Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Tests of an operational algorithmic urban heat island scheme Barton, Mark 2002

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
831-ubc_2002-0333.pdf [ 13.24MB ]
Metadata
JSON: 831-1.0090356.json
JSON-LD: 831-1.0090356-ld.json
RDF/XML (Pretty): 831-1.0090356-rdf.xml
RDF/JSON: 831-1.0090356-rdf.json
Turtle: 831-1.0090356-turtle.txt
N-Triples: 831-1.0090356-rdf-ntriples.txt
Original Record: 831-1.0090356-source.json
Full Text
831-1.0090356-fulltext.txt
Citation
831-1.0090356.ris

Full Text

TESTS OF AN OPERATIONAL ALGORITHMIC URBAN HEAT ISLAND SCHEME by Mark Barton B . A . (Hons.), York University, 1997 A THESIS S U B M I T T E D I N P A R T I A L F U L F I L L M E N T O F T H E R E Q U I R E M E N T S F O R T H E D E G R E E O F M A S T E R OF S C I E N C E in T H E F A C U L T Y OF G R A D U A T E S T U D I E S (Department of Geography) We accept this thesis as conforming to the required standard T H E U N I V E R S I T Y OF B R I T I S H C O L U M B I A M a y 2002 © Mark Barton, 2002 In presenting t h i s thesis i n p a r t i a l f u l f ilment of the requirements for an advanced degree at the University of B r i t i s h Columbia, I agree that the Library s h a l l make i t free l y available for reference and study. I further agree that permissio n for extensive copying of t h i s thesis for scholarly purposes may be granted by the head of my department or by his or her representativ es. It i s understood that copying or publication of t h i s thesis for f i n a n c i a l gain s h a l l not be allowed without my written permission. •J The University of B r i t i s h Columbia Vancouver, Canada Department of ABSTRACT This study presents the tests of an algorithmic scheme to calculate the magnitude of the heat island of the urban canopy layer, on an hourly basis throughout the year. The algorithmic scheme has been developed so that operational climatologists can apply it in all cities. Input requirements for the scheme are readily available meteorological data and simple numerical descriptors of the urban canyon geometry and rural surface conditions. The scheme comprises four components. The first, (AT u . r (m a x)), is an empirical relation that gives the absolute maximum urban heat island, experienced in a city as a function of its canyon height-to-width ratio in the urban core (Oke, 1981). Three 'phi ' terms, ( O m O w <J>t), adjust this maximum downward as a result of the combined effects of rural surface thermal admittance, weather and time of day, respectively. The rural surface thermal admittance effect term, (3>m), is derived from the results of an energy balance model (Oke etal., 1991). Two datasets from Uppsala, Sweden and one from Lodz, Poland are used to test the scheme. The nocturnal performance of the scheme is promising. Average predicted nocturnal heat island magnitudes are typically within plus or minus half a degree of the measured heat island, ( M A E 0.5 °C), in Lodz, Poland (1998). However, the daytime utility of the scheme is poor. The regular occurrence of small, (~1 °C), daytime heat and cool islands is largely missed. Degradation of performance appears to be due, in part, to the inability of the scheme to produce negative heat islands (cool islands). Given that, applied (pollution) climatologists are primarily concerned with heat island effects when they attain magnitudes of several degrees; this is not considered a significant limitation. TABLE OF CONTENTS Page Abstract i i L i s t of Tables vii L i s t of Figures vi i i Acknowledgements x Chapter 1 Introduction 1 1.1 The Urban Heat Island 2 1.1.1 The nature of heat islands 2 1.1.2 Causes of heat islands 6 1.1.3 Urban heat island models 11 1.2 The Value of an Operational Urban Heat Island Scheme 16 1.3 Description of the Algorithmic Scheme (AS) 19 1.3.1 The rural surface thermal admittance factor ( O m ) 22 1.3.2 The weather factor (O w ) 31 1.3.3 The temporal factor (3>t) 32 1.4 Objective of the Research 33 Chapter 2 Methods and Data 35 2.1 The Choice of Modelling Scheme and Testing Approaches 35 2.2 Rationale for the Selection of Data for Testing 35 2.3 Physiographic and Climatic Setting of Uppsala, Sweden 36 2.3.1 Normal climate of Uppsala 40 2.3.2 City morphology of Uppsala, 1948-1949 42 2.3.3 City morphology of Uppsala, 1976-1977 43 iv 2.4 The Uppsala Data 48 2.4.1 The 1948-1949 traverse data 48 2.4.2 The 1976-1977 hourly fixed screen thermograph data 53 2.5 Physiographic and Climatic Setting of Lodz, Poland 56 2.5.1 Morphology of the urban site 59 2.5.2 Morphology of the rural site 62 2.6 The Lodz Data 62 Chapter 3 Testing the AS with the Uppsala Data Sets 65 3.1 Testing the A S with the 1948-1949 traverse data 68 3.1.1 Daytime performance of the A S 68 3.1.2 Nighttime performance of the A S 69 3.1.3 Daytime-nighttime variability of scheme performance 71 3.2 Testing the A S with the 1976-1977 fixed screen data 72 3.2.1 Observed and A S UFA from February 15 t h to 17 t h, 1977 (Figure 3.2) 74 3.2.2 Observed and A S U H I from Apr i l 8 t h to 14 t h, 1976 (Figure 3.3) 75 3.2.3 Observed and A S UFA from June 9 t h to 12 t h , 1976 (Figure 3.4) 76 3.2.4 Observed and A S U H I from September 28 t h to October 2 n d , 1976 (Figure 3.5) 77 , 3.2.5 Results of the test with the 1976-1977 Uppsala data 78 3.3 Summary of Tests 80 Chapter 4 Testing the AS with the Lodz Data Set 81 4.1 Performance of the A S 87 4.1.1 Observed and A S U H I from January 1 s t to 5 t h , 1999 (Figure 4.3) 87 4.1.2 Observed and A S U H I from February 8 t h to 13 t h , 1999 (Figure 4.4) 88 4.1.3 Observed and A S U H I from May 7 t h to 11 t h , 1998 (Figure 4.5) 90 v 4.1.4 Observed and A S UFA from July 5 t n to 9 m and July 11 t h to 14 t h, 1997 (Figures 4.6 and 4.7) 91 4.1.5 Observed and A S U H I from August 16 t h to 20 t h , 1997 (Figure 4.8) 93 4.1.6 Observed and A S U H I from September 6 t h to 10 t h , 1999 (Figure 4.9) 94 4.1.7 Observed and A S U H I from December 2 n d to 6 t h , 1998 (Figure 4.10) 96 4.2 Results of the Tests 97 4.2.1 Results from the 1997 Lodz data 97 4.2.2 Results from the 1998 Lodz data 97 4.2.3 Results from the 1999 Lodz data 98 4.2.4 Month-by-month variability of the A S 98 4.3 Potential Errors in the Observed Heat Islands 99 4.4 Analysis of A S Sensitivity to Error in A S Inputs 100 Chapter 5 Discussion and Conclusion 102 5.1 Discussion 102 5.1.1 Limitations posed by the data sets 102 5.1.2 Universality of the scheme 103 5.2 Conclusion 104 5.3 Suggestions for Future Studies 104 L i s t of Symbols and Abbreviat ions 106 Appendix 107 References 110 vi LIST OF TABLES Table Page 1.0 Commonly hypothesized causes of the urban canopy layer heat island 8 1.1 Table of rural surface thermal admittance (u,-) values 30 2.0 Date and number of data collection traverses per day during the Uppsala May 1948 - May 1949 data collection period 50 2.1 Date and number of data collection hours per day at R A during the Uppsala January 1976 - February 1977 data collection period 54 3.0 Uppsala 1948-1949 Algorithmic Scheme (AS) statistical evaluation 66 3.1 Uppsala 1976-1977 Algorithmic Scheme (AS) statistical evaluation 67 4.0 Lodz 1997 Algorithmic Scheme (AS) statistical evaluation 82 4.1 Lodz 1998 Algorithmic Scheme (AS) statistical evaluation 83 4.2 Lodz 1999 Algorithmic Scheme (AS) statistical evaluation 84 4.3 Nocturnal 'd ' Statistics and observed average nocturnal U H I magnitude 99 4.4 Sensitivity of A S to likely maximum errors in A S inputs 101 A2.0 Uppsala, mean air temperature and precipitation 1948-1949 and 1976-1977, and temperature and precipitation normals 108 A2.1 Lodz climate data for the 1997-1999 data collection period and average temperature and average precipitation 109 vii LIST OF FIGURES Figure Page 1.0 Schematic representation of the urban atmosphere illustrating a two-layer classification, urban canopy layer (UCL)/urban boundary layer ( U B L ) of thermal modification 3 1.1 Hypothetical representation of the U C L urban heat island 5 1.2 Typical temporal variation of urban and rural: (a) air temperature, (b) cooling/warming rates, and (c) the resulting heat island intensity (AT u _ r ) under 'ideal' weather conditions 7 1.3 Relation between maximum heat island intensity and height to width ratio (H/W) of the street canyons in the centers of 31 cities for equation 1.2 20 1.4 Relation between maximum heat island intensity and sky view factor of the street canyons in the centers of cities 24 1.5 Relation between the thermal admittance reduction factor (<E>m ) and rural surface thermal admittance based on results from the SHTM numerical model 25 1.6 Diurnal variation of O t 34 2.0 Location map of Uppsala 37 2.1 Uppsala air-photograph 1950 38 2.2 Uppsala air-photograph 1977 39 2.3 Photographs of the Marsta rural air temperature measurement site 41 2.4 Uppsala sun path diagram for a level site with an unobstructed horizon 42 2.5 (a) Uppsala, Stora Torget 1953 view to east (b) Uppsala, Stora Torget 1998 view to southeast along Kungsangsgatan 45 2.6 Uppsala parking lot adjacent to the RA (Radhuset) site 47 2.7 Uppsala 1946 map, showing the areal urban/rural temperature representativity of the air temperature measurement sites 49 2.8 Sundborg's traverse thermometer setup 51 2.9 Map of Taesler's 1976-1977 Uppsala data collection sites 55 2.10 Lodz map 57 2.11 Lodz sun path diagram for a level site with an unobstructed horizon 59 vi i i 2.12 (a) Urban air temperature measurement screen site at Lipowa, Lodz (b) Hemispheric sky view photograph of the site 61 2.13 Photographs of the rural air temperature measurement screen site at Lublinek, Lodz 64 3.0 Uppsala 1948-1949 observed vs. A S nighttime & daytime UHIs 71 3.1 Uppsala 1976-1977 observed vs. A S nighttime & daytime Uffls 73 3.2 Uppsala, 15-17 February 1977 U H I 74 3.3 Uppsala, 8-14 Apr i l 1976 U H I 75 3.4 Uppsala, 9-12 June 1976 U H I 76 3.5 Uppsala, 28 September-2 October 1976 U H I 77 4.0 Observed vs. A S nighttime and daytime UHIs for Lodz 1997 85 4.1 Observed vs. A S nighttime and daytime UHIs for Lodz 1998 85 4.2 Observed vs. A S nighttime and daytime UHIs for Lodz 1999 86 4.3 Lodz, 1-5 January 1999 Urban Heat Island (°C) 88 4.4 Lodz, 8-13 February 1999 Urban Heat Island (°C) 89 4.5 Lodz, 7-11 May 1998 Urban Heat Island (°C) 90 4.6 Lodz, 5-9 July 1997 Urban Heat Island (°C) 91 4.7 Lodz, 11-14 July 1997 Urban Heat Island (°C) 92 4.8 Lodz, 16-20 August 1997 Urban Heat Island (°C) 93 4.9 Lodz, 6-10 September 1999 Urban Heat Island (°C) 95 4.10 Lodz, 2-6 December 1998 Urban Heat Island (°C) 96 ix ACKNOWLEDGEMENTS This thesis represents the main product of more than two years work. The successful completion of this work was made possible by the generous assistance of many people. To my supervisor, Dr. Timothy Oke, I owe many thanks for his Algorithmic Scheme, continuous encouragement and good-humoured support. He regularly found time to discuss ideas, answer questions and suggest a reasonable limit to my endless search for, and analysis of data and ideas. I also thank Dr. Ian McKendry, my committee member, for his thoughtful reading of this thesis. Dr. Roger Taesler of the Swedish Meteorological and Hydrological Institute, (SMHI), provided 952 hours of fixed screen level data and vital assistance in the analysis of his data. Equally, I thank Dr. Ake Sundborg for the second Swedish data set of 207 vehicle based temperature measurement traverses used in the scheme's testing. Dr. Kazimierz Klysik and Dr. Krzysztof Fortuniak, Department of Meteorology and Climatology, University of Lodz, Poland are thanked for the 1997 to 1999 (26,280 hours) data set that was used to further test the scheme. Many thanks go to Maria Furburg and Henryk Modzelewski for translation help. Funding for this research has been provided via Dr. Timothy Oke, by the Natural Sciences and Engineering Research Council of Canada. Personal funding was provided through University of British Columbia, via Teaching and Research Assistantships in the Department of Geography. x CHAPTER 1 INTRODUCTION Urbanization modifies the landscape, and thereby alters the surface energy balance, and this in turn changes the climate. Many uniquely urban climatic features such as the urban heat island, urban wind circulation, downwind precipitation enhancement, and air pollution have been documented as resulting from these modifications (Landsberg, 1981). The observed increase in average air temperature, the urban heat island, (UHI), is by far the most studied aspect of urban climate. Thousands of scientific papers have been written about urban heat islands since the publication of Luke Howard's pioneer work on London's urban heat island (Howard, 1833). Most publications focus on unique aspects of individual cases. This develops our understanding in a case specific way. Many other papers describe the various 'models' used to expand our understanding of the UHI . Such models cover a wide spectrum from complex energy balance models to simple site-specific multiple regression models. This thesis presents the results of the testing of Oke's Algorithmic Scheme (AS) in an attempt to test and refine a simple and potentially widely applicable model (Oke, 1998). The A S is an empirical scheme based on contemporary understanding of the factors that give rise to, and modulate the heat islands of cities. The general U H I patterns of (mid-latitude) cities are the base upon which scheme components are developed. Testing of the A S is accomplished with data sets from Uppsala, Sweden and Lodz, Poland. 1 1.1 The Urban Heat Island 1.1.1 The nature of heat islands In the strictest sense, the U H I is the difference between the air temperature at a site that is now urban and its pre-urban conditions, stratified by weather type (Lowry, 1977). However, even pre-urban air temperatures are of limited use due to the changes in mean global air temperatures (0.5 to 4.5 °C/century), and inexorable changes in both urban and rural land cover (D?CC, 1995). Furthermore, air temperature measurements for pre-urban conditions are rarely available. Therefore, air temperatures measured in the rural area outside the area influenced by the urban air mass are commonly used as a surrogate for the pre-urban conditions. In this study, and most others referred to herein, the urban heat island magnitude is taken to be the difference between the -1.5 m level air temperature in the urban core and a corresponding 'representative' rural air temperature. This is known as the urban canopy layer, (UCL) , heat island magnitude. The term 'urban canopy layer' is in analogy to the forest canopy and is typically taken to mean the air volume below mean roof level (Figure 1.0). However, where the width of the space between buildings exceeds about three times their height the upper boundary of the U C L is poorly defined (Oke, 1976). Screen level urban-rural temperature differences are often taken uncritically, as a measure of the urban effect on climate. The reliability of such estimates of urban effects is very questionable unless pre-urban values of air temperature are used and account is taken of the possibilities for climatic change (Lowry, 1977). The urban-rural temperature difference is a good measure of the urban effect on air temperature if several conditions are satisfied. These include the requirements that the urban and rural sites are subject to the 2 same background climate, and the same landscape controls, and that measured temperatures are truly representative of their urban and rural areas respectively. When both the urban and rural thermometers are well sited in areas subject to the same regional weather, the differences in synchronous measured temperatures are a reasonable measure of the urban effect upon temperature (Sundborg, 1951). Figure 1.0 Schematic representation of the urban atmosphere illustrating a two-layer classification, urban canopy layer (UCL)/urban boundary layer (UBL) of thermal modification Source: (Oke, 1976) The critical issue, which has been studied by relatively few climatologists, is one of thermometer siting. The choices of sites that are representative of urban and rural land are covered in the pioneering work by Sundborg (1951) in Uppsala, Sweden. Mizuno's (1991) study in Osaka, Japan provides detail about the variable "radius of influence" (Mizuno et 3 al., 1991). This is the size of the area influencing urban air temperature as wind speed varies (Mizuno et al, 1991). These works support the general conclusion that the urban site should be located in a street canyon in the most densely developed part of the city's core. While it is relatively easy to locate the warmest area of air in the 'concrete jungle' of the central business district (CBD) of a city, it is often difficult to pick a rural site that is "representative" of the rural area. A i r temperatures in rural areas are strongly affected by the microscale effects of soil moisture and plant growth related changes. It is often necessary to take an average of two or more stations due to the inherent variability of rural areas. The difficulties associated with rural variability are explained well in the classic work, The Climate Near The Ground (Geiger, 1966). Appropriate location of sites in rural areas requires an understanding of seasonal and the diurnal air temperature variations. Multiple traverses monitoring air temperature in all-weather conditions, time of day and season can facilitate a logical choice of site(s). Suhdborg was among the first to develop this technique from work by Schmidt and Peppier in the late 1920's (Schmidt, 1927; Peppier, 1929; Sundborg, 1951). Mid-latitude urban heat islands commonly exhibit several key characteristics. They are most clearly expressed on clear calm nights following dry almost cloud-free days. Figure 1.1 graphically shows the characteristic cross-section of such a heat island. A i r temperature is shown to rise sharply at the urban/rural border and then remain at a constant level with minor variations in response to variations in urban density. The peak air temperature is observed in the street canyons of the urban core. It is the difference between 4 this peak and the rural air temperature that is taken as the measure of the U C L heat island (AT u . r ) . Figure 1.1(a) shows a cross-section, A - B , of the heat island. Part (b) shows the plan view of the urban area along with the pattern of isotherms. There is a clear 'island shape' to the contours of air temperature, hence the use of the descriptive term 'heat island.' A light wind (-1.5 m s"1) is shown to displace the isotherms slightly downwind of their urban source areas. Figure 1.1 Hypothetical representation of the U C L urban heat island DISTANCE wMmi Built-up area Hypothetical representation of the spatial and temporal features of the urban heat island in the canopy layer of a mid-latitude city with 'ideal' (calm, clear) weather. Spatial pattern (a) along cross-section A-B and (b) in relation to the plan outline of the city. Source: (Oke, 1982) The heat island is usually found to have its largest magnitude several hours after sunset on clear calm nights. This is because the temperature differences arise as a function of differing cooling rates between urban and rural areas. The resultant typical diurnal cycle of heat island growth and decay under ideal conditions, difference in air temperature, cooling/heating rate and the resultant heat island intensity are illustrated in Figure 1.2. The fundamental difference is that urban cooling is less than in the rural surroundings in the early part of the night. 