Journal of Power and Energy Engineering, 2013, 1, 7-15 Published Online October 2013 (
Copyright © 2013 SciRes. JPEE
Addressing the Challenge of I nterpreting Microclimat i c
Weather Data Collected from Urb an Sites
L. Bourikas1, T. Shen2, P. A. B. James1, D. H. C. Chow2, M. F. Jentsch3, J. Darkwa2, A. S. Bahaj1
1Sustainable Energy Research Group (SERG), University of Southampton, Southampton, UK; 2Centre for Sustainable Energy Tech-
nologies (CSET), University of Nottingham, Ningbo, China; 3Urban Energy Systems, Bauhaus-Universität, Weimar, G ermany.
Received August 2013
This paper presents some installation and data analysis issues from an ongoing urban air temperature and humidity
measurement campaign in Hangzhou and Ningbo, China. The location of the measurement sites, th e positioning of the
sensors and the harsh conditions in an urban environment can result in missing values and observations that are unre-
presentative of the local urban microclimate. Missing data and erroneous values in micro-scale weather time series can
produce bias in the data analysis, false correlations and wrong conclusions when deriving the specific local weather
patterns. A methodology is presented for the identification of values that could be false and for determining whether
these are “noise”. Seven statistical methods were evaluated in their performance for replacing missing and erroneous
values in urban weather time series. The two methods that proposed replacement with the mean values from sensors in
locations with a Sky View Factor similar to that of the target sensor and the sensor s closest to the target’s location per-
formed well for all Day-Night and Cold-Warm days scenarios. However, during night time in warm weather the re-
placement with the mean values for air temperature of the nearest locations outperformed all other methods. The results
give some initial evidence of the distinctive urban microclimate development in time and space under different regional
weather forcings.
Keywords: Urban Microclimate Observations; Installation Challenges; Weather Data Time Series Analysis; Missing
1. Introduction
In the current decade from 2010 to 2020 the urban popu-
lation of China is expected to exceed the rural one for the
first time in history [1]. In the Zhejiang Province, cities
such as Hangzhou and Ningbo are developing into finan-
cial centres attracting still more people from the countr y-
side. Over the next decade the borders of these urban
agglomerations are expected to expand further. The new
population will put pressures on the existing transporta-
tion, water supply, sewage and energy infrastructures.
The expansion of city borders and the changes in land
use as well as building density usually have a prominent,
and often ominous, effect on the air temperature devel-
opment within the urban canopy layer (the part of the
atmosphere from ground level up to average building
height) [ 2] .
An informed mitigation strategy at th e initial stages of
urban planning is, therefore, of paramount importance.
Common practice in making such informed urban design
decisions is to rely on simulation results. Simulation
models are typically dr iven by weather data for the place
of study. Commonly available weather data files have a
typical meteorological year format that is based on his-
torical hourly data which were usually collected at air-
ports. However, these historical datasets largely underes-
timate the effects of the urban microclimate and do not
represent of locations within the city where the local sp e-
cific microclimates develop [3].
Despite the “urbanisation” of regional weather predic-
tion models [4,5] and the development of micro-climatic
models that solve the urban energy balance [6, 7], there is
still a need for simple urban weather forecasting models
[8,9] that would produce urban and building simulation
ready weather data-series adapted to the local micro-
climate within an acceptable accuracy.
This paper reports on an on-going air temperature and
humidity measurement campaign in Hangzhou and
Ningbo, China and details key issues regarding the
weather data time series analysis as a baseline for pro-
ducing a methodology for the micro-climatic adaptation
of commonly available weather data files.
Addressing the Challenge of Interpreting Microclimatic Weather Data Collected from Urban Sites
Copyright © 2013 SciRes. JPEE
1.1. Local Climate
The measurement network consists of 55 air temperature
and relative humidity (RH) sensors installed at a height
of 3 to 5 meters above ground level in the cities of
Hangzhou (30˚15'N 120˚10'E) and Ningbo (29˚52'N
121˚33'E) in Zhejiang Province, China (Figure 1).
Both Hangzhou and Ningbo have a humid subtropical
climate with a Cfa classification in the Köppen-Geiger
climate system [10]. The Cfa classification denotes a warm
temperate (C), fully humid (f) climate with hot summer
(a) [10].
