Spatio temporal dynamics of the urban heat island in singapore 3

All min-of the minimum values min-of UHIraw in urban stations occur during daytime, which is likely due to shading effects from buildings.. Table 4.7: Maximum UHI intensities and their t

While all stations posted a minimum UHIraw that is less than 0◦C, for imum UHImax, 12 stations had positive values The fact that less stations havenegative minimums is expected as UHImax does not consider daytime values All

min-of the minimum values min-of UHIraw in urban stations occur during daytime, which

is likely due to shading effects from buildings Three stations posted minimumUHImax values greater than 1.5◦C (S44, S45 and S46), although little can be estab-lished from this as all three stations had less than eight valid nights of data andthus may not be representative

Table 4.6: Mean, minimum and maximum values of UHImax and UHIraw

Summary across all stations UHImax UHIraw

Mean of station minimums -0.40 -2.98Mean of station means 2.71 1.81Mean of station maximums 4.83 5.09

Overall minimum -2.44 -4.44Overall maximum 6.46 6.70

UHIM AX intensity and time of occurrence

The maximum UHIraw value measured across all stations and time intervals

is 6.70◦C, occurring on 30th July 2009 at 03:10 local time at S22 (Table 4.7) Themaximum of UHImax was registered at 6.46◦C at 22:20 local time on 24th April

2009, at S22 The latter value is deemed to be the UHIM AX due to its stricterfiltering requirements Although non-ideal conditions usually dampen rather thanincrease UHI intensities, unequal antecedent conditions (Φa) can result in higherthan expected intensities (as discussed earlier)

Table 4.7: Maximum UHI intensities and their time of occurrence for all tions and their available dataset across the entire study period (Feb 2008 to Jun2011), Refer to Appendix B for the UCZ classes

Date and time

of occurrence

LCZ S01 Coastal 4.75 2008-05-05 22:50 4.55 2009-07-31 06:00 B G

S02 Industrial 5.22 2009-04-25 00:10 - - 3 B

S03 Forest 2.26 2008-07-31 05:10 2.10 2008-05-17 23:40 A S04 Park 3.50 2011-05-10 01:10 - - B S05 Low-rise residential 5.42 2008-09-25 20:20 5.03 2008-05-19 03:30 3 S06 High-rise residential 4.76 2008-05-19 06:30 - - 4 D

S07 High-rise commercial 6.44 2009-07-31 07:30 6.36 2009-07-31 06:20 1 S08 High-rise residential 5.93 2009-04-24 22:20 - - 1 S09 Coastal 5.36 2008-05-19 07:30 5.28 2009-04-24 22:20 G B

S10 Forested/agricultural 4.04 2009-08-03 08:00 3.76 2008-05-18 04:10 A 9

S11 Clearing 5.00 2008-05-19 07:20 4.95 2008-05-19 06:50 B F

S12 Industrial 5.69 2009-07-30 01:50 5.55 2009-07-31 05:40 2 S13 Low-rise residential 6.19 2009-04-24 21:50 - - 3 S14 High-rise residential 6.10 2009-07-31 07:30 6.06 2009-04-24 22:20 1 2

S15 Low-rise residential 5.95 2009-07-31 05:20 - - 3 S17 High-rise residential 6.30 2009-04-24 22:40 - - 1 E

S18 Low-rise commercial 5.35 2008-04-01 21:10 5.30 2008-10-11 02:00 3 S19 Low-rise residential 5.72 2009-04-24 23:00 - - 3 S20 Open 4.86 2008-07-03 07:50 4.74 2008-05-19 06:40 5 D

S21 Low-rise residential 5.17 2009-04-24 23:10 - - 6 S22 High-rise commercial 6.70 2009-07-30 03:10 6.59 2009-07-31 06:50 1 2

