UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Energy storage in a densely-built mediterranean city centre Roberts, Sarah Marie 2003

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

Item Metadata

Download

Media
831-ubc_2003-0356.pdf [ 23.19MB ]
Metadata
JSON: 831-1.0052339.json
JSON-LD: 831-1.0052339-ld.json
RDF/XML (Pretty): 831-1.0052339-rdf.xml
RDF/JSON: 831-1.0052339-rdf.json
Turtle: 831-1.0052339-turtle.txt
N-Triples: 831-1.0052339-rdf-ntriples.txt
Original Record: 831-1.0052339-source.json
Full Text
831-1.0052339-fulltext.txt
Citation
831-1.0052339.ris

Full Text

ENERGY STORAGE IN A DENSELY-BUILT MEDITERRANEAN CITY CENTRE by SARAH MARIE ROBERTS B.S. (Atmospheric, Oceanic, and Space Sciences), The U n i v e r s i t y of Michigan, Ann Arbor, Michigan, 1999 A THESIS SUBMITTED IN PARTIAL FULFILMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE i n THE FACULTY OF GRADUATE STUDIES (Department of Earth and Ocean Sciences; Atmospheric Science Programme) We accept t h i s t h e s i s as conforming to the r e q u i r e d standard THE UNIVERSITY OF BRITISH August 2003 © Sarah Marie Roberts, COLUMBIA 2003 In presenting t h i s thesis i n p a r t i a l f u l f i l m e n t 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 f r e e l y a v a i l a b l e for reference and study. I further agree that permission for extensive copying of t h i s thesis for s c h o l a r l y purposes may be granted by the head of my department or by his or her representatives. It i s understood that copying or p u b l i c a t i o n 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. Department The. University of B r i t i s h Columbia Vancouver, Canada ABSTRACT The complex, three-dimensional nature of an urban area and the challenges inherent in directly measuring urban energy storage flux (AQs) have led to this heat flux being an understudied component of the urban surface energy balance. The primary goal o f this research is to compare the relative ability o f several different methods to estimate the magnitude and temporal variation of AQs in an urban environment against those obtained as in observed energy balances. The investigation is based on results from a site in the center o f Marseille, France. This locale provides an ideal environment for this study, because it has a warm, dry climate (hence sensible heat dominates) and massive urban development (hence a large thermal mass), so that heat storage is likely to be a significant part of the overall surface energy balance. Estimates of AQs obtained from tower-mounted instruments (the energy balance residual approach) are compared to results from a parameterization scheme (Objective Hysteresis Model, OHM) , a local-scale numerical model (Town Energy Balance, TEB) , and a bulk heat transfer approach (Thermal Mass Scheme, TMS) . Meteorological and urban construction data are used as inputs to the methods. Two-dimensional and three-dimensional formulations of the O H M do not sufficiently handle surface-atmosphere sensible heat exchanges in this highly urbanized and windy environment while T E B shows good agreement with the residual approach. T M S values are comparable to those from the other two methods but the laborious nature of the approach renders it impractical in such a complex setting. TEB ' s good performance at this site, as well as at other dry urban settings (central Mexico City and a light industrial site in Vancouver, British Columbia), suggest it possesses promise as a basis for further analyses and sensitivity tests designed to better understand the impacts of varying f low regimes, building geometry and materials on local-scale urban surface-atmosphere energy exchanges. Such an analysis was performed and the results reveal that the wind regime plays a dominant role in surface-atmosphere energy partitioning in this environment. Varying urban geometry and surface radiative properties also result in appreciable changes whereas alterations to surface thermal parameters generate the least impact. ii TABLE OF CONTENTS ABSTRACT ii T A B L E OF CONTENTS iii LIST OF TABLES vi LIST OF FIGURES viii LIST OF SYMBOLS AND ACRONYMS xiii ACKNOWLEDGEMENTS xvi 1 INTRODUCTION 1 1.1 Population and Urbanization Trends 1 1.2 The Urban Surface Energy Balance 1 1.3 Urban Energy Storage Flux 4 1.4 Estimation of Urban Energy Storage Flux , 5 1.4.1 Residual Method 8 1.4.2 Parameterization 10 1.4.3 Numerical Model ing 14 1.4.4 Thermal Mass Scheme 18 1.5 Research Obj ectives .23 2 METHODS 24 2.1 Research Site 24 2.2 Tower Measurements ,26 2.2.1 Observation Strategy 26 iii 2.2.2 Tower Instrumentation 29 2.3 Surface Temperature Survey 33 2.3.1 Observation Strategy 33 2.3.2 Surface Temperature Survey Instrumentation 37 2.4 Flux Source Areas 40 2.4.1 Surface Cover Survey 41 2.4.2 Upwell ing Radiative Flux Source Areas 43 2.4.3 Turbulent Flux Source Areas .45 2.5 Generic Neighborhood Model 47 2.6 Data Acquisition/Processing 48 2.7 Weather Conditions During the Observation Period 50 2.8 Comparative Statistics • 55 3 R E S U L T S 57 3.1 Observed Flux Behavior 57 3.1.1 Radiation Balance 57 3.1.2 Surface Energy Balance.. 59 3.1.3 The Context o f Marseil le's Energy Regime 64 3.2 Objective Hysteresis Model 70 3.2.1 O H M Input Parameters 70 3.2.2 O H M Results 71 3.3 Town Energy Balance Model 76 3.3.1 T E B Input Parameters and Initialization 76 3.3.2 T E B Results : 80 iv 3.2.2.1 Modeled Flux Behavior 80 3.2.2.2 Modeled Energy Storage Flux 82 3.4 Thermal Mass Scheme 87 3.4.1 Observed Surface Temperatures :. 8 7 3.4.1.1 Observed Canyon Temperatures 88 3.4.1.2 Observed Roof Temperatures 94 3.4.2 T M S Results 97 3.4.3 Sensitivity o f the T M S 100 4 MODELED SENSITIVITY OF AQs 104 4.1 Sensitivity to Wind Speed 104 4.2 Sensitivity to Urban Geometry 114 4.2.1 Sensitivity to Canyon Aspect Ratio (HIW) 114 4.2.2 Sensitivity to Bui lding Plan Area 117 4.3 Sensitivity to Surface Radiative and Thermal Parameters 120 5 CONCLUSIONS 126 5.1 Summary of Conclusions 126 5.2 Recommendations for Further Research 130 REFERENCES 133 A INFRARED THERMOMETER CALIBRATION 143 B STAR MODEL OVERVIEW AND INPUT 144 C EVEREST INFRARED THERMOMETER SITE INSTALLATIONS 147 D ROOF SURFACE TEMPERATURE PARAMETERIZATION SCHEME 154 LIST OF TABLES 1.1 Measured values of QQ and AQs for different surfaces .. 6 1.2 Summary of available a coefficients for urban surface types 14 2.1 Heights at which tower instruments were mounted 30 2.2 Instrumentation mounted at the C A A tower site 32 2.3 Surface emissivities for common urban surface materials. 34 2.4 Description of surfaces and periods of measurement for the fixed IRT instrument network in Marseille .3 8 2.5 Up welling radiative flux source areas for the down-facing radiation sensors on the C A A tower for different view factors. 44 2.6 E S C O M P T E wind categorizations r 52 3.1 Hourly averages of mean daytime and daily observed fluxes and flux ratios for all-sky conditions under synoptic and sea-breeze conditions in Marseille 64 3.2 Hourly averages of mean daytime and daily observed fluxes and flux ratios for all-sky conditions at seven North American cities and Marseille 66 3.3 Statistical evaluation of the goodness of fit o f the hysteresis pattern for a variety of urban study sites 67 3.4 Fraction of plan area surface cover by increasing area that is vegetated, for various urban sites 68 3.5 O H M a coefficients selected for use in the Marseille formulation 71 3.6 Statistical performance of the O H M in Marseille and at other urban sites 75 3.7 Cover fractions and T E B input parameters for the static modeling domain used in the present study 78 3.8 Thermal properties for roofs, walls, and roads used in T E B for the C A A site 79 3.9 Performance statistics for mean values of the surface energy balance modeled by T E B 80 3.10 Summary of performance statistics of T E B for heat fluxes at the Mexico City, Vancouver light industrial, and Marseille sites 86 vi 3.11 Summary of diurnal measured surface temperatures ranges of all canyon facets 90 3.12 Statistical summary of canyon surfaces temperatures derived from a hand-held IRT street-level survey 94 3.13 Summary of diurnal range of measured surface temperatures of five roof surfaces ; 96 4.1 Summary table of T E B A Qs simulations run with wind speed modifications.... 107 4.2 Summary of average overall, daytime, and nighttime AQs bias over varying wind speed regimes , 107 4.3 Coefficients fitted to a second-order polynomial equation which describes the dependence of A Qs on modified wind speeds 109 4.4 Summary table of T E B QH simulations run with wind speed modifications 112 4.5 Summary of average overall, daytime, and nighttime QH bias over varying wind speed regimes. 112 4.6 Summary of daytime, nighttime and diurnal average flux ratios over varying wind speed regimes 113 4.7 Sensitivity analysis to varying canyon HIW for the Marseille C A A site 115 4.8 Sensitivity analysis to varying plan area of built surfaces for the Marseille C A A site 118 4.9 Sensitivity analysis to varying surface radiative parameters (albedo and emissivity) for the Marseille C A A site 121 4.10 Summary of sensitivity analysis to varying surface thickness ,...122 5.1 Summary of results and comparative statistics for three methods used to approximate AQs at a site in the city center o f Marseille, France ; 128 A . 1 IRT calibration results 143 B. 1 STAR inputs used in the formulation of O H M a coefficients for clay tile roofs. 146 C. 1 Description of each site within the Everest infrared thermometer network used in Marseille 148 D. 1 Equation coefficients and performance statistics o f the parameterization scheme to estimate unmeasured roof surface temperatures 154 vii LIST OF FIGURES 1.1 Schematic illustration of various definitions of the urban surface 3 1.2 Schematic illustration of energy balance fluxes within an urban building-air volume 7 1.3 The a regression coefficients as descriptors of the hysteresis relationship between AQswaAQ* 12 1.4 Relative size and structure of the urban atmosphere over the meso- and local scales ...15 1.5 Representation of the individual surfaces for which energy budgets are resolved by T E B ...16 1.6 Representation of approximate dimensions of the local-scale 'box' used in formulating a thermal bulk mass approach of estimating AQs 21 2.1 Map of Provence showing the location of the study site.......; 24 2.2 Photo of the city center of Marseille, depicting the densely built-up urban environment of uniform building heights and little greenspace. 25 2.3 Example of an urban canyon near the C A A site 26 2.4 Aerial view of Marseille showing the location of the C A A instrumented tower..27 2.5 Schematic showing the blending height above which the influence of individual surface elements are integrated into a local-scale signal 28 2.6 Photograph of the C A A tower in the 'up' position 30 2.7 Mean and standard deviation of ratios of u*ilu*\ and QmlQm 31 2.8 Example of an image created from ground-based infrared scanner thermometry 35 2.9 Instrument array at the C A A gravel roof site 36 2.10 Three Everest 1RT's mounted on a rooftop and measuring surface temperatures of two walls and a road. 37 2.11 Schematic o f the complete block over which hand-held IRT sampling was conducted : 39 viii 2.12 Aerial photograph of area surrounding the C A A site, overlain with square grids depicting plan area occupied by clay tile roofs 42 2.13 Daytime turbulent and up well ing radiative flux source areas 46 2.14 One in a series of infrared images used in surface temperature analysis 49 2.15 Ensemble time series plots of weather variables over the course of the IOP 53 2.16 Ensemble averages of the dimensionless surface-scaling parameter <;.... 54 3.1 Time series of individually measured surface radiative flux densities 57 3.2 Ensemble 15-minute averages of radiation flux density terms under all-sky conditions 58 3.3 10-day time series of measured surface energy balance terms at the C A A site over all sky and wind conditions 59 3.4 Ensemble hourly averages of surface energy balance terms over all conditions...60 3.5 Diurnal ensemble patterns of QH/Q*, QE/Q* and t±Qs/Q* under all conditions...58 3.6 Observed hysteresis relationship between sensible heat flux and net radiation and energy storage flux and net radiation 61 3.7 Ensemble plot of A Qs versus QH- 62 3.8 Ensemble mean diurnal values of measured (OB S) and modeled (OHM) storage heat fluxes at the Marseille site. 72 3.9 Scatter-plot of hourly modeled vs. measured storage heat flux at the C A A site...73 3.10 Ensemble hysteresis loops of energy storage fluxes derived from measurements (residual) and O H M 74 3.11 Hourly 2D and 3D O H M bias ( O H M - OBS) plotted against wind speed 76 3.12 Ensemble comparisons between the observed (OBS) and simulated (TEB) surface energy balance ; , 81 3.13 Time series of simulated (TEB) and observed (residual) energy storage flux 83 3.14 Scatter-plot o f hourly modeled (TEB) vs. measured (residual) storage heat flux at the C A A site 83 ix 3.15 Ensemble mean measured (OBS) and simulated (TEB) hysteresis loop at the C A A site 84 3.16 Ensemble mean measured (OBS) and modeled (TEB) hysteresis relationship between heat storage flux and sensible heat flux 84 3.17 Time series of observed wall temperatures. 88 3.18 Time series of canyon street surface temperature 90 3.19 Ensemble plot of average surface temperature of the three facets which comprise a north-south canyon 91 3.20 Ensemble plot of average surface temperature of the three facets which comprise an east-west canyon 92 3.21 Average ensemble canyon temperature, weighted according to the fraction of surface area occupied by each canyon component 92 3.22 Time series of measured and parameterized roof surface temperatures 95 3.23 Ensemble average measured roof surface temperatures over the 10-day 96 3.24 Ensemble mean diurnal cycle of AQs estimates from the Thermal Mass Scheme (TMS) : .98 3.25 Scatter-plot of hourly measured (Residual) vs. modeled (TMS) storage heat flux at the C A A site ....99 3.26 Ensemble hourly AQs o f the three built surface (roads, roofs, and walls) resolved by the T M S t 100 3.27 Diurnal ensemble of T M S bias with changes to wall thickness and composition... 101 3.28 Diurnal ensemble of T M S bias with changes to road thickness and composition 102 4.1 Time series of reference wind speed (U) and the range of modified wind speeds ( 0.25*U, 2*U) used to force T E B 105 4.2 Ensemble time series o f simulated A s u n d e r a weaker f low regime than measured... 105 4.3 Ensemble time series of simulated A s u n d e r a stronger f low regime than measured 106 4.4 Ensemble diurnal time series of simulated AQs bias with weaker relative wind speeds •. 108 4.5 Ensemble diurnal time series of simulated AQs bias under greater relative wind speeds 108 4.6 Dependence of AQs on modified wind speed over the diurnal, daytime, and nighttime periods 109 4.7 Ensemble time series of simulated QH under reduced wind speeds 110 4.8 Ensemble time series of simulated QH under enhanced wind speeds I l l 4.9 Average energy storage flux ratio AQS/Q* versus ( C T M O D / ^ R E F ) , for the nocturnal, daytime, and diurnal periods 113 4.10 Average turbulent sensible heat flux ratio QHIQ* versus ( C 7 M G D / C / R E F ) , for the nocturnal, daytime, and diurnal periods 114 4.11 Ensemble diurnal time series of simulated AQs bias under varying canyon aspect ratio (HIW) 116 4.12 Ensemble diurnal time series of simulated AQs resulting from modifications to the plan area occupied by buildings and impervious ground .....117 4.13 Diurnal ensemble time series of simulated AQs bias for varying plan area of buildings 119 4.14 Diurnal time series of simulated A Qs biases for varying surface albedos 121 4.15 Diurnal time series of simulated AQs biases for varying surface thickness 123 C I Aerial photograph of central Marseille, showing the relative locations of the five sites at which stationary IRTs monitored surface temperatures 149 C2 Everest array at the school site 150 C3 The two IRT arrays at the Croix Rouge site 150 C4 The IRT array on the roof of the Licensing Building, south of the C A A site.... 151 C5 The tripod array on the C A A roof 151 C6 Three IRT's measuring surface temperatures of an east-facing wall and a north-and south-facing new clay tile roof at the C A A site 152 xi C7 A n IRT at the C A A site measuring surface temperature along the spine of a roof with clay tiles slanted in the east and west directions 152 C8 The IRT array at the Jurexfi building 153 C9 Wal l surfaces (south-facing and east-facing) sensed at the Jurexfi site 153 xii LIST OF SYMBOLS AND ACRONYMS A area 3 1 C heat capacity (J m" C" ) C A A Cours d'Appel Administrative c ; specific heat of material i (J kg" 1 K" 1) CSI Campbell Scientific, Inc. d Wilmott agreement factor dz material thickness D roughness element spacing (m) F view factor (%) F O V instrument field-of-view F S A M Flux Source Area Model HIL building aspect ratio HIW canyon aspect ratio IOP Intense Observation Period I RGA infrared gas analyzer I SBA Interface Soil-Biosphere-Atmosphere Model IRT infrared radiation thermometer ISL inertial sublayer k thermal conductivity (W m"1 K"1) K1 reflected shortwave radiation (W m") K\. incoming shortwave radiation (W m") L Obukhov scaling length (m) L A T Local Apparent Time U latent heat o f vaporization (J kg"1) L\ upwelling longwave radiation (W m" ) L[ downwelling longwave radiation (W m") M A E mean absolute error M B E mean biased error O H M Objective Hysteresis Model Q* net all-wave radiation flux density (W m"2) AQA net heat advection (W m") QE turbulent latent heat flux density (W m"2) QF anthropogenic heat flux density (W m") QG 2 ground heat flux density (W m") QH turbulent sensible heat flux density (W m") AQs net storage heat flux (W m ' ) r radius (m) R H relative humidity (%) R M S E root mean squared error R S L roughness sublayer Ta air temperature (°C) To surface temperature (°C) Tr apparent surface temperature (°C) T E B Town Energy Balance Model T M S Thermal Mass Scheme u horizontal (x-plane) wind speed (ms ' 1 ) u* friction velocity (ms ' 1 ) U C L urban canopy layer v horizontal (y-plane) wind speed (ms ' 1 ) V P D vapor pressure deficit (hPa) w vertical (z-plane) wind speed (m s"1) w* free convection velocity (m s"1) Y D day of the year z* blending height (m) Zd zero-plane displacement length (m) ZH mean building height (m) z0 roughness length (m) zs sensor height (m) a albedo B Bowen ratio (= QHIQE) E0 infrared surface emissivity Pt density o f material / (kg m"3) 0 standard deviation a Stefan-Boltzmann constant (W m"2 K" 4) av standard deviation of lateral wind speed fluctuations (m T momentum flux (N m"2) q surface layer scaling parameter (dimensionless) A C K N O W L E D G E M E N T S This work would not have been possible without the Financial and academic support of my supervisor, Dr. T. R. Oke. Besides overwhelmingly fulf i l l ing the role of supervisor and mentor, T im also served as an understanding and good-natured friend to me throughout these three years, a role which I cherish greatly. I endeavored to enter graduate school not solely to learn the ins-and-outs of urban climatology but more importantly, to become a conscientious and thoughtful scientist. I can think of no better person with whom to start down that path than Tim. Grateful thanks are also due to my committee members, Drs. Ian McKendry and T. A . Black, for their helpful feedback on this thesis. Many others served important roles over the course of this project, contributing critical data and patient assistance, both in and out of the field. Dr. James Voogt (University of Western Ontario) provided imperative guidance and data concerning everything related to surface temperatures, seemingly endless ribbing, and (not nearly enough) beef. Prof. Sue Grimmond (Indiana University) kindly provided helpful comments, direction and everything energy balance-related. Dr. Jenny Salmond's (University of Birmingham) transition from fellow U B C graduate student to confident researcher to whom I turn for guidance did much to inspire and steer my development and for that, I am happy to endure her quick "wit." Deep thanks also to Dr. Valery Masson and Aude Lemonsu (Centre National de Recherches Meteorologique, Meteo-France) for the use of the Town Energy Balance model and the helpful support in its implementation. I am also indebted to the large collection of scientists involved in the E S C O M P T E - C L U campaign, in particular: G. Pigeon, J.P. Lagouarde, M . Irvine, P. Mesteyer, and N. Long. Thanks also to the Cours d'Appel Administrative de Marseille, College Longchamp, Croix Rouge, and Jurexfi, all of whom provided building access for field measurements. Dr. T. A . Black of the U B C Soi l Science Department made available the temperature calibration facility, with the kind help of E lyn Humphreys. A lso vitally important in the completion of this thesis are the amazing support staff members in the Departments of Geography and Earth and Ocean Sciences at U B C , whose laughter, resourcefulness, and helpfulness allowed me to maintain my sanity on a daily basis. I'd also like to acknowledge my fellow Okies that have passed through the lab in these ^  three years. In particular, thanks to Steph Meyn, whose supportive recommendations ultimately convinced me to come to U B C and also, for conducting STAR simulations used in this project. Andres Soux, Kathy Runnalls, and Steph Meyn served in a variety of important roles that lent to my academic development, including (but not restricted to): becoming my surrogate Geography buddies, gin horticulturalists, and truly resourceful and spirited colleagues. It goes without saying that they are all due scanned copies of this work. And where would I be without the expert ninja field assistance of Phoebe 'Tito ' Jackson? Marseille was much more bearable with her admirable dart-throwing and Guinness consuming skills. More importantly, this thesis would not be nearly as voluminous without her T E B simulations and all-around brilliant know-how. xv i Funding for this research has been provided to Prof. T. R. Oke by Natural Sciences Engineering Research Counci l of Canada, the Canadian Foundation for Climate and Atmospheric Sciences, Centre National de Recherches Meteorologique (Meteo-France), and Centre National de Recherches Scientifiques de France. Personal funding was provided through University o f British Columbia Teaching and Research Assistantships in the Department o f Geography. And of course, none of this would have been possible without the support, laughter, love, and patience of my parents, sister, grandmother, and extended family and friends. Many have played important and necessary roles in my life during these three years and I know that no part of this experience could have been as rewarding or fun without them. xv i i CHAPTER 1; Introduction 1.1 Population and Urbanization Trends The Earth's population of over six bi l l ion people is expected to continue to increase for at least a century. O f the current population, it is estimated that nearly one-half live in urban areas (Oke, 1991). Assuming current population and migration trends persist, the sheer number of urban dwellers is expected to double in the next twenty-five years, raising the proportion of the global population who live in urban areas to three-fifths (UNFPA, 1999). Most of the expected urban growth wi l l occur in developing regions, such as Afr ica and Asia, while in already-developed countries a huge urban shift occurred a century ago, the result of which is that now over 70% of the population in those areas live in cities. A current trend in developed areas is a population shift from concentrated urban areas to sprawling/metropolitan suburban areas or to smaller- or intermediate-sized cities. Worldwide, therefore, the number o f cities is increasing, as are their populations and encompassing areas (World Resources, 1996). These population patterns dictate that natural resources are disproportionately and increasingly consumed in and around urban areas. Accordingly, recognition of and concern about environmental and climatic problems resulting from urbanization also increases. 1.2 The Urban Surface Energy Balance The alteration of surface and atmospheric environments caused by urbanization leads to the creation of distinct urban climates. Climatic features unique to urban areas such as the urban heat island, urban-induced wind circulation, and precipitation enhancement downstream of urban areas are well-documented (for example, Lowry, 1998; Vakeva et al.,. 1999; Shephard et al, 2002). Such urban climate effects are 1 ultimately due to differences in the budgets of mass, heat, and momentum between the urban and pre-urban landscape (Lowry, 1977 and Oke, 1982). In order to fully understand urban climate effects it is therefore necessary to develop an understanding of the urban surface energy budget and apply that knowledge to the wealth of existing information about the boundary layer meteorology and climatology of rural areas (Grimmond and Oke, 1995). With this understanding wi l l inevitably come enhanced modeling capabilities and applications, allowing the urban environment to be better represented in numerical meso-scale models (Taha, 1999). In addition, more general concerns such as public health issues, urban planning, architectural design, and resource management may be more appropriately addressed (Oke et al., 1989). In order to gain an appreciation of urban surface-atmosphere interactions, it is important to first define the urban surface. Doing so is pivotal, because surface properties control the partitioning of the net radiant energy and therefore greatly impact the behavior of the atmospheric layer adjacent to the surface. Depending on the scale and process(es) under study, there are a variety of definitions of the urban surface (Grimmond, 1988; Voogt, 1995). Various representations are shown in Figure 1.1 but the most relevant surface definition is that of the complete urban surface (A), which considers all elements that serve as a boundary between the surface system (building walls, roof, trees, etc.) and the atmosphere (Voogt and Oke, 1997). 2 1 Figure 1.1 Schematic illustration of various definitions of the urban surface. Source: Voogt and Oke (1997). Once the urban surface is sufficiently assigned, its energy balance can be expressed theoretically as: Q*+QF=QH+QE+AQS+AQA ( W i n 2 ) , (1.1) 3 where Q*is the net all-wave radiant energy flux, QF'IS the anthropogenic heat flux, QH is the sensible heat flux, QE is the latent heat flux, AQs is the net storage flux, and AQA is the net horizontal heat advection (Schmid et al, 1991). In most contemporary urban climate studies, Q* is measured directly by radiometry and QH and QE are simultaneously directly measured using eddy covariance techniques. The anthropogenic heat flux (QF) is typically not directly measured, as it is likely implicitly included in the measured QH, QE, and Q* terms as plumes of warm air and water vapor released by vehicles and building vents, infrared radiation from the warmer urban setting, and alterations to the heat conducted into and out of buildings. Net horizontal heat advection, AQA, can be considered negligible i f energy budget measurements are conducted at carefully selected sites with an ideal extensive horizontal homogeneous fetch. There is presently no accurate, reliable method o f directly measuring AQs in an urban environment so this term is approximated by various alternative methods. 