Ceilometer observations of Vancouver’s urban boundary layer Validation and mixed-layer height estimation by Derek van der Kamp, B.Sc., The University of Victoria, 2005 A THESIS SUBMITTED IN PARTIAL FULFILMENT OF THE REQUIREMENTS FOR THE DEGREE OF Master of Science in The Faculty of Graduate Studies (Geography) The University of British Columbia (Vancouver) August, 2008 c© Derek van der Kamp, 2008 Abstract A mini-lidar system, Vaisala’s CL31 ceilometer, was installed within a sub- urban area of Vancouver, BC, for the purpose of making continuous observa- tions of the boundary layer over a period of 11 months. Initial validation of the ceilometer for use in boundary layer observations was undertaken. This involved the comparison of ceilometer data with eight months of ground- level particulate matter measurements, as well as with 16 vertical profiles of particulate matter and meteorological data. Once a variety of persistent noise structures within the data were accounted for, it was found that the ceilometer data showed good agreement with the particulate matter data, suggesting its usefulness for assessing air-quality throughout the bottom 1km of the atmosphere. Additionally, two algorithms were developed in order to estimate the height of the convective boundary layer, or the mixed-layer height, from the ceilometer data. One involved the fitting of an ideal-profile to the measured data, while the other involved the location of a minimum- gradient in the backscatter profile. The performance of these two techniques were assessed and compared, and it was found that the ideal-profile method was the more robust of the two. Finally, mixed-layer heights were estimated for fair weather, convectively active days. In order to isolate such conditions, an automatic flagging algorithm was developed. However, additional man- ii Abstract ual assessment was needed to avoided un-suitable conditions. Mixed-layer heights were estimated for 19 days over an 11 month period. the estimates presented here were found to agree with previous observations. Daily max- imum mixed-layer heights ranged from 650m in July to 350m in December, indicating that the height of the convective boundary layer within Vancouver is significantly suppressed due to the city’s coastal location. iii Table of Contents Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ii Table of Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . iv List of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viii List of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ix Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . xiv 1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 1.1 Impetus for study . . . . . . . . . . . . . . . . . . . . . . . . 1 1.2 Project overview . . . . . . . . . . . . . . . . . . . . . . . . . 4 2 Background and methods . . . . . . . . . . . . . . . . . . . . . 8 2.1 Geography of the Lower Fraser Valley . . . . . . . . . . . . . 8 2.2 Climatology and meteorology of the Lower Fraser Valley . . 10 2.3 Mesoscale circulations within the Lower Fraser Valley . . . . 11 2.4 Particulate matter within the Lower Fraser Valley . . . . . . 14 2.5 The CL31 ceilometer . . . . . . . . . . . . . . . . . . . . . . 16 2.5.1 Basic lidar theory . . . . . . . . . . . . . . . . . . . . 18 iv Table of Contents 2.5.2 Description of the CL31 ceilometer . . . . . . . . . . 19 2.6 Field site description . . . . . . . . . . . . . . . . . . . . . . 21 2.6.1 Ceilometer field installation . . . . . . . . . . . . . . 21 2.7 Processing of ceilometer data . . . . . . . . . . . . . . . . . . 23 2.8 Metro Vancouver air quality monitoring stations . . . . . . . 24 2.9 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25 3 Ceilometer validation . . . . . . . . . . . . . . . . . . . . . . . 27 3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 27 3.2 Ceilometer observations of the boundary layer . . . . . . . . 28 3.3 Ceilometer validation via measurements of ground-level par- ticulate matter concentrations . . . . . . . . . . . . . . . . . 31 3.3.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . 31 3.3.2 Methods . . . . . . . . . . . . . . . . . . . . . . . . . 32 3.3.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . 34 3.4 Ceilometer validation via direct tethersonde observations . . 40 3.4.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . 40 3.4.2 Methods . . . . . . . . . . . . . . . . . . . . . . . . . 40 3.4.3 Meteorology and ceilometer observations . . . . . . . 43 3.4.4 Relation between particulate matter concentrations and backscatter . . . . . . . . . . . . . . . . . . . . . 46 3.5 Development and optimization of a noise removal technique . 51 3.5.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . 51 3.5.2 Description of technique . . . . . . . . . . . . . . . . 53 3.5.3 Optimization of technique . . . . . . . . . . . . . . . 53 v Table of Contents 3.6 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55 4 Mixed-layer height estimations . . . . . . . . . . . . . . . . . 58 4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 58 4.2 Boundary layer theory . . . . . . . . . . . . . . . . . . . . . . 60 4.2.1 Classification of the atmospheric boundary layer . . . 60 4.2.2 Theoretical convective boundary layer development . 64 4.2.3 The thermal internal boundary layer . . . . . . . . . 65 4.3 Previous measurements of mixed-layer heights within the Lower Fraser Valley . . . . . . . . . . . . . . . . . . . . . . . . . . . 67 4.4 Development of mixed-layer height algorithms . . . . . . . . 71 4.4.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . 71 4.4.2 Discussion of issues inherent in mixed-layer height es- timations . . . . . . . . . . . . . . . . . . . . . . . . . 71 4.4.3 The minimum gradient method . . . . . . . . . . . . 74 4.4.4 The ideal-profile method . . . . . . . . . . . . . . . . 75 4.4.5 Assessment of algorithms . . . . . . . . . . . . . . . . 78 4.5 Flagging algorithm . . . . . . . . . . . . . . . . . . . . . . . 83 4.6 Mixed-layer height estimations . . . . . . . . . . . . . . . . . 85 4.6.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . 85 4.6.2 Results . . . . . . . . . . . . . . . . . . . . . . . . . . 86 4.7 Effects of the thermal internal boundary layer . . . . . . . . 95 4.7.1 Case study: August 15, 2007 . . . . . . . . . . . . . . 96 4.8 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99 4.9 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102 vi Table of Contents 5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107 5.1 Project outcomes . . . . . . . . . . . . . . . . . . . . . . . . 107 5.2 Future work . . . . . . . . . . . . . . . . . . . . . . . . . . . 110 Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113 vii List of Tables 2.1 1971-2000 Climate data for Vancouver International Airport (Source: Environment Canada) . . . . . . . . . . . . . . . . . 12 2.2 Vaisala CL31 Ceilometer performance specifications . . . . . 20 3.1 Pearson correlation coefficients for ground-level PM2.5 con- centrations and 45m ceilometer backscatter intensity values. . 34 3.2 Pearson correlation coefficients for ground-level PM10 concen- trations and 45m ceilometer backscatter intensity values. . . . 34 4.1 Theoretical estimates of the height of the thermal internal boundary layer for the Sunset Tower site on August 15, 2007. 98 viii List of Figures 2.1 The Lower Fraser Valley with the locations of urban areas outside the Greater Vancouver Regional District. . . . . . . . 9 2.2 An example of ceilometer data. Of note is the convective boundary layer development within the first 600m as well as the increasing noise with altitude. . . . . . . . . . . . . . . . . 17 2.3 Schematic of Ceilometer optics. Taken from Münkel et al. (2007) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20 2.4 A map of the Greater Vancouver Regional District. The star indicates the ceilometer field site, while points 1 and 2 indi- cate the Westham Island site and the site used for validation of the ceilomter, respectively. The locations of the GVRD air quality monitoring stations are also shown (the lettered points): A - Kitsilano (T2), B - North Burnaby (T4), C - South Burnaby (T18) and D - Vancouver Airport (T31). . . . 22 ix List of Figures 2.5 Examples of erroneous artifacts within the raw ceilometer data. A: An example of a persistent zig-zag feature (light- grey area) within the bottom 100m of the data. B: An ex- ample of the persistent fluctuations found within the bottom 50m of the ceilometer data. . . . . . . . . . . . . . . . . . . . 25 3.1 Correlation between 45 metre ceilometer backscatter and PM concentrations as the “post-rain” period in lengthened. PM data for July 2007 to February 2008 taken from station T18 (South Burnaby). . . . . . . . . . . . . . . . . . . . . . . . . . 36 3.2 Scatter plots comparing 45 metre ceilometer backscatter lev- els to ground level PM concentrations for all four GVRD air quality stations. Data is shown for both the daytime (Red) and nocturnal situation (Blue). . . . . . . . . . . . . . . . . . 38 3.3 Daytime and nocturnal correlation between 45 metre ceilome- ter backscatter intensity and PM concentrations for July 2007 to February 2008 (Data from Station T18 used). The shaded area indicates the range of daytime correlation values present between the four air quality stations themselves. . . . . . . . 39 3.4 Ceilometer observations from 1600, August 14 to 0800, Au- gust 15 (PST) Tethersonde flight locations are also shown along with flight numbers. PM2.5 measurements from the South Burnaby air quality station (T18) are shown in the lower panel. Sunrise and sunset times are indicated by the vertical dashed lines. . . . . . . . . . . . . . . . . . . . . . . . 43 x List of Figures 3.5 Meteorological and ceilometer observations for Flight 2. (As- cent: solid line, Descent: dotted line). The horizontal dotted line indicates the estimated altitude of the inversion . . . . . 45 3.6 Potential temperature profiles from selected tethered balloon soundings during case study. . . . . . . . . . . . . . . . . . . . 46 3.7 Meteorological, PM2.5 and ceilometer observations for flight 7. (Ascent: solid line, Descent: dotted line) . . . . . . . . . . 48 3.8 Scatterplots of PM1.0 (top) and PM10 (bottom) and ceilome- ter backscatter for all vertical profiles conducted with the GRIMM aerosol spectrometer (14 Profiles in total). Regres- sion lines are shown for the entire dataset (Blue) as well as for only the data deemed by the noise-removal technique to be physically based (Red). . . . . . . . . . . . . . . . . . . . . 50 3.9 Five hours of ceilometer data before noise removal is per- formed (top left), and the average profile of the five hour dataset (top right). The same data is shown after noise re- moval (bottom left), with the resulting average profile (bot- tom right). . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52 4.1 Diurnal development of an idealized boundary layer during fair weather conditions . . . . . . . . . . . . . . . . . . . . . . 61 4.2 An example of an ideal backscatter profile fitted to a mea- sured profile with the MLH (zi in Equation 4.5) indicated. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76 xi List of Figures 4.3 MLH estimates from both the ideal-profile method (red) and the gradient method (black) applied to a single convectively active summer day. . . . . . . . . . . . . . . . . . . . . . . . . 79 4.4 A: A comparison of MLH estimates from the ideal-profile method and the minimum gradient method for 22 fair weather days. B: Same as A, but only for days with an average differ- ence in the ML and FA backscatter values of 30 A.U. C: The Pearson correlation coefficient between the two algorithms as a function of the minimum ML to FA backscatter difference to which the algorithms were applied. . . . . . . . . . . . . . 80 4.5 Results from ten individual runs of the ideal-profile algorithm on a single day along with the 10-member ensemble average. The standard deviation of the ensemble is also indicated for each time step. . . . . . . . . . . . . . . . . . . . . . . . . . . 82 4.6 The mean daily trend in the MLH from 9 individual fair weather summer days, along with 4 of these individual daily trends. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87 4.7 The mean daily trend in the MLH from 9 individual fair weather summer days in which BL cumulus clouds formed, along with 3 of these individual daily trends and the clear MLH trend from Figure 4.6. . . . . . . . . . . . . . . . . . . . 89 4.8 Morning growth period for clear, fair weather summer days. A theoretically-estimated growth rate of 138 mh−1 is indi- cated with the parallel dotted lines. . . . . . . . . . . . . . . . 90 xii List of Figures 4.9 Three MLH estimates from sodar measurements (solid lines) and the corresponding model predictions (dotted lines) pre- sented by Steyn (1980): July 30, 1978: blue; August 2, 1978: grey; August 8, 1978: red. These measurements were made at the Sunset Tower. An representative day of MLH esti- mates from ceilometer data (black squares) is provided for comparison. Altered from Steyn (1980) . . . . . . . . . . . . 92 4.10 Mean daily trends in the MLH are shown from summer, au- tumn, and winter. The standard deviation for the mean trends are indicated by the dotted lines. . . . . . . . . . . . . 93 4.11 Daily maximum MLHs for 19 fair weather days with no BL cumulus formation. . . . . . . . . . . . . . . . . . . . . . . . . 95 4.12 A comparison of the theoretical, non-advective growth in the MLH for August 15, 2007 with the estimated MLH from the ceilometer data (using the ideal-profile method). . . . . . . . 97 xiii Acknowledgements I would like to extend my sincere gratitude to Dr. Ian McKendry for pro- viding such great guidance and tutelage throughout my degree program. I would also like to thank Dr. Tim Oke and Dr. James Voogt who, as prin- cipal investigators of the EPiCC project, provided me with the opportunity to work on this engaging and novel project. I am grateful to The Canadian Foundation for Climate and Atmospheric Sciences, the Natural Science and Engineering Research Council of Canada and British Columbia Hydro for supporting this research, as well as to Geoff Graham of Transport Canada for the permissions to fly the balloon. This project also greatly benefited from the meteorological and air quality data provided by Al Percival of Metro Vancouver. xiv Chapter 1 Introduction 1.1 Impetus for study Observational assessment of the atmospheric boundary layer (BL) is a signif- icant field of study, as ground-level and near-ground atmospheric processes have an obvious impact on both society and the environment. For example, knowledge of BL characteristics is crucial in developing an understanding of air pollution dynamics and possible detrimental impacts. As well, proper numerical weather forecasting requires accurate parameterization of the BL. By definition, the BL is the portion of the atmosphere directly affected by surface characteristics. Consequently, BL properties, like the surface, are highly variable and it is crucial that observations of the BL are undertaken in a wide range of locations. Yet, historically, there has been a bias in BL observations towards relatively simple, homogeneous terrain. It is therefore important to gain observational data of more complicated areas such as urban and suburban regions. The presence of the Urban Heat Island and the ubiquity of significant roughness elements within urban areas lead to unique meteorological and climatological characteristics within the BL such as increased convective activity, often sustained into the night (Oke, 1995). Consequently, the urban 1 Chapter 1. Introduction BL needs to be understood in order to address public health issues revolving around air pollution as well as accurate weather forecasting within cities. To this end, a suburban area of Vancouver, BC was chosen for the location of this project. Vancouver is situated within the Lower Fraser Valley (LFV) in Southwest BC. Vancouver’s coastal location, as well as its proximity to the complex to- pography of BC’s Coastal Mountain Range, leads to additional complexities within the region’s BL. For instance, conditions with high levels of insola- tion often result in a thermally-driven sea breeze in coastal locations such as Vancouver. This meso-scale flow has a significant impact on pollutant transport and dispersion. The morning onset of a sea breeze within Van- couver often acts to transport pollutants up the LFV. Additionally, on-shore advection can suppress the growth of the BL, which has large implications for pollutant dispersion. A number of processes can also occur at the steep valley walls of the LFV, leading to the formation of elevated pollutant lay- ers (McKendry & Lundgren, 2000). Density-driven drainage flows are also common during the night among mountainous terrain, and can lead to signif- icant pollutant transport. An in-depth analysis of these complex processes requires continuous and detailed observations of the BL. Although observations of the BL are an important field of study, the logistics associated with the field usually prove to be quite complicated. Di- rect, in-situ measurements of the BL are often performed using tethered weather balloons or single-use radiosondes. However, safety becomes an is- sue when making these types of measurements within an area of high air traffic such as Vancouver. Additionally, due to logistics, continuous observa- 2 Chapter 1. Introduction tions via weather balloons are rarely carried out for longer than a few days, and are only able to provide “snap-shots” of BL structure. Remote sensing is another technique used for BL observations. This includes acoustic so- dar techniques, radar observations and lidar observations. Although these instruments are able to provide a signficant amount of information both ver- tically and over time, their operation is costly and logistically intensive. For instance, the use of audible acoustic pulses means that sodar measurements need to be performed at a reasonable distance from residential areas. Lidar observations involve the operation of a powerful laser which is not eye-safe, presenting obvious safety issues. Additionally, these remote sensing instru- ments often require supervision by a trained technician. Consequently, they are usually only run for a few days at a time, not continuously. Given both the importance and the difficulty of BL observations, this project involves the assessment of the ceilometer as a possible cost-effective and simple option for making detailed and continuous observations of the BL. The ceilometer is a low-power, eye-safe mini-lidar system which was originally designed to measure cloud heights on a continuous, unsupervised basis. However, in recent years it has begun to be used for determining BL structure by analysing the signal in the data attributed to ground-sourced aerosols within the BL. A significant property of the BL is the mixed-Layer height (MLH) which is the height to which ground-sourced convective activity persists. The MLH is an important variable with regards to air quality modelling within cities: it dictates the volume into which ground-sourced pollutants can disperse. It is also important for parameterizing the BL in general, as the MLH determines 3 Chapter 1. Introduction the maximum scales of convective activity within the BL. Several techniques have been developed by others for determining the MLH from ceilometer data. Further independent assessment of these techniques will allow for a refinement of these methods and are undertaken in this project. As well, because BL observations via the ceilometer are based on backscat- ter signals from ground-sourced particulate matter (PM), it is possible that the instrument could be used to make assessments of air quality. In that case, a ceilometer would be a useful addition to an air quality monitoring network, as it could provide information not just on ground-level PM con- centrations, but also on elevated pollutant layers. This would prove useful in a location such as Vancouver, as the coastal setting along with the nearby mountain range often leads to a complicated vertical structure of aerosols and pollutants. Therefore, the second portion of this project involves the comparison of ceilometer data to direct PM concentrations. 1.2 Project overview This project is part of a larger observational project: Environmental Predic- tion In Canadian Cities (EPiCC), which is funded by the Canadian Founda- tion of Climate and Atmospheric Sciences. The goal of the EPICC project is to increase our understanding of the urban atmosphere, with applications which include better air quality modelling and numerical weather forecasting within Canadian cities. This project has three main objectives: 1. To assess the ability of Vaisala’s CL31 ceilometer to make detailed and 4 Chapter 1. Introduction continuous BL measurements over the period of several years; 2. To validate the CL31 ceilometer via comparisons with both PM data and meteorological data; 3. To estimate MLHs from the ceilometer data and develop a MLH cli- matology for Vancouver. To this end, Vaisala’s CL31 ceilometer was installed within a suburban area of Vancouver beginning in July 2007. Its location was chosen to be at 49th Ave. and Knight St. in East Vancouver, next to the Sunset Meteoro- logical Tower which has been the site of meteorological and climatological observational work for the last three decades. Although no ancillary obser- vations were taken at the Sunset Tower during this project, future work will provide a comparison of ceilometer data to meteorological and energy flux data measured at the tower. The ceilometer was set to run continuously. Two projects were undertaken in order to test the validity of the ceilome- ter for use as a tool for air quality assessment. The first component of this validation involved the comparison of near-ground ceilometer observations with ground-level PM concentrations measured at nearby air quality moni- toring stations. Eight months of data were used for this assessment, which required the flagging of portions of the ceilometer data in which precipita- tion, fog, or haze are present in order to improve the correlation between the two datasets. The second component involved the comparison of BL struc- ture as interpreted from the ceilometer data to in-situ, co-located vertical tethersonde profiles of both PM and meteorological data. Eight tethersonde flights were made during a single night in August 2007. In order to make 5 Chapter 1. Introduction co-located measurements, the ceilometer was re-located during this night to a park 1 km northwest of the permanent field site. Finally, using compar- isons of PM and ceilometer data, a noise-removal algorithm was developed to remove a persistent and erroneous signal from the ceilometer data. In the second portion of this project, two different algorithms for esti- mating the MLH from ceilometer data were developed. The first technique involves locating the largest gradient in the backscatter signal, which is as- sociated with the top of the convective BL. The second technique estimates the MLH by fitting an ideal backscatter profile to the measured profiles. The results and performance of both of these algorithms were assessed and compared to each other. Before applying these algorithms continuously to the ceilometer data, a flagging algorithm was developed which removes data characterized by fog, precipitation, saturation due to low-level clouds, or poor signal quality. Once this flagging was complete, MLH estimations were made using the ceilometer data collected over an 11 month period. From these estimates, average diurnal trends were calculated for the summer, win- ter and spring. The yearly trend in daily maximum MLHs were assessed as well. This thesis is divided into five chapters: Following this introductory chapter, Chapter 2 will provide a regional description of the LFV, which will include information on the climatology, meteorology, and geography of the region. A description of PM within the LFV including sinks, sources, and transport characteristics will also be included. The second chapter will also provide a description of the field site used for this project and its location, along with a description and technical summary of the CL31 ceilometer. 