1.1.2 Causes of heat islands Several processes are put forward as being potentially responsible for the urban-rural temperature difference (Table 1.0 gives a concise summary). Urban canopy geometry in the central portion of a city, (as measured by the sky view factor, SVF) , is a relevant variable in producing nocturnal urban heat islands due at least in part to its role in regulating net longwave radiation (Oke, 1981). The canyon sky view factor is the proportion of the overlying hemisphere that is sky as 'seen' from the center of the canyon floor. It is therefore a measure of the maximum radiative infrared cooling potential of the canyon. A high sky view factor is indicative of an area where nocturnal cooling can be rapid due to the large area of cold sky that is visible from the center of the canyon, in the absence of cloud cover. The role of the relative proportions of cold sky and relatively warm buildings in the longwave cooling process can be significant (Oke, 1981). However, many other factors such as the type of construction materials, the active surface area, the amount of vegetated area and the amount of anthropogenic heat released by combustion of fuels, also indirectly vary with S V F . Combined they explain the 6 significance of central city canyon geometry to heat island formation (Oke, 1981). Empirical evidence strongly suggests that the S V F of the canyons in the urban core is related to the maximum urban heat island in settlements with populations from -10 to 10 7 (Oke, 1981). Figure 1.2 Typical temporal variation of urban and rural: (a) air temperature, (b) cooling/warming rates, and (c) the resulting heat island intensity (AT u - r ) under 'ideal' weather conditions T 1 1 1 1 1 r I i i i i i i i i • ' • i 12 18 24 06 12 T I M E (h) 7 Source: (Oke, 1987) Table 1.0 Commonly Hypothesized Causes of the Urban Canopy Layer Heat Island Altered energy balance terms leading to positive thermal anomaly Features of urbanization underlying energy balance changes 1 Decreased long-wave radiation loss from within street canyons due to the complex exchange between buildings and the screening of the skyline Canyon geometry - reduction of sky view factor 2 Increased storage of sensible heat in the fabric of the city Construction materials - increased thermal admittance (stone, asphalt, concrete and steel) 3 Anthropogenic heat source Building, traffic and metabolic heat sources 4 Increased absorption of short-wave radiation Canyon geometry - increased surface area, reduced albedo and multiple reflection 5 Increased long-wave radiation from the sky A i r pollution - greater absorption and re-emission 6 Decreased evapotranspiration Construction materials - increased 'water proofing' and improved surface drainage 7 Decreased total turbulent heat transport Canyon geometry - reduction of mean wind speed Source: (Oke, 1987) Daytime heat island conditions should also be related to S V F . Daylight conditions are more complex than those at night due to the addition of shortwave (solar) radiation to the energy budget, and increased instability and wind speed that promote convection and advection respectively. Shading and complex multiple reflection of the shortwave radiation from the many facets of the urban canyon complicates the daytime situation. The relation between canyon geometry and daytime heat and cool islands (urban site cooler than the rural) has yet to be significantly explored. It has been suggested that cool island situations 8 arise, at least in part, due to the shading of the surface in urban areas and the thermal inertia of the urban fabric (Oke, 1987). Cool islands occasionally occur in the morning and early evening i f the rural area warms more rapidly than the urban area. There is some evidence that this occurs under clear sky conditions and not under overcast skies (Hilberg, 1978). The thermal properties of the surface are significantly altered by urbanization. Porous 'natural surfaces' (soil, vegetation) are replaced by impermeable asphalt, concrete, stone, wood and glass surfaces. Precipitation no longer has such a dramatic effect on surface thermal properties due to much lower water retention. Surface thermal admittance is generally increased by urbanization. This increase may only be significant in the dry season because wet soils have thermal properties similar to those of construction materials (Oke, 1988). The high proportion of the urban surface covered by asphalt and concrete has a potential to increase nocturnal air temperature due to the increased ability of the urban surface to store and subsequently release heat (Asaeda et al, 1996; Anandakumar, 1999; Grimmond and Oke, 1999; Hoyano, 1999). Anthropogenic heat due to combustion from buildings, vehicles and animal metabolism has an obvious potential to increase the heat island magnitude. However, during the daytime and summer this potential is usually small relative to the much larger solar input. Winter conditions in many mid-latitude urban areas are such that anthropogenic emissions of about 46 - 69 W m" 2can be about twice the average net all-wave radiation (Oke, 1987; Klysik, 1996). The effect of this energy source on air temperature is likely to be significant but in the canopy layer, it is not easy to discern or calculate. 9 The impact of anthropogenic heat on air temperature appears to be quite small on average. A study in the very densely developed, Otemachi core area of Tokyo, suggests that anthropogenic heat, (averaging -200 W m" 2 winter and -120 W m" 2 summer), raises air temperature by an average of -1.0 °C in winter and -0.5 °C in summer (Ichinose et al, 1999). At 9 pm in winter, the air temperature is raised by 2.5 °C, this occurs long after the 9 am peak anthropogenic heat input, of more than 400 W m" 2. Study results are based on modelling where all anthropogenic heat is released at ground level. The Tokyo study makes the point that "[there is a general] failure to consider the emission height of anthropogenic heat" (Ichinose et al. 1999), a point made earlier by Oke (1988). It also points out the many areas where our understanding of anthropogenic heat emission and its effect on air temperature is lacking, including emission height, timing of emissions, diurnal and seasonal storage and release and the interaction between breezes and heat emission (Ichinose et al, 1999). Urban areas in mid-to-high-latitude cold winter climate conditions may display large heat islands, -12 to 14 °C, in conditions where solar input is close to zero and anticyclonic conditions occur immediately after passage of a cold front (Bowling and Benson, 1978; Rastorgueva, 1979; Klysik and Fortuniak, 1999). Clearly, the rural snow-cover and anthropogenic heat emissions (Table 1.0, items 2 & 3) play a leading role under such conditions. Snowfall heightens urban-rural differences because urban snow is quickly removed and soiled whereas rural snow typically remains as an undisturbed highly reflective and insulating blanket (Oke, 1987). As little as, 15 cm of fresh snow can dramatically reduce the surface thermal admittance and thereby practically eliminate any 10 release of ground heat (Oke, 1987). Given that urban surfaces are little affected by snowfall, this creates conditions where heat island magnitude can reach 6 °C due to urban-rural thermal admittance differences alone (Oke et al, 1991). There is some evidence that, in agreement with the model results of Oke et al. (1991), the magnitude of the U C L heat island does have an inverse relation with the rural surface thermal admittance. A study of the canopy layer heat island of Vancouver, B . C . , has found that after normalization for weather effects, as a rural silt-loam soil becomes more moist thereby increasing rural thermal admittance, the mean weekly heat island declines (Runnalls and Oke, 2000). 1.1.3 Urban heat island models Models are created in part to overcome the difficulties and expenses involved in full-scale measurement programs. Urban climatologists also use models to develop their understanding of complex systems under a wider range of conditions than is generally possible in fieldwork. A range of urban heat island models has been created since 1950 (e.g. Oke, 1986; A S H R A E , 1993). These models range from simple site-specific statistical models to complex three-dimensional models that require great computing power and large data inputs for initialization. Few urban climatologists apply models in an operational situation, i.e. where forecasts are based on model output. Road weather forecasting is an exception, since it involves application of simple physical principles to the prediction of road surface temperature. It is a fast-developing field in many mid-latitude countries (Bogren, 1998; McClean and Oke, 1998; Wood et al, 1998; Bogren et al, 2000). Road surface 11 temperature is typically predicted from a surface energy balance relation, which allows for site-specific variables: surface material, cloud cover, wind, elevation and sky-view-factor (Thornes and Shao, 1991; Lindqvist, 1992; Shao and Lister, 1995). The accurate prediction of when, and even if, a road is likely to reach the freezing point allows highway departments and airport authorities to save financial and environmental costs of road salting. Models that have application in the forecasting of specific urban effects upon weather are relatively new. Most mesoscale weather forecast models do not accurately incorporate the microscale effects of urban areas and certainly not the canopy layer. A n exception to this is Masson's T E B (Town Energy Balance) model, which can be coupled to an existing mesoscale forecast model nested in a standard synoptic forecast model (Masson, 2000). Large cities such as London significantly modify the local climate (Chandler, 1965; Atkinson, 1977) but only since 1998 has this modification been incorporated into the U K Meteorological Office's forecast process (Best, 1998). Best's simple model forecasts the surface temperatures of vegetated and paved surfaces to within 1-2 °C of the measured temperature via a crude energy balance model. Inputs to the model are weather parameters and mean estimated thermal and radiative properties of the surfaces in question. The modern approach to studying urban heat islands took a big step forward in 1951 with the publication of Ake Sundborg's study of Uppsala's thermal climate (Sundborg, 1951). His work is considered the classic urban climate study from the immediate post WWII period (Oke, 1995). This early multiple-regression heat island model is noteworthy for many reasons. In particular, Sundborg was the first to develop a 12 statistical model of heat island-weather interactions and he was the first to cast the causation of the heat island overtly in terms of the energy balance (Oke, 1995). Sundborg found that variation of urban-rural temperature differences, ( A T u . r ) , were due largely to weather conditions (mainly wind and cloud) and their interaction with urban and rural surfaces. Temperature difference predictors were selected in the expectation that they serve as surrogates for the important processes governing the creation and destruction of temperature contrasts. Cloudiness (1 to 10), wind velocity (m s"1), air potential temperature (°C), and vapour pressure (mm of mercury), were chosen as inputs to a statistical urban heat island model. Overall the nighttime model predicted the urban-rural temperature difference better than did the daytime model, r 2 = 0.44 and 0.24 respectively (Sundborg, 1951). A similar multiple-regression heat island model was used by Chandler in his comprehensive study of the climate of London, England (Chandler, 1965). However, Chandler used a much larger data set of 1,462 records, which was facilitated by the availability of punched cards and a computer. Different regression equations were fitted to summer and winter data. These seasonal heat island data sets were further sub-divided into daytime and nighttime data sets. Chandler's regression model is essentially the same as Sundborg's except that temperature range is substituted for the atmospheric vapour pressure term in the regression equation. Overall, this model performed about as well as Sundborg's model. Both models had a coefficient of determination (r 2) of about 0.36 for the summer nighttime heat island. 13 The statistical modeling approaches of Sundborg and Chandler are empirical and assume that relations are linear. Linearity is not necessarily true, especially for the full range of wind speeds recorded. Extreme wind speeds are therefore not accurately modelled. Both models more accurately predict the heat island at night when the atmospheric conditions are less complex. The work of Sundborg and Chandler also illustrates the point that there is a significant diurnal and seasonal pattern to heat island magnitude and it is difficult to model these aspects. Empirical statistical models of this type have simple meteorological variables as inputs. However, the coefficients are site specific. Sundborg's model has a coefficient of determination of 0.24 in the summer and winter daytime (Sundborg, 1951), Chandler's performs poorly (r 2 = 0.08) in the summer daytime (Chandler, 1965). These models have the merits of simplicity, cheapness and that they predict the urban-rural air temperature difference. Their major drawback is that a full set of antecedent data is needed to build the statistical relations. Once calculated the regression coefficients are not transferable, a new model is required for every place of interest. More process-orientated U H I models were developed in the 1960's, starting with one-dimensional energy balance models, based on the physics of the surface energy budget (Myrup, 1969; Carlson et al, 1981; Johnson et al, 1991; Swaid, 1991). One-dimensional energy balance models work reasonably well in cities with a uniform structure and they are more physically meaningful than statistical models. However, input requirements are significantly higher and the output is surface temperature and not the air temperature difference. 14 Three-dimensional (3-D) dynamic differential models allow the full 3-D variability of the city to be accounted for (Zdunkowski et al, 1976). This capability comes at a very high price in terms of large input requirements and a need for great computing power. Masson (2000) uses a generalisation of local canyon geometry instead of the usual bare soil formulation currently used to represent cities in atmospheric models. The scheme is aimed to be as general as possible, in order to represent any city in the world, for any time or weather condition (heat island cooling by night, water evaporation after rainfall and snow). Model output gives air temperature for several layers including one in the U C L (Masson, 2000). Of the mesoscale models, it gives the most realistic output for the U C L energy balance (Masson et al., 2002). Few researchers have the computational resources and experience required to run such models. Another significant limitation of three-dimensional models is the need for extensive meteorological data and boundary condition specification. Very few models have been fully validated, largely due to the lack of suitable data sets. Of the few models that have performed acceptably, fewer still are of practical use due to their significant input requirements. Even fewer apply to the U C L . For all but the few well-funded studies or large agencies, this type of model is too complex and expensive to use. There is a need for an U C L model that can predict the heat island for any urban area at any time of day. Such a model should ideally be able to give accurate predictions with only modest demands for readily obtainable input variables. Ideally, the input 15 meteorological variables should be available from principal weather stations or at most require estimates of surface parameters and the computation of some derived terms. 1.2 The Value of an Operational Urban Heat Island Scheme The ability to accurately estimate the U C L urban heat island magnitude is important for many reasons other than scientific curiosity. Concern about the thermal comfort and air quality for people who live in cities is growing rapidly (Hoppe, 1991; Swaid et al, 1993; Bridgman et al, 1995; Kassomenos et al, 1995; Bretz et al, 1998; Fenger, 1999). It is estimated that about three-quarters of the world's population lives in urban areas, and that 25 conurbations of over 10 million people, and 330 cities of over 1 million exist (United Nations, 2000). Much of the current population growth is concentrated in cities that already experience significant heat island exacerbated problems, such as, photochemical smog. Human thermal comfort and smog are significantly worsened by urban heat islands in many parts of the globe (Fenger, 1999; Becker, 2000). A scheme able to predict canopy layer heat island magnitude has value in a wide range of applications. Key uses include urban weather forecasting, urban planning, road weather forecasting, fog prediction, mixing depth estimation, plant growth, air conditioning and heating engineering and the correction of air temperature records for urban bias. Human comfort is strongly affected by summertime thermal stress and the coincident surface air pollution that is indirectly increased by the thermal effect that cities have on the urban boundary layer, (UBL) , pollution chemistry (Fenger, 1999) (Figure 1.0). 16 High magnitude urban heat islands, and periods with high levels of surface ozone, are both strongly associated with warm sunny weather found in high-pressure (anticyclonic) systems (Heidorn and Yap, 1986; Sillman, 1990; Sillman, 1995). High levels of ozone are a serious problem in urbanized areas throughout the world (Fenger, 1999). The rate at which volatile organic compounds (VOCs), especially hydrocarbons, are released, and the rate of photochemical production of ozone that ultimately reaches ground level are both increased as temperature increases (Comrie, 1990; Rosenfeld et al, 1998; Fenger, 1999). Urban heat islands contribute to increased ozone damage to vegetation, as was first noted in the Los Angeles area in the 1940's (Haagen-Smit, 1952). Studies link microscale heat islands in shopping center parking lots to the enhanced emission of hydrocarbons due to the extra volatilization from the gas tanks of hot cars (Quattrochi et al, 1998; Rosenfeld et al, 1998; Tahaetal , 1999). The nocturnal release of heat stored in urban building and paving materials elevates nocturnal air temperatures and reduces or eliminates the opportunity to recover from daytime heat stress (Asaeda et al, 1996; Asaeda and V u , 1997). Deaths result when summer heat-wave conditions are exacerbated by nocturnal heat island conditions (Karl and Knight, 1997). Five hundred deaths in Chicago in July 1995 were attributed to heatwave conditions when minimum (nocturnal) apparent temperature remained above 31.5 °C for more than 48 hours (Karl and Knight, 1997). Higher urban temperatures significantly increase electricity consumption for air-conditioning. It has been estimated that electricity demand in five American cities, (Los Angeles, C A ; Washington, D C ; Phoenix, A Z ; Tucson, A Z ; and Colorado Springs, CO), rises by 2-4 % for each 1 °C rise in daily maximum 17 temperature above a threshold of -17 °C (Rosenfeld, 1995). There is the potential to save, in the U S , annually $10 billion in energy and equipment costs, and to eliminate 27 million metric tons of C 0 2 emissions (Rosenfeld, 1995). Recent work by Rosenfeld suggests that there is the potential to reduce urban smog by 20% and US health costs by $5 billion (Rosenfeld etal., 1998). Current interest in estimating past and present climate change has greatly increased interest in the development of schemes to estimate the effect of urbanization on long-term temperature records (Karl et al., 1986; Karl and Jones, 1989; Nicholls et al., 1996; Bohm, 1998; Brunetti et al, 2000). It is clearly a challenge to disentangle the rate of global warming from temperature records that incorporate the local thermal effects of urbanization, especially, given that such effects are recognized as being of similar magnitude to those of the global temperature trend. It is possible that the present scheme wil l have utility in the process of correcting 'urban bias' in air temperature records via retrodiction of urban heat island effects. One of the longest air temperature records to receive detailed analysis and correction for urban bias and other non-homogeneities is the long-term (1722-1995) record for Uppsala, Sweden (Moberg and Bergstrom, 1997). Moberg's urban bias correction techniques rely heavily on statistical homogeneity tests rather than the application of a more physically-based analysis of the underlying urbanization and resultant heat island development. It is suggested that where urban morphological change is well documented, the present heat island scheme could potentially provide an estimate of the urban air temperature bias in the record. 