1.3 Urban Energy Storage Flux Of particular relevance in the urban environment is the role of the net storage heat flux, AQs, which has been shown to account for over 50% of daytime net radiation at highly urbanized sites such as downtown St. Louis, Missouri (Ching, 1985) and central Mexico City (Oke et al, 1999). The storage heat flux is defined as the net uptake or release of energy from an urban system. It includes latent and sensible heat changes in the air, buildings, vegetation, and ground within the layer and spatial scale of interest (Grimmond et al, 1991). This surface flux depends on urban surface materials and because of the differences in thermal properties and structural configurations between 4 urban areas and those of rural or suburban areas, the nocturnal energy release from storage is a major contributor to the urban heat island effect (Sakakibara, 1991; Johnson et al., 1991). Knowledge of AQs is required in a host of other applications; for example, to model evapo-transpiration, the convective sensible heat flux, and boundary layer growth (Grimmond, 1988; Taha, 1997; Roth and Oke, 1994). It is important to recognize the distinction between AQs and QQ. while AQs describes changes in heat storage content within layers or volumes, Qc traditionally refers to heat conduction through a plane in a simple surface (typically soils, concrete, snow, etc.; Camuffo and Bernardi, 1982). As is the case with both storage terms, the ability of a medium to conduct and store energy through and within single and multiple layers depends on the density and heat capacity of the material(s). In turn, these parameters depend on material composition, temperature, and moisture content (Dol l et al., 1985). The presence of moisture is certainly more pertinent when permeable materials (snow, vegetation, and porous surficial materials such as clay, sand, and soil) are considered and is less of a mitigating feature with impermeable surfaces (e.g. concrete, asphalt blacktop). The above factors all combine to impact the thermal response of a surface to a given heat flux (Oke, 1988). It follows, then, that differences in surface material composition between a simple environment (rural) and a complex setting (urban) can be significant enough to warrant very different thermal responses. A n example of the range of measured values of QG and AQs for different surface types is shown in Table 1.1 1.4 Est imat ion of Energy Storage F lux Given the complex, three-dimensional nature of the urban surface and the inherent challenges associated with the direct measurement of AQs, the storage heat flux has been 5 Table 1.1 Measured values of go and AQS for different surfaces. Surface Source QG or AQs (MJ m 2 day1) Bare soil Stanhill (1965) 0.2 Whitened soil Stanhill (1965) 0.2 Coniferous forest Gay and Stewart (1974) McNaughton and Black (1973) -0.2-1.0 Tropical ocean Holland (1971) 1.2 Melting snow Granger and Male (1978) -0.11 Grassland Cleughand Oke (1986) 0.3 Suburban (Vancouver, B. C.) Cleughand Oke (1986) 1.2 Urban (daytime, Mexico City) Oke et al. (1999) 5.03 Urban (24 hr, Mexico City) Oke etal. (1999) -0.54 an understudied component o f the urban surface energy balance (Grimmond et al, 1996). Unl ike the biometeorology community's methods of measuring turbulent sensible and latent heat fluxes, the straightforward ways of obtaining energy storage fluxes in a simple environment cannot be easily adapted to the complex urban system (van Loon et al, 1998). Heat flux plates or integrating thermometers cannot be realistically deployed within an urban setting in order to represent the myriad of surface types and orientations of which urban areas are constructed. Attempts to sample these surfaces have been conducted; however, they are restricted to subsets of distinct urban surfaces such as asphalt, concrete, and various roof assemblies (Terjung et al, 1971; Yap, 1973; Taesler, 1978; Do l l et al, 1985; Kerschgens and Hacker, 1985; Kerschgens and Drauschke, 1986; Kerschgens and Kraus, 1990; Berdahl and Bretz, 1997; Anandakumar, 1999; Asaeda and Ca, 2000, Meyn, 2000). The heterogeneous nature of the urban surface has necessitated the use of a soil-building-air volume concept, to compensate for the logistical impracticalities of direct energy storage flux measurements (Bornstein and Craig, 2002). This model can be thought of as a volumetric box (Fig. 1.2), the bottom of which is the depth below the 6 surface at which zero net heat flux over the period of concern occurs. The top of this theoretical box is the height of the urban canopy layer (UCL) , which is typically chosen to be a height just above roof level at which roughness elements within the underlying layer contribute to the overall energy balance of that layer (Kerschgens, 1990). Such a formulation has the useful benefit of allowing the complex spatial arrangement of individual energy sources and sinks to be neglected, since only energy fluxes through the top of the volume need to be considered (Oke, 1988). The net storage heat flux, therefore, includes heat conduction into or out of, and temperature changes by, every component of the volume (e.g. roofs, walls, roads, vegetation), as wel l as latent and sensible heat changes in the air Volume. Figure 1.2 Schematic illustration of the fluxes in the energy balance of an urban building-air volume. Source: Oke (1987), p. 275. Since the direct measurement of AQs is so impractical within the urban setting, it is not surprising that alternative methods that attempt to quantify the urban energy 7 storage flux have evolved. The description and theoretical framework of four such contemporary methods follows. 1.4.1 Residual Method A s a matter of convenience and partly for lack of a better method, the storage heat flux has been calculated as the residual to the energy balance equation. Disregarding the anthropogenic heat flux (on the grounds that it is incorporated in the other measured fluxes) and neglecting the net advective heat flux term, the theoretical energy balance equation (1.1) can be expressed in practical, or measured, terms as: Q*=QH + QE + AQS (Wm" 2), (1.2) where Q*, QH, and QE are measured directly using standard radiometric and eddy correlation techniques and AQs is calculated as the residual to the equation (Oke and Cleugh, 1987). While this method is obviously straightforward, its primary drawback is the accumulation of potential errors from both the neglected and remaining terms of (1.1) into the residual energy storage flux term (Ching et al., 1983). Potential contributions from anthropogenic heat flux QF and horizontal advection AQA are difficult to determine, as the magnitude of these fluxes depends on the spatial pattern of their sources. In urban areas, the most obvious origins of anthropogenic contributions are combustion from stationary (infrastructure) and mobile (roadways) sources. The plausible magnitudes of QF values found in the literature are on the order of a few tens of W m"2 (Grimmond and Oke, 1995; Oke et al., 1999). Since the instruments used to measure Q*, QH, and QE 8 l ikely sense most of the anthropogenic contributions to the radiative, convective, and conductive flux terms, also including an independent anthropogenic flux contribution would, in a sense, be "double counting" this source (Grimmond and Oke, 2002). It is therefore assumed that any anthropogenic contributions not sensed by the instruments are small enough to be negligible. This error, probably on the order of a few W m" , does, however, accumulate in the residual AQs estimation. To minimize the contribution of errors associated with potential horizontal heat and moisture transport, careful site selection is conducted to ensure the measurements are representative of the land use type selected for study and that spatial variability does not create large flux variability with changes in wind direction (Grimmond, 1988). To make certain this is the case, the measurement tower should ideally be placed in an area with extensive horizontal homogeneity and the flux sensors should be mounted at a height above the roughness sub-layer, so that spatial variability o f individual surface elements is no longer discernible (Oke, 1988). To determine possible advective contributions to a data set from the coastal city of Vancouver, British Columbia, Steyn (1985) performed a statistical analysis which showed that even under sea-breeze conditions, the proper positioning of an observation tower can eliminate any significant advective. flux contributions. Pigeon et al. (2003) also assessed potential advective contributions at a central site in Marseille, France, using field observations and simulations from the non-hydrostatic M E S O - N H model coupled with the Town Energy Balance Model (see Section 1.4.3). The authors found that advection of cool and moist maritime air occurred over the measurement site, resulting in an underestimation of the measured sensible heat 9 flux and an overestimation of the latent heat flux. These terms are thought to balance at this site, thus reducing any residual errors amassed in AQs. Random errors associated with turbulent flux measurements typically range on the order of 10% in the daytime for both the sensible and latent heat fluxes, while nighttime random errors are even larger (in percentage terms, but less in absolute terms, Mahrt, 1998). Standard net radiation measurement errors are usually smaller, around 5% (Offerle etal, 2002). In summary, the sum of errors that could potentially accumulate in the net storage flux term is probably in the range of about 15-25% (Grimmond, pers. comm.). 1.4.2 Parameterization Estimates of the urban storage heat flux have also been made via parameterization schemes, in which values of the storage heat flux are parameterized in terms of a point-source net radiation value and a description of the surface material characteristics. Such schemes are based upon the simple relation used in many meso-scale and global circulation models that express the proportion of available energy used to heat the substrate materials as a fraction of the net all-wave radiation, i.e. QQ = aQ*. Further refinement to this expression by Oke et al. (1981) used published relations between net all-wave radiation and heat storage flux for several urban surface materials and combined them in a composite equation which weighted the role o f each according to their plan coverage in the study area. It also incorporated an observed offset, resulting in the following expression: 10 AQs = £ a,(a,Q * +bi) (1.3) (=1 where a, is the fraction of urban area covered by the rth surface. This objective linear relationship performs satisfactorily for periods of a day or more, but because it does not accurately describe observed phase^ shifts between AQs and Q*, values at a smaller (hourly) resolution could not be satisfactorily modeled. Several studies of heat storage into and out of homogeneous surfaces have observed a non-linear relationship between the energy storage flux and net all-wave radiation (Fuchs and Hadas, 1972; Camuffo and Bemardi, 1982). The modeling scheme that seems to accurately describe this distinct hysteresis pattern between the radiative forcing and storage change for urban areas is the Objective Hysteresis Model (OHM) of Grimmond et al. (1991). O H M uses hourly-averaged values of net radiation along with surface properties to obtain an estimation o f energy uptake/release over a source area. The equation used in this parameterization scheme is: In this expression, the subscript / identifies n types of surfaces, such as roofs, walls, lawns, or roads. The time derivative of net radiation is approximated as 0.5[Q*t+j - Q*t-i]. Statistical coefficients, a\\, a^, and a^ are empirically derived from independent studies relating AQs to Q* over surfaces known to form the urban area (Figure 1.3). The coefficient a\ describes the mean slope of the dependence o f the storage energy flux on AQs = Y, auQ*+a2. n dQ* dt + a 3 , (1.4) 11 Figure 1.3 The a regression coefficients as descriptors of the relationship between AQs and Q*-Source: Meyn (2000) net radiation and is simply the magnitude of AQs when Q* becomes negative in the evening and positive the next morning. The parameter 02 indicates the degree and direction of hysteresis between AQs and Q*. When #2 is positive, the diurnal peak in AQs precedes the peak in Q*. Correspondingly, when 02 is zero, there is no hysteresis and the AQs and Q* curves are exactly in phase while a negative (X2 value indicates a peak in Q* preceding the peak in A Qs. The larger the value of 02 the greater is the hysteresis loop showing the asymmetry between the heating and cooling portion of the daily cycle. To employ this equation at a specific site, an inventory of the areal coverage of different surface types likely to be in the flux source area must be compiled. From the literature, a list of the a\, «2, and a-$ coefficients for each of the surface types is composed. Fol lowing that, site-specific coefficients are calculated by weighting coefficients of each surface type according to the proportion of total area occupied by that surface type. These values may be dynamic; varying with wind direction and stability. The appropriate source area location and surface make-up can be calculated using a source area model (e.g. Schmid, 1997) in combination with an urban geographic information system (e.g. Grimmond and Oke, 1999). The final step is to estimate the storage flux from net radiation by summing 12 the contributions by each surface type over all surface types within the source area of interest. Taha (1999) used this approach in a meso-scale model and found it to significantly improve the storage estimation and the resulting heat island magnitude. Grimmond and Oke (1999a) evaluated O H M using residual energy storage estimates from seven North American cities and showed that it performed well in both urban (downtown) areas and suburban areas in neutrally-stable conditions with light wind (< 2 m s"1). The authors point out, however, that problems in O H M ' s performance could be explained partly by the equation's potential conflict in scale; that is, the coefficients derived for individual surface types use a Q* value which is specific to only that surface type. Whereas the net radiation measurement used in the overall summation is a local-scale average, often taken tens of meters above roof level. Schmid et al. (1991) showed that although differences between net radiation measured at point sites versus a fixed site 25-30 m above the surface were small (less than 5%), bias is l ikely introduced i f the source area is made up of surfaces with contrasting radiative exchanges. Therefore, O H M potentially does not allow for spatial or temporal variability o f convective fluxes, which are driven at least partly by net all-wave radiation. Variabil ity in convective fluxes could also be attributed to differences in surface moisture, wind speed, and synoptic conditions within the study site, which are also factors not explicitly resolved by O H M . Another drawback to the O H M scheme is the unfortunate lack of available data representing the relationship between AQs and Q* for certain surface types. In particular, Grimmond and Oke cite not only the scarcity o f rooftop data, but also the significant 13 difference between them. A summary of available measured a coefficients for different urban surface types is presented in Table 1.2. Table 1.2 Summary of available a coefficients for urban surface types. Surface type City Material Source « 2 (h) a 3(W m"2) Roof Vancouver tar and gravel Yap (1973) 0:17 0.1 -17 Uppsala not specified Taesler(1980) 0.44 0.57 -28.9 Kyoto membrane and concrete Yoshida et al. (1991) 0.82 0.34 -55.7 Vancouver gravel Meyn (2000) 0.28 1.3 -34 Vancouver asphalt shingles Meyn (2000) 0.13 0.20 -6 Vancouver tar and gravel Meyn (2000) 0.20 0.83 -19 Paved/Impervious concrete Doll et al. (1985) 0.81 0.48 -79.9 concrete Asaeda and Ca (1993) 0.85 0.32 -28.5 asphalt Narita et al. (1984) 0.36 0.23 -19.3 asphalt Asaeda and Ca (1993) 0.64 0.32 -43.6 Vienna asphalt -summer Anandakumar (1999) 0.72 0.54 -40.2 asphalt -winter Anandakumar (1999) 0.83 -0.83 24.6 Greenspace mixed forest McCaughey (1985) 0.11 0.11, -12.3 short grass Doll et al. (1985) 0.32 0.54 -27.4 Soil bare soil Novak (1981) 0.38 0.56 -273 bare soil - wet Fuchs and Hadas (1972) 0.33 0.07 -34.9 bare soil - dry Fuchs and Hadas (1972) 0.35 0.43 -36.5 soil Asaeda and Ca (1993) 0.36 0.27 -42.4 1.4.3 Numerical Modeling Attempts have also been made to model the energy storage flux within the urban environment, most often using one- or two-dimensional heat transfer approaches. Unfortunately, however, most of these efforts have been performed at the micro-scale (see, for example, Terjung and O'Rourke, 1980; Sievers and Zdunkowski, 1985; Asaeda and Ca, 1993; Mills 1997; Arnfield and Grimmond, 1998; Arnfield et al, 1998), which is relevant to buildings or urban canyons, rather than at the local scale. The local-scale (102 - 104 m) considers processes that are representative of the integrated response of an array of buildings, vegetation, and paved surfaces making up the urban district (Fig. 1.4). At this scale, spatial variability across a city reflects different neighborhoods with differing a) Mesoscale "dome" Urban "dome" < ; £ V \ C f r y ; -P B - I l l l l l l l l l l * mm Rural Urban Rural b) Mesoscale "plume" 1 f Urban " p l u m e " ' Mixing layer " • , / / UBL , L_ * >^ I 0 i i i i l i im. Surface layers R u i a l B L Rural Rural Urban c) Local scale A d) Microscale Inertial Surface sublayer layer M i l l i Roughness J..... sublayer U C L ^ _ d) UCL I Roughness _ j r ' sublayer 0 • • • • Figure 1.4 Relative size and structure of the urban atmosphere over (a) the meso-scale and (b) local scale. Source: Modified after Oke (1997). 15 combinations of land cover (built and vegetated) and structural configurations. The local-scale represents the upper and lower boundary for micro- and meso-scale models, respectively. Knowledge of local-scale variability is increasingly important in meso-scale modeling, as these models are gaining enhanced spatial resolution (Taha, 1999). It follows, then, that the need to represent the coupling between the urban surface and the atmosphere in meso-scale models is becoming more pertinent. A recent local-scale model, the Town Energy Balance (TEB) model (Masson, 2000), seeks to couple the micro- and meso-scales and to accurately represent the urban energy budget in meso-scale atmospheric models (Figure 1.5). The T E B scheme performs coupling between the urban surface and atmosphere and its primary aim is to simulate turbulent fluxes into the atmosphere at the lowest level o f a meso-scale Q, Hroof U T Q _ n t-t 3 ® H top ® F industry J « £ / o p Qf/ncfcraftJ, wall t t JL_. Q f traffic f traffic ^Hroad a , § I A t m o s p h e r i c m o d e l TEB Figure 1.5 Representation of the individual surfaces for which energy budgets are resolved by TEB. The scale on the left shows the model's scale, relative to atmospheric layers and meso-scale models. Source: Modified after Masson et al., 2002. 16 atmospheric model, i.e., the atmospheric model views the constant flux layer as its lower boundary. Rather than representing urban areas as a bare soil or concrete plate, as is currently the case with most atmospheric models, T E B uses local canyon geometry to simulate the effects produced by the presence of buildings. With the exception of Mart i l l i et al.'s (2002) urban exchange parameterization, no other scheme explicitly addresses canyon effects on the overall surface energy budget. T E B can be run on its own for highly urbanized sites or it can incorporate Noilhan and Planton's (1989) Interface Soil-Biosphere-Atmosphere (ISBA) scheme for vegetated urban areas. The model is forced with atmospheric and radiation data from above-roof level, which can be observed or are the output from a meso-scale forecast model. T E B incorporates detailed representations of the urban surface to simulate individual energy balances for walls, roads, and roofs. This allows for many of the physical effects associated with the urban heat, mass, and momentum balances to be accurately reproduced including: canyon radiative trapping (both shortwave and longwave), niomentum fluxes, turbulent sensible and latent heat flux exchanges, energy storage fluxes, and even water and snow interception. When T E B was coupled with a meso-scale atmospheric model (Meso-NH, Lafore et al., 1998), Lemonsu and Masson (2002) were able to simulate the nocturnal urban heat island, atmospheric humidity and temperature (to within 1 K ) during an anti-cyclonic summertime period in the Paris area. Masson (2000) validated the radiative portions of T E B against published data from Nunez and Oke (1976, 1977) as wel l as simulated output from the Surface Heat Island Model (SHIM) of Johnson et dl. (1991). T E B was shown to handle the evolution of nocturnal fluxes and road surface temperatures well, 17 with differences in net radiation of less than 10 W m" and surface temperatures to within 1 K of measured values. In addition, when validated with data from A ida (1982) T E B accurately captured the trapping of short-wave radiation by canyon geometry, which led to a fair parameterization of surface albedo. Whereas Masson (2000) was able to complete only sensitivity experiments of the T E B scheme itself, Masson et al. (2002) independently evaluated the performance of T E B using directly measured surface temperatures and surface energy balance fluxes for two dry urban sites - the downtown colonial district of Mexico City and a light industrial site in Vancouver. Because both sites contain a small amount of vegetation (less than 5% plan area), T E B could be assessed in the absence of a coupled vegetation scheme, because the latent heat flux term was negligible at both sites. In Mexico City and Vancouver, the model was shown to simulate net radiation to within less than 10 W m" as well as its partitioning into turbulent and storage fluxes within a few tens of W m" . Simulated surface temperatures were always within 3 K. 1.4.4 Thermal Mass Scheme A final estimate of the urban energy storage flux can be derived from basic concepts of heat conduction through an interface or surface plane. Fourier's Law describes the conductive heat flux (QG) through a single-layer medium (plane) of known thermal conductivity k, thickness (dz), and temperature difference (dT) between its upper and lower boundaries (Burmeister, 1993): Qo = -k^ (1.5) 18 The practical deployment of this relationship is generally uncomplicated: the single layer thickness is easily measured, thermocouples positioned on the lower and upper faces are typically used to calculate the temperature difference across the layer, and the material conductivity can be found from references in the literature concerning building materials. If the storage heat flux within a multi-layered system is to be resolved, the equation for the conductive heat flux density is the same for each of the individual layers. A n analytical solution to this scheme is complicated, however, by the thermal conductivities of each layer, as well as the inherent need to measure the temperature differences at each of the layer interfaces. Although an obvious experimental answer to such logistical hurdles could involve the use of heat flux plates, this, too, proves impractical, for invasive measurements between layers are often unacceptable or impossible (Meyn, 2000). Hedlin (1985) has shown that surface-mounted sensors are also not ideal for flux measurements, for they are prone to errors as a result of exposure to radiation, wind, and moisture. Since the direct measurement of heat f low through the interfaces between building material layers is not feasible, the most obvious solution is to adopt a more theoretical approach. This can be done with the continuity equation, which is used to calculate the temperature variation with depth and time within one layer (a volumetric description), rather than between different layers (Campbell and Norman, 1998): C ^ = - ^ £ (1.6) dt 8z 19 T 1 where C is the heat capacity (J m" °C" ) of the medium (which is equal to density, p, multiplied by specific heat, c), dTldt is the rate of temperature change (°C s*1), and dQcj/dz (J s" m" ) is the rate of change of heat flux density with depth through the medium. Integrating this equation to obtain QG results in a description of heat storage change within a volume made up of layers. Followdng the method used by McCaughey (1985) in a mature mixed forest in Ontario, Moore and Fisch (1986) in an Amazonian tropical forest, and Peikorz (1987) in Bonn, Germany, a Thermal Mass Scheme (TMS) can be constructed whereby surface temperature and building construction information are used to derive an approximation o f AQs. The equation used to describe this relationship is: Mis = ^ Q s , = ^-lc,— dvl ( 1 .7) where the index / identifies n surface types, At is the surface area o f cornponent i within the urban system, C, is the heat capacity (MJ m" K " ) of material i, dT/dt is the change in temperature over a given time period (K s"1) and dVj is the material volume through which the surface temperature wave propagates. When integrated over the volume of the urban canopy layer, this equation describes changes in heat storage out o f the top o f the U C L volume due to the net outcome of conductive fluxes into and out of each underlying component within the urban system. v The application of this equation to the physical environment requires that a survey of surface types (similar to that performed in implementing the Objective Hysteresis Model) be conducted. Further, it requires similar information on the construction of the 20 roofs, walls, roads, etc. so that equation 1.7 can be applied to all the layers involved in the change of heat content (i.e. storage). From this, a generic box model (Figure 1.6) of a neighborhood representative of the surface cover and structure (external and internal) within the source area can be derived. Wind -<*r-Depth of 'box' Measurement Height ^ •Length of'box'(102-104 m) Figure 1.6 Representation of approximate dimensions of the local-scale 'box' used in formulating a thermal bulk mass approach of estimating AQS. Source: Adapted from Masson et al., 2002. Once source area calculations have highlighted where suitable boundaries o f the local-scale theoretical box should be drawn, an inventory o f surface types and orientations is performed, with the use of aerial photographs o f the city as wel l as architectural plans. The volume of the box can then be described by weighting the fraction of each surface type and orientation present (i.e. E-facing limestone walls, N-S oriented streets, S-facing tile roof, etc.). The U C L volume, therefore, can be thought of as consisting o f four different volumes in various configurations: buildings, asphalt or paved areas, greenspace (trees, grass) between buildings, and the canopy layer air. Because the mass contributions of air and vegetation to the overall mass of the volume is 21 very small in comparison to the other "sol id" components of the urban canopy layer volume, these need hot be considered, but can be i f completeness is required (Kerschgens and Hacker, 1985). Surface temperatures and building material information for each facet can then be combined as described in equation 1.7 to' obtain an approximate value for the storage heat flux into or out of the top of the box within an appropriate time period. 22 1.5 Research Objectives This thesis aims to compare and contrast the aforementioned four methods used to estimate the local-scale net urban energy storage flux, AQs- To do so, the present work utilizes data gathered during a field campaign in the city center o f Marseille, France during the summer of 2001. This study locale provides an ideal environment in which to conduct this research, as its warm, dry climate and massive urban development dictates that energy storage is an essential element to the overall energy balance of the site. Approximations of AQs resulting from the parameterization (OHM) and modeling (TEB) schemes, in addition to the lesser-used bulk thermal mass-surface temperature approach (TMS), are compared to AQs estimates obtained from tower-mounted fast response instruments (the residual approach). A comprehensive statistical analysis of hourly, diurnal data from each outlined method wi l l determine their relative representativeness. For purposes of comparison, the 'measured' (i.e. residual) estimates wi l l be assumed closest to the 'truth'. In addition, T E B is employed to gain insight into the relative sensitivity of AQs to various meteorological, town geometry, and surface thermal/radiative forcing parameters. Details of the research methodology are outlined in Chapter 2. 23 CHAPTER 2: Methods This chapter describes the field site, research methods and analysis approach used in this project. It is designed to investigate the four methods outlined in Chapter 1, to approximate local-scale urban energy storage flux, AQS, in the city center of Marseille, France. 2.1 Research Site Marseille (43° 20' N, 5° 40'E) is situated on the Mediterranean coast in the French region of Provence (Fig. 2.1). It is the second largest city in France, with a metropolitan Figure 2.1 Map of Provence showing the location of Marseille. The relative location of the downtown study site is given in blue. 24 population o f over two mil l ion people and was the site o f a large multi-institutional European air quality study called E S C O M P T E (English translation: Field Experiment to Construct Models of Atmospheric Pollution and Emissions Transport), which took place 5 June - 15 July, 2001 ( Y D 156 - 196). E SCOMPTE ' s primary objective is the creation of a database appropriate to the task o f testing urban energy exchange schemes and high-resolution meteorological and chemistry-transport models (Cros et al., 2003). Because of data constraints, however, the present work focuses on results from an intense observational period (IOP) of 2-12 July, 2001 ( Y D 183 - 192). Marseille contains a densely built-up city center, characterized by low vegetation cover (-16% plan area), mostly as trees in courtyards and street boulevards. Buildings constitute the highest roughness elements within the study area and are, on average, approximately 16 m in height (Fig. 2.2, Grimmond et al., 2002b). The city center, around which the present research is focused, is comprised primarily o f 19 t h century massive Figure 2.2 View from the southwest of the city center of Marseille, depicting the densely built-up urban environment of uniform building heights and little greenspace. The CAA site used herein is located slightly to the left foreground. 