6 Chapter 1. Introduction The locations of the air quality stations used for analysis in Chapter 4 will also be provided. Following this, Chapter 4 will describe the two ceilometer validation methods undertaken. This chapter will also include a description of a noise-removal algorithm developed for this project as well as a review of previous ceilometer BL observations. Chapter 5 will detail the development and results of the MLH algorithms and will also include basic BL theory and a review of previous MLH estimates within the LFV, followed by a final concluding chapter. 7 Chapter 2 Background and methods 2.1 Geography of the Lower Fraser Valley Vancouver is located on the southwest coast of BC on the western edge of the Lower Fraser Valley (LFV). It is part of a larger metropolitan area with a population of approximately two million. East of Vancouver, the valley is comprised mainly of a flat agricultural flood plain throughout which a number of other urban areas are interspersed, such as Langley, Abbotsford and Chilliwack. Directly north of the city are the Coast Mountains which form a steep northern limit to the LFV; in some locations the range rises to 2000m within 10km of the flood plain. This steep valley wall has a large effect on precipitation within the LFV: precipitation on these slopes is significantly higher than in the middle of the valley (Oke & Hay, 1994). A map of the LFV is shown in Figure 2.1. 8 Chapter 2. Background and methods Figure 2.1: The Lower Fraser Valley with the locations of urban areas outside the Greater Vancouver Regional District. 9 Chapter 2. Background and methods 2.2 Climatology and meteorology of the Lower Fraser Valley During spring, the sub-tropical high moves northward, causing most low- pressure disturbances to be diverted north of Southwest BC. Therefore, summers in Vancouver are usually characterised by a series of sustained, fair weather anti-cyclonic events in which a high pressure ridge forms over the southwest coast. Yet, the influence of the sub-tropical high does not reach much further north then the mid-coast of BC. Consequently, periods of unsettled conditions are not uncommon during the summer due to the collapse of the high-pressure ridge and the passage of low-pressure distur- bances, bringing moist marine air (Oke & Hay, 1994; Mckendry, 1994). During autumn and winter, the jetstream begins to occupy more southerly latitudes again, which steers numerous low-pressure disturbances moving in from the Pacific towards the southwest coast of BC. During the winter, Van- couver experiences an average of one low-pressure disturbance per week (Oke & Hay, 1994). Compounded with the topographical effects of the Coastal Mountain range, this leads to relatively high precipitation throughout au- tumn and winter. Periods of fair weather during autumn and winter are often associated with the southern movement of the continental arctic air- mass across southwest BC (and the accompanying region of high-pressure), or the development of a stationary high pressure ridge which is often referred to as a “blocking high”. Such events often lead to elevated pollution levels due to the development of strongly stable nocturnal surface inversions (Mck- endry, 1994). Despite its mid-latitude location and it being situated on the 10 Chapter 2. Background and methods west coast, wind direction in Vancouver is predominantly from the east and southeast. This is due to topographic forcing: Vancouver is protected from direct westerly winds by both Vancouver Island and the Olympic Penin- sula. Climatological data for the Vancouver International Airport is shown in Table 2.1. 2.3 Mesoscale circulations within the Lower Fraser Valley During anti-cyclonic conditions in the LFV, mesoscale circulations tend to dominate regional winds, as the synoptic flow is generally weak. An im- portant example of such a process within the LFV is the sea/land breeze. During the day, the land is heated more rapidly than the adjacent water (in this case the Georgia Strait) due to the large thermal inertia of the water body. Therefore, enhanced convective activity and positive vertical motion over the land creates a horizontal pressure gradient (assuming hydrostatic conditions) which leads to a surface wind blowing inland from the sea. At night, the situation reverses. The land cools down more rapidly than the strait, the thermal and pressure gradients reverse and a land breeze develops with winds flowing out to sea. The sea-breeze is generally around 3 m/s (the land breeze is usually around 2.2 m/s) and reaches Abbotsford in 60% of cases, while sometimes penetrating as far as Chilliwack (Oke & Hay, 1994). It should be noted that synoptic forcing often opposes the land/sea breeze regime, leading to a weakened meso-scale signal (Steyn, 1998). The mountainous terrain of the LFV causes a variety of flows to develop 11 Chapter 2. Background and methods J a n F eb M a rc h A p ri l M a y J u n e J u ly A u g S ep t O ct N o v D ec Y ea r A v er a g e T em p . (o C ) 3 .3 4 .8 6 .6 9 .2 1 2 .5 1 5 .2 1 7 .5 1 7 .6 1 4 .6 1 0 .1 6 3 .5 1 0 .1 P re ci p it a ti o n (m m ) 1 5 3 .6 1 2 3 .1 1 1 4 .3 8 4 6 7 .9 5 4 .8 3 9 .6 3 9 .1 5 3 .5 1 1 2 .6 1 8 1 1 7 5 .7 1 1 9 9 M o st F re q . W in d D ir ec ti o n E E E E E E E E E E E E E T ab le 2. 1: 19 71 -2 00 0 C lim at e da ta fo r V an co uv er In te rn at io na l A ir po rt (S ou rc e: E nv ir on m en t C an ad a) 12 Chapter 2. Background and methods over the period of a fair-weather day. During the day, air directly above the valley slopes becomes warmer than air within the middle of the valley of a similar altitude. This leads to both an upwards buoyancy force acting on the air next to the slope surface and a horizontal pressure gradient force perpendicular to the along-valley axis. Both these forces lead to what’s referred to as an up-slope or ’anabatic’ flow. These flows have scales on the order of the individual slopes that create them. During the night the opposite process occurs, as increased radiative cooling at the slope surfaces lead to negative buoyancy and down-slope, or ’katabatic’ flows occur which tend to be shallow in depth (less than 20m) (Barry, 1992). These types of mesocale flows have been observed within the LFV on several occasions (Reuten et al., 2005; Banta et al., 1997; Strawbridge & Snyder, 2004a). Moving beyond the scale of individual slopes to the tributary valleys of the LFV in general, larger valley-scale wind structures occur. An upwards- moving ’valley flow’ develops during the day, the mechanism behind which some argue to be identical to the differential-heating process responsible for up-slope flows (Atkinson, 1981; Oke & Hay, 1994). During the night, valley-scale drainage flows occur within the LFV, which can be considered as an accumulation of smaller down-slope flows (Oke & Hay, 1994). This is often referred to as a mountain-flow. As is the case with slope flows, both mountain and valley flows occur along-side reverse flows or ’anti-flows’ aloft due to continuity requirements. The depth of these mountain-flows depend on the size and geometry of the valley as they generally occupy a significant portion of the valley cross-section, while the reversal flows are of a similar depth (Atkinson, 1981). 13 Chapter 2. Background and methods 2.4 Particulate matter within the Lower Fraser Valley Boundary layer (BL) structure is inferred from the ceilometer by analyzing the intensity of backscatter due to ground-sourced aerosols. The majority of suspended fine solid or liquid particles, or aerosols, is comprised of partic- ulate matter (PM) during fair-weather conditions. Therefore, the sources, sinks and transport of PM within the LFV pertains directly to this project. During unsettled, cyclonic events, above-average winds and/or precipitation acts quickly in reducing PM concentrations throughout the LFV. Therefore, the main focus of this section will be upon calm anti-cyclonic conditions. There has been numerous studies of PM within the LFV. The Pacific 2001 and Pacific ’93 observational projects produced extensive information pertaining to PM within the region, which has been analysed in detail by Strawbridge & Snyder (2004b), Strawbridge & Snyder (2004a), Boudries et al. (2004), Brook et al. (2004) and Hoff et al. (1997). McKendry (2000) provided analysis of PM measurements taken over several years at various air-quality monitoring stations within the LFV. Vertical profiles of PM were taken during an anti-cyclonic event by Maletto et al. (2003). Mean PM10 concentrations (i.e., PM with a diameter of less than 10 µm) for a three year study period in the LFV ranged from 12.8 to 17.6 µg m−3, depending on the station. These levels are considerably lower than other urban areas in North America and Europe. Maximum concen- trations occur during the summer at most air quality stations within the LFV (McKendry, 2000). The main source of PM within the LFV is from 14 Chapter 2. Background and methods the transportation sector, the majority of which is road dust, followed by particles from direct vehicular emissions (McKendry, 2000). Consequently, PM concentrations often show a diurnal cycle corresponding to the morn- ing and afternoon periods of enhanced vehicle traffic. Industrial sources are also present. Hoff et al. (1997) emphasise the Fraser River estuary and the Burrard Inlet north of Burnaby as significant sources of industrial PM. In general, however, there are few areas of heavy industry within the LFV, making it a secondary source. Another significant source is emissions from sea-traffic. A major sea-port is located in downtown Vancouver, resulting in high levels of freighter traffic along the coast of the LFV. Emissions from this international fleet are generally unregulated. Due to its west-coast location, it is rare that PM from sources outside the LFV arrive in Vancouver. How- ever, exceptions do occur, including inter-continental transport from major dust storms in Asia and Africa (McKendry et al., 2007). As previously mentioned, BL dynamics have a large effect on PM concen- trations. For instance, the mixed-layer height (MLH) determines the volume into which aerosols can disperse, which in turn affects concentrations. It is often the case that PM concentrations are highest in the morning, corre- sponding to a time of day in which the MLH is low and pollution emissions from morning rush-hour traffic are high. The formation of a nocturnal SBL also has obvious implications with regard to aerosol concentrations: the stagnant, stable layer suppresses dispersion and often leads to significantly elevated pollution levels. With regard to valley-wide transport, a fair weather summer morning in the LFV would likely be characterized by the development of a sea-breeze. 15 Chapter 2. Background and methods The westerly flow would also be enhanced by the developing plain-mountain winds. By mid-day, when the sea-breeze and plain-mountain winds have fully developed, PM originating from car and marine traffic around Van- couver is often blown up-valley (Strawbridge & Snyder, 2004a). Towns and cities in this area such as Chilliwack and Abbotsford often receive high lev- els of pollution during anti-cyclonic summer conditions, while Vancouver is relatively free of pollutants by the end of the afternoon. As well, the sea-breeze circulation may also send pollutants aloft during the day and eventually transport them out towards the Georgia Strait via the elevated return-flow. This type of process is discussed by McKendry & Lundgren (2000). It was mentioned previously that easterly valley-scale drainage flows of- ten develop during the night. In this case, PM that was transported up the valley during the day is transported back down the valley by this noc- turnal drainage flow and possibly out into the Georgia Strait. After the sun rises, the pollutants that were carried out over water may return on a developing sea-breeze, and continue up the valley. Both Hoff et al. (1997) and Strawbridge & Snyder (2004a) observed this type of transport using airborne lidar systems during the Pacific ’93 and Pacific 2001 observational projects, respectively. 2.5 The CL31 ceilometer The Vaisala CL31 Ceilometer is a low-power lidar system originally de- signed to measure cloud heights. By sending rapidly pulsed laser light into 16 Chapter 2. Background and methods the atmosphere and timing the returned laser light backscattered from liquid droplets, information on cloud locations can be obtained. However, the pres- ence of other classes of atmospheric aerosols will also result in a backscatter signal. Therefore, a technique has recently been developed (Rasanen et al., 2000; Münkel et al., 2007; Emeis & Schafer, 2006) in which information on the structure of the BL, which is associated with relatively large concentra- tions of ground-sourced aerosols, is obtained using a ceilometer. A cloudless day of ceilometer data is shown in Figure 2.2 in which the convective BL signal is readily apparent within the first 600 metres. Time of Day H ei gh t a bo ve g ro un d (m ) B ac ks ca tte r ( a.u .) 06:00 12:00 18:00 00:00 200 400 600 800 1000 1200 1400 1600 1800 2000 0 50 100 150 Figure 2.2: An example of ceilometer data. Of note is the convective bound- ary layer development within the first 600m as well as the increasing noise with altitude. 17 Chapter 2. Background and methods 2.5.1 Basic lidar theory Like all lidar systems, the CL31 ceilometer consists of a laser source and receiver. A laser pulse with a duration on the order of nanoseconds is sent through the atmosphere. As the beam travels through the atmosphere, tiny fractions of the light are scattered by aerosols. A small component of this scattered light is directed back to the lidar receiver. In the case of the CL31 ceilometer, the wavelength of the transmitted beam (910 nm) is on the same order as the scattering PM. Therefore, the majority of the scattering is via Mie scattering (Schwiesow, 1986). The timing of the received signal can be transformed into a spatial range, z, and the energy received as a function of range, P (z), is given by the lidar equation: P (z) = Eo· c2 · A z2 ·β(z)· τ2(z) (2.1) where Eo is the effective pulse energy of the transmitted signal, A is the area of the receiver aperture, c is the speed of light, β(z) is the volume backscatter coefficient, and τ2(z) is the two-way attenuation value of the beam for the entire air column given as: τ2(z) = exp { −2 ∫ z 0 σ(z′)dz′ } (2.2) Here σ(z) is the extinction coefficient. By accounting for the instrument- dependent terms in the lidar equation, the ceilometer is able to report the range-corrected attenuated backscatter profile: β(z)· τ2(z), which has units of m −1 sr−1 (Rogers et al., 1997; Münkel et al., 2007). 18 Chapter 2. Background and methods 2.5.2 Description of the CL31 ceilometer Generally, lidar systems are costly and logistically intensive. Safety also be- comes an issue due to the high-power laser operation. However, as the origi- nal function of the ceilometer is to acquire information about cloud heights, the output power can be significantly reduced as the presence of clouds re- sult in a strong, often saturated return signal. Consequently, ceilometers are designed to be an eye-safe, cost-effective lidar system which is able to run continuously with little supervision. In order to achieve a reasonable signal to noise ratio, the reduced power of the ceilometer requires a reduction in the temporal resolution and a larger number of individual profiles must be averaged together. As a consequence, the CL31 ceilometer has a minimum reporting interval of three seconds for the high resolution mode in which an average of 24,000 individual profiles is reported. The CL31 ceilometer is of a single-lens design in which the central area of the lens is used for collimating the outgoing laser beam, while the remaining area focuses the backscattered signal. A schematic of the ceilometer optics is shown in Figure 2.3. With the alternative design of a separate transmitter and receiver, the range-dependent overlap of the receiver and transmitter beam leads to a reduction in efficiency. And the first reportable altitude bin is not at ground-level due to the lack of overlap within the first section of the atmosphere. The single-lens design of the CL31 avoids these issues. Operational details of the CL31 ceilometer are provided in Table 2.2. A detailed description of the ceilometer is presented in Münkel et al. (2007). 19 Chapter 2. Background and methods Figure 2.3: Schematic of Ceilometer optics. Taken from Münkel et al. (2007) Laser source Indium Gallium Arsenide (InGaAs) Diode Laser Centre wavelength 910 ± 10 nm Energy 1.2 µWs ± 20 % Peak power 11 W typical Width, 50% 110 ns typical Repetition rate 10.0 kHz Beam divergence ± 0.4 mrad ×± 0.7 mrad Vertical resolution 10 m (Low-res mode) / 5 m (High-res mode) Max. vertical range 7.7 km (Low-res mode) / 7.5 km (High-res mode) Table 2.2: Vaisala CL31 Ceilometer performance specifications 20 Chapter 2. Background and methods 2.6 Field site description The Ceilometer was installed for continuous operation within the BC Hydro Mainwaring substation at 49th Ave. and Knight St. in East Vancouver (See Figure 2.4) which is also the location of UBC’s Sunset Meteorological Tower. A variety of meteorological data have been gathered at this tower over the last three decades including a full suite of turbulent and radiative fluxes. The ceilometer data can therefore be placed within the context of this previous work. The site is also useful as it is not accessible to the general public. It should be noted that for the validation field-work involving comparisons with tethersonde data the ceilometer was re-located to the Mountain-View Cemetery, one kilometer northwest of the permanent site (see Figure 2.4 for the exact location). 2.6.1 Ceilometer field installation As only the first few kilometres of the atmosphere are of significance to this project, the ceilometer was run using the high resolution mode (5m). The sampling time was set to 15 seconds. Although this is not the highest reso- lution available, it was found that shorter sampling periods would often lead to unreliable performance of the instrument during continuous operation. A 15 second sampling time was also deemed to be well within the timescales of interest to this project; a minimum characteristic convective timescale is on the order of 10 minutes. The ceilometer was also placed in a tilted position with an angle of 14 degrees from the vertical oriented towards the north in order to reduce 21 Chapter 2. Background and methods Figure 2.4: A map of the Greater Vancouver Regional District. The star indicates the ceilometer field site, while points 1 and 2 indicate the Westham Island site and the site used for validation of the ceilomter, respectively. The locations of the GVRD air quality monitoring stations are also shown (the lettered points): A - Kitsilano (T2), B - North Burnaby (T4), C - South Burnaby (T18) and D - Vancouver Airport (T31). 22 Chapter 2. Background and methods background radiation levels and increase the signal to noise ratio. Although the ceilometer was not sampling exactly vertical, the horizontal displacement of the beam was effectively ignored in the context of this project (beyond the adjustment of the altitude values by the cosine of the angle). The horizontal scale of the structures studied in this project was assumed to be much larger than the displacement of the beam from the vertical, which at most is a few hundred metres. 