18 1.3 Description of the Algorithmic Scheme (AS) The Algorithmic Scheme (Oke, 1998) predicts/retrodicts urban-rural air temperature differences A T u . r at any time in any weather as: AT u -r = A T u . r ( m a x ) O m O w 0>t (1.0) The first term on the right hand side ( A T u . r ( m a x ) ) gives the absolute maximum U H I likely to be recorded for a particular city (Oke, 1981). This term is predicted by an empirical relation, (Figure 1.3), based on data for a total of 31 North American, European and Australasian cities in 'ideal' (calm, clear) conditions at the time of the nocturnal maximum (Oke, 1981). These results are from summer conditions with little anthropogenic heat and with relatively dry rural soils. The critical measure underlying the value of A T u _ r ( m a x ) in these conditions is the sky view factor (\|/s), which can be obtained using photographs from the center of these canyons. In the absence of such photographs, a reasonable surrogate for \|/ s is obtained using the average canyon aspect ratio Xs (= H / W , where H is height and W is width of streets) in the urban core (Figure 1.3). This ratio is calculated for the center of the canyons via estimates of average building height and street widths. The forms of the two relations are: A T u . r ( m a X ) = 15.27 - 13.88 (\|/s) (1.1) A T u . r ( m a x ) = 7.54 + 3.97 In (ks) (1.2) where A T u . r ( m a x ) i s the absolute maximum urban-rural temperature difference that would be measured by a vehicle traverse of the city on calm, cloudless nights preceded by cloudless sunny weather (Oke, 1981). Equations 1.1 and 1.2 relate maximum urban-rural 19 Figure 1.3 Relation between maximum heat island intensity and height to width ratio (H/W) of the street canyons in the centers of 31 cities for equation 1.2. Data from the study of Oke, (1981) (a) O o < Source: Oke, (1987) (b) 12 10 < O Europe • N.America + Australasia D.5 1.0 1.5 ' 2.0 2,5 3.0 IS Xs = H/W xs = nrw 10 5 3 2 1 0.5 0.25 12 10 -r—i—r 28#\ O Australasia jg.> o Europe so* • N. America • 3 t Q1 1 1 1 1_ 0.2 0.4 0.6 0.B 1.0 Source: (Oke, 1986) temperature difference to a measure of canopy geometry with a coefficient of determination (r2) of 0.88 and a standard error of estimate of ± 0.92 °C (Oke, 1981). The work of Mizuno et al, (1991) provides some guidance on the subject of urban air temperature and the size of the surrounding urban area that influences it. Measurements in Osaka, Japan show that the 'influence radius' in an urban area is -75 m with urban wind speeds at -10 m of 1.0 - 2.5 m s"1. Wind speeds in urban areas are generally less than the speeds at the same height in their rural surroundings (Oke, 1987). Therefore, it is reasonable to suggest that air temperatures measured in urban areas are typically a measure of the microscale temperature of an area with a radius of about the length of a typical city block (-75 m) when winds are light (< 3 m s"1) (Mizuno et al, 1991). Clearly, assigning the appropriate Xs ratio for the city being modelled is critical since it sets the maximum heat island. If a temperature measurement site is fully enclosed or is in the center of a large open area of at least 75 m radius then the scheme uses the site Xs. When the site is connected to canyons within 75 m then the effective X s is calculated from equation 1.3. Where the urban canyons in the core of a city have uniform dimensions this is an easy task, however, in many cities the choice requires some averaging or estimation. It is anticipated that the microclimate of the canyon airspace surrounding the point of temperature measurement has the largest influence on the recorded temperature. Secondarily, adjoining canyons exercise some influence on the recorded temperatures, but this wi l l decline with distance and degree of connectedness. This leads to the need to calculate an average Xs ratio for the combination of airspaces surrounding 21 the point of measurement up to a radius of say 75 m, as suggested by the work in Osaka. A weighting of canyon Xs ratios has been used consisting of half the value of the ratio of the measurement site, (ks sae), and in addition, half the averaged values of the adjoining canyons within a -75 m radius of the site (A,s adj i): N K = 0.5 (A s site) + 0.5 Z ( X s a d j i ) / N (1.3) i=l In the A S the A T u . r ( m a X ) term is followed by three 'phi' reduction terms; O m O w O t . Each term has a value that can range between zero and one. This gives a scheme where the maximum heat island magnitude is reduced by the product of three terms each of which expresses the degree to which the actual magnitude is less than 'ideal' nocturnal value because of the prevailing rural surface thermal admittance, weather or time of day. 1.3.1 The rural surface thermal admittance factor (<Pm) The term to account for thermal admittance effects, (O m ) , is developed from the results of a surface heat island model (Oke et al, 1991). This work shows that thermal admittance of the rural surface, (Ur), can be a significant control on the surface heat island magnitude. Thermal admittance is the square root of the product of volumetric heat capacity (C) and thermal conductivity (k). Large values of \i (-2000 J m"2 s"1/2 K" 1 ) indicate wet surface conditions that allow the rapid take-up or release of stored heat and result in a 2 1/2 1 relatively small diurnal range of surface temperature. Small values of u. (-600 J m" s" K" ) indicate dry surface conditions that allow the very slow take-up or release of stored heat and a large diurnal range of surface temperature. 22 Modell ing suggests that after canyon aspect ratio the rural surface thermal admittance, and the urban-rural admittance difference, is potentially the strongest physical controls on heat island magnitude (Oke et al, 1991). The derived O m term here is given by: O m = [ AT u _ r ( r n a X ) ) / \ | / s ] actual / [ A T u . r ( m a x ) ) / \ | / s ] dry (dimensionless) (1.4) where [ A T u . r ( m a x ) ) / \|/ s ] d r y = 15 °C is based on empirical data in Oke (1981) and SHTM (Oke et al., 1991) suggests this corresponds to u,. = -600 J m"2 s"1/2 K " 1 . S H I M further suggests that the numerator of equation (1.4) is -10.0 °C at 1000, 7.4 °C at 1400, 5.6 °C at 1800 and 4.0 °C at 2200 J m"2 s"1/2 K" 1 . Here, the numerator is the slope of the line linking the maximum heat island magnitude under the prevailing rural soil moisture conditions to the S V F of the canyons in the city centre (Oke, 1998; see equation 1.1 & Figures 1.3 & 1.4), and the denominator is the slope of the relation linking the maximum heat island magnitude to the S V F when the soil is effectively dry (i.e. the largest values possible for that location). Modelling work, suggests that this slope is about -15 °C in dry conditions and about -4 °C when waterlogged (Oke et al., 1991). The Oke (1998) pre-print paper erroneously reported the slope as -8 °C in waterlogged conditions. Inspection of the lowest line in Figure 1.4 of (Oke et al., 1991) clearly suggests a slope of -4 °C is what was intended. This gives an operational range of about 0.27 to 1.00, for O m , in snow-free conditions. Using the values in Figure 1.4, the O m term can be calculated from rural soil thermal admittance values ( U j ) via equation 1.5, which in turn are calculated from soil moisture values parameterized in the form O m = a - b In (u..r) {equation 1.5 & Figure 1.5}: O m = 4.6 - 0.56 In (u. r) (1.5) 23 where the units of b are [J m" 2 s"'/2 K" 1 ]" 1 and |ar is the measured or estimated, (Table 1.1), mean rural thermal admittance (J m"2 s"1/2 K" 1 ) . Soil thermal admittance can be estimated from an energy balance-water balance calculation. This approach uses a measured or Figure 1.4 Relation between maximum heat island intensity and sky view factor of the street canyons in the centers of cities. The curves for five values of rural thermal admittance at increments of 400 units from 600 (upper) to 2200 2 1/2 1 J m - Z s " u K" ' (lower) are based on numerical output from S H I M . The sketched straight lines are first approximations to the curves. The solid symbols are values from 31 North American, European and Australasian cities, Open symbols are from Japanese and Korean cities. 12 | 1 1 f i 1 1 1 1 1 1 1 • . Australasia • , Europe Okeet a l , (1991) v|/s Sky View Factor 24 estimated energy balance to determine the latent heat flux that in combination with rainfall data can provide an estimate of soil moisture content. Soil moisture content for a given soil type is then used to calculate the soil thermal admittance via published relations such as those of Campbell & Norman (1998), and this value of n is entered into equation 1.5 to give 3>m. Figure 1.5 Relation between the thermal admittance reduction factor ( O m ) and rural surface thermal admittance based on results from the SHTM numerical model (Okeetal., 1991). o ta c o o 3 a> —. • I o © c CO E •o CO "Jo E k_ CU JC 1.0 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0.0 I « I Om= 4.6 - 0.56 Ln (|ir) < 200 400 600 800 1000 1200 1400 1600 1800 2000 2200 •2 -1/2 -1 Rural surface thermal admittance ((i r) (J ni" s" K" ) The energy balance-water balance approach involves estimating the net radiation, Q*, available because that is the amount of energy available to evaporate water from the rural soil. If precipitation is measured, then the soil moisture content can be estimated by 25 calculating the water balance, and u. can be estimated with equation 1.5. This approach was used for both the Uppsala (1976-1977) and the Lodz (1997, 1998 & 1999) data with minor differences. The following are the details of how this method is implemented. The surface radiation balance: Q* = Ksl-KT+Lsl-LT ( W m 2 ) (1.6) forces the surface energy balance. In equation 1.6, each term is a flux density and Q* is the net all-wave radiation, K-l is the incoming shortwave radiation, K Tis the reflected shortwave radiation depending on the surface albedo, a, L-lis the incoming longwave radiation emitted by the atmosphere and cloud, and L T, is the outgoing longwave LT=e0CTT(t+{\-e^Li ( W m 2 ) (1.7) radiation emitted or reflected by a surface as a function of the surface temperature (T0) and emissivity (£(>), (e0 here is taken to be 0.93). Where measured solar radiation is not available it can be calculated. One of the simplest approaches is given by Holtslag and Van Ulden (1983). They use a simple formula to estimate the clear sky Kj- which requires only one input, solar elevation angle <j>: KJ- = a,sm0 +a2, " ( W m 2 ) (1.8) where aj and a 2 are empirical coefficients. Holtslag and Van Ulden give mid-latitude temperate city values for aj and a 2 values of 990 and -30 W m"2, respectively. Solar elevation angle, 0, is available from standard astronomical tables or an astronomical web site such as http://www.usno.navy.mil/. If cloud cover significantly affects the incoming 26 solar radiation, Holtslag and Van Ulden (1983) provide a simple formula for the calculation of all-sky (cloudy conditions) KJ- with cloud cover N varying between 0 (clear sky) to 1 (overcast). Kl = (a, sm<f> + a2) (1 - b,Nbl) (W m"2) (1.9) where the first bracketed term is equation 1.8, and bi and b2 are 0.75 and 3.40, respectively. Reflected solar radiation is given: Kt=aKJ- ( W m ' 2 ) (1.10) where, a, is the rural surface albedo, here a value of 0.20 is used for the rural areas in Uppsala and Lodz (Oke 1987, p. 12). The formula derived by Prata (1996) is used here to estimate incoming clear sky longwave radiation, LJ-. It only requires atmospheric vapour pressure ea (hPa) and screen level temperature (Ta), a is the Stefan-Boltzman constant (5.67 x 10"8 W m"2 K" 4 ) : hi = [1 - (1 + ( 4 6 . 5 ( e f l / r f l ) ) ) exp{-(1.2 + 3 .0 (46 .5 ( e a / r a ) ) ) 1 / 2 }]c r7 ; 4 . (W m"2) (1.11) If vapour pressure is not available, two other formulae such as those of Idso and Jackson (1969), and Swinbank (1963) can be used. They only require screen level temperature, Ta, as an input. Incoming longwave radiation under cloudy skies can be estimated by applying a correction to the above clear sky results using a correction formula due to Bolz (1949): Lively = L i c l e a r { \ + aN2) (W m"2) (1.12) where a is 0.220 for low cloud, 0.185 for middle level cloud, 0.060 for high cloud and 0.250 for fog. Middle level clouds are those with a cloud base elevation of 2,000 to 6,000 m. 27 Values of 'a' for different cloud types at all three levels are given by (Oke 1987, p.374). Cloud cover /V varies between 0 (clear sky) to 1 (overcast). In this study the cloud layer of interest for this adjustment was the one with the greatest coverage. In cases where cloud at two or more heights had equal coverage, the lowest was the one used. In summary, the surface radiation balance (equation 1.6) is calculated as: Q* = [{(a; s in^ + a 2 ) ( l - bjN^-aKJ-MLJ-il + aN2) - {e0 aT04+ (1 - £0) LJ-}] (1.13) The Rural surface energy balance is: Q* = QC + QH + QE + QA ( W m 2 ) (1.14) where Qc is the net heat flux density stored in or released from the rural surface, QH is the turbulent sensible heat flux density, QE is the turbulent latent heat flux density, and AQA is net horizontal heat advection. Given that the Uppsala and Lodz rural areas are sufficiently extensive and uniform for several hundred meters around the thermometer sites, the A.QA term is considered negligible. Hence in the present study the rural surface energy balance is: Q* = QG + QH + QE ( W m 2 ) (1.15) QG is estimated as a simple linear fraction of Q* (e.g. Hanna and Chang, 1992): Qc = 0.1 Q* ( W m 2 ) (1.16) and QE is estimated via the method given by Hanna and Chang (1992): QE= (aPM 11+ S)(Q*-QG) + aPM 20 (W irf 2) (1.17) where CCPM is the Penman-Monteith surface moisture availability factor. A value of 0.4 was chosen for the Uppsala and Lodz sites. The parameter S = cpl{Le dqJdT) varies with 28 temperature, where Le is the latent heat and dqJdT is the slope of the saturation specific humidity vs. temperature curve. Using the estimated value of QE, the depth of water (mm) evaporated in an hour is calculated to be = QE(3,600 / 2,477,000) (mm H 2 0 /h ) . The net effect on soil moisture content is given by a simple running water balance of the soil column. This means setting initial soil moisture content at the start of a period and reducing/increasing the amount based on the net effect of the evaporation losses and precipitation gains. The saturation water capacity of the clay soil in Uppsala was taken to be 320 mm (Jarnefors, 1956) and of the sandy loam soil in Lodz to be 300 mm (Campbell and Norman, 1998). Once the soil moisture estimate is known, the soil thermal admittance is calculated based on the relation between soil moisture and thermal admittance (ju r) given in Campbell and Norman (1998) Figure 8.5, p. 125. For the Uppsala clay soil: f*r = 440.3 + [54.62*(sm/l,000)] + (24630*(sm/l,000) 2] - [67770*(sm/l,000) 3)] + [57510*(sm/l,000) 4] (J m " V / 2 K ' 1 ) (1.18) and for the Lodz sandy loam soil: [ i , = 560.0 + [5785*(sm/l,000)] - (4625*(sm/l,000) 2] (Jm" 2 s" 1 / 2 K" 1 ) (1.19) where 'sm' is the soil moisture content in mm for a.given hour. With the value of jur defined, the soil moisture reduction factor <J>m can then be calculated with equation 1.5. <£m is treated as a non-dimensional scaling factor. This lengthy estimation produced realistic values for thermal admittance, 600-1281 (J m" 2s" 1 / 2 K" 1 ) in Uppsala (1976-1977), and 600-2 1/2 1 1896 (J m" s" K" ) in Lodz. However, this energy balance-water balance approach to estimation of thermal admittance was not tested because measured thermal admittance values were not available. 29 Table 1.1 Table of rural surface thermal admittance (Ur) values S U R F A C E C O V E R Pore space % Thermal admittance values (J m 2 s"1/2 K" 1 ) SNOW New fluffy New dense Average Aged Old dense Ice 140 250 400 600 900 2080 S A N D Y SOIL Dry Quite Dry Moist Very moist Wet Very wet Saturated 20 • 700 1400 1950 2100 2250 2300 2700 30 650 1200 1700 1800 2000 2100 2300 40 550 1100 1550 1950 1400 1900 2550 60 (+Organic) 400 600 1000 1100 1200 1500 1600 L O A M SOIL 20 700 1100 2100 2150 2250 2300 2700 30 500 900 1600 1800 2000 2100 2300 40 440 600 800 1200 1400 1900 2100 60 (+Organic) 200 400 600 850 1200 1350 1500 C L A Y SOIL 20 700 1100 1950 2100 2250 2300 2700 30 600 900 1600 1800 2000 2100 2300 40 500 600 1300 1400 1650 1900 2200 55 450 600 950 1200 1450 1700 1900 60 (+Organic) 250 450 600 850 1200 1350 1500 ORGANIC SOIL 60 100 250 300 650 850 1100 1300 70 100 300 350 700 900 1200 1400 80 100 350 550 700 900 1200 1400 Sources: (Campbell and Norman, 1998; Monteith and Unsworth, 1990; Oke, 1987; Pratt and Ellyett, 1979) Light grey shading indicates the (u,) values used for tin- 1948-1949 Uppsala dataset. 30 1.3.2 The weather factor ( &w) Many studies of UHI's suggest that wind and cloud both exert a strong control on heat island magnitude. Wind is a surrogate for advection and turbulent mixing, both of which increase with increasing wind speed. Cloud is a surrogate measure of the infrared cooling potential. It is commonly accepted that based on the several previous studies of the effects of weather upon UFA magnitude, wind and cloud, in that order, are the most significant controls (Oke, 1987). Where studies have focused on cloudless conditions, wind speed (u) has been shown to show a relation with heat island magnitude that is well described by a (u"'/2) function (Oke, 1976; Uno et al, 1988). A similar, but logarithmic, relation was found for Goteborg, Sweden (Eliasson, 1996). These functions give a relation where A T u . r ( m a x ) is most affected by wind speed changes at low wind speeds when microclimatic differences across the landscape are best developed (Oke, 1998). Few studies have looked at the relation between heat island magnitude and cloud cover, and even fewer at the effect of cloud type (height). Most previous studies simply used cloud cover in tenths or octas without consideration of cloud base height (temperature) and its role in longwave cooling. Pioneering work by Field (1973) provided the basis for taking into account the significant effect of cloud height. Field found that, all other factors being equal, heat island magnitude decreases, as cloud base height becomes lower. Oke (2000, pers. comm.) found the Bolz (1949) correction formula for net longwave radiation (L*) is a good parameterization of Field's results. 31 Similar to equation 1.12, the Bolz formula that incorporates the effects of both cloud amount (n in tenths) and type (cloud height) is: L* = L*(o)(l-bn2) (1.20) where L * ( 0 ) is net longwave with cloudless skies, and values of 'b' are given in Oke (1987, p.374). Oke (1998) proposes that an appropriate form for the "weather factor" (<J>W), that combines the effects of wind and cloud is <Dw = u ' / 2 ( l -bn 2 ) (1.21) This equation expresses the degree to which weather controls reduce the maximum U H I on a given night below the maximum possible, A T u . r ( r n a x ) (Oke, 1998). Values range from zero in windy, low overcast conditions to unity when near calm and cloudless. To prevent values becoming unmanageable, wind speeds are never allowed to be < 1 m s"1. Antecedent weather effects on air temperature are taken into account by calculating the numerical value of <E>W as a running two hour average of the current and preceding hour's O w value. O w is treated as a non-dimensional scaling factor despite its actual units (m s - y / 2 . 1.3.3 The Temporal Factor (&t) Heat island studies usually find the largest U H I magnitudes at night. B y day, there are small heat islands or even cool islands. Oke (1998) took the innovative step of investigating the notion that i f weather and day length controls were held constant there might be a simple underlying temporal form to the U H I . He selected fine weather data from four mid-latitude cities Edmonton, London, St. Louis and Uppsala and plotted the diurnal variation of the U H I (using the normalized value A T n ) against a normalized time 32 scale (tn). To do this, the day was divided into day (heating) and night (cooling) periods based on the time of sunrise and sunset. Then each period was divided into 10 equal parts; the real time length of which therefore depends on day length. After normalization the different U H I , which originally had different magnitudes and phases, fell closely upon a single shape (Figure 1.6) given by: = 0 .004 / (0 .004+(0 .060 ( 4 0 6 4 ( R T n ) ) ) ) for noon to midnight (1.22) O t = 0.011 / (0.01 l + ( 1 4 . 