25 limestone and granite administrative, residential and commercial buildings with clay tile or pebble-topped roofs. Streets and sidewalks are asphalt and concrete pavement, respectively (Figure 2.3). Using Ellefsen's (1985) scheme, Marseil le's urban terrain zone designation is best described by the ' A 2 ' categorization. Figure 2.3 Example of an urban canyon near the CAA site. Canyon widths are on the order of 7-10 m, resulting in H/W ratios of approximately 2:1. 2.2 Tower Measurements Energy balance data used in the residual, parameterization and modeling techniques of AQs estimation was gathered using eddy covariance and radiometric measurement techniques. The fast- and slow-response instruments were mounted on a tower which was erected on the roof o f the Cours d'Appel Administrative ( C A A ) building in a densely built-up district of Marseille (Figure 2.4). 2.2.1 Observation Strategy In order to obtain observations of energy balance fluxes that are representative area averages at the local scale it is necessary to be at a height sufficient to ensure that the influence of individual surface roughness elements is not 26 Figure 2.4 Aerial view of Marseille's city center showing the location of the CAA instrumented tower. [Photo courtesy of Dr. Marc-Andre Velay-Dabat, Laboratory ABC, Ecole d'Architecture Marseille-Luminy.] evident. Measurements taken within the urban roughness sublayer are not satisfactory because by definition turbulent movement within that layer is dynamically influenced by individual buildings and trees (i.e. micro-scale influences and processes; Oke, 1987). For example, a transfer o f heat within the roughness sublayer could actually be the circulation of heat within a near-surface eddy, rather than the desired transfer of heat between the surface and the atmosphere. To be certain that measurements represent an integrated land-use response at the local-scale, it is therefore imperative that instruments be mounted within the inertial sublayer (ISL), which is just above the roughness sublayer (RSL) (Figure 2.5). 2 7 The lower boundary of the ISL, used as the 'surface' in many modeling studies, is known as the blending or roughness sublayer height, z*. Wind tunnel studies suggest this height is approximately 2D to 3D, where D is the horizontal spacing of the roughness elements (Garratt 1980, Raupach et al, 1980), or 2.5 to 4.5 ZH, where ZH is the mean height of the elements (Pasquill, 1974, Garrett 1978). Raupach et al (1980) combine the ideas of height and horizontal spacing to derive a useful relation for blending height, z* = zH + 1.5D. Inertial sublayer (ISL) Blending height, .-'•Roughness sublayer.(RSL) 1 § G ii O 60 N O "3 ^ o 2 N E "C o Figure 2.5 Schematic showing the blending height, z*, above which the influence of individual surface elements are integrated into a local-scale signal. Source: After Oke et al. (1989). Eddy covariance involves the direct measurement of fluctuations in vertical wind speed, and an atmospheric property, e.g. temperature and specific humidity, which is related to the flux of interest (e.g., sensible and latent heat). The time-averaged product of the instantaneous fluctuation of vertical wind speed and the air temperature or specific humidity can then be used to calculate QH or QE (Foken and Wichura, 1996). The resulting equations for sensible and latent heat fluxes are, respectively: 28 QH = QE = CaW'T' Lvw'' pJ (2.1) (2-2) where Ca is the heat capacity of air, w'T' is the time average of the instantaneous covariance of vertical wind speed and air temperature, Lv is the latent heat o f vaporization, and w'p\' is the time average of the instantaneous covariance of vertical wind speed and water vapor density (Kaimal and Finnigan, 1994). As is typically the case in most climate studies, ancillary slow-response measurements of wind speed and direction, surface and atmospheric moisture, and air temperature are also recorded. This additional meteorological information provides critical knowledge relating to the prevailing synoptic conditions and near-surface atmospheric stability that are often used in modeling and parameterization applications. 2.2.2 Tower Instrumentation With the exception of the net storage heat flux, each component of the urban surface energy balance (equation 1.2) o f Marseille was individually measured. Instruments were mounted on an adjustable pneumatic tower, the base of which was placed on the roof o f the C A A building 20.7 m above street level (Figure 2.6). The mast could be raised and lowered. The tower was lowered to avoid excessive wind loading in strong wind conditions. / During Mistral drainage f low episodes of high wind speeds (gusts often greater than 7 m s"1 and up to 14 m s"1), the tower was lowered. This permits representative local-scale measurements to still be made. Most days, however, were characterized as 29 moderate or light sea breeze or synoptic flow days, thereby allowing the mast to be extended to its optimum height. Figure 2.6 Photograph of the CAA tower in the 'up' position. Energy balance measurements were conducted between 28.5 - 43.9 m above street level (Table 2.1). Given that the average horizontal spacing D between roughness elements (here taken to be the average urban canyon width) was on the order o f 7 m, Raupach et al.'s (1980) estimation of z* the lower bound o f the roughness sublayer, is Table 2.1 Heights at which tower instruments were mounted. Tower Position Level 1 Level 2 Z L I / Z H Up 43.9 m 37.9 m 1.83 1.58 Down 34.6 m 28.5 m 1.44 1.18 30 approximately 35 m above street level. Closer examination of sensible heat fluxes and friction velocities between the two instrumentation levels in both the 'Up ' and 'Down' positions show that measurement heights greater than 1.5 ZH are appropriate at this site (Figure 2.7). When the tower is ful ly extended in the 'Up ' position during the day, the Q H ' D Q H U 4 8 12 16 20 Time (Local) 4 8 12 16 20 Time (Local) Figure 2.7 Mean and standard deviation of ratios of « * 2 / " * i a n d QHIIQHX- Data are stratified for tower position, time of day and wind direction. Source: Grimmond et al. (2002b) Q\H ratio between the two levels is close to one, suggesting that fluxes from both levels are similar and are therefore within a constant flux layer, the inertial sublayer. When in the 'Down' position, the lower eddy correlation system measures smaller fluxes, resulting in ratios of less than one. In this configuration, it seems the lower instrument package is sufficiently below 1.5 ZH that it should not be used in analysis. Based on rule-of-thumb zci = 0.7 ZH recommended by Grimmond and Oke (1999b) of zero-plane displacement length and field survey estimates ZH of approximately 15.6 m, Zd is expected to be ~ 10.9 m. The H/W ratios of ~2 suggest the flow type is 'skimming.' This implies interactions between the canyons and above roof-level are not optimal. Using turbulence and stratified wind data to calculate Zd yields higher estimates of approximately 26 m. Given 31 a Zd between 10.9 and 26 m and the logarithmic wind profile in neutral conditions (equation 5, Grimmond et al., 1998) a reasonable estimate of roughness length zo is ~ 1 m (Grimmond et al., 2002b). Sonic anemometers were used in conjunction with infrared gas analyzers ( IRGA) to measure the turbulent fluxes of sensible heat (QH), momentum (T), and latent heat (QE) at the two heights. A l l components o f net all-wave radiation were sensed with a net radiometer mounted on a boom extending horizontally 2.7 m from the tower. A i r temperature, relative humidity and mean wind speed and direction were measured with standard meteorological equipment (shielded temperature/relative humidity probe and cup anemometer and vane) at the same height as the lower sonic anemometer/gas analyzer package. A n array of fine-wire thermocouples provided a vertical temperature profile at nine heights up the tower. A summary of measured variables and instruments is presented in Table 2.2 (from Grimmond et al., 2002b). Table 2.2 Instrumentation mounted at the CAA tower site. Variable Instrument Model Level U , V , W , U«, T Sonic RM Young 81000, 1,2 QH, Ta, anemometer ATI H20 Infrared Gas Analyzer Licor-7500 1,2 Kl Kl L], Ll Q* Radiometer Kipp and Zonen CNR1 1 TA Thermocouple Omega T-type 36 awg 9 levels Ta,RH T/RH probe Vaisala HMP35C with radiation shield 2 Surface moisture Moisture sensor Weiss-type rooftop U, direction Anemometer and wind vane RM Young 2 32 2.3 Surface Temperature Survey The bulk thermal mass-surface temperature approach to estimating AQs requires surface temperatures over an array o f surface types and orientations to be sampled. Infrared thermometry techniques were used in a survey of the surface temperatures in the Marseille study area. 2.3.1 Observation Strategy Surface temperatures are often measured directly by constructing a series of thermocouples that can be bonded to the surface (Fairey and Kalaghchy, 1982). This technique, however, can be problematic, in that it is prone to error from exposure and intrusion of the thermocouple (i.e. the object's temperature is influenced by the presence of the sensor). Further, even i f an array of thermocouples is mounted on a surface, it is unlikely that a satisfactory spatial sample results (Oke, 1987). A better alternative to thermocouples is the use of an infrared radiation thermometer (IRT). These instruments remotely sense temperatures over a patch of surface area lying in the field-of-view of the IRT and produce no intrusion error. Infrared thermometers operate using the principle that all objects emit radiation in the infrared portion of the electromagnetic spectrum (0.75 - 1000 um). The amount of energy emitted is proportional to the object's surface temperature, given by the Stefan-Boltzmann Law: E = e0o-T04 (2.3) where E is the radiation flux density (W m" ), e0 is the infrared surface emissivity (dimensionless), a is Stefan's constant (= 5.67 x 10"8 W m"2 K" 4) and T0 is the surface temperature of the body (K). IRT optics collect the incident radiation from the measured object and focus the sample on an infrared detector, which converts it to a proportional 33 electrical signal. A linear voltage-temperature relation is conditioned by the internal circuitry, producing the final temperature analog output (Everest Interscience, 1991). To ensure the IRT optics 'see' the desired surface, instrument field-of-view calculations are performed, the results of which give the dimensions of the sensed surface area. In addition to their technological advantages, the ease with which infrared thermometers deployed is also ideal when an intensive survey over many surfaces is desired. Infrared surface temperature measurements include thermal emissions from the surface, as wel l as reflected radiance. Because naturally-occurring surfaces have emissivities of less than unity, apparent temperatures Tr of viewed surfaces are lower than the actual surface temperature, To. Voogt (1995) shows that corrections must be made to account for surface emissivity variations within an urban environment, which can range from less than 0.7 for treated windows to over 0.98 for dark painted surfaces and vegetation (Table 2.3), and the correction to the Tr of a composite urban surface is substantial because of geometry (Voogt and Oke, 2003). Table 2.3 Surface emissivities for common urban surface materials. Source: Modified after Oke (1987). Surface Material Emissivity, e Concrete 0.71-0.9 Asphalt 0.95 Brick 0.9-0.92 Stone 0.85 - 0.95 Wood 0.9 Glass 0.87 - 0.94 Black Paint 0.9 - 0.98 White/Red/Brown/ 0.85 - 0.95 Green Paint Roofing Shingles 0.9 - 0.92 Tar-gravel Roof 0.92 Tile Roof 0.9 Slate Roof 0.9 34 A further advance in infrared thermometry is the development o f thermal scanner systems, which produce snapshot infrared images o f multiple facets at once, thus allowing for fine spatial variability of surface temperatures to be detected (Voogt and Oke, 1997). Infrared scanners operate in conjunction with the manufacturer's software to provide high precision digital images (e.g. Figure 2.8) and to conduct statistical analyses such as area-averaging, spatial transects across the image, standard deviations of Tr of selected areas and cross-sections, etc. This application is especially ideal when the target surfaces include a wide range of temperatures, or when a high sampling frequency (e.g. 5-7 Hz) is required. Infrared thermal imaging scanners operate in theory much like infrared radiation thermometers: incident radiation is directed to a detector, the output of which is converted into a digital signal by a processor. Images are created from a set of scanned lines with a set number of pixels per line. Figure 2.8 Example of an image created from ground-based infrared scanner thermometry. The temperature scale on the right allows for convenient analyses of facet temperatures. Since the primary objective of this portion of the research is to sample a diverse set of surfaces, the observation strategy reflects that. Barring access restrictions, every 35 effort was made to measure surface temperatures over as long a time period as possible. Roads oriented north-south and east-west were sensed, as were canyon walls facing in approximately the north, south, east and west directions. Also, because roofs comprise a significant portion o f Marseil le's plan area (approximately 56%), infrared thermometers were mounted to measure temperatures of two roof types: a flat pebble-topped roof (Figure 2.9) and clay tile roofs pitched at -17° facing in the four cardinal compass directions. Depending on their age, rooftop tiles varied in color. Clay tiles o f uniform light red color and with little to no damage are categorized as "new" while roofs containing tile creating a speckled pattern from discolored and/or disjointed tiles (see Figs. 2.2 and 2.10) are considered "old." Figure 2.9 Instrument array at the CAA pebbled roof site. In addition to IRT surface temperature measurements (the instrument angled down at -45°, air temperature, relative humidity, K*, Q*, and sub-surface temperature measurements were conducted. 36 Figure 2.10 Three Everest infrared thermometers mounted on a rooftop edge, measuring surface temperatures of two walls and a road in an adjacent street canyon. 2.3.2 Surface Temperature Survey Instrumentation The primary source of surface temperature data was fixed Everest Interscience, Inc., Model 4000A or Model 4000.4GL infrared radiation thermometers (Figure 2.10). These instruments with either a 15° or 60° field-of-view were mounted a sufficient distance above the surface so as to optimize the surface area 'seen' by the sensor. The exterior o f each IRT was covered with highly reflective metallic tape that incorporates a layer o f insulation so that radiation incident on the instrument casing did not contribute to the warming o f its internal circuitry. Such absorption has been shown to bias temperature readings (Voogt, 1995). Table 2.4 is a summary o f the various observed surface types, the measurement period for each respective surface as well as the size o f its surface source area (see section 2.4). 37 Table 2.4 Description of surfaces and period of measurement and instrument source areas for the fixed IRT instrument network in Marseille. See Appendix C for a detailed description of the relative locations of the IRT site installations. Surface Type Description/Orientation Measurement Period (YD) Source Area (m2) Roofs Flat pebble roof 186-194 0.52 Southwest-facing new clay tile 183 -184, 191 - 192 0.31 North-facing new clay tile 190- 192 2.94 West-facing old clay tile 183 - 185, 190- 193 12.6 South-facing new clay tile 187-194 2.04 South-facing . old clay tile 173-194 N/A East- and west-facing old clay tile 179-194 33.6 Roads North-south oriented road and sidewalk 176-193 44.0 North-south oriented road and sidewalk 176-193 4.7 East-west oriented road and sidewalk 172 - 194 18.2 Walls East-facing windowless limestone wall 183 - 193 N/A East-facing limestone wall with windows 176-193 132.7 East-facing limestone wall with windows 183 - 186 9.9 South-facing limestone wall with windows 176-193 53.3 South-facing limestone wall with windows 176-193 484.7 North-facing limestone wall with windows 172-194 5.6 West-facing limestone wall with windows 176- 179,183- 193 8.9 West-facing limestone wall with windows 176 - 193 40.8 Surface temperature surveys were conducted at street-level using hand-held portable Minolta/Land Cyclops Compac 3 infrared radiation thermometers. Such measurements were made at the same time as both A S T E R satellite overpasses (giving infrared images of the E S C O M P T E domain) and airborne surveys using an infrared 38 scanner (Mestayer and Durrand, 2002). These surveys provided an intense sampling of unique surfaces at several times through the day, including limestone walls o f varying hues, glass window panes, streets and sidewalks, all in both sunlit and shaded conditions. The hand-held infrared thermometry data were gathered primarily within four street canyons that comprise a one-block area directly south o f the C A A site (Figure 2.11). These observations were used in conjunction with the fixed Everest infrared thermometers to calculate average facet temperatures. • Rue Sylvabelle Rue Saint Jacques • wall or window measurement, one per floor Road/sidewalk measurement 20 meters Figure 2.11 Schematic of the complete block over which hand-held IRT sampling was conducted. This idealization depicts shadows being cast towards the northeast, due to the position of the afternoon sun. Source: Voogt. pers. comm. A n infrared scanner system (FLIR Systems ThermaCam SC500) measured an assortment of surface types and orientations during the E S C O M P T E field campaign, most often in conjunction with A S T E R satellite and airborne scanning overpasses. For these measurements, the scanner was positioned on a rooftop five blocks south of the C A A site and manually adjusted so as to scan a 360° view of rooftops and building walls. Images looking down into a north-south oriented canyon were also taken. In addition, the scanner system was set-up on the C A A rooftop through a weekend and programmed so as 3 9 to take snapshots at five-minute intervals of a sector southwest o f the tower. This dataset is particularly useful when examining the diurnal variation of wall and roof surface temperatures and was employed specifically to analyze an east-facing tile roof. Fol lowing the field campaign, preliminary calibrations of the fixed Everest and hand-held Minolta IRT's were performed in Marseille against the standard of a Campbell Scientific, Inc. calibration plate. This device provides an accurate temperature reading of a black (near blackbody) plate, allowing inter-instrument comparisons to be made. In addition, a concrete block surface was outfitted with three thermocouple loops. IRT's were pointed at the surface and the mil l ivolt output from the instruments was compared with the corresponding output from the thermocouples. More rigorous IRT calibrations were conducted later using a stirred water bath calibration chamber in the Soil Science Laboratory of the University of British Columbia. These calibrations produced the equations given in Appendix A . 2.4 F lux Source Areas Sensing instruments used to measure surface energy fluxes are influenced by a specific portion of the underlying surface surrounding the measurement tower. Schmid and Oke (1990) define this surface area of contribution as the "source area." Even though the measured signal is derived from one point in space, it can be conceptualized as being the spatial average of the influence of individual surface elements, i.e., a composite surface influence. The size and shape of source areas depend on many factors, including the type of energy flux under study (radiation, convection), the 'f ield of view' of the instrument, the configuration of surface elements, and the prevailing f low conditions (Schmid et ah, 1991). Calculation of radiation and turbulent flux source areas 40 are important in part because they allow better understanding of where instruments should be mounted in order to optimize the study of desired surface influences. Further, knowledge of the degree of over-lap between radiation and turbulent flux source areas is pivotal in assessing the closest possible degree of energy balance closure. 2.4.1 Surface Cover Survey In calculating turbulent flux source areas, knowledge of several surface and roughness parameters (e.g. roughness length, height of zero-plane displacement, surface element heights) is a prerequisite. The acquisition of such information requires a comprehensive investigation of the study area, using a variety of techniques. Fol lowing the method of Grimmond and Souch (1994), aerial photographs depicting the urban district within a 2-5 kilometer radius (Figure 2.4) surrounding the measurement site are first analyzed to determine fractional plan cover of land cover types (roofs, vegetation, water, impervious ground). Special attention is given to those sectors corresponding to dominant wind directions; see Section 2.4.3 for explanation. Within those sectors, a detailed analysis of aerial photographs provides local-scale details within each category. A t the micro-scale, important three-dimensional information of the surface such as building heights, construction materials, vegetation type and density, and canyon dimensions is completed based on random sampling from foot surveys, photographs, historical and architectural records, and urban plans. For this study, 4761 randomly-selected points within an aerial image of the study area (Figure 2.4) was entered into an ArcGIS program. From there, each of the points were classified using eleven surface cover categories. From this procedure, plan area surface cover descriptions for 100 m X 100 m grid squares within the study domain, were 41 derived (see, for example, Figure 2.12, Grimmond et al., 2003). Once source area dimensions are known (see sections 2.4.2 and 2.4.3), average hourly surface cover weightings are generated. Analyses of the turbulent flux source areas over the source of the IOP find four primary surface cover categories with the following average weightings: 56% plan area of roofs, 26% plan area of impervious ground, 16% plan area of vegetation, and 2% plan area of water. Complete, three-dimensional surface facet calculations (see section 2.5) reveal different weightings: 30% complete area of roofs, 13% complete area of impervious ground, 8% complete area of vegetation and 49% complete area of walls. Figure 2.12 Aerial photograph of area surrounding the CAA site, overlain with 100 x 100 m 2 square grids depicting plan area occupied by clay tile roofs. Similar images are generated for other surface cover categories. Source: Grimmond, pers. comm. 42 2.4.2 Upwelling Radiative Flux Source Areas Reifsnyder (1967) discusses the upwelling radiation source area, as sensed by a down-facing standard dome radiometer. The surface area 'seen' by the radiometer is of a disk shape and the majority of the radiative flux contribution is found directly beneath the sensor. It is important to note that the radiometer is also exposed to radiation from the entire underlying surface, extending to infinity in all directions. Reifsnyder then defines a view factor (F) as the ratio of received radiation from the surface disk to the amount received from the remaining area surrounding the disk stretching to infinity and shows that this proportion can be written as a simple mathematical expression which obeys Lambert's Cosine Law: r +Zs where r is the radius of the source area circle, zs is the sensor height, and F is the proportion of the measured flux for which that area is responsible. Schmid et al. (1991) and Reifsnyder (1967) re-write this expression to use a defined view factor to calculate the circular source area (the center of which is the tower) for upwelling shortwave and longwave radiative fluxes (K\ and Z f ) : r = z^-l)-°- 5 (2.5) F Table 2.5 shows the dimensions of the upwelling radiative flux source area dimensions for different values of view factor and radiometer heights at the C A A site in 43 Marseille. The actual area encompassed by these source areas is seen superimposed on an aerial photograph of Marseille (Figure 2.13). Table 2.5 Upwelling radiative flux source areas for the down-facing radiation sensors on the GAA tower for different view factors. Up position (z, = 43.9 m): View factor Radius Plan Source area (%) (m) (xlO'm2)1 80 87.8 2.42 85 104.5 3.43 90 131.7 5.45 95 191.4 11.5 99 436.8 59.9 99.5 619.3 120.5 Down Position (z,, = 34.6 m): 80 69.2 1.50 85 82.4 2.13 90 103.8 3.38 95 150.8 7.15 99 344.3 37.2 99.5 488.1 74.8 Note: the 3-dimensional source area, including vertical facets of the surface, can be calculated as in Soux et al. (2003). Lambert's Cosine relationship is obvious when considering Table 2.5 and Figure 2.13 together. For example, when examining the 95% and 99% radiative flux source areas when the tower is in the 'Up ' position, it is important to remember that only 4% o f the upwelling radiation sensed at the tower originates from the area between the two isopleths. The plan surface area of this 4% contribution constitutes approximately 48,400 m 2 . This highlights the importance o f considering upwelling radiative flux source areas when determining site locations. Since a significant portion of the upwelling radiation sources are probably found within just a few hundred meters o f the tower (given reasonable tower heights), care must be taken to ensure "characteristic" land-use categories are 'seen.' Offerle et al. (2003) note that radiation measurements have more stringent site requirements than those of turbulent transfer measurements. 2.4.3 Turbulent F lux Source Areas Calculation of source areas of turbulent latent and sensible heat fluxes is not as straight-forward a task as it is for radiative fluxes. Whereas radiative flux source areas depend on measurement height, field-of-view and surface geometry they are essentially static over time. Turbulent flux source areas on the other hand, depend on turbulent diffusion and therefore vary in size and position according to measurement height, wind speed and direction, the heterogeneity of the underlying surfaces, atmospheric stability, and surface roughness (Kljun et al, 2002). Though flux source area calculations in urban areas are complicated by the inhomogeneous terrain, the literature nonetheless contains several examples of numerical models which attempt to do just that. Most use a modified Lagrangian trajectory scheme or are based on Eulerian analytical solutions of the advection-diffusion equation (Schmid, 1994). One such numerical model is the Flux Source Area Model ( FSAM) of Schmid (1994, 1997), which can be used to calculate the dimensions of turbulent flux surface source areas given input values of sensor height (zs), roughness length (z 0), Obukhov length (Z), the standard deviation of lateral wind speed fluctuations (av), and friction velocity (w»). Schmid uses these parameters to define a three-dimensional source-weight function (also referred to as a footprint function; Schmid and Oke, 1988, 1990) which, when integrated over a specified domain, gives the surface source area. Incorporating measured wind directions, the output gives the dimensions and horizontal position of elliptical isopleths of source areas centered on the measurement tower, in increments of 10%. These are commonly calculated on an hourly time scale. These isopleths can be overlain on a GIS database to better visualize what proportion of the actual surface 45 contributes to the overall weighting functions. Figure 2.13 shows the final radiative and turbulent flux source areas from the Marseille campaign. Figure 2.13 Daytime turbulent and upwelling radiative flux source areas. The top of the figure is north. The outer-most white circle represents the area responsible for 99% of the upwelling radiative fluxes observed at the top of the CAA tower in the 'Up' position. The inner white circle denotes the area responsible for 95% of the upwelling radiative flux. Daytime turbulent flux source areas under sea-breeze conditions are denoted by the purple area to the south; daytime turbulent flux source areas under Mistral conditions are represented by the orange/yellow area to the north. Areas shown are calculated for the 90% source areas footprint for 30-min flux data and averaged over the entire IOP. Individual hourly source areas are elliptical. Darker areas have more influence on the flux than lighter areas. Source: Grimmond, pers. comm. When the radiative and turbulent flux source areas are considered together, it becomes apparent that the common area encompassed by them is not ideal. The lack of overlap potentially raises concerns regarding the spatial representativeness of the two sets of measurements. For example, when the area from which 95% of upwelling radiative fluxes is considered, almost none of the daytime turbulent flux source area under Mistral conditions coincides. Significant turbulent flux contributions under sea-breeze and synoptic f low regimes are, however, fairly represented within the 95% upwelling 46 radiative flux source area. This is encouraging, since all of the observation days considered in the present work fall into the synoptic or sea-breeze categories (see section 2.7). A further mitigating fact is that spatial distribution of urban Q* is known to be conservative (usually < 5%, Oke, 1988; Schmid et al., 1991). The bulk of the surface temperature surface occurred directly south of the tower, an area encompassed by both the 95% radiation source area and turbulent flux source areas. 