2.7 Processing of ceilometer data A variety of initial data processing steps were applied to the data before any analysis was undertaken. The angle of the beam was compensated for by reducing the reported altitude bins by a factor of the cosine of the beam’s angle from the vertical. For the purpose of this project, the temporal reso- lution of the original data was reduced from 15 seconds to 10 minutes. This resolution was deemed to be sufficient for resolving the diurnal development of the MLH, and to compare to air quality data, which had a resolution of one hour. All of the data processing undertaken for this project was done using the Matlab 7.1 software package (Math Works Inc.). There were a number of erroneous, non-physical elements to the raw ceilometer data which are artifacts of the optical and data processing ele- ments of the ceilometer. For instance, overlaid on the physical signal is a wave-type profile which is consistently present within the ceilometer data. Schafer et al., (2004) report a similar finding when operating the LD-40 Vaisala ceilometer. A technique was developed for removing this noise struc- 23 Chapter 2. Background and methods ture by analysing instances of “clean” data. This noise-removal procedure is discussed in detail in Chapter 3. Because this noise-removal procedure was, to a certain extent, subjective, the backscatter units will be given as “Arbitrary Units” throughout this thesis. Additionally, the bottom 50 metres of the profile consistently exhibited non-physical, semi-periodic fluctuations in the signal (see Figure 2.5B). Over a large enough averaging time, these fluctuations become insignificant. How- ever, for time scales less than an hour these fluctuations have a serious effect on the data. Consequently, the bottom 50m of the ceilometer data was ig- nored for the purpose of MLH estimations. As well, there is a persistent ’zig-zag’ pattern to the signal over several altitude bins (See Figure 2.5A). For the purpose of MLH estimates, the data was filtered vertically with a 35-point running average, which effectively removed this pattern from the bottom 10 altitude bins (50m to 100m). 2.8 Metro Vancouver air quality monitoring stations Hourly averages of PM concentrations from four air quality monitoring sta- tions were used for validation purposes. The locations of these four stations are shown in Figure 2.4. Hourly meteorological data were provided by these stations as well. The concentration measurements were made using tapered element oscillating microbalances (TEOM). Both PM10 and PM2.5 concen- trations were provided (GVRD, 2006). 24 Chapter 2. Background and methods Time of Day (LST) H ei gh t a bo ve g ro un d (m ) B ac ks ca tte r ( A. U. ) 12:00 13:00 14:00 50 100 150 200 50 100 150 −50 0 50 100 150 200 250 300 100 200 300 400 500 600 Backscatter (A.U.) H ei gh t a bo ve g ro un d (m ) A B Figure 2.5: Examples of erroneous artifacts within the raw ceilometer data. A: An example of a persistent zig-zag feature (light-grey area) within the bottom 100m of the data. B: An example of the persistent fluctuations found within the bottom 50m of the ceilometer data. 2.9 Conclusions The ceilometer field site is well positioned to analyse the various BL pro- cesses which have been reviewed here. For instance, it is likely that the 25 Chapter 2. Background and methods ceilometer will be able to observe the effects of the sea-breeze, as well as the formation of elevated pollutant layers at the nearby valley walls or the development of polluted nocturnal drainage flows. These processes, among others, have a significant impact on air quality within Vancouver. There- fore, the field location is ideal for assessing the ability of the ceilometer to determine regional air quality, as it is centrally located within a network or pre-existing air quality monitoring stations. The assessment of the ceilome- ter as both a possible addition to an air quality network, and a tool for determining elevated aerosol structure within the BL will be undertaken in the following chapter. As previously mentioned, determining the MLH within urban areas is crucial for the validation of both air quality and numerical weather fore- casting models. The central location of the ceilometer field site within the Greater Vancouver region is ideal for MLH observations. The process of estimating the MLH from ceilometer data will be detailed in Chapter 4. 26 Chapter 3 Ceilometer validation 3.1 Introduction The use of the ceilometer for observations of the boundary layer (BL) is a rel- atively new technique. Although it was originally designed to measure cloud heights, Rasanen et al. (2000) suggested that the ceilometer was able to re- solve structure within the BL due to the presence of aerosols. Ceilometer observations of the BL have been subsequently pursued by others (Schafer et al., 2004; Münkel et al., 2004; Emeis et al., 2004; de Haij et al., 2006; Wiegner et al., 2006). Validation of the ceilometer for this use has been undertaken by a number of researchers, including comparisons with direct particulate matter (PM) measurements (Münkel et al., 2007) and compar- isons with sodar and wind profiler data (Emeis et al., 2004). Ceilometer data has also been used to estimate the height of the convective BL (Schafer et al., 2004; Eresmaa et al., 2006; Emeis & Schafer, 2006). This technique will be discussed in detail in the next chapter. This chapter begins with a review of the literature involving ceilometer observations of the BL. Following this, two different validation projects are described. Results from a comparison of the ceilometer data with ground- level PM data will be presented. The ceilometer data was found to correlate 27 Chapter 3. Ceilometer validation well with PM data obtained at a number of air quality sites which surround the ceilometer field site. Further validation was undertaken through the comparison of BL structure as reported by the ceilometer with direct mea- surements of PM and meteorological data gathered via tethersonde flights. Results from this work are also presented in the chapter. Finally, a noise- removal algorithm is described. This algorithm was developed to remove a persistent non-physical signal from the ceilometer data. PM data taken during the thethersonde flights were used to optimize this technique. 3.2 Ceilometer observations of the boundary layer Measurements of the height and structure of the BL using traditional lidar systems is a well-developed technique (Cooper & Eichinger, 1994; Menut et al., 1999; Strawbridge & Snyder, 2004b; Schwiesow, 1986). A major goal of these types of measurements is the determination of the mixed-layer height (MLH) as well as the location of elevated aerosol layers. There are a variety of techniques for determining the MLH from lidar data (Seibert et al., 2000; Menut et al., 1999; Davis et al., 2000). These techniques are discussed in de- tail within the following chapter. Additionally, lidar systems have been used to obtain detailed information of turbulence structure within a convectively active BL (Hageli et al., 2000). The use of a ceilometer for observing the BL is a more recent proce- dure. In 2000, the Finnish instrumentation company Vaisala published a conference paper (Rasanen et al., 2000) which stated that the improving 28 Chapter 3. Ceilometer validation signal-to-noise ratio of commercial ceilometers allowed for these instruments to be used to determine the height of the BL through visual interpretation. This technique was reported to be particularly robust within turbid condi- tions. The authors also reported that their new single-lens design allowed for accurate analysis of the first few 100m of the atmosphere, allowing for the development of a shallow, nocturnal stable layer to be identified. This was seen as an advantage over traditional lidar systems which, due to their dual lens design, are not able to make observations within the first few hun- dred metres of the BL. The paper concluded that more accurate BL height measurements could be achieved through the development of a quantitative algorithm. Subsequently, Münkel et al. (2004), Münkel & Rasanen (2004), Emeis et al. (2004) and Schafer et al. (2004) presented additional work detailing BL observations from both the CL31 ceilometer and its predecessors, the CT25k and the LD-40. Münkel et al. (2004) compared backscatter intensity to in situ PM10 measurements. Further results from this study were published by Münkel et al. (2007). In this study, PM monitors measuring both PM10 and PM2.5 were placed 20 metres above a CT25K ceilometer and run at a 30min resolution. A year of observations were obtained from all three instruments. After removing instances of wet haze, fog and precipitation, the remaining 5736 datapoints for PM10 concentrations showed a significant correlation with an average of the first two altitude bins (15m and 30m) of ceilometer data(R2=0.8351). Münkel et al. (2004) also presented comparisons of ceilometer profiles with sodar and wind-profiler data. This work was presented in greater detail 29 Chapter 3. Ceilometer validation by Emeis et al. (2004). There was convincing correlation between the BL structure reported by all three instruments on a variety of days. This work concluded that the potential of the ceilometer to ascertain BL structure was on par with single-frequency lidar systems. Schafer et al. (2004) detailed a MLH algorithm which was applied to data from the dual lens LD-40 ceilometer (7.5m vertical resolution). The algorithm was based on the identification of a maximum negative gradient in the backscatter signal and reported MLH values every ten minutes. Details of this algorithm are presented in the next chapter. The results of the algorithm were compared to measurements taken by Sodar located 53km from the ceilometer. Of note in this study is the simultaneous identification of a stable BL by both instruments during a strong katabatic flow event due to high PM concentrations. A similar study is presented by Emeis & Schafer (2006), again using an LD-40 ceilometer. Eresmaa et al. (2006) applied an alternative MLH algorithm to data from a CT25K ceilometer (15m vertical resolution) by fitting an ideal backscatter profile to the observed data, and determining the MLH from the fit (details of this technique will also be discussed in the next chapter). The algorithm reported MLHs every 30 minutes. This technique was originally applied to lidar profiles by Steyn et al. (1999). They found that during clear-sky conditions the MLH values were in good agreement with sounding data, barring any days characterized by low aerosol concentrations. Similar to Schafer et al. (2004) and Emeis & Schafer (2006), the authors pointed out the issues inherent in applying a MLH algorithm to the nocturnal situation. Münkel et al. (2007) applied the ideal profile technique as well as the gra- 30 Chapter 3. Ceilometer validation dient method to a year of data collected by a CT25K ceilometer in Helsinki, Finland. The MLH values determined from both the ideal-profile technique and nearby radiosounde temperature profiles were compared. They found a significant agreement between the MLH estimated by the ideal-profile method and estimates from radiosonde profiles after rejecting low-signal cases (ρ = 0.9 for convective situations). Having the advantage of a previously established network of around 20 LD-40 ceilometers, de Haij et al. (2006) applied a unique (with regards to ceilometer data) MLH algorithm to six years of ceilometer data using wavelet analysis. They also developed a flagging algorithm which removed all data associated with precipitation, fog, or low signal strength. Wiegner et al. (2006) and Zephoris et al. (2005) deployed ceilometers as part of larger observational projects within urban areas in which BL obser- vations from the ceilometer were compared to and validated by a variety of remote-sensing and radiosonde measurements. 3.3 Ceilometer validation via measurements of ground-level particulate matter concentrations 3.3.1 Overview As this project assumes that PM is the main source of the ceilometer backscatter signal (other than water droplets), it is important to compare the ceilometer backscatter signal to direct PM measurements (as performed 31 Chapter 3. Ceilometer validation by Münkel et al. (2007)). Such a dataset was provided by the Metro Van- couver Air Quality Monitoring Network (see Section 2.8 for details). This section presents a comparison of ceilometer data with PM data from the four closest stations, all of which are within 12km of the ceilometer field site. The location of these four air quality stations, Kitsilano (T2), South Burnaby (T18), Vancouver International Airport (T31) and North Burnaby (T4) are shown in Figure 2.4. In this analysis only PM2.5 data was used. On a city-wide scale, PM2.5 exhibits much smaller spatial variation than PM10 which is often signifi- cantly localized due to increased deposition rates. Therefore, PM2.5 data exhibit more significant levels of agreement between individual stations and with the ceilometer data. Additionally, the ceilometer is more sensitive to the smaller PM2.5 particles due to the slight wavelength dependence of mie scattering. Summary data for PM10 is given as a means of comparison. 3.3.2 Methods Initial data processing Observations from July 5, 2007 to February 29, 2008 were used for this analysis. This data range was divided into individual months in order to analyze seasonal effects. As an initial step, a linear interpolation was used in order to achieve a common time-step for both the ceilometer and PM data. Secondly, the data was smoothed using a running mean. The resolution of the common time-step and the filter window-size was optimized in order maximize the correlation. A one hour resolution (the highest resolution 32 Chapter 3. Ceilometer validation available) along with a 5-point (5 hour) running mean filter were found to maximize the agreement between the two datasets. The ground-level PM data were compared to a single altitude bin from the ceilometer data. Due to erroneous ceilometer noise within the first tens of metres, the lowest altitude bin, 5m, was not used. The 45m altitude bin was used instead, as this is the lowest bin unaffected by the noise issue. An initial vertical smoothing of the ceilometer data was found to have a negligible effect and was therefore not used in the following analysis. Flagging of potentially erroneous data The largest potential source of error in this analysis is the accidental inclu- sion of ceilometer signals resulting from the presence of water droplets such as fog, haze or precipitation. As direct measurements of humidity and pre- cipitation were not available at the ceilometer field site, rainfall data from the GVRD air quality stations were used to create a “rainfall flag.” Data recorded during periods of rain (observed at any of the four stations) were removed from the dataset. Additionally, a “post-rain” period was also ig- nored, as the ceilometer seems to register remnant amounts of condensation in the BL after precipitation has ended. The length of this ”post-rain” pe- riod was adjusted for optimal agreement between datasets. The treatment of periods of precipitation is discussed in detail below. Additionally, a “day-time” flag was created so that comparisons could be done for either the daytime situation, the nocturnal situation, or for all 24 hours. Sunset and sunrise times were used to delineate “day” from “night.” An option for defining a sunset or sunrise “offset” was also included: this 33 Chapter 3. Ceilometer validation allowed for a shortening or lengthening of what was to be defined as the “daytime” beyond simply the sunset and sunrise time. 3.3.3 Results Tables 3.1 and 3.2 provide a summary of correlation values. Once erroneous data were removed from the dataset, the 45m ceilometer backscatter signal showed a significant correlation with ground-level PM data. Indeed, during the summer months the level of correlation is often on-par with inter-station correlation values. T18 T31 T2 T4 24 hours 0.43 0.24 0.31 0.44 Daytime 0.52 0.48 0.54 0.52 Night Time 0.51 0.30 0.29 0.40 Daytime with Precipitation -0.01 0.01 -0.01 -0.03 Table 3.1: Pearson correlation coefficients for ground-level PM2.5 concentra- tions and 45m ceilometer backscatter intensity values. T18 T31 T2 T4 24 hours 0.40 0.28 0.31 0.35 Daytime 0.44 0.48 0.45 0.43 Night Time 0.47 0.29 0.27 0.34 Daytime with Precipitation -0.09 0.05 -0.06 -0.11 Table 3.2: Pearson correlation coefficients for ground-level PM10 concentra- tions and 45m ceilometer backscatter intensity values. The most significant errors involved ceilometer data associated with pe- riods of rain. The saturated ceilometer signal resulting from such conditions led to a dramatic reduction in any physically-based correlation between the two datasets. The rainfall flag was able to remove most but not all of these 34 Chapter 3. Ceilometer validation outliers. A number of large backscatter values remained which had a sig- nificant effect on the correlation values. By setting a maximum allowable backscatter threshold value of 200 a.u, these remaining outliers were re- moved. Although this process is not based on any physical evidence, it was felt that these data could be ignored as they only represented a small per- centage of the entire dataset (≤ 1%), but had a significant negative effect on the correlation. Additionally, the correlation was significantly affected when the dura- tion of the“post-rain” period was adjusted so as to avoid remnant moisture after a rainfall. In Figure 3.1, the change in correlation levels as the “post- rain” period is increased is shown for the entire time range (July-February). The first local maximum occurs just after one day. This suggests that sig- nificant moisture remains within the BL for many hours after the rainfall has stopped. The most significant correlation occurs at six days. Once the “post-rain” period becomes this long, it is likely that data from only the driest periods have been isolated: situations in which PM concentration has been building up over several days, resulting in a strong agreement between the ceilometer and PM datasets. This effect is the most pronounced during the winter months, and is not present at all during the summer months: The increased summertime radiation removes moisture from the BL quicker than during the winter. In the following analysis a post-rain period of 19 hours is taken in order to retain good correlation as well as a large enough sample size. The flagging of data observed during either the daytime or night-time also had a significant effect on the agreement of the two datasets. By re- 35 Chapter 3. Ceilometer validation 0 2 4 6 8 10 0.4 0.45 0.5 0.55 0.6 0.65 Length of post−rain Period (days) Ce il/ PM C or re la tio n (Ju ly to Fe b.) Figure 3.1: Correlation between 45 metre ceilometer backscatter and PM concentrations as the “post-rain” period in lengthened. PM data for July 2007 to February 2008 taken from station T18 (South Burnaby). moving all nocturnal data, the average correlation was improved by 49%. As well, the average correlation of solely the nocturnal situation is 73% of the daytime data. As this analysis uses the 45 m AGL ceilometer data, it is likely that the ceilometer-PM agreement would be strongest during the presence of a vertically homogeneous mixed-layer (ML). This would ex- plain the increased agreement during the convective daytime situation when compared to the stable nocturnal situation. Indeed, the correlation levels between the four stations themselves did not change significantly whether one considers either the daytime or nocturnal situation, suggesting that it is 36 Chapter 3. Ceilometer validation the vertical homogeneity of PM that increases the daytime correlation with the ceilometer data. A two hour sunrise and sunset offset was chosen when distinguishing the day from night; the four hour periods centred around sunrise and sunset were removed from the dataset. This resulted in a better correlation than not using an offset. Scatter plots for all four stations and for the entire time range are shown in Figure 3.2. It is apparent that the daytime data exhibit much less scatter than the nocturnal data. This is especially the case for stations T31 and T2. The correlation of daytime and night-time data throughout the obser- vational period (Station T18) is shown in Figure 3.3 as an example of the seasonal trend. During Summer and Autumn, the correlation between the station and the ceilometer falls within the range of correlation values be- tween the different ground-level stations. The daytime correlation decreases significantly during the Winter. This is likely due to increased moisture in the BL as well as a decreasing number of dry days available for analysis. There are two-week data gaps in both January and February which led to much smaller sample size. This may also be the reason for the anomalous minimum inter-station correlation seen during January in Figure 3.3. 37 Chapter 3. Ceilometer validation 0 5 10 15 20 25 South Burnaby (T18) PM 2. 5 Co nc . (µ g m − 3 ) 0 5 10 15 20 25 Vancouver Int. Airport (T31) −100 0 100 200 0 5 10 15 20 25 Kitsilano (T2) Backscatter (a.u.) PM 2. 5 Co nc . (µ g m − 3 ) −100 0 100 200 0 5 10 15 20 25 North Burnaby (T4) Backscatter (a.u.) Figure 3.2: Scatter plots comparing 45 metre ceilometer backscatter levels to ground level PM concentrations for all four GVRD air quality stations. Data is shown for both the daytime (Red) and nocturnal situation (Blue). 38 Chapter 3. Ceilometer validation Jul Aug Sep Oct Nov Dec Jan Feb 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 Ce ilo m et er /P M C or re la tio n Inter−Station Range Daytime Data Nocturnal Data Figure 3.3: Daytime and nocturnal correlation between 45 metre ceilome- ter backscatter intensity and PM concentrations for July 2007 to February 2008 (Data from Station T18 used). The shaded area indicates the range of daytime correlation values present between the four air quality stations themselves. Of note is the significant correlation of night-time data during October and November which is in contrast to the low levels of night-time agree- ment during the summer months. During these months, the development of strong surface inversions along with low winds may have led to elevated PM concentrations throughout the city. It is likely that these turbid surface layers were observed by both the 45m AGL ceilometer signal as well as the ground-level PM observations. 39 Chapter 3. Ceilometer validation 3.4 Ceilometer validation via direct tethersonde observations 3.4.1 Overview In order to validate the aerosol structure within the BL as reported by the ceilometer, 12 hours of tethersonde ascents were undertaken on the night of August 14-15, 2007 in order to measure both size-segregated PM concentra- tions as well as a full suite of meteorological data. This work was performed during fair-weather anti-cyclonic conditions that were characterized by an absence of condensation within the BL. There- fore, the ceilometer signal was assumed to be due almost entirely to the pres- ence of PM. Despite the low concentrations of PM observed, a significant correlation (ρ = 0.67) with backscatter intensity was found. 