4 4 2 ( " 4 1 9 5 ( 2 ( R T n ) ) ) ) ) for midnight to noon (1.23) where (R Tn) is the reset normalized time of day. The heat island growth is well described by equation 1.22 and the decay by equation 1.23. It should also be noted that two other cities (Birmingham, England and Vancouver, Canada) have a slightly different nocturnal pattern (Oke, 1998). The reasons for this difference have not been found. O t again ranges from zero in the middle of the daylight period, to unity in the middle of the period from sunset to sunrise. 1.4 Objective of the Research The objective of this research is to test, and i f necessary, refine the Oke (1998) Algorithmic Scheme. The goal is to work towards a simple, universally-applicable scheme to predict or retrodict the hourly urban canopy layer U H I , to enable operational climatologists to estimate the canopy layer heat island magnitude for all cities, in all weather conditions, using as inputs only readily estimable or available meteorological and surface descriptors. 33 CHAPTER 2 METHODS AND DATA 2.1 The Choice of Modelling Scheme and Testing Approaches Statistical multiple regression models have simple input requirements, but their coefficients are city-specific. Therefore, you need heat island data from the city before you can have a model to predict it. Oke's algorithmic scheme (AS) has the merit of simplicity of input requirements and it purports to be universal (Oke, 1998). The chosen modelling scheme requires only simple inputs of readily available meteorological and surface description data. This empirical scheme is developed from an understanding of the main physical processes that result in urban-rural air temperature differences. The A S is however almost entirely developed with data from mid-latitude settlements. This is simply a statement of the fact that there is a lack of availability of suitable data for cities at other latitudes. 2.2 Rationale for the Selection of Data for Testing The chosen A S uses a simple approach to model the complex urban-rural temperature difference. Scheme components incorporate the modelling of the effects of urban geometry, urban-rural thermal admittance difference, weather and time of day. This means that it should only be applied or tested in cities located in areas where these factors are likely to dominate. Application or testing in cities with significant thermal effects due to large differences in urban and rural temperature measurement site elevation, adjacent water bodies or large canopy layer anthropogenic heat input should be avoided. 35 Ideally, the A S should be tested with data that satisfies the following requirements; it should include periods of several days of continuous hourly measurements of temperature (urban and rural), wind, cloud (type & amount), and soil moisture status from all seasons. Data must be from settlements where urban and rural temperature measurement sites have the same local scale climate as noted by Lowry (1977). Three data sets are used in this research. Two are from Uppsala, Sweden and one is from Lodz, Poland. The Uppsala data was collected in two separate periods. In 1948-1949 using vehicle traverses and in 1976-1977 with fixed screen thermographs. The Lodz data is hourly fixed screen data for three years 1997, 1988 and 1999. Several papers show the Uppsala and Lodz data sets are from areas where urban and rural temperature measurement sites are located in areas of flat topography and uniform local climate (Sundborg, 1951; Smedman-Hogstrom and Hogstrom, 1973; Hogstrom et al, 1978; Taesler, 1980; Klysik, 1993; Oke, 1995; Klysik, 1996; Moberg and Bergstrom, 1997; Fortuniak and Klysik, 1998; Klysik and Fortuniak, 1999). 2.3 Physiographic and Climatic Setting of Uppsala, Sweden Uppsala, 59°52 ' N , 17°39' E , 8-20 m A S L , is located just west of the center of the 'Uppsala plain.' This is a flat, predominantly grain-farming area that stretches 29 km northwest to southeast and about 8 km west to east. The relief of the Upplandia region around Uppsala is only slightly 'rougher' than the Uppsala plain. Uppsala is centered 70 km north of Stockholm in the rain shadow of the Norwegian Mountains (Figure 2.0). It is a compact, flat, almost circular city with sharp boundaries 36 (Figures 2.1 & 2.2). The Baltic Sea is about 80 km to the east. The small Fyris River flows from the N - N W through central Uppsala to the S-SE where it enters Lake Malaren 8 km to the south of the city. There is a discontinuous esker (15 to 45 m A M S L ) along the west bank of the Fyris River in central areas of Uppsala. The most prominent building, the 119 m tall cathedral, sits atop the esker in west central Uppsala (Figure 2.3(d)). Other than the esker, topography is generally very flat and the soils contain 40-60 % clay (Jarnefors, 1956). Figure 2.0 Location map of Uppsala 0 S O 100 k m 0 SO 100 m i N ft .Klruna Norwegian Sea ) J .Tarnaby Lutea.? Umea, SundsvalU Hudiksvall • FINLAND NOW WAV a/ ;Gavle Uppsala- ' ""Aland fs. K K f> (FINLAND) .Karlstad *6retro * STOCKHOLM Norrkoping... ff Llnkdping* Jdnkopirtg t *G6teborg • Gotland Halmstad" KalmarJ Kar lshamn Otend L A T V I A Helsingborg V .Kartekrona Baltic SiMvesborg Sea D E N M A R K Q) (DTNMARK) LITH RUSSIA ( Gf RMAiM V POLAND 37 Figure 2.1 Uppsala air-photograph 1950 Scale 1:30,000 Source: (Taesler, 1980) 38 Figure 2.2 Uppsala air-photograph 1977 Scale 1:40,000 Source: (Taesler, 1980) 39 2.3.1 Normal climate of Uppsala In Koppen's climate system, Uppsala's climate is classified as a snowy-forest climate with a dry winter and a warm summer (Dwb). The 30-year normal mean annual temperature is 5.6 °C, and mean annual precipitation is 526 mm (Appendix Table A2.0). Winters are quite cold with a minimum mean monthly temperature of -4.8 °C in February. Snowfall is typically light and it occasionally forms a continuous ground cover of about twenty centimetres. July is the warmest month with a mean monthly temperature of 16.4 °C. Relatively dry summers lead to widespread grain farming that is characteristic of the Uppsala Plain (Figure 2.3 (a), (c) & (d)). Normal and actual data collection period month-by-month variations in mean monthly temperature and precipitation are given (Appendix Table A2.0). Solar altitude and azimuth for a level site in Uppsala with an unobstructed horizon is given in the sun path diagram (Figure 2.4). The actual times of sunrise, sunset and direct illumination for the urban site will differ significantly from those of an open rural site due to horizon obstruction by the buildings. Sun elevation at noon ranges from 9° on December 21, to 53° on June 21, and the number of daylight hours per day varies between 8 and 16 hours. 40 Figure 2.3 Photographs of the Marsta rural air temperature measurement site (a) Marsta thermometer screen (b) Hemispheric sky view near Marsta screen July 1998 part of Photo by Oke July 1998 Photo by Oke (c) Grain growing in a field adjacent to (d) Uppsala viewed from the northeast Marsta screen July 1998 Photo by Oke near Marsta Source: (Tukler, 1997). 41 Figure 2.4 Uppsala sun path diagram for a level site with an unobstructed horizon N 2.3.2 City morphology of Uppsala, 1948-1949 At 1949 the time of the study, Uppsala had a population of about 60,000 (Uppsala City Information, pers. com.). Central urban areas around the city-core square, the Stora Torget, had a high building density and very little vegetation (Sundborg, 1951). Buildings in the urban core around the Stora Torget were from two to five floors high (Figure 2.5). Most buildings were three to five floors high. Traversing from the city-core, to the eastern fringes of the city the ground level rises gradually from 10 to 20 m A M S L (Hogstrom et al, 1978). Rooftops remain almost level because building heights decline from 2-5 floors in the core to 1-2 floors on the city fringes. Sundborg's vehicle traverse air temperature was measured at Stora Torget where the Xs ratio, (H/W), of the site was calculated with equation 1.3: 42 Xs = 0.5 (0.25) + 0.5 (1.25) = 0.75 (2.0) Height to width ratios of the Stora Torget square and the four adjoining roads are scaled from photographs of the square and a personal communication from Taesler. The maximum expected heat island is then calculated with equation 1.2: A T u _ r ( m a x ) = 7.54 + 3.97 In (0.75) = 6.4 °C (2.1) This value for the first term of A S is very close to 6.5 °C that given in (Oke, 1981) for Uppsala [1950], based on Sundborg (1950). The 1949 city was compact, circular and had a radius of about 1.5 km. There was some ribbon development along main roads leading from the core to the north and the northeast. Rural areas around Uppsala were largely covered by grain farms such as those in Gladjen and Gronvik, 2.5 km to the southeast (Figure 2.1). Pine forest covered several square kilometres immediately to the southwest of the city (Figure 2.1). Anthropogenic heating of the urban air was not assessed in a quantitative manner. However, Sundborg suggested that "this factor should appear stronger in winter than in summer. Furthermore, no significant difference should exist between overcast and clear weather" (Sundborg, 1951). The anthropogenic heating of air was recognized but considered to be "of secondary importance in comparison with other influencing factors" (Sundborg, 1951) 2.3.3 City morphology of Uppsala, 1976-1977 In 1977, Uppsala had grown to a population of about 100,000 (Uppsala City Information, pers. com.). The expansion of the urban area between 1950 and 1977 is clearly shown by comparison of the air photographs (Figures 2.1 & 2.2). The general building 43 structure and density was little changed except for the city-center, where building heights and densities increased in a few places (Taesler, 1980), (see also Figure 2.5(b)). The urban core site R A (Radhuset) was located in an area that has changed from the 1949 situation. These changes are likely to have partly offset each other in their effect upon air temperature. Adjacent to the RA site was a parking lot created when buildings were demolished (Figure 2.6). This created an open space that reduced the effective H / W ratio and thereby the U H I potential of the site somewhat. However, some buildings within a o 75 m radius of R A were replaced by taller concrete structures. One example of this increase in building density is the replacement of a two floor high detached building with a four floor high one on the N E side of the adjacent Stora Torget square (Figures 2.5 (a) & (b)). Changes such as this tend to increase the magnitude of the heat island via an increase in H / W ratio and the resultant reduction in longwave cooling potential. Taesler's air temperature was measured in a small urban canyon behind City Hal l at Stora Torget where the A,s ratio, (H/W), of the site was calculated with equation 1.3: A s = 0.5 (1.5)+ 0.5 {0.4} =0.95 (2.2) Height to width ratios of the canyon and an adjacent parking lot were scaled from a site description in (Taesler, 1980) and a contemporary map. The maximum expected heat island for Taesler's observations then calculated with equation 1.2: AT u _ r ( m a x ) = 7.54 + 3.97 In (0.95) = 7.2 °C (2.3) This is slightly larger than the 7.0 °C given by Oke (1981) for Uppsala [1975], based on (Taesler, 1980). 44 Figure 2.5 (a) Uppsala, Stora Torget 1953 view to east (Lindberg, 1998) 45 The morphological structure of the urban core area of Uppsala was inventoried and classified in 1979 according to the method of (Ellefsen, 1991). The core area was classified as type A l urban terrain zone (UTZ) (Cionco and Ellefsen, 1998). A n A l class of U T Z conforms to an old core area of heavy clad structures (Ellefsen, 1991; Ellefsen, 1994; Cionco and Ellefsen, 1998). Unfortunately a detailed presentation of this potentially very useful classification of Uppsala via many measures of morphology such as building size, surface area, structure, orientation etc. on a hectare-by-hectare basis is not available to other than US military researchers (Ellefsen, 1994). B y the end of the 1970's, building density as a percentage of total plan area declined from 60 % in the core to about 5-10 % at the periphery (Taesler, 1980). The city had become quite compact, and grown to a radius of about 2.5 km. Most of the development in the 1950-1977 period occurred in the area to the east of the Fyris River (see Figures 2.1 & 2.2). A n oil-fired central heating plant supplied 70 % of the city's space and water heating requirements in the 1970's (Taesler, 1980). Heating and hot water for the core areas was entirely supplied by this plant. At an extreme low outdoor air temperature of -20 °C, the total maximum anthropogenic heat emission ranged from 125-150 W m"2 in the core to < 10 W m"2 at the largely low-density residential periphery (Hogstrom et al, 1978). Winter temperatures rarely dip below -6 °C; therefore, the heat emissions given above are maximum 'design temperature' emissions far above those likely during all of Taesler's data collection periods. 46 Figure 2.6 Uppsala parking lot adjacent to the RA (Radhuset) site created by the removal of small buildings at the intersection of Vaksalagatan and Dragarbrunnsgatan in the core of Uppsala, -70 m northeast of Stora Torget. The photo gives the view across the parking lot and Vaksalagatan to the southeast. Source: Taesler 47 2.4 The Uppsala Data Two data sets from Uppsala were used in this study. One set from the 1940's was collected by automobile traverses and the second from the 1970's was based on fixed thermographs mounted in standard thermometer screens at urban and rural sites (Sundborg, 1951; Taesler, 1981). Both data sets were collected with great attention to detail. The choice of collection sites and the instrumentation used to measure air temperature were of the highest quality. 2.4.1 The 1948-1949 traverse data The first set of temperature measurements, from the late 1940's, consists of 207 automobile traverses (Sundborg, 1951). Data was collected in an almost random temporal basis in all types of weather, and at all times of day and year (Table 2.0). Sundborg went to great lengths to select sites along a traverse route that were the most representative of urban and rural conditions (Sundborg, 1951) (Figure 2.7). This was achieved by statistical analysis of preliminary temperature data collected with a resistance thermometer mounted on an automobile (Figure 2.8). The final long traverse route was chosen after analysis of data for 79 temperature measurement points from twenty preliminary 48 km vehicle traverses at different times and in different weather conditions. Sundborg used the standard deviation of the random component of the difference between the temperature value at a given point and the average for a selected area around the point (O"ST) as his measure of representativity. The smaller the value of Ogr the more representative the station is of the area (Figure 2.7). 48 A final short traverse route was chosen based on the foregoing analysis (Figure 2.7). Traverse speed averaged about 9 m s"1 along this 9.6 km route. The temperatures measured at 15 points along this short route were corrected for temperature change over Figure 2.7 Uppsala 1946 map, showing the areal urban/rural temperature representativity of the air temperature measurement sites. The blackened line in the centre and the southeast quadrant indicates the temperature measurement short route used. Representativity of air temperature measurements at the given sites is indicated via the scale below in (°C). The smaller the dot the more thermally representative site is of the urban or rural mean temperature. Contours are at 5 m intervals. Urban temperature measurement sites: •Stora Torget •Bangardsgatan Rural temperature measurement sites: (Gladjen •Gronvik 0.21°—0.30°: • 0 .31°—0.40°: • 0 .41°—0.50°: • 0 .51°—0.60°: • 0 .61°—0.70°: • 0 .71°—0.80°: • 0 .81°—0.90°: Source: (Sundborg, 1951) 49 Table 2.0 Date and number of data collection traverses per day during the Uppsala May 1948 - May 1949 data collection period Day May Jun Jul Aug Sep Oct Nov Dec Jan Feb Mar Apr May 1 2 3 4 5 6 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 1 2 1 2 1 t 2 2 1 1 I 1 -2. l l 23| 1 2 1 2 2 1 2 1 1 1 1 2 1 2 2 2 I 2 1 1 2 1 ' 2 1 1 3 2 2 1 • 2 H i 1 1 1 2 1 2  2 1 . Ill 1 2 3 1 2 1 1 i 1 i ! 2 12 22 33 R H P l 2 L k 6 1 3 I i i 3 4 ' - i 9! sjfj : 3 16 34 10 1 1 3 2 1 1 3 10 3 •» 1 24 4 1 1 2 2 10 50 Figure 2.8 Sundborg's traverse thermometer setup (a) Sketch of the construction of the resistance thermometer. A ) Coiled copper tube B) Electric wires to the Wheatstone Bridge C) Inner radiation shield (b) The resistance thermometer mounted 15 cm from automobile roof at a height of 1.5 m. 51 the -18 minutes the traverse took to complete. This was achieved by revisiting the same site (Stora Torget) three times during the traverse so that a suitable correction value could be determined from the measured cooling/warming rate. Traverses also made use of temperatures measured during the traverses at the Uppsala Meteorological Institute via thermograph charts. Charts allowed a comparison of automobile and thermograph measurements and a further analysis of temperature change patterns during the measurement traverse. The air temperature for the urban site was the arithmetic average of four measurements, three in the urban core square (Stora Torget) and one at another core site (Bangardsgatan) three blocks to the southeast. The corresponding rural temperature was the arithmetic average of two measurements: one at Gladjen and one at Gronvik, -2.25 and -2.75 km southeast of the city core, respectively. A i r temperatures measured at these urban and rural sites were found to be representative of their respective areas (Figure 2.7). A l l air temperatures were measured with a resistance thermometer mounted at a height of about 1.5 m on an automobile (Figure 2.8). A double radiation shield and the 6-17 m s"1 speed range of the automobile ensured that the shielding and airflow kept the instrument free of radiation induced measurement errors. This arrangement was assessed to measure air temperature to within ± 0 . 1 °C of the actual value (Sundborg, 1951). 52 2.4.2 The 1976-1977 hourly fixed screen thermograph data Temperature and other data used in this study were collected by Roger Taesler while participating in The Uppsala Urban Meteorological Project (Hogstrom et al, 1978). Taesler's air temperature data was collected by recording thermographs at several sites throughout Uppsala (Taesler, 1980). It is a collection of 952 semi-continuous hours of air-temperature measurements from thermometer screens at fixed locations. There were ten data collection periods of 2 to 16 days in the 1976-1977 data set (Table 2.1). These periods were selected to include weather conditions conducive to heat island formation (Taesler, 1981). Only air temperature data from the urban core site Radhuset (RA) and the rural site Marsta ( M A ) 8 km to the north-northwest is used here (Figure 2.9). Careful calibration and comparison procedures ensured that errors in the urban-rural temperature-difference (RA - M A ) were kept to within ± 0.2°C (Taesler, 1980). The urban core site RA was located in a courtyard at 59°51 '31" N , 17°38'35" E , and 10 m A M S L . This courtyard was surrounded by 3-4 story buildings in all directions except towards the east, where it is separated from a parking lot by a wooden hoarding (Taesler, 1980). The RA site was a narrow gravel-covered canyon with a H A V ratio of about 1.5. RA was -60 m northeast of where Sundborg had recorded urban core air temperatures in Stora Torget Square (Figure 2.5). 