2.5 Generic Neighborhood Mode l Calculations of upwelling radiative and turbulent flux source areas are helpful in constructing the idealized neighborhood model which defines the local-scale volumetric boundaries of the bulk thermal mass approach to the estimation of AQs. Ideally, the dimensions of this box model should be within the boundaries of both source areas. Figure 2.13 shows that under the sea-breeze and synoptic f low conditions relevant in this study, the 95% upwelling radiative and turbulent flux source areas are co-iocated with the area south of the tower over which the surface temperature survey was conducted. It is therefore reasonable to use the block that was sampled with hand-held infrared thermometers (Figure 2.11) and scale-up accordingly. The perimeter of the block is defined to be the mid-point of the encompassing canyon. The approximate plan area dimensions of the built-up portion of this block are 100 m x 60 m (east-west x north-south). Canyon widths are assumed to be 8 m in total, making the entire plan area o f this block 108 m x 68 m, or 7344 m 2 . The complete surface area (see Figure 1.1), however, needs to account for horizontal and vertical surfaces, both along the outside perimeter of the block as well as the inner perimeter comprising the courtyard. Assuming a mean building height of 15.6 m, the vertical 47 surface area of the outside perimeter is 4992 m . To calculate the vertical surface area of the inside courtyard, it is assumed that building depths are 20 m, resulting in a courtyard vertical surface area of 2496 m 2 . Therefore, the complete (three-dimensional) surface area of this block (vertical surface area plus plan surface area) is 14,843 m 2 . These dimensions are used in weighting surface types and orientations, along with the corresponding surface temperature measurements and parameterizations, and material thermal properties. 2.6 Data Acquisit ion/Processing The sonic anemometers were logged, using software developed at Indiana University, directly to a PC located on the Cours d 'Appel Administrative rooftop at the base of the tower. A Campbell Scientific, Inc. (CSI) 23X datalogger mounted at the base of the tower was responsible for recording signals from the infrared gas analyzers, thermocouples, and ancillary meteorological data. The raw I R G A and thermocouple data were automatically downloaded from the datalogger to the PC at hourly intervals. The remaining data were averaged and stored at 15-minute intervals on the datalogger and dumped every day at midnight onto the PC. For most o f the observation period the sonic anemometers and infrared gas analyzers were sampled at 10 Hz while the signals from the ancillary sensors and the thermocouple array were sampled at 0.2 H z (Grimmond et al., 2002a). Infrared surface temperature data gathered with the Everest instruments were logged by on-site CSI 21X or H O B O dataloggers. Sampling occurred every 15 seconds (aside from a few 20 or 30 second sampling periods adopted in the interest o f power conservation; Voogt, pers. comm.) and data were averaged and stored as 15-min means. 48 Data were downloaded every few days either onto a laptop using a serial cable and CSI PC208W software or to a CSI storage module, to be later transferred to a laptop PC. Infrared images obtained with the FLIR SC500 Scanner were generated and analyzed with the manufacturer's software (ThermaCam) and a coupled Visual Basic - Microsoft Excel spreadsheet program. Specific areas of interest and analysis specifications (in this case, average surface temperature) are defined by the user (Figure 2.14) and iterated through the sequence of images. Five-minute temperature values used in the present research were converted to hourly averages for final analysis. Figure 2.14 One of a series of infrared images looking southwest from the CAA site, with the corresponding temperature scale shown to the right. Sampling occurred every five minutes over 65 continuous hours. The boxed area labeled 'AR01' is the specific portion of east-facing roof tiles from which hourly average temperatures were obtained. Final covariance values and resulting fluxes were calculated using in-house software (IUFlux, Grimmond et al., 2002a). Corrections included the effect of air density fluctuations (Webb et ah, 1980). Wind coordinates for the sonic anemometer output were rotated from the x, y, and z wind components to stream wise (u, v, and w) coordinates by aligning u with the mean wind. Quality control o f data was conducted 49 either by hand or with automated computer programs. A l l energy (flux and slow response data were then converted from 15-miri averages to hourly averages. 2.7 Weather Conditions During the Observation Period The Mediterranean climate regime is characterized by its annual precipitation cycle (wet winters and very dry summers), warm to hot summers, unusually mi ld winters, and a high percentage of the possible sunshine for the year in the summer (Trewartha, 1954). The reason for the distinct precipitation pattern lies in the poleward movement of subtropical high pressure cells during the summer months. Mediterranean climates are located on the west coasts of continents, poleward of the dry, eastern side of subtropical high pressure cells. A s these cells move poleward during the summer months, they advect dry continental tropical air. This is the basis for the characteristically dry Mediterranean summers. Marseil le's location on the northern coast of the Mediterranean Sea facilitates a particularly strong Mediterranean climate influence on the region. This is primarily because the relative warmth of the Mediterranean Sea in the winter creates a low-pressure trough and a corresponding convergence zone, resulting in the development of concentrated fronts and cyclones. The Mediterranean Sea, therefore, does not contribute to a marked cooling of its coastal region, resulting in warmer summers (Trewartha 1954). Marseil le's wet season (mean monthly precipitation of 69 - 76 mm) occurs in October-November, and the area is driest (10 -25 mm of precipitation per month) in June-July. Mean air temperature ranges from 5.5 °C in January to 22.8 °C in August. Climatologically, the wind originates from the 270° - 360° sector in all months except December (during which the winds originate from 90°). A t the meso- and local scale, 50 Marseille is influenced on a diurnal basis (especially in the summer months) by a strong sea-land breeze system. During the day, cooler winds blow on-shore from the westerly sector while at night the land breeze develops gives a nocturnal northwest f low out to sea. Marseille also experiences a localized flow, the Mistral, which is responsible for a strong northerly f low down the Rhone Val ley towards the Mediterranean Sea. This cool, dry wind occurs predominantly during the winter months and is the result of a dynamic low pressure region generated in the wake of the A lps ridge (Guenard et al., 2003). The E S C O M P T E field season was impacted by Mistral f low (characterized as persistent northerly f low greater than 7 m s"1) on four of the 27 observation days during which energy balance measurements were conducted. A meso-scale land-sea breeze system dominated the f low regime for 55% of the time. Sea breeze conditions at C A A were characterized by a northerly to easterly f low in the early to mid-morning, changing to a westerly f low for the rest of the day. Overall, regional f low influenced the region 19 of the 27 (70%) observation days (Table 2. 6, Pigeon, pers. comm.). Given that there was no precipitation during the field campaign and relatively little cloud, the wind regime served as the prevailing meteorological factor that differentiated days. 51 Table 2.6 ESCOMPTE wind categorizations. Entries in bold refer to IOP days analyzed in the present study. Source: Pigeon, pers. comm.. Y D Wind Category Y D Wind Category 167 Other 181 Sea breeze 168 Mistral 182 Mistral 169 Mistral 183 Other 170 Sea breeze 184 Sea breeze 171 Sea breeze 185 Other 172 Sea breeze 186 Other 173 Sea breeze 187 Other 174 Other 188 Sea breeze 175 Sea breeze 189 Sea breeze 176 Sea breeze 190 Sea breeze 177 Sea breeze 191 Sea breeze 178 Sea breeze 192 Other 179 Mistral 193 Sea breeze 180 Other Ensemble hourly-averaged air temperature and vapor pressure deficit values over the IOP (JD 183 - 192) reflect a 'typical ' Mediterranean summer (Figure 2.15a). A s recorded by a Meteo-France weather observation station two km northeast of the C A A site, the average air temperature during the observation period was 22.8 °C, which corresponds exactly to the region's climatological average summertime temperature. The average diurnal temperature range was 18.7 - 26.9 °C and the average vapor pressure deficit was 17.8 hPa, with an average diurnal range of 14 - 40 hPa. The difference in wind regimes experienced over the IOP is demonstrated in Figures 2.15b and 2.15c. O f particular interest is the characteristic early-morning wind direction change (easterly to west-northwesterly) associated with a land-sea breeze circulation. Sea-breeze winds are also consistently weaker than meso- or synoptic-scale f low (referred to as "other"). The ensemble air mass characteristics (temperature and vapor pressure deficit) illustrate appreciable differences under both f low regimes, with sea-breeze air generally cooler and drier. 52 360 0 -I , , , , , , , , , , r - J 0 2 4 6 8 10 12 14 16 18 20 22 Local Time Figure 2.15 Ensemble hourly measurements of temperature, vapor pressure deficit (VPD), wind speed and wind direction during the IOP, as recorded by a Meteo-France weather station near the CAA site. Near-consistent surface static instability is appreciated when examining the diurnal ensemble (30-minute averages, Figure 2.16) behavior o f the dimensionless variable, q, defined as q = .z'IL, where z' = z - Zd and L is the Obukhov length. L is a parameter derived from the turbulence kinetic energy budget equation and is proportional to the height above the surface at which buoyant factors initially dominate over mechanical (shear) production of turbulence (Stull, 2000). The parameter c, is referred to as a "surface-layer scaling parameter", and although its magnitude is not related to the 3 . ; 0 2 4 6 8 10 12 14 16 18 20 22 24 Time Figure 2.16 Ensemble (30-min.) averages of the dimensionless surface-layer scaling parameter The persistent negative values are indicative of static instability. type or degree of stability, its sign is used to assess static stability: negative values imply statically unstable conditions and positive values generally refer to statically stable conditions (Stull, 1988). Given that, it is apparent from Figure 2.16 that at most times during the day and even the night, the static stability in Marseille is considered unstable. 54 2.8 Comparative Statistics Because this is ultimately a comparative study, the statistical analysis tools used in evaluation reflect that. It is recognized that an inherent weakness in the project is the reality that there is no 'actual' or 'true' value of the energy storage flux at this site against which comparisons can be made. In lieu of this, the residual value of AQs is chosen to be the comparative or observed 'standard.' It is probably considered to be the best available estimate (assuming careful site selection to approximate energy balance closure is conducted; Oke and Cleugh, 1987). The statistical formulae of Wilmott (1982) and Fox (1981), which are deemed the most appropriate when comparing the performance of modeled (in this case, the AQs estimates from O H M , TEB , and TMS ) versus observed (the energy balance residual) data. Unl ike standard statistical analyses, the Wilmott and Fox set of statistics relate the accuracy of prediction to the degree to which model-predicted values (P) approach the magnitudes of their observed counterparts (O), for N number of cases. The first type of difference measure is the mean absolute error (MAE) , which is the mean of the absolute values of the differences between a set o f predicted and observed values. Second, there is the mean bias error (MBE) , which is simply the difference between the mean of the predicted variable and the observed variable. Third is the root mean squared error (RJVISE) is the roof of the average of the squared differences. H\Pi-Oi\ MAE = & (2.6) . N 55 MBE = N (2.7) N Y(Pi-Oi)2 RMSE= i=l (2.8) N Of these statistics, the R M S E and M A E are the best overall measures of model performance, since both express the mean difference in the same units of the observed and predicted quantities. The fourth statistic used in this analysis is the Wilmott agreement index (d), which is a descriptive, dimensionless number between zero and unity that indicates the degree of predictive accuracy from the independent variable (Wilmott, 1981). where P,' = Pt-0 and 0, ' = Ot- O. According to Saunders and Bailey (1997), values of d at or above about 0.80 are thought to indicate good agreement between observed and predicted values. I {Pi ~ Oi)2 d = 1 - (2.9) E ^ Y | + |GV|) 2 56 C H A P T E R 3: R E S U L T S The aim of this chapter is to provide results from the four methods to estimate urban energy storage flux, AQs, as discussed in Chapter 1, using observations and characteristics of a site in the city center of Marseille, France. Input values necessary for the O H M scheme, the T E B model, and the T M S approach are given. Results generated from each method over the ten-day IOP are compared to the measured (residual) approach. To better understand the role of AQS at this site, sensitivity analyses w i l l be explored in Chapter 4. 3.1 Observed F lux Behavior Radiation and surface energy budget fluxes derived from measurements conducted at the C A A site are given. A survey of energy partitioning at this site reveals the significance and behavior of surface-atmosphere sensible heat exchanges. 3.1.1 Radiat ion Balance The surface radiation budget IOP time series from the C A A site (Figure 3.1) show 1000 L-up — Q* K-up L-down K-down 183 184 185 186 187 188 189 190 191 192 193 Y D Figure 3.1 Time series of individually measured surface radiative flux densities. Data are fifteen-minute averages recorded during all-sky conditions for the CAA site in downtown Marseille. 57 signatures typical of summertime, mid-latitude, anti-cyclonic conditions. Values presented here are similar to those found in the literature for other mid-latitude sites, both urban and rural (see Oke, 1987; Monteith and Unsworth, 1990). Aside from sporadic cumuliform clouds sprinkled over the region, cloudless skies allowed incoming solar radiation to dominate the radiant energy regime. Peak K[ values reached 900 W m" . There does not appear to be a strong correlation between cloud cover and the wind regimes discussed in Chapter 2. Both sea breeze and synoptic ('other') f low days experienced essentially cloudless skies and sporadic cloud cover. Figure 3.2 shows the ensemble diurnal averages of radiative flux densities. One notable feature is that L] appears to be almost in-phase with K[, with a relative flattening 1000 c •> O H a u. S5 -200 Figure 3.2 Ensemble 15-minute averages of radiation flux density terms under all-sky conditions over JD 183-192. of the Z | plot near mid-day. The pattern at many simple surfaces has a mid-afternoon peak in L\ coincident with the time of maximum surface temperature. Similar signatures 58 to that in Marseille have been noted in other environments such as suburban Miami (Newton, 1999) and a forested site (McCaughey, 1985). The coincident timing of the K[ and 11 curves is thought to be unique to surface environments with prominent shading elements such as trees and buildings. Although some surface facets, especially ones which are vertical, may experience a peak in I f before solar noon, the average 'complete' surface temperature is mitigated by the cooler (shaded) facets at that time. This helps to explain why the flattened L\ peak straddles solar noon by several hours and why, in general, Z | and K[ are effectively in-phase at many urban sites such as Marseille (Newton, 1999). 3.1.2 Surface Energy Balance The relative importance of each energy flux term at this highly urbanized site is demonstrated in Figure 3.3 and Figure 3.4, which show the 10-day time series and diurnal 700 T — i -400 -I 1 1 1 , , , , , , \ 183 184 185 186 187 188 189 190 191 192 193 Time (Local) Q * Q H Q E Q S Figure 3.3 10-day time series of measured surface energy balance terms at the CAA site over all sky and wind conditions. 59 hourly ensemble averages, respectively, of the terms in the measured version of the energy balance equation (1.2), under all sky and wind conditions. The ensemble time series of flux ratios (fluxes normalized by Q*) at the Marseil le site, under all sky and wind conditions, gives a clear picture of the energy partitioning at this site (Figure 3.5). -200 -I 1 1 1 1 1 1 1 , , , ,—I 0 2 4 6 8 10 12 14 16 18 20 22 LAT Figure 3.4 Ensemble hourly averages of surface energy balance terms at the CAA site over all sky and wind conditions. Standard error bars are shown. Not surprisingly, the dry summer climate and lack of local vegetation results in a latent heat flux curve which hovers about 0 W m"2 during the night until sunrise, when it gradually increases and peaks in the afternoon, but at no time does the QEIQ* ratio exceed 0.16. QE peaks about an hour after solar noon and precedes the QH by about the same period. For the most part the role of evaporation and condensation during the day therefore is small (11%, see Table 3.1). The magnitude of the sensible heat flux during the day is large, accounting for over 75% of available daytime radiant energy. It is also noteworthy that QH never dips below 0 W m"2 during the night. This suggests that the nighttime release of heat from 60 storage is strong enough to support positive QH values throughout the day and night. Indeed, at all times of day, the turbulent sensible heat flux is the primary means of non-radiative heat transfer away from the surface. A hysteresis relationship between QH and Q* is evident, because the peak in the turbulent sensible heat flux lags the Q* maximum by almost two hours (Figure 3.4). This pronounced behavior has been observed at other dry sites (Tucson and Mexico City) and indicates a convective afternoon release o f stored energy accumulated during the morning hours, due to the typical increase of wind speed (Figure 2.15) and the buoyant instability (Figure 2.16). The average daily peak in AQs precedes Q* by approximately two hours. Again, 2 -1 -2 / \ /\* ---+-QH/Q -+-QE/Q -+-QS/Q v r * * y 0 2 4 1 1 1 ~! 1 1 6 8 10 12 14 16 LAT 18 1 1 — 20 22 Figure 3.5 Diurnal ensemble patterns of Qt/Q*, Qh/Q* and AQg/Q* for the CAA site under all-sky and wind conditions. this behavior is expected, as such a hysteresis pattern (Figure 3.6) is evident at other sites including urban ones (refer to Monteith and Unsworth (1990) and Table 3.3 below). Not only is AQs the first output flux to peak during the day, it is also the first to change sign (direction) in the later afternoon, in cities this supports the appreciable QH flux through 61 the night. Negative nighttime AQs values indicate an efficient release o f heat from storage, making it the sole source of thermal energy to the urban environment at those 500 400 300 s g 200 Q ' * 100 S • E $ ° -200 — • — Q H - - • Q - - Q S > ^ % • • • . • ! > - « . JQ -200 200 Q* (W m1) 400 600 Figure 3.6 Observed hysteresis relationship between sensible heat flux and net radiation (top plot) and energy storage flux and net radiation (bottom plot). times. Given that the energy storage flux term is calculated as a residual, and because the latent heat flux is not a major component of the surface energy balance at this site, it is not surprising that the regimes of the QH and AQs fluxes are strongly related. Indeed, the plots of both flux ratio terms (Figure 3.5) are practically mirror reflections of one another. This is expected, as energy released from storage within a built-up environment (where latent heat is negligible) is most l ikely to be sensed as turbulent sensible heat flux. Accordingly, Figure 3.6 shows that both AQs and QH display similar magnitudes o f hysteresis with Q*, with the obvious off-set associated with the timing (the aj, intercept term). The strength of dependence between both terms is also evident when examining the shape of the hysteresis loop of AQs vs. QH (Figure. 3.7). The considerable time lag between the daily peaks in AQs and QH are responsible for the loop's breadth (the 02 62 -200 -* : 1 0 100 200 300 400 500 H„ (W m 2) Figure 3.7 Ensemble plot of AQS versus QH. Note the change of scale relative to Figure 3.5. term) and is indicative of the asymmetry in energy uptake by the surface (conduction) and release to the atmosphere (convection). The interplay between these terms highlights the need to understand the reasons why sensible heat sharing at the surface favors conduction over convection, or vice-'-versa. It also confirms that C A A is an ideal site to explore this. Table 3.1 shows Marseil le's energy partitioning under sea-breeze and synoptic f low regimes. Because there is ho remarkable difference in wind speeds under either f low condition, the critical role that wind direction and corresponding source areas play in this environment is demonstrated. In general, the daytime source areas under both wind regimes are similar (see Section 2.7). The difference, however, could lie in the early-morning change in wind direction associated with the sea-breeze circulation. 63 Table 3.1 Hourly averages of mean daytime (Q* > 0) and daily (24 h) observed fluxes and flux ratios for all-sky conditions under synoptic and sea-breeze conditions in Marseille (see Section 2.7). Daytime N is the number of hours when Q* > 0, daily N is the total number of hours analyzed. Flux units are in MJ m"2 day"1, all others are non-dimensional. Any missing hours within a data set are linearly interpolated. Wind Category N e* QH QE A £ v Qf/Q* Qi/Q* A f t / 0 * Daytime (Q* > 0) Synoptic ('other') 13 16.2 10.9 1.6 3.7 0.67 0.10 0.23 Sea-breeze 13 16.6 13.4 2.2 0.9 0.83 0.12 0.06 Daily (24 h) Synoptic ('other') 120 12.7 11.75 0.47 0.4 0.93 0.04 0.03 Sea-breeze 120 13.8 15.0 N 2.6 -3.8 1.12 0.16 -0.28 The easterly to southerly winds generally observed in the mornings of sea-breeze days could advect turbulent eddies from a moisture source (likely from large fountains and pools located east of the tower) that is sensed at the C A A site. This would help to explain the greater disparity in latent heat flux partitioning between the wind regimes over a diurnal period. 3.1.3 The Context of Marsei l le 's Energy Regime The energy partitioning at the Marseille site can be placed in its context versus other urban sites with information provided in Table 3.2. The special nature of the surface energetics o f Marseil le's environment and climate can be gauged in comparison with daytime (Q* > 0) and daily (24 h) mean observed fluxes and flux ratios for all-sky conditions at various North American locations. Considering its dry climate, it is not surprising that the Marseille site has the second-largest daytime and diurnal Bowen ratio (P = QHIQE) values. One striking feature from the Marseille site is the ratio of daily QH^O AQS, indicating that the magnitude of sensible heat flux <g//is over five times greater than the energy storage flux AQs during the day. Over a 24-hour period, however, Marseil le's Qt/AQs ratio does not have an extraordinary ranking against the other sites. This comparably smaller ratio is explained 64 by the large magnitude of AQs release over the course of the night. The nocturnal release of energy from the urban fabric is also responsible for the negative diurnal value of QHI AQs. As the last column in the table shows, the Marseille site is able to partition over eight times more energy into sensible (both QH and AQs) rather than latent heat during the day. This proportion is even greater over the entire diurnal period. As expected, Marseille is ranked as one of the driest sites, with more available energy being partitioned into sensible heat flux terms than is observed at the suburban sites. (The suburban sites studied are generally characterized as containing a fair amount of vegetation and subject to irrigation practices, see Table 3.4.) Not only is this a function of the Marseille site's scarcity of vegetation, but also of its massive active surface. When Marseil le's energy regime is compared to Tucson's (the location with the most similar radiant energy levels), the role of water availability becomes apparent. Because of irrigated vegetation around Tucson, more energy is partitioned into latent heat at that site. Interestingly, among the various locations, Marseille partitions nearly the least amount of daytime energy into storage. In fact, Marseil le's 24-hour flux ratios rival that of Mexico City, where site characteristics are the most similar. It should follow, then, that the relative daytime energy partitioning at these two sites should also be similar but in fact, it is not: nearly twice as much energy is partitioned into QH at Marseille. The reason for this is probably related to the fact that Marseille experienced consistently stronger daytime winds than Mexico City (see Oke et al., 1999), thereby enhancing convection of heat away from the surface. The disparity between flux ratio terms of sensible heat fluxes (QH and AQs) between the Mexico City and Marseille sites is an indication that the degree of surface-atmosphere coupling, which is influenced by 65 99 o c H <C S 1 I 24.62 9.41 8.17 3.63 3.01 2.65 2.22 2.05 1.70 9.90 8.88 6.71 2.31 2.00 1.79 1.55 1.22 1.19 cS °5 2 < N — 1 0 0 " * —< " i ^ O r ^ T j - c S w - i " - ^ ~ 2 < N — — — ~ - H * * 1 o o o o o o o o o ^ t m ^ ^ C N P O O P P o O o o p o O * Cy o o o o o o o o o O N O c - i o m N O O s " / - ) * © o ^ - H ^ H m r o m c n T ^ ^ f O O O O O O O O O * CX o o o o o o o o o o o « o o o - « f ^ v q < o > o ^ - H O O ' O ' O O O ' O c3 TI or~-f>poooot~-rnos <o r-; cr, — ; i/-) oo —; w-i ° «*i T fN CN ° o o O ~ » H <N ^ TJ"" vd * cy rr ° i o ^ ^ ^ r - 0 ; o m f ) " ^ " C 4 < ^ i m ( N m ' t - ^ o o o c i t N t - -Location ( , * <8 S -7- CO £ •£ 3" 3 3 A •§ •§ * >. . 3 tn tn 1 .