3.4.2 Methods Field setup The tethersonde ascents were undertaken alongside the ceilometer within the Mountain View Cemetery located a few kilometers northwest of the ceilome- ter’s operational field site (See Figure 2.4). The ceilometer was run with the same settings as for the continuous observations: a 15 second sample time along with a 5m vertical resolution. 40 Chapter 3. Ceilometer validation Aerosol spectrometer PMmeasurements were made by an optical aerosol spectrometer, theGRIMM 1.108 (Labortechnik Ltd.). Deployment of this instrument on a tethered balloon similar to this study is described by Maletto et al. (2003) and McK- endry et al. (2004). This instrument measures particulate diameters via op- tical scattering and demarcates these measurements into 15 volume classes ranging from 0.3 to 20 µg (assuming a spherical shape). In order to ob- tain mass concentrations, a constant density value is assumed. This series of assumptions requires calibration of the GRIMM with independent PM measurements. However, calibration of PM mass concentrations is difficult due to a large number of confounding factors inherent in any form of PM measurements (Charron et al., 2004; Maletto et al., 2003). Yet it should be noted that absolute measurements of PM concentrations are less impor- tant in this study: analysis of relative changes in concentrations is sufficient for distinguishing physical structure within the ceilometer signal from noise. For this reason, calibration against other instruments was not performed in this study. Therefore, a calibration factor of 1.0 was used which was sug- gested by Maletto et al. (2003) and McKendry et al. (2004) as a reasonable approach for a coastal urban setting. Tethered balloon The GRIMM spectrometer was slung half a metre below a 5m3 helium-filled kytoon. Half a metre below this a Kestral 4500 meteorological package was hung. The Kestral 4500 provided a full suite of meteorological data, includ- 41 Chapter 3. Ceilometer validation ing wind speed, humidity, temperature and altitude derived from pressure measurements. The kytoon used is designed to orient itself to the prevailing wind direction. Information from the internal compass within the Kestral 4500 was therefore used to obtain wind direction information. Eight balloon flights were performed between 1815 PST, August 14 and 0715 PST, August 15 resulting in a total of 16 profiles. Both instruments were flown on all flights except for Flight 2 during which the GRIMM was not used. The temporal and vertical location of these flights is shown in Figure 3.4, overlaid on top of the ceilometer data. The Kestral 4500 and the GRIMM spectrometer were run at a ten second and six second sampling rate respectively. Timing information from both instruments was used to create vertical profiles of PM concentrations. An electrical winch controlled the vertical motion of the balloon, with a typical flight lasting 45 minutes and reaching between 200-300m. This resulted in an average vertical resolu- tion of 2.4m for the meteorological data and 1.5m for the PM data. As the altitude was calculated via pressure measurements, there were slight fluctu- ations in the altitude reading at ground level. To compensate, the altitude for each flight was shifted so that the minimum altitude value was set to zero metres above ground level. In order to remove noise which is character- istic of tethered balloon ascents in general, the resulting profiles were then smoothed using a 10-point running average. 42 Chapter 3. Ceilometer validation 3.4.3 Meteorology and ceilometer observations Height Above Ground (m) Backscatter (A.U.) 16 :0 0 18 :0 0 20 :0 0 22 :0 0 00 :0 0 02 :0 0 04 :0 0 06 :0 0 0 10 0 20 0 30 0 40 0 50 0 60 0 16 :0 0 18 :0 0 20 :0 0 22 :0 0 00 :0 0 02 :0 0 04 :0 0 06 :0 0 24681012 PM 2.5 (µgm −3 ) 10152025303540 8 7 6 5 4 3 2 1 Figure 3.4: Ceilometer observations from 1600, August 14 to 0800, Au- gust 15 (PST) Tethersonde flight locations are also shown along with flight numbers. PM2.5 measurements from the South Burnaby air quality station (T18) are shown in the lower panel. Sunrise and sunset times are indicated by the vertical dashed lines. 43 Chapter 3. Ceilometer validation The validation study (August 14-15, 2007) was performed during a fair- weather anti-cyclonic event in which a ridge of high pressure developed over the southwest coast of BC. This development was preceded by the passage of a depression associated with unsettled weather conditions and relatively low PM concentrations. Temperatures at the Vancouver International Airport (YVR) reached above-average values of 23oC and 25oC on the 14th and 15th, respectively, with west to southwest daytime winds transitioning to northeast winds at night. The BL structure as reported by the ceilometer is shown in Figure 3.4. The initial collapse of the convective BL near 1800 PST is clearly suggested by the ceilometer data. A three hour window of ceilometer data after the decay of convective turbulence is shown in Figure 3.5. The meteorological data presented in this figure correlates well with the ceilometer data. At around 225 m AGL, there is a distinct transition in the ceilometer data which is associated with a distinct temperature inversion, as well as a decrease in humidity with height. This suggests that the ceilometer was able to observe the development of a layer characterized by the presence of ground-sourced PM, above which cleaner, drier air was present. After the decay of turbulent activity, the development of a stable BL throughout the night was observed and is shown in Figure 3.6. By sunrise, this stable layer grew to a depth of 200-250m. This is a significant result, as urban environments are often characterized by a neutral or even unstable nocturnal BL (Oke, 1995). However, this is clearly not the case here. Al- though the tethersonde flights were performed within a park, a large portion of the observed profiles was likely indicative of air which was in equilibrium 44 Chapter 3. Ceilometer validation Lo ca l T im e (P ST ) 19 :0 0 22 :0 0 29 0 29 5 30 0 05010 0 15 0 20 0 25 0 30 0 35 0 40 0 45 0 Po te nt ia l Te m p (K ) Height Above Ground (m) Backscatter (A.U.) 0 2 4 W in d Sp ee d (m s−1 ) 0 18 0 36 0 W in d D ir . (D eg .) 5 10 15 D ew −P oi nt Te m p (o C ) 101520253035404550 Figure 3.5: Meteorological and ceilometer observations for Flight 2. (Ascent: solid line, Descent: dotted line). The horizontal dotted line indicates the estimated altitude of the inversion 45 Chapter 3. Ceilometer validation Figure 3.6: Potential temperature profiles from selected tethered balloon soundings during case study. with the surrounding urban area rather then the cooler, moist park surface. Garrett (1992) presents a formulation for the growth of a stable internal BL as air is advected across a cooler surface such as a park. A high estimate suggests that only the first few metres would be in equilibrium with the park surface, given the rate of advection from the surrounding urban areas. 3.4.4 Relation between particulate matter concentrations and backscatter In situ measurements of PM concentrations show agreement with the BL structure as reported by the ceilometer. There were a variety of instance in 46 Chapter 3. Ceilometer validation which areas of elevated PM concentrations were also associated with elevated backscatter levels. The most obvious agreement between the ceilometer and the PM profiles is the simultaneous observation of an elevated aerosol layer on the morning of the 15th by both the ceilometer and the GRIMM. The appearance of this layer between 100 and 200m is plainly seen in Figure 3.4 beginning just before 0600 PST. This development is associated with the onset of a layer of west-by-southwest winds, suggesting that the layer is advective in origin. Figure 3.7 provides a comparison of two PM10 profiles with the ceilometer data. The presence of the layer is obvious in both instances. Although there is a slight discrepancy in the vertical location of the layer, this is likely explained by the fact that the two instruments were observing slightly different sampling volumes due to their horizontal separation. Using the 14 profiles in which PM measurements were made, the cor- relation between PM concentrations and backscatter levels was calculated. For each time step the ceilometer signal from the altitude bin nearest to the altitude of the balloon was taken in order to create a corresponding series of ceilometer signal intensities to be compared to the GRIMM data. Be- fore this selection, the ceilometer data was filtered using a 11-point vertical (55m) and temporal (165 seconds) running average. Measurements taken below 50 metres were ignored during this analysis due to the erroneous ceilometer signals present within that altitude range. Additionally, a large but temporary spike in the PM measurements recorded at around 130m AGL during the first flight has been ignored. Although the source of this peak is not known, the dramatic concentration increase of an 47 Chapter 3. Ceilometer validation Lo ca l Ti m e (P ST ) 05 :0 0 07 :0 0 0 1 2 3 W in d Sp ee d (m s−1 ) 28 0 29 0 30 0 05010 0 15 0 20 0 25 0 30 0 35 0 P ot en tia l Te m p. (K ) Height Above Ground (m) Backscatter (A.U.) 0 2 4 6 PM c on c. (µ gm − 3 ) 8 10 12 D ew −P oi nt Te m p (o C ) 0 18 0 36 0 W in d D ir . (D eg .) 101520253035404550 Figure 3.7: Meteorological, PM2.5 and ceilometer observations for flight 7. (Ascent: solid line, Descent: dotted line) 48 Chapter 3. Ceilometer validation order of magnitude over only three metres of ascent was taken to be non- physical and therefore these 22 data points (126-144m AGL) were ignored. The resulting scatter-plots for PM10 (ρ = 0.67 ) and PM1.0 (ρ = 0.82) concentrations are shown in Figure 3.8. In both plots, two linear regression lines are shown, one which includes all of the ceilometer data (blue line), and a second relationship (red line) that accounts for only the positive data. These positive values are the portion of the ceilometer signal which is deemed by the noise-removal algorithm to be physically-based. It is apparent that the noise-removal algorithm improves the correlation between the ceilometer and PM observations. This is discussed in more detail in Section 3.5. The results reported here refer to the positive, physically based data. These data suggest a significant, positive and linear relationship be- tween PM concentrations and ceilometer backscatter intensities. This agrees with the results published by Münkel et al. (2007) who reported a simi- lar linear relationship between ceilometer backscatter intensity and PM10 concentration measurements taken 30m above the ceilometer. However, the slope of the regression line calculated by Münkel et al. (2007) (0.2 µgm−3/10−5km−1sr−1) is an order of magnitude smaller than the slope in Figure 3.8 (1.1 µgm−3/10−5 km−1sr−1). This may be explained by the different PM measurement techniques used (gravimetric versus optical). It could also be a result of the smaller range of PM10 concentrations measured during this study: maximum PM10 concentrations measured by Münkel et al. (2007) were four times higher than during the tethersonde flights. 49 Chapter 3. Ceilometer validation −15 −10 −5 0 5 10 15 20 25 0 1 2 3 4 Backscatter (A.U.) PM 1. 0 Co nc . (µ g m − 3) −15 −10 −5 0 5 10 15 20 25 0 5 10 15 20 25 30 35 Backscatter (A.U.) PM 10 Co nc . (µ g m − 3) Figure 3.8: Scatterplots of PM1.0 (top) and PM10 (bottom) and ceilome- ter backscatter for all vertical profiles conducted with the GRIMM aerosol spectrometer (14 Profiles in total). Regression lines are shown for the entire dataset (Blue) as well as for only the data deemed by the noise-removal technique to be physically based (Red). 50 Chapter 3. Ceilometer validation As previously mentioned, it is possible that the ceilometer and GRIMM were observing different sampling volumes. This could explain a large amount of the scatter seen in Figure 3.8. For example, the wave-like vertical motion of the aerosol layer which was advected over the field site on the morning of August 15 could have led to a situation in which the GRIMM was located within the layer, while the ceilometer was measuring clean air at a similar altitude outside of the layer. It is apparent that the regression fit is significantly better for PM1.0 than for PM10. Below PM2.5 the correlation is fairly consistent. Above this the correlation strength falls off quickly. This is likely due to two major rea- sons. First, according to mie-scattering theory, the backscatter coefficient, β, is largest for particle sizes similar to the wavelength of the incident light (910nm in this case). Secondly, the variability in concentration reported by the GRIMM increases significantly for the size bins above two microns; the largest bin sizes are characterized by large and sudden fluctuations in concentration values which may lead to a lower correlation strength when compared to the backscatter values. 3.5 Development and optimization of a noise removal technique 3.5.1 Overview This study revealed a major issue with ceilometer data: a permanent, non- physical wave-type signal superimposed on the actual signal. Schafer et al. 51 Chapter 3. Ceilometer validation (2004) speak to this issue, stating that the wave structure is not dependent on temperature or signal amplitude. They suggest that an average profile should be calculated for the time period in question and that the wave pat- tern present in this average should be removed from the individual profiles. An example of this wave-type noise is shown in Figure 3.9 (A and B). It H ei gh t a bo ve g ro un d (m ) 200 400 600 800 1000 1200 1400 H ei gh t a bo ve g ro un d (m ) 18:00 19:00 20:00 21:00 22:00 200 400 600 800 1000 1200 1400 20 40 60 Backscatter (a.u.) Figure 3.9: Five hours of ceilometer data before noise removal is performed (top left), and the average profile of the five hour dataset (top right). The same data is shown after noise removal (bottom left), with the resulting average profile (bottom right). 52 Chapter 3. Ceilometer validation can be seen that this signal is persistent, regardless of the physical structure present. It also leads to a significant over-estimation of the aerosol content within the first 150m, which may lead to significant errors in determining the MLH, especially in the case of a low convective BL height. 3.5.2 Description of technique In order to correct for this issue, an example of this noise profile had to be extracted from the entire dataset. It was assumed that the erroneous signal is constant throughout the entire ceilometer dataset, and that instances of very low signal amplitudes consist entirely of this erroneous signal. There- fore, just over two months of data (July 5 - Sept 14) were analyzed in order to identify the “cleanest” cases in which it could be reasonably assumed that no physical structure was present within the signal. To do this, a mean signal strength for each time-step was calculated. A percentile threshold was then defined which would delineate the “clean” instances from cases in which physically-based information was present. By taking an average of the entire set of “clean” signals, a single noise profile was obtained. This noise profile was then subtracted from the ceilometer signal in question and any negative signal values were deemed to be non-physical. An example of the results of this technique is shown in Figure 3.9. 3.5.3 Optimization of technique As the night of the tethersonde measurements represents a case of low PM concentrations, it presents an opportunity to optimize the noise-removal technique. Due to the low ceilometer signal, a portion of the ceilometer sig- 53 Chapter 3. Ceilometer validation nal reported during the night of August 14-15 is likely non-physical. There- fore, the “clean-percentile” threshold was adjusted so as to maximize the correlation between the ceilometer signal and the direct PM measurements made during the tethersonde flights. The “clean-percentile” threshold was adjusted over a range of 0% to 30%. With each new threshold value the ceilometer data were compared to the direct PM measurements and a Pear- son correlation coefficient was calculated. The noise-removal procedure has a significant effect on the correlation between the ceilometer signal and the direct PM measurements. By using a clean percentile threshold of 1.5%, the correlation improves by 61%, 24% and 20% for PM10, PM2.5 and PM1.0 respectively. However, as the clean percentile threshold is increased beyond 1.5%, the correlation does not sig- nificantly improve and, in the case of PM2.5 and PM1.0, decreases slightly. Beyond 15% the correlation falls off significantly as an increasing amount of the physically-based ceilometer signal is removed. This suggests that a 1.5% clean percentile threshold is optimal. Figure 3.8 shows the effects of the noise-removal on the correlation between the ceilometer and tethersonde PM data. For PM1.0 the positive data shows a much more linear trend when compared to the entire dataset. This result was confirmed by a qualitative examination of the ceilometer data for the night of August 14-15, 2007. Resulting ceilometer data from different percentile thresholds were analyzed. At a 1.5% threshold, a large portion of the structure that was qualitatively assessed as being significant and physically based was included in the resulting signal. As the thresh- old was increased, larger portions of this physical structure were removed. 54 Chapter 3. Ceilometer validation Additionally, below a 1.5% threshold, the wave structure begins to become significant. 3.6 Conclusions Validation of the ceilometer as an instrument for observing the BL has been presented in this chapter. Analysis of data from both the tethersonde flights and surrounding air quality stations suggests that during periods without precipitation or significant condensation, ceilometer backscatter intensity provides a linear proxy for PM data. This relationship is the most significant for PM sizes on the order of one micron, which correlates with the ceilometer laser wavelength. Results from Section 3.3 show that the ceilometer data are closely re- lated to regional (i.e., city wide) PM concentrations. However, the presence of water droplets within the BL must be accounted for as it acts to saturate the ceilometer signal. This correlation is strongest during dry conditions during the Summer in which there has be significant accumulation of PM within the region. These types of conditions often lead to significant pollu- tion events, which suggests that the ceilometer may be a useful addition to a regional air quality network. Not only does it provide a reliable indicator of PM concentrations, but it also provides information about elevated lay- ers and vertical movement of pollutants which would complement existing ground-level information. McKendry & Lundgren (2000) suggest that ver- tical pollutant information is valuable in coastal and mountainous regions such as Vancouver where a variety of processes lead to vertical transport of 55 Chapter 3. Ceilometer validation pollutants. Additionally, data collected during the tethersonde flights show that the ceilometer is able to observe aerosol structure even in relatively clean con- ditions. The BL structure as reported by the ceilometer was confirmed by both the meteorology and the PM measurements during an observational period which was charactized by a mean PM10 concentration of 10.6µgm−3. These results also confirm the direct relationship between backscatter in- tensity and PM concentration which was reported by Münkel et al. (2007). Although the relationship found (see Figure 3.8) was an order of magni- tude higher than that found by Münkel et al. (2007), this is likely due to a difference in both the ceilometer used (the CT25K vs. the CL31) and the technique used for measuring the PM concentrations. Further analysis of this relationship may allow for quantitative assessments of air quality using ceilometer observations. Although previous projects have undertaken the validation of ceilome- ter data, they have all involved comparisons of either other remote sensing techniques (lidar, wind-profiler sodar, etc.), stationary PM concentrations, or radiosonde measurements which were often located kilometers away from the ceilometer. The tethersonde results represent the first time in which vertical profiles of in situ, co-located, high resolution measurements of PM concentrations and a full suite of meteorological variables have been used for validation. Therefore, the results from this chapter represent a signifi- cant addition to the growing body of evidence supporting the usefulness of ceilometers for making detailed, cost-effective measurements of the BL. The results described in this chapter allow for a confident application of 56 Chapter 3. Ceilometer validation the ceilometer data towards the estimation of MLHs, which is the subject of the following chapter. Yet, it should be mentioned that the above analysis does not provide a comparison of ceilometer data with daytime thermal structure within the convective BL. That the ceilometer is able to suggest such structure is a significant assumption with regards to the estimation of MLHs. However, several previous studies (Eresmaa et al., 2006; Münkel et al., 2007) have shown a strong agreement between the MLH estimated from the ceilometer data and the height of the elevated thermal inversion layer. This issue is discussed in detail in the following chapter. 