UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Summertime horizontal and vertical advective carbon dioxide fluxes measured in a closed-canopy Douglas-fir… Leitch, Adrian 2010

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

Item Metadata

Download

Media
24-ubc_2010_fall_leitch_adrian.pdf [ 3.54MB ]
Metadata
JSON: 24-1.0052565.json
JSON-LD: 24-1.0052565-ld.json
RDF/XML (Pretty): 24-1.0052565-rdf.xml
RDF/JSON: 24-1.0052565-rdf.json
Turtle: 24-1.0052565-turtle.txt
N-Triples: 24-1.0052565-rdf-ntriples.txt
Original Record: 24-1.0052565-source.json
Full Text
24-1.0052565-fulltext.txt
Citation
24-1.0052565.ris

Full Text

SUMMERTIME HORIZONTAL AND VERTICAL ADVECTIVE CARBON DIOXIDE FLUXES MEASURED IN A CLOSED-CANOPY DOUGLAS-FIR FOREST ON A SLOPE  by  Adrian Leitch  B.Sc., The University of British Columbia, 2008  A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF  MASTER OF SCIENCE  in  The Faculty of Graduate Studies  (Atmospheric Science)  THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver)  August 2010  © Adrian Leitch, 2010  Abstract An observational study was conducted during an eight-week period in the summer of 2009 at a forested site in British Columbia (“DF49”) to determine the incidence and magnitude of advective carbon dioxide (CO2) fluxes. The site is situated in a tall, closed-canopy (zh = 35 m, 1100 stems ha-1, LAI of 7 m2 m-2) Douglas-fir forest on a 5-10o slope. Katabatic flow occurs within the subcanopy during a significant portion of the day, especially during the evening and night. Wind vector and CO2 concentration measurements were made at heights of 1, 2.6, 8, 20 and 42 m on the main flux tower, while CO2 concentration measurements were made 73.5 m upslope from the tower at heights of 1, 2.6, 8 and 20 m, creating a two-dimensional transect. Eight thermostated non-differential infrared gas analyzers (IRGAs) were paired using a solenoid valve switching system, creating four differential analyzers. The switching system, combined with a twice-daily polynomial IRGA calibration, enabled measurement of CO2 concentration differences at the four subcanopy measurement heights to an accuracy of 0.5 µmol mol-1 (horizontal) and 1 µmol mol-1 (vertical). The system also recorded semi-continuous CO2 traces at all sample points. Horizontal advective fluxes, averaged over the main measurement period (August 2 - 25), reached 5 µmol m-2 s-1 at night and 10 µmol m-2 s-1 in the afternoon (a net source to the atmosphere) and -2 µmol m-2 s-1 during the morning (net sink). Vertical advective fluxes averaged around zero during the night, 5 µmol m-2 s-1 during the morning and -5 µmol m-2 s-1 during the afternoon. A cumulative sum of advective flux-corrected net ecosystem exchange (NEE) over the main measurement period (-57 g C m-2) was close to that of the friction velocity-corrected NEE (-60 g C m-2), both predicting less carbon uptake by the Douglas-fir forest than the storage-corrected NEE (-86 g C m-2). Additional CO2 difference measurements made after the main sampling period along the 73.5-m transect demonstrated that horizontal CO2 differences increased monotonically down slope (on average) at the 2.6-m height, despite an opposing gradient in manual chamber-measured soil CO2 effluxes.  ii  Table of Contents Abstract ............................................................................................................................ii Table of Contents ............................................................................................................ iii List of Tables ................................................................................................................... v List of Figures ..................................................................................................................vi List of Symbols, Acronyms and Abbreviations ................................................................ x Acknowledgments .......................................................................................................... xii 1 Introduction .................................................................................................................. 1 1.1  Eddy covariance................................................................................................. 1  1.2  Scalar exchange in a fluid .................................................................................. 3  1.3  Common flux values ........................................................................................... 5  1.4  Complex terrain .................................................................................................. 6  1.5  Advection studies ............................................................................................... 8  1.6  Study objectives ............................................................................................... 14  2 Methods ..................................................................................................................... 15 2.1  Site description................................................................................................. 15  2.2  Long-term monitoring instrumentation .............................................................. 16  2.3  Experimental layout .......................................................................................... 18  2.4  Sampling point placement ................................................................................ 21  2.5  Advection flux instrumentation ......................................................................... 22  2.6  High-frequency data collection ......................................................................... 24  2.7  Thermostated infrared gas analyzer box .......................................................... 24  2.8  Sonic anemometer offset calibration ................................................................ 28  2.9  Field IRGA calibration ...................................................................................... 29  2.10  System power ............................................................................................... 31  2.11  Data processing ............................................................................................ 32  2.11.1  Software IRGA calibration ...................................................................... 33  2.11.2  Wind vector coordinate system rotation ................................................. 34  2.11.3  Data quality control ................................................................................. 35  2.11.4  Advection flux computations ................................................................... 35  2.12  Comparison with long-term flux instrumentation ........................................... 37  2.13  Soil CO2 efflux measurements ...................................................................... 39 iii  3 Results and Discussion .............................................................................................. 40 3.1  First experiment ............................................................................................... 40  3.2  Second experiment .......................................................................................... 42  3.2.1  Temperature and radiation climatology...................................................... 43  3.2.1  Wind direction and speed .......................................................................... 45  3.2.2  CO2 concentration ..................................................................................... 52  3.2.3  CO2 concentration differences ................................................................... 55  3.2.4  Advective CO2 fluxes ................................................................................. 58  3.2.5  Cumulative carbon balance ....................................................................... 63  3.3  Third experiment .............................................................................................. 68  3.4  Soil efflux.......................................................................................................... 72  4 Conclusions................................................................................................................ 76 Works Cited ................................................................................................................... 79 Appendices ................................................................................................................... 89 Appendix 1: Switching system mathematics .......................................................... 89 Appendix 2: Westham Island sonic anemometer comparison ................................ 94 Appendix 3: Post-experiment IRGA calibration ...................................................... 96 Appendix 4: High-frequency time evolution of morning subcanopy CO 2 concentration.......................................................................................................... 97 Appendix 5: Understory leaf CO2 exchange ......................................................... 102 Appendix 6: Site pictures ..................................................................................... 103 Appendix 7: IRGA calibration period .................................................................... 106 Appendix 8: Transect map ................................................................................... 