§ 1 s | § -a ^  i a> &| 1| 1 l l 8 8:3 * .a a. tn y >> .3 tn tn Q qj U u o> c £ 111< | i s I l l i n o i s e s i 3 o •a 3 O 3 *C <u CL X <D J 3 O -o 0) &. cd 6 o convective and dynamic stability as well as surface thermal properties, dictates the nature of energy exchange (conduction Vs. convection). Table 3.3, showing fitted a coefficients to the hysteresis model described by Eq. 1.4, reveals the behavior o f Marseil le's hysteresis pattern in relation to other urban sites. Marseil le's fitted a\ coefficient is certainly within the reasonable range of values. However, this value, which indicates the overall strength of dependence of AQs on Q*, seems surprisingly small, given that Marseille is a very built-up site, similar to that of Mexico City. Although it has been noted at other sites that in general, a\ tends to increase with the area of roofs and impervious surfaces and decrease with area vegetated (Grimmond and Oke, 1999), Marseil le's fitted d\ value does not seem to comply with this scheme (for plan areas of surface cover for each site, refer to Table 3.4). Clearly, the lesser dependency o f AQs on Q* is impacted by other processes and/or factors, most l ikely Table 3.3 Statistical evaluation of the goodness of fit of the hysteresis pattern for a variety of urban study sites. Coefficients ah a2, and a3 are determined from fitting Eq. (1.4) to the mean diurnal data for all-sky conditions. Statistics derived from correlating Eq. (1.4) with the fitted coefficients to the hourly data. Sites are ordered by decreasing fraction of surface cover built (refer to Table 3.4). Source: Non-Marseille data from Grimmond and Oke (1999a) Site a\ « 2 "3 Slope Intercept r2 RMSE h W m 2 W m 2 Mexico City 0.740 0.069 -37.3 0.996 0.093 0.973 25.2 Vancouver industrial 0.568 0.217 -29.2 0.933 4.152 0.885 46.6 Marseille 0.363 0.596 -75.8 0.619 -9.069 0.619 79.7 Tucson 0.385 0.440 -57.2 0.930 • 7.246 0.770 65.2 Miami 0.409 0.428 -36.6 0.971 1.824 0.797 55.6 Chicago 0.288 0.664 -43.5 0.969 34.170 0.606 86.5 Vancouver suburban 0.350 0.656 -48.6 0.982 3.564 0.715 55.4 Sacramento 0.385 0.256 -39.2 0.803 0.394 0.730 51.9 Los Angeles 0.397 0.405 -32.5 1.068 1.153 0.903 41.5 67 related to the strength o f Marseil le's daily wind regime. As stated above, the amount of energy partitioned into storage at this site is highly dependent on the strength of the f low (see Section 4.1). Also important when considering in differences in surface-atmosphere energy exchange between sites are the predominant building materials. In particular, Marseil le's roof materials (predominently clay tiles) differ from those of other, North American, sites (mostly gravel or bitumen or asphalt shingles, Meyn, 2000). North American roof assemblies typically have slightly higher heat capacity values (Oke, 1988), which contributes to the greater overall diurnal energy uptake observed at those sites. Table 3.4 Fraction of plan area surface cover ordered by increasing area vegetated. A v = plan area vegetated, A R = plan area of roofs, At = plan area of impervious ground. Source: Modified after Grimmond and Oke (1999a). Site A V AR A , Mexico City 0.02 0.55 0.44 Vancouver industrial 0.06 0.50 0.44 Marseille 0.16 0.56 0.28 Tucson 0.35 0.24 0.41 Miami 0.35 0.39 0.32 Chicago 0.39 0.38 0.23 Vancouver suburban 0.46 0.30 0.24 Sacramento 0.51 0.37 0.12 Los Angeles 0.55 0.26 0.19 The parameter 02 describes the degree and direction o f the phase relations between the two variables. The positive value from Marseille, like at all sites, indicates that the peak in AQs precedes the peak in Q*, so that the hysteresis loop is clockwise. The magnitude of the hysteresis loop in Marseille ranks among the highest, thereby indicating a relatively greater phase shift, similar to those reported at suburban sites. Interestingly, Fuchs and Hadas (1972) suggest that hysteresis effects in soil increase with 68 decreasing soil moisture. This observation is supported by Grimmond and Oke (1999a), who found that hysteresis at the Vancouver and Chicago suburban sites increased as conditions at those sites dried out (indirectly due to smaller evaporation rates). Although Marseille is considered a dry environment, surface moisture conditions alone probably do not l ikely explain Marseil le's pronounced hysteresis behavior. Rather, it is proposed that the timing and magnitude of the diurnal f low regime is l ikely a definitive factor. A t sunrise, when conditions are relatively calm (average wind speeds approximately 1 4 m s ' see Figure 2.15), the urban fabric is able to accumulate energy efficiently, resulting in a late morning peak in AQs, preceding the peak in Q* by about two hours (a pronounced hysteresis signature). A lso occurring during or soon after the daytime peak in AQs is the daily wind speed maximum, with consistent moderate f low (4-6 m s"1) persisting until sunset. It is during these unstable convective conditions, when the coupling between the boundary layer and the atmosphere is the greatest, that turbulent transport of heat from storage into the atmosphere dominates. The afternoon transfer of sensible heat into the atmosphere is far more efficient than conduction into surface elements (Grimmond et al., 1991). The diurnal wind regime also helps to explain the obvious difference in hysteresis magnitude between Marseille and Mexico City, the site at which surface characteristics most resemble that of Marseille. Not only are both sites very built-up, but thermal properties of the corresponding surface materials also rival each other, allowing for a fair comparison. A t the central Mexico City site, Oke et al. (1999) report consistently near-calm flow (~ 1 m s"1) from the early morning into the late afternoon near sunset, followed by a slight increase to an average of 2 m s"1. These calm and statically-stable conditions 69 suppress surface-atmosphere coupling, thereby allowing more conduction into the urban fabric to occur throughout the day. In this case, the storage heat flux is heavily dependent on the radiation regime (refer to a\ values in Table 3.3), so much so that pronounced hysteresis behavior is not observed (in other words, the daily maximum peaks in zJgsand Q* occur at similar times). The parameter #3, the intercept term, indicates the timing when AQs turns negative. The large value observed in Marseille indicates AQs becomes negative much earlier than Q*. Less energy, therefore, is partitioned into storage in the afternoon than has been observed at other sites. Again, this behavior is l ikely to be an effect of the diurnal wind regime (see discussion above), in which energy is increasingly stripped from the built-up environment during the afternoon and through the evening hours. 3.2 Objective Hysteresis Mode l The Objective Hysteresis Model (OHM) of Grimmond et al. (1991) is implemented for the C A A site to parameterize magnitude and behavior of the energy storage flux. Input coefficients and surface weightings are given, followed by results and performance statistics of the scheme. 3.2.1 O H M Input Parameters Direct measurements of hourly-averaged net radiation and a coefficients derived or modeled for urban surfaces found in the literature, were used to initialize the O H M for this site. Because a significant portion of the plan area consisted of clay tile roofs, for which there are no measured a coefficients, Wi lkes ' Simplif ied Transient Analysis of Roofs model (STAR, 1989; see Appendix B) was used to determine these coefficients following the approach of Meyn (2000). The plan areas of tile roofs, flat roofs 70 (predominantly gravel- or pebble-topped), impervious ground, ^ vegetation, and water that contribute to the energy balance depend on the direction and size of the turbulent flux source area for each hourly period. Using F S A M calculations (Section 2.4.3), the O H M considers the dynamic nature of source areas, generating hourly energy storage flux values unique to that hour's corresponding surface cover. The range of values o f hourly plan area surface cover during the course of the IOP are: 29% - 57% clay tile roofs, 5% -17%o flat gravel roofs, 18% - 40% impervious ground, 8% - 33% vegetation, and 0% - 8% water. Since this site is characterized by its densely-packed high surface roughness elements, O H M was also run to account for the complete (three-dimensional) built surface. Surface cover weightings were recalculated so as to allow for vertical walls, that account for about 50% of the complete surface area. Table 3.5 provides the a coefficients of the individual surfaces used in running the O H M scheme for Marseille. Table 3.5 OHM a coefficients selected for use in the Marseille formulation. Surface «i fli (h) a3 (W m"2) Source Gravel roof 0.26 0.89 -24 Meyn (2000) Clay tile roof 0.07 0.06 -5.0 Meyn (pers.comm.) Impervious ground 0.70 0.41 -38.3 Average of Doll et al. (1985), Asaeda and Ca (1993), Narita et al. (1984), Anandakumar (1999) Vegetation 0.32 0.54 -27.4 Doll et al. (1985) Stone wall 0.85 0.32 -28.5 Asaeda and Ca(1993) Water 0.50 0.21 -39.1 Souche/a/. (1991) 3.2.2 O H M Results Visual assessment of the ensemble mean diurnal measured (residual method) and modeled (OHM) storage heat fluxes (Figure 3.8) for the Marseille site does not show good agreement between the two methods. Quantitative assessment of the scheme is also 71 400 300 <T 200 s & 100 o 0 -100 -200 — i 1 1 1 1 1 r— 1 1 1 r 0 2 4 6 8 10 12 14 16 18 20 22 LAT Figure 3.8 Ensemble mean diurnal values of measured (OBS) and modeled (2D OHM and 3D OHM) storage heat fluxes at the Marseille site. provided: a scatter-plot o f hourly measured vs. modeled (OHM) storage heat flux (Figure 3.9), the parameterized hysteresis loops (Figure 3.10), and the statistical performance o f the O H M in Marseille, compared with other urban sites (Table 3.6). Over the diurnal period, neither O H M formulation handles the magnitude of heat flux to and from storage well. Generally, the 2D O H M overestimates the release of energy from storage during the night by about 50 - 90 W m" while underestimating the daytime maximum uptake o f energy by 20 - 30 W m"2. Further, there appears to be an offset in the timing of the scheme's dynamics. Energy storage fluxes predicted by the scheme turn positive one hour before the measured values in the morning. Similarly, the 2D O H M maintains a positive flux one hour later into the afternoon, suggestive of the scheme's tendency tp overestimate the daytime energy uptake. This estimation is likely not offset by the underestimation of the magnitude of the maximum positive energy storage flux. 72 -400 -200 0 200 400 600 Measured AQS (W m"2) Figure 3.9 Scatterplot of hourly modeled vs. measured storage heat flux (W m"2) at the Marseille CAA site. Statistics of goodness of fit are reported in Table 3.6. The 3D formulation performs particularly poorly in the daytime hours, overestimating energy uptake by approximately 124 W m" . So although the inclusion of wall surfaces is certainly more geometrically representative o f this built-up site, this additional thermal mass results in greater surface conduction (storage uptake) than is actually observed. A t night, however, the 3D formulation does a reasonable job of •y simulating convective AQs release, underestimating the flux by approximately 36 W m" . This suggests that the additional weighting offered by the wails under more stable conditions, when the atmosphere and surface are the most de-coupled, is appropriate to describe surface-atmosphere convective exchanges in this environment. The difference in hysteresis behavior (Figure 3.10) shows the hysteresis relationship between AQs and Q* produced by O H M is not as strong as the observed pattern. Average a values generated by the 2D O H M for this site are: a\ = 0.307, a.2 = 0.335 h, and as, = -20.3 W m" . While the measured and modeled dependency terms (a\) 73 400 300 _ 200 £ ^ 100 < o -100 -200 -200 0 200 400 600 Q* (W m"2) Figure 3.10 Ensemble hysteresis loops of energy storage fluxes derived from measurements (residual) and both OHM formulations. are roughly comparable, the 2D O H M ' s considerable departure in the a-i and a-$ terms is disappointing. Average a values in the 3D O H M formulation are: a\ = 0.577, ci2 = 0.327 h, and ai = -24.4 W m" , clearly not an ideal fit to the measured hysteresis pattern. Accordingly, Figure 3.9 shows that there is little overlap between the 3D O H M and observed hysteresis loops. Because the O H M does not explicitly resolve the impact(s) of the diurnal wind regime, the timing and degree of hysteresis is not captured with this scheme. The performance statistics show that the current formulation of the O H M is not altogether suitable to the estimation o f the diurnal behavior and magnitude of energy storage flux at this site. The 2D O H M scheme in Marseille has the poorest mean discrepancy, as evident by the lowest slope statistic, while the 3D formulation generates a high R M S E statistic. A lso notable is that the highest R M S E statistics are from Tucson 74 and Marseille, which are both characterized as dry and windy sites. It is probable, then, that the sizeable R M S E values from these sites are due in part to the O H M ' s inability to inherently consider the relationship between the sharing of sensible heat flux and energy storage, which is enhanced only further by a lack o f moisture at both environments. Both O H M formulations produce agreement indices, d, o f 0.80, just barely considered satisfactory. Given its good performance at several other urban sites (Grimmond and Oke, 1999a), especially at a site in Mexico City that possessed many similarities to C A A , this poses several interesting questions for the present study. Table 3.6 Statistical performance of the OHM in Marseille and at other urban sites. Fluxes are determined for hours when all four 15-min periods are from acceptable wind directions. Sites are ordered by decreasing r2. Source: Non-Marseille data from Grimmond and Oke (1999a). Intercept r 2 RMSE Site N Slope (Wm2) (W m2) Mexico City 61 0.892 -6.8 0.961 33.6 Los Angeles 424 0.966 -4.0 0.915 29.0 Vancouver industrial 312 0.962 • -4.5 0.880 48.9 Miami 204 0.975 22.9 0.788 61.9 Tucson 75 1.206 66.1 0.748 107.4 Vancouver suburban 464 0.745 30.8 0.674 62.9 Marseille (2D) 240 0.490 35.0 0.611 95.6 Marseille (3D) 240 0.868 80.5 0.584 128.3 Chicago 163 0.806 38.8 0.562 83.3 Sacramento 222 0.548 5.0 0.556 66.0 When the errors o f both schemes are plotted against wind speed (Figure 3.11), the prevailing overprediction o f both methods, even at low wind speeds, is evident. Consistent with the analysis o f Grimmond and Oke (1999a), the greatest overpredictions are associated with higher wind speeds. Unl ike the sites investigated by Grimmond and Oke, however, the Marseille results show that at the greatest wind speeds (> 5 m s"1), there is a trend back towards less overprediction. Also contrary to simulations from other 75 8 7 6 X o o .o • • S 4 t> 3 2 O 2D OHM • 3D OHM 0 -400 -300 -200 -100 0 100 200 300 400 500 600 700 A 0 A - n M d - A 0 S o b s ( W m - 2 ) Figure 3.11 Hourly 2D and 3D OHM errors (predicted - observed) plotted against wind speed. sites is that the error associated with AQs underprediction, for both formulations, seems to be uniform over all wind speeds, rather than concentrated at the lowest wind speeds. 3.3 Town Energy Balance Model Masson's Town Energy Balance Model (TEB), coupled to the ISBA scheme, is run in off-line mode for the CAA to parameterize local-scale energy and water exchanges between the complete urban surface and atmosphere. The model was initialized with observations from the CAA site and its performance is evaluated with the observations presented above (see Section 3.1). Because ESCOMPTE-CLU investigators wish to run blind tests of dynamic and chemistry models, results from YD 183 and 184 are withheld and unfortunately unavailable at this time (Cros et al., 2003). 3.3.1 TEB Input Parameters and Initialization TEB was used to yield surface temperatures and surface energy balances of generic roads, roofs, and walls at the CAA site throughout the IOP. The model uses aerodynamic resistances to generate individual surface output which is weighted to 76 resolve the energy balance for the canyon top, as well as canyon air temperature for the middle of the street (see further discussion in Lemonsu et ah, 2003). Given the warm conditions over the course of the IOP, TEB ' s formulation was slightly modified to include the thermal production of turbulence. This was done by adding the free convection velocity term w* developed from Monin-Obukhov similarity theory (1954; see Lemonsu et al., 2003). For the purposes of the present research, T E B was run for a static modeling domain, which assumes that surface characteristics within a 500-meter radius around the measurement tower are homogeneous. This is in contrast to the dynamic modeling domain, which takes into account the varying size and location of source areas due to changing wind and stability conditions (see section 2.4.3). Sensitivity analyses by Lemonsu et al. (2003) indicate that a static modeling domain is appropriate for this site, since surface characteristics are relatively homogeneous. It is also noted that the additional computational requirements of the dynamic formulation do not yield significantly different or improved results. Tables 3.7 and 3.8 provide the input parameters used to initialize the scheme. Meteorological and incoming radiation observations taken from the top of the C A A measurement tower were used to force the model with a 30-minute time step. Given the high volume of vehicular traffic at this site, an anthropogenic heat flux term, QF, is included in the simulations. Because this term was not directly measured, it is estimated following the method of Grimmond (1992) and is typically very small (15 W m" in the day and 2 W m"2 at night). 77 Table 3.7 Cover fractions and TEB input parameters for the static modeling domain (500-meter radius around the CAA tower) used in the present study. See Table 3.8 for thermal parameters of the materials. Parameter Input Value Cover fractions Natural cover 0.160 Water 0.000 Urban cover 0.840 Building fraction 0.560 Road fraction 0.280 Geometric Parameters Building height 15.6 m Building aspect ratio: HIL 0.78 Canyon aspect ratio: HIW 2.01 Roughness length 1.90 m Road Properties Mean material asphalt and concrete over dry soil Albedo 0.080 Emissivity 0.941 Momentum roughness length 0.05 m Roof Properties Mean material tile or gravel over concrete, wood and insulation Albedo' 0.217 Emissivity 0.902 Momentum roughness length 0.15 m Wall Properties Mean material stone and wood shutters Albedo 0.200 Emissivity 0.900 Momentum roughness length 0.15 m Temperature Initialization Inside building temperature 22.0 °C Deep soil temperature 18.5 °C Roof surface temperature 20.0 °C Road surface temperature 23.5 °C Roof albedo parameter represents a weighted average of values from measurements and the literature: 0.28 for new tile roofs, 0.21 for gravel roofs, and 0.15 for darker tile roofs. Table 3.8 Thermal properties for roofs, walls, and roads used in TEB for the Marseille site. Layer sequence: 1 is nearest to the surface, d is layer thickness in meters, C is the heat capacity of the layer (MJ m"3 K"1) and k is the thermal conductivity (W m"1 K"1). Values initialized for the static modeling domain. Roofs Gravel (11.4%) Tile (88.6%) Layer 1 d 0.02 gravel 0.02 tile C 1.769 1.769 k 1.400 0.840 Layer 2 d 0.15 concrete 0.15 concrete C 1.500 1.500 k 0.930 0.930 Layer 3 d 0.12 insulation 0.12 insulation C 0.290 0.290 k 0.050 0.050 Layer 4 d 0.03 wood 0.03 wood C 1.520 1.520 k 0.190 0.190 Roads Asphalt (60%) Concrete (40%) Layer 1 d 0.04 asphalt 0.04 concrete C 1.940 1.280 k 0.750 0.250 Layer 2 d 0.20 stone aggregate 0.20 stone aggregate C 2T000 2.000 k 2.100 2.100 Layer 3 d 0.50 gravel/soil 0.50 gravel/soil C 1.400 1.400 k 0.400 0.400 Layer 4 d 0.50 gravel/soil 0.50 gravel/soil C 1.400 1.400 k 0.400 0.400 Walls Stone (80%) Wood Shutters (20%) Layer 1 d 0.01 stone 0.01 shutter C 2.250 0.450 k 2.190 0.090 Layer 2 d 0.04 stone 0.04 shutter C 2.250 0.450 k 2.190 0.090 Layer 3 d 0.15 stone 0.15 air C 2.250 0.150 k 2.190 0.0012 Layer 4 d 0.06 stone 0.06 glass C 2.250 1.660 k 2.190 0.740 3.3.2 T E B Results In the interest of brevity, simulations of canyon air temperatures, canyon and roof surface temperatures and radiation fluxes are not provided (for further discussion of these simulations performed for this site, refer to Lemonsu et al. (2003)). Rather, the present discussion will focus on the local-scale energy balance results generated by TEB. The model is run and output is generated at five-minute time steps which are then converted to hourly averages. 3.3.2.1 Modeled Flux Behavior Simulated average surface energy fluxes computed by the TEB-ISBA scheme are compared to observed fluxes in Figure 3.12. The plots show that there is generally good agreement between the modeled and measured terms, during both daytime and nighttime periods. Table 3.9 provides TEB's performance statistics (bias error and root mean square error), which also support the model's ability to accurately predict the surface energy balance at this site. Table 3.9 Performance statistics for mean values of the surface energy balance modeled by the T E B scheme. Bias = T E B - OBS. Al l units are W m"2. Q* QH QE AQs Overall Period OBS 154 159 12 -17 T E B 132 112 18 9 Bias -22 -47 6 26 R M S E 51 75 52 78 Daytime OBS 343 263 30 51 T E B 315 189 33 97 Bias -28 -74 3 47 R M S E 66 100 58 101 Nighttime OBS -69 36 -8 -97 T E B -83 22 1 -96 Bias -14 -14 9 1 R M S E 23 24 47 38 80 a «4-l o B e o £ o pa H T 3 ( L > S O (N (N O A\) ^ ! S « 3 Q x n W ^ 3 U 3 T3 s ^ — \ VI m o o c cu X> VI a o "S s o W The model contains some bias in its handling of the daily timing and magnitude of net radiation. Although T E B successfully simulates a positive nighttime sensible heat flux, QH, the model consistently underestimates the magnitude of this heat flux at all times during the day (average daytime QH bias = 74 W m"2): Lemonsu et al. (2003) speculate that this underestimation is l ikely related to a lack of roof-level mixing resolved by the scheme. Similar to the residual method of estimating energy storage flux, the model also calculates AQs to be the residual to the surface energy balance. It is not surprising, then, that TEB ' s underestimation of the daytime sensible and latent heat fluxes is accordingly reflected as an overestimation of the amount of heat taken up into storage. The overestimation of modeled AQs results in a simulated daytime AQs/Q* ratio (0.31) that is much higher than the observed ratio (0.15). During the nighttime, however, T E B does a suitable job of handling the release of stored energy (average nighttime AQs bias = 1 W m"2) . A closer examination of the modeled storage flux term follows in Section 3.3.2.2 3.3.2.2 Modeled Energy Storage F lux Considering the dynamic and convective instability observed at this very complex site, the Town Energy Balance model does a surprisingly good job of simulating observed surface energy fluxes. A visual inspection of the time series of the simulated energy storage flux term (Figure 3.13) and the corresponding scatter-plot (Figure 3.14) shows considerably more agreement than with observations than with the O H M scheme (Figure 3.8). Whereas O H M tends to estimate the timing of the diurnal peak in energy uptake to occur later (closer to solar noon) than is actually observed, T E B appears to handle the observed behavior well. The model, therefore, reasonably simulates the 82 s < e 400 300 200 100 0 -100 -200 -300 • • • v * • • • 5* • • A 1 I 1 W  1 1 !# Slope = 0.792 ^ Intercept = 22.2 (W m"2) r 2 = 0.67 J=0.89 -300 -200 •100 0 100 200 300 400 Measured AQS (W m) Figure 3.14 Scatter-plot of hourly modeled (TEB) vs. measured (residual) storage heat flux (W m"2) at the Marseille C A A site. Overall statistics of goodness of fit are reported in the plot area. diurnal relation between the AQs and Q* (i.e. hysteresis, Figure 3.14). The hysteresis loop produced by T E B is of approximately the same breadth as the observed loop, indicative of the model's accuracy in predicting the timing of the diurnal flux behavior. 83 300 200 S 100 t © 1 o -100 -200 1 - ' 1 ' ' ' ' ^r^^ - TEB OBS -200 200 6* (W m" ) 400 ,600 Figure 3.15 Ensemble mean measured (OBS) and simulated (TEB) hysteresis loop at the CAA site. The overall upward shift of the simulated loop is merely related to the model's overestimation of the magnitude of the daytime partitioning of energy into storage. TEB ' s relative ability to simulate surface-atmosphere sensible heat exchanges can also be appreciated when examining the hysteresis relationship between AQs and QH (Figure 3.16). The amount of overlap between the two curves points towards TEB ' s 300 200 "a 100 0 ©> < -100 -200 \ ' ' S 1 1 „ . • -" ' - - • - - T E B -•r^ —m— OBS -200 200 400 600 800 QH (W ni"2) Figure 3.16 Ensemble mean measured (OBS) and modeled (TEB) hysteresis relationship between heat storage flux and sensible heat flux. 84 ability to capture the asymmetry in energy uptake by the urban fabric and release to the atmosphere. Again, the scheme's overestimation of daytime energy uptake explains the upward shift of the simulated curve. The behavior of surface sensible heat sharing is correlated to local wind conditions which dictate the degree of surface conduction or convection. Therefore, TEB ' s promising surface energy flux results highlight the model's favorable ability to reproduce surface energetics under varying synoptic, meso-scale, and local f low regimes. Indeed, the model's performance statistics from other dry, built-up sites (Mexico City and a light industrial site in Vancouver, each with very different surface structure and geometry) indicate TEB ' s general adeptness in simulating surface fluxes within an urbanized environment (Table 3.10). Masson et al. (2002) tested the sensitivity of the scheme by altering a variety of input parameters (surface thickness, roughness lengths, surface albedo and emissivity, building fraction, etc.) for the Mexico City and Vancouver light industrial sites. The authors found the model to be robust at both of these sites, and most sensitive to the incoming solar radiation and roof characteristics. Lemonsu et al. (2003), running simulations for the C A A site in Marseille, support the Mexico City and Vancouver light industrial findings and further add that the model is not appreciably impacted by the inclusion of a dynamic source area modeling domain. A lso noted is the model's improved performance when a more specific knowledge of the urban surface is used, highlighting importance of correctly documenting urban surface characteristics. TEB ' s good performance at these sites, coupled with results from the aforementioned sensitivity analyses, further suggest that the model can reasonably be employed to gain more insight into the Marseille site's responses to various surface forcing parameters. 85 Table 3.10 Summary of performance statistics of TEB for heat fluxes (W m"2) at the Mexico City, Vancouver light industrial, and Marseille sites. Source: Non-Marseille date from Masson et al. (2002). Heat flux Q* QH+QE Mexico City Diurnal Period OBS 45 57 -12 TEB 55 54 1 Bias 10 -3 13 RMSE 32 25 39 Daytime OBS 257 108 . 149 TEB 252 . 113 139 Bias -5 5 -10 RMSE 41 32 45 Nighttime OBS -103 21 -125 TEB -82 12 -95 Bias 21 -9 . 30 RMSE 24 18 35 Vancouver Industrial Diurnal Period OBS 150 95 55 TEB 141 133 8 Bias -9 38 -47 RMSE 59 - 76 91 Daytime OBS 323 168 156 TEB 306 234 73 Bias -17 66 -83 RMSE 76 103 121 Nighttime OBS -59 7 -66 TEB -57 11 -68 Bias 2 4 -2 RMSE 24 12 23 Marseille Diurnal Period OBS 154 171 -17 TEB 132 130 9 Bias -22 -41 26 RMSE 51 127 78 Daytime Obs 343 293 51 TEB 315 222 97 Bias -28 -71 47 RMSE 66 188 101 Nighttime Obs -69 28 -97 TEB -83 23 -96 Bias -14 -5 1 RMSE 23 71 38 Rather than evaluating the sensitivity of the scheme itself, T E B wi l l be used in the present research as a useful tool to conduct sensitivity analyses to better understanding 86 the primary criterion affecting the relationship between heat conduction and convection at the urban surface (wind regime, surface geometry and structure). Such analyses are addressed in Chapter 4. 3.4 Thermal Mass Scheme Estimates of AQs from the C A A site using the Thermal Mass Scheme (TMS) are presented. Initial discussion focuses on the construction of the scheme and the required input data, including measured and parameterized surface temperatures o f a representative set of surfaces (see Section 2.3 and Appendix C). 3.4.1 Observed Surface Temperatures Surface temperature observations of the three surface types (streets, building walls, and roofs) that define an urban area's complete built-up surface are given. Data presented are derived primarily from fixed Everest IRT installments and are shown as time series plots and ensemble averages over the ten-day IOP. It is important to note that although surface temperatures are reported according to the compass direction towards which the facets are facing, measurements used in the T M S were conducted on a block which is actually oriented 19° west o f grid north. This has obvious ramifications insofar as solar geometry is concerned and this influence is seen in the surface temperature signatures. For the purposes of clarity, facet labels refer to the compass direction in which the facet predominantly faces (i.e. the SSE-facing wall is considered 'S-facing,' the E N E - W S W oriented road is considered ' E -W oriented,' etc.) 87 3.4.1.1 Observed Canyon Surface Temperatures Observed building wall surface temperatures are shown both as a time series plot (Figure 3.16) and ensemble averages, according to canyon orientation (Figures 3.18 and 3.19). Data are derived from fixed IRT sites near the C A A tower (Croix Rouge and Jurexfi, refer to Appendix C). Because the IRTs field-of-view measured an area covering tens of square meters, data represent the average temperature o f all the surface materials (limestone, wooden shutters, and pane glass) which constitute the wall . Evidence of the facet shading which lends to the flattened L\ curve discussed in Section 3.1.1, is illustrated above. The characteristic narrow urban canyons from which 40 | 20 1 1— 1 1 1 1 1 183 184 185 186 187 188 189 190 191 192 193 YD N-facing wall E-facingwall S-facingwall W-facingwall Figure 3.17 Time series of observed wall temperatures. Data are 30-min averages, derived from 15-min measurements. these measurements were taken facilitate the distinct timing signatures. The east-facing wall is the first vertical surface to experience direct sun exposure in the morning, leading to its relatively early primary diurnal surface temperature peak at approximately 1130 88 local time (about 1.5 hours before solar noon for horizontal surfaces). Being a mid-latitude location in the summertime, east-facing walls at this site intercept most of their daily incident solar radiation in the morning hours, at a time when other surfaces are generally shaded. Occurring at approximately solar noon, the south-facing wal l is the next vertical facet to experience its diurnal temperature peak. Again, this behavior is reflective of the sun's path across the sky throughout the course of the day. A t this latitude in the summer, solar noon occurs approximately ninety minutes after local noon. This helps to explain the relatively sharp temperature gradient observed on the south-facing wall from the time of initial exposure (at approximately 0900) until the temperature/exposure peak between 1300 - 1400 local time. The morning temperature gradient of the west-facing wall is appreciably smaller, due in part to morning shading. Radiant energy reflected by or emitted by the opposing east-facing wall could also contribute to the gradual and steady temperature increase of the west-facing wall, which achieves its daily temperature maximum between 1500 - 1600 local time. O f course, the last vertical facet to peak in surface temperature is the north-facing wall (at approximately 1900), which is the least exposed to direct sunlight. Surface temperature observations from the streets oriented approximately north-south and east-west also display notable differences in magnitude and timing (Figure 3.17). Again, data represent an areal average of both the road (typically constructed of an asphalt layer overlain on stone aggregate and gravel/soil layer) and flanking sidewalks (concrete overlain on stone aggregate and gravel/soil). A n obvious characteristic displayed in Figure 3.17 is the distinct and consistent phase shift between the curves. The north-south running road peaks in temperature first (at about 1400 local time) and 89 achieves a larger average temperature maximum. The surface temperature of east-west oriented streets peaks in the late afternoon (at approximately 1700 local time, four hours after solar noon) and does not appear to warm or cool as much as north-south oriented streets (e.g. the diurnal temperature range of east-west streets is not as large as that o f north-south streets; see Table 3.11). 