57 Chapter 4 Mixed-layer height estimations 4.1 Introduction When describing the convective boundary layer (CBL), one of the most sig- nificant variables is the mixed-layer height (MLH). The MLH is associated with the top of the convective boundary layer and the height to which con- vective turbulence persists into the atmosphere. Below the MLH the bound- ary layer is generally well-mixed. The ability to estimate or model the MLH is important as it determines not only the maximum length scales of convec- tive turbulence in the boundary layer, but it also dictates the volume into which ground-sourced pollutants can disperse. In most cases, the CBL is characterised by high aerosol concentrations relative to the FA above. This fact allows for the identification of the MLH from ceilometer backscatter profiles by identifying the distinct gradient in backscatter intensity. Estimates of the MLH using ceilometer data have been undertaken on several occasions. Initially, Rasanen et al. (2000) noted that the MLH could be visually estimated from ceilometer data. Following this, Schafer 58 Chapter 4. Mixed-layer height estimations et al. (2004) and Emeis & Schafer (2006) made estimates of the MLH by identifying the altitude with the minimum backscatter gradient. Eresmaa et al. (2006) estimated the MLH by fitting an ideal-profile to the measured backscatter profile. In this chapter MLH estimates will be made using both these methods. There have been numerous estimates of the MLH within Vancouver and the surrounding Lower Fraser Valley (LFV)(Strawbridge & Snyder, 2004b; Hageli et al., 2000; Snyder & Strawbridge, 2004; Steyn & Wallis, 1987). The most pertinent observations for this project are the sodar observations reported by Steyn (1980) which were made at the same field location as this project: the Sunset meteorological tower. The estimates from these papers will be used to compare to the MLH estimates made in this chapter. Vancouver’s coastal location has a large effect on the height of the con- vective boundary layer. On-shore breezes, which, due the presence of the fair weather sea breeze, are a significant meteorological feature during con- vective situations, act to suppress the height of the convective boundary layer through the development of a thermal internal boundary layer. This effect has been modelled by various authors (Steyn & Oke, 1982; Batch- varova et al., 1999; Cai & Steyn, 2000). MLH estimates made here will be used to analyse the effects of the sea breeze on the height of the convective boundary layer during the summer. The first portion of this chapter will provide an introduction to basic convective boundary layer theory as well as a review of the various MLH estimates made within the LFV. The second portion of this chapter will present the results of the MLH estimations. The objectives of this chapter 59 Chapter 4. Mixed-layer height estimations are three-fold. First, a presentation of the two MLH estimation algorithms will be presented. The results from these two algorithms will be assessed and compared. Secondly, a flagging algorithm will be presented which was de- veloped in order to isolate convective situations so that the MLH algorithms will not be applied to non-applicable data. The results of the ideal-profile method will then be presented and assessed. This includes a comparison of MLH estimates for summer, autumn, and winter, as well as the seasonal trend in MLH measured during fair weather, clear days. Finally, the effects of the TIBL will be discussed. 4.2 Boundary layer theory 4.2.1 Classification of the atmospheric boundary layer Stull (1988) defines the Boundary Layer as “that part of the troposphere that is directly influenced by the presence of the earth’s surface, and responds to surface forcing with a timescale of about an hour or less.” The depth of the BL can be anywhere from tens of metres to three kilometers. Above the BL is the free atmosphere (FA), which exhibits variations on much longer time scales then the BL and is only indirectly affected by the earth’s surface. The classic diurnal cycle of the BL during fair weather conditions is shown in Figure 4.1. Just after sunrise, convective activity is initiated by solar surface heating. The influence of the convective turbulence increases in height as the day progresses, creating a well-mixed layer in which the constituents and conserved meteorological variables are near constant with altitude. This layer is often referred to as the mixed-layer (ML), the height 60 Chapter 4. Mixed-layer height estimations Figure 4.1: Diurnal development of an idealized boundary layer during fair weather conditions of which is referred to as the mixed-layer height (MLH). Above the ML is the entrainment zone (EZ) in which there is a transition from the unstable mixed layer to the stable FA. The stability of the FA effectively places a vertical limit on the convective turbulence of the ML. However, as individual thermal plumes impinge on this stable air, the slight overshooting of the thermal results in downwelling of warmer air due to continuity requirements, which results in a downward heat flux throughout the EZ. In this way, ambient stable air above is ’entrained’ into the ML which allows the ML to grow in height. The MLH is always located somewhere within the EZ, regardless of the particular MLH definition. Specific definitions of the MLH 61 Chapter 4. Mixed-layer height estimations and their implications with regard to estimating the MLH from backscatter profiles will be discussed in detail in Section 4.4.2. Depending on the level of insolation, the MLH can grow to a few kilometers. The rate of the growth is also affected by the strength of the ambient inversion above the EZ. It is also important to note that the MLH has a variety of definitions. The available observations often dictate which definition is used. A theo- retical definition that is often cited involves locating the altitude at which there is a maximum downward heat flux due to entrainment. However, this definition requires the difficult task of measuring heat flux values at a vari- ety of different altitudes. Another definition labels the MLH as the altitude at which there are equal parts FA air and ML air. Yet another approach defines the MLH as the altitude characterized by the highest variability in conserved variables. The Parcel Method takes the MLH to be the point at which an adiabatically rising air parcel would finally come to rest due to its impingment on the EZ. During the afternoon, the reduction in insolation results in a plateauing of the MLH growth. Eventually the source of turbulent kinetic energy (i.e., the sensible heat flux) is no longer large enough to compensate for the viscous dissipation of turbulence. This often occurs in the late afternoon nearing sunset. At this time, the turbulent kinetic energy levels are reduced by an order of magnitude within the period of around an hour, with convective activity decaying from the ground upwards (Garratt, 1992). Once the ML becomes effectively non-turbulent, it is reclassified as the residual layer (RL) which shares the same vertical distribution of variables as the ML but is now non-turbulent. The inversion of the EZ often remains as well, which at night 62 Chapter 4. Mixed-layer height estimations is referred to as a Capping Inversion. At the surface after sunset, an inversion begins to form due to radiative cooling of the surface. This non-turbulent nocturnal stable boundary layer (SBL) grows in height throughout the night reaching a few hundred metres by sunrise. The stability strength also increases and effectively de-couples the atmosphere above it (the RL) from the surface. The stability strength of the SBL on any given night is affected by a variety of non-negligible factors including radiative divergence, subsidence and advective effects. Af- ter sunrise, convective activity again acts to break down the stable layer. Within this stable layer there is often an accumulation of pollutants which frequently leads to nocturnal pollution events. A complete description of the SBL is significantly more complicated than for the CBL. Firstly, the developmental timescales of the SBL are often very long, leading to a situation in which the nocturnal BL is rarely in a state of equilibrium. Additionally, the SBL is often characterized by non-periodic, intermittent bursts of turbulence (Salmond & McKendry, 2005), which are often overlaid on large-scale wave motions such as gravity waves and Kevin- Helmholtz billows, making it difficult to assess turbulence structure within the SBL. Of particular difficulty is the estimation of the SBL depth, as there is often no obvious demarcation in the temperature profile between the SBL and the RL above: the temperature profile describes a smooth, asymptotic curve. Indeed, there is no one agreed upon definition of the SBL depth. A variety of definitions are summarized by Stull (1988). Some involve analysis of the temperature profile, others base their definition on turbulence levels, while some define the SBL depth using wind profiles. These difficulties are 63 Chapter 4. Mixed-layer height estimations significant for this project as they make it very difficult to estimate the SBL depth from ceilometer observations. Because of these complications, only daytime convective situations will be analysed in this project. 4.2.2 Theoretical convective boundary layer development By making a variety of assumptions, the development of the CBL can be approximated from theory. The Bulk ML model is one such simplifica- tion which assumes constant values of potential temperature, wind and con- stituent concentrations throughout the mixed layer. At the top of the ML is an immediate temperature drop which simulates the EZ. Above the ML the FA is often assumed to have a constant positive lapse rate. Following these assumptions, Garratt (1992) presents an equation for the development of the MLH, zi, for convective, non-advective situations: zi = √ 2 (1 + 2Ar) γ ∫ t 0 ( w′θ′ ) s dt′ (4.1) where Ar is the entrainment coefficient which specifies the ratio of the surface heat flux to the heat flux at the EZ, γ, is the ambient lapse rate, and ( w′θ′ ) s is the surface kinematic heat flux. If one assigns a sinusoidal shape to the heat flux, then as a first order approximation, the initial morning hours of CBL growth follow a linear trend (Reuten, 2006): zi = √ pi 2 (ω′θ′)s,max tdγ (4.2) Where (ω′θ′)s,max is the daily maximum heat flux and td is the diurnal heating time scale, which is taken as 7.75 hours by Reuten (2006).Therefore, 64 Chapter 4. Mixed-layer height estimations according to this basic theory, the two factors affecting the growth rate of the CBL are the strength of both the stratification and the sensible heat flux. However, there are a variety of processes not included in this basic the- ory which can have a significant impact on CBL development. For instance, large-scale subsidence or upwelling due to synpotic forcing often has a sig- nificant effect on the above processes. Adiabatic warming due to subsidence can create a ’subsidence inversion’ which suppresses the vertical develop- ment of convective activity. Adiabatic warming due to subsidence can also counteract nocturnal radiative cooling, decreasing the strength of the SBL. As well, the basic bulk-layer theory ignores advective effects. Yet, in a coastal region as complex as Vancouver, advection plays a large role in determining the properties of the BL. For instance, air-masses which advect over a mountain range such as the Coastal Mountains north of Vancouver often descend on the lee-side. Adiabatic warming of this descending dry air can lead to the creation of what is referred to as a valley inversion. 4.2.3 The thermal internal boundary layer Advection plays a large role in BL development. When air is advected over a boundary between two surfaces with significantly different properties, a boundary is formed which grows in height downwind of the interface, below which the air is in equilibrium with new surface, while above the air is still unaffected by the properties of the new surface and is in equilibrium with the former surface. The lower layer is referred to as an internal boundary layer (IBL). 65 Chapter 4. Mixed-layer height estimations Advection over a coastline is one such example. During warm daytime conditions, there is often a significant temperature difference between the sea and warmer land due to the large thermal inertia of the sea, which also acts to maintain a stable BL above the sea which is in contrast to the unstable situation on land. As this stable air is advected over land, it suppresses the land-based vertical convective turbulence, resulting in an IBL which grows in height down wind of the coast and is often referred to as a thermal internal boundary layer (TIBL). The top of the TIBL often manifests itself as an elevated inversion the height of which is usually described using a square root dependence on the distance along the mean-wind direction from the coast. In fact, the TIBL is a direct spatial analog to the temporal morning growth of a CBL in non-advective situations. If one assumes Taylor’s Hypothesis of frozen turbulence to be true, then the independent time variable in equation 4.1 is simply replaced by a down-wind fetch distance, x, by assuming a constant advective velocity, va: htibl = √ 2 (1 + 2Ar) γ ∫ t′+x/va t′ ( w′θ′ ) s dt′ (4.3) where t′ is the time at which the the air begins to advect over the coastline (Garratt, 1992). The lapse rate of the Stable Marine Layer (γ), and the advective velocity both have an inverse effect on the TIBL height at a given point inland. This simple one-dimensional model assumes a constant wind speed and direction as well as a spatially constant surface heat flux. More complicated two and three dimensional models of TIBL development have been developed and applied to the LFV by Steyn & Oke (1982), Batchvarova 66 Chapter 4. Mixed-layer height estimations et al. (1999) and Cai & Steyn (2000). Within the LFV, fair weather condi- tions along with the associated sea breeze often lead to the development of a TIBL at the top of which an elevated inversion suppresses the growth of the CBL. Therefore, the MLH in Vancouver is significantly affected by the presence of on-shore advection. 4.3 Previous measurements of mixed-layer heights within the Lower Fraser Valley There has been a variety of previous estimations of the MLH within the LFV. The example which is most applicable to this project is the numerous days of sodar observations taken at the Sunset Tower site which was presented by Steyn (1980). On most days the inversion height estimated from the sodar data reached a maximum altitude sometime between 1200 and 1400 local solar time, with the height passing 200m sometime between 0600 and 0800 local solar time. As well, most days showed a declining inversion height throughout the afternoon. This is contrast to the predictions of the simple bulk-model theory discussed in Section 4.2.2 which predicts that the MLH will plateau during the afternoon. The daily maximumMLH achieved during these fair weather conditions ranged from 400 to 700 metres. The same study also involved a number of tethersonde ascents. The inversion heights estimated from the resulting temperature profiles showed good agreement with the sodar-based estimates. It is important to note that of the 38 inversion heights estimated from the tethersonde flights, only two were above 600m. 67 Chapter 4. Mixed-layer height estimations A similar study was presented by Steyn & Wallis (1987) in which MLH estimates were made in three different sites: a coastal location, East Surrey, and Delta. These observations were performed during the summer of 1985. The coastal location showed no significant sign of a thermally driven mixed layer, being so close to the coast. The other two inland sites showed daily trends in the MLH which were similar to the results of Steyn (1980). namely, a MLH which peaks sometime between 1200 and 1500 LST, and declines in the afternoon. As these two sites were further inland than the Sunset Tower site, they exhibited slightly higher MLH: the values were often above 700m. Cai & Steyn (2000) attempted to model the spatial variation of the TIBL within the LFV which they validated with observations recorded in mid-July 1985. Tethersonde profiles were measured in Surrey, Delta, and at Queen Elizabeth Park which is 3 km northwest of the Sunset Tower site. At Queen Elizabeth Park the estimated MLH rose to a maximum height between 500m and 600m, with a declining height in the afternoon. The maximum MLH in Delta rose to just above 600m while in Surrey the MLH rose past 800m on one day and was consistently higher than the other two sites. During August 1993, an intensive observational project, “Pacific ’93”, was undertaken within the LFV in order to gain a further understanding of ozone pollution events within the region (Hayden et al., 1997). A va- riety of MLH estimates resulted from this study, including estimates from both tethersonde ascents as well as aircraft observations. Significantly, MLH values were also estimated from aircraft-based Lidar observations using the minimum gradient method. These lidar measurements demonstrated the spatially complex nature of the MLH within the LFV: the presence of a con- 68 Chapter 4. Mixed-layer height estimations voluted coastline and major topographical features create significant spatial variability in the MLH. A number of lidar transects were used by Batch- varova et al. (1999) in order to validate a model of the MLH within the LFV. During anti-cyclonic conditions the transect nearest to the ceilometer field site ( 11 km to the east at nearest approach) indicated a MLH of around 600m. McKendry et al. (1997) presented a day of tethersonde data observed near Langley during anti-cyclonic conditions (see Figure 2.1 for location). The measurements showed a MLH growing to a height of around 580m by 1400 and remained at this height throughout the rest of the afternoon. Hayden et al. (1997) presented MLHs for a variety of locations through- out the LFV during the Pacific ’93 observational period. The maximum MLH observed in Langley via an ozonesonde ranged from 1100m to around 500m over a week-long period. MLH estimates for Abbotsford, which is fur- ther inland, were slightly higher, agreeing with the concept of an increasing TIBL with distance from the coastline. MLH values were also estimated from airborne lidar observations. Significantly, the MLH values estimated from the lidar data were strongly correlated with the aircraft observations. Another large-scale field campaign was undertaken during August 2001 (“Pacific 2001”) within the LFV. During this study, two radiosondes were installed in Langley and Chilliwack. A third tethersonde station was set up in Slocan Park which is located 3 km northeast of the Sunset Tower site. These observations are summarized by Snyder & Strawbridge (2004). During the first period of the study, a blocking high in the eastern Pacific created stable, dry conditions in which subsidence acted to suppress the 69 Chapter 4. Mixed-layer height estimations height of the CBL. During this period the MLH observed at Slocan Park stayed below 750m. Following this, a series of troughs passed over the region. This allowed MLHs to rise, and the inversion heights observed at Slocan Park rose past 1000m on several occasions. Again, the MLHs observed further up the valley in Langley and Chilliwack were generally higher than what was observed at Slocan Park. Strawbridge & Snyder (2004b) utilized a ground-based scanning lidar system to estimate the MLH near Langley during the Pacific 2001 cam- paign. Their results confirm those of Snyder & Strawbridge (2004), that is, the initial few days of significant subsidence saw suppressed maximum MLH values (the scanning lidar observed MLH in the range of 600 to 900m) while the later unsettled period saw higher values: 800 to 1200m. Another significant finding by Strawbridge & Snyder (2004b) is that the MLH con- sistently exhibited a sharp drop between 1800 and 1900 PDT. This suggests a sudden collapse of convective activity as insolation levels decrease in the late afternoon. Steyn & Oke (1982) also made note of this feature. They suggested that it may be a result of subsidence due to the collapse of the sea breeze. Overall, during fair weather anti-cyclonic conditions, MLH values over Vancouver tend to be lower than 1km and are usually within the range of 400-700m. This general range of heights was observed via tethersonde as- cents and aircraft measurements, as well as sodar and lidar observations. The agreement in observations between these different techniques offers fur- ther confidence in the results. Further inland, the MLH height increases as the effects of the coastline diminish; further up the valley, around Abbotsford 70 Chapter 4. Mixed-layer height estimations and Chilliwack, the MLH often rises above 1km. By making along-valley transects of the LFV with an airborne lidar, Strawbridge & Snyder (2004a) were able to observe a pronounced increase in the MLH further up the valley. 