107  iv  List of Tables Table 1: Survey of studies reporting horizontal and vertical advective CO 2 fluxes measured in forest control volumes. ...................................................................... 10 Table 2: Sonic anemometer zero offsets determined in the laboratory both before and after the summer of 2009. ...................................................................................... 29 Table 3: FHA and FVA (PFM) statistics for the second experiment (µmol m-2 s-1). .......... 62 Table 4: Cumulative carbon balance values, with various corrections (g C m-2). .......... 68 Table 5: Sonic anemometer zero offsets calculated from Westham Island, spring of 2009. ...................................................................................................................... 95 Table 6: Sonic anemometer zero offsets calculated from Westham Island, spring of 2009. Analysis performed by Kate Liss and Andreas Christen. ............................ 95 Table 7: Post-experiment IRGA calibration results. ....................................................... 96  v  List of Figures Figure 1: Map of Vancouver Island, B.C., Canada. ....................................................... 16 Figure 2: Topographic map of DF49.............................................................................. 17 Figure 3: First experiment. ............................................................................................ 19 Figure 4: Second experiment. ....................................................................................... 20 Figure 5: Third experiment. ........................................................................................... 20 Figure 6: Thermostated box in the laboratory................................................................ 25 Figure 7: IRGAs in thermostated box at DF49. ............................................................. 25 Figure 8: Outside air and thermostated box temperatures. ........................................... 26 Figure 9: Control signal to the heat sink fans. ............................................................... 27 Figure 10: IRGA system component diagram ............................................................... 31 Figure 11: Vertical velocity corrected with the PFM and TAM methods. ....................... 34 Figure 12: Comparison of CSAT3 tower profile with site RM-Young. ............................ 38 Figure 13: Wind direction and speed. ............................................................................ 41 Figure 14: Wind direction and speed for a subset of the first experiment. ..................... 41 Figure 15: Air temperature. ........................................................................................... 43 Figure 16: Above-canopy net radiation.......................................................................... 44 Figure 17: Ensemble average of the diurnal variation in potential temperature. ............ 45 Figure 18: Vertical profile of buoyancy flux. ................................................................... 46 Figure 19: Vertical profile of buoyancy flux, showing the subcanopy only. .................... 46 Figure 20: Ensemble average vector mean wind direction. ........................................... 47 Figure 21: Subcanopy ensemble average of the standard deviation in wind direction. . 48 vi  Figure 22: Ensemble average of wind speed. ............................................................... 49 Figure 23: Ensemble average of wind speed within the subcanopy. ............................. 49 Figure 24: Standard deviation of wind speed in the subcanopy. ................................... 50 Figure 25: Ensemble average friction velocity profile. ................................................... 50 Figure 26: Half-hour average CO2 concentration. ......................................................... 53 Figure 27: Ensemble average CO2 concentration. ........................................................ 53 Figure 28: Ensemble average of the standard deviation of CO 2 concentration. ............ 54 Figure 29: Subcanopy horizontal CO2 concentration differences on August 4 th, 2009. . 56 Figure 30: Subcanopy horizontal CO2 concentration differences on August 9 th, 2009. . 56 Figure 31: Ensemble average subcanopy horizontal CO 2 concentration. ..................... 57 Figure 32: Standard deviation in horizontal CO2 concentration. .................................... 57 Figure 33: Ensemble average vertical CO2 concentration differences. ......................... 58 Figure 34: Ensemble average FHA and FVA (PFM) as well as storage-corrected NEE... 59 Figure 35: Ensemble average FHA shown with error bars.............................................. 60 Figure 36: Ensemble average FVA (PFM) shown with error bars. .................................. 60 Figure 37: Histogram of FHA. ......................................................................................... 61 Figure 38: Histogram of FVA (PFM)................................................................................ 61 Figure 39: Cumulative FHA fluxes by subcanopy level. .................................................. 63 Figure 40: Cumulative subcanopy and EC storage fluxes. ............................................ 64 Figure 41: Ensemble average storage flux, advection corrected- and u*-corrected NEE. ............................................................................................................................... 65 Figure 42: Cumulative carbon balance over the 23-day main experiment, with corrections.............................................................................................................. 66  vii  Figure 43: Cumulative carbon balance over the 23-day main experiment, showing various wind velocity tilt corrections. ...................................................................... 66 Figure 44: Cumulative FVA using the PFM and TAM. .................................................... 68 Figure 45: Ensemble average horizontal CO 2 differences. ............................................ 70 Figure 46: Standard deviation of horizontal CO 2 concentration differences. ................. 70 Figure 47: Standard deviation of CO2 concentration. .................................................... 71 Figure 48: Soil CO2 efflux during the three experiments. .............................................. 73 Figure 49: Auto-chamber soil CO2 efflux. ...................................................................... 73 Figure 50: Cumulative soil CO2 efflux over the second experiment. .............................. 74 Figure 51: Cumulative trace of FHA over the course of the second experiment. ............ 74 Figure 52: Example of CO2 concentration traces from individual IRGAs while switching. ............................................................................................................................... 90 Figure 53: Switching difference comparison, 6 minute timescale. ................................. 92 Figure 54: Switching difference comparison, 30 minute timescale. ............................... 92 Figure 55: Ensemble average difference comparison, 6 minute timescale. .................. 93 Figure 56: Ensemble average difference comparison, 30 minute timescale. ................ 93 Figure 57: Westham Island sonic anemometer calibration setup. ................................. 94 Figure 58: Subcanopy CO2 concentration profile, 5:30 – 6:00 PST............................... 97 Figure 59: Subcanopy CO2 concentration profile, 6:00 – 6:30 PST............................... 98 Figure 60: Subcanopy CO2 concentration profile, 6:30 – 7:00 PST............................... 98 Figure 61: Subcanopy CO2 concentration profile, 7:00 – 7:30 PST............................... 99 Figure 62: Subcanopy CO2 concentration profile, 7:30 – 8:00 PST............................... 99 Figure 63: Subcanopy CO2 concentration profile, 8:00 – 8:30 PST............................. 100  viii  Figure 64: Subcanopy CO2 concentration profile, 8:30 – 9:00 PST............................. 100 Figure 65: Subcanopy CO2 concentration profile, 9:00 – 9:30 PST............................. 101 Figure 66: Understory net leaf exchange. ................................................................... 102 Figure 67: Instrumentation table with calibration gas tanks. ........................................ 103 Figure 68: Flow control and electronics for the IRGA system. ..................................... 103 Figure 69: DF49 canopy with upslope 1-m and 2.6-m sampling points. ...................... 104 Figure 70: Upslope co-located sampling points during the third experiment. .............. 104 Figure 71: Calibration gas distribution box. ................................................................. 105 Figure 72: Calibration period IRGA CO2 differences and error between the four systems. ............................................................................................................................. 106 Figure 73: Plan view of the advection transect during the third experiment. ............... 107  ix  List of Symbols, Acronyms and Abbreviations  Symbol  Units  Description  °  degrees  Slope or compass angle  AC  --  Alternating current  C  µmol mol-1  Scalar mixing ratio  CO2  --  Carbon dioxide  CSI  --  Campbell Scientific, Inc.  