45 40 2 a. 30 E H 25 20 I 4 . 11 i | n H I ii r v / 1 — • — 1 1 1 1 - , 1 , 183 184 185 186 187 188 189 190 191 192 193 YD N-S Road E-WRoad Figure 3.18 Time series of canyon street surface temperature. Data are 30-min averages. Table 3.11 Summary of diurnal measured surface temperatures ranges of canyon facets. Overall average canyon surface temperatures are calculated by weighting the total surface area occupied by each canyon facet type. Data in this table are derived from the network of fixed IRT's. Surface Average daily maximum (°C) Average daily minimum (°C) Diurnal range ( ° Q N-S road 40.2 24.6 15.6 E-W road 39.3 24.7 14.6 E-facing wall 32.5 23.7 8.8 W-facing wall 33.8 24.1 9.7 S-facing wall 34.5 23.7 10.8 N-facing wall 32.7 22.7 10.0 N-S canyon system 34.7 24.1 10.6 E-W canyon system 34.9 23.5 11.4 90 Figures 3.19 and 3.20 show ensemble temperature plots o f the three surfaces (road/sidewalk and two flanking building walls) which define an urban canyon. These 45 1 Time (Local) Figure 3.19 Ensemble plot of average surface temperature of the three facets which comprise a north-south canyon. Data are from the fixed IRT network near the CAA si plots give a good understanding of the directional variations in overall canyon response to incident solar radiation. Regardless of surface type and canyon direction (and the consequent variations o f sun exposure), the average daily surface temperatures of all three canyon facets suggest upon first glance, that both canyon orientations behave similarly over the diurnal period (Table 3.11). In the case of the north-south canyon, regardless of the timing and magnitude of the behavior o f each surface component diurnal temperature, al l three curves nearly converge during the night, until sunrise. Surface temperature gradients o f east-west canyon surfaces are not as dramatic as those within the north-south canyons, which suggests more subtle sunlit-shaded transitions. 91 45 40 4 20 N-facing wall S-facing wall - — E-W road Air 0 2 4 6 8 10 12 14 16 18 20 22 Time (Local) Figure 3.20 Ensemble plot of average surface temperature of the three facets which comprise an east-west canyon. Data are from the fixed IRT network located near the CAA site. Total canyon thermal response is perhaps better resolved when examining diurnal curves of average canyon temperature, after incorporation of the relative surface area occupied by each surface component (Figure 3.20). Assuming similar construction, east-west canyons seem to achieve a slightly warmer average surface temperature during the 35 r 3 o E 20 E-W Canyon •N-S Canyon •Air -r- I 1 1 1 1 1 — | — 2 4 6 8 10 12 14 16 18 20 22 Time (Local) Figure 3.21 Average ensemble canyon temperature, weighted according to the fraction of surface area occupied by each canyon component (two walls and a road). 92 day, and subsequently cool to a lower temperature at night. Since each wal l surface occupies approximately the same relative surface area in its respective canyon (39%), the ability of the east-west canyon to achieve and maintain a warmer average surface temperature is l ikely due to the relatively disproportionate amount of sun exposure intercepted by the south-facing wall . The overall east-west canyon warming resulting from persistent south-facing exposure is apparently not significantly off-set by cooler north-facing wall shading. To gain better understanding of the spatial variability of surface temperatures over each canyon facet measured by the fixed infrared thermometer network, Table 3.12 gives a statistical summary of results from a street-level survey of canyon surface temperatures conducted from 1545 - 1630 local time on Y D 191 (refer to Figure 2.11 in Section 2.3.2 for sensed locations within the measurement site). This table shows a large range of measured temperatures over sunlit and shaded surfaces (in this case, stone walls, wooden shutters, and glass windows). Although the data in Table 3.12 describe surface thermal responses near solar noon, a considerable degree of shading is apparent. North- and east-facing walls and north-south oriented streets appear to be completely shaded at this time. East-west street surfaces record the largest temperature values and west-facing walls are the warmest vertical surface. In terms of wall materials, under similar sun exposure, stone walls are consistently warmer than wood shutters or windows. While resolving spatial and temporal variability of temperature over various surface types which comprise a particular facet is intriguing, for the purposes of the Thermal Mass Scheme, only average surface temperatures of each facet wi l l be employed. 93 Table 3.12 Statistical summary of canyon surfaces temperatures (°C) derived from a hand-held IRT street-level survey. Data was gathered in the early- to mid-afternoon of YD 191. All data are average results from two measured facets (e.g. measurements were done over two north-facing walls, two east-west roads, etc.), with the exception of just one south-facing wall being represented. West-facing wall South-facing wall Average G Max Min Average a Max Min Sunlit wall 33.4 2.2 37.2 28.2 30.6 1.3 32.4 28.6 Sunlit window 30.0 3.8 35.4 ' 25.8 28.7 1.7 30.5 26.5 Shaded wall 28.1 : 1.6 31.1 25.9 27.0 2.1 30.6 22.3 Shaded window 28.7 . 2.6 32.6 26.0 Shaded shutters 27.8 1.6 30.6 26.4 East-facing wall North-facing wall Average o Max Min Average a Max Min Sunlit wall Sunlit window Shaded wall 27.8 1.2 30.0 25.7 28.5 1.7 33.5 25.9 Shaded window 27.2 1.3 29.4 25.2 27.8 0.6 28.4 27.1 Shaded shutters 26.2 0.5 26.9 25.8 North-South road East-West road Average o Max Min Average o Max ' Min Sunlit road 40.5 4.8 46.4 33.5 Shaded road 31.2 3,0 35.9 27.3 29.4 3.6 35.0 26.6 3.4.1.2 Observed Roof Surface Temperatures Roof surface temperatures used in the present analysis are derived from fixed Everest infrared thermometers as wel l as from infrared thermal scanner images (see Section 2.3.2). Complete measurements of all the roof surfaces of interest (flat gravel-topped roof and clay tile roofs pitched 17° in the four compass directions) could not be conducted over the course of the entire IOP. To compensate for missing measurements, a simple parameterization scheme was derived and applied. Refer to Appendix D for details of the scheme. Figure 3.22 is the time series plot of measured and parameterized roof surface temperatures over the IOP. Ensemble diurnal plots derived from measured values from each surface are shown in Figure 3.23./ 94 183 184 185 186 187 188 189 190 191 192 193 Y D Gravel N-facingtile W-facingtile E-facingtile - — S-focing tile Figure 3.22 Time series of measured and parameterized roof surface temperatures over the duration of the IOP. Data are 15-minute averages. Unlike walls and roads, roof surfaces do not appear to have very different peak timing signatures. Regardless of orientation, all of the roof surfaces appear to achieve their maximum diurnal temperature close to solar noon (1400 local time). This is because at this site, roofs are, for the most part, unobstructed surfaces and therefore not susceptible to a large range of different sun exposures over the course of the day. The tiled roofs are, however, pitched at 17°, so that regardless of their open exposure, not all tile roof surfaces intercept exactly the same amount of incident solar radiation. Assuming consistent thermal properties between clay tile roofs, Figure 3.22 and Table 3.13 show that this fairly slight roof slant does give differential heating: south-facing tile roofs receive appreciably more radiant energy than north-facing tile roofs, accounting for the fifteen-degree difference in the average daily maximum surface temperature. 95 0 2 4 6 8 10 12 14 16 18 20 22 Time (Local) Figure 3.23 Ensemble average measured roof surface temperatures over the 10-day IOP. Table 3.13 Summary of the diurnal range of measured surface temperatures of five roof surfaces. Data in this table are derived from the network of fixed infrared radiation thermometers. Average daily Average daily Diurnal range Surface maximum ("C) minimum (°C) (°Q Flat gravel roof 46.7 18.0 28.6 E-facing tile roof 38.5 18.3 20.2 W-facing tile roof 34.3 20.5 13.7 S-facing tile roof 53.1 19.4 33.7 N-facing tile roof 37.5 16.9 20.6 Regardless o f the roof surfaces' diurnal temperature range, all surfaces cool to within 3 - 4 °C of one other at night. Interestingly, all roof types efficiently release energy at a rate which allows them to cool to a point below measured air temperature. On the other hand, canyon wall and road surfaces remain warmer than the nocturnal air. This observation lends support to the claim that canyon surfaces l ikely play a larger role than roofs in the surface-atmosphere energetics at this site, releasing sufficient energy to support an upward sensible heat flux during the night, in the day (see Table 3.8). This point wi l l be explored further in the following section and Chapter 4. 96 3.4.2 Therma l Mass Scheme Results Measured surface temperatures, combined with user-defined surface layer thermal properties and thickness, are used as inputs to calculate AQs according to equation 1.7. To achieve consistency with the Town Energy Balance model, a few assumptions are made: 1. Each surface (roof, road, and wall) is sub-divided into two surface cover sub-categories (refer to Table 3.8). The proportion of surface area occupied by each sub-category is defined by the user. In the case of the current study, wall surfaces are defined to be covered by 80% stone and 20% wooden shutters. Road/sidewalk surfaces are 60% asphalt and 40% concrete. Roof surfaces are assumed to be homogeneous and either gravel or clay tile. To account for both roof types within the study area, a generic roof surface is employed, the composition of which is defined by the observed plan areas of each: 11% gravel and 89% clay tile. 2. Similar to TEB ' s thermal property inputs (Table 3.8), the depth and material composition of the active layer of each surface sub-category is user-defined. Each layer can be constructed of up to four different materials and are assumed to be flush against one another (unless air is a pre-defined layer). 3. Corresponding to each defined layer, heat capacity values of construction materials from the literature are integrated wdth respect to volume to compute an average C for each active layer. Finally, average surface (road, wall, and roof) heat capacity values are calculated by weighting sub-category (e.g. asphalt vs. concrete, stone vs. wooden shutters, gravel vs. clay tile) values with respect to occupied surface coverage. 97 The ensemble mean diurnal cycle of AQs estimated with the T M S approach is shown in Figure 3.24, with the scheme's 10-day, hourly-averaged performance statistics 300 0 2 4 6 8 10 12 14 16 18 20 22 LAT Figure 3.24 Ensemble mean diurnal cycle of AQS estimate from the Thermal Mass Scheme (TMS), plotted as hourly averages with measured (OBS) values. included in the scatter-plot (Figure 3.25). T M S results are not in very good agreement with those from the residual approach. The scheme underestimates the nighttime release 500 ^ 400 S 300 ^ 200 100 0 & 5 •o | -100 1 -200 -300 -400 • • -400 -200 0 200 400 600 Modeled AQS (W m"2) Figure 3.25 Scatter-plot of hourly modeled (TMS) vs. measured (residual) storage heat flux (W m"2) at the Marseille CAA site. Overall statistics of goodness of fit are reported in the plot area. 98 of stored energy by about 65 W m"2 as well as underestimates the daytime uptake by approximately 27 W m"2. The late-morning spike in storage uptake (beginning at about 1000 LAT ) and the subsequent early-afternoon dive towards energy release are interesting and notable features. A n explanation o f this peculiar behavior is perhaps understood when the average storage uptake/release by each surface type is considered (Figure 3.26). 300 I ROOF -200 0 2 4 6 8 10 12 14 16 18 20 22 LAT Figure 3.26 Ensemble (hourly) A Qs of the three built surface (roads, roofs, and walls) resolved by the TMS. The sum of the three surfaces is shown in the TMS curve in Figure 3.24 above. A l l three surfaces maintain a negative AQs through the night, the sum of which is not as large as the observed AQs release. Roofs are the first surface type to peak in storage uptake and do so at sunrise, when initial solar exposure causes the largest temperature gradients. The roofs' relatively low average C (0.995 M J m"3 K" 1) counteracts the fact that roofs constitute the largest built-up volume considered (-35%). Considering the larger average C of walls (1.894 M J m" K" ) , these surfaces do not play a 99 predominate role in energy exchanges. This is because compared to roads and roofs, wal l surfaces do not experience a large diurnal temperature range, thereby tempering the degree of storage change within the wal l volume. In this formulation, roads are the most thermally responsive to surface temperature forcing, experiencing dramatic surface temperature gradients (on the order of ~8 °C over a 15-minute sampling period) upon initial sun exposure and shading. These large surface temperature gradients cause the scheme to, in a sense, 'blow-up' and create the large late morning spike in energy uptake by the road volumes and subsequent dip towards an early energy release. In order to avoid such misrepresentative occurrences, higher resolution surface temperature input values are recommended. 3.4.3 Sensitivity of the T M S This section briefly discusses the relative sensitivity o f user-defined building construction input parameters to the Thermal Mass Scheme. Measured and parameterized surface temperatures presented in Section 3.4.1 are used to run the modified simulations. Results are presented as ensemble plots of the diurnal bias of AQs estimates for each surface type (roof, wall and road). The bias is defined to be the departure of the modified simulated flux from the reference value (reference run shown in Figure 3.26). The scheme was found to be the most sensitive to wall thickness (Figure 3.27), with walls twice as thick as the reference value (see Table 3.8) absorbing tens of W m"2 more energy in the morning. A t night, the positive bias indicates a greater convective release of energy to the atmosphere. The opposite behavior is noted when walls are one-half the reference thickness. Very little bias is reported when walls are constructed 100 completely of stone (no shutters). This lack of bias, however, is l ikely to be more a function of the forcing surface temperature measurements, which are conducted over a surface area dominated by the stone surface. Figure 3.28 shows the bias when modifications to road surfaces are considered. Clearly, the magnitude of the diurnal biases is not as great as those reported for wall surface modifications. Roads twice as thick as the reference run (see Table 3.8) are shown to store only < 10 W m" more energy. A s expected, road surfaces one-half the thickness of the reference run see less heat conduction (on the •y order of a few W m") during the day, with the same proportion of energy lost to the atmosphere at night. Changes to the plan area of road surfaces (plotted as '100% plan 'y area asphalt' and '100% plan area concrete') result in biases within 2 W m" and are 101 § - - road thickness / 2 H -4 -road thickness x2 -6 -100% plan area asphalt -8 -100% plan area concrete I - 1 0 "I 1 1 1 1 1 ] 1 — 1 I 1 1 1 0 2 4 6 8 10 12 14 16 18 20 22 LAT Figure 3.28 Diurnal ensemble of TMS bias (= reference - modified) with changes to road thickness and composition. therefore deemed negligible. Related to canyon solar geometry is the timing of the curves' maxima and minima, which are later than those pertaining to wal l surface AQs biases, due to road surface exposure occurring later in the day. Among the four modifications, very little bias is seen during the night (< 2 W m" ). The fact that biases from road surface alterations are a relatively small fraction o f AQs flux into/out of these surfaces lends support to the claim that the most important forcing in this scheme are surface temperature measurements. The construction of and percent cover of roof surfaces contribute the least sensitivity to the T M S scheme (so little, in fact, that a plot of individual biases is not relevant). When roof thickness is simulated to be twice that or one-half that of the * 2 reference value, resulting biases are within 1 W m" over the entire diurnal period. Even 102 less bias (on the order of 0.5 W m") is reported when roof surface plan area is altered (100% clay tile, 100% gravei, 50% clay tile/50% gravel). 103 CHAPTER 4: MODELED SENSITIVITY OF AOs: The Town Energy Balance (TEB) numerical model is implemented to examine the relative sensitivity of AQs to various meteorological, geometric, and surface thermal / properties. Fol lowing the recommendations of Lemonsu et al. (2003), the model is run using a static modeling domain so that surface characteristics within a 500-m radius around the C A A site are held constant. Caution is required in interpreting the results that follow, because they are numerical simulations and have not been validated against measurements. 4.1 Sensitivity to Wind Speed The degree to which AQs behavior is impacted by increased or decreased measured wind speed U is examined. Aside from wind speed modifications, all surface, radiometric, and meteorological input parameters presented in Section 3.3 are held constant in the following simulations. The model is forced with measured hourly wind speed data over the eight-day period. Modifications to the original forcing file are done by multiplying every hourly measurement by 0.25, 0.50, 0.75, 1.25, 1.50, 1.75, and 2.0 of its original value. Figure 4.1 shows the reference wind speed (LO as wel l as the range of wind speeds over which simulations were performed (0.25*U, 2*U). T E B output final energy balance data as 30-minute averages over the eight-day period, which are converted to hourly averages for analysis. Data are presented as ensemble hourly means over the period Y D 185 - 192. Figures 4.2 and 4.3 show the time series of AQs partitioning as a result of altering the strength of the f low regime at this site. The general behavior in the plots is as expected, with more energy uptake by the urban surface under weaker daytime f low and 104 18 -, 0 -I ——i 1 1 1 1 1 1 ' 185 186 187 188 189 190 191 192 193 Time (Local) Figure 4.1 Time series of reference wind speed (U) and the range of modified wind speeds ( 0.25*U, 2*U) used to force TEB. All data are hourly averages. -200 i 1 1 1 1 1 1 1 1 1 1 r-0 2 4 6 8 10 12 14 16 18 20 22 LAT Figure 4.2 Ensemble time series of simulated AQS under a weaker flow regime than measured. Here, REF = the reference simulation against which the modified simulations are analyzed and U = reference (measured) wind speed. less conduction to the substrate with stronger daytime flow. Conversely, at night, weaker wind speeds and a greater accumulation of daytime energy facilitate enhanced exchange of surface-atmosphere sensible heat, i.e. a greater magnitude o f nocturnal AQs release. Consistently stronger nighttime flow, on the other hand, results in less energy being released to the atmosphere, due to reduced daytime energy uptake by the substrate. It is, however, the magnitude of the simulated fluxes (shown in Table 4.1) which is perhaps more intriguing. For example, simulations performed with wind speeds 75% less than measured values (^MOC/L^REF = 0.25) results in an overall average AQs value almost twice that of the reference simulation. Increasing the wind speed by the same magnitude, however, results in an overall average AQs value that is only 65% of the magnitude of the reference value. Comparing the magnitudes of the average daytime and nighttime flux biases (= reference - modified AQs) at a given wind specification (Table 106 4.2) shows that more variability generally occurs in the daytime, with flux differences sometimes twice the magnitude of those simulated at night. Table 4.1 Summary table of TEB simulations run with wind speed modifications. REF =' the reference simulation against which the other runs are analyzed. All flux values represent hourly averages over the eight-day modeling period. Average AQs (W m"2) % change ^MOD/C^REF Diurnal Daytime Nighttime Diurnal Daytime Nighttime 0.25 16.3 120.7 -107.0 87 24 11 0.50 13.0 111.0 -102.9 49 14 7 0.75 10.5 103.5 -99.3 21 6 3 R E F 8.7 97.5 -96.2 1.25 7.4 92.6 -93.3 -15 . -5' -3 1.50 6.4 88.5 -90.6 -26 -9 -6 1.75 5.7 84.9 -87.9 -35 -13 -9 2.00 5.0 81.6 -85.4 -42 -16 -11 Table 4.2 Summary of average diurnal, daytime, and nighttime AQS bias (REF - MOD) in varying wind speed regimes. Bias (W m"2) {/MOD/^REF Diurnal Daytime Nighttime 0.25 -7.6 -23.2 . 10.9 0.50 -4.3 -13.5 6.7 0.75 -1.8 -6.0 3.1 1.25 1.3 4.9 -2.9 1.50 2.3 9.0 -5.6 1.75 3.1 12.6 -8.3 2.00 3.7 15.9 -10.8 Time series plots o f the AQs bias helps to point towards times of the day/night when the effects o f wind speed are the most (or least) pronounced (Figures 4.4 and 4.5). Nighttime biases under reduced f low conditions are relatively small and for the most part, remain positive, indicating greater nocturnal release of AQs. With enhanced nighttime flow, the 107 biases are negative and the magnitudes grow progressively larger, indicative of an inverse relationship between nocturnal energy release and increasing nocturnal wind speeds. 30 i -60 \ 1 1 1 —r- 1 1 1 1 1 1 1 ' 0 2 4 6 8 10 12 14 16 18 20 22 LAT Figure 4.4 Ensemble diurnal time series of simulated AQS bias under wind speeds weaker than the reference. -30 -I 1 1 - r - 1 1 1 1 1 1 i r-0 2 4 6 8 10 12 14 16 18 20 22 LAT Figure 4.5 Ensemble diurnal time series of simulated AQS bias under greater relative wind speeds. Note the change of scale relative to Figure 4.3. 108 Figure 4.6 suggests that the relationship between wind speed and simulated AQs is fairly straight-forward and numerically stable. The dependence of AQs on modified wind speeds is best described with a second-order polynomial of the form: AQs = a(UUo\VUKE¥)2 + &(T/MOD/£W +C (4.1) where the a, b, and c coefficients, along with the r2 statistic for goodness of fit are given in Table 4.3. E < 150 100 50 0 -50 -100 -150 0.25 —r-$ €> -~ ST $ < -ffr • — • * — • — — 1 — • — 1 • h 1 1 1 0.50 0.75 •Overall REF 1.25 U MOl/U REF -0—Daytime — 1.50 1.75 2.00 -Nighttime Figure 4.6 Dependence of AQS on modified wind speed over the diurnal, daytime, and nighttime periods (data from Table 4.1). A second-order polynomial (given in equation 4.1) is generally found to be a best fit to the curves. Table 4.3 a, b, and c coefficients fitted to a second-order polynomial equation (4.1) which describes the dependence of A Qs on modified wind speeds. a b c r2 Overall* 0.1240 2.6689 18.009 0.9862 Daytime 0.3033 -8.1806 126.650 0.9955 Nighttime -0.0879 3.8449 -110.380 0.9996 *A slightly better fit to the data set (r1 = 0.9973) was found with a logarithmic relation of the form: AQS = -5.3867ta(IW</REF) + 16.349 109 Since urban surface energy exchanges are presumably dictated by the degree of coupling between the surface and atmosphere, it should fol low that simulated turbulent sensible heat fluxes (QH) under varying wind speed regimes are inversely related to the simulated effects seen with AQs. This exchange is also resolved in the T E B simulations, as the time series of simulated QH (Figures 4.7 and 4.8) show. 400 350 300 .O 250 a £ 200 OI 150 100 50 0 0 2 4 6 8 10 12 14 16 18 20 22 LAT Figure 4.7 Ensemble time series of simulated sensible heat flux under reduced wind speeds. When the f low is weaker than the observed pattern, correspondingly less energy is partitioned into turbulent sensible heat. These conditions foster less atmosphere-surface coupling, which gives greater efficiency of heat conduction into storage. Interestingly, even under less ambient flow, simulated convective sensible heat fluxes remain positive at all times during the day and night, supported by the conductive energy release from the urban fabric. Wind speeds 75% less than the reference value are capable o f giving an 85 W m" variation in peak QH. 110 450 400 REF 350 1 ' 2 5 * U / ^ ^ K 'a 300 250 -1.50*U / / / 200 1.75*U \ 150 100 2.00*U / / \ 50 0 -50 1 1 1 1 1 1 1 1 1 1 1—-0 2 4 6 8 10 12 14 16 18 20 22 LAT Figure 4.8 Ensemble time series of simulated turbulent sensible heat flux QH under enhanced wind speeds. Note the change of scale relative to Figure 4.7. Increased wind speeds considerably enhance the convective exchange between the surface and the atmosphere, especially when coupled with the general static instability at this site. During the night, at the highest forcing wind speeds, turbulent sensible heat flux briefly dips below 0 W m" (over the entire nighttime period, however, an average positive sensible heat flux is simulated; refer to Table 4.4). In this case, the enhanced f low effectively acts to strip the surface o f heat at all times during the day and night. When the simulated wind speed is twice that of the reference simulation, approximately 60 W m" more energy is partitioned into peak turbulent sensible heat flux. A consequence of this regime, then, is a lack o f accumulated stored energy during the day. On the other hand, under a lighter wind regime, energy accumulated during the daytime acts as an enhanced energy source which supports nocturnal unloading. 111 Table 4.4 Summary table of TEB QH simulations run with wind speed modifications. REF = reference simulation against which the other runs are analyzed. All flux values represent hourly averages over the eight-day modeling period. Average Q „(Wm- 2) % change E^MOD/^REF Diurnal Daytime Nighttime Diurnal Daytime Nighttime 0.25 90.7 146.8 24.4 -19 -22 10 0.50 99.5 163.6 23.6 -11 -13 6 0.75 106.7 177.5 23.1 -5 -6 4 REF 112.3 188.6 22.3 1.25 116.7 197.6 21.1 4 5 -5 1.50 120.1 205.2 19.6 7 9 -12 1.75 122.8 211.6 17.8 9 12 -20 2.00 124.9 217.3 15.7 11 15 -29 Table 4.5 Summary of average overall, daytime, and nighttime QH bias (REF - MOD) over varying wind speed regimes. Bias (W m"2) f^MOlV^REF Diurnal Daytime Nighttime 0.25 21.7 41.8 -2.2 0.50 12^ 9 25.0 -1.4 0.75 5.6 11.1 -0.8 1.25 -4.4 -9.0 1.2 1.50 -7.8 -16.6 2.6 1.75 -10.4 -23.0 4.5 2.00 -12.5 -28.7 6.5 Another way to conceive the degree of surface-atmosphere sensible heat sharing under varying wind regimes is to consider the proportion of available radiant energy (Q*) accounted for by the heat storage flux and convective sensible heat flux terms, respectively (Table 4.6). It is encouraging that at all times during the day and night, regardless of the wind speed, T E B maintains a consistent relative amount of total sensible heat flux (QH + AQs): 91-92% of Q* overall, 91% ofQ* during the day, and 89% of Q* at night. Therefore, 112 Table 4.6 Summary of daytime, nighttime and diurnal average flux ratios over varying wind regimes. REF = reference simulation. AQWQ* QH/Q* C^MOD/^REF Diurnal Daytime Nighttime Diurnal Daytime Nighttime 0.25 0.14 0.41 1.16 0.77 0.50 -0.27 0.50 0.11 0.37 1.16 0.81 0.54 -0.27 0.75 0.08 0.33 1.16 0.83 0.57 -0.27 REF 0.07 0.31 1.16 0.85 0.60 -0.27 1.25 0.05 0.29 1.14 0.86 0.62 -0.26 1.50 0.05 0.27 1.13 0.87 0.64 -0.24 1.75 0.04 0.26 1.11 0.88 0.65 -0.22 2.00 0.04 0.25 1.08 0.88 0.66 -0.20 while the degree of hysteresis between net radiation and either of the sensible heat fluxes varies according to f low strength, the hysteresis relationship between energy storage and turbulent sensible heat flux remains constant, as it should in this fairly dry environment (Masson et al., 2002). Over the range of simulations, wind speed alters daytime energy partitioning by 16% of available net radiation. The dependence of sensible heat flux ratios on wind speed is illustrated in Figures 4.9 and 4.10. A A A • • • w Overall — * " • ^ —0—Daytime A Nighttime 0.25 0.50 0.75 REF 1.25 1.50 1.75 2.00 ^ M O D ^ R E F Figure 4.9 Average energy storage flux ratio AQS/Q* versus fraction of reference (measured) wind speed ({/MOD/^REF), for the nocturnal, daytime, and diurnal periods. 