These two cities combined had a total urban popula-
tion of 6.65 million people in 2011 with a mere 7.85 mil-
lion documented in the broader Municipality areas [12,
1.2. Urban Context
Recent urbanization trends in combination with a climate
change induced warming effect led to a noticeable tem-
perature rise in developed Chinese cities [14]. According
to China’s national meteorological statistics for the pe-
riod 1981 to 2010, Hangzhou experienced on an average
27.2 days per year with average temperatures higher than
35˚C. Since 2003, this figure has increased to more than
35 days annually, making Hangzhou the third hottest city
in China [15,16].
From an urban planning perspective, Hangzhou’s ur-
ban grid spans across 3070 km2 [12] and contains two
large water bodies, namely the West Lake close to the
city centre and Xixi Wetland. Ningbo occupies an area of
2460 km2 [13] and is characterised by three main rivers,
the Fenghua, Yao and Yong Rivers, which meet in the
city centre. It is interesting to examine whether these
natural ecological resources could be utilised to alleviate
the current urban heat island effects and if so, to what
2. Measurement Equipment and Set-up
At each measurement location (Figure 2) air temperature
Figure 1. The geographic location of Hangzhou and Ningbo
in China. World Image Source: [11].
and relative humidity iButto n sen sor s [17] with miniature
data loggers were installed on lamp posts (Figure 3) . The
data loggers record hourly values for air temperature at a
11-Bit (0.0625˚C) resolution and relative humidity (RH)
at a 12-Bit (0.04%) resolution [17]. The memory capaci-
ty of 8kb allows to store up to 110 days of hourly data.
At the end of each 110 day period the data are down-
loaded to a PC and the sensors repositioned at the same
location to continue recording.
All sensors were calibrated prior to installation and
their readings were inter-compared in an environmental
test chamber at the Centre for Sustainable Energy Tech-
nologies (CSET) at the University of Nottingham Ning-
The consultation report on meteorological observations
at urban sites [18] that complements the World Meteoro-
logical Organisation’s (WMO) Guide to Meteorological
Instruments and Methods of Observation [19] standard
provides extensive guidelines for the location selection
and on-site installation of weather stations. The network
installation and maintenance of the sensor network dis-
cussed here has proved so far that applying all the WMO
recommendations is quite a challenging task in practice.
In situ installations had to strike a balance between
being representative of the urban canopy layer (UCL)
and at the same time ensuring accessibility to the site and
easy maintenance.
The sensors are expected to be representative of the
temperature and RH of an area ranging from 100 to sev-
eral hundred meters in a direction upwind and around
each sensor [18]. The locations were carefully selected to
have homogeneous characteristics and the sensors were
installed at sites with a reasonable distance to the fringe
of different surface types.
With regard to radiation shielding, the shield design
and the iButton sensors were tested to assure that the
logged air temperature data are reliable and similar with-
in an accepted accuracy to the data collected with the
commonly used Stevenson screen design [23].
2.1. Experiences from th e Installation and First
Measurement P eriod
The first problem that was encountered was that initially
the local government did not consent to install objects on
lamp posts on the grounds that solar radiation shields
would impact on street aesthetics and pose a risk to pe-
destrians. An experimental setting was prepar ed to prove
the firmness and safety of the installation. In the end,
owing to the safety test report and the government’s
support to the research project, the Urban Administration
Bureau of Hangzhou and Ningbo approved the applica-
tion for the network’s installation.
Despite the permissions and public information notes
provided with the sensors, six solar radiation shields and
Addressing the Challenge of Interpreting Microclimatic Weather Data Collected from Urban Sites
Copyright © 2013 SciRes. JPEE
Letter Index: C = commercial, R = residential, PG = public green, E = educational, I = indus-
trial, ER = Ecological Reserve, F = Farmland, SU-A = Special use-Airport.
Figure 2. Sensor network maps in Hangzhou (top) and Ningbo (bottom). The boxes give an eye-estimation of the albedo and
the bullet points describe the Sky View Factor (SVF) as calculated on site. The letters nearby the sensors indicate the Land
Use type as given in the Local Planning maps [20,21]. The dot circles show the distance from the city centre in kilometres.
Background Image Source: [22].
Figure 3. Sensors were installed on lampposts at a height of
3 to 5 meters due to security concerns and local authority
restrictions (left). The radiation shield (centre) and the
iButton sensor with the miniature data logger that are posi-
tioned inside the radiation shield (right). Central Image
Source: [23].
the sensors within have been removed since September
2012 (fou r in Ningbo and two in Hang z hou).
There were also some cases of private sites such as
factory grounds and residential compounds where per-
mission to install sensors was deni e d by the owne rs .
