<%BANNER%>

Record for a UF thesis. Title & abstract won't display until thesis is accessible after 2013-04-30.

Permanent Link: http://ufdc.ufl.edu/UFE0042558/00001

Material Information

Title: Record for a UF thesis. Title & abstract won't display until thesis is accessible after 2013-04-30.
Physical Description: Book
Language: english
Creator: KERTESZ,RUBEN A
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2011

Subjects

Subjects / Keywords: Environmental Engineering Sciences -- Dissertations, Academic -- UF
Genre: Environmental Engineering Sciences thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Statement of Responsibility: by RUBEN A KERTESZ.
Thesis: Thesis (Ph.D.)--University of Florida, 2011.
Local: Adviser: Sansalone, John.
Electronic Access: INACCESSIBLE UNTIL 2013-04-30

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2011
System ID: UFE0042558:00001

Permanent Link: http://ufdc.ufl.edu/UFE0042558/00001

Material Information

Title: Record for a UF thesis. Title & abstract won't display until thesis is accessible after 2013-04-30.
Physical Description: Book
Language: english
Creator: KERTESZ,RUBEN A
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2011

Subjects

Subjects / Keywords: Environmental Engineering Sciences -- Dissertations, Academic -- UF
Genre: Environmental Engineering Sciences thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Statement of Responsibility: by RUBEN A KERTESZ.
Thesis: Thesis (Ph.D.)--University of Florida, 2011.
Local: Adviser: Sansalone, John.
Electronic Access: INACCESSIBLE UNTIL 2013-04-30

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2011
System ID: UFE0042558:00001


This item has the following downloads:


Full Text

PAGE 1

1 ENGINEERING URBAN TRANSPORATION INFRASTRUCTURE TO MITIGATE THERMAL POLLUTION IN STORMWATER RAINFALL RUNOFF USING SOURCE CONTROL METHODS By RUBEN ALEXANDER KERTESZ A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVE RSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2011

PAGE 2

2 2011 Ruben Kertesz

PAGE 3

3 To everybody who has encouraged me and supported my desire to explore our r elationship in the global environment and to God for giving me the chance to share it with others.

PAGE 4

4 ACKNOWLEDGMENTS I thank my family for supporting my move into engineering. I thank Dr. Lindner for bringing me to the University of Florida and I thank Dr Heaney for encouraging me to build my understanding of water conservation and computational techniques. I thank Dr. Sansalone for allowing me to take classes to become a licensed engineer and for encouraging me to pursue thermal pollution. I thank Dr. Huber for his guidance and flexibility. I thank Dr. Bloomquist for his instruction and his enlightening comments I thank John Mocko for giving me access to campus weather data and to Demetris Athienitis for assistance in statistical analysis. I thank the Florida Education Fund for providing financial support I thank my lab mates, my friends and my significant other who have listened to me share my findings.

PAGE 5

5 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ .. 4 LIST OF TABLES ................................ ................................ ................................ ............ 7 LIST OF FIGURES ................................ ................................ ................................ .......... 9 LIST OF ABBREVIATIONS ................................ ................................ ........................... 12 ABSTRACT ................................ ................................ ................................ ................... 13 CHAPTER 1 GLOBAL INTRODUCTION ................................ ................................ ..................... 15 2 HYDROLOGIC TRANSPORT AND FIRST FLUSH OF THERMAL L OAD FROM ASPHALTIC PAVEMENT ................................ ................................ ....................... 17 Background ................................ ................................ ................................ ............. 17 Objectives ................................ ................................ ................................ ............... 19 Methodol ogy ................................ ................................ ................................ ........... 19 Data Collection Methods ................................ ................................ .................. 20 Calculation Methods for Temporal Distribution of Heat Transfer to Runoff During Event ................................ ................................ ................................ 21 Method Components of Heat Balance Models ................................ ................. 22 Radiation ................................ ................................ ................................ .... 22 Heat loss by evap oration ................................ ................................ ............ 24 Sensible heat loss ................................ ................................ ...................... 25 Heat loss by convection ................................ ................................ ............. 25 Substit ution of Runoff Temperature for Pavement Surface Temperature ......... 26 Results and Discussion ................................ ................................ ........................... 26 Heat Transfer to Runoff during an Event ................................ .......................... 26 Impact of hydrologic parameters on heat transfer ................................ ...... 27 Relationship between antecedent pavement temperature and heat transfer ................................ ................................ ................................ ... 28 Impact of event date and start time on heat transfer ................................ .. 29 Heat Balance Model Comparison ................................ ................................ ..... 29 Discussion ................................ ................................ ................................ .............. 31 Summary ................................ ................................ ................................ ................ 33 3 CYCLIC TEMPERATURE PROFILES FOR ASPHALTIC PAVEMENT AS A FUNCTION OF TREE CANOPY SH ADING AND VEHICULAR PARKING FREQUENCY ................................ ................................ ................................ ......... 49 Background ................................ ................................ ................................ ............. 49

PAGE 6

6 Objective ................................ ................................ ................................ ................. 51 Methodology ................................ ................................ ................................ ........... 51 Parking Stall Data Collection Methods ................................ ............................. 52 Simulated Driving Activity Data Collection ................................ ........................ 54 Tree Canopy Shade Data Collection Methods ................................ ................. 55 Results and Discussion ................................ ................................ ........................... 57 Thermal Results of Park ing Stall Shade Treatments ................................ ........ 57 Pavement Temperature Shift Under Simulated Parking Activity ....................... 58 Thermal Trends on Shaded Roadway ................................ .............................. 61 Summary ................................ ................................ ................................ ................ 64 4 MITIGATING URBAN HEAT: TEMPORAL TEMPERATURE PROFILES FOR PAVEMENT MATERIALS ................................ ................................ ....................... 81 Background ................................ ................................ ................................ ............. 81 Objective ................................ ................................ ................................ ................. 83 Methodology ................................ ................................ ................................ ........... 84 D ata Collection Methods ................................ ................................ .................. 84 CFD Model Components of Heat Transfer with Solar Radiation ...................... 86 Simulation Methods for Temporal Distributi on of Heat Transfer Under Solar Radiation ................................ ................................ ................................ ....... 89 Results and discussion ................................ ................................ ........................... 90 Measured Heat Balance on Pavement ................................ ............................. 90 Heat Balance Simulation Model ................................ ................................ ....... 97 Summary ................................ ................................ ................................ ................ 98 5 COMPUTATIONAL MODELING OF OVERLAND FLOW A ND HEAT TRANSFER IN ASPHALTIC PAVEMENTS ................................ .......................... 116 Background ................................ ................................ ................................ ........... 116 Objective ................................ ................................ ................................ ............... 120 Methodology ................................ ................................ ................................ ......... 120 Physical Experiments ................................ ................................ ..................... 121 Modeling Methodology ................................ ................................ ................... 123 Heat Transfer Calculation of Flow Over a Flat Plate ................................ ...... 128 Results and Discussion ................................ ................................ ......................... 130 Summary ................................ ................................ ................................ .............. 135 6 GLOBAL CONCLUSION ................................ ................................ ....................... 146 LIST OF REFERENCES ................................ ................................ ............................. 149 BIOGRAPHICAL SKETCH ................................ ................................ .......................... 159

PAGE 7

7 LIST OF TABLES Table page 2 1 Selected properties of asphalt pavement from various studies .......................... 35 2 2 Storm event da ta for measured rainfall events and Kolomogorov Smirnov test for goodness of fit ................................ ................................ ........................ 36 2 3 Correlations between storm event parameters ................................ .................. 37 2 4 Tabular pavement and subgrade temperature profiles at beginning and end of storm. ................................ ................................ ................................ ............. 38 2 5 Total NHT for various modeling methods compared to measured values. Negative values represent heat gain by pavement. ................................ ............ 38 3 1 Weather conditions during 18 September and 19 September calibration days. 65 3 2 Weather data during parking experiment performed on 4 October, 2010. .......... 65 3 3 Parametric statistics for hysteretic loop equations for 19 October, 2010 experiment. ................................ ................................ ................................ ......... 66 3 4 Parametric statistics for hysteretic loops equations for 28 October, 2010 experiment ................................ ................................ ................................ ......... 66 3 5 Hourly asphalt pavement temperatures across east west transect. ................... 67 3 6 Daily solar radiation, air temperature, wind, and shadow patterns. .................... 68 3 7 Shadow patterns over transect, measured from west curb ................................ 69 3 8 Average annual bene fits of four tree sizes over 40 year period. ......................... 69 4 1 Thermal and physical properties of pavement ................................ .................. 100 4 2 Model parameters for computational simulation ................................ ............... 100 4 3 Properties of air and expanded polystyrene (EPS) ................................ ........... 100 4 4 Median values of pavement heat cycle for all measured days. ........................ 101 4 5 Integration of pavement heat cycle heat for 8 September to 1 0 September. .... 101 5 1 Thermal and physical properties of pavement ................................ .................. 136 5 2 Material parameters used in computational fluid dynamics simulation ............. 137

PAGE 8

8 5 3 Model parameters for computational simulation ................................ ............... 138 5 4 Analysis of error between modeled and measured results. .............................. 139 5 5 Analysis of error between modeled and measured results with implicit body force and specified operating density. ................................ .............................. 139 5 6 Analysis of error between modeled and measured results with 50% evaporation/condensation threshold. ................................ ................................ 140

PAGE 9

9 LIST OF FIGURES Figure page 2 1 Historical monthly distribution of weather dat a for Gainesville, FL and Portland, OR ................................ ................................ ................................ ....... 39 2 2 Lake Alice watershed including subject catchment (~450 m 2 ). ........................... 39 2 3 Plan and cross se ctional view of thermocouples (TC) for catchment pavement system in Lake Alice watershed. ................................ ........................ 40 2 4 Conceptual pavement heat balance model with nominal thermocouple installation depths. ................................ ................................ .............................. 40 2 5 Low flow rate storm event data recorded on June 23, 2008 .............................. 41 2 6 Moderate flow rate storm event data recorded on June 30, 2008 ...................... 42 2 7 Storm event data recorded on Augus t 21, 2008 (Tropical Storm Fay) ............... 43 2 8 Distributions of cumulative heat and cumulative flow f or 12 storms that are similar according to K S tests ................................ ................................ ............. 44 2 9 Modeled storm event data showing only best f it models for A) 14 July 2008 and B) 12 August 2008. ................................ ................................ ...................... 45 2 10 Modeled storm event data showing only best fi t models for A) 21 August 2008 and B) September 10 2008 ................................ ................................ ........ 46 2 11 Residual values for four models. ................................ ................................ ....... 47 2 12 Median temperature at two depths in a 38mm asphalt pavement with a forced wind velocity of 2. 2 m/s over the pavement surface ............................... 48 3 1 Lake Alice watershed including parking lot catchment, transect, and parking spaces investigated herein. ................................ ................................ ................ 70 3 2 Vehicle body and asphalt surface thermocouple installation diagram.. .............. 71 3 3 Vehicular surface temperatures measured in direct sunlight for the A) roof, B) hood and C) trunk during calibration period. ................................ ...................... 72 3 4 Pavement surface temperatures beneath engine (front) and gas tank (rear) of vehicles A and B exposed to direct sunli ght during calibration period. ............... 73 3 5 Comparison of average surface and pavement temp eratures between shaded and unshaded vehicles between the hours of 10:00 and 17:00. ............ 74

PAGE 10

10 3 6 Pavement temperature A) before, B) during, and C) after driving test vehicle to observe effect of warm engine on 4 October, 2010. ................................ ....... 75 3 7 Pavement surface temperature under f requent parking A) on 19 October and B) on 28 October ................................ ................................ ................................ 76 3 8 Pavement surface temperature hysteretic loops on 19 October 2010 beneath front and rear of vehicle. Three cycles are shown. ................................ ............. 77 3 9 Pavement surface temperature hysteretic loops on 28 October 2010 beneath front and rear of ve hicle. Three cycles are shown. ................................ ............. 78 3 10 Graphic analysis of shadow patterns over pavement surface for daytime hours. ................................ ................................ ................................ ................. 79 3 11 Plot of heat transfer to runoff compared to pavement temperature before storm. ................................ ................................ ................................ ................. 80 4 1 Comparison of rainfall pattern frequency by hour from 10 years of hourly rainfal l data collected in two climates in the United States. .............................. 102 4 2 S chematic of simulation geometry. ................................ ................................ .. 103 4 3 Comparison of temperatur es at surface and interior of pavements, 15 September, 2010. ................................ ................................ ............................. 104 4 4 Relative distribution of rainfall event occurrence and total rainfall depth by day hour during the rainy season in Gainesvil le, FL. ................................ ........ 105 4 5 Mean hourly temperature and heat absorption with standard deviation. KJ are per unit area 1m 2 ................................ ................................ ....................... 106 4 6 Relative im pact index (RII) for pavement heat storage reduction in Gainesville, FL (negative is better). ................................ ................................ .. 107 4 7 Comparison of c umulative heat storage in pavement and atmospheric conditions between 8 Sept ember and 11 September, 2010.. ........................... 108 4 8 Comparison of pavement temperature before, during, and after two rain events of differing intensity and time of day. ................................ ..................... 109 4 9 Comparison of thermal heating pattern on two dry days of differing radiation on A) 17 September and B) 10 September ................................ ...................... 110 4 10 Concrete temperature and asphalt t emperature at A) east side of road and B) west side of road ; C) difference between concrete and asphalt at both locations ................................ ................................ ................................ ........... 111

PAGE 11

11 4 11 Modeled pavement temperature for control asphalt and white asp halt pavements on 18 August, 2010. ................................ ................................ ....... 112 4 12 Comparison of modeled pavement temperature results under for current, low, and high thermal conductivity (k) values for reflective asphalt simulation ........ 113 4 13 Measured vs. modeled asphalt temperatures for two days in August, 2010. .... 114 4 14 A comparison of measured and modeled a sphalt and concrete temperatures on 6 September, 2010. ................................ ................................ ..................... 115 5 1 Installation of thermocouples in pavement specimen ................................ ....... 141 5 2 CFD mesh dimensions and statistics. ................................ ............................... 142 5 3 Measured and modeled asphalt specimen temperature and effluent temperature. ................................ ................................ ................................ ..... 143 5 4 Measured a nd modeled concrete specimen temperature and effluent temperature.. ................................ ................................ ................................ .... 144 5 5 Effluent temperature modeled using flat plate method. ................................ .... 145

PAGE 12

12 LIST OF ABBREVIATION S B MP best management p ractice CDF cumulative distribution function CFD computational fluid d ynamics EPS expanded polystyrene EST eastern standard t ime FDA functional data analysis FEA finite element analysis HRIC high resolution interface capturing HSPF hydr ologic simulation program in fortran LID Low Impact Development NHT Net Heat Transfer PIP Peak Insolation Period PISO pressure implicit with splitting operators PRESTO pressure staggering option QUICK quadratic upwind interpolation RHT relative heat t ransf er RMSE root mean squared e rror RPD relative percent difference RPE relative percent e rror TC t hermocouple TMDL total maximum daily l oad TRMPAVE thermal runoff model for pavement TURM thermal urban runoff model

PAGE 13

13 Abstract of Dissertation Presented to the G raduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy ENGINEERING URBAN TRANSPORATION INFRASTRUCTURE TO MITIGATE THERMAL POLLUTION IN STORMWATER RAINFALL RUNOFF USING SOURCE CONTROL METHODS By Ruben A. Kertesz May 2011 Chair: Sansalone Major: Environmental Engineering Sciences Research in the field of thermal pollution in urban areas has traditionally been relegated to studies on the urban heat island effect or glob al climate change. Little research has been performed to test for the effect of pavement temperature on stormwater runoff. The research presented herein focuses on the measurement and simulation of heat transfer to pavement by radiation and of heat transfe r from the pavement to rainfall runoff. Four stu dies are performed to provide an understanding of the mechanisms to limit thermal pollution The first study involves the measurement and simulation of heat transfer to rainfall runoff from a n in situ parkin g lot surface Results from applying a series of published heat balance models indicate that evaporation and long wave radiation are important runoff event based heat transfer mechanisms The second study is designed to determine the effect of shading and vehicular activity on pavement surface temperatur e in a n asphaltic parking lot Results show that pavement temperature does not differ significantly beneath a shaded and an unshaded vehicle, that there is a

PAGE 14

14 demonstrable effect of vehicle operation on pavem ent temperatur e, and that it is most critical to shade pavement during the daily peak insolation period. The third study provides a thermal comparison between the daytime temperatures of three pavement specimens of differing material selection and surface treatments A computational analysis is compared to measured data. CFD model results are not statistically significantly different from measured data for each pavement material. Results indicate that adding a reflective coating to asphalt or utilizing conc rete in lieu of asphalt results in a 20% reduction in pavem ent heat load through the day. Concrete pavement stores up to 55% less heat than asphalt between 12:00 and 19:00. The fourth study investigates the applicability of a computational fluid dynamics simulation to model heat transfer to overland flow from two pavement surfaces with the intent of enhancing knowledge of the rainfall runoff heat transfer relationships for various pavement mix designs Results from 300 seconds of simulation are compared t o measured results. Findings indicate that evaporation may only be critical within the first seconds of runoff. The best CFD result is exhibited by the turbulent concrete simulation with a 50% air/water threshold for evaporation/condensation to occur.

PAGE 15

15 CHAPTER 1 GLOBAL INTRODUCTION The series of investigations herein are developed as an exploration into the contribution of urban rainfall runoff pollution from urban surfaces. Akbari et al. (2003) reported that pavement covers 29% of Houston and 45% of S acramento with 60% and 29% of these areas attributed to parking, respectively. Converting vegetated areas to impervious areas reduces groundwater fed streamflow, compounding thermal impacts (Janke et al. 2009; Ferguson and Suckling 1990; Leith and Whitfiel d 2000; Horner et al. 1994). Much research has already been performed on nutrient, metal, an d hydrocarbon pollution sources. V arious treatment mechanisms have been proposed, some of which are commonly used today. The most commonplace mechanisms involve temporarily or permanently impounding water allowing various physical and chemical processes to remove pollution from receiving waters. H owever, in many parts of the United States, stormwater is still discharged directly to receiving waters, whether they be lakes, streams, the ocean, or, to a lesser extent direct discharge to groundwater. This dissertation focuses on a novel pollutant: h eat. Heat pollution is novel for two reasons. Most importantly, the effects of heat pollution on receiving water biota a re only recently being documented but construction practices have not yet advanced in accordance with these findings. Secondly, heat is a transient property rather than a persistent pollutant. In fact, many of the traditional methods of impoundment that re move persistent pollutants can actually increase exposure to sunlight and therefore heat content of the water. The transient nature of thermal pollution also makes it difficult to determine the magnitude and timing of pollution discharge in urban areas wit hout

PAGE 16

16 having intimate knowledge of the contributing source areas as well as surface and subsurface flow routing connectivity. Many low impact development methods have been proposed to minimize the energy and land area required for traditional treatment, su ch as the use of bioretention areas, subsurface exfiltration basins, both of which are often coupled with filter media using porous building materials, or simply disconnecting source areas from conduit networks. By focusing on the source area, stormwater pollution, and particularly heat pollution can be controlled systematically and successfully mitigated. It is even possible to additionally treat more well understood pollutants while controlling for thermal pollution. It is within the context of Low Impac t Development (LID) that the following chapters are written. The testing sites are located in North Central Florida As a heat conductive interface, impervious asphalt pavement serves as a thermal reservoir for climates with diverse conditions such as annu unique from Wisconsin (Roa Espinosa et al. 2003), Ontario, CA (Van Buren et al. 2000; James and Verspagen 1995), or Oregon (Haq and James 2002); locations of previous thermal runoff studies. T he warmest months; illustrating an inverted pattern to that of Oregon. Florida storms typically occur during the mid afternoon when pavement temperature is hottest but rainwater is at dew point te mperature. Hence, the studies benefit by a high signal to noise ratio due to the very high pavement temperatures t hat are reached in the sunlight

PAGE 17

17 CHAPTER 2 HYDROLOGIC TRANSPORT AND FIRST FLUSH OF THERMAL LOAD FROM ASPHALTIC PAVEMENT Background Since th e Industrial Revolution, thermal loads from urban environs have increased (Sansalone 2002). Recently, impacts of imperviousness on thermal load and causal mechanisms have been identified (Oke 1982; Mestayer and Anquetin 1994; Langford 199 0). Akbari et al. (2003) reported that pavement covers 29% of Houston and 45% of Sacramento with 60% and 29% of these areas attributed to parking, respectively. Converting vegetated areas to impervious areas reduces groundwater fed streamflow, compounding thermal impacts ( Janke et al. 2009; Ferguson and Suckling 1990; Leith and Whitfield 2000; Horner et al. 1994). Asphalt can emit 130 W/m 2 of radiation and 200 W/m 2 sensible heat at mid day, significantly above vegetated cover levels (Thanh Ca et al. 1997). Asaeda et al. ( 1996) report ed that asp halt temperatures can exceed 65 C. As a heat conductive interface, impervious asphalt pavement serves as a thermal reservoir even for diverse climates. For example, as shown in Figure 2 1, the n is coincident with the warmest months; an inverted pattern to that of Oregon. Thermal load is a concern due to impacts on water chemistry and ecosystem integrity of receiving waters such as increases in cold water stream temperatures (Langford 1990; Gall i 1990) and fish distress (Coutant 1987; Nakatani 1969; Paul and Meyer, 2001). Urbanization and increased receiving water temperature are related (Langf ord 1990). Galli (1990) reported that a 1% increase in imperviousness is related to a 0.09C increase in cold water stream temperature with local extinction of trout and stoneflies. Trout and salmon stressed by water above 21C will change habitat

PAGE 18

18 (Coutant 1987). From 1979 1999, an increase of 0.83C had a deleterious impact on the Upper Rhone River base d on indicator species (Daufresne et al. 2004). Armour (1991) found increased Escherichia coli. levels due to thermal load. Thermal load can reduce dissolved oxygen needed for fish and plant survival (Nakatani 1969; James and Xie 1998; Paul and Meyer 2001 ) and can lead to increased metal toxicity (Davies 1986). Few studies have measured pavement and runoff temperature during uncontrolled transient event loadings. Studies focus ed on pavement temperature (Minhoto et al. 2005; Asaeda et al. 1996; Yavuzturk e t al., 2005), thermal load of pavement runoff (Krause et al. 2004; Haq and James 2002), and heat fluxes to and from pavement surfaces (Anandakumar 1999; Than Ca et al. 1997; Herb et al. 2008). While steady loadings have the advantage of a controlled load r esponse, the response to uncontrolled transient loadings is also required. However, researchers report ed that study of actual rainfall runoff events can be challenged by spatial, temporal, event frequency and number constraints (Roa Espinosa et al. 2003, Janke et al. 2009, Van Buren 2000). In my study it is hypothesized that the transport of temperature and thermal load by source area pavement runoff has analogs to the transport of constituent concentration and mass, respectively. It has been shown that transport concepts such as the first flush commonly utilized for design, regulation and control can be distilled from many previous studies into either concentration or mass definitions (Sansalone and Cristina 2004). Specifically, with respect to the tra nsport of pollutant load, Sheng et al. (2008) demonstrate d by categorical analysis that the limiting transport classes for