D  µmol mol-1  CO2 difference  DC  --  Direct current   mn  --  Kronecker delta  ε  µmol mol-1  Offset error   ijk  --  Alternating unit tensor  EC  --  Eddy covariance  FC  µmol m-2 s-1  Vertical turbulent flux of CO2  FS  µmol m-2 s-1  Storage flux  FHA  µmol m-2 s-1  Horizontal advective flux  FVA  µmol m-2 s-1  Vertical advective flux  g  m s-2  Acceleration due to gravity  IRGA  --  (Non-dispersive) infrared gas analyzer  LAI  m2 m-2  Leaf area index  NEE  µmol m-2 s-1 or g C m-2  Net ecosystem exchange  j  s-1  Angular velocity vector of the earth‟s rotation  PST  --  Pacific Standard Time  x  Symbol  Units  Description  PAR  µmol m-2 s-1  Photosynthetically active radiation  p  Pa  Pressure  PFM  --  Planar-fit method  ρ, ρda  kg m-3  Air density, dry air density  S  m s-1  Arithmetic mean wind speed  SC  µmol mol-1 s-1  Source of scalar C  σCO2, σWD  µmol mol-1, °  Standard deviation  SN  --  Serial number  t  s  Time   ij  N m-2  Viscous shear stress tensor  θ  degrees  Wind direction  TAM  --  Tilt angle method  Uj  m s-1  Wind speed, three-dimensional vector  U  m s-1  Vector mean wind speed  u*  m s-1  Friction velocity  u  m s-1  Wind speed along x  v  m s-1  Wind speed along y  w  m s-1  Wind speed along z  x  m  Distance, x coordinate (horizontal along-wind)  y  m  Distance, y coordinate (horizontal cross-wind)  z  m  Distance, z coordinate (vertical)  xi  Acknowledgments This work would not have been possible without the financial support of the Salt Spring Apple Festival, the University of British Columbia, Environment Canada, the Canadian Carbon Program, the Canadian Foundation for Climate and Atmospheric Sciences, and the Natural Sciences and Engineering Research Council (NSERC). Support from NSERC included a CGS-M grant to Adrian Leitch, Discovery Grants to Andreas Christen (342029-07) and Andy Black, an RTI grant to Andreas Christen (344541-07) and a Strategic Grant to Andy Black and others. Mike Novak, Rachhpal Jassal, Douw Steyn, Andreas Christen and Andy Black provided essential academic guidance. Christian Brümmer provided both academic and technical support, as well as experience from his own advection experiment. Silvia Renata-Motta, Nick Grant, Kate Liss, Andrew Hum, Dominic Lessard, Rick Ketler and Zoran Nesic made the experiment happen. Finally, family members were a constant source of support.  xii  1 Introduction This thesis is concerned with measuring the exchange of matter, energy and momentum between the land surface and the atmosphere due to wind. The technique involved (eddy-covariance or EC) can theoretically be applied over any surface, whether a water body, agricultural field, forest or city. EC measures the transport, or flux, of scalars or vectors (e.g. water vapour and momentum) through a plane above the surface in question. EC can also be used to estimate the source or sink strength of these quantities beneath the measurement plane. The observational study reported in this thesis focuses on the exchange of matter between a forested slope and the atmosphere, with the twin goals of determining fluxes of carbon dioxide (CO2) and the cumulative carbon balance of the forest. In particular, this work focuses on a suspected systematic error that occurs when the EC technique is used over a forested slope, especially at night.  1.1  Eddy covariance  The EC technique, developed in the middle of the previous century (e.g. Baldocchi et al., 1988) can be applied to investigate an ecosystem‟s carbon balance at a range of scales, though a square kilometer is most common. Continuous measurements of CO2 exchange between an ecosystem and the atmosphere is possible at a minimum time scale of around half an hour. The time series of CO2 flux can be correlated with simultaneously measured energy and mass budgets (e.g. 1  photosynthetically active radiation (PAR) and soil water content) over several days, and yearly carbon sequestration totals can be compared with climatic variability on a decadal scale. Though it is difficult to control the parameters that govern an ecosystem‟s carbon uptake, natural variations in environmental variables on diurnal and annual cycles provide an opportunity to discover their effects on an ecosystem‟s balance between photosynthesis and respiration.1 Ecosystem fluxes (including CO2, water vapour and heat) using the EC technique have been measured at several sites since the early 1990s (e.g. Wofsy et al., 1993; Goulden et al., 1996; Black et al., 1996), and many newer flux towers have since been established („FLUXNET‟, Baldocchi et al., 2001; Christen et al., 2009). The global FLUXNET network consists of over 500 sites that use or have used the EC technique to measure ecosystem fluxes around the world (Baldocchi, 2008). While the EC technique adequately characterizes the carbon balance of many sites, measured fluxes may not be accurate under certain conditions. These include a wind direction from an undesirable sector (Kljun et al., 2004), stable atmospheric stratification (Massman and Lee, 2002), or complex topography (Finnigan and Belcher, 2006; Wang, 2009). It is also inherently an atmospheric measurement technique, and so may not capture other fluxes of carbon into or out of the system of concern (e.g. through hydrological processes – Johnson et al., 2006). More specifically, though established theory holds that an EC site must be flat with a homogeneous vegetation cover, several monitoring sites are situated on slopes  1  Experiments that manipulate water and nutrient availability are common exceptions.  2  or near significant land-use changes. In the case of a forested slope, advective flow down the incline may occur in the layer of air near the forest floor, leading to horizontal transport of respired CO2 from the measurement (or “control”) volume under consideration beneath the EC sensors. As photosynthesis (ecosystem CO2 gain) occurs primarily in the crown space and respiration (CO2 loss) occurs mainly in the soil, failure to account for advection could lead to a systematic measurement error. The respiration signal may therefore be underestimated, resulting in an overestimate of the ecosystem‟s carbon gain. This error and others can be more fully understood by examining the scalar conservation equation underlying the EC method, outlined in the following section.  1.2  Scalar exchange in a fluid  Newton‟s second law, which describes the conservation of momentum, has the following form when applied to air (Stull, 1988):  U i U i 1 p 1  ij Uj   i 3 g  2 ijk  j U k   t x j  x i  x j I  II  III  IV  V  (1)  VI  (Symbols are defined in the List of Symbols, Acronyms and Abbreviations.) Transport of a quantity by wind occurs due to the forces that control the wind itself. The main processes at the canopy scale include diffusion due to viscous stress (term VI) and convection due to buoyancy („free convection‟, term III) or pressure gradients („forced convection‟, term V). The Coriolis force is frequently neglected (term IV).  3  The scalar conservation equation expresses these processes more concisely as:  c c  2c Uj   C 2  SC t x j x j I  II  III  (2)  IV  Term I indicates storage of a quantity C within a volume, term II refers to transport of C due to its gradient along some direction by a wind U, term III refers to diffusion and term IV represents a source or sink of C. The atmosphere is usually at least somewhat in motion, leading many scientists to neglect the diffusion term (III) on the assumption that it is small with respect to the others. Motion or scalar concentration is also currently subdivided into long-term and short-term components, achieved through „Reynolds averaging‟. Separation is achieved by taking the mean of a quantity over some time interval and subtracting this mean from the time series to generate instantaneous departures from the mean over the averaging interval. With the above two modifications, removing Einstein notation and considering the source term as the unknown, the scalar conservation equation becomes (Loescher et al., 2006):  SC   c u ' c ' v ' c ' w ' c ' u c v c w c       t x y z x y z  I  II  III  IV  V  VI  (3)  VII  The overbar represents a mean and the prime a departure from the mean. Term I is again the storage term, terms II - IV are turbulent (or „eddy‟) fluxes, and terms V - VII are advective fluxes. Integrating this equation with respect to z from the surface to a  4  measurement height zm gives the source strength of quantity C below zm, termed net ecosystem production (NEP) or net ecosystem exchange (NEE). For certain conditions (e.g. a horizontally homogeneous landscape), the horizontal transport terms may be negligible (II, III, V and VI). Vertical transport is restricted to the eddy flux if the surface is flat as the mean vertical velocity will average to zero (negating term VII). Rate of change of storage is also minimal in the air column below the instruments under strong winds (term I). Under such „ideal‟ conditions, the vertical turbulent flux (term IV) is used to estimate SC.  1.3  Common flux values  NEE has a clear diurnal trend within most forests – negative during the day when the photosynthetic flux is larger than the respiratory flux (the forest acting as a carbon sink) and positive at night when only ecosystem respiration occurs (with the forest as a CO2 source). In other words, CO2 is transported from the atmosphere in to the forest during the day (negative NEE) and from the forest to the atmosphere during the night (positive NEE).  