4.4 Development of mixed-layer height algorithms 4.4.1 Overview Two different algorithms were developed in order to estimate MLH values from the ceilometer data: The ideal-profile method suggested by Steyn et al. (1999) and the minimum gradient method which has been applied to lidar data by various authors and to ceilometer data by Eresmaa et al. (2006). This section first presents a discussion of issues inherent in estimating MLH values from ceilometer data. Following this, detailed descriptions of both MLH algorithms will be presented. Finally, results from both algorithms will be assessed and a comparison of the techniques will be made. 4.4.2 Discussion of issues inherent in mixed-layer height estimations A fundamental assumption made in this project is that aerosol concentra- tions within the BL can be an accurate proxy for BL temperature structure. Comparisons of ceilometer data with temperature profiles by Eresmaa et al. (2006), Emeis & Schafer (2006), Münkel et al. (2007) and Schafer et al. (2004), suggest that MLH estimations from ceilometer data do agree with 71 Chapter 4. Mixed-layer height estimations temperature structure. Eresmaa et al. (2006) compared 56 MLH estima- tions from radiosonde data to ceilometer MLH estimations. These data were well-correlated and significant (ρ = 0.90). Yet, this comparison by Eresmaa et al. (2006) required the removal of 15 data points due to low aerosol concentrations. This is an example of a significant issue with estimating MLH from backscatter data: namely, turbid conditions are required in order to make any kind of estimation. It will be shown later that the turbidity of the ML directly affects the reliability of the MLH estimation algorithms used in this project. This requirement of a turbid ML is of particular concern in the case of Vancouver. Due to the city’s coastal location it is often the case that little or no BL structure is observed by the ceilometer due to relatively clean on-shore winds. Another obvious issue is the presence of water droplets which act to saturate the ceilometer signal. However, periods of precipitation are of less interest, as they do not represent cases of either convective activity, or high pollutant concentrations. What is of greater significance is the presence of smaller amounts of water droplets within the BL due to the presence of haze. Haze is often present in Vancouver, a major cause of which is the presence of sea-salt particles within the atmospheric column. Sea-salt is known to be hygroscopic, i.e., individual particles are capable of absorbing significant amounts of moisture which causes an expansion in the size of the particle. This issue is of primary concern when attempting to estimate the MLH from ceilometer data operationally, especially as there are no vertical profiles of humidity regularly available for the LFV region. It is very difficult to distinguish a signal due to ground-based aerosols from a signal due to the 72 Chapter 4. Mixed-layer height estimations presence of haze without a qualitative manual analysis of the data. This issue will be discussed in more detail in the following sections. Yet, even with a sufficient aerosol concentration, and lack of haze or precipitation, backscatter profiles are not always a reliable proxy for ther- mal structure. One must be aware of the possibility of advected sources of elevated aerosols which may significantly alter the ceilometer signal within the BL (Seibert et al., 2000). Strong slope flows present within the LFV during the day may act to eject aerosols into the FA. These ejected aerosols often form elevated layers above the city. McKendry & Lundgren (2000) discuss this process of mountain venting. Aerosols may also be ejected into the FA or into the stable EZ by strong thermals which over-shoot the EZ at the top of the CBL (McKendry & Lundgren, 2000). The ejection of PM into the FA can lead to a possible positive bias when estimating the MLH from ceilometer backscatter profiles. Indeed, Seibert et al. (2000) found that MLH values determined from Lidar data are systematically higher than val- ues determined from temperature profiles or sodar measurements. This is an important issue to note when applying MLH algorithms to ceilometer data. Also of note is the possibility that aerosol structure within the CBL does not develop as quickly as the thermal structure. By analyzing simulta- neous MLH estimates from both sodar and ceilometer observations, Emeis & Schafer (2006) were able to identify a delay in the MLH development determined via the ceilometer when compared to the sodar-based estimates (which are based on thermal activity). The authors suggest that this de- lay is evidence of a slower response time of ground-sourced aerosols when compared to the thermal development of the CBL. 73 Chapter 4. Mixed-layer height estimations 4.4.3 The minimum gradient method A simple way of estimating the MLH from backscatter profiles is to locate the altitude which exhibits the largest negative backscatter gradient. This is then taken to be associated with the temperature inversion of the EZ. Emeis & Schafer (2006) and Schafer et al. (2004) have applied this technique to data from an LD40 ceilometer. The authors applied a two dimensional running average to the data before estimating the MLH: ten minutes temporally (40 data points) and 180m in the vertical (25 data points). This same averaging technique was used in this project: a ten minute temporal average (40 data points) and a 180m vertical window which, in the case of the CL31 ceilometer, is 36 vertical bins at a 5m resolution. In order to find the gradient value at each altitude bin, G(t, h), from the smoothed backscatter profile, B(t, h), the backscatter value 90m below the altitude in question is subtracted from the backscatter value 90m above. I.e., G(t, h) = B(t, h+ 90m)−B(t, h− 90m) (4.4) The MLH, zi is then taken to be the altitude bin associated with the minimum value of G(t, h). It should be noted that this algorithm only al- lows for analysis of the profile above 140m. The first 50m of the ceilometer was characterized by erroneous noise, and therefore was ignored. In addi- tion, unlike Emeis & Schafer (2006), the averaging window of 180m was not allowed to be reduced in order accommodate the first 90m of the profile. It was felt that the reduction of the window size would result in erroneous 74 Chapter 4. Mixed-layer height estimations MLH estimates. For these two reasons, the first 140m of the profile was ignored in this algorithm. 4.4.4 The ideal-profile method In general, backscatter profiles taken during turbid, convective situations exhibit a particular trend: a fairly constant backscatter level within the ML, above which is a distinct transition (which is associated with the MLH) to a constant low backscatter value associated with the clean FA. An example of such a profile is shown in Figure 4.2. This consistent profile shape can be used to estimate the MLH. Steyn et al. (1999) have suggested fitting an ideal-profile to measured profiles. This fitted profile can then be used to estimate the MLH, along with the EZ thickness. As previously mentioned, Eresmaa et al. (2006) applied the ideal-profile method to data from a CT25K ceilometer. They found that the during clear sky convective situations, the results agreed with MLH estimations from radiosonde temperature profiles. The EZ can be considered as a region in which there is a diminishing probability of observing ground-sourced aerosols with height. Therefore, the authors suggest that a cumulative error function would provide the most accurate representation. This function can be presented in terms of four parameters: the MLH, zi; Bm, and Bu, which are associated with the mean backscatter value of the mixed-Layer and FA, respectively; and s which is associated with the EZ thickness. B(z) = (Bm +Bu) 2 − (Bm −Bu) 2 erf ( z − zi s ) (4.5) 75 Chapter 4. Mixed-layer height estimations 0 20 40 60 80 100 0 500 1000 1500 Backscatter (A.U.) H ei gh t a bo ve g ro un d (m ) Ideal Profile Measured Profile MLH Figure 4.2: An example of an ideal backscatter profile fitted to a measured profile with the MLH (zi in Equation 4.5) indicated. An example of a fitted profile is shown in Figure 4.2. From this equation it is apparent that the MLH is defined as the height at which the ideal-profile is equal to the average of Bm, and Bu. Steyn et al. (1999) suggest that this particular technique is more robust than the gradient method, as the estimation involves the entire profile rather than identifying a single critical point. Additionally, this technique can be used to estimate the EZ thickness as well as the entrainment coefficient, 76 Chapter 4. Mixed-layer height estimations which determines the ratio of ground-level sensible heat flux to downward heat-flux via entrainment. Description of algorithm The use of an entire profile for MLH estimations results in an increase in computational time required. A four-parameter fitting algorithm had to be used. For this project a simple genetic algorithm was developed to perform the optimization. For each time-step, an initial set of 100 profiles is randomly generated and then compared to the measured profile. The profiles are then ranked according to their “level of fitness,” i.e., how well they approximate the measured profile. Then the top 50 profiles are retained while the bottom 50 are discarded and replaced with “mutated” versions of the top 50. That is, each of the top 50 profiles produces an “offspring”, which varies slightly from its predecessor. Numerically, this is done by adding a small variation (negative or positive) to the four parameters of each of the top 50 profiles. Once this replacement is complete, the process is started again: another 50 profiles are retained and “offspring” of these profiles are created. Each iteration of this process is referred to as a generation. Initially, the best agreement possible amongst the 100 profiles increases significantly with each generation, as the poorest estimations are removed. Eventually, through random mutations, a profile is created which exhibits a high level of agreement with the measured profile, and this profile remains the best fit over multiple generations. In order to decrease required computation cost, the process is stopped either when the hundredth generation is reached or when the estimated MLH from the profile with the highest level of fitness 77 Chapter 4. Mixed-layer height estimations does not change over 20 generations. For each time step, the 100 initial profiles are calculated using a range of values for the four parameters: zi, Bm, Bu and s. The range of values used is centered around an initial estimation. For sequential profiles, the parameter values of the previous fitted profile are used for this initial estimation. For the first profile in a dataset, backscatter values, Bm and Bu, are estimated by calculating the mean backscatter value of first 50m and top 500m of the measured profile while the initial values of zi and s are taken to be 300m and 50m, respectively. Eresmaa et al. (2006) suggested that this type of initial estimation increases the speed and accuracy of the algorithm. As the noise level grows significantly with height, only the bottom 1200m of the measured profiles are used. It was found that this choice of altitude cut-off did not have a large effect on the results of the algorithm, as long as the maximum altitude remained above 1000m. 4.4.5 Assessment of algorithms The results of both algorithms were found to agree with visual estimations of the MLH in most cases. This is especially the case for periods characterized by a turbid ML and relatively clean FA. If this is not the case, then the results of the algorithm begin to exhibit significant noise due to a weak ML signal in the backscatter profiles. The presence of noise is most prominent in the results from the minimum gradient method. Often, when the ML is clean, backscatter gradients associated with noise structure within the FA become comparable to the gradient associated with the EZ. An example of these erroneous estimates of the minimum gradient method is seen in Figure 4.3. 78 Chapter 4. Mixed-layer height estimations The two different MLH estimates in this Figure agree well with one another, except for a period with a low ML backscatter signal. Moving beyond this single example, a comparison of MLH estimates from both algorithms for 22 days is shown in Figure 4.4A. Although the majority of data are in good agreement, there still is significant scatter. Time of Day H ei gh t A bo ve G ro un d (m ) B ac ks ca tte r ( A. U. ) 06:00 08:00 10:00 12:00 14:00 100 200 300 400 500 600 700 800 900 1000 1100 1200 0 10 20 30 40 50 60 70 80 90 100 Figure 4.3: MLH estimates from both the ideal-profile method (red) and the gradient method (black) applied to a single convectively active summer day. In order to analyse the influence of the turbidity of the ML on this noise issue, the difference between the mean ML backscatter and the mean FA backscatter was calculated for each data point plotted in Figure 4.4A. The 79 Chapter 4. Mixed-layer height estimations 0 500 1000 0 500 1000 M in . G ra d. M LH (m ) Ideal Profile MLH (m) 0 500 1000 0 500 1000 M in . G ra d. M LH (m ) Ideal Profile MLH (m) 0 20 40 60 80 100 0.7 0.8 0.9 1 Co rre la tio n Co ef . ( ρ ) ML Backscatter − FA Backscatter All Data 30 A.U. Threshold A B C Figure 4.4: A: A comparison of MLH estimates from the ideal-profile method and the minimum gradient method for 22 fair weather days. B: Same as A, but only for days with an average difference in the ML and FA backscatter values of 30 A.U. C: The Pearson correlation coefficient between the two algorithms as a function of the minimum ML to FA backscatter difference to which the algorithms were applied. larger this difference, the more likely the minimum gradient algorithm will be able to identify the MLH. Indeed, by removing any instances in which this difference was below a certain threshold value, the correlation between the two algorithms was dramatically improved. Figure 4.4B shows data for all cases in which the difference between the mean ML backscatter and mean FA backscatter was greater than 30 A.U. It is clear from this plot that the agreement has improved significantly. In Figure 4.4C, the correlation 80 Chapter 4. Mixed-layer height estimations of results is shown as a function of the backscatter difference between the ML and the FA. It is clear from this figure that the agreement between the two algorithms improves dramatically when cases of clean ML’s are removed. It also indicates that once the ML backscatter is 30 A.U. above the FA backscatter, the minimum gradient method is no longer susceptible to significant noise. It was mentioned in the description of the ideal-profile algorithm that in order to reduce computational time, the genetic algorithm was not always allowed to converge completely. This decision has resulted in a certain lack of precision in the results of the genetic algorithm; multiple MLH estimations of the same profile will produce slightly different results. An example of this is shown in Figure 4.5. This day (August 1, 2007) presents examples of a turbid situation (0800 to 1100 LDT), as well as a much cleaner ML. The genetic algorithm was run ten times for this day, providing a 10-member ensemble. Calculating both the ensemble-mean and standard deviation for each time step, it is apparent that the algorithm precision is the highest in the cases with a relatively high signal. This result is to be expected and was also found by Steyn et al. (1999), although they used a self-annealing algorithm. In general, the standard deviation of the algorithm results represents an error not larger than 10%, which was deemed to be acceptable for this project, as no direct comparisons to other forms of MLH estimations will be made. This allowed for reasonable computation times. As is often the case, a trade-off must be made between computational time and precision of the results when making MLH estimates. Although the minimum gradient method is a significantly simpler algorithm, the more 81 Chapter 4. Mixed-layer height estimations H ei gh t a bo ve g ro un d (m ) B ac ks ca tte r ( A. U. ) 05:00 07:00 09:00 11:00 13:00 15:00 17:00 100 200 300 400 500 600 700 800 900 1000 05:00 07:00 09:00 11:00 13:00 15:00 0 20 40 Time of Day (LDT) St an da rd D ev ia tio n (m ) 0 20 40 60 80 100 Figure 4.5: Results from ten individual runs of the ideal-profile algorithm on a single day along with the 10-member ensemble average. The standard deviation of the ensemble is also indicated for each time step. robust ideal-profile method leads to less erroneous results. The algorithm is also able to provide information about the EZ thickness. Additionally, the precision of the results from the ideal-profile method can be increased by increasing computational time. For this project, the ideal-profile method, run with at 100 iterations per profile was found to be most suitable and will be used for the remaining analysis. 82 Chapter 4. Mixed-layer height estimations 4.5 Flagging algorithm As this project is limited to the analysis of convective daytime situations, a flagging algorithm was developed in order to remove any other types of situations. This algorithm removed periods of precipitation and fog, all night-time data, as well as periods of low or poor backscatter signals. As the two MLH algorithms were to use the bottom 1200m of data, the flagging algorithm also removed instances of clouds within the bottom 1200m of the profile, as these saturated signals would lead to erroneous MLH estimates. In order to remove periods of rain, fog or cloud formation, a backscatter intensity threshold was set. Any backscatter profile exhibiting a signal above this value was removed from the dataset. A value of 200 A.U. was chosen. This value was based on the fact that the ceilometer signal was bimodal in its distribution. That is, the presence of water droplets resulted in a signal which was orders of magnitude larger than the signal resulting from PM, which remained below 200 A.U. This analysis is also supported by the comparison of the ceilometer data with ground-level PM data presented in Chapter 3, where 200 A.U. was also found to be an optimal threshold value for the removal of erroneous data due to the presence of condensation. To avoid non-convective situations, all night-time data was removed. Sunset and sunrise times were used to delineate daytime from the nighttime. An additional one hour was added to the sunrise time, as it was found that the convective signal was not present in the ceilometer data during the first hour after sunrise. Cases which exhibited low PM concentrations were also removed, as the 83 Chapter 4. Mixed-layer height estimations ceilometer signal would not be significant enough to provide a reasonable MLH estimation. To accomplish this, a mean backscatter value for the bottom 300m of each profile was calculated. If this value was lower than 5 A.U., the profile was removed from the dataset. As well, any instances exhibiting a backscatter signal within the FA which was as large as that of the BL were also ignored, as these cases would not exhibit a prominent gradient in the backscatter profile. For this analysis the FA was assumed to be represented by the signal from 800m to 1200m, while the ML signal was represented by the bottom 300m. It is important to note that these two flagging techniques are, to a certain extent, arbitrary. However, the results of this flagging algorithm were manually checked for a variety of different situations. Through this manual assessment, the threshold and altitude values used here were optimized in order to decrease noise within the MLH estimations, but at the same time retain significant cases. By flagging nocturnal data, instances of low MLH signals, and periods of saturation, the flagging routine was able to isolate convective periods. In ret- rospect, it is likely that larger sunrise and sunset offsets would have removed more non-convective situations. Observations have suggested that convec- tive activity disappears within the last hour before sunset (Stull, 1988). Additionally, due to erroneous data within the first 100m, the ceilometer is usually unable to register evidence of convective activity for at least two hours after sunrise, when the MLH rises above 100m. A major issue with this flagging technique is the presence of haze. As mentioned previously, the backscatter signal due to haze is very similar to that of ground-sourced aerosols. Therefore, it is difficult to remove periods of 84 Chapter 4. Mixed-layer height estimations haze from the dataset using the flagging algorithm. Remnant moisture from cloud formation is one such example. Elevated layers of moisture which often remain after BL cloud formation has ceased may lead to relatively high MLH estimates, even though the vertical extent of convective turbulence may be much lower. The flagging algorithm would have no reason to remove such a case. In other cases, haze may be present throughout the entire atmospheric column up to a few kilometres. In this situation, the backscatter signal within the FA is comparable to the signal within the ML, leading the flagging algorithm to remove this data, even though the period may be characterised by fair weather and active convection. Therefore, the flagging results are improved be removing any day in which saturation occurs within the bottom 1200m. This increases confi- dence that the algorithm is not including periods in which the backscatter signal is significantly altered by remnant haze in the BL. Such precaution is supported by the analysis presented in Chapter 3, where it was found that water droplets which remain after periods of precipitation decreased the cor- relation between the ceilometer data and ground-level PM measurements. This reduction in agreement was avoided if all data measured within the first 19 hours after periods of precipitation were removed. 4.6 Mixed-layer height estimations 4.6.1 Overview In this section MLH estimates from the ideal-profile algorithm are presented for fair weather convective situations. The data were split into three periods: 85 Chapter 4. Mixed-layer height estimations summer (July to September), Fall (October to December), and winter/early spring (January to May). Data from July 2007 through to May 2008 were used. In order to make legitimate comparisons, only days which exhib- ited significant insolation and no cloud formation within the BL were used. Following these restrictions, along with the initial removal of days by the flagging algorithm, 19 days were chosen for comparison. 