PAGE 19

19 dissolved or particulate matter mass are mass limited (first order mass or heat transport) or flow limited (zero order mass or heat t ransport) Objectives The primary objective of my study is to measure and model the intra event distribution of temperature and transport of thermal load in runoff from an asphaltic pavement source area. The study hypothesizes that (1) thermal load delive ry is controlled by hydrology and can be primarily flow limited; (2) for a rainfall runoff event, the seasonal event date, event duration, antecedent weather parameters, and pavement temperature are correlated with net heat transfer (NHT) to runoff; (3) fo r a rainfall runoff event, the subgrade temperature and intra event weather conditions are correlated with NHT. A second objective is to reproduce measured results utilizing heat balance models. As part of this second objective, the study hypothesizes th at: (1) pavement heat conduction is a surrogate for overall heat transfer to runoff; and (2) that runoff temperature is an appropriate substitute for pavement surface temperature. The study combines measurement and modeling to illustrate the transport and potential of a first flush of thermal load for an asphalt paved source area, illustrating the coupling of hydrology and heat transfer Methodology In my study, an outfall appurtenance located at 29.644098 N, 82.348404 W drains an asphalt paved catchment used for surface parking as shown in Figure 2 2. The catchment is loaded by approximately 708 vehicles per weekday and 84 vehicles per weekend day. The contributing drainage area is approximately 450 to 500 m 2 determined using l i ght d etection a nd r angi ng (LIDAR) data and onsite surveying, and is dependent on rainfall intensity. The hot mix asphalt pavement has a concrete curb and

PAGE 20

20 gutter. Trees surround the catchment, with two dense foliage trees on the west side of the catchment and magnolia trees imm ediately east of the catchment Data Collection Methods Thermal Thermal measurements are made using type T Omega Inc. {5TC PVC} thermocouples (TCs). The catchment primary flow path is ground truthed and a 5.6 m transect of TCs is installed in the path of the sheet flow. Measurements are taken at 0.1 m 1.2 m 2.6 m 4.1 m and 5.3m from the east end (headwater) of the transect, and concrete gutter measurements at 0 m and 5.6 m from the east end of the transect for respectively. Figures 2 3 and 2 4 illustrate the spatial and depth locations of the TCs. Surface temperature is approximated as a function of subsurface pavement temperature as shown in Equation 2 1. ( 2 1) In this equation, is the mean surface temperature ( o C), is the temperature in the pavement at 13mm ( o C), is the temperature at location A5 and depth of 1 mm and is the temperature at location A5 a nd depth of 13 mm Runoff temperature is measured with two TCs placed at the invert of a 150 mm PVC pipe conveying pavement flows at the catchment outfall. Tipping bucket rain data (increments of 0.254 mm ) are collected at 29.642891 N, 82.34864 W. At 29.6 39461 N, 82.345293 W a Texas Weather Instruments WRL 25 records solar radiation, ambient temperature, cloud cover, and wind. An AM25T multiplexer measures TC data and a Campbell Scientific CR 800 logs data. A calibration

PAGE 21

21 curve is generated for each TCs by logging temperature of boiled water as it cools as represented in Equation 2 2. ( 2 2) In this equation, TC is the thermocouple reading (C) and T t is the temperature (C) recorded using an alcohol thermometer. Run off is measured using a 25.4 mm (1 inch) calibrated Parshall flume. Flow depth is measured using a 24 volt ultrasonic sensor and recorded. From the calibration the relationship between flow ( Q ) and depth in the flume is given in Equation 2 3, for Q (L/s) and D depth in the flume (inches). Intra event TC data are logged at five second intervals. ( 2 3) Calculation Methods for Temporal Distribution of Heat Transfer to Runoff During Event NHT from the pavement to the runoff is calculated by the convection e quation (Herb et al. 2008) as shown in Equation 2 4 where q c is the pavement net heat export to runoff (W/m 2 ), is the runoff temperature ( ), is the dewpoint temperature ( ), as a surrogate for rainfall temperature (U.S. Army Corps of Engineers, 1956), is the flow (m 3 /s), is the specific heat of runoff (J/kg K), is the runoff density (kg/m 3 ), and A s is the contributing area (m 2 ). The Kolomogorov Smirnov (K S) test is performed for goodness of fit between cumulative runoff volume and cumulative NHT to the runoff. This test is chos en due to the non normal distribution of intra event flows. ( 2 4) A heat based first flush is defined as an event where there is a disproportionate heat transfer as NHT (analogous to mass ) in relation to runoff volume early in the event. In contrast, a flow limited event is an event in which NHT is proportional to flow; heat

PAGE 22

22 transferred to runoff is linearly proportional to flow volume. A temperature based first flush is defined where the re is a disproportionate increase in runoff temperature (analogous to concentration) in relationship to runoff volume early in the event, followed by a rapid decline in runoff temperature Method Components of Heat Balance Models Simulation using heat bal ance models requires pavement characterization, atmospheric data, and pavement and runoff temperature data during a storm event. The models are validated by comparing intra event modeled results to measured NHT. Heat balance model components are utilized from Janke et al. (2009), Herb et al. (2008), Van Buren et al. (2000), Kim et al. (2008), Thompson et al. (2008), and Sansalone and Teng (2005). Models incorporating these components are compared with a heat budget on rainfall runoff generated from measu red rainfall and runoff temperatures. The governing heat balance equations used in this study are shown in Equation 2 5 for the Van Buren et al. method (2000) and in Equation 2 6 for the other methods. In these equations, q t is the total heat stored in the pavement Thompson et al. (2008) further includes pavement subgrade conduction ( q sub ) as a loss term. All balances are in W/m 2 Table 2 1 presents thermal properties based on published results. ( 2 5) ( 2 6) Radiation Net radiation q rad may be calculated as shown in Equation 2 7 where q r,s is net direct and diffuse solar radiat ion where q r,lw is net longwave radiation (W/m 2 ). Solar radiation is calculated in the same manner for each method, shown in Equation 2 8.

PAGE 23

23 ( 2 7) q r,s = r s ( 1 ) ( 2 8) In Equation 2 8 r s is the total inco ming solar radiation at the surface (W/m 2 albedo. In contrast to solar radiation, methods for net long wave radiation are more variable. Janke et al. (2009) calculates net longwave radiation as summarized in Equations 2 9 and 2 10. ( 2 9) ( 2 10) In these equations is amospheric emissivity, is cloud cover fraction, is surface emissivity, T a,k is air temperature (K), T s,k is surface temperature (K), e s,kPa is saturated vapor pressure (kPa), and is the Stefan Boltzmann constant (J 1 K 4 m 2 sec 1 ). Net longwave radiation from Herb et al. (2008) is summarized in Equa tion 2 11 where e a,Pa is surface vapor pressure (Pa). Equation 2 12 where e a,Hg is surface pressure (mm Hg). ( 2 11) ( 2 12) Equation 2 13 shows the calculation method for Sanalone and Teng (2005) where atmospheric emissivity is calculated as shown in Equation 2 1 4 where is the vapo r pressure at 2 meters (mbar). ( 2 13 ) ( 2 14 )

PAGE 24

24 Heat loss by evaporation Evaporative heat loss summarized in Equations 2 15 2 16, and 2 17. In these equations, r is runoff water density (kg/m 3 ), and D v are the latent heat of vaporization (J/kg) and evaporation rate (m/s), T r is r unoff temperature (C), is wind speed (m/s), and RH is relative humidity. Herb et al. (2008) utilizes Equation 2 18. ( 2 15) ( 2 16) ( 2 17) ( 2 18) In Equation 2 18 is the air density (kg/m 3 ), and are published without reference to units, is the diff erence in virtual temperature between the surface and air (C) (Ryan et al. 1974), and q is specific humidity (kg/kg). Virtual temperature is the equivalent dry air temperature if pressure and density equal measured moist air co nditions. Specific humidity is shown in Equation 2 19. ( 2 19) In this expression q x is either the saturated or surface specific humidity, is saturated or surface vapor pressure and p is atmospheric pressure, all of the same units. Kim et al. (2008) report heat loss by evaporation to be a function of wind speed and vapor pressure. The heat loss equation is derived from the form discussed in Edinger (1974) as shown in Equation 2 20. ( 2 20)

PAGE 25

25 Kim et. al. present the following values for wind function coefficients: a 0 = 57; a 2 = 2.85. Thompson et al. (2008) publish a similar expression shown in Equation 2 21. ( 2 21 ) In this equation a o = [7.2 to 13.6], a 1 = [3.1 to 4.9], a 2 = [0.0 to 0.66], and e s,Hg is in mm Hg. An alternative method (Sansalone and Teng 2005) is based on Penman Monteith (Monteith 1980) Sensible heat loss Sensible heat loss is explicitly added to the heat balance by Van Buren et al., Herb et al., and Kim et al. Van Buren et al. calculate sensible heat as a function of evaporation by multiplying by the Bowen ratio as shown in Equation 2 22. ( 2 22) In this expression is atmospheric pressure in kPa, and temperature is recorded in C. This ratio is also used to calculate sensible heat los s as a function of q evap using the Sansalone and Teng method (2005). Herb et al. utilize Equation 2 23 to calculate heat transfer by sensible heat. ( 2 23) In this expression is the specific heat of the air (1.005 J/kg K) and T s is surface temperature (C). Kim et al. use a similar method shown in Equation 2 24. ( 2 24) In this expression, c 1 mm Hg/C Heat loss by c onvection Convection is calculated as the remainder of the heat balance equation and does not include heat loss of evaporation or sensible heat; hence it is defined as net heat

PAGE 26

26 transfer (NHT) Results are compared to values calculated explicitly using the rainfall runoff temperature differential method described previously. Figure 2 4 demonstrates pavement reside nce time to better correlate runoff temperature readings with NHT calculated by pavement response. The methodology by which convection is solved for in the heat budget is as shown in Equations 2 25 and 2 26, written to express heat gain by radiation and h eat loss by other terms. ( 2 25) Tpav i+1 = Tpav i + ( )* ( 2 26) Substitution of Runoff Temperature for Pavement Surface Temperature Herb et al. and Janke et al. indicate that turbulence generate a uniform runoff temperature equal to pavement temperature at the start of a given time step. Therefore this study examines if substituting runoff temperature for pavement surface temperature impacts model predicti ons. Results from the substitution of runoff temperature for pavement surface temperature are compared to results from the same events where the models do not substitute runoff temperature for surface temperature. R esults and Discussion Heat Transfer to Ru noff during an Event Table 2 2 summarizes event data while Table 2 3 summarizes correlation coefficients between storm event parameters. There is a positive correlation (r = 0.96) between peak flow and NHT. The correlation with NHT for rainfall is 0.64; f or initial air temperature is 0.14; and for continuous flow duration is 0.24. Table 2 2 illustrates the

PAGE 27

27 positive correlation between peak flow and NHT reflected by the K S test for similarity between cumulative flow and NHT in 12 of 17 events Impact of hydrologic parameters on heat t ransfer Figures 2 5 and 2 6 illustrate relationships between NHT and runoff volume for low and medium flow storms as defined in Table 2 4. K S tests between cumulative runoff volume and cumulative NHT indicate a statisticall y significant difference ( p > = 0.05). While these events illustrate a temperature first flush, with respect to NHT both events are flow limited with respect to thermal load. There is a linear relationship between cumulative NHT and volume. The net flux of heat to runoff continues throughout each event and dilution occurs during peak flows. Instantaneous NHT and instantaneous flow follow similar temporal patterns, suggesting lack of a distinct heat based first flush. In contrast, Figure 2 7 illustrates the only heat limited event (Tropical Storm, TS Fay) in the database, where cumulative heat transfer proceeds cumulative flow. The maximum difference between cumulative runoff and cumulative NHT is 33.2% ( p < = 0.05). All other events are flow limited where heat is not exhausted from the pavement. Of the 17 storms, only five produce a significant difference in trajectories between cumulative flow and NHT as shown in Table 2 2. For the remaining 12 storms, cumulative NHT shows an approximate linear trajectory when plotted against cumulative flow as shown in Figure 2 8. Results indicate that hydrology drives NHT for a given pavement source area. Relative heat transfer (RHT, defined as NHT divided by rainfa ll depth) is conceptually similar (ignoring losses) to an event mean concentration (EMC); in this case, dividing NHT by rainfall depth is similar to dividing constituent load by runoff volume. Results in Figure 2 8 indicate for high intensity events, ther e is a

PAGE 28

28 lower RHT and by proxy a lower unit heat transfer as compared to the short duration, lower flow events. The negative correlation between MPRT and NHT indicates that events with longer pavement residence time have lower NHT from pavement to runoff. Parameters other than hydrologic parameters have the potential to influence NHT and RHT. Correlations for RHT events, tabulated in Table 2 3, initial radiation levels show no correlation with NHT (r = 0.05). However, Figure 2 7 is an example where solar radiation between rainfall bands of TS Fay results in pavement temperature increasing despite moderate wind during the storm. Wind speed before the onset of rainfall is observed to have no correlation with RHT (r = 0.08) but does have a moderate negative correlation with NHT (r = 0.48). In a separate experiment, air flow over the surface of 38 mm thick asphalt at 2.2 m/s resulted in 6% drop in surface temperature but 11% in the pavement interior, after 8 minutes of airflow as shown in Figure 2 12. This suggests that wind does affect surface temperature, however with a corresponding slow rate of interior heat loss, supporting the moderate correlation with NHT measured in situ Results illustrate that antecedent air temperature (immediately before rainfall) exhibits a weak correlation with RHT (r = 0.42) Relationship between a ntecedent pavement temperature and heat t ransfer Antecedent asphalt temperature correlates with RHT (r = 0.74) more strongly than with NHT (r = 0.45) a nd has the greatest correlation of any non hydrologic factor for NHT and RHT. Antecedent subgrade temperature has a weak correlation with NHT (r = 0.25) and RHT (0.28), noting that subgrade is buffered from surface temperature and hydrologic parameters. R esults indicate that initial concrete temperatures are lower

PAGE 29

29 than asphalt and subgrade. As a reflective surface, concrete does not correlate strongly with either NHT or RHT. Impact of event date and start time on heat t ransfer There is a weak correlatio n between event date and heat transfer, as between event date and other initial conditions (air, subgrade, and pavement temperature). Similarity of the intra event phenomena at different seasonal points suggests a lack of seasonal correlation. Event start time has little correlation with NHT (r = 0.16) or RHT (r = 0.04). Results shown in Table 2 4 suggest that shading of locations A1 and A5 confounds any correlation between event date and pavement temperature patterns. This may also cause the difference in East Concrete and West Concrete pavement temperatures shown in Figures 2 5 through 2 7. Heat Balance Model C omparison Ta ble 2 5 summarizes results of cumulative net heat transfer (KJ/m 2 ) measured directly by heat gain in runoff as well as modeled using th e heat transfer components q lw (hereafter modified q lw (hereafter q v (hereafter Thomps on), and Kim q v q r,lw (hereafter modified Thompson). Additionally, all events are modeled with the substitution of runoff temperature for pavement surface temperature. Figures 2 9 and 2 10 summariz e modeling results for four storms where pavement surface temperature is measured. The two closest fitting models are shown. In addition, these figures also summarize the mean differential produced by the two

PAGE 30

30 closest models when runoff temperature ( T ro ) is substituted for pavement surface temperature ( T surf ) in each model; assuming T ro at the discharge location equals T surf The rationale for applying net longwave radiation from Janke et al (2009) in the Herb model is two fold: (1) when applying q lw as c alculated by Herb, the net flux of longwave radiation away from the pavement is lower under clear sky conditions than under cloudy conditions; (2) the Boltzmann constant is reported in non standard units in the Herb model, possibly leading to modeling erro r. The Janke method for calculating q lw is of similar origin to the Herb method and provides results consistent with Sansalone and Teng (2005). The Kim et al. (2008) method for calculation of net longwave radiation has been modified to substitute Sansalo q lw because Kim et al. refer to longwave radiation leaving the water surface but provide no equation for calculation. Results calculated without this term are opposite in sign and 10x the magnitude of the Sansalone and Teng (2005) and the Ja nke et al. (2009) methods as shown in Table 2 5. The Thompson model is very similar to the Kim model but presents a different calculation metho d for evaporative heat transfer. T he same longwave radiation modification made to the Kim model is applied to th e Thompson model. The distribution of residuals in Figure 2 11 illustrate that both the modified Kim and the Sansalone and Teng methods represent measured data (mean normalized residuals closest to 0). For example, the 14 July event is best represented us ing the modified Kim method. This method is also closest to measured total NHT for the 12 August event, followed by the Sansalone and Teng method. The 10 September event is also best predicted using the same methods. In contrast, the modified Herb metho d

PAGE 31

31 over predicts NHT and the Thompson model under predicts NHT for the measured events. During the 12 August event, all models generate a greater magnitude increase in heat transfer during peak flow (5 L/s) than measured values. The 21 August event has a very low instantaneous NHT and all models perform poorly. There is a difference in calculated NHT when substituting runoff temperature for asphalt surface temperature as shown in Figure 2 9 however the difference is relatively small. The maximum differenc es for each of the four events are 16.7, 19.9, 4.1, and 41.8 W/m 2 for the 14 July, the 12 August, the 21 August and the 10 September events, respectively. The mean differences for the same storm events are 0.7, 2.1, 2.3, and 4.52 W/m 2 Discussion Re sults of this thermal pollution study for an asphalt paved source area illustrate a temperature first flush and lack of a heat based first flush. This finding suggests that thermal pollutant transport can be analogous to particulate or solute transport fr om urban source areas (Sheng et al. 2008). Sheng et al. also suggest that there is a need to capture and treat the entire event rather than a first flush or water quality volume (WQV) that is designated a priori. This link between hydrology and pollutant transport is also supported by the correlation between NHT and rainfall runoff flow volume and by the statistical analysis of the same Results demonstrate that pavement temperature exhibits a strong correlation with NHT. For the same ambient conditions, low rainfall depth events can exhibit a more significant temperature increase in runoff than high rainfall depth events for asphalt paved source areas. However, for the same ambient conditions the NHT for a high rainfall depth event will be greater than a low rainfall depth event. In contrast to capturing a first flush or WQV, a more effective management strategy may be to

PAGE 32

32 minimize the storage of heat in the pavement through design and material changes. This strategy also remedies the disproportionate i mpact of thermal pollution on perennial, low volume, or ephemeral systems compared to streams with significant base flow. Radiation is the dominant mechanism by which the pavement warms; hence, although a low correlation is measured between radiation and NHT/RHT, it is particularly useful to minimize radiation that reaches or is absorbed by pavement. For example, the uses of shading and concrete pavement have well known thermal benefits and are passive strategies. The thermal discontinuity between the sub grade (composed largely of sand) and the asphalt is shown clearly in Figure 2 6. The implications of a thermal disconnect are multi fold. It suggests that models do not need to focus on sub pavement heat content; at the same time, it implies that better coupling may be achieved by using engineered pavement and ground media to enhance thermal connectivity between the pavement and the subgrade. There are multiple mechanisms that impact the temperature of receiving waters due to urbanization. The critical component of thermal pollution in urban streams is direct discharge. While there are deviations between the Sansalone and Teng, modified Kim, and modified Herb models, all of the aforementioned models are observed to approximate measured NHT following the same temporal pattern. Results suggest that existing models may benefit by performing more tests under real storm events, validating parameters such as longwave radiation with measured values, and focusing more discretely on evaporation early in the stor m event.

PAGE 33

33 Substitution of runoff temperature for pavement surface temperature provides cumulative NHT values that compare nearly as closely to measured surface temperature as non substituted NHT calculation. However, initial runoff temperature misrepresen ts initial pavement temperature because it is cooler than the asphalt pavement (Figures 2 5 through 2 7). It is important to accurately model initial heat transfer because of the rapid convection and evaporation processes unique to event beginnings Summa ry T hermal load transport in runoff from urban asphalt pavement is measured for 17 events at a Gainesville, FL catchment and results are simulated with a series of published models. Hypothesizing that thermal load delivery is driven by hydrology and is pr imarily flow limited, a K S statistical analysis is performed that demonstrates that for 12 out of 17 storms normalized cumulative runoff is an appropriate surrogate for normalized cumulative NHT. Correlation results between these parameters also support this conclusion. The thermal load transport is predominately flow limited with no first flush in relation to NHT. While pavement temperature is strongly correlated to NHT, results indicate that seasonal event date, event duration, and antecedent weather parameters are not correlated to NHT. Results do not support the hypothesis that pavement heat conduction is an appropriate estimation of heat transfer to and from the pavement based on measured pavement and pavement subgrade temperatures during runoff eve nts. Governing equations for pavement heat balance models described by Herb et al. (2008) and Kim et al. (2009) are applied in this study and evaluated with measured NHT. These models are modified to include heat balance components from Janke et al. (2009 ), Sansalone

PAGE 34

34 and Teng (2005), Thompson et al. (2009) and Van Buren et al. (2009). Results indicate heat transfer is modeled equally well with more than one model but that the heat transfer predicted by each model early in an event requires further refinem ent. Utilization of runoff temperature as a surrogate for asphalt surface temperature has little effect on simulated NHT based on models presented but provides a lower NHT early in the event

PAGE 35

35 Table 2 1 Selected properties of asphalt pavement from vario us studies Study Density (kg/m 3 ) Thermal Conductivity (W/m o C) Specific Heat (J/kg o C) Thermal Diffusivity (m 2 /s) Albedo Emissivity Van Buren et al. ( 2000 ) 2250 (1760) 1.21 (1.3) 921 (837) 5.86x10 7 (8.79x10 7 ) NR NR Janke et al. ( 2009 ) 2100 2400 (13 00 1500) 1.4 1.8 (0.4 1.2) 1120 1370 (900 1400) NR 0 .12 0 .94 Herb et al. ( 2008 ) NR NR NR 4x10 7 (6x10 7 ) 0 .12 0 .94 Kim et al. ( 2008 ) NR NR NR 6.98x10 7 0 .05 NR This Study 1850 1.3 (0.6) 1050 6.69 x10 7 0.12 0.94 Note: Values in parent heses are for pav ement subgrade NR: not reported