113 -0.4 H r : , 1 1 1 1 1 0.25 0.50 0.75 REF 1.25 1.50 1.75 2.00 U MOl)IU REF Diurnal 0 Daytime —•—Nighttime Figure 4.10 Average turbulent sensible heat flux ratio QHIQ* versus fraction of reference (measured) wind speed (CMOD/^REF), for the nocturnal, daytime, and diurnal periods. 4.2 Sensitivity to Urban Geometry TEB ' s user-defined input parameters allow for sensitivity of energy storage flux to town geometry criteria (in the case of the current research, canyon aspect ratio and plan area occupied by built surfaces) to be tested. 4.2.1 Sensitivity to Canyon Aspect Ratio (H/W) The first set of analyses correspond to variations in canyon aspect ratio (H/W). For simplicity, the plan areas o f vegetation, water, buildings and streets are held constant at 16%, 0%, 56% and 28%, respectively. Canyon H/W are modified by adjusting the average building height (the reference height is taken to be 15.6 m), and then entering the desired H/W (the reference H/W =2.0) . This procedure ensures that canyon width remains constant between runs, so that only changes to the wal l surface area in contact 114 with the atmosphere are considered. As Table 4.7 demonstrates, adjusting the canyon H/W does not impact the amount o f energy taken up or released from storage remarkably. . Table 4.7 Sensitivity analysis to varying canyon HlWiov the Marseille CAA site. Average AQs (W m-2) Bias (W m"2) H/W Diurnal Daytime Nighttime Diurnal Daytime Nighttime 0.5 9.9 99.6 -96.0 -1.2 -2.1 -0.2 1.0 8.7 97.6 -96.4 0.0 -0.1 0.2 1.5 8.5 97.3 -96.3 0.2 0.2 0.1 REF 8.7 97.5 -96.2 2.5 9.0 97.9 -96.0 -0.3 -0.4 -0.3 3.0 9.3 98.3 -95.9 -0.6 -0.8 -0.3 3.5 9.6 98.7 -95.7 -0.8 -1.2 -0.5 4.0 9.8 99.0 -95.6 -1.1 -1.5 -0.4 These results are consistent with those of Masson et al. (2002), who found that modifications to canyon aspect ratios in simulations of central Mexico City and Vancouver (light industrial) did little, on the order of ~2%, to impact diurnal energy uptake/release. A s at the Mexico City and Vancouver light industrial sites, diurnal variations in flux ratios (AQs/Q*) were very small (± 0.01 - 0.02 from the reference value, not shown). Modeled flux variations are so small they are barely discernible so that displaying an ensemble time series of AQs as a function of H/W is not relevant. The diurnal pattern of the bias (AQSREF - AQSMOU) does, however, demonstrate some interesting features (Figure 4.11). A t most times through the night, H/W values less than the reference simulation show positive biases, indicative o f enhanced nocturnal storage releases. This is l ikely related to less nocturnal trapping of heat within the urban canyon, due to considerably smaller wal l surfaces and increased sky view factors. In this configuration, heat is more effectively stripped from the surface to the atmosphere above, as the local-scale f low is 115 2 LAT Figure 4.11 Diurnal ensemble time series of AQS bias (REF - MOD) for TEB simulations with varying canyon aspect ratio (H/W). Per field observations, the reference///UK =2.0. better able to penetrate down into the canyon airspace, i.e. the f low can likely be categorized more appropriately as 'wake interference,' rather than 'skimming,' flow. Although smaller H/W ratios mean less thermal mass involved in energetic exchanges, an explanation of the slighter higher daytime energy uptake is l ikely related to solar geometry. In this configuration, lower building heights lead to greater canyon sky view factor and more wall exposure to available radiant energy Q* (which, over all runs and times of day and night, is within 1 W m"2 of the reference simulation). In turn, this generally results in higher surface temperatures, especially road surface temperatures, simulated by the T E B (Masson, 2000). Also possibly related to solar geometry is the slight phase shift in daytime AQs bias associated with H/W values progressively increasing from 2.0. Overall canyon shading effects increase with building heights, so 116 that maximum energy uptake by the overall canyon system occurs later and later in the day, after all canyon surfaces have undergone maximum solar exposure. 4.2.2 Sensitivity to Bu i ld ing P lan A rea Altering the percent plan area occupied by buildings, while holding the average building height and canyon aspect ratio (15.6 m and 2.0, respectively) constant, is shown to have a larger impact on local-scale AQs partitioning than modifications to canyon geometry alone (Figure 4.12). In this set of simulations, surface thermal and radiative parameters, along with plan area o f vegetation and water, are unchanged, so that the proportion of impervious ground and buildings used in the final local-scale surface energy balance are adjusted. The reference plan area of buildings is 56%. 250 REF 200 25%Bldg. J ^ ^ ^ \ 150 35% Bldg. / X ^ X a 100 45% Bldg. / 50 65% Bldg. / 0 75% Bldg. / \m < -50 -100 -150 r— 1 1 r , , _ I I I I I I 0 2 4 6 8 10 12 14 16 18 20 22 LAT Figure 4.12 Ensemble diurnal time series of simulated AQS resulting from modifications to the plan area occupied by buildings and impervious ground. There is a stable, positive relationship between AQS behavior and plan area o f built surfaces (see Table 4.8 below). When plan area o f buildings is smaller than the reference (actual) value and road surfaces constitute the difference, the amount o f overall 117 available radiant energy partitioned into storage is smaller. Since the relative proportion of simulated built surfaces is static, these results show that, when integrated over the local scale, building volumes within the C A A source area play a larger role in energy storage partitioning than do road volumes. On the one hand, these results are somewhat unexpected, as increasing the plan area of buildings consequently increases the roof surface area computed in the integrated (local-scale) surface energy balance. One would expect that, to some degree, the increase in wall surfaces (which have relative high heat capacities) would be tempered by the somewhat smaller heat capacity of gravel/tile roof surfaces (refer to Table 3.8 for values). Table 4.8 Sensitivity analysis to varying plan area of built surfaces for the Marseille CAA site. Average AQs (W m"2) Bias (W nf2) % Bldg. Diurnal Daytime Nighttime Diurnal Daytime Nighttime 25 5.8 91.6 -95.7 2.9 5.9 -0.5 35 5.8 92.5 -96.8 2.9 5.0 0.6 45 6.7 94.5 -97.0 2.0 3.0 0.8 REF 8.7 97.5 -96.2 65 11.1 100.3 -94.4 -2.3 -2.8 -1.8 75 14.3 103.0 -90.5 -5.6 -5.5 ' -5.6 Although not shown here, the proportion of available radiant energy which is partitioned in to storage (AQslQ*) changes by no more than 5% over all simulations and times of the day and night, with the greatest departure occurring at night, when plan area of buildings = 75%. Asymmetry in energy uptake/release when plan area of buildings is greater than or less than the reference value is seen more clearly in the plot of the bias (Figure 4.13). Bui lding plan areas greater than the reference value take up less energy in the morning but then compensate for that deficit with an afternoon positive energy 118 0 2 4 6 8 10 12 14 16 18 20 22 LAT Figure 4.13 Diurnal ensemble time series of AQS bias (REF - MOD) for TEB simulations with varying plan area of buildings. Per field observations, the reference plan area of buildings = 56%. storage flux that is greater than the reference simulation. The opposite behavior occurs when the plan area o f buildings is smaller than the reference simulation. A configuration containing less buildings and more road surfaces stores slightly more early to mid-morning energy but then decreases its storage capacity in the afternoon hours (positive bias). Such a diurnal response is the combined effect of material thermal behavior and solar geometry. In essence, when more building volumes/roof surfaces are considered in the final integrated surface energy balance, more energy uptake, occurs primarily in the afternoon hours. Fewer canyons over the modeling domain thereby enhances the role of roofs in controlling sensible heat sharing between the built surface and the air. A configuration such as this, when the majority of horizontal or near-horizontal surface areas considered are not susceptible to shading, minimizes the impacts of solar geometry 119 resolved by TEB. Differential surface heating caused by shading was illustrated by Masson et al. (2002), who showed that increasing the plan area of buildings decreases average road temperature while increasing average roof temperature (changes to wall temperatures are not reported). At night, when the effects of shortwave radiative forcing are removed, the simulated AQs between runs nearly converge (bias ~ 0 W m" ). The results of Section 4.2 reveal that although some bias exists when urban geometry at this site is modified, the simulated impacts are not as pronounced as in the case of wind speed modifications. 4.3 Sensitivity to Surface Radiative and Thermal Parameters In this section, modifications to TEB input parameters pertaining to surface material characteristics (primarily roof, wall, and road thickness, albedo and emissivity Values) are examined. When considering the relative impacts of surface radiative parameters, the largest biases occur in the daytime hours and are associated with higher surface albedo values (Figure 4.14). Less absorption of incident solar radiation by the surface means that less total energy is accumulated in storage over the day. Consequently, proportionately less energy is lost to the atmosphere at night (Table 4.9). Variations to radiative parameters result in differences of less than 5% irt AQslQ* between simulations, with changes to roof albedo showing the most sensitivity, for each daytime and nighttime period. Although not shown here, higher roof albedos resulted in average roof temperature over 1 K less than the reference value between simulations, consistent with results of simulations in Mexico City and Vancouver (Masson et al, 2002). This degree of sensitivity is expected, as the highly exposed and nearly horizontal roof surfaces are not involved in canyon 120 trapping (Masson et al. 2002). The built-up nature of this site lessens the influence of the road surface albedo, since roads are very rarely exposed to direct incident solar radiation. 30 i 10 12 14 16 18 20 22 LAT Figure 4.14 Diurnal time series of energy storage biases (AQSREF - AQSMOD) for varying surface albedos. Here, 'Town' refers to complete surface area of built elements (roads, wall, roofs). Refer to Table 3.7 for reference values. Table 4.9 Sensitivity analysis to varying surface radiative parameters (albedo and emissivity) for the Marseille CAA site. Average AQS (W m~2) Bias (W m2) Diurnal Daytime Nighttime Diurnal Daytime Nighttime REF 8.7 97.5 -96.2 Roof albedo +0.10 8.0 93.9 -93.4 0.7 3.6 -2.7 Wall albedo +0.10 7.1 94.3 -95.9 1.6 3.2 -0.3 Road albedo +0.10 9.0 96.9 -94.8 -0.3 0.6 -1.4 Town albedo +0.10 7.2 91.0 -92.0 1.5 6.5 -4.2 Roof albedo+0.20 7.8 91.1 -90.7 0.9 6.4 -5.5 Wall albedo +0.20 6.4 92.6 -95.4 2.3 4.9 -0.8 Road albedo +0.20 10.5 98.3 -93.2 -1.8 -0.8 -3.0 Town albedo +0.20 6.2 85.5 -87.5 2.5 12.0 -8.7 Roof emissivity -0.05 8.3 96.4 -95.7 0.4 1.1 -0.5 Wall emissivity -0.05 8.7 97.3 -96.1 0.0 0.2 -0.1 Road emissivity -0..05 7.9 96.4 -96.6 0.8 1.1 0.4 121 Similarly, solar geometry is the controlling mechanism insofar as a higher wall albedo is concerned. The time series plot o f AQs bias associated with higher wal l albedos shows a distinct bimodal signature, with peaks in positive biases (signifying the greatest departure from the reference value and less AQs uptake) occurring at sunrise and then again in the late afternoon. These maxima correspond to times when the vertical wal l surfaces are the most exposed to direct incident solar radiation, so that the effects of surface albedo are realized. Figure 4.14 and Table 4.9 show that the greatest departures from the reference simulation occur with the lightest surfaces (a = +0.20 from reference albedo) and when the integrated town (complete built surface area or roads, walls, and roofs) albedo value is increased. Lower surface emissivity does little to significantly impact the local-scale surface energy balance at this site, with biases < 3 W m" at all times during the day and night (bias plot not shown). Greater sensitivity was found, however, when facet surface thickness was altered. Modifications were chosen so as to maintain consistency with simulations performed by Masson et al. (2002) for the central Mexico City and Vancouver light industrial sites. Table 4.10 and Figure 4.15 show some noteworthy features. For the most part, bias associated with halving the road thickness straddles 0 W m" until the late afternoon-early Table 4.10 Summary of sensitivity analysis to varying TEB surface thickness. Average AQs (W m~2) Bias (W m2) Diurnal Daytime Nighttime Diurnal Daytime Nighttime REF 8.7 97.5 -96.2 Roof thickness x 2 10.0 94.3 -89.6 -1.3 3.2 -6.6 Wall thickness / 2 7.9 91.6 -90.9 0.8 5.9 -5.2 Road thickness / 2 4.5 91.6 -98.6 4.2 5.9 2.4 122 evening, when positive biases are reported. The substantial positive bias in the afternoon, at a time when AQSRE? is still positive, denotes a time when the thinner road surface configuration stores a few tens o f W m"2 more than the reference case. In the evening, when AQSREF is negative, the positive bias until about midnight means that more than the reference energy is released to the atmosphere. Thinner road surfaces, then, are more 25 , 0 2 4 6 8 10 12 14 16 18 20 22 LAT Figure 4.15 Diurnal time series of energy storage biases (AQSMOD - AQs^^) for varying surface (wall, roof, and road) thickness. Refer to Table 3.8 for reference values. thermally responsive later in the day, when the oblique solar path allows for lower-canyon exposure. Thinner roads decrease the integrated AQSIQ* by 2% in the daytime and make up that difference by releasing 2% less available radiant energy at night. Overall, reducing the road thickness by one-half results in 4% less o f available radiant energy being partitioned into storage. 123 Modifying the wall thickness to one-half the reference value results in approximately the same magnitude of bias in the daytime as during the nighttime period. Consequently, there is very little overall diurnal bias. The approximate symmetry of bias about solar noon is a function of TEB ' s simple urban geometry formulation, which treats all possible canyon orientations with the same probability of occurrence. Thinner walls do little to modify the amount of available radiant energy partitioned into storage: 1% less uptake overall, with 2% less daytime uptake, and 4% less nocturnal release. When roofs are altered to be twice as thick as in the reference simulation, slightly more energy (just over 1 W m"2) is partitioned into storage over the diurnal period. Interesting, the magnitude of these results are similar to those of the Thermal Mass Scheme sensitivity analysis presented in Section 3.4.3. When the proportion of available radiant energy taken up by storage (+1% in the daytime, - 1% during the night) is considered, the impacts are less than those simulated in Mexico City (-4% in the daytime) and Vancouver light industrial (+2% in the daytime). This is probably due to the fact that roof surfaces at those sites are constructed of materials with slightly greater heat capacities. The sensitivity analyses discussed in this chapter reveal that the wind speed at this site plays the most significant role in surface-atmosphere sensible heat exchanges. Presumably, this is because the f low regime greatly dictates the degree of surface-atmosphere coupling, which in turn, determines the relative importance of energy conduction/convection processes. AQs partitioning was impacted to a lesser extent by increased built surface albedos and modifications to thermal parameters (represented as \ 124 surface thickness). Similar to the conclusions of Masson et al. (2002), adjustments to surface emissivity was found to generate very little bias between simulations. 125 CHAPTER 5: CONCLUSIONS This thesis presents the results of a field campaign designed to investigate current methods to approximate local-scale energy storage flux {AQs) in cities. Field measurements of meteorological parameters, turbulent energy fluxes, and remotely-sensed surface temperatures taken within a densely built-up urban center (Marseille, France) are used to implement four AQs estimation techniques: the energy balance residual approach, a parameterization scheme (Objective Hysteresis Model), a local-scale numerical model (Town Energy Balance model), and a bulk heat transfer approach (Thermal Mass Scheme). The primary goal of this research is to assess the performance of the three latter schemes against residual estimates so, in effect, the observed energy balance residual estimates are the comparative standard. Additional analyses which look at the impacts of varying wind regimes, urban geometry and surface thermal and radiative parameters on the urban energy storage flux are also conducted, by running numerical simulations using the Town Energy Balance model. 5.1 Summary of Conclusions • Observations of AQs derived from tower-mounted fast response instruments (the energy balance residual approach) are a useful tool to gain understanding of sensible heat partitioning at this site. Marseil le's warm, dry climate and massive urban development dictate that surface sensible heat sharing (turbulent sensible and energy storage flux) plays the dominant role in the energy balance of this site, accounting for approximately 90% of available radiant energy Q* over the diurnal period. As a consequence of a fairly robust meso-scale f low regime, surface-atmosphere energy exchanges generally favored convective over conductive heat transfer processes. 126 The Objective Hysteresis Model of Grimmond et al. (1991), which uses hourly-averaged values of measured net radiation and known surface characteristics (Camuffo-Bernardi coefficients) to parameterize AQs, does not perform as well at the Marseille site as at other urban sites. While giving fair estimates overall, the scheme is not able to capture the observed magnitude or degree of hysteresis between AQs and Q*, at almost any time of the day or night (see Table 5.1). This is probably because the current formulation of the OHM does not explicitly incorporate the influence of wind on zl (2s. Considering the complex flow regime and surface static instability observed at this site, Masson's Town Energy Balance model does a good job of simulating AQs, especially at most times during the night (Table 5.1). TEB has good ability to reproduce surface energetics under these varying synoptic, meso-scale and local flow regimes. In addition, the model accurately simulates the asymmetry in energy conduction to the urban fabric and its convective release to the atmosphere. The Thermal Mass Scheme, while generating results that are comparable to those from the other methods (Table 5.1), is so laborious in its measurement and surface description requirements that it is rendered impractical for use in cities on a routine basis. The scheme is particularly sensitive to large temporal surface temperature changes; so much so, in fact, that a 15-minute averaging period is considered to be too long a sampling frequency. Although not independently examined, surface 127 temperature gradients are thought to impose the largest impacts on integrated AQs estimates with this scheme. Changes to urban surface specifications (material thickness, surface plan area, etc.) do not significantly impact final T M S AQs estimates. Table 5.1 Summary or results and comparative statistics for three methods used to approximate AQsata site in the city center of Marseille, France. Here, the OHM, TEB, and TMS methods are compared to results derived from the residual (OBS) method. AQs MBE RMSE d Diurnal Period OBS* -18 OHM (2D) 26 44 95 0.80 OHM (3D) 64 83 129 0.80 TEB 9 26 78 0.90 TMS -2 16 114 0.76 Daytime OBS* 52 OHM (2D) 84 32 109 0.80 OHM (3D) 175 123 166 0.67 TEB 97 47 101 0.82 TMS 27 -25 136 0.74 Nighttime OBS" -102 OHM (2D) -42 60 77 0.75 OHM (3D) -65 37 60 0.47 TEB -96 1 38 0.74 TMS -37 65 82 0.47 % For the eight-day TEB IOP, average observed AQS values used in analysis are: diurnal period = -17 W m"2; daytime = 51 W m"2; nighttime = -97 W m"2. Numerical sensitivity analyses performed with the Town Energy Balance model reveal the relative degree o f dependency convective sensible heat fluxes have on varying wind speed regime, town geometry, and surface thermal and radiative parameters at the Marseille site. The greatest simulated impacts result from modifying the roof-level flow, which can be viewed as a proxy for the level of surface-atmosphere coupling. Slower-than measured wind speeds generally result in 128 more daytime energy conduction to the surface followed by greater nocturnal release to the atmosphere. Conversely, stronger f low generally results in more convective surface-atmosphere exchange so that less available radiant energy is taken up by the surface. The dependence of the sensible heat terms (AQs and QH) on the intensity of the wind speed regime at this site is shown to be numerically stable, suggesting the possibility of predictive utility. In accordance with analyses performed for central Mexico City and a light industrial site in Vancouver, British Columbia (Masson et al., 2002), modifications to simulated town geometry (canyon aspect ratio and plan area of buildings/roads) have relatively little impact on surface-atmosphere sensible heat sharing. Over a range of H/W, biases are on the order o f ±6 W m" over the entire diurnal period. Greater sensitivity (amounting to tens of W m" for the diurnal period) is found when the plan area of buildings was adjusted. The least amount of relative sensitivity, both in terms of absolute flux magnitude and proportion of radiant energy partitioned into storage, is associated with changes to surface radiative and thermal (here, represented by surface thickness) parameters. Negligible changes occur when surface emissivity values are reduced while larger impacts are seen with surface albedo changes. Among the three considered built surfaces, higher roof albedos generate the greatest bias. Notable impacts to AQs partitioning occurs when built surface albedo is simulated to be 0.20 greater than the reference value and when built surface albedos are simultaneous altered.. The built-129 up nature of the site minimizes the impacts of modifying road surface parameters. Changes to surface thickness modifications with the existing thermal properties had little impact. 5.2 Recommendations for Fur ther Research One of the primary objectives of this work was to examine the relative ability of several methods to estimate the magnitude and temporal variation of AQs. In doing so, this study confirms that estimating the magnitude and behavior of heat storage uptake and release by the urban fabric is not a simple task. For a start, there is no known 'standard' against which estimates can be tested (here the residual estimates were decided to be used as the base). The different schemes presented are limited in their relative ability to approximate this important term in the urban surface energy balance. Unt i l the nature of heat sharing and surface-atmosphere coupling involving AQs is able to be better captured, considerable uncertainty in our ability to state the local scale urban surface energy balance remain. Given that, the results of this study serve to highlight some potential for further work in this area. Because efforts to devise a method of directly measuring energy storage flux in an urban environment are likely to be futile and certainly not within the scope of contemporary measurement methods, focus should instead be centered on other ventures which seek to understand the degree of surface-atmosphere sharing of sensible heat. To that end, observation and modeling efforts which endeavor to approximate the magnitude and behavior o f this important energetic exchange should be conducted. Before many of these understandings can come to fruition, however, the ) intricacies of the urban surface at a variety of scales needs to be better represented. 130 Seeing that an important goal of this work is to account properly for the urban environment in meso- and synoptic-scale general circulation models, a careful description of the urban surface and its interactions with the atmosphere is prerequisite. Undertakings such as the Local-scale Urban Parameterization Scheme (LUMPS; Grimmond and Oke, 2002) seek to address this. There is a place for more inter-scale model coupling (using, for example; GIS platforms, the STAR model, the Objective Hysteresis Model, etc.), which act to, in a sense, 'scale-up' from the micro-scale to the local scale (the extent of which is often used as the lower boundary of meso-scale atmospheric models). \ Furthermore, the development of more comprehensive GIS databases (e.g., Brown et al., 2002), which portray the complete, rather than plan, surface area of cities would aid better surface description. The capabilities of such databases could be expanded to allow for the incorporation of solar geometry. From that, local-scale models, such as the TEB, and parameterization schemes (e.g., LUMPS) can be more appropriately initialized. Observational studies, although allowing for general awareness of urban surface-atmosphere energetic interactions, are often limited in their applicability to other urban sites and/or processes. While this work has supported the local-scale numerical modeling capability of the TEB, further model validation using micro- and local-scale measurements would garner a more thorough understanding of the model's sensitivity. Some possible refinements to the TEB could include: allowing for the inclusion of higher-resolution surface and meteorological input parameters, and enhancing the degree 131 of complexity o f town geometry parameters so that variable canyon geometries are considered. In regards to the further examination of urban surface-atmosphere sensible heat exchanges, there is certainly a place for the development of more sophisticated sensitivity analyses using T E B or perhaps the parameterization scheme developed by Mart i l l i et al. (2002). For example, the model could be run for an idealized, simple urban environment and forced with constructed radiative and meteorological data (for example, cloudless skies and a constant wind speed). Increasing levels of complexity can be added to the scheme so that the relative importance o f surface, meteorological, and radiative parameters at the local scale can be examined. O H M ' s accuracy at various urban locales, as wel l as its role in the L U M P S , suggests that it is worthwhile to modify the scheme to explicitly include the effects o f wind and moisture in its parameterization of AQs. Doing so would inevitably require validation against measured estimates. Given the heterogeneity of urban surface types, further measurement and modeling efforts (i.e. the STAR model) to expand the number of surfaces for which Camuffo-Bernardi coefficients are available for use in the O H M , is encouraged. 