The difficulty to convince the authorities and private
owners to allow the use of the lamp posts in conjunction
with the need for sites with easy access which at the
same time, have to be free of obstructions make the cur-
rent selection of sites the best possible choice. However,
there are some cases where the sensors had to be posi-
tioned very close to airflow obstructions or in open park
locations that may affect the representativeness of their
readings for the local urban micro-climate conditio ns.
False measurements or missing data due to malfunc-
tion of the sensors will distort the signal of site specific
weather development, cause bias during the development
of urban weather adaptation models and produce invalid
correlations. Consequently, any false or missing values in
Addressing the Challenge of Interpreting Microclimatic Weather Data Collected from Urban Sites
Copyright © 2013 SciRes. JPEE
the time series must be identified and effectively replaced.
2.2. Data Analysis and Interpretation in the
Urban Context
The weather data time series collected from the sensors’
network were, in a first step, examined to identify any
noise in the form of outliers in the datasets. It was found
that RH values were often above 100% regardless of the
sensor and its location. A simple algorithm was applied
to cap the RH values at 100% levels. The scatter and
frequency of the erroneous RH data indicates that the
most probable cause is a sensor’s drift due to pollution,
water spray and general degradation.
In a second stage, boxplots of the hourly air tempera-
ture distribution per week were created with IBM SPSS
Statistics Ver.19 [24] for each sensor’s location in
Hangzhou (see Figure 2). All outliers were traced back
to the data sample. Outliers that showed up in groups of
subsequent hours or were common to all sensors at a
specific time and day were not treated as potential errors.
The remaining outliers were compared against the av-
erage hourly air temperature for the same week and the
coldest and hottest days of this week respectively (Fig-
ure 4). The range of hourly changes on each measure-
ment site (Figure 5) was also assessed against the hourly
changes in weather data collected at Mantou Mountain,
National Principle weather Station (30˚13'N, 120˚26'E).
The interpretation of these tests’ results in view of site
specific attributes (i.e. Sky View Factor (where SVF is
an index (0 to 1) of the unobstructed sky area which can
be seen from a given point), albedo, building height to
street width aspect ratio, impermeable to permeable sur-
face ratio, surrounding materials’ properties) can lead to
conclusions for the weight of each parameter and their
role in the local urban micro-climate’s development in
time and space.
3. Case Study for the Replacement of
Missing/False Data in Urban Observations
Seven cases (Table 1) have been investigated for replac-
ing false data readings and missing values in urban weath-
er measurements.
The first three cases involved the use of statistical pre-
diction models with the existing dataset and application
of these models to the forecast of missing data (Cases 1-3
in Table 1). These models estimate the level, seasonal
change and trend for the data that are put into the models
and predict the time series’ development [24].
The remaining four cases were based on the replace-
Figure 4. Comparison of hourly air temperature to the weekly trend and extreme days in the week.
Addressing the Challenge of Interpreting Microclimatic Weather Data Collected from Urban Sites
Copyright © 2013 SciRes. JPEE
Figure 5. The box shows the range of air temperature change for the 50% of the values. The line inside the boxes is the me-
dian of the air temperature change distribution. The red stars denote the extreme outliers, changes in temperature at least
three times larger than the range of change in the 50% of the values. The whiskers extend 1.5 times the height of the box or if
there are not any values in that range then to the minimum and maximum values [24].
Table 1. Summary of the temperature estimation methods for missing data in the case study.
Cases Estimation Method Dataset from Sensors (see in Figure 6) Scenarios
Case 1 Forecast model 1 1 (00:00 to 05:00)
Case 2 Forecast model 1 to 6 2 (12:00 to 17:00)
Case 3 Forecast model 2,7,8
Case 4 Mean (2 hours before and after missing data) 1
Case 5A Linear Interpolation Average from sensors 1 to 6 Days
Case 5B Mean Average from sensors 1 to 6 13/01/2013
Case 6A Linear Interpolation Average from sensors 2,7,8 09/03/2013
Case 6B Mean Average from sensors 2,7,8
Case 7 Long-time Mean (10 years) Average from Mantou mountain, WMO station
ment of missing values with the mean value of the two
hours before and after the missing data (Case 4 in Table
1), replacement with the mean temperature from loca-
tions with similar to the target site’s SVF (marked with
blue points in Figure 6) (Case 5B in Table 1) and with
the mean temperature from the nearest sites to the target
site’s location (marked as 2, 7, 8 in Figure 6) (Case 6B
in Tab le 1). In Cases 5A and 6A the missing values were
replaced with the results from linear interpolation of the
average temperature in locations with similar to the tar-
get site’s SVF and linear interpolation of the average
temperature of the three nearest sites respectively. In
Case 7 it is proposed to fill gaps in the dataset with 10
year-long average temperature values from data collected
at Mantou Mountain, National Principle Station (marked
as 9 in Figure 6).