PAGE 36

36 Table 2 2 Storm event data for measured rainfall events and Kolomogorov Smirnov (K S) test for goodness of fit between normalized cumulative heat and time and normalized cumulative flow and time Event Date (2008) (MM DD) Start Time of Rainfall (HH:mm) (t o ) Duration (H:mm) Rainfall (mm) Peak Flow (L/s) Initial Air Temperature ( o C) Initial Pavement Temperature ( o C) Initial temperature of soil ( o C) Runoff T max ( o C) Continuous Flow Duration ( H :mm)* Previous Dry Hours Net Heat T ransfer to Runoff (KJ) Relative Heat Transfer (KJ/mm of rainfall) MPRT** (min) D (K S test), P ++ 7 31 10:59 0:42 1.27 .15 30.6 33.6 29.1 32.5 0:04 37 2,035 1,602 4 0.044,1 7 14 22:11 1:19 2.03 .15 27.2 31.2 28.7 27.5 0:28 75 3,785 1,865 6 0.033,1 10 23 14:58 0:51 3.56 1.6++ 25.6 28 24.9 26.5 0:15 340 19,216 5,398 3 0.3, 0.043 (n) 6 22 14:38 2:25 1.78 0.07 31.7 33.2 28.3 31.0 0:06 25 2,248 1,263 5 0.283, ~ 0.0 6 3 15:26 0:55 2.03 0.82 33.9 39.3 29.9 34.2 0:15 600 14,814 7,298 4 0.l22, 0.832 9 20 13:4 4 0:47 3.30 1.01 27.8 36.5 28.4 30.3 0:16 45 15,055 4,562 3 0.0857, 0.99 8 21** 12:34 7:09 54.6 5.94 ++ 26.1 27.2 28.1 27.8 2:47 2 74,700 1,368 2 0.332, ~ 0.0 10 09 14:08 1:41 20.8 9.2 ++ 29.4 31.6 26.7 26.8 0:26 20 131,048 6,300 3 0.40, ~ 0.0 8 12 14:29 1:30 16.8 4.6 27.8 31.3 30.3 28.7 1:10 2 45,771 2,724 5 0.0737, 0.951 6 30 14:42 0:31 5.58 3.17 30.0 38.9 27.4 32 0:13 45 39,277 7,039 4 0.111, 0.994 6 11 13:22 1:54 21.6 11 ++ 29.4 41.7 29.4 33.4 0:30 12 218,622 10,121 0.5 0.351, ~ 0.015 7 15 13:08 1 :40 62.2 13.2 ++ 29.4 35.7 28.6 31.1 0:54 12 170,047 2,734 1 0.180, 0.514 9 10 16:13 0:58 6.10 1.96++ 32.8 37.4 29.8 31.1 0:42 120 38,022 6,233 3 0.204, 0.19 6 10 14:02 1:21 22.6 10.7++ 32.8 42.2 31.5 31.7 1:00 600 195,427 8,647 4.5 0.0405, 1.0 7 29 11 :43 0:43 5.08 3.64++ 31.1 37.8 30.6 32.8 0:25 330 45,930 9,041 5 0.18, 0.51 6 21 11:45 1:10 13.7 3.8 30.0 27.3 28.1 26.4 0:10 61 35,808 2,614 3 0.0465, 1 (n) 6 23 10:35 2:27 7.87 0.52 25.6 28.2 28.1 0.52 1:30 18 26,022 3,306 3 0.0417, 1 Excludes gutt er flow; **MPRT (Median Pavement Residence Time); ++ (n) = normally distributed; D is maximum difference, P is p value for

PAGE 37

37 Table 2 3 Correlations between storm event parameters. Note that correlation coefficie nts for wind velocity and radiation are not shown Event Date Start of Rainfall (t o ) Duration of Event Rainfall Depth Peak Flow Antecedent Air Temperature Antecedent Asphalt Temperature Antecedent Subgrade Temperature Maximum Runoff Temperature ( T max ) Continuous Flow Duration (CFD) Previous Dry Hours (PDH) NHT (KJ) RHT (J/mm runoff) MPRT (minutes) Event Date 1.00 Start of Rainfall (t o ) 0.08 1.00 Duration of Event 0.02 0.17 1.00 Rainfall Depth 0.00 0.23 0.62 1.00 Peak Flow 0.09 0.19 0.18 0.77 1.00 Air T (t o ) 0.41 0.03 0.39 0.20 0.09 1.00 Asphalt T (t o ) 0.37 0.09 0.42 0.09 0.33 0.69 1.00 Subgrade T (t o ) 0.52 0.03 0.10 0.02 0.17 0.58 0.59 1.00 Runoff T max 0.01 0.25 0.22 0.04 0.22 0.58 0.56 0.29 1.00 CFD 0.04 0.19 0.85 0.66 0.27 0.43 0.32 0.11 0.37 1.00 PDH 0.17 0.09 0.10 0.11 0.02 0.43 0.36 0.33 0.25 0.00 1.00 NHT 0.18 0.16 0.15 0.64 0.96 0.14 0.45 0.25 0.19 0.24 0.08 1.00 R HT 0.10 0.04 0.56 0.16 0.36 0.42 0.74 0.28 0.29 0.24 0.49 0.49 1.00 MPRT 0.13 0.44 0.33 0.53 0.50 0.22 0.01 0.28 0.09 0.27 0.23 0.53 0.19 1.00 Note: Units are as defined in the previous table. MPRT = Mean Pavement Residence Time; T max Runof f = Maximum Runoff Temperature

PAGE 38

38 Table 2 4 Tabular pavement and subgrade temperature profiles at beginning and end of storm. Event Date (2008) (MM DD) Initial Pavement Profile Final Pavement Profile Initial Subgrade Profile Final Subgrade Profile Start Time (HH:mm) Runoff Volume (L) Q 50 (L/s) Percent ile (%) 6 10 3>5>1 3>5>1 3>5>1 3>5>1 14:00 8000 1.195 75 100 6 21 3>5>1 3>5>1 3>5>1 3>5>1 11:40 3568 0.391 0 25 7 29 3>5>1 3>5>1 3>5>1 3>5>1 11:42 1406 0.656 75 100 7 31 3>5>1 3>5>1 3>5>1 3>5>1 10:56 66 0 .061 0 25 7 15 3>5>1 3>5>1 3>5>1 3>5>1 13:03 22380 3.451 75 100 7 14 3>5>1 3>5>1 3>1>5 3>1>5 21:25 248 0.005 0 25 8 21 3>5>1 3>1>5 3>5>1 3>5>1 11:05 20409 0.310 50 75 6 23 3>5>1 3>1>5 3>1>5 3>1>5 10:35 1373 0.184 25 50 6 11 3>1>5 3>5>1 3>5>1 3>5>1 13: 11 6678 1.560 75 100 6 22 3>1>5 3>5>1 3>1>5 3>1>5 14:33 29 0.006 0 25 9 20 3>1>5 3>5>1 3>1>5 3>1>5 13:36 502 0.200 25 50 10 23 3>1>5 3>5>1 3>1>5 3>1>5 14:50 916 0.194 25 50 6 3 3>1>5 3>1>5 3>1>5 3>1>5 15:25 293 0.073 0 25 6 30 3>1>5 3>1>5 3>1>5 3>1>5 14:38 1028 0.359 50 75 8 12 3>1>5 3>1>5 3>1>5 3>1>5 14:24 3861 0.216 25 50 10 9 3>1>5 3>1>5 3>1>5 3>1>5 13:56 8467 0.707 75 100 9 10 1>3>5 3>1>5 1>3>5 3>1>5 16:07 1540 0.217 50 75 Note : T hermal profiles are in order from hot to c old. Thermal profile sy mbols are 1=A1, 2=A2, 3= A3 as illustrated in Figure 2 3. The 25, 50, 75 th percentile = 0.184, 0.217, 0.656 L/s, respectively. Flow less than 25% is defined as low flow; less than 75% is moderate flow; greater than or equal to 75% is high flow Table 2 5. Total NHT for various modeling methods compared to measured values. Negative values represent heat gain by pavement. Event Date (Day/Month/2008) 6/10 6/23 7/14 8/12 8/21 9/10 Model Components Heat Transfer to Runoff (KJ) Sansalone and Teng 28 54 34 104 209 101 Modified Herb 63 214 94 134 157 122 Van Buren 377 411 122 69 290 198 Kim 909 2014 684 653 3770 941 Thompson 1025 2021 662 610 3497 895 Modified Kim 77 54 19 99 151 83 Modified Thompson 192 46 4 56 425 37 Measured 258 68 9 83 51 76

PAGE 39

39 Figure 2 1 Historical monthly distribution of weather data for Gainesville, FL from August 1998 to July 2008 (NCDC, 2009) and for Portland OR (Oregon Climate Service, 2010) from January 1998 to December 2008 Figure 2 2 Lake Alice watershed including subject catchment (~450 m 2 )

PAGE 40

40 Figure 2 3 Plan and cross sectional view of thermocouples (TC) for catchment pavement system in Lake Alice watershed. Figure 2 4 Conceptual pavement heat balance model with nominal thermocouple installation depths T dew represents rainfall temperature ( o C); T R.O. is runoff temperature ( o C); q r,lw is net longwave radiation; q r,s is net shortwave radiation; q conv is convective heat transfer; q v is evaporative heat transfe r; q s is sensible heat transfer; T surf is surface temperature ( o C); T 13 is asphalt temperature ( o C) measured at ~13mm depth; T 38 is asphalt temperature ( o C) measured at ~38mm depth; T sub is subgrade temperature ( o C) measured at ~76mm depth; T pav is average pavement temperature.

PAGE 41

41 Figure 2 5 Low flow rate storm event data recorded on June 23 2008. Q: Flow; V: Volume; T: Temperature O C; Heat n : n orm a lized heat; V n : normalized volume

PAGE 42

42 Figure 2 6 Moderate flow rate storm event data recorded on June 30, 2008. Q: Flow; V: Volume; T: Temperature O C; Heat n : normalized heat; V n : normalized volume

PAGE 43

43 Figure 2 7 Storm event data recorded on August 21, 20 08 (Tropical Storm Fay). Q: Flow; V: Volume; T: Temperature O C; Heat n : normalized heat; V n : normalized volume

PAGE 44

44 Figure 2 8 Distributions of cumulative heat and cumulative flow for 12 storms that are similar according to K S tests of difference between normalized values of the former. The heat response is stronger during small storms and shallow under larger events, with the exception of the 6 10 event

PAGE 45

45 A B Figure 2 9 Modeled storm event data showing only best fit models for A) 14 July 2008, B using a substitution of runoff temperature for pavement surface temperature

PAGE 46

46 A B Figure 2 10 Modeled storm event data showing only best fit models for A) 21 August is the difference between heat transfer modeled using a substitution of runoff temperature for pavement surface temperature

PAGE 47

47 Figure 2 11 Residual values for four models. Kim and Thompson models are corrected to use q r,lw from Sansalone and Teng model. Herb is modified to use q r,lw from Janke model. A mean of 0 with a normal distribution about the mean indicates a close estim ation of total heat transfer with a good fit to the measured NHT. Use of runoff temperature in place of pavement surface temperature for NHT model calculations results in trending similar mean residual values.

PAGE 48

48 Figur e 2 12. Median temperature at two depths in a 38mm asphalt pavement with a forced wind velocity of 2.2 m/s over the pavement surface. There is an 11% reduction in surface temperature and 6% reduction in the interior temperature. 95% confidence interval is shown in light gray for surface measurements and dark gray for interior measurements

PAGE 49

49 CHAPTER 3 CYCLIC TEMPERATURE PROFILES FOR ASPHALTIC PAVEMENT AS A FUNCTION OF TREE CANOPY SHADING AND VEHICULAR PARKING FREQUENCY Background The temperature of urban r unoff is fast becoming a concern in many locations throughout the United States, most of which have sensitive cold water habitats (Langford 1990; Galli 1990) and some of which exhibit fish distress (Coutant 1987; Nakatani 1969; Paul and Meyer 2001). If no t mitigated, runoff temperature can have an impact on the ecology of receiving waters (Daufresne et al. 2004; James and Xie 1998). The clean water act, as amended by the water quality act of 1987, has established total maximum daily loads whereby states m ust identify locations where controls on thermal discharges to waters cannot assure protection of biota in those waters. Thermal TMDLs have been established in states ranging from the Northwest (Oregon DEQ 2008) to the Southeast (Louisiana DEQ 2001). Par king lot surfaces dominate the urban landscape in urban environments, making up more than 29% of paved area in Houston and Sacramento (Akbari et al. 2003) and between 39% and 64% of commercial areas in Olympia, Washington (City of Olympia 1994). Asaeda (1 996), Celestian, and Martin (2004), and Grimmond and Oke (1999) have demonstrated a contribution to the urban heat island effect from parking lots. Urban drainage areas used for parking generate a thermal input into stormwater run off that is comparable w ith roadways with high speed and high intensity traffic (Hanh and Pfeifer 1994). Low impact development best management practices (BMP) mitigate thermal load to receiving waters in addition to meeting other stormwater criteria or ancillary benefits such as metal, nutrient, or volumetric reduction, or even energy production (Golden

PAGE 50

50 2007). One such BMP is to reduce the area dedicated to parking. Most municipalities maintain minimum parking space requirements, such as 2 spaces per single family home, 0.25 sp aces per movie theater seat or 6.8 spaces per 100m 2 of health spa leasable area (Davidson and Dolnick 2002). Some requirements vary wildly between regions or municipalities. A pool hall may vary between 1 space per billiard table in North Ogden, Utah to 4 spaces per table in Platte County, Missouri (Litman 2006). There also is a very complex relationship between available parking and patronage (Shoup 1997) and few definitive numbers are available of typical parking lot patronage (Institue for Traportation Engineers 1987). Wilson (1995) found that peak parking demand is only 56% of total capacity at 10 office buildings in CA. According to the Urban Land Institute, shopping malls only receive 100% parking space patronage for 19 hours/year (Shoup, 1997). L itman (2006) produced a table from data gathered by Gould (2003) that finds an average occupancy of <50% across a wide cross section of land uses with a maximum occupancy of 82.5 %. Thermal pollution mitigation strategies include multi level parking struct ures, alternative pavement materials, treatment or infiltration of runoff, and the implementation of shade structures on parking lots (McPherson, 2001; Noguera,2005; Laverne and Winson Geideman 2003). While McPherson and Muchnick (2005) and Heisler & Gran t (2000) found that that tree shade is partially responsible for reduced pavement fatigue and increased lifetime, few studies have previously compared pavement temperatures beneath vehicles when shaded and unshaded, but vehicles are parked for up to 23 hou rs of the day, as determined by a study of approximately 11,000 persons in Atlanta (Frank et al. 2004) and may serve to lower pavement temperature.

PAGE 51

51 Scott et al. (1999) measure d a 2.1 3.7 C drop in vehicle chassis temperature when parked in shaded parking lot in Sacramento, CA, however they did not document pavement temperature Objective My study first investigates the relative impact of tree canopy shade on pavement temperature beneath parked vehicles; the hypothesis put forth is that there exists a demon strable and statistically significant difference in day time pavement surface temperature beneath a vehicle that is shaded by tree canopy and beneath a vehicle that is not shaded. The second objective is to determine the cumulative impact of parking activ ity on pavement surface temperature in a parking space under varied initial conditions; the hypothesis put forth is that pavement exposed to insolation for 8 hours before treatment will cool when repeatedly parking and removing vehicles over the space whil e pavement that is shaded before the experiment will warm instead. In cases where a parking lot is not filled to capacity, multiple parking spaces may be exposed to direct solar radiation unless another form of shade is provided. The third objective of th e study is to investigate the relative influence of tree shading on roadway temperature at the surface parking facility. The study hypothesizes that the presence of medium to large foliage trees (as defined in McPherson et al. 2005) east and west of a N S road lowers peak pavement temperature and that the thermal disconnect between asphalt and subgrade is visible as a difference in the gradient of temperature response in the two materials Methodology In my study, a student union parking lot on the Univer sity of Florida campus located at 29.644098 N, 82.348404 W is composed of hot mix asphaltic concrete

PAGE 52

52 (density=1850 kg/m 3 conductivity=1.3 W/m o C, specific heat=1050 J/kg o C, and albedo=0.12) and is used for surface parking as shown in Figure 3 1, receiv ing approximately 708 vehicles per weekday and 84 vehicles per weekend day. Two dense foliage trees of canopy diameters > 9.1m (30ft) are located directly west and one Magnolia Grandiflora tree (diameter >6.1m (20ft)) is located directly east of a catchba sin that drains a 450m catchment shown in Figure 3 1. Due to the N S orientation of the parking spaces, most automobiles receive little to no shade from nearby foliage. A parking stall 6m northeast of the catchbasin is shaded by the magnolia and is used for the vehicular shade experiment Parking Stall Data Collection Methods A central component of my investigation is the analysis of pavement temperature beneath vehicles. A vehicle shade experiment is performed to determine the relative impact of tree ca nopy shade on the pavement temperature beneath the vehicle. Temperature data collection methods include point measurements of temperature taken on the exteriors of two vehicles (on the hood, roof, and trunk) and on the pavement beneath the vehicles as sho wn in Figure 3 2, on the parking space centerline, 1.22m (4 ft) interior of the front and rear of the vehicle. Parking space dimensions are measured to be 2.74m wide by 6.1m long (9x20 ft). Type T Omega {5TC TT T 30} thermocouples (TC) are used to measur e vehicle and pavement surface temperatures. TCs are calibrated by heating water in a beaker over 30 minutes until boiling. Water temperature is measured simultaneously using an alcohol thermometer every minute while a datalogger measures water temperatu re via TC s to generate a calibration curve for the TCs. All experimental temperature data are logged at 2 minute intervals using a Campbell Scientific CR 10x logger with AM25T multiplexer. Tests for significant

PAGE 53

53 difference are performed using the Mann Whit ney rank sum test due to the non normal nature of the data. Vehicle models used in the investigation are a 2005 Lexus RX300 (burnished gold metallic), denoted Vehicle A, and a 2001 Toyota Corolla (silverstream opalescent), denoted Vehicle B. Vehicles are not modified from factory condition. Temperature data collected on 18 September and 19 September, 2010 are used to calibrate temperature measurements including the hood, roof, trunk, and front and rear pavement temperatures. The calibration method invol ves placing both vehicles in parking spaces unobstructed from sunlight, with the front end of the vehicle facing south (same direction as in the experimental trials), over a two day weekend period, separated by 10m to prevent interference. Afterwards, the thermocouple readings measured on the warmer vehicle are calibrated to the cooler readings on the other by a coefficient of multiplication, normalizing temperatures recorded at vehicle B to those at vehicle A. The converse method is used to normalize the cooler asphalt temperature measurements (vehicle A) to those measured beneath the other vehicle (vehicle B). Each of the five measurements locations is independently calibrated. Two parking stalls are included in the shade investigation. One stall is par tially shaded from the southwest by the aforementioned magnolia tree, leaving the rear 33% of the parking space exposed to solar radiation. An unshaded stall is located 14 meters directly east of the shaded stall. Vehicle A is parked in the unshaded stall and vehicle B in the shaded stall between 4 September, 2010 and 16 September, 2010. Temperature measurements are made between 10:00 and 17:00. Upon parking the vehicle, the

PAGE 54

54 thermocouples used to measure pavement surface temperature are affixed to the pave ment surface using thermal pa ste Simulated Driving Activity Data Collection Three driving experiments are performed to determine the effect of engine and drivetrain use on the pavement temperature. The first experiment is designed to measure the impact of vehicle operation on parking space surface temperature after being parked and shut off. The second experiment is designed to measure the cumulative impact on pavement temperature from parking, removing, and reparking vehicles over a parking space not e xposed to radiation before the experiment. The third experiment is similar to the second experiment but over a parking space previously under insolation. The first test, performed on 4 October, 2010, simulates typical workweek parking lot driving activit y by parking both vehicles in sunlit spaces until 13:30 then removing and driving a test vehicle (vehicle B) for 30 minutes, measuring the temperature increase of the pavement while the pavement is exposed to sunlight. The vehicle is driven with maximum ai r conditioning for 10 minutes and then the air conditioning is shut off for the remainder of the vehicle operation. A control vehicle (vehicle A) is simultaneously removed from its parking space, driven to a location 6m outside of the experimental area an d then turned off for the duration of the test while the surface temperature of the exposed parking space is measured. The vehicles are then re parked and all temperature measurements are continued for hour. The second driving experi ment is set up by first parking vehicle A in its space in the morning of 19 October, 2010. The vehicle is then temporarily removed to secure the pavement surface measurement TCs at the same locations used in the previous

PAGE 55

55 experiments and then reparked, mar king the beginning of the experiment at 15:43. The experimental procedure involves cycling the vehicles through the space. While vehicle A is parked in the stall and turned off for approximately 10 minutes, vehicle B is driven on university roads. Vehic le A is then started, immediately removed, and left running nearby but outside the experimental area. The space is empty for approximately 4 minutes, after which vehicle B is parked in the space and turned off. Vehicle A is then driven while vehicle B is parked. This cycle continues until the end of the experiment. Air conditioning is used as needed to maintain a comfortable cabin temperature. In order to draw comparisons between the hysteretic cycle of a cool pavement to a warm pavement, the third driv ing experiment is the same as the second experiment but is performed on a parking space that is exposed to sunlight for 8 hours, warming it until the beginning of the experiment at 14:15 on 28 October, 2010 Tree Canopy Shade Data Collection Methods The tr ee canopy shade investigation is performed prior to the other experiments in this publication but serves to address temperature phenomena of pavement that is constantly exposed, a phenomenon that is not observed at the university parking lot but one that m ay be more typical of retail locations described by sources in the introduction of this publication. Roadway transect thermal measurements are made using Omega {5TC PVC 24} type T TCs calibrated in the same manner described in the first experiment. A 5.6 m transect of TCs is installed as show n in in Figure 3 1 which illustrates the horizontal and vertical placement of TCs in the pavement. TCs are installed at various depths by drilling a vertical shaft through the asphalt using a carbide tip 6.3mm (1/4in) bit. After placement, thermocouples are sealed into the pavement using elastic crack filler followed by a coal tar emulsion sealcoat.