53 Table 2.1 Date and number of data collection hours per day at R A during the Uppsala January 1976 - February 1977 data collection period day Jan Feb Mar Apr May Jun Jul A u g Sep Oct Nov Dec Jan Feb sum 1 24 j 24 2 4 ' ' d i k J 12 3 12 1 12 4 0 5 0 6 0 7 - 2 8 22 T l 25 9 24 24 10 58 10 24 24 24 72 11 24 24 14 62 12 24 10 34 13 24 22 46 14 24 -24 1 48 15 24 24 12 60 16 24 24 23 71 17 24 ' 22 46 18 24, j 24 48 19 j 12 j 24 36 20 13 7 20 21 23 2 4 : 47 22 ">4 24 48 23 7 .12 24 43 24 20 ^12,J 42 25 7 19 26 0 27 5,-" > 5 28 24 24 29 24 24 30 24 24 31 0 2 34 0 22 276 0 337 0 0 168 32 0 0 0 83 952 54 Figure 2.9 Map of Taesler's 1976-1977 Uppsala data collection sites Scale: 1:75,000 M A (Marsta) rural temperature measurement site ^16, airfield hourly meteorological data Site. T Z I X ™ ~ ( c l ° u d height and amount, wind and precipitation) . R A (Radhuset) urban temperature *» measurement site Source: (Lantmateriet i Uppsala, 1992) 55 The rural site Marsta ( M A ) was located at 59°53 '31" N , 17°36'00" E and 20 m A M S L (Figure 2.9). Air-temperature data measured here was from the permanent Uppsala University meteorological site located 8 km north-northwest of R A , in flat fields of grass and grain crops (Figure 2.3). Hourly wind speed, cloud height, cloud type, and precipitation data are available from an airfield meteorological site, called F 16, at 59°55'41" N , 17°35'21" E , and 20 m A M S L , approximately mid-way between R A and M A (Figure 2.9). This data was used to calculate the weather term of the Algorithmic Scheme. Precipitation amount and snow depth data from F16 was used in the estimation of rural surface thermal admittance. 2.5 Physiographic and Climatic Setting of Lodz, Poland Lodz located at 51°46 ' N . , 19°27' E . , 200 m A M S L (Figure 2.10), is the second largest Polish settlement. Lodz's population was 825,000 in 1995 (United Nations, 2000). It is located 120 km southwest of Warsaw. Flat farmland with some trees surrounds the city. Within the urban area, land elevation varies gradually. No more than a 55 m range of elevation is found within the urbanized (79 km 2 ) part of the administrative area (224 km 2 ) . The region is devoid of water bodies. The climate of Poland is dominated by the effects of continental air masses from the east, and maritime air masses from the west. This meeting of distinctly different air masses over Poland results in weather that is very changeable. Great weather differences are frequently experienced from hour-to-hour, month-to-month and from year-to-year (Appendix Table A2.1). 56 Figure 2.10 Lodz map Poland Source: (Klysik, 1999) 57 In Koppen's climate system, the climate of Lodz is classified as a warm temperate climate without a dry season, and with a warm summer (Cfb). The 30-year normal mean annual temperature is 8.0 °C, and mean annual precipitation is 564 mm (Appendix Table A2.1). Winters are quite cold with a minimum mean monthly temperature of -4.8 °C in January. Snowfall is typically light but it occasionally forms a continuous ground cover of 10-20 cm. July is the warmest month with a mean monthly normal temperature of 18.9 °C. Rainfall is highest during the summertime. This rainfall distribution results in conditions where grass and other crops grow and ripen in the fall, when soils are relatively dry. Normal month-by-month variations in mean monthly temperature and precipitation are given in the Appendix Table A2.1 . Corresponding data for the data collection period is also given in the Appendix Table A2.1 . Solar altitude and azimuth for a level site with an unobstructed horizon at Lodz are given (Figure 2.11). The actual times of sunrise, sunset and direct illumination for the urban site wil l differ significantly from those of an open rural site due to shading by the urban structures. Mean monthly wind speeds in Lodz are typically in the range 2.5-5.0 m s"1 (the overall annual mean for Lodz 1997-1999 was 3.3 m s"1). Mean nocturnal wind speeds (-3.0 m s"1) are about 90 % of the daytime mean value. 58 Figure 211 Lodz sun path diagram for a level site with an unobstructed horizon N S 2.5.1 Morphology of the urban site Urban air temperature was measured at a site known as 'Lipowa' (LIP) located at the southwest corner of Sklodowskiej-Curie and Lipowa streets at 51°45'45" N . , 19°26'30" E. , 205 m A M S L (Figure 2.12).. This site is in a somewhat open, partially vegetated urban canyon with mean building height of 10.5 m (Fortuniak, pers. com.). Urban canyon morphology in terms of percentage of plan area is: 22 % roofs, 40 % asphalt and concrete paving and 38 % grass and trees. A i r temperature and humidity are measured by a Vaisala H M P 35 temperature and humidity probe located at a height of about 2 m in a Stevenson 59 screen (Figure 2.12). The central sky view factor of the Lipowa canyon is 0.66 from May-October, and 0.85 from November-April, due to seasonal changes in the foliage of the deciduous trees in the canyon (Fortuniak, pers. com.). Anthropogenic heat emissions play a significant role in heat island formation in some core areas of Lodz during exceptional anticyclonic synoptic conditions in winter months (Klysik, 1999). At this time of year emissions range from about 8 W m"2 in suburban residential areas on the periphery to -70 W m"2 in the mixed industrial and high-density residential urban core (Klysik, 1996). The same author considers the heat island magnitude to be significantly affected by anthropogenic heat emissions under clear sky and light wind conditions during winter. In these conditions heat islands of 8 °C are possible (Klysik, 1996). A n extreme, 12 °C, heat island was recorded in February 1996 under highly favourable synoptic conditions with a blanket of insulating snow covering rural areas (Klysik and Fortuniak, 1999). This was a rare transient condition brought about by frontal conditions that strongly favoured heat island formation (Klysik and Fortuniak, 1999). The urban site, Lipowa, is located in an area with much lower building density than the city core area that has the highest anthropogenic heat input. Based on the relative building floor area to ground area, the anthropogenic heat is estimated to average 25 W m"2 at Lipowa during the coldest winter period. Much of the heat loss wil l be through poorly insulated roofs into the urban boundary layer. Klysik wrote in an earlier paper "In Lodz the difference between monthly average temperature of the city and its surroundings is 0.5 °C in winter and 0.8-1.0 °C in summer" (Klysik, 1996). There is little or no evidence that the air 60 temperatures measured at Lipowa are affected by anthropogenic emissions (Klysik and Fortuniak, 1999). Figure 2.12 (a) The urban air temperature measurement screen site at Lipowa, Lodz (b) Hemispheric sky view photograph of the site 2.5.2 Morphology of the rural site Rural air temperature was measured at Lublinek airport, 3.5 km to the southwest of the Lipowa site, at 51°43'25" N . , 19°24'23" E . , 187 m A M S L . A Vaisala H M P 35 air temperature and humidity probe measured temperature at a height of about 2 m height in a Stevenson screen within a standard meteorological station grass-covered compound (Figure 2.13) (Fortuniak and Klysik, 1998). The Stevenson screen at Lublinek had a sky view factor of 0.998 with a trivial 0.002 occupied by the terminal building (Klysik and Fortuniak, 1999). The closest urbanized land is an area of low-density housing 0.6 km to the east where the percentage of total plan area occupied by rooftops is about 6 %. Higher density urban areas are located at more than 1.5 km to the northeast. No urban effects upon air temperature are found in the airport weather station record (Klysik and Fortuniak, 1999). The city is most likely to influence air temperatures at the airport when the wind is from the northeast, but such directions occur only 6 % of the time in an average year (Klysik and Fortuniak, 1999). 2.6 The Lodz Data The Lodz data is a temporally continuous set of 26,280 hourly values for the complete 1997-1999 period. From January 1997 to December 1999, air temperature was measured in the urban site at Lipowa and at Lublinek airfield in the adjacent rural area (Figure 2.10). A i r temperatures measured at Lipowa were used in all heat island calculations except for a few hours where gaps in data were filled with temperatures measured at the nearby meteorological station, M S M (Figure 2.10). At all measurement 62 sites Vaisala temperature and humidity probes were used in conjunction with Campbell Scientific data loggers to record air temperature and humidity in standard screens at a height of about 2 m. July 1997 was exceptionally wet with 269 mm of precipitation, this is 4.3 standard deviations above the normal monthly mean of 84 mm, and is the wettest month on record for Lodz. This may make the summer 1997 data good as a test of the A S ' s ability to model cloud and thermal admittance effects upon the heat island magnitude. 63 Figure 2.13 Photographs of the rural air temperature measurement site at Lublinek, Lodz (a) Vaisala FfMP temperature and humidity probe (just left of center) in the thermometer screen (b) Lodz, Lublinek thermometer screen enclosure with airport terminal building in the background CHAPTER 3 TESTING THE AS WITH THE UPPSALA DATA SETS Algorithmic Scheme (AS) performance is initially assessed with the temporally discontinuous individual vehicle traverse (1940's) data, and then with semi-continuous fixed-screen (1970's) data. Pearson's (r2) and Willmott 's (d) statistical measures of observed vs. modelled temperature difference (heat island magnitude), Root Mean Squared Error with systematic and unsystematic components ( R M S E , R M S E S , and R M S E U ) and the Mean Absolute Error ( M A E ) are used to assess the accuracy of the scheme. Of these measures, Willmott's index of agreement (d) is suggested to be superior to the (r2) measure. However, some evaluations are done with just the (r2) statistic because these are available for the Chandler (1965) and Sundborg (1951) models and hence can be used in comparisons. A detailed explanation of this (d) statistic and a general discussion of model validation are contained in Willmott (1981). Assessments are based upon vehicle traverse and fixed screen data to show the performance of the scheme using the very different types of data (discontinuous, near random, individual traverses and semi-continuous screen records). Statistical measures of scheme performance give an accurate overall measure of the scheme's ability to replicate the heat island pattern (Tables 3.0 & 3.1). Hour-to-hour scheme performance with the continuous data is discussed with the aid of straightforward visual comparisons of time series plots of observed vs. scheme temperature differences. This permits analysis of the performance of the scheme on an hour-hour timescale. Many interesting short-term, hour-hour, nocturnal, diurnal, and sunrise-sunset phenomena are available for discussion. Table 3.0 Uppsala 1948-1949 Algorithmic Scheme (AS) statistical evaluation (a) Observed (Obs) and AS UHI values for Sundborg's 207 (all, i.e. day & night) traverses Obs Mean AS Mean Obs StDev AS StDev MAE RMSE RMSES RMSEU r r2 d (°C) (°C) (°C) (°Q (°Q (°C) (°Q (°C) 1.23 0.97 1.10 1.07 0.65 0.85 0.42 0.74 0.72 0.52 0.83 (b) Observed (Obs) and AS UHI values for Sundborg's 96 (daytime) traverses Obs Mean AS Mean Obs StDev AS StDev MAE RMSE RMSES RMSEU r r2 d (°C) (°C) (°C) (°C) (°C) (°Q (°C) (°C) 0.69 0.18 0.57 0.30 0.56 0. 70 0.65 0.25 0.53 0.28 0.57 (c) Observed (Obs) and AS UHI values for Sundborg's 111 (nighttime) traverses Obs Mean AS Mean Obs StDev AS StDev MAE RMSE RMSES RMSEU r r2 d (°C) (°C) (°C) (°C) (°C) (°C) (°C) (°C) 1.71 1.65 1.22 1.02 0.71 0.96 0.56 0.78 0.64 0.41 0.79 66 Table 3.1 Uppsala 1 9 7 6 - 1 9 7 7 Algorithmic Scheme (AS) statistical evaluation (a) & (d) all times of day, (b) & (e) daytime, (c) & (f) nighttime; (d), (e) & (f) give statistics for the mean monthly values based on variable data set sizes, given as ' n ' ; Collection dates (hours/day) for this data was presented earlier (Table 2 . 1 ) (a) Observed and AS UHI values for 952 hours (all times of day) Obs Mean AS Mean Obs StDev AS StDev M A E R M S E R M S E S R M S E U r r 2 d (°C) (°C) (°C) (°C) (°C) (°C) (°C) (°C) 1.69 1.08 1.50 1.28 1.00 1.25 0.86 0.63 0.70 0.50 0.78 (b) Observed and AS UHI values for 559 hours (daytime) Obs Mean AS Mean Obs StDev AS StDev M A E R M S E R M S E S R M S E U r r 2 d (°C) (°C) (°C) (°C) (°C) (°C) C O C O 1.09 0.39 1.18 0.56 1.03 1.30 1.19 0.48 0.39 0.15 0.38 (c) Observed and AS UHI values for 393 hours (nighttime) Obs Mean AS Mean Obs StDev AS StDev M A E R M S E R M S E S R M S E U r r 2 d (°C) (°C) (°C) (°C) (°Q (°C) (°C) C O 2.54 2.06 1.51 1.37 0.97 1.18 0.70 0.56 0.73 0.53 0.81 (d) Observed and AS mean heat island (°C), 9 5 2 hours (all times of day) Jan-1976 Mar-1976 Apr-1976 Jun-1976 Sep-1976 Oct-1976 Feb-1977 AToBS 0.9 1.9 1.3 1.7 2.2 2.9 2.0 A T A S 0.7 0.7 0.8 0.9 2.0 2.3 0.8 n 34 22 278 337 166 32 83 (e) Observed and AS mean heat island (°C), 5 5 9 hours (daytime) Jan-1976 Mar-1976 Apr-1976 Jun-1976 Sep-1976 Oct-1976 Feb-1977 A T O B S 0.6 1.9 1.0 1.3 0.5 0.8 1.9 A T A S 0.3 0.4 0.3 0.4 0.5 0.4 0.3 n 10 13 165 253 76 12 30 (f) Observed and AS mean heat island (°C), 393 hours (nighttime) Jan-1976 Mar-1976 Apr-1976 Jun-1976 Sep-1976 Oct-1976 Feb-1977 AToBS 1.0 2.0 1.7 3.0 3.6 4.1 2.0 A T A S 0.8 1.1 1.6 2.1 3.2 3.4 1.1 n 24 9 113 84 90 20 53 6 7 3.1 Testing the AS with the 1948-1949 Traverse Data When testing the A S with the 1948-1949 traverse data, the energy balance-water balance approach to calculate the thermal admittance factor, (O m ) , could not be applied due to the lack of required continuous precipitation data. Instead, an estimated water balance method is used. The thermal admittance values (jxr) for the rural surface of Uppsala used to calculate <E>m are chosen from a selection table of thermal admittance values for four common soil types at several moisture contents and for snow cover (Table 1.1). Values selected are for the clay soil (30 % pore space) in the rural area, based on information found in the soils literature, and precipitation is estimated according to the normal climate (Taesler, 1972). The rate of evaporation was based on Campbell and Norman (1998, Table 9.2, p 138). The accuracy of the chosen (|xr) values is probably not high due to the discontinuous precipitation data. 3.1.1 Daytime performance of the AS The A S is tested with ninety-six measured heat island values collected during daytime in a near random fashion from the 1940's Uppsala set of traverse data. The A S retrodicts heat island magnitude poorly in the mean with r = 0.28 for the 96 traverses. This performance is better than Sundborg's regression model's performance, where r 2 = 0.24. In addition, Sundborg's model requires empirical regression coefficients based on prior observations of urban-rural temperature difference, whereas the A S requires only sky view factor, soil type and meteorological observations. Root mean square error ( R M S E ) statistics show that mean systematic error (0.65 °C) is much larger than the unsystematic (0.25 °C). On average, the scheme under-predicts the mean observed 68 0.69 °C daytime heat island by 0.51 °C, hence the proportionately large R M S E systematic component of the overall R M S E error. Performance is likely degraded due to the inaccuracy of the estimation of the rural thermal admittance and of the average urban sky view factor (H/W = 0.75) for the points at which the urban temperatures were measured. Temperature data collected by automobile are averaged in this case, for four points, three at Stora Torget and one at Bangardsgatan. Measurements are made a few minutes apart, and adjusted for temperature change with time, to generate an overall mean value for an intermediate time. Individual temperature measurements are affected by slight variations in weather and measurement point (vehicle position) variation over the 20-30 minute traverse period. The absence of a means of accurately determining sky view factors via hemispheric sky photos has introduced a degree of uncertainty about the correct value for the maximum observable heat island, the first term of the scheme. Weather conditions are measured hourly and interpolated linearly to generate the 'observed weather' for traverse times that rarely coincide with weather observation times. This leads to inaccuracies in the wind and cloud inputs for the weather term <J>W. 3.1.2 Nighttime performance of the AS Nocturnal performance, r 2 = 0.41, is better than that in the daytime. This is as expected given that Sundborg's own statistical model performs better for nocturnal than daytime conditions: night r 2 = 0.44, daytime r 2 = 0.24 (Sundborg 1951). It is likely that errors in the estimation of the sky view factor wi l l have a more serious effect on nocturnal results (see section 1.1.2). The scheme appears to predict the small number of large heat island events quite well, but performance with average heat islands in the 1.0-1.7 °C range is poor (Figure 3.0). Only 111 nocturnal heat islands were sampled by Sundborg. They include examples from all times of night and cover a wide range of weather conditions during the 367 day period. This makes it difficult to analyze the mean situation in order to relate observed phenomena to sunrise, sunset or any other repetitive factor in a statistically meaningful manner. Sundborg developed two simple statistical expressions for urban-rural temperature differences. One a full multiple regression form with several variables, which under-estimated 2-6 °C heat islands, by 2-3 °C. A second simpler expression, for nocturnal conditions, is based on his understanding of the key role of cloud and wind in nocturnal cooling: D = 4.6-0.28 N / U (3.0) where D is urban-rural air temperature difference (°C), N is suburban cloud cover fraction on a scale from 1 to 10 and U is the 10 m suburban wind speed in m s"1. When applied to the 111 nocturnal traverses the fit to equation 3.0 gives an r 2 = 0.32. This level of performance is mid-way between the day and night performance of Sundborg's full statistical model. The simple expression was largely free of bias in its ability to replicate heat islands of any particular magnitude. Clearly, this performance (r 2 = 0.32) is less than desired. Sundborg's report suggests it may be preferable for the wind term to be a non-linear function (Sundborg, 1951). A non-linear wind term is incorporated in the Oke (1998) A S <I>W term. 70 Figure 3.0 Uppsala, 1948-1949 Observed vs. A S nighttime & daytime UHIs Diagonal line is the 1:1 line O T3 C co CO a> X (fi < o -4 c b - o o - p o — o • Nightt ime o Dayt ime Observed Heat Island (°C) Nighttime (n = 111): r 2 =0.41 d = 0.79 Obs. Mean = 1.71 (°C) A S Mean = 1.65 (°C) M A E = 0.71 (°C) M A E / Obs. Mean = 0.42 Daytime (n = 96): r 2 =0.28 d = 0.57 Obs. Mean = 0.69 (°C) A S Mean = 0.18 (°C) M A E = 0.56 (°C) M A E / O b s . Mean = 0.81 3.1.3 Daytime-nighttime variability of scheme performance The A S Mean Absolute Errors ( M A E ) for daytime and nighttime were 0.56 °C and 0.71 °C, respectively. The M A E as a proportion of the mean observed heat island magnitude is 81 % of the 0.69 °C mean daytime, and 42 % of the 1.70 °C mean nighttime, heat island. The A S performs better than Sundborg's full statistical model in the daytime (r 2 = 0.28 vs. 0.24) but slightly poorer at night (r 2 = 0.41 vs. 0.44). Sundborg's full statistical model cannot be reproduced here because his report does not provide sufficient detail concerning derivation of the temperature input. His temperature input values were, "partly obtained by the investigator's own observations and partly from recorded data taken at the Meteorological Institute" (p.80, Sundborg, 1951). The 71 non-linear wind term in the A S weather expression is probably responsible for most of the improved performance, especially in the daytime when winds are normally stronger. A n overall impression of the strengths and weaknesses of the scheme can be obtained by studying the scatter plot of observed vs. A S heat island magnitude (Figure 3.0). Many of the smaller (mostly daytime) U H I values of less than 1.5 °C are under-estimated by the scheme. A tendency to over-estimate middling UHIs by about 1.0-2.5 °C is also evident. Several of the larger nighttime, 3-5 °C, events are relatively accurately estimated. These tests are based on data collected in a near random fashion in all weather conditions and times of day and season. This allows an assessment of general scheme performance but does not permit the assessment of the scheme's ability to replicate continuous hourly data. 3.2 Testing the AS with the 1976-77 Fixed Screen Data The