4.6.2 Results Summer Eight days within the summer period were chosen for MLH estimations. These days were all characterized by anti-cyclonic conditions and relatively high levels of insolation. Figure 4.6 shows the diurnal variation of the MLH estimates for four of these days along with a mean MLH calculated from all eight days. On average, the MLH tends to rise until 5 hours after sunrise, which places the growth period between sunrise and 0900 to 1045 during the summer months. In the afternoon the MLHs tended to either diminish or plateau. During the early evening, the MLH was often observed to falloff quickly. This feature was also observed within the LFV by both Strawbridge & Snyder (2004b) and Steyn & Oke (1982). Steyn & Oke (1982) suggested that this could be due to subsidence as the convectively-driven sea breeze structure begins to collapse. An exception to this behaviour is seen in the August 24 data where there is considerable growth during the afternoon. Considering that the insolation levels were decreasing during this period and that August 24 was 86 Chapter 4. Mixed-layer height estimations 01:00 03:00 05:00 07:00 9:00 11:00 13:00 15:00 0 100 200 300 400 500 600 700 800 900 Hours after sunrise H ei gh t a bo ve g ro un d (m ) Aug 1 Aug. 24 Aug. 29 Sept. 13 Mean MLH Figure 4.6: The mean daily trend in the MLH from 9 individual fair weather summer days, along with 4 of these individual daily trends. characterised by a hazy afternoon, it is likely that this peak is an erroneous result of the algorithm which was caused by slight condensation within the atmosphere. This is an example of erroneous MLH estimates due to the presence of haze, which was discussed in the introduction to this section. During fair weather convective conditions, it is common for BL cumulus clouds to form, especially, during the morning growth period. The presence of these clouds acts to saturate the ceilometer signal, making the ideal-profile algorithm unusable in these instances. Additionally, the release of latent heat within the clouds acts to increase the upwards buoyancy of individual 87 Chapter 4. Mixed-layer height estimations thermals, resulting in a higher vertical extent of convective turbulence. For these reasons, days in which BL clouds were present were not consolidated with data from cloudless days. However, it is illuminating to compare the MLH from these two different situations. In the presence of clouds, the MLH was taken to be the altitude at which the signal is no longer saturated, i.e., at the top of the cumulus clouds. Eight different days were used. Four of these days are shown as examples in Figure 4.7 along with the mean MLH for all eight days. For the sake of comparison, the mean MLH for the cloudless days is included. It is apparent from Figure 4.7 that the formation of clouds within the BL has a significant effect on the MLH which, on average, rises about 200m higher than the cloudless days. The morning growth is also stronger and earlier. This is likely due to the latent heat release associated with conden- sation and the resulting increase in positive buoyancy and convective energy. However, by late afternoon, as the presence of clouds decreases, the MLHs for these days begins to approach the MLHs exhibited during the cloudless days. It was discussed in Section 4.2.2 that basic bulk ML theory can provide an estimate of the MLH growth rate during the morning hours. Reuten (2006) provided a formulation for the growth rate which depended on the ambient inversion strength and the maximum sensible heat flux reached on the day in question. By estimating values for these two variables, one can achieve a range of growth rates that are likely to occur within the LFV. Chen & Oke (1994) and Batchvarova et al. (1999) suggest a typical in- version strength of 0.007 K/m for the LFV. Reuten (2006) suggests a range 88 Chapter 4. Mixed-layer height estimations 01:00 03:00 05:00 07:00 09:00 11:00 13:00 15:00 0 200 400 600 800 1000 1200 Hours After Sunrise H ei gh t A bo ve G ro un d (m ) July 6 Sept. 17 Aug. 5 Mean MLH Mean MLH (No BL clouds) Figure 4.7: The mean daily trend in the MLH from 9 individual fair weather summer days in which BL cumulus clouds formed, along with 3 of these individual daily trends and the clear MLH trend from Figure 4.6. of 0.006 to 0.007 K/m. Steyn & Oke (1982) presented measurements of the ambient inversion strength within the LFV for an entire day. Based on these observations, they used a similar value of 0.01 K/m for their models. As no direct and continuous observations of the buoyancy flux were taken alongside the ceilometer, physically sensible ranges for these two forcing vari- ables were estimated and used to predict the range of possible CBL morning growth rates in non-advective situations. The chosen range of maximum sen- 89 Chapter 4. Mixed-layer height estimations sible heat flux, 150-300 wm−2, was chosen based on previous energy budget measurements taken at the ceilometer field site which are presented by Oke (1995). A typical maximum heat flux value of 225 Wm−2 and an ambient inversion strength of 0.007 K/m were used to estimate an average morning growth rate for the LFV using Equation 4.2. A value of 0.038 m/s or 138 metres per hour was calculated. This growth rate is compared to the mean MLH growth rate for sum- mer, cloudless days in Figure 4.8. On most days the convective growth was observed starting around two hours after sunrise when the MLH grew past 00:00 01:00 02:00 03:00 04:00 05:00 06:00 07:00 0 100 200 300 400 500 600 700 Hours after sunrise H ei gh t a bo ve g ro un d (m ) Figure 4.8: Morning growth period for clear, fair weather summer days. A theoretically-estimated growth rate of 138 mh−1 is indicated with the parallel dotted lines. 90 Chapter 4. Mixed-layer height estimations the region of erroneous data within the first 100m, and continued to grow until around five hours after sunrise. During this period the mean growth rate observed agrees well with this middle estimate of 0.038 m/s. Low and high estimates of the theoretical growth rates were also cal- culated. For a maximum sensible heat flux of 300 Wm−2 and an ambient inversion of 0.005 K/m, a MLH growth rate of 0.052 m/s (187 m h−1)was calculated as a high estimate while a low estimate of 0.021 m/s ( 75.6 m h−1) was calculated using values of 150 Wm−2 and 0.015 K/m. This range of growth rates was able to account for the majority of the range in morning growth rates observed on the eight individual fair weather summer days. As mentioned previously in Section 4.3, sodar measurements of Vancou- ver’s urban boundary layer were obtained at the Sunset Tower site during the summer of 1978. These observations were presented by Steyn (1980). MLH estimates obtained from the sodar for clear, fair weather summer days were analysed and compared to results obtained from the ceilometer data. An example of such a comparison is shown in Figure 4.9. The daily MLH trend observed on August 29, 2007 was used to compare to the data from Steyn (1980) as it is close to the mean MLH trend for the summer months (see Figure 4.6. The three days taken from the 1978 dataset were chosen to represent the range in MLH estimated from the sodar. It is apparent from Figure 4.9 that the MLHs estimated from the ceilome- ter agree well with the MLHs estimated from the sodar at the same site. Generally, the 1978 dataset exhibits maximum MLHs between 400 at 650m, which agrees well with the range of MLHs observed by the ceilometer on clear days during the summer months. One discrepancy seems to be the 91 Chapter 4. Mixed-layer height estimations Figure 4.9: Three MLH estimates from sodar measurements (solid lines) and the corresponding model predictions (dotted lines) presented by Steyn (1980): July 30, 1978: blue; August 2, 1978: grey; August 8, 1978: red. These measurements were made at the Sunset Tower. An representative day of MLH estimates from ceilometer data (black squares) is provided for comparison. Altered from Steyn (1980) . growth of the MLH to an altitude of around 700m on July 30, 1978. It is possible that this day was characterised by cumulus cloud formation within the BL. The increased growth seen on this day is similar to the growth rates observed by the ceilometer on days with cumulus formation (Figure 4.7). Significantly, the modelled MLH for July 30, which does not account for la- tent heat release, exhibits a more conservative growth rate which is similar to the ceilometer MLH estimate. 92 Chapter 4. Mixed-layer height estimations Autumn, winter and year-long trends Fair weather days characterized by high levels of insolation and no BL clouds were also identified for the autumn and winter months. Three actively con- vective days within the autumn period (October 1, 2007 to January 1, 2008) were identified, while five days within the winter period (January 1, 2008 to May 12, 2008) were chosen for analysis. Figure 4.10 shows the mean MLHs for both the autumn and winter period. The mean summer MLH is included for comparison. 00:00 02:00 04:00 06:00 08:00 10:00 12:00 14:00 0 100 200 300 400 500 600 Hours after sunrise H ei gh t a bo ve r g ro un d (m ) Summer mean MLH Autumn mean MLH Winter mean MLHJ Standard deviation Figure 4.10: Mean daily trends in the MLH are shown from summer, au- tumn, and winter. The standard deviation for the mean trends are indicated by the dotted lines. 93 Chapter 4. Mixed-layer height estimations In the case of the autumn period, the mean MLH did not differ all that much from the summer MLHs. However, considering the small sample sizes for both periods, it is likely that this is not a statistically significant result and was influenced by possibly erroneous MLH estimates. However, the mean MLH growth for the winter is noticeably less energetic than the other two periods. The mean winter MLH grows more slowly and reaches a lower daily maximum height. Statistical analysis of the mean MLH for both the summer and winter months shows that the difference between the two datasets are statistically significant only for the morning growth period. The summer and autumn mean MLH values were not found to be statistically different. Hopefully, as future MLH estimates are added to the dataset, more significant seasonal changes in the diurnal MLH trend will be apparent. The maximum MLHs attained on each of the 19 convective days were estimated in order to obtain a sense of the seasonal trend in MLHs. These results are shown in Figure 4.11. Although there is significant spread in the data, a seasonal trend is apparent in which the maximum MLH decreases from the summer into the winter, with a rise again throughout the spring. 94 Chapter 4. Mixed-layer height estimations July Aug Sept. Oct. Nov. Dec. Jan. Feb March April May300 350 400 450 500 550 600 650 700 D ai ly M ax im um M ix ed −L ay er H ei gh t ( m) Figure 4.11: Daily maximum MLHs for 19 fair weather days with no BL cumulus formation. 4.7 Effects of the thermal internal boundary layer As previously mentioned in Section 4.2.2, the presence of a TIBL due to advection from the Georgia Strait has a large influence on the growth of the MLH within Vancouver. Indeed, if this advective effect was not present, the levels of sensible heat flux within Vancouver would, according to ba- sic bulk-layer ML theory (Section 4.2.2), lead to MLH values on the order of 1 km during the summer, which is much higher than the MLH values 95 Chapter 4. Mixed-layer height estimations suggested by the ceilometer and other observations. Daily trends of MLHs shown previously suggest that on most fair weather days, the MLH exhibits predictable growth within the morning, but very little growth during the af- ternoon. This trend is likely due to the onset of the sea breeze. An example of such a day is August 15, 2007. 4.7.1 Case study: August 15, 2007 Due to the validating tethersonde work performed on the night of August 14-15, early morning temperature profiles up to 350m were available for August 15th. These profiles were used to calculate the theoretical growth of the CBL as suggested by the Bulk ML model (Equation 4.1) which was then compared to the ceilometer data. For these estimates the encroachment approximation was used which assumes that there is no downward sensible heat flux through the EZ (i.e., Ar = 0 in Equation 4.1). This is an acceptable approximation for early morning growth of the CBL through the surface inversion (Garratt, 1992). The day’s heat flux values were estimated using the data from the West- ham Island field site (see Figure 2.4 in Chapter 2 for location). This is a rural, coastal field site which was established as part of the EPiCC observational project. This rural data was augmented with a “suburban adjustment.” Observations presented by Oke (1995) show that around 3 hours after sunrise the rural/suburban difference in measured heat flux val- ues jumps from zero to around 35 Wm−2 during fair weather conditions. This“suburban adjustment” was therefore added to the Westham Island data in order to simulate the heat flux at the ceilometer field site for August 96 Chapter 4. Mixed-layer height estimations 15. Above the extent of the tethersonde profile (about 350m), the lapse rate was estimated by extending the measured lapse rate of the profile’s top 50m, which was found to be around 0.006 K m−1. This agrees with measurements presented by Reuten (2006). However, 1200m profiles measured within East Vancouver during the Pacific 2001 field campaign (Snyder & Strawbridge, 2004) exhibited lapse rates of around 0.01 K m−1 during anti-cyclonic con- ditions. Therefore, this lapse rate was also used. A comparison of these results in Figure 4.12 shows general agreement between theory and observations during the early morning. However, the Figure 4.12: A comparison of the theoretical, non-advective growth in the MLH for August 15, 2007 with the estimated MLH from the ceilometer data (using the ideal-profile method). 97 Chapter 4. Mixed-layer height estimations results show significant divergence after around 1100 PST. On this day, a strong sea breeze signal was observed at Westham Island and the Vancouver Airport, with the initial onset occurring just after 0800 PST. It is therefore likely that the divergence in the theoretical and actual MLH seen in Figure 4.12 at this time is due to the onset of the sea breeze and suppression of the MLH due to the presence of the TIBL. In order to provide further confirmation of this assertion, the theoretical TIBL height for August 15 was calculated using Equation 4.3. The heat flux values were taken from the Westham Island data. The wind direction and speed were estimated from both Vancouver Airport observations and Westham Island data. The wind direction is important as it informs the value of the fetch distance, x, between the coast and the ceilometer field site As neither the speed of the sea breeze front, the ambient lapse rate, nor the actual fetch distance were known precisely, a range of values were chosen in order to produce a range of possible TIBL heights at the ceilometer field site. The results are shown in Table 4.7.1. The middle estimate correlates well with the MLH height observed by the ceilometer on August 15 after 1000 PDT. Lapse rate Sea Breeze Fetch dist. TIBL height (Km−1) velocity (ms−1) (km) (m) High Estimate 0.005 2.5 11.5 595 Middle Estimate 0.006 3.5 10.5 425 Low Estimate 0.01 4.5 9.5 275 Table 4.1: Theoretical estimates of the height of the thermal internal bound- ary layer for the Sunset Tower site on August 15, 2007. 98 Chapter 4. Mixed-layer height estimations 4.8 Discussion There are a variety of short-comings with regards to estimating MLHs from ceilometer data. An obvious drawback is the relatively small portion of days that are available for analysis. This issue is most significant in the case of fair weather days characterized by low aerosol concentrations within the ML. Of the two algorithms, the ideal-profile method is less susceptible to this problem. There are numerous instances in which the flagging routine removed entire days of data due to a low signal even though those days were clearly convectively active. Precipitation also causes the data to be unusable. However, days such as these are of less interest with regards to air quality modelling. The issue of haze has been addressed several times throughout this chap- ter. It is likely the largest source of erroneous MLH estimates. Numerous days could not be used due to the presence of haze both within the ML and in the FA. Therefore, the typical signal of a turbid ML and clean FA, upon which the algorithms depend, was not present, leading to erroneous MLH values. It is likely that a further manual analysis of the data will be required in conjunction with the automatic flagging algorithm in order to remove pe- riods of haze, as the flagging algorithm is not capable of distinguishing these periods. Another drawback of the ceilometer, which is discussed in detail by de Haij et al. (2006), is the problem of poor signal-to-noise ratios within altitudes above a kilometre: the authors often found SNR ratios of less than unity, which resulted in poor MLH estimates. Fortunately, MLHs within 99 Chapter 4. Mixed-layer height estimations Vancouver generally remained within the bottom one kilometre. Yet, this shortcoming should be acknowledged here, as any application of the ceilome- ter within a non-coastal location will likely face this problem. Yet it may be possible to utilize ground-level PM observations in order to flag periods in which haze may be present. By following a similar analysis to that discussed in Chapter 3, the correlation between the ceilometer data and the ground-level PM data could be calculated for each day. It may be possible to define a correlation threshold value, below which the backscatter signal is assumed to be significantly influenced by moisture. In this way, days characterized by haze may be automatically removed by the flagging algorithm. The results described in this chapter would have also benefited from a comparison with vertical profiles of direct meteorological observations. In this way, the assumption that backscatter profiles can be used to suggest thermal structure, which is central to this project’s results, could be con- firmed. It is likely that this assumption is not always the case, due to several reasons such as the possible ejection of aerosols into the FA or the possibility of an advective origin of elevated aerosols. Additionally, information on hu- midity and moisture could have allowed for a more rigorous analysis of the effects of haze on the MLH estimates. These sources of erroneous data could have been identified if, for instance, regular radiosonde data were available for the region. However, it should be mentioned that previous MLH estima- tions from ceilometers have been confirmed by data from radiosonde ascents (Eresmaa et al., 2006; Emeis & Schafer, 2006). Additionally, ground-level, co-located observations would have also helped 100 Chapter 4. Mixed-layer height estimations in the confirmation of the MLH estimates. Specifically, turbulent and ra- diative energy fluxes would have allowed for a comparison of convective turbulent energy with MLH growth rates, allowing for a more accurate con- firmation of theory. This absence of complementary datasets has led to subjectivity within this chapter. For instance, on several occasions MLH estimates were deemed to be erroneous, based solely on a qualitative analysis of the ceilometer data without the use of any other independent observations. Additionally, the flagging routine developed to remove erroneous data was based on a series of threshold values which, to a certain extent, were arbitrary. The values finally chosen represented a compromise between avoiding noisy data which would lead to erroneous results and retaining as much of the convective data as possible. As an example, sensible heat flux measurements at the Sunset Tower would allow for an ’active convection’ flag to be developed so that MLH estimates could be confidently applied to only convective situations. Also, periods of fog could be avoided using humidity observations. Despite these issues, the results from this chapter show that the ceilome- ter is able to provide estimates of the MLH that agree with previous obser- vations of the BL within the region. Significantly, this is are the first study in which continuous estimates of the MLH have been available over the pe- riod of an entire year in Vancouver. This has allowed for an assessment of the yearly trend in the MLH for Vancouver during convective situations (Figure 4.11). Additional data, accumulated over several years, will provide a much stronger confirmation of this trend. The future development of a MLH climatology using several years of data, similar to the one developed 101 Chapter 4. Mixed-layer height estimations by de Haij et al. (2006) for a coastal location in the Netherlands, will be valuable for validating regional air quality models and regional forecasting models. 4.9 Conclusions Estimations of the MLH within Vancouver, BC, were presented in this chap- ter, along with an analysis of two different algorithms for estimating the MLH, and a flagging algorithm for the automatic removal of non-convective or unsuitable situations. Both daily and yearly trends in MLH values were analysed. Development of the two algorithms for estimating the MLH, the mini- mum gradient method and the ideal-profile method, was presented in detail. The minimum gradient method is much less computationally intensive than the ideal-profile method and simpler in its execution. Both algorithms pro- vided erroneous MLH estimates for situations in which the ML was relatively clean. As well, a relatively turbid FA also led to erroneous estimates. The combination of these types of non-ideal situations with periods of precip- itation, fog and low clouds led to a significant portion of the data being deemed unsuitable for MLH estimates. However, the periods for which the MLH could be estimated were often the periods which were pertinent to air quality modelling: calm, anti-cyclonic conditions in which a dry, warm air mass is stationary over the LFV. The ideal-profile method is the more robust of the two MLH algorithms as it considers the entire backscatter profile and not just one critical point. 