PAGE 56

56 Pavement thickness is measured to be 0.0381m (nominally 1.5in). Pavement subgrade is identified as a clean sand backfil l. At 29.639461 N, 82.345293 W, directly west of the catchment, a Texas Weather Instruments WRL 25 is installed to collect solar radiation using its included pyranometer, ambient temperature using both wet and dry bulb thermometers, and wind velocity us ing an anemometer. A Campbell Scientific AM25T thermocouple multiplexer and a CR800 datalogger are used to log TC temperatures. Dry period weather and pavement temperature data are recorded discontinuously from 16 May, 2008 to 6 September, 2008, along th e transect shown in Figure 3 1. Measurements are made and recorded at 5 minute intervals, grouped by daytime hour, and tabulated. Horizontal and vertical temperature profiles are analyzed for trends in time. The aforement ioned pyranometer has dimensions of 305x102x61mm (12x4x2in) and a spectral range of 300 1100nm from 0 W/m 2 up to 1500 W/m 2 radiation. This spectral range allows for the capture of energy associated from the near UV range to part of the near infrared range. Shade coverage is measured by photographing the test site hourly, from 07:00 to 19:00 between 7 June and 9 June, 2008 and retaining photos that most clearly illustrate shading. Photographs are taken with a tripod mounted 7.2 megapixel digital camera from a point due south from the cro wn of the road. All photos are taken from the same vantage point and viewing angle. Visual editing software is then used to quantify the areal coverage of shadows with the included pixel area measurement algorithm which is then normalized to the maximum extents of the asphalt pavement (maximum width equal to the distance between concrete curbs and maximum length equal to the northernmost and southernmost records of pavement shade). Results are then entered

PAGE 57

57 into the spreadsheet and are presented along wit h the hourly pavement temperature and radiation records in the results section Results and Discussion Thermal Results of Parking Stall Shade Treatments Calibration results are shown in Figure 3 3 and Figure 3 4. Table 3 1 shows the weather data during th e calibration period. Thirteen days of experimental results are shown in Figure 3 5. There is > 20C difference between treatment and control peak hood and roof temperatures but not trunk temperatures. Visual observation confirms that the trunk of the s haded vehicle is only partially shaded during the early afternoon and evening. The maximum temperature difference between treatment methods is observed at the vehicle hoods (23.4C), followed by roofs (22.5C) and trunks (14.8C) at 14:00. The difference in pavement surface temperatures is < 0.7C. Results also show that there is minimal difference between pavement temperature measured at the front and rear of the vehicle. Interestingly, sub vehicular pavement temperature continues to climb after 15:00 when the temperature of continuously exposed asphalt drops. The pavement surface most likely continues to heat beneath the vehicle after 15:00 because it is influenced by conduction from nearby exposed pavement and by sensible heat while the exposed aspha lt is strongly controlled by radiation. T he ambient air temperature peak s at 17:00 as shown in Table 3 6, which supports this posit. The non normal data are statistically analyzed using the Mann Whitney rank sum test; results indicate a significant diffe rence between the shaded and control roofs, trunks, and hoods (P < = 0.05), but not the asphalt surface temperatures (P > = 0.05). The results suggest that vehicles parked without shade during the PIP should be

PAGE 58

58 parked in high density such as by using fewer roadways, sizing lots for typical usage rather than peak patr onage, or if possible by double parking vehicles using a valet system. Hot vehicle surfaces would still contribute to a first flush, however, the duration of which is a function of material heat capacity Pavement Temperature Shift Under Simulated Parkin g Activity While the results of the shading experiment show little difference in sub vehicular pavement temperature due to canopy shade, three parking experiments do show significant influence on pavement temperature (p < a = 0.05). Exposure to solar radi ation increases pavement surface temperature while parking a vehicle over the pavement cools pavement surface temperature. The first experiment documents a stronger thermal influence from a warm engine and drivetrain on pavement temperature than the later experiments do, likely because the vehicle is operated for 3 times as long before reparking in the first experiment as it is in the second and third experiments. Results from the first experiment are shown in Figure 3 6 and Table 3 2. The average pavemen t temperature when first exposed to sunlight a 14:14 is 26.2C, increasing to 45.6C for both test and control vehicles at 15:00. The vehicles are reparked at 15:04, two minutes after which the pavement surface temperature drops by 2.2C in the front of t he test vehicle and 9.6C in the rear, suggesting an exponential decay in surface temperature and that heat does not penetrate deeply into the pavement within 30 minutes of insolation. The front and rear pavement temperatures beneath the control vehicle differ by 0.2C after 15:26 but the front of the test vehicle is approximately 5C hotter than the rear of the test vehicle in the same period. In addition, both front and rear temperatures of test vehicle are hotter than beneath the control vehicle at 15: 28, after which the test

PAGE 59

59 vehicle remains under monitor for 10 hours. The front and rear pavement temperatures beneath the test vehicle do not reach within 0.5 degrees of each other (not shown), suggesting a thermal influence from the vehicle engine or dri vetrain. Results suggest that the benefits of shade outweigh the costs of a hot drivetrain on pavement heat balance during sunlight hours. Results from the second vehicle parking experiment are shown in Figure 3 7 A which illustrates the difference betwee n temperatures measured front and rear of the vehicle over three cycles. After 16:32, without direct insolation, removing vehicle B drops the pavement surface temperature by more than 2C in < 4 minutes and when vehicle A is returned to the parking space the temperature increases in both the front and rear of the vehicle for 2 minutes before cooling off. Hence drivetrain temperature produces a measured but transient impact on pavement surface temperature. Figure 3 8 represents the hysteretic heating and cooling cycles observed before 16:32 in the 19 October experiment, recorded on a 2 minute timestep, generated from data shown in Figure 3 7A with fit statistics shown in Table 3 3. In the first cycle, the front pavement temperature rises exponentially to 38.7C (+6.7C) when exposed to solar radiation while the pavement at the rear rises in an near linear manner to 36C (+5C). Pavement temperatures sharply plateau in the second cycle when reaching approximately 40C. The third cycle exhibits a linear t emperature increase and exponential temperature decay in both the front and the rear. While the observed exposure/shade cycle does not result in a systematic increase in peak pavement temperatures, base temperatures do show a consistent increase. The ver y large increase in front pavement temperature during the first heat phase lays credence to the

PAGE 60

60 influence of engine temperature. After radiation diminishes due to shade from nearby vehicles, the influence of drivetrain temperature is strong enough to show a warming effect on pavement surface temperature. The cumulative effect of this phenomenon on stormwater runoff during evening events a parking lot is unknown and only of possible concern in parking lots serving high frequency afternoon and evening traffic The third parking scenario is designed to determine if exposure of a pavement surface to solar radiation before parking and reparking vehicles would dampen the trends observed in the previous experiment. Results are shown in Figure 3 7B Initial pavem ent temperature is 52C, 18C higher than the previous experiment. The pavement surface temperature increases during insolation and cools when covered by a vehicle. All three heating cycles observed between 14:53 and 15:38 exhibit less change in pavement temperature, however, than the previous experiment. As shade begins to cover the site at 15:30, the temperature data exhibit noise for three cycles (possibly due to patchy shade) after which the heat/cool patterns shift at the front of the vehicle in the same manner as the previous experiment. The three cycles shown in Figure 3 9 are generated from data shown in Figure 3 7B recorded on a 0.5 minute timestep to better capture rapid pavement temperature decay; statistics are shown in Table 3 4. In compari son to the previous experiment, there are two distinct components to the cooling cycle during this exp eriment: a rapid decrease for between 1 minutes and 2 minutes, followed by a slower loss rate. The cooling patterns are modeled using a 5 parameter expon ential function. The heating patterns follow power law relationships rather than the linear or exponential increases observed in the previous experiment. Conclusions that can be drawn from the

PAGE 61

61 reparking experiments are that exposure of a shaded pavement to sunlight results in a rapid increase in temperature whose shape may not be known a priori (ranging from a power function to exponential). In comparison to Figure 3 8, Figure 3 9 shows very sharp contrasts between pavement temperature measured at the fr ont and rear of the vehicle. While the patterns at the rear are consistent and repeated, base temperatures are inconsistent at the front of the vehicle but still higher in the front of the vehicle than at the rear, by >1C. The heat already stored in the pavement before the third experiment is a likely cause for the stable base temp erature at the pavement rear. Note that water is observed on the pavement in close proximity to the thermocouple when removing the vehicle after the third cycle, indicating th at condensation may have cooled the pavement. In general, the heat cycles at the front of the vehicle consistently exhibit a slower rate of heat gain (lower slope) than at the rear but the heat cycles also begin at warmer temperatures when compared to the rear. Figure 3 8 exhibits a rise in the base temperature during consecutive cycles, both in the front and rear graphs. Figure 3 9 shows a consistent base temperature at the rear while results in the front are influenced by condensation Thermal Trends o n Shaded Roadway A roadway on the parking lot is instrumented so as to monitor subsurface temperatures and determine the impact of shade on the temperatures measured. Results from measurements made beneath the pavement surface between 16 May 2008 and 6 Se ptember, 2008 are summarized in Table 3 5. Surface temperatures indicate a distinct peak and trough in daily temperatures occurring between 12:00 to 15:00 and 6:00 to 7:00, respectively. Peak temperatures occur at different times for different locations a long the transect. For example, peak temperature at location A1 s

PAGE 62

62 occurs between 14:00 and 15:00 but the peaks for A4 s and A5 s oc cur between 12:00 and 13:00. The subscript s denotes surface, b indicates bottom of pavement, and sub indicates subgrade. The rate of change with respect to time is most strongly positive between 10:00 and 13:00 and most negative between 14:00 and 17:00. Note that the positive rate of change at the surface occurrs sharply at locations A1 s and A2 s as well as A4 s and A5 s but more gradually at A3 s located in the center of the transect and furthest from shade trees. Results at 38 mm of depth also show rapid heating and cooling. Both peak gradient and maximum temperatures are lower in A4 b and A5 b as compared to the surface, and the y occur later than at the surface. A1 b A2 b and A3 b temperatures all peak when their respective surface temperatures peak. In the subgrade, peak gradient occurs 1 hour later than peak temperature, except at A3 b The magnitude of peak temperatures and peak gradients are notably lower in the subgrade than at the surface. Still, the observation that subgrade temperatures do change according to diurnal patterns indicates some thermal connectivity between the pavement and the subgrade while simultaneously highlighting the lack of thermal conductivity between the subgrade and the thermal mass of the ground beneath. Table 3 6 shows atmospheric conditions as a function of time as well peak gradient and peak temperature location vs. time. Shade coverage shift s from A1 and A2 at 11:00 to A3, A4, and A5 at 13:00 as the sun is blocked by east tree in the morning and the west tree in the afternoon. It appears that if a region of the transect is shaded during the peak insolation period (PIP), the thermal peak also occur at this time. Table 3 7 and Figure 3 10 show shadow extension over the more than 5m wide pavement surface as a function of time. Shading from the west provides shadow

PAGE 63

63 coverage during the PIP. A2 and A3 peak in temperature gradient from 11:00 to 12 :00, during the PIP, but A4 and A5 peak long before the time of maximum solar radiation shown in Table 3 5. The exposure of shaded TC locations to radiation after the PIP does not result in a thermal peak, strongly suggesting that it is critical to shade p avement during the peak period of insolation to minimize pavement heat storage. It is recommended that surface parking lots in the North Florida region be oriented with rows in the N S direction with trees planted west of the parked vehicles. This would be most advantageous when combined with angled parking and 1 lane roads such that every car may receive maximum benefit from canopy shade. This also has a positive impact on aesthetics, vehicle surface temperature, and likely on vehicle interior temperature. As illustrated in Kertesz and Sansalone (2011), shown in Figure 3 11, there is a demonstrable difference in the heat transfer potential of a pavement at 40C and a pavement at 35C. In addition to the previously mentioned methods to improve pavement shade alternative methods include but are not limited to parking garages, alternative pavement material, application of reflective pavement coatings, and runoff retention. Parking garages provide vehicle shading and increase the effective water quality of urb an runoff per parking space. Alternative material parking lots such as porous concrete effectively treat runoff by increasing onsite infiltration but may introduce pollutants into the ground if not properly engineered to treat pollutants. Some solutions also provide ulterior benefits. Sansalone et al. (2009) redesigned a University of Florida parking lot to mitigate stormwater pollution while also minimizing phosphorus and nitrogen runoff, demonstrating that LID designs and retrofits are potentially more cost effective than mitigating contaminated roadway runoff.

PAGE 64

64 Given that soft BMPs such as tree canopy shade are context sensitive, it is critical to select plant species that maintain coverage during the hot season and to properly maintain them. Scott et al (1999) found that 41% of the shaded lot trees in their study site are Chinese elm which are defoliating due to drought stress. Kjelgren and Montague (1998) found that two tree species, Green and Norway maple, exhibited reduced transpiration over aspha lt surfaces while flowering pear showed increased transpiration. If installed correctly, the benefits of trees can outweigh their costs as shown in Table 8 (McPherson et al. 2005). However, improper vegetation maintenance may result in increased biogenic material entering the urban drainage network Summary This research illustrates the impact of shade on pavement surface temperature, whether provided by a vehicle, tree canopy, or both. Results from the first experiment reject the hypothesis that tree ca nopy shading of vehicles provides a significant decrease in pavement temperature beneath the vehicle compared to pavement temperature beneath an unshaded vehicle. Results do not reject the hypothesis that pavement temperature increases during repeated rep arking cycles over an initially cool (31 o C) pavement while pavement temperatures decrease under repeated cycles when the pavement is initially warm (43 o C). The investigation does not reject the hypothesis that shade from the west lowers peak pavement temp erature more than shade from the east, on a N S roadway. The hypothesis that temperature gradients can be used to illustrate a thermal disconnect between pavement and subgrade material cannot be rejected, however it is not strongly supported

PAGE 65

65 Table 3 1 Weather conditions during 18 September and 19 September calibration days Date 9/18/2010 9/19/2010 Statistic Range Mean Median Range Mean Median Temperature ( o C) 33.9 22.8 32.8 22.8 Time 14:48 6:00 15:59 5:04 Humidity (%) 51.0 63.3 62.0 56.0 69.1 73.0 Pressure (Pa) 339 101592 101592 339 101592 101592 Rainfall (mm) 0.0 0.0 0.0 0.0 0.0 0.0 Wind speed (m/s) 4.02 0.72 0.0 4.92 0.54 0.0 Solar radiation (W/m 2 ) 650.0 208.4 140.0 660.0 144.2 10.0 Table 3 2 Weather data during parking exper iment performed on 4 October, 2010. Time (HH:mm) Temperature ( o C) Humidity (%) Pressure (Pa) Rainfall (mm) Wind speed (m/s) Solar radiation (W/m 2 ) 13:30 25.0 48 101659 0 3.13 540 13:40 25.6 47 101626 0 1.34 540 13:50 25.6 42 101626 0 0.89 540 14:00 25 .6 42 101626 0 2.68 530 14:10 26.1 42 101626 0 0.00 530 14:20 26.1 42 101592 0 2.24 530 14:30 26.1 39 101592 0 0.00 520 14:40 26.1 39 101592 0 0.00 510 14:50 26.1 40 101592 0 0.89 500 15:00 26.1 39 101592 0 1.79 490 15:10 26.1 38 101592 0 1.79 480 15:20 25.6 41 101558 0 0.45 470 15:30 26.7 40 101558 0 2.24 450

PAGE 66

66 Table 3 3 Parametric statistics for hysteretic loop equations for 19 October, 2010 experiment Location Time Description Equation Formulation r 2 Standard Error Front 15:48 Rising Limb E quation f= 0.743+0.744*exp(0.377*x) 1 0.004 Falling Limb Equation f=2.951+0.614*exp(0.282*x) 0.939 0.391 16:04 Rising Limb Equation f=3.495*x^0.21 1 0 Falling Limb Equation f= 0.921+1.833*exp(0.287*x) 0.998 0.12 16:18 Rising Limb Equation f=0.008 +1.213*x 1 0.02 Falling Limb Equation f=1.161+1.258*exp(0.265*x) 0.991 0.159 Rear 15:48 Rising Limb Equation f= 5.51+5.37Eexp(0.105*x) 0.979 0.512 Falling Limb Equation f=0.321+1.272*exp(0.196*x) 0.984 0.223 16:04 Rising Limb Equation f=5.706*x^0. 214 1 0 Falling Limb Equation f=0.566+1.162*exp(0.452*x) 0.996 0.22 16:18 Rising Limb Equation f=0.008+1.538*x 1 0.02 Falling Limb Equation f=0.737+2.658*exp(0.176*x) 0.737 0.375 Equations are developed to model temperature difference from the bas e temperature measured at the beginning of the rising limb. The independent axis is net exposure time in minutes. The rising limb moves forward in net exposure time while the falling limb reduces net exposure time. Table 3 4 Parametric statistics for h ysteretic loop equations for 28 October, 2010 experiment Location Time Description Value r 2 Standard Error Front 14:58 Rising Limb f=1.288*x^0.737 0.992 0.104 Falling Limb f= 19.238+20.46*exp(0.012*x) 0.993 0.085 15:12 Rising Limb f=1.241x^0.915 0.9 86 0.159 Falling Limb f= 3.666+4.936*exp(0.025*x)+0.007 *exp(1.832*x) 0.973 0.124 15:24 Rising Limb f=0.969*x^0.749 0.948 0.212 Falling Limb f= 1E8+1.5E8*exp(0)+0.107 *exp(0.731*x) 0.985 0.161 Rear 14:58 Rising Limb f=2.36*x^0.562 0.988 0.188 Falling Limb f= 2.425+3.487*exp(0.071*x) 0.995 0.092 15:12 Rising Limb f=2.579*x^0.457 0.797 0.991 Falling Limb f= 0.753+2.044*exp(0.124*x)+0.001 *exp(2.709*x) 0.793 0.112 15:24 Rising Limb f=2.073*x^0.525 0.952 0.993 Falling Limb f= 1.881+2.858* exp(0.086*x) 0.326 0.081 Equations are developed to model temperature difference from the base temperature measured at the beginning of the rising limb. The independent axis is net exposure time in minutes. The rising limb moves forward in net exposure ti me while the falling limb reduces net exposure time.

PAGE 67

67 Table 3 5 Hourly asphalt pavement temperatures across east west transect Time Temperatures under Pavement Surface (C) Temperatures at Pavement Bottom (C) Subgrade Temperatures (C) Scale (HH) A1 s A2 s A3 s A4 s A5 s A1 b A2 b A3 b A4 b A5 b A1 sub A2 sub A3 sub (C) 00 01 28.2 29.3 29.7 28.7 28.2 28.6 29.3 29.9 30.0 28.7 30.1 31.4 29.5 7.0 01 02 27.6 28.6 29.0 28.1 27.6 27.8 28.6 29.2 29.3 28.2 29.5 30.7 28.9 6.4 02 03 27.2 28.2 28.6 27.7 27.2 27.4 28.1 2 8.8 28.9 27.8 29.1 30.3 28.5 5.9 03 04 26.8 27.8 28.2 27.4 26.8 27.0 27.8 28.4 28.6 27.4 28.7 29.9 28.2 5.3 04 05 26.5 27.5 27.8 27.1 26.5 26.7 27.4 28.1 28.2 27.1 28.4 29.5 27.8 4.7 05 06 26.3 27.2 27.5 26.8 26.2 26.4 27.1 27.8 27.9 26.8 28.1 29.2 27.6 4.2 06 07 26.1 26.9 27.3 26.6 26.0 26.2 26.9 27.6 27.7 26.6 27.8 28.9 27.3 3.6 07 08 26.5 27.5 27.7 27.2 26.2 26.5 27.4 27.8 27.8 26.6 27.7 28.8 27.2 3.0 08 09 28.0 28.9 28.9 28.5 27.2 27.9 28.9 28.9 28.5 27.3 27.9 29.1 27.4 2.5 09 10 29.6 30.1 30.2 3 0.5 29.0 29.4 30.1 30.0 29.6 28.4 28.5 29.6 28.0 1.9 10 11 31.2 32.3 34.0 36.6 34.0 30.8 32.0 33.1 32.8 31.6 29.4 30.8 29.7 1.4 11 12 33.2 38.5 39.3 41.6 38.9 33.0 38.0 38.1 36.8 35.0 30.3 33.4 32.2 0.8 12 13 39.5 44.1 43.7 46.0 43.3 38.2 44.1 42.4 40.6 38.3 32.1 36.7 34.9 0.2 13 14 44.6 46.0 45.4 45.5 41.2 45.4 46.0 44.3 42.1 38.0 35.2 39.1 36.0 0.3 14 15 46.6 47.1 44.1 40.8 38.3 47.4 47.2 43.6 40.0 36.3 37.5 40.1 35.2 0.9 15 16 45.5 43.7 41.0 40.0 37.7 46.0 44.2 40.8 38.6 35.7 38.5 39.1 34.7 1.5 16 17 40.2 40.9 39.6 38.0 36.1 40.2 41.1 39.6 37.7 34.8 37.9 38.5 34.3 2.0 17 18 36.8 38.0 37.6 36.4 35.2 37.1 38.0 37.6 36.6 34.3 36.4 37.6 34.0 2.6 18 19 34.5 35.8 35.7 34.7 33.8 35.0 35.8 35.9 35.3 33.4 35.0 36.4 33.4 3.2 19 20 33.0 34.4 34.5 33. 4 32.6 33.4 34.3 34.7 34.2 32.5 34.2 35.5 32.8 3.7 20 21 31.3 32.6 32.9 31.7 31.1 31.4 32.5 33.1 33.0 31.4 33.0 34.4 31.9 4.3 21 22 29.9 31.3 31.6 30.5 29.9 30.2 31.2 31.8 31.8 30.3 31.8 33.3 31.0 4.9 22 23 29.1 30.4 30.7 29.7 29.1 29.2 30.3 30.9 30. 9 29.6 31.1 32.4 30.3 5.4 23 24 28.5 29.7 30.0 29.0 28.4 28.7 29.6 30.2 30.3 29.0 30.4 31.7 29.7 6.0

PAGE 68

68 Table 3 6 Daily solar radiation, air temperature, wind, and shadow patterns. 10:00 to 14:00 is the peak insolation period. Time ( <= HH) Wind (m/s ) Air Temp (C) Solar Radiation (W/m 2 ) % Shadow Coverage Gradient Peak Location Thermal Peak Location Shaded TCs 01:00 0.44 25.8 0.0 100 All 02:00 0.10 25.0 0.0 100 All 03:00 0.16 24.6 0.0 100 All 04:00 0.08 24.3 0.0 100 All 05:00 0.0 7 24.0 0.0 100 All 06:00 0.04 23.7 0.0 100 All 07:00 0.03 23.5 7.3 20.4 A1 08:00 0.03 23.4 90.9 82.4 A2 A5 09:00 0.11 24.1 214.1 93.7 All TCs 10:00 0.21 25.6 318.1 33.6 None None A1, A2 11:00 0.38 27.3 410.9 10.8 A4, A5 None A1, ( A2) 12:00 0.69 28.6 487.1 0.7 A2, A3 None A1 13:00 0.96 29.8 549.1 15.4 A1 A4, A5 A5, A4, (A3) 14:00 1.24 30.3 540.9 47.6 None A4, A3 A5 A3, (A2) 15:00 1.53 30.6 510.4 76.6 None A1, A2 A5, A3, A2 16:00 1.47 30.8 415.8 71.3 A1, A2 A5 17:00 1.27 31 .3 318.7 90.7 All 18:00 1.57 31.2 246.9 95 All 19:00 1.86 30.7 148.3 69.1 A3 A5 20:00 1.40 29.8 66.0 100 All 21:00 1.17 29.2 9.1 100 All 22:00 0.61 27.6 0.0 100 All 23:00 0.32 26.8 0.0 100 All 24:00 0.19 26.1 0.0 100 All An explanation for the migration of peak temperature from A5 to A1 throughout the PIP is due to shade patterns, as shown in the far right column. Shade patterns reverse from A1 and A2 to A5 A3 from 11:00 to 13:00, at same time as peak gradient shifts from A5 to A1 (column 5). There is a relationship between shading of a TC and time of thermal p eak (two rightmost columns). TC: thermocouple

PAGE 69

69 Table 3 7 Shadow patterns over transect, measured from west curb Time of Day Covered Portion of Asphalt Pavement Transect (m (ft)) 07:00 5.5 6.1 (18 20) 08:00 0.3 0.9 (1 3) 09:00 0.9 6.1 (3 20) 10:00 4.9 6.1 (16 20) 11:00 >5.8 (>19) 12:00 <0.3 (<1) 13:00 0 1.8 (0 6) *{large coverage area 0 11ft directly south of transect} 14:00 0 1.2, 1.8 5.5 (0 4, 6 18) 1 5:00 0.3 1.2, 3.7 6.1 (1 4, 12 20) 16:00 Full coverage 17:00 Spotty full coverage 18:00 Full coverage 19:00 3.7 4.0 (12 13) 20:00 5.2 5.8 (17 19) Distances mentioned are perpendicular to the concrete curb shown in Figure 3 10. Table 3 8 Average annual bene fits of four tree sizes over 40 year period. Tree Size Representative Species Stormwater Retention (gal) [$] Cooling Energy Saving (kWh) [$] Heating Energy Saving (kWh) [$] CO 2 offset (lbs) Increased Property Value ($) Small Cornus florida 1,2 65 [$12.52] 44 [$3.36] 278 [$2.91] 168 [$1.26] $7.29 Medium Magnolia grandiflora 2,566 [$25.40] 53 [$3.99] 298 [$3.12] 128 [$0.96] $13.44 Large Deciduous Acer rubrum 4,778 [$47.30] 89 [$6.74] 415 [$4.34] 340 [$2.55] $41.02 Large Conifer Pinus taeda 3,88 8 [$38.49] 66 [$4.98] 337 [$3.53] 227 [$1.71] $23.08 Note: Values are generated from data presented by McPherson et al. ( 2005)

PAGE 70

70 Figure 3 1. Lake Alice watershed including parking lot catchment, transect, and parking spaces investigated herein.

PAGE 71

71 F igure 3 2 Vehicle body and asphalt surface thermocouple installation diagram. Vehicle length, width, and height (from ground) is 4.58m, 1.816m, and 1.669m for vehicle A and 4.768m, 1.76m, and 1.466m for vehicle B. Pavement temperature is measured 1.2 m in ward from the front and rear for the respective measurements. Parking space dimensions are measured to be 6.1m long and 2.74m wide (9x20 ft).