S23 Rural 2.88 2008-04-03 08:50 2.56 2008-03-28 01:50 B A

S24 Low-rise residential 6.20 2009-07-31 08:00 6.14 2009-07-31 06:30 5 S25 Low-rise industrial 6.32 2009-07-31 07:30 6.30 2009-07-31 05:20 8 S27 Rural 0.88 2008-05-19 22:00 - - B A

S36 Open grass 5.30 2009-08-03 07:50 5.16 2009-07-31 04:30 D 8

S37 High-rise residential 5.62 2009-07-31 05:00 - - 1 D

S38 High-rise residential 6.03 2009-07-31 03:40 - - 1 S39 Open grass patch 2.86 2009-01-19 08:50 1.94 2009-02-19 06:50 D S40 Low-rise residential 6.43 2009-07-31 07:30 5.80 2009-02-19 06:50 2 S41 Med-rise commercial 6.01 2009-07-31 05:40 - - 3 S42 Low-rise residential 6.20 2009-07-31 07:00 6.18 2009-07-31 06:50 2 1

Table 4.7 includes both “rural” and “urban” sites in an effort to captureheat island differences across the entire spectrum of land use A few points to noteinclude (i) UHImax filtering see less unexpected time of maximum UHI values thanUHIraw filtering; (ii) stations found in rural and vegetated areas tend to have max-imum UHIraw occurring close to sunrise; and (iii) most of the maximum values forboth UHIraw and UHImax are found during the April-May (Pre-SW monsoon) andJuly-Sept (SW monsoon) period

Oke (1981) states that UHIM AX typically occurs 3 to 5 hours after sunset,which in Singapore’s case, would be 22:00 to 00:00 hrs This is consistent with themaximum of UHImax, which occurred at 22:20 hrs local time The peak occurrence

of UHI in Singapore, however, varies between stations and time of year It is notedthat some stations and months see UHI values that peak around sunrise However,care has to be taken when interpreting values close to sunrise and sunset as rapidchanges to meteorological conditions mean that artefacts may arise (Oke, 2006).Detailed discussion of temporal variations in time of maximum UHI occurrence iscovered in Section 4.3.2

The UHIM AX value in this study (6.46◦C) is approximately half a degreeless than that reported by Chow and Roth (2006), which was 7.07◦C Apart fromthe slight difference in magnitude, the time of occurrence (present study: 22:20hrs, Chow and Roth: 22:00 hrs), time of year (24th April, 17th May) and location

of occurrence (both in Orchard Road) are almost identical Unsurprisingly, paststudies in Singapore have all identified the location of UHIM AX to be within theCBD (Nieuwolt, 1966; Singapore Meteorological Services, 1986; Goh and Chang,1998; Wong and Chen, 2005; Chow and Roth, 2006; Priyadarsini et al., 2008) Ofinterest is the general increase in reported UHIM AX intensities with time, from 3

to 5◦C in the earlier studies to approximately 7◦C in more recent studies, a trendechoed in the neighbouring country of Malaysia (Elsayed, 2011; Roth and Chow,2012)

en-vironment

There are several temporal scales at which climatological and meteorological ability operate The focus of this section on distinctive temporal patterns andtherefore the diurnal, seasonal and inter-annual scales will be of greatest interest;they represent cyclical patterns as opposed to, for example, a weekly scale UHIintensities referred to in this section are UHIraw values unless otherwise stated

vari-4.3.1 Diurnal variability of air temperature

An hourly box-and-whiskers plot of ensemble mean air temperature (across allstations and the entire study period) illustrates that the range of measured airtemperature across all the stations changes throughout the course of a day Theinterquartile range (75th

percentile - 25th

percentile) begins to increase as the sunsets and is largest in the night (19:00 - 06:00 hrs), exceeding 1◦C during each hour(Figure 4.7) The same applies for the full range (max - min), even after takingoutliers into consideration During daytime (07:00 - 18:00 hrs), the interquartilerange decreases as the sun rises and most hours have a sub-1◦C range, especiallyaround solar noon (13:00 hrs)

