
over the study period. Consequently, irrespective of the
physical mechanism, Toronto is expected to have weaker
choice of the urban station (Toronto or Pearson), the patterns
of seasonal Tu-r are comparable for both urban stations UHI in summer. The result from the frequency analysis
shows a winter dominating UHI at Toronto, which is
when KST is used as a rural station. This is not the case for
observed for rural stations Millgrove and KST but not for
the other two rural stations, Millgrove and Albion Hill, for
which different patterns of seasonal Tu-r are observed for Albion Hill.
the two urban stations. From this perspective (and in
3.3 Criteria suggested from this study to choose a rural
combination with the correlation analysis), the choice of
KST as a rural station to estimate Tu-r is a preferred choice station
for Toronto and Pearson.
The observations from the seasonal analysis of Tu-r are From the correlation analysis (Section 3.1), KST was suggested
to be a preferred choice for Toronto as a rural station to
clarified further from the frequency analysis of the seasonal
compute Tu-r. However, it is critical to understand whether
time series for Tu-r. The seasonal UHI intensities are
Tu-r computed with Albion Hill or Millgrove or that
computed by taking the difference of the mean temperature
for each season between an urban and a rural station. For computed with KST is also consistent with the climatic and
Toronto using Albion Hill as the rural station, a higher geographical features of Toronto and Pearson. Previous
frequency of UHI intensities of 2 3 C is observed in studies on the analysis of Toronto s heat island suggested that
summer and autumn than in winter and spring. However, Toronto experience dominant UHI effect in winter compared
the magnitudes are found to be highest in winter (>3 C) to summer (Munn et al. 1969; Gough and Rozanov 2001;
compared to other seasons for Albion Hill. A similar Mohsin and Gough 2010). The results from the current study
observation is also evident for Pearson. With Millgrove, the also exhibit winter dominating UHI effect depending on the
maximum frequencies of UHI of 2 3 C are found in winter choice of the rural station that is used to compute Tu-r. A
trend analysis was done for the time series of seasonal Tu-r. A
compared to the other seasons at both Toronto and Pearson.
The frequency of UHI intensities with KST as the rural detailed look at the results from the trend analysis of seasonal
Tu-r (Tables 5 and 6) suggest that statistically significant
station is dominant in the cooler part of the year than in the
warmer part with a higher magnitude of 3 C to 4 C, at increasing trend of UHI intensity exists at Toronto mainly for
Toronto compared to that of 1 C to 2 C at Pearson. A winter with Millgrove and KST, while for Albion Hill no
trend for seasonal Tu-r is detected. The choice of Millgrove
possible explanation for larger UHI effect in winter at
as a rural station, which shows winter dominating Tu-r for
Toronto than at Pearson is the differences in the intensity of
114 T. Mohsin, W.A. Gough
Table 5 Estimated trends for seasonal Tu-r for Toronto using Mann Kendall test (degree Celsius per year)
Tu-r_winter Tu-r_spring Tu-r_summer Tu-r_autumn
Stations
0.009 0.008
Tor-Alb 0.03 0.0009
0.003
Contact this candidate