PAGE 72

72 Figure 3 3. Vehicular surface temperatures measured in direc t sunlight f or A) the roof B) the hood and C) the trunk during calibration period. A B C

PAGE 73

73 Figure 3 4. Pavement surface temperatures beneath engine (front) and gas tank (rear) of vehicles A and B exposed to direct sunlight during calibr ation period. These results are used to calibrate the pavement measurements between vehicles, performing a temperature correction for the front of the vehicles, and a separate correction for the rear.

PAGE 74

74 Figure 3 5. Comparison of average surface (A C) and D) pavement temperatures between shaded and unshaded vehicles betwe en the hours of 10:00 and 17:00 Error bars represent 1 standard deviation from mean A C B D

PAGE 75

7 5 Figure 3 6. Pavement tem perature A) before, B) during, and C) after driving test vehicle to observe effect of warm engine on 4 October, 2010.

PAGE 76

76 Figure 3 7. Pavement surface temperature under frequent parking on A) 19 October and B) 28 Octobe r. The symbol [A] denotes vehicle A (Lexus); [B] denotes vehicle B (Toyota). Up arrows denote removing a vehicle; down arrows denote placing a vehicle in the space. A B

PAGE 77

77 Figure 3 8 Pavement surface temperature hystereti c loops on 19 October 2010 beneath front and rear of vehicle. Three cycles are shown. Parenthetical time is duration since start of experiment ( H :mm). Arrows show the cycle trajectory from 0:00. Experiment is started at 15:48 (H :mm) EST Note the use of th e x axis for net duration of insolation and shade While the experiment has progressed for 44 minutes, there have been 16 minutes more shade than exposure as shown by the x axis at time 0:44. A B

PAGE 78

78 Figure 3 9 Pavement su rface temperature hysteretic loops on 28 October 2010 beneath front and rear of vehicle. Three cycles are shown Experiment commences at 14:58 EST ( H :mm :ss ). Parenthetical time is duration since start of experiment ( H :mm :ss ). Arrows show the cycle trajecto ry from 0:00. The x axis is used for net duration of insolation and shade. A B

PAGE 79

79 A B Figure 3 1 0. Graphic analysis of shadow patterns over pavement s urface A) Purple 07:00 green 08:00 blue 09:00 yellow 10:00 orange 11:00, red 12:00 ; B ) Red 13:00 orange 14:00 yellow 15:00 blue 16:00 purple 17:00 white 18:00 green 19:00; time is in HH:mm.

PAGE 80

80 Figure 3 11 : Plot of heat transfer to runoff compared to pavement temperature before storm. A 2.5mm precipitation event ov er the hotter pavement can release a net (after evaporation, convection, etc.) 12500 KJ more heat to runoff. Polynomial regression y = 49.2x 2 2805x+44479. *Heat load is per unit depth of rainfall = 1mm.

PAGE 81

81 CHAPTER 4 MITIGATING URBAN HEAT: TEMPORAL TEMPERATUR E PROFILES FOR PAVEMENT MATERIALS Background As part of demographic changes across North America the constructed environment continues to undergo expansion and reconstruction. For example, in Florida, 1.54 million hectares of land have been converted into the constructed environment from 1960 to 1997 (Reynolds 2001). Paved parking and roadways are prominent features of the constructed environment and can dominate the urban landscape Parking lots alone represent approximately 30% of the paved area in Hous ton and Sacramento (Akbari et al. 2003). Pavement surface area in urban environments has been documented to cause the urban heat island (UHI) effect (Akbari et al. 2003; Thanh Ca et al. 1997; Pomerantz et al. 2002 ; Chudnovsky et al. 2004). Asphalt pavement is the predominant pavement surface type in the United States. Approximately 94 95% of paved roads in America are asphalt (Takamura 2002; Anderson et al 2009). While m any formulations have been made in asphalt design such as the incorporation of recycled rubber (Choubane et al. 1999) and while asphalt thermal properties make it a heat sink for solar radiation during periods of insolation and a heat source at night and d uring rainfa ll runoff events. A sphalt thermally augment s rainfall runoff during precipitation events (Hanh and Pfeifer 1994) The environmental effect of transient and long term temperature changes in receiving waters have been thoroughly documented (Langford 1990 ; Galli 1990 ; Coutant 1987 ; Nakatani 1969; Paul and Meyer 2001; Daufresne et al. 2004; James and Xie 1998). There are more recent examples of thermal total maximum daily load

PAGE 82

82 (TMDL) values for receiving waters (Oregon DEQ 2008), indicating a regulatory reco gnition of the issue. Thermal augmentation of runoff is a nascent concern but is a growing issue in light of global climate change and TMDLs. The transfer of heat to rainfall runoff depends on the seasonal and daily distribution of rainfall as illustrated in Figure 4 1 for Portland, OR and Gainesville, FL. Whereas a rainfall event is likely to occur in the afternoon during the wet season (summer) in Gainesville, there is an approximately equal probability of a rain event occurring at any hour of the day in Portland. The potential for thermal loadings based on daily distribution of rainfall alone would be greater for an asphalt pavement in Gainesville in the summer as compared to Portland. Pavement h eat gain can be mitigated by changing physical propertie s or surface reflectivity (albedo) P revious studies have examined the relationship between albedo and building temperature (Oleson et al. 2010 ; Akbari and Taha 1992 ; Bretz et al. 1998 ; Synnefa et al. 2006) as well as the effect of reflectiv e coatings on p avement temperature ( Akbari et al. 2001 Levinson and Akbari 2002 ; Kinouchi et al. 2004 ). Leadership in Energy and Environmental Design ( LEED ) credits (Haselbach 2008 ; U.S. Green Building Council 2009 ) are available for coatings under the sustainab ility ( credit 7.1) and the green neighborhood development rating system (credit 9). Santero and Horvath (2009) concluded that changing surface reflectivity can be an effective method to lower the environmental impact of parking lots Their study also suggested that road ways with higher average daily vehicle traffic (ADT) may not heat up as much as those with low er traffic ; hence it may be more beneficial to treat low ADT areas such as parking as compared to highways. Surface reflective treatments such as reflec tive

PAGE 83

83 paints have been utilized for decades in building rooftop applications (Oleson et al. 2010) where there is little abrasion When applied to pavements, surface coating can have the undesirable impact of reducing depression storage or infiltration for p ermeable pavements. The results of m odifying parameters such as pavement reflectivity, heat storage and heat transfer properties can b e examined with physical models but such parameters can also be examined using computational resources under conditions an d variability that can be for more challenging with a physical model. For example, finite e lement m odels (FEM) have been developed by Hermansson (2001) and Gui et al. (2007) for pavement temperature as a function of energy flux. In addition, c omputational f luid d ynamics (CFD) while similar to FEM (Onate and Idelsohn 1992), is well suited for modeling fluid and energy fluxes for dry and wet weather scenarios CFD can be expanded into a 3 D environment with a solar radiation routine and u ser d efine d f uncti ons (UDF) This current study is designed to simulate pavement heat flux. Objective The objective of my study is to physically measure and model the temporal temperature gradients of concrete pavement as well as asphalt pavement materials of differing surf ace treatments subject to wet and dry ambient atmospheric conditions. A primary hypothesis is that changing the pavement material from asphalt to concrete reduces cumulative pavement energy in North Central Florida summer weather conditions during the day time period of maximum potential rainfall. The secondary hypothesis is that changing pavement surface reflectivity mitigates potential storage of solar radiation in the pavement matrix, therefore reducing the heat storage while maintaining land use functio n. Another objective is to measure and compare pavement

PAGE 84

84 responses under rainfall loads if a rainfall event occurs during the investigation. A final objective is development of a CFD model based on physical model data to simulate pavement temperature. Metho dology The methodology consists of two experiments, an uncontrolled physical experiment where pavement specimens are exposed to ambient weather conditions, and an experiment where a computational model is used to approximate measured data. Physical experim ents consist of investigating and comparing pavement surface temperatures to interior temperatures, assessing hourly rainfall frequency and pavement temperature to create a thermal relative impact index (RII), investigating pavement temperature during a st orm event, and comparing pavement response rates under changing weather conditions. The computational model is used to simulate pavement temperature under simulated weather conditions reproduced from measured weather data. Data Collection Methods Physical experiments are performed at an urban environment in Gainesville located at coordinates 29.643006 N, 82.34902 W. There are existing buildings 10 m to the south, 20 m to the east and 100 m to the southeast. Trees are located 7 m to the north and 10m to th e west. The asphalt specimens are taken from an asphalt pavement wearing course that was in service for three years on the Center Drive roadway in Gainesville FL The Portland cement concrete specimen is taken from a concrete slab that has been in service for two years. The properties of the pavement materials are shown in Table 4 1.

PAGE 85

85 Pavement specimen mass and volume are measured to the nearest 10 mg and 10 cm 3 respectively utilizing a Mettler Toledo type electronic balance and volumetric displacement in a 1 0 L rectangular polycarbonate vessel that was volumetrically calibrated to the nearest 10 mL Bulk d ensity is calculated based on measured dry mass and volume. Specific heat capacity is measured using a calibrated low heat capacity, low density expanded po lystyrene calorimeter. Calorimetric tests of the specimens are performed by measuring the temperature ch ange of a known volume of liquid water (initially at ambient temperature) after placing a pavement specimen uniformly heated to 60C in the calorimeter. Conductivity values are estimated from pavement thermal diffusivity as measured by the time to reach equilibrium in the calorimeter, based upon the methods of Army Corps of Engineers ( 1949 ). Due to the uncertainty in measuring this model parameter, a sens itivity analysis is also performed to determine the impact of conductivity on simulated pavement temperature as discussed in the results Of the two instrumented asphalt specimens, one is used as a control ( control ) with no reflective coating, and another specimen ( reflective asphalt ) is painted with two coats of reflective white paint. This reflective paint is applied as an aerosol from a distance of 250 mm for 3 seconds (3 passes, 1 second per pass). This painting method was used to minimize the effect th at multiple or thick layers of paint may have on the heat transfer to the pavement and also to minimize the infill of depression storage. No surface treatments are applied to the Portland cement concrete ( concrete ). A third asphalt sample cut from the same asphalt roadway is instrumented and measured but not modeled. The specimen ( sealed asphalt ) is sealed with a commercially available

PAGE 86

86 petroleum asphalt silica crystalline filler/sealant. The sealant is applied in two coats to visibly seal the pavement surfa ce cavities Calibrated Omega {5TC TT T 30 ST} type T thermocouples (TC) are inserted into 4mm diameter boreholes drilled into the asphalt specimen in duplicate at both 5 and 30 mm depths from the pavement surface. TCs are then encapsulated in a cyanoacryl ate bonding agent (k = 0.2 W/m K) and fully inserted into the drilled boreholes, securing the TC to the pavement and sealing each borehole. Temperature is monitored every 5 minutes using a Campbell Scientific CR 5000 datalogger and AM25T thermocouple multi plexer. Local atmospheric data are monitored through proximate weather stations. Solar radiation is measured only at the weather station located at 29.6395 N, 82.3453 W, north of the urban location where the specimens are monitored, using a 305 mm x 102 mm x 51 mm pyranometer (CdS photocell manufactured by Advanced Phototnix Inc.) wi th a spectral range of 300 1100 nm (up to 1500 W/m 2 radiation), affixed to a Texas Weather Instruments WRL 25 weather station. This spectral range allows for the capture of energ y associated from the near ultraviolet (UV) range (300 400 nm) to part of the near infrared range (750 1400 nm). CFD Model Components of Heat Transfer with Solar Radiation Modeling of environmental phenomena requires validation (Thacker et al. 2004) and in this study CFD simulations are validated with physical model data. In order to simulate heat transfer under simulated weather conditions by the CFD model, an unsteady pressure based solver is used with an absolute velocity formulation and Green Gau ss cell based gradient solution under compressible air flow Fundamental equations used for compressible air flow in this stu dy are those of mass continuity and

PAGE 87

87 momentum. The continuity or conservation of mass equation is as shown in Equation 4 1 (Patankar 1980) ( 4 1) In this equation is the velocity field under laminar conditions (m/s), is the rate of change of density per unit volume, where density (kg/m 3 ) can be related back to mass fraction and temperatur e (K) via an equation of state. S m is the mass (kg) added to the continuous phase from a secondary phase (in this case, S m = 0). Conservation of momentum is written for an inertial ( 0 acceleration) frame of reference (Batchelor 2000 ) as shown in Equation 4 2. ( 4 2) In this equation, p is static pressure (Pa), is the gravitational force (kg/m 2 s 2 ), is an external body force (in this case = 0), and is the stress tensor ( Equation 4 3), where is molecular viscosity, I is the unit tensor, volume dilation is accounted for by the loss term on the right hand side. (4 3) A fundamental equation of energy conservation is used because the employed mode l also simulates conduction, convection and radiation. T he equation for the conservation of energy can be written as shown in Equation 4 4. ( 4 4 ) In this equation k is lam inar conductivity (W/m C), J j is diffusion flux of species j (kg/m 2 s), and h j is enthalpy (J/kg); where T is temperature (K) and S h is the heat of che mical

PAGE 88

88 reaction plus radiation. The first term on the right hand side of the equation is en ergy transfer due to conduction. T he second term is species diffusion T he third term is viscous dissipation. Viscous heating is important when the Brinkman number shown in Equation 4 5 is 1 or greater. (4 5) In this equation, is the fluid dynamic viscosity, is the velocity, k is thermal conductivity, T 0 is the bulk fluid temperature, T is the wall temperature. In this case, the viscosity of the ai r (nominal 1.8E 5 Kg/m s) is too low to necessitate the inclusion of this term and asphalt can be considered to have zero velocity and thus negligible viscous heating in the simulated time duration. In Equation 4 4, E is defined as shown in Equation 4 6. I n the same equation, h is enthalpy ( J/kg) defined for compressible fluids as shown in Equation 4 7. ( 4 6 ) ( 4 7 ) In Equation 4 7 Y j is the mass fraction of species j Specific enthalpy is defined as shown in Equation 4 8. In this equation T ref is 298.15 K. ( 4 8 ) Equations 1 8 can be solved simultaneously for compressible flow when coupled with equations of state (Versteeg and Malalasekara 1995 ; Batchelor, 2000)

PAGE 89

89 Simulation Methods for Temporal Distribution of Heat Transfer Under Solar Radiation The 3 D numerical model requires a model domain and a mesh composition illustrated in Figure 4 2 The cubic enclosure i s designed as a large air domain such that surrounding air is circulated within the domain. A velocity inlet is used to specify wind mag nitude according to measured values using a transient profile while a transient temperature profile is used to specify the measured air temperature data, recorded on one minute intervals An outlet zone boundary is added downstream of the inlet The model incorporates a simplification where wind direction is constant. However, because temperature measurements are numerically measured along a linear transect perpendicular to the direction of wind, bisecting the pavement, this is a reasonable simplification. In addition to the upstream inlet and downstream outlet, the domain enclosure consists of shear free boundaries at the top and side walls, a no lip wall boundary at all other walls to simulate interaction with the pavement surface and insulation chamber Wall temperatures are specified using the transient profile specified for upstream air temperature The pavement top surface participates in solar ray tracing. Simulations are performed using solar ray tracing in CFD where a user defined function (UDF) i s generated to lookup measured radiation (W/m 2 ) from an array using a binary search algorithm ( Knuth 1997) and apply the radiation to participating surfaces at 1 minute increments between 0 7:00 and 19:00. The methods of Michalsky (1988a; 1988b), Iqbal (198 3), and Spencer (1971) are used to track solar inclination. Ray tracing methods are discussed in Cook et al. (1984) and Weghorst et al. (1984). S imulation parameters are shown in Table 4 2 Air and insulating materials are specified in Table

PAGE 90

90 4 3 The pave ment is simulated as a dense liquid of high dynamic viscosity (50000 kg/m s) much higher than a semi solid mixture (Barbe et al. 2000). While the statistical method used to assess significant difference between different measured data is the Mann Whitney rank sum test, the statistical methods used to determine significant difference between measured and modeled results are part of a F unctional D ata A nalysis (FDA). Functional data are often data that can be represented as a curve over a continuum such as t ime. FDA allow for a comparison of nonfunction al or partially functional data (Ramsay and Silverman 2005). It achieves this by approximating measured and modeled data using piecewise curves. While there are many aspects to FDA, the basic process applied he re is to transform the measured and modeled data to a normal distribution using a box cox transformation, divide the data into 27 segments, fit a piecewise continuous curve to the data, perform a principle component analysis to determine proper of design m atrix, group the measured data by forcing their matrix vectors to sum to zero, then perform an F test between the measured an modeled curves with =0.05. Results and discussion Measured Heat Balance on Pavement Measured densities, specific heat, albedo and conductivities for the asphalt and concrete pavements are shown in Table 4 1 alongside published values for asphalt and concrete. Due to variati on in specific heat, comparison between concrete and the other materials are conducted using heat storage as the dependent variable. The specific heat for the asphalt and concrete used in this study are below the range of typical asphalt and concrete. In the case of asphalt the limestone aggregate and asphalt is oxidized and the higher air content (approximately 10%) in the concrete

PAGE 91

91 the measured concrete density, specific heat, and conductivity are all lower than published values for concrete (Bentz et al. 2010 ; Newman and Choo 2003). The concrete unconfined compressive strength is 18.6 0.14 MPa. Temperatures measured at the pavement surface are not statistically significantly different (p < = 0.05) from the interior of the pavement according to the Ma nn Whitney rank sum test as shown in Figure 4 3. The relative percent difference (RPD) between the interior and surface temperature measurements are 1.96, 1.24, 1.91, and 1.89 for the control (conventional asphalt), reflective asphalt, concrete, and sealed asphalt, respectively. A test of significant difference with respect to the control is performed on the data presented in Figure 4 3 after transforming the temperature data into energy storage (W/m2) where cumulative heat flux = T t C p V) for t = 0 to 720 minutes at a 5 minute timestep (t 0 = 7:00). The results of the Mann Whitney rank sum test indicate that interior pavement temperatures for the reflective asphalt treatment is significantly different (p < = 0.05), the c oncrete is significantly different (p < = 0.05), and the sealed asphalt treatment is not significantly different (p > = 0.05) from the control. It is hypothesized that changing the pavement material from asphalt to concrete reduces heat storage during the highest probability of hourly summer rainfall in Gainesville. Analyzing historical rainfall data collected at the Gainesville regional airport between 1998 and 2008 (inclusive) for daily trends yields slightly different results for peak event frequen cy hour and time of peak precipitati on within the wet season (June to September) as shown in Figure 4 4. Events are defined as rainfall greater than 0.01 inches (the minimum rain gage sensitivity). The probability of a rainfall event occurring

PAGE 92

92 between mi dnight and 12:00 (noon) is low, a total of approximately 19.7%. The chance of an event occurring between 12:00 and 13:00 is almost double that of the previous hour, increasing every hour until after 15:00 and finally dropping below 4% at 21:00. The frequen cy distribution of rainfall is similarly higher during the same time interval but more normally distributed about 17:00. Events occurring after 16:00 tend to generate higher precipitation depth per event than events before 16:00. A cumulative distribution function (CDF) of precipitation probability is shown in Figure 4 4, indicating that after 19:00 approximately 80% distribution of rainfall has been accounted for. After 20:00, the CDF reaches 90%. According to the distribution of heat in Figure 4 5, after 19:00 there is <18% difference between reflective asphalt and the control (dropping to 16% at 20:00) and <3% between the concrete and the control. The low chance of a rainfall event before noon combined with the diminishing temperature differential between treatment methods after 19:00 indicates that this is a critical period in which to minimize the potential for heat pollution in rainfall runoff. Temperature data collected for the control, reflective asphalt, coated asphalt, and concrete are analyzed for temperature and heat flux trends as shown in Figure 4 5 and summarized Table 4 4. Results are utilized to (1) test the hypothesis that changing pavement reflectivity significantly mitigates potential storage of solar radiation in the pavement matrix and (2 ) test the hypothesis that concrete reduces potential heat transfer to runoff when compared to the control asphalt specimen during hours of peak rainfall frequency. The Mann Whitney rank sum test of significant difference indicates that both the reflective asphalt and concrete (but not the sealed asphalt) are significantly different lower than the control (p < = 0.05), supporting hypothesis (1) above. The

PAGE 93

93 performance of the reflective asphalt between 12:00 and 19:00 is 80% of the control while the concrete is 87% of the control as shown in Table 4 4 (lower is better), supporting hypothesis (2) above. Results f rom an analysis of three consecutive days of temperature data (8 September 11 September) are summarized in Table 4 5 and Figure 4 7. Results support aforementioned hypotheses. Concrete and reflective asphalt perform equally between 12:00 and 19:00, achie ving 78% 81% of the control. The daily concrete heat pattern is significantly different from the control over the three days shown (p < = 0.05). Reflective asphalt is not significantly different from the control on 8 September (p > = 0.05). By observation, Figure 4 7 also shows a delay in the rising limb of the concrete, 30 35 minutes later than the reflective asphalt curve (also obser ved in Figure 4 6). The falling limb is similarly shifted, however the peak for the concrete is observed to occur at the same time as the reflective asphalt. In two of the three days, heat is not lost from the concrete after the peak as fast as it is for t he reflective asphalt (similar to Figure 4 5). The exception is 8 September, where the light concrete performs similarly to the reflective asphalt, during concurrent wind gusts. In general, the daily results support the findings based upon averaged tempera ture data. Multiplying the probability of rainfall by the difference in heat storage for various pavement treatment methods provides the relative impact index (RII) for each hour of the day where 0 indicates no improvement and 1 indicates 100% mitigation of heat. This result is normalized to a maximum potential heat loss from the pavement that can occur during the maximum frequency of precipitation (600 KJ/m 2 x 11.75%) as shown in Figure 4 6A or maximum probability of a rainfall event (600 KJ/m 2 x 12.0%) as shown in