Typical daytime NEE values from temperate forest ecosystems (measured with EC, assuming ideal conditions) are between 0 and -25 μmol m-2 s-1, and vary depending on soil water content, photosynthetically active radiation, soil temperature, atmospheric vapour pressure deficit and other variables. Peak averaged (summer) uptake is usually around -10 to -20 μmol m-2 s-1 during the middle of the day. In contrast, nighttime fluxes are typically under 10 μmol m-2 s-1 with an average around 5 μmol m-2 s-1. Early 5  morning and late evening transition periods occur as photosynthesis and respiration fluxes balance.  1.4  Complex terrain  Researchers situating instruments above a tall forest canopy, especially in complex terrain and over the course of at least several hours, must occasionally consider the importance of other terms in Equation 3. A common technique is the replacement of suspect measurements (especially at night) with those made during more ideal conditions using the friction velocity (u*) filter method (e.g. Goulden et al., 1996; Falge et al., 2001; Papale et al., 2006).  The storage flux (term I) is also commonly measured at many sites due to accumulation of CO2 beneath the measurement height at night when wind speeds are low (Yang et al., 1999; Finnigan, 2006; de Araujo et al., 2009). Morning and afternoon transition periods show the largest storage change values, due to a buildup of CO2 concentration during the evening as winds die down, and a corresponding reduction of the built-up CO2 as winds pick up again in the morning. The advective and horizontal flux divergence terms (V, VI, VII, II and III), in contrast, are not routinely measured at FLUXNET sites; scientists have recently suggested a change in this practice (Lee, 1998; Finnigan, 1999; Massman and Lee, 2002) to replace the u* filter method. Several observational studies to determine the feasibility of measuring the advection terms specifically (V, VI and VII) within the context  6  of FLUXNET have been carried out (e.g. Staebler and Fitzjarrald, 2004; Feigenwinter et al., 2004; Aubinet et al, 2005; Feigenwinter et al., 2008, Yi et al., 2008). Fewer measurements exist quantifying the horizontal turbulent flux divergence terms (II and III, e.g. Finnigan, 1999; Staebler & Fitzjarrald, 2004; Moderow et al., 2007; Montagnani et al., 2009). Estimates for the magnitude of horizontal turbulent fluxes range between 5-100% of the horizontal advective terms (Staebler & Fitzjarrald, 2004; Moderow et al., 2007; Sun et al., 2007), but no research group has published estimates of these terms so far with CO2 gas analyzers. Instead, the estimations have been done with thermocouples measuring sensible heat transfer (Staebler & Fitzjarrald, 2004; Moderow et al., 2007) and hygrometers measuring latent heat transfer (Sun et al., 2007). The current assumption that horizontal turbulent divergence of CO2 is small enough to ignore is therefore questionable; however, discussion in the following sections will focus on the advective terms. Equation 3 can therefore be given in terms of symbols for each commonlymeasured flux: NEE = FS + FC + FHA+ FVA  (4)  where FS is the storage flux, FC is the vertical turbulent flux, and the last two terms are horizontal (FHA, terms V and VI in eq. 3) and vertical (FVA, term VII in eq. 3) advective fluxes. A positive flux is directed away from the ecosystem in question; in other words, a positive FC indicates that turbulent eddies are transporting CO 2 out of the forest.  7  1.5  Advection studies  A number of observational studies have been conducted that have attempted to measure horizontal and vertical advective fluxes under the measurement conditions that make these terms important. The history of these studies and the results they have achieved are summarized in this section. Modeling and laboratory studies of wind flow in complex terrain (e.g. Mahrt, 1982; Fitzjarrald, 1984; Gudiksen et al., 1992; Finnigan and Belcher, 2004; Sun et al., 2006a and 2006b; Poggi et al., 2008; Dupont et al., 2008; Wang, 2009) are not covered. Field experiments considering drainage flow alone (e.g. Fitzjarrald 1986; King, 1989; Sun et al., 1998; Papadopoulis and Helmis, 1999; Mahrt et al., 2001; Staebler and Fitzjarrald, 2005; Pypker et al., 2007) are also neglected. CO2 advection experiments covered include those intended to capture two-dimensional and three-dimensional wind fields, measuring either in a horizontal network or a vertical network of layers. All literature reviewed here reports data from forested sites.  The first paper that explored advective flux divergence did not report data from a specifically designed sampling plan. Rather, Lee (1998) analyzed vertical velocity data from two EC towers to investigate FVA. This was done by changing the common coordinate-rotation method of the time that set the average vertical velocity to zero (implicitly assuming zero vertical advection; e.g. Tanner and Thurtell, 1969; McMillen, 1988) to a „non-planar fit method‟ (see also Wilczak et al., 2001; Heinesch et al., 2007) that calculated the average vertical wind vector for one degree azimuthal increments (out of 360°) during the entire measurement period. This method determined the mean wind streamline for every wind direction, allowing for detection of significant vertical  8  velocity departures from this mean for any particular half hour average. Lee showed that vertical velocities calculated this way averaged around zero for the day, but were biased towards negative values (i.e. sinking air over the forest) at night. Though the velocities calculated were close to the error of the sonic anemometer, they suggested that the FVA term could be significant. Lee assumed that FHA was small enough to be negligible – i.e., he was still working within a one-dimensional (vertical) framework for the scalar conservation equation.  Finnigan (1999) argued, in a response to Lee (1998), that FHA could not be neglected except under very special circumstances (i.e., if the tower were located underneath the vertical stagnation streamline of a recirculating flow). Finnigan stated that while the change in vertical CO2 concentration (away from a surface) is usually much greater than in the horizontal, the average vertical wind speed is (in contrast) much less than that in the horizontal, making it likely that the FHA and FVA terms are of similar magnitude in a typical forest control volume. While two later papers (Baldocchi et al., 2000 and Paw U et al., 2000) only explored the magnitude of FVA, almost all other advection studies subsequently presented in the literature provide an analysis of FHA (Mamarella et al., 2007 being an exception).  Over the next ten years, sampling designs progressed from a horizontal network of CO2 intakes allied with a 5-m high CO2 profile and the flux tower itself (Staebler and Fitzjarrald, 2004) to four canopy-height towers arranged in a rectangular prism around the flux tower, also connected by a horizontal network (Feigenwinter et al., 2008, 2010a and 2010b; Montagnani et al., 2009; Aubinet et al., 2010). Researchers at the Niwot  9  Table 1: Survey of studies reporting horizontal and vertical advective CO2 fluxes measured in forest control volumes. Study or Review  Dimensionality, Transect Length  FHA (+/–, % of NEE)  FVA (+/–, % of NEE)  Baldocchi et al., 2000  1D– tower top & vertical profile  ---  0 to 3 μmol m-2 s-1  Paw U et al., 2000  1D– tower top & vertical profile  ---  100% at night, small during the day.  Aubinet et al. 2003  2D, 55 metres  Measured at greater than 200%  Staebler and Fitzjarrald, 2004  3D, 60m square  50%  5%  Feigenwinter et al., 2004  3D, 50m length, triangle  –, same order of magnitude  +, same order of magnitude  Wang et al., 2005  2D, 20 km  -10%, only one wind direction  20%, only one wind direction  Aubinet et al., 2005  2D/3D, ~90 metres [6 sites]  0.1 to 5 μmol m s-1  Marcolla et al., 2005  2D, 90 metres  50%  Mammarella et al., 2007  1D – tower top & vertical profile  Heinesch et al. 2007, 2008  2D, 89 metres  -50% from NE, 0% from SW  Sun et al., 2007; Yi et al, 2008  3D, roughly 300m square  FHA +, FVA -, combination leads to 10% decrease in NEE monthly & 65% decrease in NEE annually.  Feigenwinter et al., 2008  3D, 100m rectangular prism  [3 sites], + except for flat site, same order of magnitude  [3 sites], +, same order of magnitude  Leuning et al., 2008  3D, 50m square  35%  125%  Montagnani et al., 2009  3D, 100m rectangular prism  Combined advective term: 9% during the day and 190% during the night  Etzold et al., 2010  2D, 60 metres  40%  -2  ---  -2  -1  -5 to -10 μmol m s  22%  ---  0 to 435%, depending on u* 500% from NE, 0% from SW  Daily average close to zero  10  Ridge site in Boulder, CO, USA have used eight towers arranged in an alpine forest in an attempt to quantify advective flows (e.g. Sun et al., 2007 or Yi et al., 2008), and a study performed in Tumbarumba, NSW, Australia, used a network of irrigation tubing to estimate line averaged advective fluxes in the lowest 6 m of a Eucalyptus forest (Leuning et al., 2008). Though these studies measured advection in a three-dimensional context, there have also been several attempts to measure advection using a reduced twodimensional design. This has been possible due to constrained flow regimes that develop at certain sites – e.g. land/sea or mountain/valley breezes – that allow for a reduction in the number of instruments required to measure an along-wind CO2 concentration difference. Two dimensional advection flux studies have been performed at the Vielsalm site in the Walloon Region, Belgium (e.g. Aubinet et al., 2003 and 2005; Heinesch et al., 2007 and 2008) and the Renon (Ritten) site in Bolzano, Italy (Marcolla et al., 2005). The issue of directional wind shear within the canopy (e.g. Kondo and Akashi, 1976; Lee et al., 1994; Pyles et al., 2004) is of some concern for these studies. The term describes a „turning‟ of the average wind direction with height through the upper canopy, due to the shear or drag forces of the canopy on the wind (not shown in Equation 1). In such a situation, instruments attempting to measure the concentration difference between “upwind” and “downwind” locations may only accomplish this task at one or two levels in the subcanopy, with the others characterizing unrelated CO2 concentrations. Heinesch et al. (2008) discount directional wind shear as negligible; see Figure 20). 