A S is now tested with continuous hourly data collected in ten short periods, of about two to sixteen days, in the period from 23 January 1976 to 16 February 1977. Approximately 60 % of the available 952 fixed screen hourly measurements (559) were collected in the daytime (Table 3.1b). Rural surface thermal admittance is estimated by using the energy balance-water balance calculation method described in chapter 1. Rainfall was exceptionally low during this period with only a total of 23 mm measured on the 52 data collection days. Water retentive properties of the clay soil result in gradual changes of 72 soil moisture and thermal properties, and this in turn gives a slowly varying thermal admittance term, O m , in the A S scheme. General A S strengths and weaknesses are shown in the scatter plot of observed vs. A S heat island magnitude (Figure 3.1). In general, daytime heat island values are poorly estimated by the scheme and nighttime heat islands are much better replicated; but many of the larger nighttime 4-7 °C UHIs are under-estimated. Figures 3.2 to 3.5 illustrate the hour-to-hour and seasonal relations between the observed U H I and the A S retrodicted U H I for periods where several days of continuous data are available. The chosen sequences demonstrate the typical strengths and weaknesses of the A S when applied to this specific data set. Small gaps in graphed lines on some of the charts are due to data gaps. Figure 3.1 Uppsala, 1976-1977 Observed vs. A S nighttime & daytime UHIs Diagonal line is the 1:1 line Nighttime (n = 393): r 2 =0.53 d = 0.81 Obs. Mean = 2.54 (°C) A S Mean = 2.06 (°C) M A E = 0.97 (°C) M A E / Obs. Mean = 0.38 Daytime (n = 559): r 2 = 0.15 d = 0.38 Obs. Mean = 1.09 (°C) A S Mean = 0.39 (°C) M A E = 1.03 (°C) M A E / O b s . Mean = 0.94 73 3.2.1 Observed and AS UHI from February 15th to 17th, 1977 (Figure 3.2) A large 5-6 degree heat island on the afternoon of the 15 t h occurs at a time when seven octas of cloud cover effectively excluded the small potential solar input during midday.. This U H I centered in the early afternoon is exceptionally large for the conditions and is not well replicated. This is clearly an anomalous condition possibly brought about by a frontal passage at a time when a blanket of snow insulated the ground outside the city. Rural temperatures remained in the -9 to -7 °C range throughout, whereas the urban temperature warmed from -6 to -2 °C when the wind dropped from 3 to 0 m s"1 in the early afternoon. In the following 18 hours, cold northerly winds, overcast skies and snow cover appear to be the dominant conditions. Winds are 3-6 m s"1 with 6-8 octas of low cloud. This period shows an almost constant 1.5 degree heat island that is also not well replicated by the scheme. Figure 3.2 Uppsala, 15-17 February 1977 UHI (°C) 74 3.2.2 Observed and AS UHI from April 8th to 14th, 1976 (Figure 3.3) Figure 3.3 shows the variation of the measured and A S predicted hourly UFTJ from Apr i l 8 t h to 14 t h . Some of the sudden observed heat islands, such as the 3.2 °C one on the 9 t h , are probably due to local effects. This particular spike is possibly due to the 7 m s"1 S W wind circulating air warmed by the sun at an adjacent parking lot through the screen. The A S is not able to duplicate it well. Another observed heat island of both greater magnitude (5 °C) and duration (~4 hours) centered at 14:00 on the 13 t h, illustrates what is probably the greatest weakness of the A S . Since the O t scheme term always reduces the maximum possible A S value to zero at noon, it is unable to reproduce any observed heat islands at midday. Figure 3.3 Uppsala, 8-14 April 1976 UHI(°C) 75 3.2.3 Observed and AS UHI from June 9th to 12 "', 1976 (Figure 3.4) Conditions on the nights of June 8-9* and 9-10 t h are nearly cloud and wind free and therefore favour U H I development. The A S under-predicts the peak values of the two large heat islands by about 30%. The observed midday heat island values of 2 °C on the 9 t h and 10 t h, under clear skies with light winds, are not replicated by the scheme. On the night of 11-12*, the cloud height changes from low (cloud base height < 2,000 m) to high (cloud base height > 6,000 m) and back to low. The A S performs poorly. It illustrates the sensitivity of the scheme to sudden cloud changes, especially at the middle of the night. However, it is not known if the cloud observed at the rural site also applies to the urban one. The daytime heat island effect may be due to heat from vehicles using the parking lot adjacent to the urban site. Figure 3.4 Uppsala, 9-12 June 1976 UHI(°C) 7 5 2 4 3 4 6 0 1 4 Observed UHI • — --•AS UHI 6/9/76 0 : 0 0 6/10/76 0 : 0 0 6/11/76 0 : 0 0 6/12/76 0 : 0 0 -1 76 3.2.4 Observed and AS UHI from September 28th to October 2nd, 1976 (Figure 3.5) Between September 28 t h and October 2 n d , 1976, the correspondence between the observed and A S UHI ' s is generally good, especially in the middle of the night. However, there is an inability to generate negative values in the daytime. On the night of the 28-29 t h, the A S over-predicts by about 1.5 °C. Again, it shows that A S responses to wind and cloud variations, (O w ) , are too sensitive to sudden weather changes, especially at night, when the <E>t term constrains the scheme's maximum possible U H I magnitude far less. Figure 3.5 Uppsala, 28 September-2 Oct. 1976 UHI(°C) 77 3.2.5 Results of the test with the 1976-1977 Uppsala data Overall statistical measures of the A S performance when retrodicting the daytime heat island magnitude are given for the 1970's data set (Table 3.1). In Table 3.1, many statistical measures of the relationship between measured and modelled heat island magnitudes are presented. A separate table (Table 2.1) is also included to show the highly variable number of hours of data for each month. The A S predicts very poorly in the mean with r 2 = 0.15 for the 559 daytime hours. Sundborg's own simple expression (equation 3.0) performs equally poorly (r = 0.15) when applied to the same data. A great month-by-month variability in the ability of the scheme to retrodict the observed temperature difference is found (Table 3.1 (e)). This observation is based largely on the meaningful number of data points available for only three months, A p r i l , June and September. The mean monthly error ranges from an under-prediction of 0.8 °C in June, for day and nighttime hours together, to an accurate prediction of the 0.5 °C daytime heat island in September. Again, daytime performance is not good, in part due to the inaccuracy of the estimation of the sky view factor for the canyon in which the urban temperatures were measured. The sky view factor is estimated from a scale map and the description of the site in the original report. The absence of a hemispheric sky photo has introduced a degree of uncertainty about the correct value for the maximum observable heat island (the first term of the AS) . The urban site is also a car park with considerable activity nearby. A factor was developed to correct for the mean observed error at each hour of the day for each month. However, given the relatively sparse data for some hours, of most 78 days, in all but A p r i l , June and September, these corrections became little more than an exercise in analysis of diurnal effects in the three months with the most data. Site-specific corrections for particular sites can easily be developed, but pursuit of this approach would have changed the A S into a more statistical and site-specific scheme. The Mean Absolute Error for the 559 daytime hours is 1.03 °C. The M A E as a proportion of the mean observed 1.09 °C daytime heat island magnitude is quite high at 94%. The A S is further assessed for the 393 hours of semi-continuous fixed screen Uppsala data gathered at night (Table 3.1(c)). The A S predicts these nocturnal observations moderately well in the mean, with r 2 = 0.53. This is much better than in the daytime performance. Sundborg's own simple expression (equation 3.0) for nocturnal hours, r 2 = 0.51, performs not quite as well as the A S . Again month-by-month variability in the ability of the scheme to retrodict the observed temperature difference is large (Table 3.1 (f))- The mean monthly error ranges from an under-prediction of 0.9 °C in June (n = 84 hours) to an under-prediction of 0.1 °C in Apr i l (n = 113 hours), for nighttime hours on a mean monthly basis. The Mean Absolute Error for nighttime is 0.97 °C. M A E as a proportion of the mean observed heat island magnitude is a low 38 % at night. This observation is consistent with the results of another statistical heat island model (Chandler, 1965) that shows similar difference in daytime and nighttime accuracy when applied to London, U K . There is a general tendency for the A S to slightly overreact to changes in cloud and wind. It is more obvious at night when the <E»t term constrains possible U H I magnitude far less. 3.3 Summary of Tests Testing the A S with the discontinuous 1940's data is of limited value for a scheme that is to be used to give hourly continuous values, but it does indicate a moderately good overall correlation, 0.72, between observed and A S U H I . The 1970's data consists of more hours and is from several short periods of weather conducive to heat island formation. Testing the A S with such data permitted the analysis of hour-to-hour heat island variation, and the strengths and weaknesses of the A S became apparent. Large 3-4 degree nighttime heat islands, not associated with frontal passage, were generally well predicted; but heat islands that are larger than 4 degrees are often under-predicted by the scheme. The A S has a tendency to overreact to rapid weather changes. It is also unable to retrodict cool islands and midday heat islands. Overall, the scheme performs far better at night than in the daytime with the 1970's hourly data, where O m can be calculated in an objective manner. 80 CHAPTER 4 TESTING THE AS WITH THE LODZ DATA SET After testing the A S with the Uppsala data, the scheme is again statistically assessed for the continuous hourly January 1, 1997 to December 31, 1999 Lodz data set (Tables 4.0, 4.1 & 4.2). The energy balance-water balance approach to rural surface thermal admittance estimation was used to calculate the thermal admittance reduction factor <J>m. The overall quality of the A S is shown in three scatter plots of observed vs. A S heat island magnitude for the three years (Figures 4.0, 4.1 and 4.2). In these three figures, data points are centred roughly on the 1:1 line. Daytime heat and cool islands are not well replicated, partly due to the damping effect of the O t term. The hourly variations of the observed and A S heat islands of selected periods are graphed and presented with discussions. The performance of the A S in these examples demonstrates the scheme's typical strengths and weaknesses in various weather conditions and with various soil moisture contents when applied to this data set. Days with abnormal weather or interesting phenomena are also included to show how effectively the A S can handle extreme cases. Because the weather for all three years was quite similar and close to normal except for July 1997, appropriate examples may be drawn from any of the three years and the chosen sequences are presented by the order of the day of the months, and not by the year. 81 Table 4.0 Lodz 1997 Algorithmic Scheme (AS) statistical evaluation (a) all, i.e. day & night hours, (b) daytime hours, (c) nighttime hours (a) Observed & A S (all 8,760 hours, i.e. day & night) monthly mean heat island (°C) Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Ann. mean A T 0 B S 0.7 0.5 0.8 0.6 0.6 0.8 0.8 1.0 0.6 0.3 0.2 0.3 0.59 A T A S 0.8 0.6 0.9 0.6 0.7 0.9 0.5 0.8 0.8 0.5 0.3 0.2 0.63 error -0.1 -0.1 -0.1 0.0 -0.1 -0.1 0.3 0.2 -0.2 -0.2 -0.1 0.1 -0.04 n (hours) 744 672 744 720 744 720 744 744 720 744 720 744 8760 Observed and A S values for all 8,760 hours Obs Mean AS Mean Obs StDev AS StDev M A E R M S E R M S E S R M S E U r r 2 d (°C) (°C) (°C) (°C) (°C) (°C) (°C) (°C) 0.59 0.63 0.90 0.83 0.50 0.68 0.33 0.30 0.69 0.48 0.82 (b) Observed & AS (4,380 daytime hours) monthly mean heat island (°C) Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Ann. mean A TOBS 0.4 0.2 0.2 0.3 0.3 0.4 0.3 0.3 0.1 0.0 0.1 0.1 0.24 A T A S 0.2 0.1 0.2 0.2 0.3 0.3 0.2 0.3 0.2 0.1 0.1 0.1 0.20 error 0.2 0.1 0.0 0.1 0.0 0.1 0.1 0.0 -0.1 -0.1 0.0 0.0 0.04 n (hours) 254 268 361 406 481 484 489 448 374 321 255 239 4380 Observed and AS values for 4,380 daytime hours Obs Mean AS Mean Obs StDev AS StDev M A E R M S E R M S E S R M S E U r r 2 d (°C) (°C) (°C) (°C) (°C) (°C) (°C) (°C) 0.24 0.20 0.63 0.35 0.45 0.60 0.51 0.28 0.35 0.12 0.46 (c) Observed & AS (4,380 nighttime hours) monthly mean heat island (°C) Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Ann. mean A T Q B S 0.9 0.7 1.2 1.0 1.0 1.6 1.1 2.1 1.1 0.6 0.3 0.3 0.91 A T A S 1.1 0.9 1.4 1.1 1.4 2.0 1.0 1.5 1.4 0.9 0.4 0.2 1.02 error -0.2 -0.2 -0.2 -0.1 -0.4 -0.4 0.1 0.6 -0.3 -0.3 -0.1 0.1 -0.11 n (hours) 490 404 383 314 263 236 255 296 346 423 465 505 4380 Observed and AS values for 4,380 nighttime hours Obs Mean AS Mean Obs StDev AS StDev M A E R M S E R M S E S R M S E U r r 2 d (°C) (°C) (°C) (°C) ( ° Q (°C) (°C) (°C) 0.91 1.02 0.95 0.93 0.54 0.74 0.32 0.30 0.70 0.49 0.82 82 Table 4.1 Lodz 1998 Algorithmic Scheme (AS) statistical evaluation (a) all, i.e. day & night hours, (b) daytime hours, (c) nighttime hours (a) Observed & AS (all 8,760 hours, i.e. day & night) monthly mean heat island (°C) Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Ann. mean A T O B S 0.6 0.3 0.4 0.5 0.7 0.8 0.6 0.6 0.6 0.2 0.2 0,2 0.47 A T A S 0.5 0.4 0.4 0.4 0.7 0.6 0.7 0.8 1.0 0.5 0.3 0.5 0.56 error 0.1 -0.1 0.0 0.1 0.0 0.2 -0.1 -0.2 -0.4 -0.3 -0.1 -0.3 -0.09 n (hours) 744 672 744 720 744 720 744 744 720 744 720 744 8760 Observed and AS values for all 8,760 hours Obs Mean AS Mean Obs StDev AS StDev M A E R M S E R M S E S R M S E U r r 2 d (°C) (°C) (°C) (°C) (°C) (°C) (°C) (°C) 0.47 0.56 0.82 0.76 0.47 0.65 0.32 0.29 0.67 0.45 0.80 (b) Observed & AS (4,380 daytime hours) monthly mean heat island (°C) Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Ann. mean A T 0 B S 0.1 0.1 0.2 0.2 0.4 0.5 0.3 0.2 0.1 -0.1 0.1 0.0 0.19 A T A S 0.1 0.1 ' 0.1 0.1 0.2 0.2 0.2 0.3 0.3 0.1 0.1 0.1 0.18 error 0.0 0.0 0.1 0.1 0.2 0.3 0.1 -0.1 -0.2 -0.2 0.0 -0.1 0.01 n (hours) 254 268 361 406 481 484 489 448 374 321 255 239 4380 Observed and AS values for 4,380 daytime hours Obs Mean (°C) 0.19 AS Mean (°C) 0.18 Obs StDev (°C) 0.60 AS StDev (°C) 0.32 M A E (°C) 0.41 R M S E (°C) 0.58 R M S E , (°C) 0.49 RMSE,, (°C) 0.26 0.33 0.11 0.43 (c) Observed & AS (4,380 nighttime hours) monthly mean heat island (°C) Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Ann. mean A TOBS 0.8 0.5 0.7 1.0 1.3 1.4 1.3 1.1 1.1 0.4 0.3 0.3 0.74 A T A S 0.7 0.6 0.7 0.6 1.4 1.4 1.5 1.4 1.7 0.8 0.4 0.6 0.91 error 0.1 -0.1 0.0 0.4 -0.1 0.0 -0.2 -0.3 -0.6 -0.4 -0.1 -0.3 -0.17 n (hours) 490 404 383 314 263 236 255 296 346 423 465 505 4380 Observed and AS values for 4,380 nighttime hours Obs Mean AS Mean Obs StDev AS StDev M A E R M S E R M S E S R M S E U r r 2 d (°C) (°C) ( ° Q (°C) ( ° Q ( ° Q (°C) (°C) 0.74 0.91 0.87 0.85 0.50 0.69 0.33 0.30 0.70 0.48 0.81 83 Table 4.2 Lodz 1999 Algorithmic Scheme (AS) statistical evaluation (a) all, i.e. day & night hours, (b) daytime hours, (c) nighttime hours (a) Observed & A S (all 8,760 hours, i.e. day & night) monthly mean heat island (°C) Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Ann. mean ATOBS 0.5 0.4 0.5 0.6 0.8 0.6 0.7 0.9 0.7 0.3 0.5 0.3 0.55 A T A S 0.5 0.5 0.4 0.4 0.7 0.5 0.8 1.1 1.3 0.8 0.7 0.4 0.67 error 0.0 -0.1 0.1 0.2 0.1 0.1 -0.1 -0.3 -0.6 -0.5 -0.2 -0.1 -0.12 n (hours) 744 672 744 720 744 720 744 744 720 744 720 744 8760 Observed and A S values for all 8,760 hours Obs Mean AS Mean Obs StDev AS StDev M A E R M S E R M S E S R M S E U r r 2 d (°C) (°C) (°C) (°C) C O (°C) C O CO 0.55 0.67 1.00 0.86 0.53 0.75 0.42 0.35 0.69 0.48 0.81 (b) Observed & A S (4,380 daytime hours) monthly mean heat island (°C) Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Ann. mean ATQBS 0.1 0.2 0.0 0.2 0.3 0.3 0.3 0.2 -0.1 -0.1 0.1 0.1 0.16 A T A S 0.2 0.1 0.1 0.1 0.3 0.2 0.3 0.4 0.4 0.2 0.2 0.1 0.22 error -0.1 0.1 -0.1 0.1 0.0 0.1 0.0 -0.2 -0.5 -0.3 -0.1 0.0 -0.07 n (hours) 254 268 361 406 481 484 489 448 374 321 255 239 4380 Observed and A S values for 4,380 daytime hours Obs Mean AS Mean Obs StDev AS StDev M A E R M S E R M S E S R M S E U r r 2 d (°C) (°C) ( ° Q (°C) (°C) (°C) ( ° Q (°C) 0.16 0.22 0.67 0.38 0.47 0.64 0.54 0.30 0.37 0.14 0.48 (c) Observed & A S (4,380 nighttime hours) monthly mean heat island (°C) Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Ann. mean ATQBS 0.6 0.5 0.9 1.1 1.6 1.2 1.4 1.9 1.6 0.6 0.6 0.4 0.93 A T A S 0.7 0.7 0.7 0.8 1.6 1.2 1.7 2.2 2.2 1.2 0.9 0.6 1.12 error -0.1 -0.2 0.2 0.3 0.0 0.0 -0.3 -0.3 -0.6 -0.6 -0.3 -0.2 -0.19 n (hours) 490 404 383 314 263 236 255 296 346 423 465 505 4380 Observed and A S values for 4,380 nighttime hours Obs Mean AS Mean Obs StDev AS StDev M A E R M S E R M S E S R M S E U r r 2 d (°C) (°C) (°C) (°C) (°C) (°C) C O C O 0.93 1.12 1.12 0.98 0.60 0.84 0.48 0.40 0.70 0.49 0.81 84 In figures 4.0, 4.1 and 4.2, the diagonal line is the 1:1 line. Figure 4.0 Observed vs. A S nighttime and daytime UHIs for Lodz, 1997 o o c re (0 CO < O) • N igh t t ime o D a y t i m e Z $ + 5 * Observed Heat Island (°C) Nighttime (n = 4,380): r 2 = 0.49 d = 0.82 Obs. Mean = 0.91 (°C) A S Mean = 1.02 (°C) M A E = 0.54 (°C) M A E / Obs. Mean = 0.59 Daytime (n = 4,380): r 2 = 0.12 d = 0.46 Obs. Mean = 0.24 (°C) A S Mean = 0.20 (°C) M A E = 0.45 (°C) M A E / O b s . Mean= 1.88 Figure 4.1 Observed vs. A S nighttime and daytime UHIs for Lodz, 1998 o o TJ C re CO < oo • N igh t t ime o D a y t i m e Observed Heat Island (°C) Nighttime (n = 4,380): r 2 = 0.48 d = 0.81 Obs. Mean = 0.74 (°C) A S Mean = 0.91 (°C) M A E = 0.50 (°C) M A E / Obs. Mean = 0.68 Daytime (n = 4,380): r 2 = 0.11 d = 0.43 Obs. Mean = 0.19 (°C) A S Mean = 0.18 (°C) M A E = 0.41 (°C) M A E / O b s . Mean = 2.16 85 Figure 4.2 Observed vs. A S nighttime and daytime UHIs for Lodz, 1999 -4-0™ o o x> c TO tn CO a> I CO < a> • N ight t ime o Day t ime 2 3 4 $ $ 8 $ 1 Observed Heat Island (°C) Nighttime (n = 4,380): r 2 = 0.49 d = 0.81 Obs. Mean = 0.93 (°C) A S Mean= 1.12 (°C) M A E = 0.60 (°C) M A E / Obs. Mean = 0.65 Daytime (n = 4,380): r 2 = 0.14 d = 0.48 Obs. Mean = 0.16 (°C) A S Mean = 0.22 (°C) M A E = 0.47 (°C) M A E / Obs. Mean = 2.94 86 4.1 Performance of the AS Figures 4.3 to 4.10 exemplify the ability of the A S to replicate U H I . Because small heat islands are proportionately more affected by the magnitude of cumulative errors and are not the primary concern for applied climatologists, only periods that contain heat islands that are larger than one degree are chosen. In addition, sequences are selected as good examples of how the O terms impact on scheme performance, and at least one example is drawn from each season. Since both <I>m and O w change magnitudes due to the effect weather has on their input variables (equations 1.5 & 1.21), relevant weather information is included for the purpose of understanding how weather may affect the performance of the A S . 4.1.1 Observed and AS UHI from January 1st to 5th, 1999 (Figure 4.3) Foggy conditions are found throughout these five days. The fog clears shortly after sunrise on all days except on the fourth where fog persists rather than clearing to conditions of 3-7 octas of high cloud as on the other days. No snow lies on the ground and light rain falls on the morning of the fifth suppressing heat island growth even more than the conditions did on the preceding days. In January, daylight period is about 7 hours long, and solar input barely reaches 200 W m" 2 at noon with a solar elevation of only 15 degrees (Figure 2.11). In general, the A S performs reasonably well in foggy conditions and this sequence presents a good example. The scheme's predictions are quite good at night on the first four nights when foggy conditions are recorded. The night of the 4-5 t h is also quite well replicated by the scheme when overcast windy conditions essentially obliterate any urban-rural temperature difference. 