PAGE 94

94 Figure 4 6B Results indicate that concrete minimizes heat storage as compared to asphalt. Median performance of the concrete is 0.58 on the volumetric RII and 0.60 on the event frequency RII between the hours of 12:00 and 19:00 (Figure 4 6 ). Reflective asphalt performs both 5% and 6% better than concrete during the same time period for the volumetric and event frequency based RII, respectively. Neither the daily performance nor the performance between 12:00 and 19:00 is statistically signif icantly different according to the Mann Whitney rank sum test and the t test respectively (p < = 0.05). Figure 4 8 illustrates pavement heat loss during two rainfall events, on the 5 September and the 24 August. Figure 4 8 shows a heat loss of 773 KJ/m2, 941 KJ/m2, and 1062 KJ/m2 from the reflective asphalt, concrete, and control specimens between the onset of rainfall at 15:40 and the end of the event at 17:15. Figure 4 8 B shows a heat loss between the start of the event at 10:45 and 14:00 of 550 KJ/m2 a nd 670 KJ/m2 for the reflective asphalt and control, respectively. In comparison to the control, the reflective asphalt performs 9.3% better during an afternoon rain event ( Figure 4 9A ) than a morning event where the initial pavement heat storage is half t hat of the afternoon ( Figure 4 9 B ). The higher heat loss from concrete compared to asphalt in ( Figure 4 8 A ) may suggest that reflective asphalt performs better than concrete, however the rapid and strong response of the control and reflective asphalt to a change in solar insolation after 16:00 as well as the stronger heat loss from the control and reflective asphalt due to wind and radiation between 15:15 and 15:40 (600 KJ/m 2 499 KJ/m 2 ) compared to concrete (273 KJ/m 2 ) suggests that at parking lot surfaces where pre event wind was previously found not to be correlated to heat transfer to runoff as

PAGE 95

95 shown in Chapter 2, concrete and reflective asphalt would perform comparably. This also emphasizes a benefit of concrete, that it is observed to release heat mor e slowly due to convection than asphalt does. Figure 4 9 presents a narrow focus on pavement energy under two different radiation patterns, on the 10 September and 17 September. The pavement interiors are observed to respond to changes in radiation within 5 10 minutes of the change in solar radiation. One interesting finding is that the smoother solar radiation curve observed in Figure 4 9B corresponds to a higher peak pavement temperature for all specimens compared to results observed during the less stead y insolation in Figure 4 9A The reflective asphalt specimen more rapidly loses heat following its thermal peak when compared to the concrete, an observation that is consistent with Figure 4 5 and Figure 4 7. In a separate experiment, concrete curb tempera ture was previously measured at a University of Florida parking lot, along a transect described in the second chapter Concrete temperature was measured in a 6.1 m wide north south asphalt road and 300 mm wide concrete curb next to the road. Measurements we re made 15 cm on either side of the asphalt concrete seam for the asphalt and concrete measurements, respectively, and 15 mm below the pavement surfaces. This experiment was performed in duplicate, with one location at the east side of the road, and the othe r at the west side of the road. A graphical comparison of concrete to asphalt pavement temperature is shown in Figure 4 10. Note that the temperatures of the concrete readings on the East and the West are similar, only 0.87 C difference on average (3.3% d ifferent during peak hours) with the west curb higher in te mperature than the east curb. The difference

PAGE 96

96 in temperature between concrete and asphalt is similar to the difference between concrete and the control in Figure 4 5, suggesting that the application of concrete may provide similar thermal mitigation without the need to alter existing mix designs or apply special paints or coatings. Heat, rather than temperature, is the continual focus of this investigation because it allows for comparison of specimen s of differing material composition and because peak it is useful to consider the potential heat transfer to rainfall runoff. The RII heat results stress the importance of accounting for regional rainfall patterns in mitigation techniques. While the asph alt performs better than concrete as shown in Figure 4 6, because asphalt and concrete perform s imilarly over a 24 hour period as shown in Table 4 4 locations such as Portland OR with a consistent probability of rainfall, shown in Figure 4 1, may achieve better performance using concrete than Gainesville, FL. While many obs ervations are made from the 24 A ugust and 5 September rainfall events, one of the most interesting results is that changes in wind and radiation appear to affect asphalt specimens more than concrete. This may be a function of pavement thickness or a function of the reflective paint increasing convection, which is also supported by sealed asphalt results presented in Figure 4 8 A The second chapter found that wind before a rain event h ad little effect on heat transfer to runoff, suggesting that, in parking lots, more stored energy may be transferred to rainfall runoff than presented here. The use of concrete thus would likely result in reduced heat transfer to runoff t, thereby increas ing the performance of concrete relative to asphalt. During data collection for the roadway experiment, shadows were present at various points in the day which will have affected pavement temperature in both asphalt

PAGE 97

97 and concrete results due to the close p roximity of the measurements. Future studies would benefit by analyzing when shading is critical to pavement heat content in order to make recommendations as to proper placement of shade trees in parking lots. Heat Balance Simulation Model Utilizing the physical model data, a heat balance CFD model is created. Figure 4 11 illustrates the model fit of measured data for the control (r 2 = 0.986) and reflective asphalt (r 2 = 0.982) pavements for 18 August 2010. The F test indicates no significant difference (p < = 0.05) between measured and modeled results for both the control and reflective asphalt. A second series of simulations illustrates the effect of changing thermal conductivity within the ranges presented for asphalt and concrete in Table 4 1. Results a re shown in Figure 4 12. The difference in models after varying conductivity from 1.2 to 1.8 W/m K is not statistically significant (p < = 0.05) and resulting curves overlap for the entire simulation. Given the small variation observed when changing co nductivity, a second simulation of asphalt temperature is performed to test model performance subject to dry weather conditions using 19 August data. Results are summarized in Figure 4 13 illustrating the overall model fit (r 2 = 0.987 control, r 2 = 0.99 r eflective asphalt). A third series of measured and modeled comparisons are made using 6 September weather data. The fit (r 2 ) between measured and modeled for each pavement treatment is greater than 0.95, and modeled and measured results are not significa = 0.05). The falling limb of the model deviates from measured results in the August simulations while the rising limb deviates in the September simulation. Deviation in the falling limb may occur because heat stored in the rooftop it self is radiating up to the

PAGE 98

98 specimens. It may also be a function of the higher pyranometer elevation such that it receives more shading at 17:00 than the specimens. A likely cause for the need to utilize reduced wind speed in the simulation is because of l ocal obstructions. Previous publications (Touma 1977, Blackadar 1962) have also measured lower wind speeds at pavement surface than at >10 m. Simulation results show that it is possible to use 2006 computational technology to simulate pavement temperature. Run times are 3.3 times as fast as a physical measurement experiment. Performing simulations offers the opportunity to rapidly test multiple pavement types, such as warm mix asphalt and pavement with softeners or various surface treatments in a under cons trained or unconstrained weather scenarios. The model presented herein can also be used to enhance existing models such as the NCAR urban canyon model (Oleson et al. 2010) or to build on work by Wu et al. (2008) who augmented heat transfer in asphalt pave ment by adding graphite. This model can also be expanded to simulate thermal connectivity with the subgrade and optimize thermal properties of an engineered subgrade. It can additionally be used to assess the impact of various surface treatments and mix de signs on ice formation in cold climates under varying conditions of humidity and temperature. Summary Through experimental research, three major hypotheses are investigated. The first, that changing pavement composition from asphalt to low density concrete reduces pavement heat storage during the period of maximum daily rainfall frequency, cannot be rejected. In North Central Florida, the precipitation frequency is highest time period between 12:00 and 19:00. Results show a statistically significant differe nce between control asphalt and concrete heat storage patterns during these hours. Results from a

PAGE 99

99 separate roadway experiment indicate that traditional concrete may provide similar thermal mitigation to the low density concrete measured here without alteri ng mix designs or applying special paints or coatings. A thermal relative impact index (RII) is created to better compare the relative performance of concrete and reflective asphalt. According to the RII, reflective asphalt performs 5% to 6% better than co ncrete. The second hypothesis, that pavement albedo can be changed to reduce heat storage in pavements, cannot be rejected. Results show a statistically significant difference in heat storage patterns by analyzing the entire population of gathered data for concrete and reflective asphalt. The third hypothesis, that CFD can be used to simulate heat storage using physical pavement properties and weather data as model inputs, is not rejected. Simulations successfully model measured weather conditions and gener ate pavement temperature results that are not statistically significantly different from measured data for three example days. Two of the three simulations do, however, begin to depart from measured pavement temperature data after 17:00 and possible causes of this departure are discussed

PAGE 100

100 Table 4 1 Thermal and physical properties of pavement Item Control Reflective Asphalt Concrete Asphal t** Concrete +, Density (kg/m 3 ) 2203 2203 2079 2100 2400 1600 2000 Cp (J/kg K) 950 950 915 1000 1400 920 1004 K ( W/m K) 1.7 1.7 1.0 1.2 1.8 0.2 1.0 Viscosity (kg/m s)* 50000 50000 50000 NA NA Albedo .22 .6 0.5 0.1 0.2 0.4 0.5 Thickness (mm) 55 55 65 NA NA Area (m 2 ) 0.02076 0.02581 0.01761 NA NA *Specified for purposes of model function ality; + Bentz et al. (2010) ; Newman and Choo (2003) ; ** Van Buren ( 2000 ) and Janke et a l ( 2009 ) Table 4 2 Model parameters for computational simulation Item Value Solver Transient Gravity 9.81 m/s 2 on x axis Equations used Energy, flow Models used Solar load, energy Pressu re velocity coupling Simple scheme Gradient discretization Green Gauss node Pressure discretization Body force Momentum discretization 1 st order upwind Energy discretization 2 nd order upwind Transient formulation 1 st order implicit Table 4 3 Pro perties of a ir and e xpanded p olystyrene (EPS) Item Air EP S* Density (kg/m 3 ) 1.18 15 C p (J/kg K) 1006.4 1300 K (W/m K) 0.0242 0.038 Viscosity (kg/m s) 1.789E 05 NA Glicksman et al ( 1987 ); Yajnik and Roux (1990)

PAGE 101

101 Table 4 4 Median values of p avement heat cycle for all measured days. Energy stored in pavement Daily Median 12:00 19:00 Control ( KJ/m 2 ) 668 2617 Reflective asphalt (KJ/m 2 ) 574 2089 Sealed asphalt (KJ/m 2 ) 623 2539 Concrete (KJ/m 2 ) 572 2279 Reflective asphalt (% of control) 0. 86 0. 80 Sealed asphalt (% of control) 0.93 0.97 Concrete (% of control) 0.8 6 0. 87 Note: Data are not normally distributed Table 4 5 Integration of pavement heat cycle heat for 8 September to 10 September Daily Median 12:00 19:00 Median Energy stored in pavement (KJ/m 2 ) 8 Sept 9 Sept 10 Sept 8 Sept 9 Sept 10 Sept Control ( KJ/m 2 ) 813 935 1153 3070 3373 3771 Reflective asphalt ( KJ/m 2 ) 699 804 992 2399 2740 3083 Sealed asphalt ( KJ/m 2 ) 730 826 1039 3075 3397 3864 Concrete ( KJ/m 2 ) 621 743 933 2404 2720 3033 Reflective asphalt (% of control) 0.86 0.86 0.86 0.78 0.81 0.82 Sealed asphalt (% of control) 0.90 0.88 0.90 1.00 1.01 1.02 Concrete (% of control) 0.76 0.80 0.81 0.78 0.81 0.80 Sealed asphalt results in less daily heat storage when compared to t he control asphalt but slightly higher storage during peak hours Reflective asphalt reduces heat storage by 18% or more and concrete reduces heat storage by 19 % or more duri ng the critical hours of 12:00 to 19:00.

PAGE 102

102 Figure 4 1 Comparison of rainfall pattern frequency by hour from 10 years of hourly rainfall data collected in two climates in the United States A B

PAGE 103

103 Figure 4 2 Schematic of simulation geometry. A) plan view, B) side view of cropped mesh C) side vi ew geometry. N ominal representation of value (see Table 4 1 for dimensions per specimen) The outer box (air enclosure) represents the domain extent for the simulation. The mesh contains 165,123 elements (average skewness = 0.240 + 027). A curvature size function with medium smoothing and a slow transition setting was used with a smooth transition inflation function, a transition ratio of 0.272, 3 layer maximum, and growth rate of 1.2 to produce cells between 4.3E 4 meters and 8.6E 2 m eters B A C

PAGE 104

104 Figure 4 3 Comparison of temperatures at surface and interior of pavements, 15 September, 2010. Average relative percent difference between interior and surface temperature measurements for the A ) control, B ) reflective aspha lt, C ) sealed as phalt, and D ) concrete are 1.96%, 1.24%, 1.91%, and 1.89%, respectively. C A B D

PAGE 105

105 Figure 4 4 Relative distribution of rainfall event occurrence and total rainfall depth by day hour during the rainy season (Ju ne September, inclusive) from 10 years of historical data collected in Gainesv i lle, FL. The distribution of events based upon the onset time of rainfall is denoted event onset hour and the distribution of rainfall depth by hour is denoted precipitation d uring hour.

PAGE 106

106 Figure 4 5 Mean hourly temperature and heat absor ption with standard deviation. KJ are per unit area 1m 2 Control, reflective, and sealed asphalt data collection period is between 17 August and 22 S eptember Concrete data collectio n period is between 4 September and 22 September. Standard error bars are shown. C A B D

PAGE 107

107 Figure 4 6 Relative impact index (RII) for pavement heat stora ge reduction in Gainesville, FL (negativ e is better). Results are nor malized to the product of maximum difference from control heat content and either A) rainfall depth or B) event frequency Median concrete performance is 0.58 on the volumetric RII between 12:00 and 19:00 and 0.60 on the even t frequency RII The reflective asphalt performs 5% and 6% better than the concrete for the volumetric and event frequency based RII, respectively. Standard error bars are shown

PAGE 108

108 Figure 4 7 Continuous measurement of A) Cumulative heat storage in pavement and B) atmospheric conditions between 8 September and 11 September, 2010. KJ are per unit are 1m 2 The areas under the curve illustrate that there is a hysteretic cycle of heat gain in the pavement where a minimum cumulative heat level is reached between 7:00 and 7:30 each day, a fter all the heat is exhausted. A B

PAGE 109

109 Figure 4 8 Comparison of pavement temperature before, during, and after two rain events of differing intensity and time of day. It is observed that wind cools the pavements before the storm onset. The rate of thermal recovery in the pavements is not proportional to the rate of change of solar radiation, following the event, suggesting evapo ration is mitigating heat ga in. A B

PAGE 110

110 Figure 4 9 Comparison of thermal heating pattern on two dry days of differing radiation A ) o n the 17 September and B ) o n the 10 September. A B

PAGE 111

111 Figure 4 10 Concrete t empe rat ure and asphalt temperature at A ) east side of road, B ) west side of road, and C ) difference from asphalt thermocouple (TC) temperature at both locations. There is a reduction in temperature of >5 degrees between 12:00 and 16:00. These results are similar to the 4 C to 6C difference duri ng these hours shown in Figure 4 4 A B C

PAGE 112

112 Figure 4 11 Modeled pavement temperature for control asphalt and white asphalt pavements on 18 August, 2010. Tests for goodness of fit: r 2 =0.986 0.981 for control and white asphalt, respectively. No significant difference is found between measured and modeled results using functional data analysis ( p < =0).

PAGE 113

113 Figure 4 1 2 Comparison of modeled pavement temperature results under for current, low, and high thermal conductivity (k) values (k=1.7, 1.2, 1.8 W/m K, respectively) for white asphalt simulation. Results indic ate r 2 = 0.999 and no significant difference using functional data analysis ( p < =0)

PAGE 114

114 Figure 4 13 Measured vs. modeled asphalt temperatures for two days in August, 2010. The rising limb deviates more during this even t th an during the 18 August. Not e that it also shows a drop in radiation at approximately the same time of day. The model fit statistics are r 2 =0.987, 0.99 for control and white asphalt, respectively. No significant difference is found between measured and modeled results using functional data analysis ( p < =0)

PAGE 115

115 Figure 4 14 A comparison of measured and modeled asphalt and concrete temperatures on 6 September, 2010. In comparison to the aforementioned simulations, the model fit is better at the end of the day than befor e 12:00. Model fit statistics r 2 =0.954, 0.944, 0.980 with for control, white asphalt, and light concrete, respectively. No significant difference is found between measured and modeled results using functional data analysis ( p < =0)

PAGE 116

116 CHAPTER 5 COMPUTATIO NAL MODELING OF OVER LAND FLOW AND HEAT T RANSFER IN ASPHALTIC PAVEMENTS Background Asphalt concrete is the predominant pavement surface type in the United States. Approximately 94% of roads in the United States are asphalt (Takamura 2002; Anderson et al. 20 00). However, asphalt has been found to contribute to increased stormwater runoff temperature in Florida in Chapter 2 The rate at which heat is gained or lost from an asphalt pavement is a function of mix design, additives, and/or coa tings, as discussed i n Chapter 4 A commonly used alternative paving material to asphalt is Portland cement concrete. North America has used using Portland cement concrete to construct roads since 1881 (Snell and Snell 2002). The most direct method to measure heat transfer pot ential to stormwater runoff is by performing highly controlled ex situ physical experiments but it is time consuming and challenging to compare thermal responses from the variety of pavement materials to stormwater runon. However, with knowledge of basic m aterial properties, a computational model can rapidly prototype the thermal response of different pavements by digitally altering material properties. Previous studies have performed simulations of asphalt pavement temperature as a function of overland flo w (Janke et al. 2009 ; Van Buren et al. 2000 ; Roa Espinosa et al. 2003 ; Minhoto et al. 2005 ; Yavuzturk et al. 2005 ; Krause et al. 2004). Most of them account ed for evaporation usi ng empirical methods. Janke used an unsteady 1 D model that required coefficie n ts of convection. Van Buren used a 1 D finite difference model called the t hermal r unoff m odel for p avement (TRMPAVE) Roa t hermal u rban r unoff m odel (TURM) was on a la rge watershed scale. Krause used a h ydrologic s imulation p rogram f ortran

PAGE 117

117 (HSPF) which wa s applied to a large watershed and focused on s tream temperature. Minhoto used a custom 3 D finite element method for calculating asp halt temperature. Yavuzturk used a 2 D finite difference model and calculated a convection heat transfer coefficient using the flat plate method (Incropera and DeWitt 2002). The models were designed and calibrated for use at large scales. Some but not all of the models account ed for evaporation. Janke used the heat flux equation put forth by Ste fan et. al (1980) to account for thermal transfer and did not account for mass transfer. Van Buren used 1964) and an equation by Linsley et al. (1975) for heat flux. Roa Espinosa did not provide a method used to c alculate evaporation. Minhoto, Yavuzturk, and Krause also did not discuss evaporation. Part of the reason evaporation was not included in all th e aforementioned investigations is because evaporation has historically been a challenging phenomenon to model. Fundamental models of evaporation/condensation stem from kinetic and statistical rate theory (Rahimi and Ward 2005). The equations applied in this study are k inetic. A widely known k inetic equation i s by Schrage (1953), who focused on the interface between water liquid and vapor simple motion prevails near the interface. The rate of mass transfer at the interface is the sum of condensation and vaporization, each of which are calculated separately as shown in Equation 5 1 (Schrage 1953 ; Marek and Straub 2001). ( 5 1) In this equation, w is the mass flux vector (kg/m 2 sec) s is the evaporation coefficient, M is the molar mass (kg/mol), R is the universal gas constant (J/mol K), is the Sc hrage

PAGE 118

118 correction factor which accounts for net velocity of vapor molecules under non steady conditions, p v is the vapor pressure (N/m 2 ), p l is the liquid vapor pressure (N/m 2 ), T v is the liquid temperature (K) and T l is the vapor temperature (K). This equa tion is a modification of the Hertz Knudsen formula that allows for a non stationary flow of vapor (Barrett and Clement 1991) ( Equation 5 2), which assumes a Maxwellian distribution of molecules at the interface. ( 5 2) In this equation, e is an evaporation coefficient, e is the condensation coefficient, p is the vapor pressure far away from the interface (N/m 2 ), T is the temperat ure far away from the interface (K). Assuming only minor departure from equilibrium conditions, the equation can be written as shown in Equation 5 3 (Eames et al 1997; Schrage 1953; Kucherov and Rikenglaz 1960). ( 5 3) Assuming the temperature of the gas is approximately the same as the temperature of the liquid, the H K equation ( Equation 5 2) can be further simplified and rearranged as Equation 5 4, which has been used in numerous inve stigations (Alty and Mckay 1933; Alty and Mckay 1935 ; Bowman and Briant 1947; Carman 1948; Eames et al.1997). ( 5 4) In this equation, 1 is the evaporation coefficient, T s is t he temperature at the surface of the liquid/air interface (K), P s is the saturated vapor pressure (N/m 2 ), and P 0 is the

PAGE 119

119 current vapor pressure (N/m 2 ). This is only valid as w/ws+ approaches zero (far away from the interface) This approximation gives rise Equation 5 5). ( 5 5) In this equation W is mass transfer (kg/sec) and A is interface area (m 2 ). The derivation of this equation is well described in Nabavian and Bromley ( 1963). Eames et al. (1997) applied the modified correction factor from Schrage to Equation 5 4, resulting in Equation 5 6. ( 5 6) Marek and Straub (2001) review ed a number of published coefficients and proposed that evaporation and c ondensation coefficients are higher for moving thin films ( >0.1) than they are for quasi static surfaces ( <0.1). Through experimentation, Nabavian and Bromley (1963) found that >0.35 for water. Hardt and Wondra (2008) use d = 1 because they state d t hat it creates a more numerically challenging situation to model. Eames et al. (1997) review ed the literature and conclude d data suggest >0.5 with little deviation in evaporation rate when 0.5 < < 1. Computational fluid dynamics (CFD) have been previou sly developed to model evaporation using Kineti c theory (Hardt and Wondra 2008; Welch and Wilson 2000). Both Hardt and Wondra and Welch and Wilson use d CFD with a volume of fluid (VOF) scheme to model the multiphase water liquid/vapor interface. Hardt and Wandr a applied their own user defined functions to a commercially available CFD package to simulate inter phase mass and energy flux. Both studies claim ed good agreement with measured results. Both studies focus ed on the micro scale. Computational power is such that