The goodness-of-fit of the estimated values to the ob-
served data was examined under two scenarios; the pre-
diction and replacement of 6 subsequent hours during
night-time (00:00 - 05:00) (Scenario 1) and daytime
(12:00 - 17:00) (Scenario 2). The target sensor selected
for this case study (marked as 1 in Figure 6) is located in
a commercial area 8 km North West of Hangzhou’s
commercial city centre.
Addressing the Challenge of Interpreting Microclimatic Weather Data Collected from Urban Sites
Copyright © 2013 SciRes. JPEE
Figure 6. Locations of the sensors used for the case study. The target location for the case study is marked as 1. The blue
points denote locations with similar to the target site’s SVF. The National Principle Station, Mantou Mountain’s location is
marked as 9. Image Source: [22].
There were two days selected in the scenarios; one
cold cloudy day in January (13/01/2013) and one warm
sunny day in March (09/03/2013). On the 13th of January
the sky conditions at Mantou Mountain’s National Prin-
ciple S tatio n were reported as mostly cloudy with light rain
showers while the winds were blow ing from a North West-
North direction with speeds from 1 to 3 m/s. The 9th of
March was the fifth day in a row with a clear sky, the
wind at the Mantou Mountain was blowing from a South
West-South direction with speeds from 1 to 4 m/s [25].
Statistical Forecast Modelling
Three prediction models were created with historical ob-
servations from the target sensor (marked as 1 in Figure
6, Case 1), the average temperature from locations with
the same summer time SVF (blue bullets in Figure 6,
Case 2 in Tab le 1) and the average temperature from the
three nearest sites to the target sensor’s location (marked
as 2, 7 & 8 in Figure 6, Case 3 in Table 1).
Hourly weather observations cannot be considered as
stationary (statistical properties such as the mean and
variance would be time independent [26]) with respect to
the time scale of weather development within the urban
canopy layer. For this reason autoregressive integrated
moving average (ARIMA) models were excluded from
the SPSS built-in forecasting procedure. Only seasonal
exponential smoothing models have been considered and
the most appropriate one was selected with the Expert
Modeler component of IBM SPSS Ve r. 1 9 [ 24].
The data observations currently available are for the
period from 01/01/2013 to 10/03/2013. Control tests were
carried out with 7, 5, 3 and 2 days’ samples for January,
February and March. Ta bles 2 and 3 show the Root Mean
Square Error (RMSE) and the Mean Absolute Error
(MAE) for a single day ’s temperat ure forecast each month.
The results in this preliminary analysis show that for
the dates and months studied in the case study the models
should be built with 7 days’ of data in January and with 5
days’ of data in Mar ch (Tab les 2 and 3). It also needs to
be pointed out that the weather conditions in the last few
days before the missing values have a larger effect to the
model’s output than the monthly weather trend. A stable,
homogeneous weather conditions’ pattern and a strong di-
urnal signal in the prediction dataset (data put into the mo-
dels) seem to increase the accuracy of the first few fore-
casted hours and give a 24-hour periodicity trend to the
It should be noted that the models created for the pur-
poses of this study are simple and that they are only eva-
luated for their capability to fill missing values in an ur-
ban weather dataset. A weather forecasting model would
need to include more variables such as rainfall, wind
speed and solar radiation and closely map their effects on
the site specific weather development.
4. Results
The goodness-of-fit of all methods was evaluated with
the fit of t he predicte d va l ue s to the 6 h ou rs observat ions
Addressing the Challenge of Interpreting Microclimatic Weather Data Collected from Urban Sites
Copyright © 2013 SciRes. JPEE
Table 2. Root Mean Square Error results indicating the goodness-of-fit of the predicted values to observed data.
RMSE Model - 2 days data Model - 3 days data Model - 5 day s data Model - 7 days data
January 2.361 2.014 1.654 1.442
February 1.179 0.619 1.006 0.870
March 1.950 1.464 1.374 1.643
Table 3. Mean Absolute Error serving as an additional indication of the model results’ fit to data.
MAE Model - 2 days data Model - 3 days data Model - 5 days data Model - 7 days data
January 2.10 1.73 1.35 1.08
February 0.98 0.52 0.89 0.80
March 1.70 1.16 1.11 1.26
that were assumed as missing in the scenarios (Table 1).