11  Table 1 shows a survey of the current advection literature, and relates the average magnitude of FVA and FHA measured in each study to the average NEE flux during the same period. The experimental design is also listed, with the dimensionality of the sampling plan and the transect length (if applicable). Most control volumes listed are on the order of 100 m (i.e. the “plot scale”, Feigenwinter et al., 2010a). Staebler and Fitzjarrald (2004) suggested, when defending the distance between horizontal sampling points at Harvard forest (60 m maximum, Petersham, MA, USA), that fluctuations in CO2 at the points used to calculate gradients must be well correlated in order to qualify as a “subcanopy network”. They therefore maintained that in order to measure an advection term, which is defined as a mean flow, sampling point turbulent fluctuations must be correlated. However, all researchers measuring mean advective transport have so far assumed that the turbulent contribution is a negligible mass transport term in the horizontal. One might expect that advective transport, in contrast, could be measured up to the scale of the slope where the control volume is situated, rather than only at the scale determined by correlated measurements. This last, of course, also assumes that source strength gradients or other significant heterogeneities are not present at the scale of the slope (e.g. Heinesch et al., 2007 and 2008; Marcolla et al., 2005). A better criterion for the control volume size may be the scale of the terrain producing the advective flow, rather than simply the maximum distance at which turbulent fluctuations are reasonably correlated. Staebler and Fitzjarrald (2005) show topographic maps with the most significant terrain elements highlighted at the Harvard and Borden (Ontario, Canada) forests. When considering flow down a long, fairly even 12  slope, however, advection may be occurring at the regional scale (Goulden et al., 2006, see Figure 8). Feigenwinter et al. (2010a and 2010b) postulate that measuring advective fluxes at the 100 m control volume size misses significant scalar transport at smaller and larger scales. However, due possibly to current limitations in sampling tube or instrument cable length, few advection studies (other than Wang et al., 2005) have considered advection at scales other than the common 100 m. Inspection of the last two columns in Table 1 shows that, not only is the sign of the average horizontal and vertical advective fluxes (measured over the plot scale) different depending on the site, but the magnitudes also vary widely. Advection flux divergences can be negligible during certain hours of the day and from certain wind directions, but become the same order of magnitude (or even larger) than NEE fluxes under specific conditions – normally at night. Aubinet (2008) developed a site classification system that attempts to explain the sign of advective fluxes based on source gradients and convergence or divergence of the wind field around the measurement tower. Researchers have also reported large uncertainty in flux values due to measurement error of vertical wind speed and horizontal CO2 gradients (Heinesch et al., 2007; Dellwik et al., 2009). Certain authors (e.g. Feigenwinter et al., 2004 and 2008) refrain from correcting NEE values based on their calculated advection fluxes as the large magnitudes and extremely large variability in the FHA and FVA terms would result in ecosystem fluxes that fail to represent the clear diurnal pattern that has come to be expected from natural ecosystems. Yi et al. (2008), in contrast, present a multi-year analysis of the carbon balance at Niwot Ridge inclusive of the advection terms measured at the site. 13  1.6  Study objectives  The objectives of the present study, in light of the above, were: 1. To improve advective flux measurement accuracy, especially with regard to horizontal CO2 differences. 2. To investigate advective flux magnitudes at a sloped but otherwise relatively homogeneous FLUXNET forest stand. 3. To relate advective fluxes to the carbon balance of DF49 in an attempt to determine their importance in the scalar conservation equation at the site.  14  2 Methods This section describes the site at which the UBC advection study was performed as well as the instrumentation and calculation procedures used to determine advective fluxes. Instrumentation includes the long-term monitoring instrumentation permanently installed at the site and the short-term advection instruments installed during the summer of 2009. The instruments and the experimental layout are described; calibration procedures for wind and CO2 concentration measurements are detailed.  2.1  Site description  A mature (i.e., near-end-of-rotation) Douglas-fir forest situated near Campbell River on Vancouver Island, B.C., Canada, was chosen for this study (49.87° N, 125.33° W). The stand was planted in 1949, and while primarily Douglas-fir (80%), it also contains a small percentage of cedar (17%) and hemlock (3%) (Morgenstern et al., 2004). The stand height is roughly 35 m, with a density of 1100 stems ha -1 and a leaf area index of between 7 and 10 m2 m-2 (Humphreys et al., 2006). The site is situated 9 km west of a major water body (Georgia Strait, Figure 1), on a 5-10° slope that steepens upslope of the tower to the southwest (Figure 2). Hence, it experiences topographyinduced circulations, including land-sea and slope breezes, when not otherwise influenced by synoptic weather systems.  15  Figure 1: Map of Vancouver Island, B.C., Canada.  2.1  Long-term monitoring instrumentation  The site, „DF49‟, was established in 1998, and has measured continuous vertical CO2 fluxes from a 45-m tall tower using the EC method since that time. Monitored environmental variables include air and soil temperature, soil water content, all components of the radiation balance including PAR, atmospheric humidity, and soil respiration. The latter is measured continuously using several automatic chamber systems; one located at the base of the tower and eight distributed 100 m to the north.  16  Figure 2: Topographic map of DF49 (Drewitt, 2002).  17  EC instrumentation includes a 3-dimensional sonic anemometer (model R3-50, Gill Instruments Ltd., Lymington, UK) and a thermostated non-dispersive infrared gas analyzer (IRGA) (model LI-6262, LI-COR Inc., Lincoln, NE, USA), both mounted on the tower above the canopy at 42 m. Air is drawn through a 4-m long heated tube from an air intake located 20 cm from the measurement volume of the R3-50 into the LI-6262. The IRGA is calibrated once a day at midnight using a zero (N2) and span (~380 µmol mol-1 in air) gas. A profile system also exists to measure the change in CO2 storage in the air column beneath the eddy-covariance sensors. The system includes four sampling heights (2, 12, 27 and 42 m) measured sequentially throughout a half-hour period using an IRGA (model LI-840, LI-COR, Inc.) at 7-min intervals. A more complete description of the EC and profile system can be found in Morgenstern et al. (2004).  2.2  Experimental layout  Advection flux instruments were deployed four times during the summer of 2009. Six sonic anemometers (model CSAT3, Campbell Scientific Inc. (CSI), Logan, UT, USA) were calibrated on Westham Island, British Columbia (south of Vancouver), over a field covered with tall grass between May 27 and June 15. Three experiments were then conducted at DF49, all within a two-dimensional (2D) transect framework.  18  Figure 3: First experiment to determine the consistency and direction of katabatic flow. Four sonic anemometers were first set up at DF49 to determine both the mean wind direction during katabatic flow and its consistency along a 2D transect through the forest. This was the first of three experiments performed at DF49 (Figure 3); it ran between July 9 and July 21. The second experiment, intended as the main observational period, ran between July 21 and August 25 (Figure 4). A vertical profile of CO2 concentration and wind vector measurements was established on the flux tower, and a similar profile was established 73.5 m upslope. Subcanopy CO2 concentration data was continuously collected starting July 30, and full calibration of the IRGA system began on August 2. Between August 25 and 26, the IRGA instrumentation system was checked (Appendix 7) while sampling points were reconfigured as shown in Figure 5 and Figure 73 (Appendix 8). This last DF49 experiment was intended to capture CO2 differences at an increasing distance down the slope; it ran between August 26 and September 3.  