PAGE 120

120 increasingly complex phenomena can be simulated in domains of increasing size. My study endeavors to apply a user defined function to a commercially available CFD package in order to simulate the effects of heat transfer from asphaltic and Portl and cement concrete pavement under controlled benchtop scale conditions while accounting for the generation of turbulence and the effects of evaporation/condensation. Objective The goal of my research is to simulate heat transfer from two pavement surfaces under constant rate overland flow of water and validate each model against ex situ pavement tests. I seek to demonstrate that the flow of stormwater over a pavement surface can be modeled using computational fluid dynamics (CFD). A secondary hypothesis is that evaporation measurably affects pavement and runoff temperature. It is hypothesized that flow can be modeled using a laminar regime for travel lengths of 2 feet (0.61m). A fourth hypothesis is that the flat plate model for forced convection does not p rovide as accurate of an approximation of measured heat transfer to stormwater runoff as does the CFD evaluation. Methodology This study is completed in two phases, beginning with physical testing and ending with simulations of the experiment. Physical te sting consists of calibrating thermocouples (TC), obtaining pavement materials, instrumenting the pavement with TCs, performing the overland flow experiment, and then measuring the thermal properties of the pavement after the completion of the overland flo w experiments. Simulations consist of developing CFD models (with appropriate material properties and experimental conditions) for both asphalt and concrete tests, followed by a comparison with the classical flat plate model for heat transfer in overland f low

PAGE 121

121 Physical Experiments Omega {TC PVC T 24} 0.5mm TCs are used to instrument the asphalt specimen. Omega {TC TT T 30} 0.25mm TCs are used to instrument the concrete specimen. TCs are calibrated by simultaneously recording the t emperature of water using T Cs and an alcohol thermometer every minute as the water is heated. C hanges in temperature are recorded with a timestamp that is used to generate a calibration curve for the TCs. All thermal data are logged at 2 minute intervals using a Campbell Scientific CR 10x logger. The asphalt experiment is constructed by compacting asphalt (FDOT FC 5) into a 38mm x 305mm x 610 mm wooden form by hand tamping, allowing the pavement to cure for 10 weeks. Calibrated TCs are then placed into the pavement by drilling a hole o Figure 5 1, towards the pavement surface so as not to disturb the surface of the pavement. Drill depth for a surface TC is 2mm while interior TC borings reach 19mm beneath the pavement surface. One TC is installed in each boring by inserting the tip upwards from the pavement bottom, and backfilling the boring with pavement filler (<30% silica crystalline, <25% petroleum asphalt, <15% latex polymer). The asphalt bottom and si des are then filled with the pavement filler and smoothed. The TC wires are pressed into the curing filler material. The cured specimen is then placed into a 37mm x 305mm x 610mm, 18 gage steel tray before testing. The concrete experiment is constructed by troweling a 3000 psi Type II cement with a 4 5 in. slump 3 % air, and size 67 coarse aggregate into an 18 gage steel tray of internal dimensions 38mm x 305mm x 610mm, then placing the TCs into the wet concrete by burying them to the same locations as the TCs in the asphalt specimen, also shown in Figure 5 1. Surface TCs are buried 2 mm beneath the concrete surface

PAGE 122

122 and interior TCs are buried to a depth of 19mm. The pavement is troweled to a smooth surface and the concrete is allowed to hydrate for 8 weeks b efore performing experiments. Pavement properties are determined as follows, with results shown in Table 5 1: Mass is measured using a Melner scale, and bulk density is calculated as mass per volume. Specific heat capacity is measured for both pavements u sing a calibrated, low heat capacity expanded polystyrene calorimeter. Calorimetric tests of the specimens are performed by measuring temperature change of a known volume of liquid water at ambient temperature after placing a pavement specimen heated to 60 C in the calorimeter. Conductivity is estimated from pavement thermal diffusivity as measured by the time to reach equilibrium in the calorimeter, based upon the methods of Army Corps of Engineers (1949). After heating either the asphalt or concrete speci men in an oven at 65 o C, the tray with pavement is placed into a 51mm thick expanded polystyrene (EPS) insulation bed (R value = 7) which is also heated to reduce the heat gradient between the pavement and the insulation bed. The insulation bed height is pr eviously trimmed at the entrance to fit a 3mm thick polyvinyl chloride splash plate (not shown) at the pavement entrance to transition water flow from the influent pipe to the chamber to smoothly transition flow across the entire width of the pavement. The downstream end of the insulation bed is also trimmed to channel flow to a central outlet. An EPS cover with closed top and sides (interior width = 310 mm ) is placed atop the insulation bed and pavement, creating a flow chamber. The assembly (including insu lation bed) is then placed on a 2% downslope.

PAGE 123

123 Before water flow is turned on, Omega {TC PVC T 24} TCs are placed in duplicate at the both water influent pipe and the water effluent channel (not shown). A Campbell Scientific thermocouple multiplexer (AM25T) and a Campbel l Scientific logging device (CR 10x) are used to record temporal data on a 5 second intervals. Air temperature is also measured in the lab using an Onset HOBO U12 logger. Influent flow rate is volumetrically calibrated by measuring time to fil l a flask calibrated to 4L + 10mL and determined to be 0.485 L /s (kg/s). With the datalogging equipment operational, flow is turned on and the start time is recorded. After flowing through the experimental domain, flow is discharged into a floor drain. The experiment is conducted for >5 minutes. Data are then uploaded to a computer and analyzed for thermal patterns. Modeling Methodology The hydrodynamics and heat transfer dynamics of the pavement runoff system can be approximated using a 2 dimensional (2 D) spatial environment. A 2 D CFD analysis is developed in Fluent using a k kl Defined Functions (UDF) are incorporated into the model to assist in simulating physical phenomena. An initialization UDF is used to produ ce an x y distribution of the initial temperatures created from the pavement TC measurements made at time zero of the physical experiment. This is necessary because initial temperature gradients within the pavement are unavoidable. The UDF is invoked durin g initialization of the simulation. Using a cell loop, it steps through a [3,256] array of the x position, y position, and pavement temperature (K). For threads i n the solid domain, Equation 5 7 holds true where the measured temperature at the index of the array with the shortest linear path from the cell location at the current position in the loop is applied to the cell at the current position.

PAGE 124

124 ( 5 7) In this equation i is the current cell in the cell loop, j is the current index of the array, x i is the current x coordinate in the cell loop, y i is the current y coordinate in the cell loop, X j is the X location at the current index in t he array, Y j is the Y location at the current index in the array. D is the linear distance between the current array X,Y location and the current x,y cell coordinate, T j is the temperature at index j of the array and ti is the temperature of the current ce ll in the cell loop. After initialization, a different UDF is called during each iteration to make adjustments before the next iteration commences. This is used to calculate evaporation/condensation of H 2 O using a mass transfer mechanism. The simulation is performed under compressible flow, utilizing fundamental equations of mass continuity and momentum. The continuity or conservation of mass equation is shown in Equation 5 8 (Patankar 1980). ( 5 8) In this equation is the rate of change of density per unit volume, where density (kg/m 3 ) can be related back to mass fraction and temperature (K) via an equation of state, is the velocity field (m/s). S m is the mass (kg) added to the continuous phase from a secondary phase. Conservation of momentum is written for an inertial frame of reference as shown in Equation 5 9 for the x direction. ( 5 9)

PAGE 125

125 In this equation, u is directional velocity, is the viscosity (Pa s), p is static pressure (Pa), B x is the directional body force per unit volume, and V x represents additional viscous terms (Patankar 1980). Heat transfer is physically modeled using energy conservation as shown in Equation 5 10. No te that in the solid zone, heat transfer can be simplified as a function of conductivity and radiation. (5 10 ) The first term on the right hand side of the equation is ener gy transfer due to conduction. The second term is species diffusion; the third term is viscous dissipation. In this equation k is laminar conductivity (W/m C), J j is diffusion flux of species j (kg/m 2 s), h j is enthalpy (J/kg), and S h is the heat of chemi cal reaction plus radiation. is the visco us stress, shown in Equation 5 11 where I is the unit tensor. E is the total energy, defined as shown in Equation 5 12. ( 5 11) ( 5 12) In this equation, h is def ined for compressible fluids as shown in Equation 5 13 where Y j is the mass fraction of species j Specific enthalpy is defined as shown in Equation 5 14. ( 5 13) ( 5 14)

PAGE 126

126 In thi s equation T ref is 298.15 K. These equations can be solved simultaneously for compressible flow when coupled with the equation of state for gases (Versteeg and Malalasekara 1995 ; Batchelor, 2000). The mass transfer UDF operates in the fluid domain, where the gradient surface area of the liquid/air interface is reconstructed at each timestep when the UDF is invoked (Rider and Kothe 1998). Molar fraction of the water vapor ( y vap ) is calculated using Equation 5 15 if the mass fraction of vapor is greater than zero, otherwise y vap is zero. ( 5 15) In this equation, MW air = 28.0 kg/kmol, MW vap = 18.0 kg/kmol, X vap is mass fraction of vapor (kg vapor/kg air). Saturation pressure ( P sat Pa) is calculated based on the local cell temperature following an eight term polynomial equation published by Reynolds (1979). Vapor pressure is calculated as shown in Equation 5 16. ( 5 16) In this equation, P vap is vapor pressure (Pa), gas is the gas density (kg/m 3 ), R is the universal gas constant (J/kmole K), T c is the cell temperature (K). Evaporation occurs if P sat > P vap but if the volume of gas is less than 10% of the cell, th en evaporation is not expected to occur and is set equal to zero. This is necessary to reduce numerical instability. Evaporation flux is calculated according to Equation 5 17, reproduced below. ( 5 17)

PAGE 127

127 In this equation A i (m 1 ) is calculated as shown in Equation 5 18, l is the density of the water liquid (kg/m 3 ), V l is the volume of liquid in the cell (m 3 ) and V c is the volume of the cell (m 3 ), TS is the timestep (seconds), and W is the interfacial mass transfer rate per unit of volume (kg/m 3 s). ( 5 18) If the volume of liquid is not less than 10% of the cell volume, and if P sat < P vap then condensation occurs as shown in Equation 5 19, where x vap is the mass fraction of vapor in the gas, V g is the volume of gas in the cell (m 3 ), and V c is the volume of the cell (m 3 ). The total mass transfer to or from the liquid phase ( M ) at the interface between liquid and gas during e ach timestep is shown in Equation 5 20. ( 5 19) ( 5 20) The CFD model is shown in Figure 5 2. Properties for the air mixture, water vapor, air, water liquid, and steel used in the simulation are shown in Table 5 2. Model parameters are shown in Table 5 3. The following boundary conditions are specified in the simulation: The tray is simulated as a solid boundary along the bottom and sides of the pavement with a wall thickness of 1.024mm. Pressure boundaries are defined to have 1E 6 m 2 /s 2 laminar kinetic energy, no turbulent kinetic energy, a 1s 1 specific dissipation rate, and a vapor mass fraction of 0.0143 kg H 2 O/kg dry air. The water inlet is d efined to have a water flow rate of 0.485 kg/s (l/s), a hydraulic diameter of 0.002m, and no turbulence. Liquid temperature is specified using measured data every second at the mass flow influent boundary using a transient profile and air temperature is

PAGE 128

128 sp ecified every second at the pressure boundaries. The water air surface tension is specified as 0.04 N/m. The geometry is tested for grid convergence. Simulation monitors are used to export runoff temperature and pavement temperature over time. Heat T rans fer C alculation of F low O ver a F lat P late Incropera and DeWitt (2002) provide d a rigorous method for the calculation of convection over a flat plate. The method is applied to stormwater flow over a pavement surface shown below and model results are compare d to measured results The basic convection equation is shown in Equation 5 21. ( 5 21) In this equation, T is the free stream temperature ( C), and the average heat transfer coefficient is calculated as shown in Equation 5 22. ( 5 22) In this equation, k is pavement conductivity (W/m C) and L is the length of pavement flow (m). The Nusselt number (unitless) is calculated differently under laminar and transitional flow. Under laminar flow [Re x < 5X10 5 ] and Equation 5 23 is applied. Under transitional flow [ 5X10 5 <= Re x < 1X10 7 ] and Equation 5 24 is applied. ( 5 23) ( 5 24) In the previous two equations, Re Pr is the Prandtl lated as shown in Equation 5 25. ( 5 25)

PAGE 129

129 In this equation, V is the cross sectional runoff velocity (m/s), L is the length of the pavement transect (m), and is the kinematic viscosity (m 2 /s). Velocity is calculated as shown in Equation 5 26. ( 5 26) In this expression, V t is the runoff cross sectional velocity at time t (m/s), Q t is the flow at time t (m 3 /s), B is the path width (m), and H t is the depth of water over the pavement at time t (m). Given a known flow rate ( Q ), H t equation with substitution as shown in Equations 5 27 to 5 32. ( 5 27) ( 5 28) ( 5 29) width perpendicular to flow ( 5 30) ( 5 31) ( 5 32) In these equations, n S is the slope of the water surface (m/m), V is the cross sectional average velocity (ft/s m/s), k is a conversion constant equal to 1.486 for U.S. customary units or 1.0 for SI units, a nd A is the cross sectional area of the flow.

PAGE 130

130 The water runoff temperature is calculated as a function of heat transfer on a 10 second timesteps where the flat plate method described above is used to calculate the heat transfer during the timestep. The hea t capacity of water is used to calculate the temperature change in water as a function of volumetric flow rate of runoff in that timestep as shown in Equation 5 33. ( 5 33) In this equation T i is the influen t water temperature and T o is the effluent water temperature (C), Q l is the liquid flow rate, and C p is the specific heat capacity of water (4200 J/kg C). Results and Discussion CFD model performance for 300 seconds of flow time is shown in Table 5 4. Si mulation time for the most complicated simulations (evaporation, turbulent flow) ranges between 60 70 hours on an 8 core workstat ion with Intel Nehalem design. Results indicate that laminar simulations perform more poorly than their respective turbulent s imulations. This is most strongly shown in the difference between measured and modeled runoff temperature. There is little difference in the model RMSE or RPE between evaporation coefficients of 0.5 and 0.1 for the turbulent simulations. The turbulent conc rete simulation with =0.5 performs better than =0.1 except for the upstream internal temperature. The turbulent asphalt simulation with =0.1 shows less error in runoff temperature than =0.5 but almost identical error in pavement temperature. Interestingly, the turbulent mo dels that do not include evaporation provide a better estimation of runoff temperature than the simulations with evaporation.

PAGE 131

131 In multiphase simulations, body forces and pressure gradients can dominate convection and viscosity, causing poor convergence. T he CFD software utilized in this study allows for the use of an implicit body force that provides better pressure field stability during initial iterations. It requires the specification of operating density for VOF simulations. Due to the compressibility of the water vapor species, operating density is set to zero for simulations that perform evaporation, otherwise operating density is specified as 1.225 kg/m 3 Results of asphalt simulations that incorporate the aforementioned changes are shown in Table 5 5. They indicate that the model fit is worse for all the simulations with respect to pavement temperature but better with respect to error in runoff temperature. Convergence also occurs as a much faster rate (approximately 3 times as fast) and with 1 order of magnitude better continuity residual error. A third analysis was performed to determine the effect of changing the threshold for evaporation or condensation to take place from 10% to 50%. Hence, for condensation to occur in an interface cell, the cel l must have at least 50% water present; for evaporation to occur the cell must have at least 50% air present. The analysis performed for asphalt and concrete simulations with =0.1. Results show that error in runoff temperature is the lowest of all simulations for concrete and asphalt as shown in Table 5 6. In addition, in the asphalt simulation the RPE and RMSE for the upstream and downstream internal pavement temperature are lower than in the turbulent asphalt simulation shown at the top of Table 5 4 but error i n surface temperature is worse. G raphical representation s of the results for the third analysis are shown in Figure 5 3 and Figure 5 4.

PAGE 132

132 Results from the simulation of pavement runoff show that the simulated effluent temperature approaches the measured effluent temperature after 240 seconds. In addition, the first 20 seconds of the simulation follow measured effluent results ( Figure 5 1). However, the interim simulation results are up to 1.4 degrees higher than measured at the point of largest difference. The measured upstream interior temperature of the asphalt experiment cools off more rapidly than downstream until it meets the interior temperature measured downstream. At the asphalt surface, downstream simulation results also do not decay as quickly as upstream simulation results. The concrete simulation differs in the following manner: the upstream interior temperature is observed to cool much more rapidly than downstr eam and the shape of the temperature decay curve is similar to those shown for the surface temperatures. The specimen was examined and it was determined that both the upstream and downstream interior TCs floated closer to the surface while the specimen was hydrating. The upstream TC rose approximately 9mm closer to the surface while the downstream TC rose approximately 4mm. These corrections were made in the CFD models before simulations were performed. The results of the flat plate model are shown in Figur e 5 5. While the CFD model tends to over predict runoff temperature, the flat plate method tends to under predict runoff temperature. The flat plate model does not fit as well as the CFD model results in Table 5 6 for the concrete simulation, while the opp osite is true for asphalt. Previous research in the second chapter describes a more rapid cooling of upstream pavement temperature with a slower decay in downstream pavement temperature. This is observed in both measured and simulated results in this

PAGE 133

133 expe riment. The poor fit of the laminar simulations suggests that a laminar simulation is not appropriate, contrary to the hypothesis. However, the poor results are only present during the first 100 seconds of simulation time, after which the pressure field st abilizes and simulated runoff temperature approaches measured values (not shown). The asphalt simulation with the specified operating density and implicit body force stabilizes much earlier in the simulation (after 16 seconds of simulation). The use of imp licit body force is very effective at improving results for simulations without turbulence. The similarity in performance between turbulent simulations that incorporate evaporation and those that do not suggests that evaporation is not critical to measur ing runoff temperature during the first 5 minutes of runon over a hot pavement. It is even less important as time progresses due to the cooler pavement surface. Interestingly, the concrete (zero evaporation) simulation effluent starts approximately 1C war mer than the measured value but in 2 seco nds it drops by approximately 1 C and follows the measured curve closely. Effluent temperature in the zero evaporation asphalt is approximately 1C warmer for 8 seconds before following measured results for 10 secon ds. In both of the simulations shown in Figure 5 3 and Figure 5 4, effluent temperature starts at the measured temperature, suggesting that evaporation may only be critical within the first seconds of a runoff event. It is possible that the experimental de sign is minimizing observed evaporation, leading to the very small difference between accounting for and not accounting for evaporation. The closed top design was thought to be necessary at the outlay of the experiment because it is used to block extra ex perimental air currents and to allow for a smaller modeling domain. Evaporation was observed to occur because following the

PAGE 134

134 experiment, droplets were observed on the underside of the cover. There is no forced airflow in the domain and it is likely that eva poration may have a larger e ffect over pavement surfaces in situ The shorter time to runoff from the asphalt ( Figure 5 3 and Figure 5 4) may be due to increased surface depression storage in the asphalt, whereas the concrete is smoother than the asphalt. Water was initially observed to flow through channels along the asphalt pavement surface due to surface tension. It took approximately 30 60 seconds for the liquid to cover the entire pavement surface. This may also be a factor in the increased heat loss observed from the measured internal asphalt temperature compared to the modeled temperature during the initial 200 seconds of the simulation. Depression storage increases the effective surface area of the asphalt, increasing heat transfer. However, this wo uld likely also lead to increased runoff temperature. The flat plate method is observed to under predict concrete runoff temperature for the entire duration of the 300 seconds shown in Figure 5 5 B but it simulated asphalt pavement temperature much better than any of the CFD simulations between 50 300 seconds. The flat plate method requires knowledge of the pavement surface temperature as a function of time in order to calculate runoff temperature, which limits its utility, however if those data are availa ble, these results suggest that it may be a good approximation of temperature from a pavement surface. Previous researc h, shown in Chapter 4, has successfully modeled ex situ pavement temperature over varied weather conditions using CFD. It did not includ e the simulation of runoff temperature during rain events. The results presented herein suggest that the previously designed model and the model presented herein can be

PAGE 135

135 combined into a unified model to simulate heat exchange from concrete pavements under b oth wet and dry periods. Summary A model is successfully created to simulate the flow of stormwater over heated concrete for 5 minutes (300 seconds) within less than 0.5% RPE of measured pavement and runoff temperature. A model is also created for asphalt flow with less than 0.3% RPE of measured pavement tem perature and runoff temperature but asphalt model performance decreases after 40 seconds. Conversely, the flat plate method is observed to perform well after 100 seconds of flow over the asphalt pavement which does not support the hypothesis that CFD modeling is more accurate than the flat plate method for measuring runoff temperature from an asphalt pavement. This hypothesis is supported, however, by the concrete results. Results for both concrete and a sphalt simulations are not strongly improved by accounting for evaporation, however models that incorporate turbulence (via the k kl transitional model) are observed to perform Results suggest that there is potential for com parison between concrete pavement of different mix designs to help identify the effect of mix design on pavement runoff temperature. The tools presented herein can be used in part to evaluate the efficacy of these solutions and possibly provide future TMDL BMPs and better understand heat transfer during the early period of rainfall runoff events.

PAGE 136

136 Table 5 1 Thermal and physical properties of pavement Property Asphalt Concrete Density (kg/m 3 ) 2393 2252 Cp (J/kg K) 1008 1104 K (W/m K) 1.8 2.19

PAGE 137

137 Table 5 2 Material parameters used in computational fluid dynamics simulation Species Density (kg/m 3 ) Specific heat (j/kg k) Thermal conductivity (w/m k) Viscosity (kg.m s) Mass diffusivity (m 2 /s) Molecular weight (kg/kgmol) Enthalpy (j/kgmol) Air vapor mi xture Ideal gas Mixing law Mass weighted Mass weighted 2.88E 05 --Water vapor Ideal gas 2014 0.026 1.34E 05 -18.02 4.07E+7 Water liquid UDF 4182 0.600 0.0018 -18.02 0 Air 1.225 1006.43 0.024 1.79E 05 -28.97 0 Steel 8030 502.38 16.270 ----*When applying the UDF, w ater density and speed of sound are defined using the Tait equations ( Dymond et al. 1988 ).