This finding is in agreement with earlier discussions on the effects of urbaninfluence on the diurnal air temperature cycle Temporal variations are most dis-

Figure 4.8: Ensemble mean hourly air temperatures for selected stations acrossall weather conditions and for the entire study period (Feb 2008 to Jul 2011).The selected stations are representative of sites of common land use types, such

as forest (S03), rural (S16), commercial (S22 & S07), low-rise residential (S15),high-rise residential (S08) and urban park (S29)

residential areas, which may be shaded One striking behaviour is that the ruralstation has air temperatures not unlike the urban sites during solar noon (13:00 hrs)

Relationship between air temperature, cooling rate and UHI intensity

The hourly cooling rate can be mathematically defined as the rate of change

of air temperature between each hourly interval (Equation 4.1)

rates) occur after sunrise The highest warming rates are observed before solarnoon for all three stations, at 09:00 hrs at the rural and low-rise residential stationsand 11:00 hrs at the commercial station This is possibly due to greater solar accessfrom a nearly overhead sun and hence warming potential is greatest closer to noon

in the high-rise commercial area

As discussed previously, a distinct difference between urban and non-urbansites is the magnitude of their rates of cooling This can be seen in the much higherrates of cooling for the rural site between 14:00 to 23:00 hrs compared to the twoother urban sites (Figure 4.9) The openness of the rural site with low re-absorption

of L↑ and lower daytime heat storage flux, unlike in urban areas, contribute to thehigher cooling rates Furthermore, reduced rural µ during drier periods contributes

to increased cooling potential The cooling rate differential gives rise to the lishment of the UHI When the cooling rate of the rural station exceeds that of theurban stations (approximately 14:00 - 23:00 hrs), the effective UHI intensity (notethat ∆Tu−r used here is unfiltered for weather effects as discussed in Section 2.1)increases, in this case peaking at around 01:00 and 02:00 hrs

estab-Conversely, the UHI intensity will begin falling after sunrise when warmingrate becomes larger in rural areas than urban areas In the case of the commercialstation (S07), UHI intensities are negative from 12:00 to 14:00 hrs, creating anurban cool island This is attributed to shading from direct K ↓ by tall buildingsduring daytime The low-rise residential station (S15) also experiences lower UHIintensities during the day but the intensities remain positive due to a lack of shading

4.3.2 Seasonal change in UHI characteristics

The focus of this section is to explore the seasonal variation of UHIraw while specificrelationships between synoptic weather elements and heat island intensity will becovered in a later section (Section 4.7) As discussed in Section 3.1, the weatherconditions in Singapore follow a monsoonal pattern (refer to Tables 3.1 and Figure3.2) Figure 4.10 shows the monthly mean values for the entire study period Thereappears to be a usual trend, which sees the pre-south-west (PSW) intermonsoonand south-west (SW) monsoon periods having the highest heat island intensities.The pre-north-east (PNE) usually experiences a drop in UHI intensity before thelowest values occur during the north-east monsoon (NE)

26.5 27.0 27.5 28.0 28.5 29.0

Figure 4.10: Boxplot of mean monthly nocturnal (19:00 - 07:00 hrs) UHIraw

for all stations together with air temperature measured at Changi MeteorologicalStation (red dashed line) Only stations with more than 24 months of data areused in order to prevent skewed results Note that there is no data for September

2009, October 2009 and June 2011

The warmer months have higher nocturnal mean UHIraw intensities while thewet monsoon (NE) sees lower intensities High precipitation rates during the coolermonths lead to an increase in the rural thermal admittance (µ) This is despite thefiltering out of intervals with rainfall as surface moisture levels may still be high

due to build up of soil moisture during the rainy season As discussed in Section2.1, this contributes to the “dampening” of UHI intensities (Runnalls and Oke,2000; Chow and Roth, 2006) Warmer months also tend to have less cloud coverand lower wind speeds, which may contribute to higher UHIraw intensities Spe-cific relationships between the weather variables and UHI are covered in Section 4.7