19  Figure 4: Second experiment, including CO2 concentration, to calculate advective fluxes.  Figure 5: Third experiment, examining CO2 differences along the 2D transect.  20  Four sampling points were identical to those during the first DF49 experiment (Figure 3), and one additional location was added 13 m upslope from the tower.  2.3  Sampling point placement  During the first experiment, anemometers were aligned using a compass, and placed 2.6 m above the ground on tripods (CM110, CSI) guyed with nylon rope. Wire netting protector screens (1-cm mesh) were placed 1.5 m above the anemometers to shield them from falling branches (see Appendix 6). For the second experiment, the four flux tower sonic anemometers were placed on 2-m booms to situate their measurement volume away from the tower itself. The (horizontal) booms were angled 20° into the slope to favor the accuracy of down-slope wind directions (due to the geometry of the CSAT3). The tower sonic anemometers were aligned by placing the four booms against the same side of the triangular flux tower. The upslope sonic remained where it was located during the first experiment, and the lower two upslope CO2 sampling points were placed on the same tripod. The 8and 20-m upslope CO2 sampling points were positioned using a rope and pulley system hanging from a boom attached to a nearby tree. The rope stretched during the experiment, leaving the upper 20 m sampling point displaced upwards by 1 m on August 25 (the rope being raised to maintain the 8-m sampling point roughly in position). During the third experiment, the four upslope sampling points shown in Figure 4 were collapsed to the same point at 2.6 m above the ground in Figure 5, while the four 21  tower sampling points were distributed down the slope. Specifically, the 1 m downslope point in Figure 4 was placed at x = 53 m in Figure 5, the 20 m point was placed at x = 0 m, and the other two were distributed in between. Sonic anemometer protector screens were again installed 1.5 m above all five sonic anemometers. CO2 sampling points for all experiments were placed on the sonic anemometer booms roughly 30 cm away from the anemometer measurement volume; thermocouples were taped to the anemometer within 15 cm of the transducer heads (Appendix 6).  2.4  Advection flux instrumentation  Wind vector measurements were made using five laboratory-calibrated CSI CSAT3 sonic anemometers. CO2 concentration measurements were made using eight thermostated LI-COR Inc. LI-7000 and LI-6262 IRGAs; they were run in absolute mode, and N2 reference gas was fed individually to each IRGA at 100 cm3 min-1 (LI-6262s) and 120 cm3 min-1 (LI-7000s). Four flow meters (model RMB-BV, Dwyer Instruments Ltd., Michigan City, IN, USA) were used for this purpose, with a single meter controlling flow to a paired set of IRGAs. Reference N2 tank output pressure was maintained using a single stage regulator (model PRS2122321-01-580, Praxair Inc., Danbury, CT, USA) set at 10 psi (70 kPa). Additionally, the instruments were paired using a solenoid switching system placed close to the IRGAs to create four differential analyzers (Figure 7). The system, designed to remove instrument offset error (Black and McNaughton, 1971), was set to a 22  3 minute switching interval, generating differences between an upslope and downslope sampling point at a minimum time of 6 minutes (Appendix 1). Four solenoid valves (model mouse EV-2M-12 V DC, Clippard Instrument Laboratory, Inc., Cincinnati, OH, USA) were attached to a single aluminum block, switching air flow from two incoming sample lines to two IRGAs. The valves were controlled by a relay board (model ADU 200, Ontrak Control Systems, Sudbury, ON, Canada) connected to a mini-PC (model ultraclient, Norhtec Corporation (Thailand) Ltd., Pakkred, Nonthaburi, Thailand). The switching system was combined with a twice-daily IRGA calibration (2:00 and 14:00 h PST2) consisting of 6 CO2 concentrations (0, 360, 380, 400, 450 and 500 ppm) injected via solenoid valves attached to each sampling point (section 2.7). CO2 sampling tubes (Eaton Synflex, distributed by Wirex Controls Ltd., Brampton, ON, Canada) were of equal length (60-m long, inner diameter of 4 mm); air was drawn through each of the 8 tubes at a rate of 2 L min-1 using a single linear pump (model SPP-15EBS, Gast Manufacturing Inc., Benton Harbour, MI, USA). Sample air was passed through 1.0 µm PTFE filters (model Acro 50, Pall (Canada) Ltd., Mississauga, ON, Canada) placed between the solenoid valve switching systems and the IRGAs themselves. Sample flow control was achieved using two flow meters (model RMB-BV, Dwyer Instruments Ltd.); one for the four LI-6262 analyzers and one for the four LI-7000s. Tube fittings were stainless steel and brass (Swagelok Co., Solon, OH, USA). Pictures of the instruments deployed at the site are in Appendix 6.  2  Pacific Standard Time  23  2.5  High-frequency data collection  IRGA data was collected via two 4-port RS-232 to USB hubs (model US09ML24P, B&B Electronics Mfg. Co., Ottawa, IL, USA) connected to the mini-PC at a frequency of about 1.6-1.7 Hz (nominally 2Hz, LI-6262s) and exactly 2 Hz (LI-7000s). The system recorded semi-continuous CO2 traces at all sampling points, with a 5.6% loss due to discarded data from a 10-s pressure equilibration after each solenoid valve switch. Sonic anemometer data was collected at 10 Hz using a datalogger (model CR3000, CSI) with 60-m lengths of 12-conductor (6-pair) aluminum-foil-shielded wire (type 8778, Belden Inc., Richmond, IN, USA) using the SDM data transfer protocol. Air temperature data (obtained using 75 µm chromel-constantan thermocouples, Omega Engineering Inc., Stamford, CT, USA) was additionally collected at 10 Hz using a signal multiplexer (model AM25T, CSI) connected to a datalogger (model CR5000, CSI) at the four subcanopy measurement heights at both ends as well as in the middle of the advection transect. Sonic anemometer and thermocouple data was collected on flash cards connected to the respective logger and downloaded periodically to a laptop computer.  2.6  Thermostated infrared gas analyzer box  The IRGAs were placed on metal rail shelves in a plywood box insulated with 2.5 cm Styrofoam (R value of 5, type SM-C, Dow Chemical Canada ULC, Calgary, AB, Canada). Air was mixed within the box using four vertically-oriented 120-mm diameter 24  Figure 6: Thermostated box in the laboratory, showing the aluminum heat exchanger.  Figure 7: IRGAs in thermostated box at DF49, showing solenoid valve switching system and mixing fans.  25  Figure 8: Outside air and thermostated box temperatures during the second and third experiments. fans placed behind the IRGAs (64 CFM, model CFA1212025MS, Circuit-Test (division of R.P. Electronic Components Ltd.), Burnaby, BC, Canada) (Figure 7 and Figure 8). The temperature difference between the edges and centre of the box was around 5 °C (Figure 8). Air temperature within the box was kept relatively constant by four additional proportionally controlled 120 mm AC fans blowing warmed air from the instruments through an aluminum heat exchanger situated on top of the box (Figure 6). The fans were 106 CFM 115 V AC 12 W ball-bearing tubeaxial fans (model CFB11512038 HB, Circuit-Test) driven by a 120V/25A variable output solid state relay AC source (model MCPC1225A, Crydom Inc., San Diego, CA, USA) itself controlled by a variable DC output from a datalogger (model 23X, CSI). The DC output kept the fans either off or 26  between 48 and 100% of maximum load (Figure 9). The temperature of a single thermocouple at the base of the box (“T”) queried at 1-s intervals was used to generate a signal between 2400 and 5000 mV (corresponding to 48-100% of 5 V DC) via the following equation:  fanCTRLmV  T  37 7800  2400 2400   fanCTRLmV  fanCTRLmV  5000   (5)  fanCTRLmV  2400 2400  fanCTRLmV  5000 fanCTRLmV  5000  The (arbitrary) factor of 7800 changed the sensitivity of the system to deviations from the temperature set point (in this case 37 °C). The set point was changed twice during the experiment, from 37 to 35 and then 33 °C (Figure 8; „Bottom‟ thermocouple).  Figure 9: Control signal to the heat sink fans during the second and third experiments.  27  The datalogger also recorded 1-min average thermocouple temperatures at other locations distributed within the box and the heat exchanger (Figure 8). The thermostated box itself was placed along with the datalogger enclosures on a plywood table situated midway along the advection transect. The table was covered by a sheet of plywood and a green PVC tarpaulin, intended to protect the table from both rain and sun (Figure 67).  2.7  Sonic anemometer offset calibration  The sonic anemometers were first set up in a laboratory environment before their comparison on Westham Island. In the laboratory, a plastic bag was placed over each sensor head to reduce turbulence within the sampling volume, and an estimate of the sonics‟ zero offset in x, y and z was estimated from their average output over several minutes (Table 2). The averaging time used was 12 minutes at an output frequency of 1 Hz before the advection experiment; this was increased to 10 Hz over 21 minutes after the experiment. Zero offset estimation was repeated four times with a single instrument (serial number (SN) 0126) after the advection experiments to gain an idea as to the variability in zero offsets measured with the “plastic bag” procedure; standard deviation of the four estimates in x, y and z was 0.2, 0.3 and 0.4 cm s-1, respectively. Note that SN 0126 had a larger offset in the vertical than the other four instruments. An average of the „before‟ and „after‟ laboratory zero offset values (e.g. -1.81 cm s-1 for x, SN 0126) was removed from each high frequency data point during analysis. It should be noted that any offset correction in the vertical is modified by a tilt correction 28  algorithm such as the planar fit method, which takes into account sensor tilt due to improper field installation (Heinesch et al., 2007).  