PAGE 138

138 Table 5 3 Model parameters for computational simulation Item Value Solver Transient, pressure based Gravity 9.81 m/s 2 on y axis Models used VOF multiphase, k kl turbulence, species transport Pressure velocity coupling PISO Gradient discretization Least squares cell based Pressure discretization PRESTO! Momentum discretization QUICK Density discretization QUICK Volume fraction Modified HRIC Turbulent kinet ic energy QUICK Laminar kinetic energy Second order Specific dissipation rate QUICK Energy discretization 2 nd order upwind Gas Phase Water Vapor QUICK *P ISO : pressure implicit with splitting of o perators ; P RESTO! : pressure staggering o ption ; Q UICK : qu adratic upwind i nterpolation ; H RIC : high resolution interface c apturing

PAGE 139

139 Table 5 4 Analysis of error between modeled and measured results. Pavement Turbulence Model Evap Model C evap Upstream Surface Downstream Surface Upstream Interior Downstream Inter ior Effluent Water RMSE RPE RMSE RPE RMSE RPE RMSE RPE RMSE RPE Asphalt Turbulent None NA 0.882 0.19% 1.319 0.27% 0.118 0.03% 0.262 0.07% 0.402 0.11% Asphalt Turbulent UDF 0.1 0.941 0.20% 1.458 0.31% 0.113 0.03% 0.329 0.09% 0.853 0.18% Asphalt Tur bulent UDF 0.5 0.924 0.20% 1.394 0.29% 0.116 0.03% 0.290 0.08% 0.534 0.16% Asphalt Laminar None NA 1.668 0.43% 2.340 0.65% 0.169 0.05% 0.410 0.11% 1.689 0.44% Asphalt Laminar UDF 0.1 1.681 0.44% 2.290 0.62% 0.288 0.08% 0.568 0.15% 2.342 0.64% Asphalt La minar UDF 0.5 1.717 0.44% 2.470 0.68% 0.274 0.08% 0.577 0.16% 1.926 0.50% Concrete Turbulent None NA 3.482 0.35% 3.304 0.50% 0.654 0.12% 0.833 0.22% 0.166 0.04% Concrete Turbulent UDF 0.1 3.449 0.40% 2.942 0.50% 0.738 0.15% 0.727 0.19% 0.278 0.05% Concr ete Turbulent UDF 0.5 3.429 0.39% 3.003 0.50% 0.742 0.15% 0.681 0.18% 0.355 0.10% Concrete Laminar None NA 3.802 0.87% 4.353 1.18% 1.623 0.46% 0.733 0.18% 3.289 0.93% Concrete Laminar UDF 0.1 3.363 0.81% 4.100 1.13% 1.731 0.49% 0.738 0.20% 1.216 0.25% C oncrete Laminar UDF 0.5 3.512 0.93% 3.875 1.11% 1.742 0.53% 0.522 0.14% 3.014 0.91% Table 5 5 Analysis of error between modeled and measured results with implicit body force and specified operating density. Pavement Turbulence Model Evap Model C evap Upstream Surface Downstream Surface Upstream Interior Downstream Interior Effluent Water RMSE RPE RMSE RPE RMSE RPE RMSE RPE RMSE RPE Asphalt Turbulent None NA 2.165 0.57% 2.757 0.77% 1.138 0.30% 1.324 0.35% 0.383 0.11% Asphalt Turbulent UDF 0.1 2 .305 0.68% 2.999 0.90% 1.702 0.45% 1.870 0.49% 0.538 0.16% Asphalt Turbulent UDF 0.5 2.157 0.65% 3.251 0.97% 1.426 0.37% 1.627 0.42% 0.511 0.15% Asphalt Laminar None NA 2.186 0.58% 2.860 0.81% 0.719 0.19% 1.139 0.30% 0.509 0.10%

PAGE 140

140 Table 5 6 Analysis of error between modeled and measured results with 50% evaporation/condensation threshold. Pavement Turbulence Model Evap oration Model C evap Upstream Surface Downstream Surface Upstream Interior Downstream Interior Effluent Water RMSE RPE RMSE RPE RMSE RPE RMSE RPE RMSE RPE Asphalt Turbulent UDF 0.1 0.865 0.19% 1.307 0.28% 0.125 0.03% 0.225 0.06% 0.368 0.10% Concrete Turbulent UDF 0.1 3.314 0.39% 3.073 0.49% 0.775 0.16% 0.301 0.08% 0.117 0.03%

PAGE 141

141 Figure 5 1 Installation of thermocouples in pav ement s pecimen. A) Top view and side view and B ) front view. Note that air cavity chamber and EPS insulation bed are not shown in top or side view.

PAGE 142

142 Figure 5 2 CFD mesh dimensions and statistics. Mesh Type: MapPave ; e lement Nodes: 19925 ; c ell thickne ss at water inlet = 0.5mm, 4 cell thickness ; c ell thickness above water inlet = geometric growth edge sizing (bias factor = 2), 8 cell thickness ; s olid domain cell size = 4mm ; f luid domain cell width = 2mm at pressure inlet and bottom wall at water inlet, 4mm in main domain ; c ell thickness at pressure inlet = 2mm ; w edge cell size = 4 mm

PAGE 143

143 Figure 5 3 Measured and modeled asphalt specimen temperature and effluent temperature. Model shown incorporated 50% evaporation/con densation threshold, turbulent, and =0. 1

PAGE 144

144 Figure 5 4 Measured and modeled concrete specimen temperature and effluent temperature. Model shown incorporated 50% evaporation/condensation threshold, turbulent, and =0. l.

PAGE 145

145 Figure 5 5 Effluent temperature m odeled using flat plate method for both A) asphalt and B) concrete A B

PAGE 146

146 CHAPTER 6 GLOBAL CONCLUSION Four investigations are performed to better understand the heat transfer phenome na to and from pavement bodies, before and during storm events. The first two investigations focus on in situ temperature measurement. R esearch findings from an investigation of 17 rainfall events at a University of Florida faculty parking lot indic ate that event heat transfer from an asphalt pavement surface during the rainy season in North Central Florida is flow limited, with cumulative flow as an appropriate surrogate for cumulative heat transfer to the rainfall runoff for 12 of 17 storms. It is also found that the average pavement temperature before a rain event is very strongly correlated with heat transfer and that concrete temperatures before an event are lower than asphalt temperatures. Heat balance models are able to approximate measured da ta R esults from the first investigation suggest that storm water runoff temperature is not equal to pavement surface temperature when sampling flow from a large contributin g area. It is posited that the flow regime may diminish heat transfer, creating an i nsulating boundary layer if flow is laminar or increasing air entrainment if turbulent. It is also observed that there is a sharp difference between subgrade thermal response and pavement temperature. The second investigation consists of a series of sub e xperiments performed to determine the impact of shade on pavement temperature Pavement surface temperature is measured when exposed to sunlight, shaded by a parked vehicle, and shaded both by a vehicle and by tree canopy. There is a peak difference of mor e than 16 o C between shaded and unshaded surface temperature but no significant difference between the pavement temperature beneath vehicles. It is also determined that vehicle

PAGE 147

147 surface temperatures reach more than 60C. A first flush may occur on full parki ng lot due to the potential energy release from vehicle surfaces during an event, however this investigation is not performed. Radiation from the chassis of a recently operated vehicle (before parking) is observed to dampen the cooling capacity of vehicula r shading and it is suggested that frequent removal and replacement of a vehicle from a cool parking spac e may lead to a gradual increased in pavement temperature over time. An investigation of the temperature along a temporally shaded transect illustrates the effect of shadows on pavement temperature but also illustrates that localized horizontal conduction to cool pavements has a demonstrable effect on surface and subsurface temperatures. It is recommended to orient to face east with shade trees reducing radiation on empty and vehicle occupied parking spaces during a peak insolation period. The third and fourth investigations are performed ex situ to aid in the creation of computational fluid dynamics models. In addition, the third investigation compares the relative performance of pavements of different material composition and surface reflectivity. Results of the third investigation show that concrete performs comparably to an asphalt to which a reflective coating is applied The CFD model is found to si mulate measured internal pavement temperature with no significant difference between measured and modeled results. The model also performs this successfully using weather data input and pavement material composition properties. The fourth chapter details an investigation in the relationship between pavement temperature and runoff temperature. Similar to in situ results, a temperature first flush is seen. A model is created to simulate the measured results and it is concluded that flow

PAGE 148

148 can be simulated as t urbulent over the surface. The CFD model performs better when simulating concrete than asphalt. Theories are postulated as to why there is higher error in simulating asphalt temperature. An interesting finding from the investigation is that there is little improvement by modeling evaporation, suggesting that, for the conditions of the physical experiment, evaporation may not be significant. These results support the future investigation of engineered systems that can achieve multiple goals, such as porous concrete which allows for groundwater recharge, runoff reduction, and reduced thermal storage. Studies have shown that using a BMP such as permeable paving provides reducti ons in runoff temperatures of 2 C to 4 C in comparison to asphalt streets (Haq and J ames 2002). Simply changing the pavement color can also have profound impacts outside the field of stormwater such as the urban heat island effect (Akbari et al. 2009) for a lower cost while addressing the issue of runoff temperature, however it does nothi ng to minimize peak flow. Shading by natural foliage allows for the reduction in pavement temperature and peak runoff volume, critical to maintaining cool stream temperatures (Roy 2005), however if not incorporated correctly, it can cause increased nutrien t loading to receiving waters. The tools presented herein can be used in part to evaluate the efficacy of these solutions and possibly provide future BMPs. More advanced tools can be used to evaluate the more intricate mechanisms occurring during rainfall runoff to better understand heat transfer mechanisms during the early period of rainfall runoff events.

PAGE 149

149 LIST OF REFERENCES wide urban albedos to offset CO 2 J ournal (J ) C limatic Change 94(3), 275 286. Solar Energy, 70(3), 295 310. g the land cover of an urban environment using high J. Landscape and Urban Planning 63(1), 1 14. heating and cooling energy use in four Canadian c Energy, 17(2), 141 149. Proceedings of the Royal Society of London A149, 104 116. Philosophical Magazine 15, 82 103. J. Atmospheric Environment 33(24 25), 3911 3918. Anderson, D., Youtcheff, J., Zupanick, M. (2000). Asphalt b inders Transportation Review Board Committee on Characteristics of Bituminous Materials (A2D01), Washington D.C., 6pp. Biological Report, 90(22), 1 13. U. S. Fish and Wildlife Service, Washington, D.C. Handbook for Concrete and Cement CRD C44 63, Army Corps of Engineers, Mt. Vernon, NY. accessed 3 22 11 258. port of heat and moisture and its Boundary Layer Meteorology 65(1), 159 179. J. Atmospheric Enviro nment 30(3), 413 427.

PAGE 150

150 solid J. Physics: Condensed Matter, 12, 2567 2577. co J. Colloid and Interface Science 150(2), 352 364. Batchelor, G. (2000). An Introduction to Fluid Dynamics Cambridge University Press, Cambridge, UK. 615. Bentz, D., Petlz, M., Duran proper ties of high J. Building Physics 34(3), 263 275. Journal of Geophysical Research 67(8), 3095 3102. Bowman, J Industrial and Engineering Chemistry 39(6), 745 751. reflective materials to mitigate urban h Atmospheric Environment 32(1), 95 101. Buildings and the Environment International Conference Assessment methods, natural resources 2, CSTB, France, 309 318. Transactions of the Faraday Society 44, 529 536. Celestian, S., Martin, C. (2004). "Rhizosphere, surface, and air temperaeture patterns at parking lots in Ph oenix, Arizona, U.S." Journal of Arboriculture 30(4), 245 252. Choubane, B., Sholar, B., Musselman, J., Page, G. (1999). "Ten Year Performance Evaluation of Asphalt Rubber Surface Mixes." J. Transportation Research Board, 1681(2), 10 18. Chow, V.T. (1964) Handbook of applied hydrology: a compendium of water resources technology, McGraw Hill Book Co., New York, 1500 pp. Chudnovsky, A., Ben Energy and Buildings 36(11), 1063 1074. City of Olympia, (1994). Impervious surface reduction study technical and policy analysis final report Olympia Public Works Department, Olympia, Washington..

PAGE 151

151 ACM SIGGRAPH Computer Graphics 18(3), 137 145. Environmental Biology of Fishes 18(3), 161 172. Daufresne, M., Roger, M., Capra, H., and Lamouroux, N. (20 term changes within the invertebrate and fish communities of the Upper Rhone River: effects of Global Change Biology 10(1), 124 140. Davidson M., Dolnik, F. (2002). Parking Standards, Planning Advisory Service Report 510/511 American Planning Associateion, Chicago, IL, 181 pp. Davies, P.H. (1986). "Toxicology and Chemistry of Metals in Urban Runoff." Urban Runoff Quality Impact and Quality Enhancement Technology B. Urbonas and L.A. Roesner, eds., American Society of Civil Engineers, New York, NY, 60 78. International J. Thermophysics 9(6), 941 951. International J. Heat Mass Transfer 40(12), 2963 2973. 049 00 3, Electric Research Alto, CA. EPA Environmental Protection Agency. (1987). Clean Water Act. 40 C.F.R 104 149, Office of the Federal Register runoff relationships in the J. American Water Resources Association 26(2), 313 322. Frank, D. American Journal of Preventative Medicine 27(2), 87 96 Galli, J. (1990). Thermal Impacts Associated With Urbanization and Stormwat er Best Management Practices Metropolitan Washington Council of Governments, Washington, DC. 188 pp. ASME Journal of Heat Transfer 109, 809 812. Golden, J. (2 006) "Photovoltaic canopies: thermodynamics to achieve a sustainable systems approach to mitigate the urban heat island hysteresis lag effect." International Journal of Sustainable Energy 25(1), 1 21.

PAGE 152

152 Golden, J., Carlsonb, J., Kaloushc, K., Phelan, P. (20 thermal and radiative impacts of photovoltaic canopies on pavement surface Solar Energy 81(7), 872 883. Transportation Planning 28(1), 3 11. Grimmond, C., Oke, scale observations Journal of Applied Meteorology 38(7), 922 940. properties on su J. Materials in Civil Engineering 19(8), 683 690. Hahn, H., Pfeifer, R. (1994). "The contribution of parked vehicle emissions to the pollution of urban run off." Science of the Total Environment 146 147, 525 533. Haq, R.U., and James Global Solutions for Urban Drainage, 9 th International Conference on Urban Drainage ASCE, Portland OR. 195 205. flows based on a continuum J. Computational Physics 227, 5871 5895. Haselbach, L. (2008). The engineering guide to LEED new construction: Sustainable construction for engineers McGraw Hill Professional, NY, NY 392pp. Urban Ecosystems 4, 193 229. simulation Journal of Hydrology 356(3 4), 327 343. J. Transportation Research Record 1764(2001), 180 188. Horner, R.R. (1994). Fundamentals of Urban Runoff Management: Technical and Institutional Issues. Terrene Institute, Washington, D.C., 302 pp. Incropera, F. P., De Witt D. P. (2001). Fundamental of Heat and Mass Transfer, 5th ed. Wiley, New York, 944 pp. Institute of Transportation Engineers (1987). Parking Generation, 2 nd Edition ITE, Ishington, D.C., 350 pp. Iqbal, M. (1983). An Introduction to solar radiation Academic Press, NY, NY, 390pp.

PAGE 153

153 nt of stormwater by urban Advances in Modeling the Management of Stormwater Impacts, Vol. 5 W. James, Ed., CHI, Canada, 155 175. heating of local urban stormwat New Applications in Modeling Urban Water Systems W. James, Ed., CHI, Canada, 39 157. rainfall J. Hydrology, 365(3 4), 195 212. Kim, K from an asphalt Water Resources Research 44(W11405), 1 13. Kinouchi, T., Ibaraki, T., Yoshinaka, T., Fukae, N., and Session 4, Mitigation of Urban Heat Islands. Fifth Conference on Urban Environment American Meteorological Society, Vancouver, BC, Canada. (Mar. 5, 2011). Kjelgren. R., Montague, T. (1998). "Urban tree transpiration over turf and asphalt surfaces." J. Atmospheric Environment 32(1), 35 41 The art of computer programming, sorting and searching, third edition Addison Wesley, Boston, MA, 409 426. Krause, C., Lockard, B., Newcomb, T. J., Kibler, D., Lohani, V., and Orth, D. J. (2004). J. American Water Resources Association 40(6), 1645 1658. Doklady Akademii Nauk SSSR (Phys. Chem. Section), 133, 735 737. Langford, T.E. (1990). Ecologi cal effects of thermal discharges Elsevier Science Pub., New York, NY, 470 pp. Laverne, R., Winson Geideman, K. (2003). "The influence of trees and landscaping on rental rates at office buildings." Journal of Arboriculture 29(5), 281 290. Leith, R., and Northwest Science 74, 69 75. reflectance of Cement and Concrete Research 32(11), 1679 1698.

PAGE 154

154 Linsley R., Kohler M., Paulhus J. (1975). Hydrology for Engineers McGraw Hill Inc, New York. 482 pp. Litman, T. (2006). Parking Management Best Practices American Planning Assoc iation, Chicago, Illinois. pp12,20 Louisiana Department of Environmental Quality (2001). Indian Creek and Indian Creek Reservoir TMDL for Temperature: Subsegment 060206 US EPA Region 6, Baton rouge, LA. 12 pp. International J. Heat and Mass Transfer 44, 39 53. mental 123. McPherson, E., Muchnick, J. (2005). "Effects of street tree shade on asphalt concrete pavement performance." Journal of Arboriculture 31(6), 303 310. Mestayer, P., and An Diffusion and Transport of Pollutants in Atmospheric Mesoscale Flow Fields A. Gyr, F. Rys, Eds., Kluwer Academic Publishers, Norwell, 165 189. m for approximate solar position (1950 Solar Energy 41(1), 113. Solar Energy 40(3), 227 235. Minhoto, M., Pais, J., Pereira, P., and Pica do pavement temperature with a three J. Transportation Research Record 1919, 96 110. Applic ations of Soil Physics D. Hillel, Ed., Academic Press, Orlando, FL, 247 253. Biological Aspects of Thermal Pollution P. Krenkel, F. Parker, Eds., Vanderbilt Univ. Press, Nashville. 294 317. Chemical Engineering Science 18, 651 660. National Climatic Data Center (2009). "COOP Select State." COOP Data/Record of Climatological Observations, US, (Dec. 6, 2009)

PAGE 155

155 Newman, J., Choo, B.S. (2003). Advanced concrete technology: Processes Elsevier Ltd. Burlington, MA, 704pp. Papers in Regional Science 84, 145 158. Quarterly Journal of the Royal Meteorological Society 108, 1 24. temperature in a global climate mo Geophysical Research Letters 37(L03701), 1 7. Proceedings of the European Computational Fluid Dynamics Conference, 1st, Brussels, Belgium, 93 100 Oregon Climate Service (2010). "Services: Site specific climate reports/summaries." Oregon Climate Service, (Feb. 18, 2010) Oregon Department of Environmental Quality (2008). Molalla Pudding Subbasin Total Maxi mum Daily Load (TMDL) and Water Quality Management Plan (WQMP), Portland, OR. 243 pp. Patankar, S. (1980). Numerical heat transfer and fluid flow Taylor & Francis, London, UK. Paul, M.J., and Meyer, Annual Review of Ecology and Systematics 32(1), 333 365. Pomerantz, M., Pon, B., Akbari H., Chang, S.C. (2002). temperatures on air temperatures in large cities Lawrence Berkeley N ational Laboratory, Berkeley, CA, 20pp. International J. Thermodynamics 8(1), 1 14. Ramsay, J. S ilverman, B.W ( 2005 ). Functional Data Analysis 2 nd Edition Spring er, New York, NY. 450pp. J. Agricultural and Applied Economics 33(2), 271 281. J. Computational Physics, 141 112 1 52.

PAGE 156

156 Roa impact of urban development on stream temperature using a thermal urban runoff U.S. EPA National Conference on Urban Stormwater Proceedings: Enhancing Prog rams at the Local Level EPA, Chicago, IL, 369 389. scale effects of riparian Canadian Journal of Fisheries and Aquatic Sciences 62(10), 2312 2329. Ryan, P Water Resources Research 10(5), 930 938. Roa impact of urban development on stream temperatu re using a thermal urban runoff U.S. EPA National Conference on Urban Stormwater Proceedings: Enhancing Programs at the Local Level EPA, Chicago, IL, 369 389. f Wet Weather Flow in the Urban Watershed: Technology and Management R. Field, D. Sullivan, Eds., CRC Press, Boca Raton, FL, 43 63. Dissolved Solids in Small Impe Journal of Environmental Engineering 130(11), 1301 1314. Runoff Loadings to a Partial J. Environmental Engineeri ng 131(8), 1155 1167. Sansalone, J., Kertesz, R., Maccarone, K., Raje, S., Seltzer, K., Siminari, M., Simms, P. (2009). "Green Infrastructure and LID Design for a Florida Constructed Environs Subject to Rainfall Runoff Loadings" Proceedings, Australia: St ormwater 2009 Joint Conference of the Stormwater Association of New S. Wales and Victoria Sydney, 2009. Environmental Research Letters 4(3), 1 7. Schrage, R. A theoretical study of interphase mass transfer Columbia Univ. Press, NY New York, 1953. 103pp. Scott, K., Simpson, J., McPherson, E. (1999). "Effects of tree cover on parking lot microclimate and air quality." J. Arboriculture 25, 129 142. Sheng, Y., Ying, G., and Sansalone particulate and dissolved water chemistry load indices in rainfall runoff from urban Journal of Hydrology 361(1 2), 144 158.

PAGE 157

157 Journal of Planning Education and Research. 17(1), 3 20. Shoup, D. (2005). The High Cost of Free Parking Planners Press, American Planning Association, Chicago, Illinois. Concrete Interna tional, March, 72 74 Search 2(5), 172. Stefan, H.G., Gulliver, J., Hahn, M.G., Fu, A.Y. (1980). Water temperature dynamics in ng. Project Report 193, St.Anthony Falls Lab., University of Minnesota, Minneapolis, MN, 217 pp. properties, and thermal performance of cool colored coatings for the urban Solar Energy 81, 488 497. Polymer Dispersions and Their Industrial Applications D. Urban, K. Takamura, Eds., Wiley VCH, Weinheim, Germany, 301 306. Thacker, B., Doebling, S., Hem ez, F., Anderson, M., Pepin, J., Rodriguez, E. (2004). Concepts of model verification and validation. LA 14167 MS, Los Alamos National Laboratory, CA, 41pp. Stormwater Runoff fro Journal of the American Water Resources Association 44(5), 1325 1336. Journal of Air Pollution Control Association 27, 863 866 U.S. Army Corps of Engineers (1956). Summary report of the snow investigations: Snow hydrology report U.S. Army Corps of Engineers North Pacific Division, Portland, OR, 405 pp. U.S. Green Building Council, (2009). Green neighborhood development: LEED re ference guide for neighborhood development U.S. Green Building Council, 473pp. Water Research 34(4), 1359 1371.

PAGE 158

158 Versteeg, H. Malalasekera, W. (1995). An introduction to computational fluid dynamics Addison Wesley, London, UK. ACM Transactions on Graphics 3(1), 52 69. Welch, S., W J. Computational Physics 160, 662 682. American Planning Association 61, 29 42. distribution in conductive asphalt solar collector due to pavement material Materials Science Forum 1314, 575 578. Yajnik, S.J., Roux, J.A. (1990). Insulation materials, testing, and applications, ASTM STP 1030 D.L. McElroy and J.F. Kimpflen, Eds., American Society for Testing and Materials, PA 561 574. fluctuations in asphalt pavements due to thermal environmental conditions using a two dimensional, transient finite J. Materials in Civil Engi neering 17(4), 465 475.

PAGE 159

159 BIOGRAPHICAL SKETCH Ruben Kertesz was born in the Pacific Northwest, and grew an affinity to know more about why he exists, how things work, and nature. Encouraged to carefully and cautiously explore he ventured into environm ental subjects throughout his scholastic career. While a t tending a high school in the Pacific Northwest, Ruben was invited to join an environmental science club which afforded him an opportunity to perform hands on research on artificial reefs. Ruben reali zed the joy of building and testing, recording data and presenting his findings. Throughout college, Ruben has participated in numerous environmental action He obtained a m ee from the Department of Environmental Engineering Sciences at the University of Florida in 2005, focusing on water resources conservation and stormwater mitigation by low impact development Ruben is an engineering intern and his interests have focused o n sustainable construction practices. Ruben still carries a passion for integrating research, social awareness, and technology. He reminds himself every day that the wellbeing of him and humankind depend on sound and conscious environmental and social thou ght