87 These first few days of 1999 also illustrate the ability of the scheme to replicate temperature differences at night, and inability to replicate the daytime situation when the urban area is often cooler than the rural area for a few hours. Given that the value of O m is relatively constant during this period, it means that O w is functioning quite well in this case. The inability of the scheme to retrodict cool islands is due to the limitation imposed by the <Dt term. Figure 4.3 Lodz, 1-5 January 1999 Urban Heat Island (C) W 9 9 1/5/99 1/5/99 1/6, 2:00 0:00 12:00 0: 99| 00 Obs UHI AS UHI 4.1.2 Observed and AS UHI from February 8th to 13th, 1999 (Figure 4.4) How the scheme performs when frontal conditions occur is well illustrated in this sequence. A blanket of 16 cm of snow lies on the ground throughout the February 8 t h to 13 t h period. Light snowfall and foggy conditions persist until noon on the 10 t h, when the sky 88 clears to 3 octas of high cloud, and later to clear conditions at 20:00 when light winds and sudden ninety plus degree shifts in the wind direction suggest the possibility of a frontal passage. These relatively unusual conditions give rise to a very large 9.4 °C heat island that the scheme under-estimates by about 60 %. This exceptional frontal event cannot be handled by the A S because the first term of the scheme sets an absolute limit to the difference at 4.5 °C. The A S replicates the sudden decline of several degrees in an hour well when fog returns in the early morning hours of the 11 t h. The scheme's values are 1 to 2 °C too large on the night of the 11 t h . Light snowfall on the 12 t h and 13 t h obliterates any urban-rural difference and this is very well replicated. Figure 4.4 Lodz, 8-13 February 1999 Urban Heat Island (°C) 10 Obs UHI AS UHI /99 2/8/99 2/9/99 2/9/99 2/10/99 2/10/99 2/11/99 2/99 2/12/99 2/13/99 2/13/99 2/14/99 OCO 12:00 0:00 12:00 0:00 12:00 0:00 12:00 0:00 12:00 0:00 12:00 0:00 89 4.1.3 Observed and AS UHI from May 7th to 11th, 1998 (Figure 4.5) The limitations imposed by the O t term in retrodicting daytime heat islands are clearly shown in this example. During the period of M a y 7 t h to 11 t h , 2-5 octas of cloud and light winds are generally found throughout these days. The unusually large 2.1 °C daytime heat island which occurs at 11:00 on the 9 t h under clear sky and light wind conditions is not replicated by the scheme. Nocturnal A S retrodictions are substantially better than those in the daytime. In the mean the A S monthly nocturnal values in this month are only 0.1 °C too high. Figure 4.5 Lodz, 7-11 May 1998 Urban Heat Island (°C) 5/7/98 5/8/98 K 5/8/98 5/9/98 5/9/98 5/10/98 1/5/10/9.8/ 5/11/98 5/11/9$ 00 0:00 V 12:00 0:00 12:00 0:00 12:0CN 0:00 12: DO -1 -2 Obs UHI • » - - A S UHI 90 4.1.4 Observed and AS UHI from July 5th to ^and July 11th to 14th, 1997 (Figures 4.6 and4.7) July 1997 was unusually wet (a total of 269 mm of rain, four standard deviations above the normal) with two very heavy rain periods, l s t - 9 t h (191 mm) and 18 t h -23 r d (61 mm). The following two sequences (Figures 4.6 and 4.7) show how the scheme performs in such wet conditions. From July 5 t h to 9 t h , heavy rain falls daily (14, 11, 62, 42 and 37 mm for each day respectively) and foggy conditions exist. The modeled thermal admittance value has greatly increased from 1300 to 1800 (J m"2 s"1/2 K" 1 ) and the magnitude of O m decreased from 0.60 to 0.47. The scheme performs well during heavy rain but not well on the few occasions when cloud cover reduces to less than 4 octas, and it retrodicts the small heat islands and the rapid changes poorly (Figure 4.6). Peak A S values are correct at about one degree, but the timing is poor (Figure 4.6). Figure 4.6 Lodz, 5-9 July 1997 Urban Heat Island (°C) O b s U H I • A S U H I q/97^  91 The scheme performs very well after the heavy rainfall period (Figure 4.7). From July 11 t h to 14 t h, light wind, partially cloudy and very light rain are the general weather conditions during daytime; and no wind and foggy from midnight to the early hours. The A S retrodiction is very good for several hours prior to midnight on all nights, but its performance drops off from midnight until about 06:00 when fog is present (Figure 4.7). Figure 4.7 Lodz, 11-14 July 1997 Urban Heat Island (°C) 7/14/97| 12:30 Obs UHI AS UHI 92 4.1.5 Observed and AS UHI from August 16th to 20th, 1997 (Figure 4.8) B i g heat islands are often under-predicted by the A S such as those shown in this sequence. Extended, mostly clear calm weather conditions from 0:00 on the 16 t h to 08:00 on the 18 t h enable the growth of a 6 °C heat island. A S under-predicted this large heat island substantially. The poor A S performance is probably due largely to the reduction of O m from 0.60 to 0.42 by the heavy rainfall in July. The reduced O m value limits the maximum A S value to less than 2.5 °C, 42% of the 5.8 °C A S maximum. This suggests that the thermal admittance may not be correct in summer as a result of poor modelling of the rate of drying for the sandy soil. Nocturnal performance of the scheme is very good on the last three nights of the period when light wind of 1 to 3 m s"1 is present. Figure 4.8 Lodz, 16-20 August 1997 Urban Heat Island (°C) 93 4.1.6 Observed and AS UHI from September 6th to 10th, 1999 (Figure 4.9) This period contains large heat islands of various sizes, and the sequence shows the varying ability of the scheme to replicate them. Quite dry soil, largely clear sky, and light wind conditions contribute to the development of large heat islands from September 6 t h to 10 t h. As daytime heat storage potential develops and nocturnal conditions become consistently clearer and calmer, heat island magnitudes grow; but the level of A S performance decreases. The A S replicates heat islands that are 4 °C or less very well, but under-predicts the larger ones. On the night of the 8-9 t h, after a period of calm clear sky, calm foggy conditions appear from 00:00 to 06:00, and the fog is cleared by a light wind at 7 am. The largest (6.2 °C) heat island occurs on the morning of the 9 t h and peaks at sunrise (05:00). The A S under-predicts this large heat island by ~ 2.5 °C. A n interesting phenomenon worth noting appears in the same morning. Between 06:00 to 07:00, an exceptionally rapid rural warming of 6.5 degrees (12.4-18.9 °C) occurs while the urban temperature increased by only 1.2 degrees (17.4-18.6 °C). Whether fog played a role is not clear because weather conditions are only available for the rural site, and the frequency of weather observations is on an hourly basis only. Rapid warming of 6.5 °C in one hour may be due to a front passing through the rural weather station site. A reason for the variable performance of the scheme in retrodicting the heat islands is the limit imposed by the thermal admittance term O m = 0.65. This limits the peak scheme value to 3.8 °C. It is likely that several days of clear sunny weather would in reality dry the soil more than the scheme's calculation gives. This in turn causes the under-prediction of the larger heat islands in the second half of the period. 94 The <&t term also plays a role in the under-predictions of the early morning peak heat island magnitudes. Due to the clear and calm sky conditions, the <I>W term exerts little effect in this case. The largest (1.7 °C) cool island occurs just after sunrise on the 7 t h , following several hours of near-clear sky and light wind conditions. It diminishes rapidly when 4 octas of low cloud comes in during the morning at 11:00. Daytime cloud cover appears to be a significant control on the cool island magnitude. The A S cannot replicate this, or any of the early morning and evening cool island conditions that occur throughout the period due to the constraining effect of <3?e Figure 4.9 Lodz, 6-10 September 1999 Urban Heat Island (°C) 95 4.1.7 Observed and AS UHI from December 2nd to 6th, 1998 (Figure 4.10) During this period, snow cover with a low thermal admittance maximizes <E>m and it remains at the value of 1. Therefore, how well the 3>w and O t terms function in the scheme may be revealed in this sequence. On the night of 2-3 r d , light wind and clear sky persists through the day until 7 am. On the night of 3-4 t h, 6 octas of low cloud comes in at 9 pm, following a few hours of clear sky. The A S replicates these 2-degree heat islands quite well, especially the larger one of the first night. The performance of the scheme suggests that the combined effects of the <J>W and O t terms are quite good at nighttime. Figure 4.10 Lodz, 2-6 December 1998 Urban Heat Island (°C) 2/6/98 96 4.2 Results of the Tests The A S performs at a similar level for each of the three years. In general, the scheme retrodicts daytime heat islands poorly, and it does not have the capacity to replicate daytime cool islands. Nighttime performance is much better; but there is a tendency to under-predict large heat islands that are over 4 °C. The statistical results from each year are presented below for comparison. 4.2.1 Results from the 1997 Lodz data T h e ' d ' statistics indicate a poor daytime performance (d = 0.46) and a good nighttime performance (d = 0.82) for the A S (Table 4.0). The nocturnal Mean Absolute Error ( M A E ) of the A S is 0.54 °C, 59 % of the observed mean 0.91 °C heat island. The nocturnal mean of the A S is 1.02 °C. This is an over-prediction of the nighttime heat island by 0.11 °C. Nocturnal Root Mean Square Error ( R M S E ) for the A S is 0.74 °C. The nocturnal systematic and unsystematic errors are almost equal at 0.32 °C and 0.30 °C respectively. 4.2.2 Results from the 1998 Lodz data The 1998 results are similar to the results for 1997. Daytime performance of the A S has a l o w ' d ' value of 0.43, and a much higher 'd ' value (d = 0.81) for the nighttime (Table 4.1). The nocturnal M A E of the A S is 0.50 °C, 68 % of the observed mean 0.74 °C heat island. The nocturnal mean of the A S is 0.91 °C, over-predicted nighttime heat island by 0.17 °C. The R M S E for the A S (0.69 °C) has systematic and unsystematic components of about equal magnitude at 0.33 °C and 0.30 °C, respectively. 97 4.2.3 Results from the 1999 Lodz data Again, the scheme has a far better performance at nighttime (d = 0.81) then in daytime (d = 0.48) (Table 4.2). The nocturnal M A E of the A S is 0.60 °C, 65% of the observed mean 0.93 °C nocturnal heat island. The nocturnal mean of the A S is 1.12 °C, over-predicted nighttime heat island by 0.19 °C. The R M S E for the A S (0.84 °C) has systematic and unsystematic components of about equal magnitude at 0.48 °C and 0.40 °C, respectively. 4.2.4 Month-by-month variability of the AS Some month-by-month variability in the accuracy of the retrodiction of the observed heat island is found. When applying the A S to the three years of data, the average monthly error in the mean is an over-estimation of 0.16 °C. The scheme has an average error of an over-estimation of 0.11 °C, 0.17 °C, and 0.19 °C for 1997, 1998 and 1999 respectively. There is significant variability in some months from year to year. Variable performance is most noticeable in December. Months with a mean nocturnal heat island of less than 0.5 °C are poorly replicated, whereas summer months with 0.6 °C and greater urban heat islands are associated with a better performance. The average 1997-1999 nocturnal performance for each month for the three years ranges from a low of d = 0.47 (December 1997) to a high of d = 0.88 (April 1997 & June 1999) (Table 4.3). 98 Table 4.3 Nocturnal 'd ' Statistics and observed average nocturnal U H I magnitude 1997 1998 1999 average 'd ' average U H I Ranked list of average 'd ' ' d ' ' d ' • for for ' d ' for 1997-1999 1997-1999 1997-1999 (°C) January 0.74 0.77 0.86 0.79 0.8 0.83 June February 0.74 0.82 0.80 0.79 0.6 0.81 July March 0.81 0.82 0.71 0.78 0.9 0.79 May Apr i l 0.88 0.80 0.69 0.79 1.0 0.79 February May 0.82 0.74 0.82 0.79 1.3 0.79 Apr i l June 0.82 0.81 0.88 0.83 1.4 0.79 January July 0.84 0.83 0.77 0.81 1.3 0.78 March August 0.70 0.83 0.77 0.77 1.7 0.78 September September 0.80 0.78 0.77 0.78 1.2 0.77 August October 0.79 0.74 0.75 0.76 0.5 0.76 October November 0.79 0.69 0.74 0.74 0.4 0.74 November December 0.47 0.74 0.71 0.64 0.3 0.64 December Year 0.82 0.81 0.81 0.81 0.9 4.3 Potential Errors in the Observed Heat Islands The accuracy of the observed heat islands depends on many factors such as the quality of the instrument and the screen enclosure, the accuracy of the thermometer calibration, the exposure of the screen, the types of local human activities, and the kinds of manipulations of the data collection and analysis processes. The Vaisala technical data sheet shows that the H M P 35 temperature probe used at the Lodz urban and rural sites, i f calibrated annually, is accurate to ± 0.2 °C for the temperature range measured in Lodz. Therefore, the maximum instrumental error is ± 0.4 °C for an urban-rural temperature difference. The urban and rural temperatures in both Uppsala (1976-77) and Lodz are measured with instruments mounted in Stevenson screens. Instruments located at the 99 urban sites, which usually have lighter wind conditions, can be affected by a reduced air flow through the screens and result in a difference between the measured and actual urban air temperature. On the other hand, instruments located in the rural sites tend to be less affected by low wind speed due to the more open locations. Thermometers at the urban sites are also more likely to be subjected to thermal effects from nearby human activities. It is difficult to quantify these errors and their cumulative effect on the observed heat islands because the necessary meta-data are not available. Given that there are instrumental errors and the likelihood of some of the other possible errors is high, evaluation of the scheme performance has been focused on larger heat islands for the screen based studies. 4.4 Analysis of AS Sensitivity to Error in AS inputs The sensitivity of the A S to the likely maximum error expected for each of the input variables is assessed using the Lodz 1997-1999 data set. A n error of 10% is assigned to the H / W ratio as its likely maximum error between the actual H / W ratio and the estimated H / W ratio from photographs; for thermal admittance, one step in the selection table; for cloud cover, maximum trained observer error; for wind speed, maximum instrument error. Each input is alternately increased and then decreased by the likely maximum error, and the effect on the M A E is recorded for each case. The A S nighttime schemes sensitivities are given as, (a) the overall M A E and (b) as a percentage of the overall mean observed heat island magnitude (Table 4.4). 100 Scheme performance is by far the most sensitive to the under-estimation of the mean observed 2.7 m s"1 rural wind speed by 1 m s"1. It increases the M A E from 65 % to 82 % of the observed heat island magnitude. A n over-estimation of 1 m s"1 reduced the M A E from 65 % to 59 % of the observed heat island magnitude. This is because the 110 M A E is proportional to the wind speed function that is (u" ). The likely maximum errors in H / W ratio, thermal admittance and cloud cover do not cause performance to deteriorate by more than 7 % individually. Table 4.4 Sensitivity of A S to likely maximum errors in A S inputs A S L o d z 1997-1999 (Nighttime hours) sensitivity to scheme inputs M A E (°C) M A E / Obs U H I (%) Canyon Height to Width ratio 0.60 65 Canyon Height to Width ratio + 0.1 0.64 69 Canyon Height to Width ratio - 0.1 0.53 57 Thermal admittance u.r ( J m~2s~1/2 K" 1 ) 0.60 65 Thermal admittance |Xr + 200 ( J m" 2s" 1 / 2 K" 1 ) 0.54 58 Thermal admittance \ir - 200 ( J m"2 s"1/2 K" 1 ) 0.66 71 Cloud cover n (Octas) 0.60 65 Cloud cover n + 1 (Octas) 0.66 71 Cloud cover n - 1 (Octas) 0.67 72 10 m rural wind speed u (m/s) 0.60 65 10 m rural wind speed u + 1 (m/s) 0.55 59 10 m rural wind speed u - 1 (m/s) 0.76 82 101 CHAPTER 5 DISCUSSION AND CONCLUSION 5.1 Discussion 5.1.1 Limitations posed by the data sets Despite the high quality of the data sets, there are limitations presented by the data and some may affect the accuracy of the test. For an objective evaluation of the potential and performance of the scheme, the following issues need to be taken into consideration. Firstly, the urban site of Lodz, Lipowa, was not located in the heart of the heat island, and 38 % of the urban canyon is grass and trees. It is questionable i f the site was truly 'urban' and suitable for the direct application of the equation, especially during the transition periods in the spring and fall when trees gained and lost foliage. Secondly, there are uncertainties of the sky view factors (SVF) , which are required in the calculations of the first term. Publications for the Uppsala data sets do not provide enough details for calculating the SVFs , and memories of the researchers have faded. As a result, the SVFs for the Uppsala sites are only estimates. The accurate SVFs for the Lodz 's urban site are known but not for the transition periods of spring and fall when leaf coverage is changing. Thirdly, the temporal factor (<J>t) has the most significant effect on the scheme under fine weather conditions, especially during daytime. To assess the 4>t term more properly, a full diurnal period of continuous fine weather data, preferably from midnight to midnight, is needed. Unfortunately such weather periods are not found in the data sets. 102 Similarly, thermal admittance field data are not available for checking how applicable the thermal admittance factor (<3>m) is to reality. Finally, only temperature was measured at both rural and urban sites; all other weather data, except solar radiation (measured at the Lodz urban site) were taken from the rural sites. This leads to the uncertainty of how accurate the weather data are when applying them to the urban sites. Furthermore, the Lodz data shows that mean monthly nocturnal performance for all months of the year can be good, and only the month of December 1997 stands out as very poor. However, available data for this month do not show anything that is not seen in other months. It can be an indication that factors that have affected the scheme performance were simply not captured in the data set. 5.1.2 Universality of the scheme When evaluating how universally applicable the scheme is, potential bias should be considered. Even though test results indicate that A S captures the diurnal and seasonal pattern of urban heat islands quite well, it should be noted that both Uppsala and Lodz are located in the mid-latitudes. Since the A S is almost entirely developed with data from mid-latitude settlements, the scheme may be biased by giving more favourable results for such sites. The A S has only been applied to cities in simple topographic locations. Further studies on the scheme's performance for urban sites that are located in cities with more complex topography are necessary. In addition, the 1997-1999 Lodz data has a small 103 (0.54 °C) overall mean urban-rural temperature difference. Performance of the scheme for cities that have larger differences may not necessarily be the same. 5.2 Conclusion The A S is a good initial scheme and performs at least on par with the other regression models. It retrodicts nighttime heat islands far better than that of the daytime heat islands, and the performance can be good in any month. The scheme performs better in months with larger heat islands, but individual nights with large heat islands can exhibit great errors where peak values are under-predicted. Such under-predictions are mainly due to the limitations imposed by the O m term. The scheme is unable to predict daytime heat islands, and morning and afternoon cool islands because of the limitation imposed by the O t term. To improve the scheme, the functions of the O terms must be reassessed and modified. The effects of cloud cover upon heat island magnitude may need to be handled differently for day and night periods. To improve daytime performance, the scheme must permit negative heat islands. It is likely that local canyon shading effects play an important part in the daytime conditions, and it is important to investigate this issue. 