132 R E F E R E N C E S Aida, M : 1982: Urban albedo as a function of the urban stracture - A model experiment (Part I). Bound. -Layer Meteor., 23, 405-413. Anandakumar, K., 1999: A study on the partition of net radiation into heat fluxes on a dry asphalt surface. Atmos. Environ., 33, 3911-3918. Arnfield, A . J. and C. S. B. Grimmond, 1998: A n urban canyon energy balance model and its application to urban storage heat flux modelling. Energy and Buildings, 27,61-68. Arnfield, A . J., J. M . Herbert and G. T. Johnson, 1998: A numerical simulation investigation of urban canyon energy budget variations. Preprints Second Symposium on the Urban Environment. Amer. Meteorol. Soc , Boston M A , 2-5. Asaeda, T. and V . T. Ca, 1993: The subsurface transport o f heat and moisture and its effect on the environment: a numerical model. Bound-Layer Meteor., 65, 159-179. Asaeda, T. and V . T. Ca, 2000: Characteristics of permeable pavement during hot summer weather and impact on the thermal environment. Building and Environment, 35, 363-375. A S H R A E , 1993: Handbook of Fundamentals (SI), American Society o f Heating and Airconditioning Engineers, New York, N Y . Berdahl, P. and S. E. Bretz, 1997: Preliminary survey of the solar reflectance of cool roofing materials. Energy and Buildings, 25,149-158. Bornstein, R. D. and K. J. Craig, 2002: Survey of the history of the urbanization of numerical mesoscale models. Preprints Fourth Symposium on the Urban Environment. Amer. Meteorol. Soc , Norfolk, V A . Brown, M . J, S. J. Burian, S. P. Linger, S. P. Velugubantla, and. and C. Ratti, 2002: A n overview of building morphological characteristics derived from 3D building databases. Preprints Fourth Symposium on the Urban Environment. Amer. Meteorol. Soc , Norfolk, V A . Burmeister, L. C , 1993: Convective Heat Transfer, second edition. Toronto: John Wiley and Sons, Inc. Campbell, G. S. and J. M . Norman, 1998: Chapter 8: Heat F low in the Soil, mAn Introduction to Environmental Biophysics, pp. 113-128, Springer Verlag, New York. 133 Camuffo, D. and A . Bernardi: 1982: A n observational study of heat fluxes and the relationship with net radiation. Bound-Layer Meteor., 2 3 , 359-368. Ching, J. K. S., 1985: Urban-scale variations of turbulence parameters and fluxes. Bound-Layer Meteor., 3 3 , 335-361. Ching, J. K. S., J. F. Clarke and J. M . Godowitch, 1983: Modulation of heat flux by different scales of advection in urban environments. Bound-Layer Meteor., 2 5 , 171-192. Cleugh, H. A . and T. R. Oke, 1986: Suburban-rural energy balance comparisons in summer for Vancouver, B. C. Bound-Layer Meteor., 3 6 , 351-369. Cros, B., P. Durrand, E. Prejafon, C. Kottmeier, P. E. Perros, V - H Peuch, J. L. Ponche, D. Robin, F. Said, G. Toupance, and H. Wortham, 2003: The E S C O M P T E program: an overview. Atmos. Research, submitted. Dol l , D., J . K. S. Ching and J. Kaneshiro, 1985: Parameterization of subsurface heating for soil and concrete using net radiation data. Bound-Layer Meteor., 3 2 , 351-372. Ellefsen, R., 1985: Urban terrain zone characteristics. Tech. Monogr., No. 18-87, U.S. Army Engineering Laboratory, Aberdeen Proving Ground, M D , 350 pp. Everest Interscience, 1991: Model 4000A Infrared Temperature Transducer Operating Manual. Everest Interscience: Fullerton, CA . Fairey, P. and S. Kalaghchy, 1982: Evaluation of thermocouple installation and mounting techniques for surface temperature measurement in dynamic environments. Seventh National Passive Solar Conference, J. Hayes and C. B. Winn (eds.). American Solar Energy Society, Inc., 801-805. Foken, Th. and B. Wichura, 1996: Tools for quality assessment of surface-based flux measurements. Ag. and For. Meteor., 78, 83-105. Fox, D. G., 1981: Judging air quality model performance: a summary o f the A M S Workshop on Dispersion Model Performance. Bull. Amer. Meteor. Soc, 6 2 , 599-609. Fuchs, M . and A . Hadas, 1972: The heat flux density in a non-homogeneous bare ldessial soil. Bound-Layer Meteor., 3 , 191-200. Garrett, J. R., 1978: Transfer characteristics for a heterogeneous surface of large aerodynamic roughness. Q. J. R. Meteorol. Soc, 104,199-211 Garratt, J. R., 1980: Surface influences upon vertical profiles in the atmospheric near-surface layer. Q. J. R. Meteorol. Soc, 1 0 6 , 803-820. 134 Gay, L. W. and J. B. Stewart, 1974: Energy balance studies in coniferous forests. Report No. 23, Instit.Hydrol., Natural Environ. Res. Council, Wallingford, Berks. Grimmond, C. S. B., 1988: A n evapotranspiration-interception model for urban areas. Department of Geography, University o f British Columbia, Unpubl. Ph.D. dissertation, 206 pp. Grimmond, C. S. B, 1992: The suburban energy balance: methodological considerations and results for a mid-latitude west coast city under winter and spring conditions. Int.J. Climat. 12,481-497. Grimmond, C. S. B., B. D. Offerle, J. Horn, and D. Golub, 2002a: Observations of local-scale heat, water, momentum and CO2 fluxes at Cub H i l l , Baltimore. Preprints Fourth Symposium on the Urban Environment. Amer. Meteorol. Soc , Norfolk, VA. , 96-97. Grimmond, C. S. B. and C. Souch, 1994: Surface description for urban climate studies: a GIS based methodology. Geocarto Inter., 1, 47-59. Grimmond, C. S. B., H. A . Cleugh, and T. R. Oke, 1991: A n objective heat storage model and its comparison with other schemes. Atmos. Environ., 25B, 311-326. Grimmond, C. S. B., J. Salmond, B. D. Offerle, and T. R. Oke, 2002b: Local-scale flux measurements at a downtown site in Marseille during the E S C O M P T E field campaign. Preprints Fourth Symposium on the Urban Environment. Amer. Meteorol. Soc , Norfolk* VA . , 21-22. Grimmond, C. S. B., J . Salmond, B. D. Offerle, and T. R. Oke, 2003: Local-scale surface flux measurements at a downtown site in Marseille during the E S C O M P T E field campaign. In preparation. Grimmond, C. S. B. and T. R. Oke, 1995: Comparison of heat fluxes from summertime observations in the suburbs of four North American cities. J. Appl. Meteor., 34, 873-889. Grimmond, C. S. B. and T. R. Oke, 1999a: Heat storage in urban areas: local-scale observations and evaluation of a simple model. J. Appl. Meteor., 38, 922-940. Grimmond, C. S. B. and T. R. Oke, 1999b: Aerodynamic properties of urban areas derived from analysis of surface form. J. Appl. Meteor., 38,1262-1292. Grimmond, C. S. B. and T. R. Oke: 2002c: Turbulent heat fluxes in urban areas: Observations and a local-scale urban meteorological parameterization scheme (LUMPS) . J. Appl. Meteor., 41, 792-810. 135 Grimmond, C. S. B., T. R. Oke, R. Sponken-Smith, E. Jauregui, C. Souch, T. Newton, T. S. K ing, J. Voogt, M . Hubble, 1996: Heat storage in urban areas. Prepr. Vol. 12th AMS Conf. on Biometeorology and Aeriobiology, Atlanta, GA, 28 Jan. - 2 Feb., 1996. Grimmond, C. S. B., T. S. K ing, M . Roth and T. R. Oke, 1998: Aerodynamic roughness of urban areas derived from wind observations. Bound-Layer Meteor., 89,1-24. Guenard, V., J. L. Caccia, and G. Tedeschi, 2003: Numerical simulation of a Mistral wind event occurring during E S C O M P T E field experiment. Abstract submission. European Geophysical Union, Nice, 6-11 Apr i l 2003. Hedlin, C. P., 1985: Calculation of thermal conductance based on measurements of heat f low rates in a flat roof using heat flux transducers. Building Applications of Heat Flux Transducers. E. Bales, M . Bomberg and G. E. Courvil le [Ed.]. American Society for Testing and Materials, Philadelphia. 184-202. Holland, J. Z., 1971: Interim report on results from the B O M E X core experiment. BOMEXBull. No,10, N O A A , U.S. Dept. Commerce, 31-43. Johnson, G. T., T. R. Oke, T. J . Lyons, D. G. Steyn, I. D. Watson, and J . A . Voogt, 1991: Simulation o f surface urban heat islands under ' ideal ' conditions at night, Part I: theory and tests against field data. Bound-Layer Meteor., 56, 275-294. Kaimal, J. C. and J . J. Finnigan, 1994: Atmospheric Boundary Layer Flows: Their Structure and Measurement. Oxford University Press, New York, 289 pp. Kerschgens, M . J., 1990: On the energetics o f the urban atmosphere. Proc. Conf. on Urban Meteor. Belgrade, Yugoslavia, 7-11 September 1987. Kerschgens, M . J. and J. M . Hacker, 1985: On the energy budget of the convective boundary layer over an urban and rural environment. Contrib. Atmos. Phys., 58, 171-185. Kerschgens, M . J. and R. L. Drauschke, 1986: On the energy budget o f a wintry mid-latitude city atmosphere. Beitr. Phys. Atmosph., 59, 115-125. Kerschgens, M . J . and H. Kraus, 1990: On the energetics o f the urban canopy layer. Atmos. Environ., 24B, 321-328. Kl jun, N., M . W. Rotach, and H. P. Schmid, 2002: A three-dimensional Lagrangian footprint model for a wide range of boundary-layer stratifications. Bound-Layer Meteor., 103, 205-226. 136 Lafore, J., J. Stein, N . Asencio, P. Bougeault, V . Ducrocq, J. Duron, C. Fischer, P Hereil, P. Mascart, V . Masson, J. Pinty, J. Redelsperger, E. Richard, and J. Vila-Guerau de Arellano, 1998: The Meso-NH atmospheric simulation system. Part I: Adiabatic formulation and control simulation. Ann. Geophys., 16, 90-109. Lemonsu, A., C. S. B. Grimmond and V . Masson, 2003: Model l ing the surface energy balance of an old Mediterranean city core. Bound-Layer Meteor., 103, 205-226. Lemonsu, A . and V . Masson, 2002: Simulation of a summer urban breeze over Paris. Bound-Layer Meteor., 104, 463-490. Lowry, W. P., 1977: Empirical estimation of urban effects on climate: a problem analysis. J. Appl. Meteor., 16* 129-135. Lowry, W. P., 1998: Urban effects on precipitation amount. Progress in Physical Geog., 22, 477-520. McCaughey, J . H., 1985: A radiation and energy balance study of mature forest and clear-cut sites. Bound.-Layer Meteor., 32, 1-24. McCaughey, J. H., 1985: Energy balance storage terms in a mature mixed forest at Petawawa, Ontario - a case study. Bound. -Layer Meteor. ,31,89-101. McNaughton, K and T. A . Black, 1973: A study of evapotranspiration from a Douglas fir forest using the energy balance approach. Water Resources Res. 9, 1579-1590. Mahrt, L, 1998: Flux sampling errors for aircraft and towers. J. Atmos. and Oceanic Tech. 15, 416-429. Mart i l l i , A., A . Clappier and M . W. Rotach, 2002: A n urban surface exchange parameterization for mesoscale models. Bound. -Layer Meteor., 104, 261-304. Masson, V., 2000: A physically-based scheme for the urban energy budget in atmospheric models. Bound-Layer Meteor., 94, 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. J. Appl. Meteor., 41, 1011-1026. Mestayer, P. G. and P. Durand, 2002: The U B L / C L U - E S C O M P T E experiment: description and first results. Preprints Fourth Symposium on the Urban Environment. Amer. Meteorol. Soc , Norfolk, V A , 19-20. Meyn, S. K., 2000: Heat fluxes through roofs and their relevance to estimates of urban heat storage. Department of Earth and Ocean Sciences, University of British Columbia, Unpubl. M . S c Thesis, 106 pp. 137 Mi l ls , G., 1997: A n urban canopy-layer climate model. Theor. Appl. Climatol, 57, 229-244. Monin, A . S. and A . M . Obukhov, 1954: The main features of turbulent mixing in the surface atmospheric layer. Trudy Inst. Geophys. Acad. Sci., 151, 163-187. Monteith, J. L. and M . H. Unsworth, 1990: Principles of Environmental Physics. London: E. Arnold. Moore, C. J. and G. Fisch, 1986: Estimating heat storage in Amazonian tropical forest. Ag. and For. Meteor., 38, 147-169. Narita, K.-L, T. Sekine and T. Tokuoka, 1984: Thermal properties of urban surface materials: Study on heat balance at asphalt pavement. Geographical Review of Japan, 57 (Ser. A) , 639-651. Newton, T., 1999: Energy balance fluxes in a subtropical city. Department of Geography, , University of British Columbia, Unpubl. M.Sc. thesis, 140 pp. Noilhan, J . and S. Planton, 1989: A simple parameterization of land surface processes for meteorological models. Monthly Weather Review., 117, 536-549. Novak, M . D., 1981: The moisture and thermal regimes of a bare soil in the Lower Fraser Val ley during the spring. University of British Columbia, Unpubl. Ph.D. dissertation, 153 pp. Nunez, M . and T. R. Oke, 1976: Long-wave radiative flux divergence and nocturnal cooling of the urban atmosphere, II: Within an urban canyon. Bound-Layer Meteor., 10, 121-135. Nunez, M . and T. R. Oke, 1977: The energy balance of an urban canyon. J. Appl. Meteor., 16, 11-19. Offerle, B., C. S. B. Grimmond and T. R. Oke, 2003: Parameterization of net all-wave radiation for urban areas. J. Appl. Meteor., 42,1157-1173. Offerle, B., C. S. B. Grimmond, T. R. Oke, and T. Newton, 2002: Parameterization of net all-wave radiation for urban areas. Preprints Fourth Symposium on the Urban Environment. Amer. Meteorol. Soc , Norfolk, VA . , 96-97. Oke, T. R., 1982: The energetic basis o f the urban heat island. Q. J. R. Meteor. Soc, 108, 1-24. Oke, T. R . i 1987: Boundary Layer Climates, second edition. London: Routledge. Oke, T. R., 1988: The urban energy balance. Prog. Phys. Geog., 12, 471-508. 138 Oke, T. R., 1991: Bibliography of urban climate 1981-88. WCAP-15 W M O / T D No. 397. Oke, T. R., B. D. Kalanda, and D. G. Steyn, 1981: Parameterization of heat storage in urban areas. Urban Ecol., 5, 45-54. Oke, T. R. and H. A . Cleugh, 1987: Urban heat storage derived as energy budget residuals. Bound.-Layer Meteor., 39, 233-245. Oke, T. R., H. A . Cleugh, S. Grimmond, H. P. Schmid, and M . Roth, 1989: Evaluation of spatially-averaged fluxes of heat, mass, and momentum in the urban boundary layer. Weather and Climate, 9, 14-21. Oke, T. R., R. Spronken-Smith, E. Jauregui, and C. S. B. Grimmond, 1999: The energy balance of central Mexico City during the dry season. Atmos. Environ., 33, 3919-3930. Pasquill, F., 1974: Atmospheric Diffusion, second edition. Wylie, 429 pp. Peikorz, G., 1987: Die energiebilanz einer stadtischen struktur. Dip l . Meteorologie, Rheinische Freidrich-Wilhelm Universitat, Bonn. Pigeon, G;, A . Lemonsu, V . Masson, and P. Durand, 2003: Sea-town interactions over Marseille - Part II: Consequences on atmospheric structure near the surface. Preprints International Conference on Urban Climate - 5. Int. Assn. for Urban Climate, Lodz, Poland. Raupach, M . R., A . S. Thorn, and I. Edwards, 1980: A wind-tunnel study of turbulent f low close to regularly arrayed rough surfaces. Bound-Layer Meteor., 18, 373-397. Reifsnyder, W. E., 1967: Radiation geometry in the measurement and interpretation o f radiation balance. Ag. Meteor., 4, 255-265. Roth, M . and T. R. Oke, 1994: Comparison of modelled and "measured" heat storage in suburban terrain. Contrib. Atmos. Phys., 67,149-156. Sakakibara, Y., 1991: Numerical study of heat storage in a building. Energy and Buildings, .15-16, 577-586. • Saunders, I. R., and W. G. Bailey, 1997: Longwave radiation modeling in mountainous environments. Phys. Geog., 18, 37-52. Schmid, H-P., 1994: Source areas for scalars and scalar fluxes. Bound-Layer Meteor., 67,293-318. Schmid, H-P., 1997: Experimental design for flux measurements: matching scales of observations and fluxes. Ag. and For. Meteor., 87, 179-200. 139 Schmid, H-P., H. A. Cleugh, C. S. B. Grimmond, T. R. Oke, 1991: Spatial variability of energy fluxes in suburban terrain. Bound.-Layer Meteor., 54, 249-276. Schmid, H. P., and T. R. Oke,1988: Estimating the source area of a turbulent flux measurement over a patchy surface, Preprints, 8th Symposium on Turbulence and Diffusion, San Diego, Ca., April 26-29, 1988, Amer. Meteorol. Soc, Boston, Mass, 123-126. Schmid, H-P., and T. R. Oke, 1990: A model to estimate the source area contributing to turbulent exchange in the surface layer over patchy terrain. Q. J. R. Meteorol. &>c, 116, 965-988. Shephard, J. M., H. Pierce and A. J. Negri: 2002: Rainfall modification by major urban areas: Observations from spaceborne rain radar on the TRMM satellite. J. Appl. Meteor., 41, 689-701. Sievers, U. and W. G. Zdunkowski, 1985: A microscale urban climate model. Beitrage zur Physik der Atmosphare., 59, 13-40. Soux, A, 2000: Heat fluxes through roofs and their relevance to estimates of urban heat storage. Department of Geography, University of British Columbia, Unpubl. M.Sc Thesis, 101 pp. Soux, A., J. A. Voogt, and T. R. Oke, 2003: A model to calculate what a remote sensor 'sees' of an urban surface. Bound-Layer Meteor., in press. Steyn, D. G., 1985: An objective method to achieve closure of overdetermined surface energy budgets. Bound-Layer Meteor., 33, 303-310. Stanhill, G., 1965: Observations of the reduction of soil temperatures. Ag. Meteor., 2, 197-203. Steyn, D. G., H-P. Schmid, J. L. Walmsley, and J. D. Wilson, 1997: Spatial variability in surface climates. In The Surface Climates of Canada. Bailey, W. G., T. R. Oke, and W. R. Rouse (eds.), McGill-Queen's University Press, Montreal, 44-67. Stull, R. B., 1988: An Introduction to Boundary Layer Meteorology, Boston: Kluwer Academic Publishers. Stull, R. B., 2000: Meteorology for Scientists and Engineers, second edition. Pacific Grove, CA: Brooks/Cole. Taha, H., 1997: Urban climates and heat islands: albedo, evapotranspiration, and anthropogenic heat. Energy and Buildings, 25, 99-103. 140 Taha, H., 1999: Modify ing a mesoscale meteorological model to better incorporate urban heat storage: a bulk parameterization approach. J. Appl. Meteor., 38, 466-473. Taesler, R., 1978: Observational studies on 3-dimensional temperature and wind fields in Uppsala. Proc. WMO Symp. Bound. Layer Physics Applied to Specific Problems in Air Pollution. W M O No. 510, Geneva, 23-30. Taesler, R., 1980: Studies of the development and thermal structure of the urban boundary layer in Uppsala, Part II: Data, analysis and results. Meteor. Instit, Uppsala University, Uppsala, Sweden. 61 pp. Terjung, W. H. and Collaborators, 1971: The effect of a cyclonic storm on the energy fluxes at the urban interface - a preliminary experiment. Archiv. Meteor. Geophys. Bioklimatol, Ser. 13, 367-416. Terjung, W. H. and P. A . O'Rourke, 1980: Simulating the casual elements of urban heat islands. Bound-Layer Meteor., 19, 93-118. Trewartha, G., 1954: Introduction to Climate, third edition. New York: McGraw-Hi l l Book Company, Inc. UNFPA , 1999: The State of World Population 1999. United Nations Population Fund, United Nations Publications, 76 pp. Vakeva, M , K. Hameri, M . Kulmada, R. Lahdes, J. Ruuskanen, and T. Laitinen, 1999: Street level versus rooftop concentrations of submicron aerosol particles and gaseous pollutants in an urban street canyon. Atmos. Environ., 33, 1985-1997. van Loon, W. K. P., H. M . H. Bastings, and E. J . Moors, 1998: Calibration of soil heat flux sensors. Agr. and For. Meteor., 92, 1-8. Voogt, J . A., 1995: Remote sensing of urban surface temperatures. Department of Geography, University o f British Columbia, Unpubl. Ph.D. dissertation, 340 pp. Voogt, J . A . and T. R. Oke, 1997: Complete urban surface temperatures. J. Appl. Meteor., 36,1117-1132. Voogt, J. A . and T. R. Oke, 2003: Thermal remote sensing of urban climates. Remote Sensing of Environ, in press. Webb, E. K., G. I. Pearman, and R. Leuning, 1980: Correction o f flux measurements for density effects due to heat and water vapor transfer. Q. J. R. Meteorol. Soc, 106, 85-100. 141 1 Wilkes, K. E., 1989: Model for Roof Thermal Performance. Oak Ridge National Laboratory, Oak Ridge, T N . ORNL/CON-274. Bui lding Thermal Envelopes and Materials Program. 78 pp. Wilmott, C. J., 1981: On the validation of models. Phys. Geog., 2, 184-194. Wilmott, C. J., 1982: Some comments on the evaluation of model performance. Bull. Amer. Meteor. Soc., 6 3 , 1309-1313. World Resources, 1996: The Urban Environment. A combined publication of the World Resources Institute, United Nations Environment Programme, United Nations Development Programme, and the World Bank, Oxford University Press, Oxford, 365 pp. Yap, D. H., 1973: Sensible heat fluxes in and near Vancouver, B. C. Department o f Geography, University of British Columbia, Unpubl. Ph.D. dissertation, 177 pp. Yoshida, A., K. Tominaga and S. Watatani, 1991: Field measurements on the energy balance of an urban canyon in the summer season. Energy and Buildings, 1 5 - 1 6 , 417-423. 142 APPENDIX A. INFRARED THERMOMETER CALIBRATION To correct for any off sets from factory calibrations, some of the infrared thermometers used in the Marseille field campaign required calibration. Original factory calibrations for the Everest IRT's were performed for two surface temperatures (25 and 76 °C) at an air temperature of 25 °C. Subsequent calibration tests revealed differences between the Everest IRT temperature readings (rev) and the temperature o f a black body cavity (Tcav). To account for these differences, calibration curves for some o f the Everest IRT's as wel l as the hand-held Minolta IRT's were constructed at the University of British Columbia Soi l Science Laboratory (SSL) using procedures outlined by Voogt (1995). Some instruments were calibrated in the field with a CSI near black-body calibration plate. The regression relations between Tcav (s = 1 0) and Tev (e = 0.98) are presented in Table A . 1, using the equation: Tev = a + bTcav (A.l) Table A.1 IRT calibration results, 2001. Calibrations performed in the laboratory following Voogt (1995) are denoted by SSL while those performed in the field in Marseille are referred to as MRS. Instrument Serial a b Temperature Calibration Number Range (°C) Location Everest 2336-3 2.9241 0.9001 26-44 MRS Everest 2336-4 1.4236 0.951268 15-32 MRS Everest 2336-5 1.978752 0.930055 15-49 MRS Everest 2094-1 1.3621 0.9298 8-50 SSL Everest 2094-6 2.8349 0.8799 8-50 SSL Everest 2094-7 2.6163 0.9259 8-50 SSL Everest 3601-1 1.6788 0.9303 25-48 MRS Minolta 24001778 0.6861 0.99878 8-50 SSL Minolta 24004501 3.4413 1.16204 8-50 SSL Minolta 24004223 -0.2916 1.0384 8-50 SSL 143 APPENDIX B. STAR MODEL OVERVIEW AND INPUT The Simplif ied Transient Analysis of Roofs (STAR) is a numerical model developed by K. E. Wilkes at the Roof Research Center of Oak Ridge National Laboratories (1989). The model's primary purpose is to guide thermal/energy efficiency experiments and to extrapolate experimental results to other sites and conditions. STAR applies to transient one-dimensional conduction in multi-layer roof systems, it is fully coupled to weather conditions, and has been found to accurately capture the diurnal effects of radiation forcing and weather conditions. As discussed in Chapter 1, exact analytic solutions to the continuity equation (1.6) are difficult to obtain, due to the coupling o f external weather conditions, and the differing thermal properties of multi-layered roof systems. A s an alternative, STAR employs a finite difference solution method to solve the heat conduction equation for a roof system: where T is roof temperature, t is time, x is the thickness of the roof, kr is its thermal conductivity, and Cr is its heat capacity. STAR first breaks down the roof layers into a grid of nodes and then calculates the temperature at each node; i.e., equation B. l is integrated over space (JC) and time (t) for each node. Once the temperature o f each node is found, the model is able to calculate hourly values of temperature and heat flux at each interface between the roof materials. (B.l) 144 The inputs to STAR include geometric, radiative, and thermal roof layer properties, along with hourly weather data (outdoor temperature, relative humidity, incident solar radiation, wind speed, and cloud amount). STAR uses external weather conditions to solve the following heat balance equation: Ki(l-a)+eLi-eaTs4+hr(Ta-T^+QE+QG = 0 (B.2) where the terms represent absorbed shortwave radiation, absorbed incident longwave radiation, longwave radiation emitted by the surface, sensible.heat convection between the air and the surface, heat delivered to or removed from the surface by condensation or evaporation, and heat conducted into our out of the roof surface, respectively. For a detailed discussion of the generation and application of each term, refer to Meyn (2000) and Wilkes (1989). 145 Table B.l STAR inputs used in the formulation of OHM a coefficients for clay tile roofs. Variable Unit Value Slope of roof cm/m 30 Length and width of roof m 15,5.5 Layers of roof integer 4 LAYER DATA Layer 1 Layer 2 Layer 3 Layer 4 Variable Unit clay tile paperboard air plywood deck Name of material 3 char CLY PAP AIR PLY Thickness cm/m 1 0.5 2.3 1.5 Number of nodes integer 4 2 9 6 Thermal conductivity W m ' K 1 0.571 0.04 0.025 0.12 Slope of thermal cond w/ T unitless 0 0 0 0 Specific Heat J kg1 K 1 840 1400 1010 1210 Slope ofSp. Heatw/T unitless 0 0 0 0 Density kg m 3 1121 40 1.2 540 Variable Unit Value Time steps per hour of simulation time integer 10 Transient solution technique 0, 0.5, or 1 1 Boundary conditions outside Oorl 1 Solar absorptions, IR emittance .72, .90 Outside convection coefficient Oorl 1 Condensation/evaporation Oorl 0 Boundary conditions inside Oorl 1 Indoor temperature °C 22 Inside convection coefficient 0 or 1 1 146 APPENDIX C. EVEREST INFRARED THERMOMETER SITE INSTALLATIONS The network of Everest infrared thermometers used in the E S C O M P T E field campaign to measure surface temperatures of various facets (roofs, walls, and roads) in different orientations consisted of individual instrument sites grouped into five general sites, the names of which reflect the permission-granting business/organization within the buildings used. Table C . l is modified after Table 2.3 and outlines the five sites, the surfaces sensed at each site, their respective sampling periods and instrument field-of-views. Refer to Table 2.3 for individual instrument source areas. The relative location of each site is seen in the aerial photograph (Figure CI ) . Photographs of each site are also included (Figures C2-C9). 147 Table C.l Description of each site within the Everest infrared thermometer network used in Marseille. Surface Type Description/Orientation Measurement Period (YD) FOV (°) Site Name Roofs Flat gravel roof 186-194 60 CAA Southwest-facing 183 -184, 191 - 192 60 School new clay tile North-facing 190-192 60 CAA new clay tile West-facing 183 - 185, 190- 193 15 CAA old clay tile South-facing 187- 194 60 CAA new clay tile South-facing 173 - 194 15 Croix Rouge old clay tile East- and west-facing old 179-194 30 CAA clay tile (along roof spine) Roads North-south oriented road and sidewalk 176-193 15 Licensing Bldg. North-south oriented 176-193 15 Jurexfi road and sidewalk East-west oriented 172 -194 15 Croix Rouge road and sidewalk Walls East-facing windowless limestone wall 183 -193 60 CAA East-facing limestone 176- 193 60 Jurexfi wall with windows East-facing limestone 183 - 186 15 Jurexfi wall with windows South-facing limestone 176-193 15 Licensing Bldg. wall with windows South-facing limestone 176-193 60 Jurexfi wall with windows North-facing limestone 172 -194 60 Croix Rouge wall with windows West-facing limestone 176- 179, 183 - 193 60 Licensing Bldg. wall with windows West-facing limestone 176-193 15 Jurexfi . wall with windows 148 Figure CI Aerial photograph of central Marseille, showing the relative locations of the five sites at which stationary infrared radiation thermometers monitored surface temperatures. 149 Figure C2 Everest array at the school site. A southwest-facing new tile roof was sensed, as was net radiation with a mini net radiometer. Figure C3a, b The two IRT arrays at the Croix Rouge site, a) shows the surface temperature sampling of a north-facing wall and east-west oriented road and sidewalk, b) shows an Everest sampling a south-facing clay tile roof across the street. Figures C4a, b a) The IRT array installed on the roof of the Licensing Building, south of the CAA site. The tripod contains three infrared thermometers. The top instrument is looking down onto a north-south oriented street. The lower instruments are measuring surface temperatures of south-facing and west-facing walls, b) The canyon viewed from the installation. Figure C5 The tripod array on the CAA roof. The IRT is measuring gravel surface temperature. Additional instruments measured net all-wave radiation net short-wave radiation, air temperature and relative humidity, sub-surface (membrane) temperature and conductive heat flux. 151 Figure C6 A photo taken from the CAA tower platform, showing three IRT's measuring the surface temperatures of an east-facing wall and a north- and south-facing new clay tile roof. Figure C7 An infrared thermometer mounted on a railing at the CAA rooftop site, measuring surface temperature along the spine of a roof with tiles slanted in the east and west directions. 152 Figure C8 The IRT array at the Jurexfi building, one block southeast of the C A A site. The above photo shows three Everest IRT's pointed towards a north-south oriented street, and an east- and west-facing wall. Figure C9a, b Wall surfaces a) south-facing and b) east-facing sensed at the Jurexfi site. Instruments are mounted on the balcony (in shadow in both photos). 153 APPENDIX D. ROOF SURFACE TEMPERATURE PARAMETERIZATION SCHEME Logistical complications encountered during the ESCOMPTE-CLU field campaign hampered surface temperature measurements from being continuously conducted over all four clay tile roof surface orientations. Because such measurements are critical for the implementation of the Thermal Mass Scheme to estimate AQs, a parameterization approach to fill the data void is adopted. Gravel-topped, east-facing, north-facing, and west-facing clay tile surface temperatures are parameterized with respect to a representative south-facing clay tile roof, for which there is a complete data set spanning the entire IOP. Second-order polynomial regression relations are found to produce the best correlation between measured and modeled surface temperatures, accurately validating both the relative magnitude and timing behavior of each surface at 15-minute resolution. The regression relations between the parameterized surface temperature (Tp) and the measured south-facing clay tile roof (Tsouth) are presented in Table D.l using the general form: Tp = a(Twuth)2 + b(Twulh) + c (D.l) Table D.l Equation coefficients and performance statistics of the parameterization scheme to estimate unmeasured roof surface temperatures. Roof Surface a b c r2 RMSE Gravel -0.0086 1.4813 -1.661,1 0.9825 3.06 East-facing clay tile -0.0070 1.0887 0.1970 0.8441 6.97 West-facing clay tile -0.0117 1.2550 0.5969 0.9458 8.50 North-facing clay tile -0.0095 1.3003 -4.8268 0.9709 8.51 154 

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.831.1-0052339/manifest

Comment

Related Items