102 Chapter 4. Mixed-layer height estimations Noise structure within the data often led to faulty MLH estimates by the minimum gradient method. However, the ideal-profile method is much more computationally intensive. As well, a trade-off must be made between com- putational time and the precision of the algorithm: individual runs of the ideal-profile method with the same data produce slightly different results. It should be reiterated here that there are inherent issues surrounding the estimation of MLH values from backscatter profiles. The presence of haze was found to be a significant source of erroneous estimations. As well, ejection of aerosols into the FA via vigorous convective activity can cause a positive bias in the MLH estimates. Elevated layers of advected aerosols or a lag in the development of aerosol structure relative to thermal development can also lead to erroneous data. The automatic flagging algorithm described in this chapter was found to remove the majority of erroneous and non-convective data. However, the algorithm is based on a number of parameters which are subjectively determined. Often these choices are based on a trade-off between avoiding noisy or erroneous MLH values and minimizing the removal of legitimate data. The flagging algorithms were also not able to distinguish a signal due to haze from one due to ground-based aerosols. Considering these issues, a further manual flagging of erroneous data will likely be needed to achieve the best results. After the automatic flagging and additional manual removal of data, 19 fair weather days were selected for comparison over a period of ten months, from July 2007 to May 2008. All 19 days had relatively high levels of insola- tion. These results allowed for an analysis of the yearly trend in maximum 103 Chapter 4. Mixed-layer height estimations MLHs (Figure 4.11), which ranged from 650m in August to 360m in Decem- ber. The data also allowed for the estimation of the MLH for days in which BL cumulus clouds formed. It was found that due to latent heat release, the MLH grew more quickly and to higher altitudes than for clear days. Analysis of the diurnal trend in the MLH was made for the summer, autumn, and winter months separately. The MLH followed a similar trend for all three seasons. A morning growth period was observed from around 2 hours after sunrise to around 7 hours after sunrise. During the afternoon and evening, the MLH would either plateau and exhibit no major growth, or decrease in height. Of significance is the fact the the ceilometer was only able to return MLHs for altitudes above 100m. This was due in part to pervasive erroneous data within the bottom 50m of the ceilometer data, as well as the frequent presence of turbid nocturnal stable boundary layers in the early morning which often masked the growth of the CBL for the first 100m or so. Mean diurnal MLH trends for each season were compared to each other (Figure 4.10). Although there seemed to be apparent differences between the three mean trends, the small number of samples for each season (9 summer cases, 3 autumn cases, and 5 winter cases) meant that no statistically sig- nificant differences in the trends could be found beyond the morning growth period. The gathering of future cases will hopefully increase the confidence in these mean diurnal trends. None of the MLH estimates were confirmed by measured profiles of tem- perature, humidity, or wind information. Additionally, this project suf- fered from the lack of co-located, ground-based meteorological measurements 104 Chapter 4. Mixed-layer height estimations which would have allowed for an analysis of influences of the meteorological conditions on the MLH. However, the estimated MLH did show some agree- ment with theory. Specifically, the measured summer morning growth rates fell within the theoretical range of growth rates. As well, afternoon MLHs throughout the year matched the theoretical height range of the TIBL for the Sunset Tower site. Further confirmation of the MLH estimates presented in this chapter was provided by a comparison of MLH estimates from sodar measurments taken at the Sunset Tower site (Steyn, 1980). The range of sodar-based MLH estimates for fair weather summer days correlated well with the ceilometer-based MLH estimates. Overall, the CL31 ceilometer was able to provide accurate estimations of the height of the convective boundary layer for fair weather conditions throughout the year. Although the number of days in which MLH estimates could be made was limited, continuous operation of the ceilometer into the future will allow for the development of a MLH climatology for Vancouver. This information will prove useful in validating both air quality models and numerical weather forecasting models. With regard to future work, this project may also benefit from a further development of the MLH estimation algorithms. For instance, it is likely that the computational time required for the ideal-profile method can be reduced. As well, a wavelet-based algorithm similar to the technique de- veloped by de Haij et al. (2006) may add useful estimates of the MLH. As of May 2008, a full suite of meteorological data as well as turbulent and radiative flux measurements are being collected at the Sunset Tower. These co-located observations will likely prove invaluable in developing a more 105 Chapter 4. Mixed-layer height estimations detailed analysis of the ceilometer data. 106 Chapter 5 Conclusions 5.1 Project outcomes This project had three main objectives: to asses the ability of the ceilometer to make detailed and continuous observations of the BL over several years; to validate the ceilometer via comparisons to direct particulate matter (PM) concentrations and meteorological data; and finally, to estimate mixed-layer heights (MLH) from the ceilometer data using two separate estimation al- gorithms. With regards to the first objective, a permanent field site for the ceilome- ter was establish at the Sunset Meteorological tower at 49th Ave. and Knight St. in East Vancouver, BC. Setting up the ceilometer was found to be rela- tively simple, and the instrument has performed well, showing no significant equipment failure over the period of 11 months. During the study period the ceilometer has provided detailed and continuous measurements of the BL. However, it was found that the bottom 100m of the ceilometer data is characterised by persistent erroneous signals. Consequently, the bottom 50m of the ceilometer data was deemed unusable, while the data between 50m and 100m required significant vertical averaging in order to remove er- roneous vertical fluctuations. Overall, the ceilometer has demonstrated its 107 Chapter 5. Conclusions usefulness for BL observations and has several advantages over other tech- niques, as it is relatively cheap, safe, unobtrusive to the general public, and simple in its operation. The second component of this project, validation of the ceilometer data, has produced positive results. By comparing near-ground ceilometer data to ground-level measurements of PM concentrations, it was found that the ceilometer was able to provide information about regional air quality. It should be mentioned, however, that air quality assessment via the ceilome- ter requires the removal of instances of fog, rain, or periods with significant amounts of moisture. Ceilometer data was also compared to co-located, ver- tical and direct measurements of meteorological variables and PM concen- trations via tethersonde ascents. This study demonstrated the ceilometer’s ability to observe subtle aerosol structure within the atmosphere. The re- sults of both validation studies suggest that the ceilometer would be an ideal addition to an air quality monitoring network as the ceilometer would be able to provide information about not only ground-level PM concentrations, but also vertical information, which, in a geographically complex region such as the Lower Fraser Valley, is very significant with regards to pollutant transport. The final objective, estimating MLH values from the ceilometer data, was obtained with reasonable success. A significant issue with regards to these estimations is the small percentage of the data to which the MLH algorithms could be applied. The significant amount of precipitation experienced within the LFV, combined with periods of fog, haze, and instances characterised by low backscatter signals, result in the removal of a significant portion of 108 Chapter 5. Conclusions the ceilometer data. Consequently, only 19 fair-weather days with no BL cumulus formation were isolated for comparison from 11 months of data. Additionally, there was a lack of ancillary observations that would have been useful for the confirmation of the MLH estimates. Both ground-level meteorological observations as well as vertical profiles would have allowed for an objective analysis of the results presented here. As it stands, the assessment of the algorithm results were, to a certain extent, subjective. It was found that the automatic flagging algorithm developed for this project was able to remove almost all unsuitable situations from the data-set before the MLH algorithms were applied. However, the flagging algorithm was not able to distinguish backscatter signals due to the presence of haze or residual elevated condensation from that due to ground-sourced aerosols. Because of this, an additional manual flagging of the data was required in order to avoid erroneous MLH estimates. Of the two MLH algorithms, the ideal-profile technique was more robust and less likely to provide erroneous estimates. It was found that for situ- ations with a relatively “clean” ML, the minimum-gradient method often assigned the MLH to large noise structures within the higher altitudes. Yet, when considering periods with a strong ML signal, the two methods showed strong agreement. However, the ideal-profile methods was much more com- putationally expensive, and due to time constraints, the algorithm was not always allowed to converge on a final result, leading to a lack of precision in the estimates. However, for convective periods characterised by fair-weather and a tur- bid ML, both MLH algorithms produced results which agreed with a qual- 109 Chapter 5. Conclusions itative assessment of the data. This confirms the results from previous assessments of the ceilometer (Emeis & Schafer, 2006; Eresmaa et al., 2006; Zephoris et al., 2005). It should be mentioned that the issue of a poor signal to noise ratio, which was reported by previous authors, was less of a problem in this context, since the MLHs were relatively low. Overall, daily maximum MLH values within Vancouver ranged from 650m in the summer to 350m during the winter. An average day saw a growth in the MLH for the first 6 to 8 hours after sunrise, with a relatively constant or decreasing MLH during the afternoon. However, due to the small sample size (19 days), few significant statements regarding the diurnal trend could be made, other than the fact that the winter months showed a notable lag in MLH growth during the morning when compared to the summer months. The estimates made in this project are in general agreement with previous MLH estimates within the LFV. Of most significance is the agreement between the present results and the MLH estimates made by Steyn (1980) using sodar data, which were in turn confirmed by tethersonde ascents. 5.2 Future work Results from this project, and others before it, have shown the ceilometer’s ability to make uninterrupted measurements with little supervision over sev- eral years. Therefore, it is likely that the ceilometer will be able to provide significantly more information to what has already been presented in this project. This will allow for increased confidence in the yearly trends of both the daily maximum MLHs and the diurnal variations, and the development 110 Chapter 5. Conclusions of a MLH climatology for Vancouver. As it stands now, there are only a few significant statements than can be made about the yearly MLH trends. As of May 2008, A full suite of meteorological observations as well as radiative and turbulent heat fluxes are being collected at the Sunset Tower site. These co-located ancillary datasets will allow for a more accurate and objective analysis of the ceilometer data. For instance, sensible heat flux measurements will allow for a more accurate assessment of the agreement between measured MLH and theoretical MLH values, especially during the morning growth period. The turbulent heat flux measurements will also allow for an analysis of the MLH at a timescale on the order of a minute. Thermal structure suggested by the ceilometer data, such as the impinge- ment of individual thermal plumes upon the elevated inversion within the entrainment zone, may be confirmed by analysing sensible heat flux time- series. A more accurate and objective flagging algorithm may also be de- veloped. For instance, humidity measurements, which are being made at a height of 30m, may allow for the flagging of periods in which condensation within the BL would lead to erroneous MLH estimates. As well, by using sensible heat measurements, a precise “convective flag” can be developed which would accurately isolate convective periods. It may also be useful to perform a series of daytime tethersonde flights, as the MLH estimates made during this project were not confirmed via direct measurements. It may also be possible to make mobile transects of the LFV, which would allow for the spatial variation in the MLH to be analysed. For instance, the development of a thermal internal boundary layer with distance from the coast could be measured, and may prove useful 111 Chapter 5. Conclusions for model validation. 112 Bibliography Atkinson, B.W. 1981. Meso-scale Atmospheric Circulation. London: Aca- demic Press. Banta, R.M., Shepson, P.B., Bottenheim, J.W., Anlauf, K.G., Wiebe, HA, Gallant, A., Biesenthal, T., Olivier, L.D., Zhu, C.J., McKendry, I.G., et al. 1997. Nocturnal cleansing flows in a tributary valley. Atmospheric Envi- ronment, 31(14), 2147–2162. Barry, R.G. 1992. Mountain Weather and Climate. Cambridge: Routledge. Batchvarova, E., Cai, X. M., Gryning, S. E., & Steyn, D. 1999. Modelling internal boundary-layer development in a region with a complex coastline. Boundary-Layer Meteorology, 90(1), 1–20. Boudries, H., Canagaratna, M. R., Jayne, J. T., Alfarra, M. R., Allan, J., Bower, K. N., Coe, H., Pryor, S. C., Jimenez, J. L., Brook, J. R., Li, S., & Worsnop, D. R. 2004. Chemical and physical processes controlling the distribution of aerosols in the Lower Fraser Valley, Canada, during the Pacific 2001 field campaign. Atmospheric Environment, 38(34), 5759–5774. Brook, J.R., Strawbridge, KB, Snyder, BJ, Boudries, H., Worsnop, D., Sharma, S., Anlauf, K., Lu, G., & Hayden, K. 2004. Towards an under- 113 Bibliography standing of the fine particle variations in the LFV: integration of chemi- cal, physical and meteorological observations. Atmospheric Environment, 38(34), 5775–5788. Cai, X. M., & Steyn, D. G. 2000. Modelling study of sea breezes in a complex coastal environment. Atmospheric Environment, 34(18), 2873– 2885. Charron, A., Harrison, R.M., Moorcroft, S., & Booker, J. 2004. Quanti- tative interpretation of divergence between PM10 and PM2. 5 mass mea- surement by TEOM and gravimetric (Partisol) instruments. Atmospheric Environment, 38(3), 415–423. Chen, J.M., & Oke, T.R. 1994. Mixed-layer heat advection and entrainment during the sea breeze. Boundary-Layer Meteorology, 68(1), 139–158. Cooper, D.I., & Eichinger, W.E. 1994. Structure of the atmosphere in an urban planetary boundary layer from lidar and radiosonde observations. Journal of Geophysical Research. D. Atmospheres, 99, 22. Davis, KJ, Gamage, N., Hagelberg, CR, Kiemle, C., Lenschow, DH, & Sullivan, PP. 2000. An objective method for deriving atmospheric structure from airborne lidar observations. Journal of Atmospheric and Oceanic Technology, 17(11), 1455–1468. de Haij, M., Wauben, W., & Baltink, H.K. 2006. Determination of mixing layer height from ceilometer backscatter profiles. In: Slusser, J.R., Schfer, K., & Comern, A. (eds), Remote Sensing of Clouds and the Atmosphere XI, Proc SPIE, vol. 6362. 114 Bibliography Emeis, S., & Schafer, K. 2006. Remote sensing methods to investigate boundary-layer structures relevant to air pollution in cities. Boundary- Layer Meteorology, 121(2), 377–385. Emeis, S., Münkel, C., Vogt, S., Muller, W. J., & Schafer, K. 2004. Atmo- spheric boundary-layer structure from simultaneous SODAR, RASS, and ceilometer measurements. Atmospheric Environment, 38(2), 273–286. Eresmaa, N., Karppinen, A., Joffre, S. M., Rasanen, J., & Talvitie, H. 2006. Mixing height determination by ceilometer. Atmospheric Chemistry And Physics, 6, 1485–1493. Garratt, J.R. 1992. The Atmospheric Boundary Layer. Cambridge: Cam- bridge University Press. GVRD. 2006. Lower Fraser Valley air quality report. Tech. rept. Greater Vancouver Regional District. Hageli, P., Steyn, D. G., & Strawbridge, K. B. 2000. Spatial and temporal variability of mixed-layer depth and entrainment zone thickness. Boundary- Layer Meteorology, 97(1), 47–71. 29. Hayden, K.L., Anlauf, K.G., Hoff, R.M., Strapp, J.W., Bottenheim, J.W., Wiebe, H.A., Froude, F.A., Martin, J.B., Steyn, D.G., & McKendry, I.G. 1997. The vertical chemical and meteorological structure of the boundary layer in the Lower Fraser Valley during Pacific’93. Atmospheric Environ- ment, 31(14), 2089–2105. 115 Bibliography Hoff, RM, Harwood, M., Sheppard, A., Froude, F., Martin, JB, & Strapp, W. 1997. Use of airborne lidar to determine aerosol sources and movement in the Lower Fraser Valley (LFV), BC. Atmospheric Environment, 31(14), 2123–2134. Maletto, A., McKendry, I. G., & Strawbridge, K. B. 2003. Profiles of par- ticulate matter size distributions using a balloon-borne lightweight aerosol spectrometer in the planetary boundary layer. Atmospheric Environment, 37(5), 661–670. McKendry, I. G. 2000. PM10 levels in the Lower Fraser Valley, British Columbia, Canada: An overview of spatiotemporal variations and meteo- rological controls. Journal of the Air and Waste Management Association, 50(3), 443–452. McKendry, I. G., & Lundgren, J. 2000. Tropospheric layering of ozone in regions of urbanized complex and/or coastal terrain: A review. Progress in Physical Geography, 24(3), 329–354. McKendry, I. G., Sturman, A. P., & Vergeiner, J. 2004. Vertical profiles of particulate matter size distributions during winter domestic burning in Christchurch, New Zealand. Atmospheric Environment, 38(29), 4805–4813. Mckendry, I.G. 1994. Synoptic Circulation and Summertime Ground-Level Ozone Concentrations at Vancouver, British Columbia. Journal of Applied Meteorology, 33(5), 627–641. McKendry, IG, Steyn, DG, Lundgren, J., Hoff, RM, Strapp, W., Anlauf, K., Froude, F., Martin, JB, Banta, RM, & Olivier, LD. 1997. Elevated 116 Bibliography ozone layers and vertical down-mixing over the Lower Fraser Valley, BC. Atmospheric Environment, 31(14), 2135–2146. McKendry, I.G., Strawbridge, K., ONeill, N., McDonald, A-M, Liu, P., Jaegle, L., Jaffe, D., Fairlie, D., Anlauf, K, & Westphal, D. 2007. A case of trans-Pacific transport of saharan dust to Western North America. Journal of Geophysical Research, 112, 1277 – 1290. Menut, L., Flamant, C., Pelon, J., & Flamant, P. H. 1999. Urban boundary- layer height determination from lidar measurements over the Paris area. Applied Optics, 38(6), 945–954. Münkel, C., & Rasanen, J. 2004. New optical concept for commercial lidar ceilometers scanning the boundary layer. Pages 364 – 374 of: Sensors, systems, and next-generation satellites VIII, Proc SPIE, vol. 5571. Münkel, C., Emeis, S., Mller, W.J., & Schfer, K. 2004. Aerosol concentra- tion measurements with a lidar ceilometer: results of a one year measuring campaign. Pages 486–496 of: Schfer, K, Comeron, A, Carleer, M, & Pi- card, RH (eds), Remote sensing of clouds and the atmosphere VIII, Proc SPIE, vol. 5235. Münkel, C., Eresmaa, N., Rasanen, J., & Karppinen, A. 2007. Retrieval of mixing height and dust concentration with lidar ceilometer. Boundary- Layer Meteorology, 124(1), 117–128. Oke, T.R. 1995. The heat island of the urban boundary layer: Charac- teristics, causes and effects. Pages 81–107 of: Cermak, J.E. (ed), Wind Climate in Cities. Kluwer Academic. 117 Bibliography Oke, T.R., & Hay, J. 1994. The Climate of Vancouver. 2 edn. BC Geo- graphical Series, no. 50. Vancouver: Department of Geography, University of British Columbia. Rasanen, J., Lnnqvist, J., & Piironen, A.K. 2000. Urban boundary layer measurements with a commercial ceilometer. Pages 34 –35 of: Proc. 3rd Symp. on Urban Environment. Boston: American Meteorological Society. Reuten, C. 2006. Scaling and kinematics of daytime slope flow systems. Ph.D. thesis, University of British Columbia. Reuten, C., Steyn, D.G., Strawbridge, K.B., & Bovis, P. 2005. Observations of the relation between upslope flows and the convective boundary layer in steep terrain. Boundary-Layer Meteorology, 116, 37–61. Rogers, R. R., Lamoureux, M. F., Bissonnette, L. R., & Peters, R. M. 1997. Quantitative interpretation of laser ceilometer intensity profiles. Journal of Atmospheric and Oceanic Technology, 14(3), 396–411. 22 AMER ME- TEOROLOGICAL SOC Part 1. Salmond, J. A., & McKendry, I. G. 2005. A review of turbulence in the very stable nocturnal boundary layer and its implications for air quality. Progress In Physical Geography, 29(2), 171–188. Schafer, K., Emeis, S., Rauch, A., Münkel, C., & Vogt, S. 2004. Deter- mination of mixing layer heights from ceilometer data. Pages 248 – 259 of: Schfer, K, Comeron, A, Carleer, M, Picard, RH, & N, Sifakis (eds), Remote sensing of clouds and the atmosphere IX, Proc SPIE, vol. 5571. 118 Bibliography Schwiesow, R.L. 1986. Lidar measurments of boundary-layer variables. Pages 139–162 of: Lenschow, D.H. (ed), Probing The Atmospheric Bound- ary Layer. American Meteorological Society. Seibert, P., Beyrich, F., Gryning, S. E., Joffre, S., Rasmussen, A., & Tercier, P. 2000. Review and intercomparison of operational methods for the determination of the mixing height. Atmospheric Environment, 34(7), 1001–1027. Snyder, B.J., & Strawbridge, K.B. 2004. Meteorological analysis of the Pacific 2001 air quality field study. Atmospheric Environment, 38(34), 5733–5743. Steyn, D. 1998. Scaling the vertical structure of sea breezes. Boundary- Layer Meteorology, 86(3), 505 – 524. Steyn, D. G., Baldi, M., & Hoff, R. M. 1999. The detection of mixed layer depth and entrainment zone thickness from lidar backscatter profiles. Journal Of Atmospheric And Oceanic Technology, 16(7), 953–959. Steyn, D.G. 1980. Turbulence, diffusion and the daytime mixed-layer depth over a coastal city. Ph.D. thesis, University of British Columbia. Steyn, D.G., & Oke, T.R. 1982. The depth of the daytime mixed layer at two coastal sites: A model and its validation. Boundary-Layer Meteorology, 24(2), 161–180. 119 Bibliography Steyn, D.G., & Wallis, C.H. 1987. Spatial and temporal variability of the daytime mixed layer depth over Greater Vancouver. Tech. rept. Department or Geography, University of British Columbia. Strawbridge, K. B., & Snyder, B. J. 2004a. Daytime and nighttime aircraft lidar measurements showing evidence of particulate matter transport into the Northeastern valleys of the Lower Fraser Valley, BC. Atmospheric Environment, 38(34), 5873–5886. Strawbridge, K. B., & Snyder, B. J. 2004b. Planetary boundary layer height determination during Pacific 2001 using the advantage of a scanning lidar instrument. Atmospheric Environment, 38(34), 5861–5871. Stull, R.B. 1988. An Introduction to Boundary Layer Meteorology. San Diego: Academic. Wiegner, M., Emeis, S., Freudenthaler, V., Heese, B., Junkermann, W., Munkel, C., Schafer, K., Seefeldner, M., & Vogt, S. 2006. Mixing layer height over Munich, Germany: Variability and comparisons of different methodologies. Journal Of Geophysical Research-Atmospheres, 111(D13), 17. Zephoris, M., Holin, H., Lavie, F., Cenac, N., Cluzeau, M., Delas, O., Eideliman, F., Gagneux, J., Gander, A., & Thibord, C. 2005. Ceiloffi- eter observations of aerosol layer structure above the Petit Luberon during ESCOMPTE’s IOP 2. Atmospheric Research, 74(1-4), 581–595. 120