Table 2: Sonic anemometer zero offsets determined in the laboratory both before and after the summer of 2009. SN 0126 Coordinate  SN 1341  SN 1389 (ref.)  SN 1393  SN 1396  Before  After  Before  After  Before  After  Before  After  Before  After  -1  -1.86  -1.76  0.79  -0.29  -1.16  0.31  0.16  -0.31  0.41  2.05  v (cm s )  -1  -2.02  -1.85  -1.47  -0.31  -0.64  -0.96  -1.01  -0.95  -1.62  -2.25  -1  3.92  3.45  0.94  0.28  0.16  0.10  1.60  0.04  0.60  0.85  u (cm s )  w (cm s )  Zero offset values were also determined from the Westham Island sonic comparison (see Appendix 2).  2.8  Field IRGA calibration  The IRGA offset and span values for both CO2 and H2O were calibrated in the laboratory prior to and after the advection experiment (see Appendix 3 for the latter). IRGA CO2 and H2O zero as well as CO2 span values were also calibrated in the field, as mentioned in Section 2.4. CO2 calibration gases were run through all eight instruments at the same time in the following sequence: 0 (2 min), 363 (2 min), 378 (1.5 min), 395 (1.5 min), 451 (1.5 min), and 492 (1.5 min) (µmol mol-1). This sequence was run twice a day, at 2:00 and 14:00 (PST). The time of 14:00 PST was chosen as katabatic flow at the site usually starts at around that time, at least close to the ground. The calibration time of 2:00 PST was simply 12 hours offset from 14:00 PST. The CO2 concentrations were chosen to reflect the range of values expected in the subcanopy air at DF49 and the availability of calibration gases in Canada at the time of the experiment.  29  CO2 zero and span gas tanks were placed in the middle of the advection transect (in the open air) close to the instruments. Output was controlled by dual-stage regulators (model PRS201223 and PRS2122301-75-000, Praxair Inc.), and pressure was maintained at 20 psi (140 kPa) in order to ensure adequate flow at the sample point solenoid valve after flowing through roughly 62 m of 4 mm Synflex tubing (80 m during the third experiment) (Wirex Controls, Ltd.). Flow meters (model RMB-BV, Dwyer Instruments Ltd.) were used to set the calibration gas flow to 20 L min -1 immediately after the regulators. The gas flow was then split using a T-shaped fitting and run through separate 50-m long tubes to two distribution boxes placed at the upper and lower ends of the advection transect (totaling 12 tubes). Two DC relay boards in each box (ADR 2205, Ontrak Control Systems) controlled six AC solenoid valves (model DS6011, Burkert Contromatic Inc., Burlington, ON, Canada) which enabled flow from one of the six calibration gas tubes through a manifold to the four nearby sampling points. All six gases were distributed to each sampling point via a short length of clear plastic tubing (bev-a-line IV, Thermoplastic Processes, Georgetown, DE, USA) connected to 12 m of Synflex tubing (Wirex Controls, Ltd.). Solenoid valves (model DS6011, Burkert Contromatic Inc.) were also placed on the end of each 12-m line at the sampling point itself (Appendix 6). The solenoid valves in the distribution box and at each sampling point were activated simultaneously during calibration. The commands to the relay boards were made via aluminum foil shielded single pair wire using the RS485 data transfer protocol. The wire was connected to the miniPC using a RS485 to USB adapter (model uLinks USOTL4, B&B Electronics Mfg. Co.). The entire IRGA system is sketched in Figure 10. 30  Figure 10: IRGA system component diagram. Note that only one differencing pair is shown and not all Synflex lines are drawn. (Not to scale.)  2.9  System power  During the Westham Island sonic calibration, the datalogger and anemometers were powered using a 12 V battery. At DF49, the loggers, pump, fans and instruments were powered from a bank of 24 V batteries whose primary purpose was to provide power to the long-term flux instrumentation at the site. AC power was generated using the site‟s inverter and run through a series of 12-gauge AC power cords to the plywood instrumentation table on the advection transect. DC voltage (12 V) was then generated from the AC power at the table using several AC-to-DC converters, which powered the CSI dataloggers and switching solenoids in the IRGA box (model GHOF N-12 REV A, GFC Power). AC power lines (16 gauge) were also run to solenoid valves and relay  31  switches housed in two boxes that distributed calibration gases to sampling point solenoid valves at either end of the advection transect. The power used for this purpose was first passed through a 115 to 115 V isolating transformer (model 171, Hammond Mfg. Co. Inc., Guelph, ON, Canada). AC power to the sampling point solenoid valves was taken from the calibration gas distribution boxes (also through 16 gauge wire). Wires in all logger and relay boxes, as well as the thermostated IRGA box, were connected using clamping yoke screw terminal blocks (model WDU, Weidmuller Canada, Markam, ON, Canada). The entire advection instrumentation system consumed an estimated 320 W when the heat exchanger cooling fans were running during the middle of the day. To maintain the charge in the 24-V battery bank, a diesel generator was run for 6 hours every night between midnight and 6:00 PST. The generator was situated 100 m downslope (azimuthal angle of ~50°) from the flux tower and advection transect; the exhaust plume (containing CO2) was also directed away from the advection transect during katabatic flow. Power was maintained without interruption throughout the advection experiments at DF49.  2.10  Data processing  CO2 concentration and wind vector data was processed following the experiment to calibrate the CO2 output from the IRGAs, convert the sonic anemometer wind vector  32  output into a streamline coordinate system, and remove periods when the instruments were malfunctioning. 2.10.1 Software IRGA calibration Raw IRGA mV (LI-6262) and radiant power (LI-7000) output for CO2 and H2O concentration as well as analyzer cell temperature and pressure for each of the six gases was used to generate averages for each calibration period. Averages were taken from the last 15 seconds of each calibration gas interval; the time of each average was determined by taking the differential of the IRGA CO 2 concentration trace to find significant changes in CO2 concentration. This was necessary due to slight variations in calibration gas injection timing between subsequent days; the calibration gas solenoid valves were controlled by a batch file on the mini-PC, which was running under a significant computing load. These averages were then linearly interpolated between the 2:00 and 14:00 PST calibration periods to generate a sequence of averages for each half-hour period. Finally, a second-degree polynomial was created for each IRGA and each half-hour from the average CO2, H2O, temperature and pressure values which converted the raw mV and power values to CO2 and H2O concentrations in µmol mol-1 and mmol mol-1, respectively (Burns et al., 2009). All CO2 concentrations were converted to a mixing ratio for analysis (i.e., corrected for atmospheric water vapour content). Calibration values depend on temperature and pressure; a slight inaccuracy was therefore induced by the linear interpolation procedure as air temperature generally cooled a few degrees after the 2:00 PST calibration time before warming during the day.  33  The instruments were usually at a constant temperature regardless of the surrounding air temperature, but deviations from thermostatic conditions normally also occurred during the coldest part of the day (Figure 8). 2.10.2 Wind vector coordinate system rotation The planar fit method (PFM; Wilczak et al., 2001) was used to convert the sonic anemometer instrument coordinates into streamline coordinates in order to account for the sloping terrain. A separate fit was performed for each instrument during the second DF49 experiment, placing each instrument in a slightly different frame of reference. The fit was performed both for the entire dataset and for a subset, accepting only half-hours during neutrally-stratified conditions (Hunner et al., 2009). The planar fit method was  Figure 11: Vertical velocity corrected with the PFM and TAM methods.  34  also used to rotate vertical fluxes of momentum (i.e. the friction velocity), buoyancy and CO2. Another rotation procedure developed by Vickers and Mahrt (2006), termed the „tilt angle method‟ (TAM), was used to generate a parallel set of corrected data. The tilt angle method was applied both with and without the neutrally-stratified criteria. Coordinate rotation, as well as the choice of rotation method, had little impact on horizontal wind velocities but a significant impact on vertical wind speeds (Figure 11). 