5.3 Suggestions for Future Studies The scheme has good potential and requires further development. To develop it properly, the O terms need to be assessed more thoroughly. Ideally, they can be tested 104 with data that contains a full diurnal period of continuous fine weather, and data that would provide enough information on the urban sites to accurately determine the sky view factors and on rural soil moisture that are measured at frequent intervals, preferably hourly, by automated means. Testing with data sets from cities in areas outside the mid-latitudes would help to determine how widely applicable the scheme is. Furthermore, many settlements are located in complex terrain or adjacent to water bodies, and for some, anthropogenic heat is a significant term in the canopy layer energy balance. Therefore, testing with data for cities that have complex conditions may help to expand the applicability of the scheme, or show the need for new algorithms or modification of existing ones. 105 LIST OF SYMBOLS AND ABBREVIATIONS A S Algorithmic Scheme C heat capacity (J m" 3 K" 1 ) d Willmott 's index of agreement W horizontal spacing (canyon width) of surface roughness elements (m) H height of (buildings) surface roughness elements (m) k thermal conductivity (W m"1 K" 1 ) M A E mean absolute error n fractional cloud-cover (1/10's), number of data points r Pearson's correlation coefficient r Pearson's coefficient of determination R H relative humidity (%) R M S E root mean squared error T u urban air temperature (°C) T r rural air temperature (°C) Tu_ r urban-rural air temperature difference, urban heat island (°C) ATu-rfmax) maximum urban-rural air temperature difference, urban heat island (°C) u mean horizontal (10 m) rural site wind speed (m s"1) U B L urban boundary layer (above roof level) U C L urban canopy layer (below roof level) U H I urban heat island (°C) O m A S rural surface thermal admittance reduction factor O w A S weather reduction factor O t A S time of day reduction factor APPENDIX 107 Table A2.0 Uppsala, mean air temperature and precipitation 1948-1949 and 1976-1977, and temperature and precipitation normals Uppsala Meteorological Institute mean monthly air temperature 1948-1949 (°C) Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Ann mean 1948 -5.9 -3.5 1.9 6.6 10.6 14.2 17.5 14.7 11.5 5.0 0.8 2.1 6.3 1949 -0.6 0.5 -0.5 5.8 12.5 13.3 17.1 14.6 14.2 6.7 3.6 0.1 7.3 Source: (Moberg, A. and H. Bergstrom, 1997) Uppsal a Meteorological Institute mean monthly air temperature 1976-1977 (°C) Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Ann mean 1976 -6.2 -2.3 -2.9 3.9 10.9 14.2 16.6 16.0 8.5 5.0 1.7 -3.5 5.2 1977 -2.6 -5.5 0.8 2.6 10.2 14.6 13.9 14.2 9.3 7.4 2.4 -0.9 5.5 Source: (Moberg, A. and H. Bergstrom, 1997) Uppsala 30-year normal mean monthly air temperature 1931-1960 (°C) Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Ann mean mean -4.2 -4.8 -1.2 3.9 9.8 15.0 16.4 15.3 11.0 6.3 1.3 -2.1 5.6 StDev 2.8 3.4 2.6 1.6 1.3 1.3 1.5 1.4 1.3 1.4 2.0 2.7 0.8 Source: (Taesler, 1972) Uppsala Meteorological Institute monthly precipitation January 1948 - M a y 1949 (mm) Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Ann total 1948 60 8 24 45 51 44 69 113 33 33 26 22 525 1949 47 14 16 30 38 Uppsala F16 monthly precipitation January 1976 -February 1977 (mm) Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Ann total 1976 26 10 11 28 27 32 27 23 73 15 60 85 417 1977 48 21 Uppsala 30-year normal mean monthly precipitation 1931-1960 (mm) Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Ann total mean 39 26 25 30 32 46 60 73 52 51 50 42 526 StDev 16 16 15 16 24 20 37 35 29 27 21 19 92 Source: (Taesler, 1972) 108 Table A2.1 Lodz climate data for the 1997-1999 data collection period and average temperature and average precipitation Lodz, Lublinek mean monthly air temperature 1997-1999 (°C) Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Ann mean 1997 -4.1 1.9 3.4 5.2 13.8 17.2 17.7 19.9 13.3 6.4 2.8 0.4 8.2 1998 1.0 3.5 2.3 10.5 14.9 17.9 17.7 16.9 13.7 7.5 -1.1 -1.9 8.6 1999 0.4 -1.4 5.1 9.9 13.7 17.1 20.8 18.3 16.9 8.5 1.8 0.5 9.4 Source: (Fortuniak) Lodz, Lublinek average air temperature 1951-1990 (°C) Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Ann mean -4.8 -3.6 2.9 9.2 13.5 16.9 18.9 18.3 14.5 9.5 3.0 -2.0 8.0 Source: (Fortuniak) Lodz, Lublinek monthly precipitation 1997-1999 (mm) Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Ann total 1997 4 37 23 30 69 46 269 20 47 52 66 38 700 1998 41 47 45 56 49 69 100 47 51 81 38 54 678 1999 39 46 34 59 43 143 47 17 22 52 31 32 565 Source: (Fortuniak) Lodz, Lublinek average monthly precipitation 1931-1995 (mm) Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Ann total mean 30 29 31 36 50 67 84 67 48 37 46 39 564 StDev 16 15 15 17 25 35 43 33 29 30 28 20 99 Source: (Fortuniak) 109 REFERENCES Anandakumar, K . (1999). A study on the partition of net radiation into heat fluxes on a dry asphalt surface. Atmospheric Environment 33(24-25): 3911-3918. Asaeda, T., V . T. Ca and A . Wake (1996). Heat storage of pavement and its effect on the lower atmosphere. Atmospheric Environment 30(3): 413-427. Asaeda, T. and C. T. V u (1997). Effects of different pavements on the urban thermal environment. Proceed, of the A A G 93rd Annual Meeting, Fort Worth, Texas, U S A . A S H R A E (1993). Determination of climate variation within metropolitan areas; Phasel summary. A S H R A E Trans. 1993, vol.99. Part 1, Paper number 3665 (RP-606), 430-449, 8 tabs, refs. K . L . Dufner, D . T. Bailey, D . E . Wolfe and S. P. Arya. New York, A S H R A E . 99. Atkinson, B . W . (1977). Urban effects on precipitation : an investigation of London's influence on the severe storm in August 1975. London, Dept. of Geography Queen Mary College University of London. Becker, S. (2000). Bioclimatological rating of cities and resorts in South Africa according to the climate index. International Journal of Climatology 20: 1403-1414. Best, M . J. (1998). A model to predict surface temperatures. Boundary Layer Meteorology 88: 279-306. Bogren, J. (1998). Monitoring temperature variations along roads. Optimising road weather information systems by use of models. Proceed. A M S 13th Conference on Biometeorology and Aerobiology, Albuquerque, New Mexico, Amer. Meteorol. Society, Boston, 259-261. Bogren, J., T. Gustavsson and U . Postgard (2000). Local temperature differences in relation to weather parameters. International Journal of Climatology 20(2): 151-170. Bohm, R. (1998). Urban bias in temperature time series - a case study for the city of Vienna, Austria. Climatic Change 38(1): 113-128. Bolz , H . M . , 1949: Die Abhangigkeit der infraroten Gegenstrahlung von der Bewolkung. Zeitschritt fur Meteorologie, 3, 201-203. Bowling, S. A . and C. S. Benson (1978). Study of the subarctic heat island at Fairbanks, Alaska, U S E P A , Research Triangle Park, N . C . : 150 pp. Bretz, S., H . Akbari and A . Rosenfeld (1998). Practical issues for using solar-reflective materials to mitigate urban heat islands. Atmospheric Environment 32(1): 95-101. Bridgman, H . A . , R. F. Warner and J. R. Dodson (1995). Urban Biophysical Environments. Melbourne, New York, Oxford University Press. Brunetti, M . , F. Mangianti, M . Maugeri and T. Nanni (2000). Urban heat island bias in Italian air temperature series. Nuovo Cimento Delia Societa Italiana D i Fisica Geophysics and Space Physics 23(4): 423-431. Campbell, G . S. and J. M . Norman (1998). A n Introduction to Environmental Biophysics. New York, Springer. 110 Carlson, T. N . , J. K . Dodd, S. G . Benjamin and J. N . Cooper (1981). Satellite estimation of the surface energy balance, moisture availability and thermal inertia. Journal of Applied Meteorology 20: 67-87. Chandler, T. J. (1965). The Climate of London. London, Hutchinson. Cionco, R. M . and R. Ellefsen (1998). High resolution urban morphology data for urban wind flow modeling. Atmospheric Environment 32(1): 7-17. Comrie, A . C. (1990). The climatology of surface ozone in rural-areas: a conceptual model. Progress in Physical Geography 14(3): 295-316. Eliasson, I. (1996). Intra-urban nocturnal temperature differences: a multivariate approach. Climate Research 7(1): 21-30. Ellefsen, R. (1991). Mapping and measuring buildings in the canopy boundary-layer in 10 United-States cities. Energy and Buildings 16(3-4): 1025-1049. Ellefsen, R. (1994). Meteorologically related urban terrain characteristics of Uppsala, Sweden. Adelphi, M D , U S A , U S A R L . Fenger, J. (1999). Urban air quality. Atmospheric Environment 33(29): 4877-4900. Field, W . J. (1973). The effect of overcast cloud on the Vancouver heat island. Geography 411 project report, Geography department, U B C , Vancouver, B . C . Fortuniak, K . and K . Klys ik (1998). Intensity of the urban heat island in Lodz under winter conditions and its simple model. Acta Universitatis Lodziensis, Folia Geographica Physica 3: 83-90. Geiger, R. (1966). The Climate near the Ground. Cambridge, Massachusetts, Harvard University Press. Grimmond, C. S. B . and T. R. Oke (1999). Heat storage in urban areas: local-scale observations and evaluation of a simple model. Journal of Applied Meteorology 38(7): 922-940. Grimmond, C. S. B . and T. R. Oke (2000). A Local-scale Urban Meteorological Pre-processing Scheme ( L U M P S ) . C O S T , Prague, Czechoslovakia. Haagen-Smit, A . J. (1952). Chemistry and physiology of Los Angeles smog. Industrial Engineering and Chemistry 44: 1342-46. Hanna, S. R. and J.C. Chang (1992). Boundary-layer parameterizations for applied dispersion modeling over urban areas. Boundary-Layer Meteorology 58: 229-259. Heidorn, K . C . and D . Yap (1986). A synoptic climatology for surface ozone concentrations in southern Ontario, 1976-1981. Atmospheric Environment 20: 695-703. Hilberg, S. D . (1978). Description of Surface Weather Conditions, Temperature Analysis. Ackerman. Urbana. ISWS Bulletin # 63. Hogstrom, U . , R. Taesler, S. Karlsson, L . Enger and A . O. Smedman-Hogstrom (1978). The Uppsala urban meteorology project. Boundary-Layer Meteorology 15: 69-80. Holtslag, A . A . M . , and A . P. Van Ulden, 1983: A simple scheme for daytime estimates of the surface fluxes from routine weather data. / . Applied Meteorology, 22, 517-529. Hoppe, P. (1991). Improving Indoor Thermal Comfort by Changing Outdoor Conditions. Energy and Buildings 16(1-2): 743-747. I l l Howard, L . (1833). The Climate of London Deduced from Meteorological Observations Made in the Metropolis and at Various Places Around It. London, Harvey and Darton. Hoyano, A . (1999). Analysis of the influence of urban form and materials on sensible heat flux - a case study of Japan's largest housing development "Tama New Town". Atmospheric Environment 33: 3931-3939. Ichinose, T., K . Shimodozono and K . Hanaki (1999). Impact of anthropogenic heat on urban climate in Tokyo. Atmospheric Environment 33(24-25): 3897-3909. Idso, S.B. and R . D . Jackson (1969). Thermal radiation from the atmosphere. Journal of Geophysical Research. 74, 3397-3403. fPCC (1995). Climate Change. Cambridge, Cambridge University Press. Jarnefors, B . (1956). Quaternary deposits in the Uppsala region. Stockholm, Swedish Geological Service. Johnson, G . T., T. R. Oke, T. J. Lyons, D . G . Steyn, I. D . Watson and J. A . Voogt (1991). Simulation of surface urban heat islands under ideal conditions at night .1. theory and tests against field data. Boundary-Layer Meteorology 56(3): 275-294. Kar l , T. and R. W . Knight (1997). The 1995 Chicago heat wave: how likely is a recurrence? Bulletin of the American Meteorological Society 78(6): 1107-1119. Kar l , T. R. and P. D . Jones (1989). Urban bias in area-Averaged surface air-temperature trends. Bulletin of the American Meteorological Society 70(3): 265-270. Kar l , T. R., C. N . J. Will iams, P. J. Young and W . M . Wendland (1986). A model to estimate the time of observation bias associated with monthly mean maximum, minimum and mean temperatures for the United States. Journal of Climate and Applied Meteorology: 145-159. Kassomenos, P., V . Kotroni and G . Kallos (1995). Analysis of Climatological and A i r -Quality Observations from Greater Athens Area. Atmospheric Environment 29(24): 3671-3688. Klysik , K . (1993). The meteorological observation records from Lodz. Geograficzne, Krakow, Poland. Klysik , K . (1996). Spatial and seasonal distribution of anthropogenic heat emissions in Lodz, Poland. Atmospheric Environment 30(20): 3397-3404. Klysik , K . and K . Fortuniak (1999). Temporal and spatial characteristics of the urban heat island of Lodz, Poland. Atmospheric Environment 33(24-25): 3885-3895. Landsberg, H . E . (1981). The Urban Climate. New York, Academic Press. Lindqvist, S. (1992). Local climatological modelling for road stretches and urban areas. Geografiska Annaler, Series A 74A(2-3): 265-274. Lowry, W . P. (1977). Empirical estimation of urban effects on climate: a problem analysis. Journal of Applied Meteorology 16: 129-135. Masson, V . (2000). A physically-based scheme for the urban energy budget in atmospheric models. Boundary-Layer Meteorology 94(3): 357-397. Masson, V . , C .S .B . Grimmond and T.R. Oke, (2002). Evaluation of the Town Energy Balance (TEB) scheme with direct measurements from dry districts in two cities. Journal of Applied Meteorology, in press. 112 McClean, A . A . and T. R. Oke (1998). Development and tests of a simple model to predict road surface temperature in urban as well as rural terrain. Proceed. A M S 13th Conference on Biometeorology and Aerobiology, Albuquerque, New Mexico, American Meteorological Society, Boston 262-265. Mizuno, M . , Y . Nakamura, H . Murakami and S. Yamamoto (1991). Effects of land-use on urban horizontal atmospheric-temperature distributions. Energy and Buildings 15(1-2): 165-176. Moberg, A . and H . Bergstrom (1997). Homogenization of Swedish temperature data. Part 3: the long temperature records from Uppsala and Stockholm. International Journal of Climatology 17(7): 667-699. Monteith, J. L . and M . H . Unsworth (1990). Principles of Environmental Physics. London, E . Arnold. Myrup, L . O. (1969). A numerical model of the urban heat island. Journal of Applied Meteorology 8: 896-907. Nicholls, N . , R. Tapp, K . Burrows and D . Richards (1996). Historical thermometer exposures in Australia. International Journal of Climatology 16(6): 705-710. Oke, T. R. (1976). The distinction between canopy and boundary-layer urban heat islands. Atmosphere 14(4): 268-277. Oke, T. R. (1981). Canyon geometry and the nocturnal urban heat island: comparison of scale model and field observations. Journal of Climatology 1(3): 237-254. Oke, T. R., 1982: The energetic basis of the urban heat island. Quarterly Journal of the Royal Meteorological Society, 108, 1-24. Oke, T. R. (1986). Urban Climatology and its Applications with Special Regard to Tropical Areas : Proceedings of the Technical Conference. Geneva, Switzerland, Secretariat of the World Meteorological Organization. Oke, T. R. (1987). Boundary Layer Climates. London ; New York, Routledge. Oke, T. R. (1988). The urban energy balance. Progress in Physical Geography 12(4): 471-508. Oke, T. R. (1995). Classics in physical geography revisited. Climatological studies in Uppsala with special regard to the temperature conditions in the urban area. Geographica 22. Progress in Physical Geography 19(1): 107-113. Oke, T. R. (1998). A n algorithmic scheme to estimate hourly heat island magnitude. Proceed. A M S 13th Conference on Biometeorology and Aerobiology, Albuquerque, New Mexico, American Meteorological Society, Boston, M A . 80-83. Oke, T. R., G . T. Johnson, D . G . Steyn and I. D . Watson (1991). Simulation of surface urban heat islands under ideal conditions at night: 2. diagnosis of causation. Boundary-Layer Meteorology 56(4): 339-358. Peppier, A . (1929). Das Auto als Hilfsmittel der Meteorologischen Forschung. Das Wetter 46: 305-308. Prata, A . J. (1996). A new long-wave formula for estimating downward clear-sky radiation at the surface. Quarterly Journal of the Royal Meteorological Society, 122, 1127-1151. 113 Pratt, D . A . and Ellyett (1979). The thermal inertia approach to mapping of soil moisture and geology. Remote Sensing of Environment 8: 151-168. Quattrochi, D . A . , J. C . Luval l , E . M . G . , C . P. Lo , K . S.Q., H . J., H . Taha, R. D . Bornstein, R. R. Gillies and K . P. Gallo (1998). Project Atlanta (Atlanta land use analysis: temperature and air quality) - A study of how the urban landscape effects meteorology and air quality through time. Proceed. A M S 13th Conference on Biometeorology and Aerobiology, Albuquerque, New Mexico. , 104-107. Rastorgueva, G . P. (1979). Temperaturnyj Rezhim Vozdukha Promyshlennykh Gorodov U S S R . Glavnava Geofiz. Observ., 436: 111-117. Rosenfeld, A . (1995). Mitigation of urban heat islands: materials, utility programs, updates. Energy and Buildings 22: 255-265. Rosenfeld, A . H . , H . Akbari , J. J. Romm and M . Pomerantz (1998). Cool communities: strategies for heat island mitigation and smog reduction. Energy and Buildings 28(1): 51-62. Runnalls, K . E . and T. R. Oke (2000). Dynamics and controls of the near-surface heat island of Vancouver, B . C . Physical Geography 21(3): 283-304. Schmidt, W . (1927). DieVerteilung der Minimumtemperaturen in der Frostnacht des Gemeindegebiet von Wien. Fortschritte d. Landwirtschaft 2(21): 5-27. Shao, J. and P. J. Lister (1995). The prediction of road surface-state and simulation of the shading effect. Boundary-Layer Meteorology 73(4): 411-419. Sillman, S. (1990). The sensitivity of ozone to nitrogen oxides and hydrocarbons in regional ozone episodes. Journal of Geophysical Research 95(D2): 1837-1851. Sillman, S. (1995). The use of N O x , H 2 0 2 , and FTN03 as indicators for ozone-NOx -hydrocarbon sensitivity in urban locations. Journal of Geophysical Research 100(D7): 14,175-14,188. Smedman-Hogstrom, A . - S . and U . Hogstrom (1973). The Marsta Micro-Meteorological Field Project., Profile Measurement System and Some preliminary Data. Boundary-Layer Meteorology 5: 259-273. Sundborg, A. (1951). Climatological studies in Uppsala with special regard to the temperature conditions in the urban area. Geographica, Geographical Institute, Uppsala, (22): 111. Swaid, H . (1991). Nocturnal Variation of A i r Surface-Temperature Gradients for Typical Urban and Rural Surfaces. Atmospheric Environment Part B-Urban Atmosphere 25(3): 333-341. Swaid, H . , M . Barel and M . E . Hoffman (1993). A Bioclimatic Design Methodology for Urban Outdoor Spaces. Theoretical and Applied Climatology 48(1): 49-61. Swinbank, W. C. (1963). Long-wave radiation from clear skies. Quarterly Journal of the Royal Meteorological Society, 89, 339-348. Taesler, R. (1972). Climatological data for Sweden. Stockholm. Taesler, R. (1980). Studies of the development and thermal structure of the urban boundary layer in Uppsala, Meteorological Institution, Uppsala University, Uppsala, Sweden. 61: 236. 114 Taesler, R. (1981). Urban meteorological studies in Uppsala. Faculty of Science. Uppsala, Uppsala: 21. Taha, H . , S. Konopacki and S. Gabersek (1999). Impacts of large-scale surface modifications on meteorological conditions and energy use: A 10-region modeling study. Theoretical and Applied Climatology 62(3-4): 175-185. Thornes, J. E . and J. Shao (1991). Spectral-analysis and sensitivity tests for a numerical road surface-temperature prediction model. Meteorological Magazine 120(1428): 117-124. United Nations, P. D . (2000). Population Statistics web page, www.unpd.org/popin/. Uno, I., S. Wakamatsu and H . Ueda (1988). Behaviour of the nocturnal urban boundary layer and air pollutants. Journal Japanese Society A i r Pollution 23: 103-114. Willmott, C . J. (1981). On the validation of models. Physical Geography 2(2): 184-194. Wood, N . L . PL, P. J. McLean and R. T. Clark (1998). The reaction times of road-surface temperatures after cloud clearance and cold advection in the county of Devon. Proceedings of the 9th S I R W E C Conference, Lulea, Sweden: 223-228. Zdunkowski, W . G. , R. M . Welch and J. Peagle (1976). One-dimensional numerical simulation of the effects of air pollution on the planetary boundary layer. Journal of the Atmospheric Sciences 33: 2399-2414. 115 

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.831.1-0090356/manifest

Comment

Related Items