It is clear that the diurnal characteristics of the heat island undergo changesacross the year, in terms of magnitude, agreement across stations (i.e SD), onset,peak and decline (Figures 4.11, 4.12 and 4.13) The plots include measurementsfrom all weather conditions To ensure consistency, stations with less than 80%observed month-hours (observations at a given hour of a given month in any year,i.e 12 months×24 hours= 288 month-hours) during the study period are removed(Table 4.8)

Table 4.8: Omitted stations and percentages of month-hour observed

Stn S06 S09 S18 S20 S27 S33 S39 S42 S44 S45 S46

% month-hour 67.0 33.7 72.2 75.0 8.7 69.1 25.4 60.4 32.6 32.3 32.3

In terms of magnitude, the months in the middle of the year, from April toOctober, typically have a higher number of stations having mean UHIraw intensi-ties that exceed 4◦C during the night (Figure 4.11) In contrast, months during thenorth-east monsoon (December to March) see lower heat island intensities (< 4◦C)

January and February see the UHI intensities increase at a somewhat slowrate starting at just before sunset, typically reaching the peak just before sunrise(Table 4.9 and Figure 4.13) This creates diurnal UHI curves that are skewed to-wards sunrise, which is unlike usual descriptions of UHI peaking a few hours aftersunset These two months are characterised by relatively strong winds throughout

The thick cloud cover and relatively high levels of surface moisture and µtowards the end of the rainy season are also important to the shape of the UHIcurve during these months The difference in cooling rate between urban and ruralareas determines the slope of the curve at the start of the night Months withhigh precipitation rate and wind speeds may see urban areas cooling at a fasterrate than usual after sunset, results in slow UHI development (i.e gentler slopesaround sunset during December, January and February) As the sun rises, the UHIintensities drop rapidly in a few hours before stabilizing at noon

Table 4.9: Time of maximum UHIraw hourly ensemble for each month of theyear

A transition occurs around March and the shape of the diurnal curves fromApril to June are rather symmetrical with peaks for most stations occurring aroundmidnight, which is five hours after sunset Synoptic air temperatures are higher,wind speeds and precipitation rates are generally lower during this period Thelower rural µ due to reduced soil moisture creates larger relative difference in cool-ing potential between rural and urban areas, creating a more “ideal” heat islandformation conditions During July and August, the typical peak heat island oc-currence moves later and takes place closer to sunrise than sunset, although thediurnal curves are still rather symmetrical compared to January and February

In September and October, the peak intensities of UHI occur earlier in thenight, prior to midnight This may be the opposite of the above-mentioned effect.Lower wind speeds and relatively drier conditions reduces the cooling rate of urbanareas relative to rural areas resulting in quicker development of UHI (i.e steeper

as discussed above) Hotter and/or drier months with lower wind speeds (such asMay, June and October) have a larger interquartile range with somewhat earlierpeaks The ‘dampening’ of UHI intensities in the NE monsoon relative to the

SW monsoon is in agreement with previous studies by Singapore MeteorologicalServices (1986), Goh and Chang (1998) and Chow and Roth (2006) The reduction

of UHI intensities attributed to a decrease in soil moisture content due seasonalrainfall changes has also been also reported in other tropical cities such as Mexicoand Gabarone (Jauregui, 1997 and Jonsson, 2004, respectively) The relatively highvariability in rural µ values is a likely driving force behind UHI variability in thetropics (Roth, 2007)

4.3.3 Inter-annual trending and cycles of UHI intensities

Beyond the seasonal variability of UHI intensities, variations may also occur atthe inter-annual scale (Pigeon and Masson, 2009) Inter-annual trends in UHI can

be identified by first removing the seasonal trends As such decomposition is verysensitive to noise and outliers, this analysis will only be conducted at the stationlevel for stations with good data continuity

Seasonal-trend decomposition procedure based on loess (STL) (Cleveland