Figure 7 shows that the replacement of missing urban
temperature measurements with the long-term average
temperature from the nearby National Principle Station
will lead to large deviations from the reality. In the case
of a hot day such as the one studied in March, the error
can be as large as 16˚C (F igure 8). The 10˚C to 16˚C air
temperature difference between the historical dataset
from the Mantou Mountain and the location in the city
strengthens the view that historical datasets collected at
airports and rural locations do not represent of locations
within the city where a distinctive micro-climate develops.
In the Cases 5B and 6B, the missing air temperature
values at location 1 were replaced with direct substitution
with the mean temperature from sensors across the city
that have similar SVF and from sensors within a small
distance (<4 km) to the site of interest.
It is important to point ou t that on a day with clear sky
(Mar_Day marked as green bar) both methods performed
well at the daily temperature peak point. The large daily
variation of air temperature during warm days and the
associated high values at noon could not be satisfactorily
predicted by the statistical models (Cases 1 to 3 in Table
1) and linear interpolation methods (Cases 5A and 6A in
Table 1). Generally, these methods appear to be applica-
ble only during days with small diurnal temperature dif-
Figure 8 shows that all the estimation methods had a
similar performance during the mostly cloudy day that
was studied in January. Even Case 4 that replaces the
missing values with the mean data of the 2 hours before
and after the gap in the dataset returns values close to the
measured data. However, in Case 4 all 6 missing values
are replaced with the same mean value. Therefore, this
method should only be used when a single value is missing
and the change in the values before and after is not large.
5. Conclusions
The air temperature data collected from an urban meas-
urement network in Hangzhou, China were examined to
Figure 7. Goodness-of-fit of the estimated January (Jan)
and March (Mar) temperature data for each method in
relation to the observed temperature.
Figure 8. Mean absolute difference between the tempera-
ture predictions and the observed air temperatures for each
identify missing or erroneous values. At the beginning of
the analysis the measurements from each location that
C ase 1
C ase 2
C ase 3
C ase 4
C ase 5A
C ase 5B
C ase 6A
C ase 6B
C ase 7
R oot Mean Square Error
J an_Night
C ase 1
C ase 2
C ase 3
C ase 4
C ase 5A
C ase 5B
C ase 6A
C ase 6B
C ase 7
Mean Absolute Erro r ( |O bserved - Estimated | )
Addressing the Challenge of Interpreting Microclimatic Weather Data Collected from Urban Sites
Copyright © 2013 SciRes. JPEE
did not fit the weekly temperature trend were isolated.
The weekly distribution of hourly air temperature and the
range of hourly changes revealed any values that could
be false. These were further compared against the weekly
average temperatures and the extreme days in the week.
Seven methods for replacing missing values were eva-
luated in their performance in the context of urban mea-
surement datasets. It was observed that the replacement
of missing and false values with the mean from nearby
sites and with the mean from sites with characteristics
similar to those of the target location returns equally
good results. There were not significant differences in the
performance of these two methods during day-night time
and on cold-warm days. However, the night-time tem-
perature estimates during warm weather when the urban
temperature difference to the rural surroundings is ex-
pected to be larger show that the mean of the nearby sites
fits the measured data the best. This might be due to the
night cooling potential of remote sites, away from the
city centre that were included in the dataset of Case 5
when they had a SVF similar to one of the site of interest.
6. Acknowledgements
L.B. would like to thank th e “Liveable Cities Proj ect” for
funding a visit to Hangzhou and Ningbo in China for
researching on the urban micro-climate and to collabo-
rate with the Centre for Sustainable Energy Technologies
at the University of Nottingha m Ningbo (EPSRC funded :
The installation work of the sensors’ network in
Hangzhou and Ningbo is supported by the Ningbo Natu-
ral Science Foundation (No. 2012A610173) and the Ningbo
Housing and Urban-Rural Development Committee (No.
[1] Population Division of the Department of Economic and
Social Affairs of the United Nations Secretariat, “World
Population Prospects: The 2010 Revision and World Ur-
banization Prospects: The 2011 Revision,” 2012, UN
[2] R. A. Memon, D. Y. C. Leung and L. Chunho, “A Re-
view on the Generation, Determination and Mitigation of
Urban Heat Island,” Journal of Environmental Sciences,
Vol. 20, No. 1, 2008, pp. 120-128.
[3] A. Mylona, “The Use of UKCP09 to Produce Weather
Files for Building Simulation,” Building Services Engi-
neering Research and Technology, Vol. 33, No. 1, 2012,
pp. 51-62.
[4] F. Chen, et al., “The Integrated WRF/Urban Modelling
System: Development, Evaluation, and Applications to
Urban Environmental Problems,” International Journal
of Climatology, Vol. 31, No. 2, 2011, pp. 273-288.
[5] A. Baklanov, et al., “Hierarchy of Urban Canopy Para-
meterisations for Different Scale Models,” In: A. Mahura
and A. Baklanov, Eds, MEGAPOLI Project Scientific Re-
port 10-04, Danish Meteorological Institute, DMI: Co-
penhagen, 2010.
[6] V. Masson, C. S. B. Grimmond and T. R. Oke, “Evalua-
tion of the Town Energy Balance (TEB) Scheme with
Direct Measurements from Dry Districts in Two Cities,”
Journal of Applied Meteorology, Vol. 41, No. 10, 2002,
pp. 1011-1026.
[7] C. S. B. Grimmond and T. R. Oke, “Turbulent Hea t Fluxes
in Urban Areas: Observations and a Local-Scale Urban
Meteorological Parameterization Scheme (LUMPS),” Jour-
nal of Applied Meteorology, Vol. 41, No. 7, 2002, pp.
[8] A. J. Arnfield, “Two Decades of Urban Climate Research:
A Review of Turbulence, Exchanges of Energy and Wa-
ter, and the Urban Heat Island,” International Journal of
Climatology, Vol. 23, No. 1, 2003, pp. 1-26.
[9] J. Bouyer, C. Inard and M. Musy, “Microclimatic Coupl-
ing as a Solution to Improve Building Energy Simulation
in an Urban Context,” Energy and Buildings, Vol. 43, No.
7, 2011, pp. 1549-1559.
[10] F. Rubel and M. Kottek, “Observed and Projected Cli-
mate Shifts 1901-2100 Depicted by World Maps of the
Köppen-Geiger Climate Classification,” Meteorologische
Zeitschrift, Vol. 19, No. 2, 2010, pp. 135-141.
[11] Esri, “ArcGIS 10. GIS Software Suite,” 2012.
[12] Hangzhou Statistical Bureau, “Statistical YearBook of
Hangzhou 2012,” China Statistical Press, Beijing, 2012.
[13] Ningbo Municipal Statistics Bureau, “Ningbo Statistical
YearBook 2012,” China Statistical Press, Beijing, 2012.
[14] Information Office of the State Council of the People’s
Republic of China, “China’s Policies and Actions for
Addressing Climate Change,” 2008.
[15] China Daily, “Summer Heat Leaves Cities Gasping as
Mercury Rises,” 2013.
[16] Best News, “Hangzhou Ranked New ‘Four Fournaces
Third Old Foundation Hangzhou Sign,” 2013.
[17] Maxim Integrated, “iButton Temperature/Humidity Log-
ger with 8 kb Data Logger Memory ”.
[18] T. R. Oke, “Initial Guidance to Obtain Representative
Meteorological Observations at Urban Sites,” Instruments
Addressing the Challenge of Interpreting Microclimatic Weather Data Collected from Urban Sites
Copyright © 2013 SciRes. JPEE
and Observing Methods (WMO/TD-No. 1250):47, 2006.
[19] World Meteorological Organisation, “Guide to Meteoro-
logical Instruments and Methods of Observation. Sixth
Edition,” WMO-No. 8, Geneva, 1996.
[20] Hangzhou Academy of Urban Planning & Design, “Hang-
zhou General Planning Map,” Hangzhou, China, 2006.
[21] Ningbo Urban Planning Bureau, “Ningbo General Plan-
ning Map (2004-2020),” Ningbo, China, 2003.
[22] Google Earth, “Satellite Images and Maps of Hangzhou
and Ningbo, China,” 2013.
[23] H. K. W. Cheung, G. J. Levermore and R. Watkins, “A
Low Cost, Easily Fabricated Radiation Shield for Tem-
perature Measurements to Monitor Dry Bulb Air Tem-
perature in Built Up Urban Areas,” Building Services En-
gineering Research and Technology, Vol. 4, No. 31, 2010,
pp. 371-380.
[24] IBM Corp., “IBM SPSS Statistics for Windows,” Version
19.0. Released 2010, Armonk, NY: IBM Corp.
[25] The Weather Underground, “Hangzhou Weather Data
from from Mantou Mountain’s National Principle WMO-
Listed Weather Station”.
[26] NIST/SEMATECH, “e-Handbook of Statistical Methods,”