2.10.3 Data quality control Individual half-hourly wind speed and direction data were scanned for the second DF49 experiment, while CO2 concentration data was screened for both the second and third experiments. When a subset of the IRGAs failed for an individual half-hour, data from all IRGAs was removed from the analysis. Similarly, when one or several sonic anemometers reported unrealistic values (i.e. a square wave), all sonic anemometer data from that half-hour was removed. IRGA data during switching (10 seconds every 3 minutes) as well as IRGA data during calibration times (12 minutes, twice a day) was removed from individual half-hour traces. Short periods when the N2 reference gas tank was being changed or researchers were breathing into CO2 inlets were also removed. For data gaps less than half an hour in length, the remaining data was used to generate a half-hour average. Data gaps of a half-hour length or longer were filled using linear interpolation. 2.10.4 Advection flux computations The following equations were used to compute FHA and FVA: 35  FHA  da  30m   0  u  (6)  c dz x  FVA   da w 42m (c 42m   1 42m  42m   cdz )  (7)  0  The integrals were approximated by finite sums; the partial differential was approximated by a finite difference. A positive gradient in CO2 concentration indicated higher CO2 concentrations at the downslope sampling point and a positive wind speed indicated katabatic or downslope flow. In equation 6, the mean wind speed term has been taken out of the horizontal differential (compared to equation 3); this indicates the assumption that wind speed does not change along the advection transect (however, see Figure 13). The dry air density was calculated from EC data at 42 m above the forest canopy for all computed fluxes, assuming that variations in wind speed, air pressure and temperature throughout the control volume had a negligibly small effect on the magnitude of air density (see Montagnani et al., 2009). A complete derivation of the FVA term (equation 7) can be found in Lee (1998). In order to generate canopy-averaged CO2 concentrations and wind speed profiles, several additional assumptions were made. First, CO2 concentrations from the upper levels (8 and 20 m), measured with LI-6262s at about 1.6 Hz, were linearly interpolated to 1 Hz using MATLAB‟s “interp1” function (MATLAB R2006b, The MathWorks, Inc., Natick, MA, USA). This procedure generated an estimated maximum error in an individual 3 minute average of 0.3 µmol mol -1 (from a half-hour period as shown in Appendix 4).  36  Wind speed, CO2 concentrations and CO2 differences were interpolated linearly with height. A no-slip condition was also applied at the forest floor (wind speed was assumed zero at 0 m). For horizontal CO2 differences, atmospheric mixing was assumed sufficient to reduce gradients to zero at 30 m (i.e. 5 m below zh). In order to calculate vertical CO2 differences for the FVA calculation, CO2 concentration was extrapolated from the 1-m height to ground level (also using the MATLAB “interp1” function). Finally, horizontal advection was computed both with and without a filter for wind direction. Instead of removing individual half-hours if their wind direction exceeded a certain acceptance range, a wind speed cosine correction was implemented as: U cos  cos(  113)  U raw  (8)  where the addition of 113° ensured the cosine of 247° (the advection transect angle) was 1; the absolute value operator ensured that the correction did not change the sign of the raw windspeed (set to positive for downslope flow and negative for upslope flow).  2.11  Comparison with long-term flux instrumentation  Calibrated CO2 concentrations, wind speed and direction values were compared with with CO2 values from the site EC and profile system and a RM-Young 81000 sonic anemometer situated on the main flux tower at the 4-m height. The wind vector data from the subcanopy CSAT3s agreed with the RM-Young, both in speed and direction ( Figure 12). The profile system, while sampling from subcanopy air at each sampling point for only 25% of each half-hour, appeared to generate similar average CO 2 concentrations 37  to the advection experiment and EC instrumentation. A slight negative offset of 1-2 µmol mol-1 was noticeable especially in the upper canopy levels (24 and 42 m) between the profile and EC systems, while the lower two profile system CO 2 concentrations (2 and 12 m) were somewhat different (up to 5-10 µmol mol-1) compared to concentrations calculated with the advection instruments due to the variability of CO2 concentration at those levels. Finally, the 27-m profile system level gave some indication of a buildup of CO2 concentration at night (possibly due to foliar respiration) that was missed by the 20-m advection system level. This may indicate that linear interpolation of the 20-m CO2 difference to zero at 30 m gave an inaccurate representation of horizontal advective CO2 fluxes within that region of the crown space.  Figure 12: Comparison of CSAT3 tower profile with site RM-Young.  38  2.12  Soil CO2 efflux measurements  Throughout the summer, manual chamber measurements of soil CO2 efflux were made along the advection transect. The system consisted of a datalogger (model 21X, CSI), gas analyzer (model LI-800, LI-COR Inc.), diaphragm pump (model NMP850KNDCB, KNF Neuberger Inc., Trenton, NJ, USA) and flow meters (model RMB-BV, Dwyer Instruments Inc.) connected to a 11 cm diameter PVC tube chamber (area = 0.0075 m2 and volume = 0.0015 m3). No collars were placed in the ground for measurement; instead, flags marked rough measurement locations and foam around the collar base stopped mixing between the chamber airspace and the surrounding atmosphere. Measurements were made for 2 minutes at each location with an IRGA measurement frequency of 1 Hz; the rate of change of the CO 2 concentration3 within the chamber was used to calculate a flux for each point. Each ramp was manually screened for data quality. Air temperature was measured with a thermocouple prior to each measurement, and half-hour averaged air pressure was taken from the site barometer; these measurements were combined to generate an air density for each flux calculation. Measurements were made along a 165 m transect straddling the flux tower for the first half of the summer (5-m measurement intervals) and along a 90 m transect with its base at the flux tower for the latter half of the summer (3-m measurement intervals). Measurements were taken between 10:00 and 15:00 PST.  3  A mole fraction in this case, with no dilution correction applied.  39  3 Results and Discussion Results from the DF49 observational periods are presented in the order that they occurred (sections 3.1 to 3.3). Soil efflux measurements (section 3.4) are presented after the three experiments, sample high-frequency CO2 data is in Appendix 4 and leaf CO2 exchange data is in Appendix 5. Horizontal axes marking a date refer to midnight and the beginning of the day indicated.  3.1  First experiment  The intent of this setup was to determine whether wind directions along the advection transect were similar and whether they had a consistent direction during katabatic flow events. This experiment was designed to validate the two-dimensional assumption of the main advection flux experimental layout. Four sonic anemometers were placed along the 73.5-m transect at roughly 25-m horizontal spacing, 2.5 m above local ground level (Figure 3). Figure 13 shows that a clear diurnal cycle in both wind speed and direction was typical at the site for at least seven of the eleven days shown (July 9-11 and 14-17). Wind direction during the night was downslope, normally around 270°; upslope wind direction during the day, while more variable, was around 90°. Highest half-hour average wind speeds were < 1 m s-1, with 30 cm s-1 being typical. Periods of upslope flow during the day were shorter in duration than the katabatic flow at night, which began late in the afternoon and continued usually until the morning. 40  Figure 13: Wind direction and speed.  Figure 14: Wind direction and speed for a subset of the first experiment.  41  Figure 14 shows a four-day period from the same experiment in order to more clearly distinguish similarities and differences between the four sonic anemometer traces. The difference in half-hour average wind direction between the anemometers along the transect was occasionally less than one or two degrees, but was more commonly between 10 and 20°. Wind speed tended to be similar between the four sonics during the day (± 5 cm s-1), and became somewhat different during katabatic flow (± 15 cm s-1) even when wind directions were similar. This difference invalidated the assumption that wind speed was constant along the advection transect. The results show that wind direction was similar along the advection transect, especially during katabatic flow. The wind angle, however, tended to vary throughout the night. During the strongest katabatic flow on July 9-11, the direction was around 250°, but tended to oscillate around 270o on July 14-17. A trend towards an increasing angle throughout the night (240° in the late afternoon to 300° in the morning) was also occasionally observed. The two-dimensional katabatic flow assumption therefore holds only tenuously at the DF49 site.  3.2  Second experiment  The main observational period was intended to provide an estimate of advective fluxes throughout the lower canopy during the summertime, when soil respiration is strongest and advective fluxes are hypothesized to be maximal. Figures outlining the climate experienced at DF49 during this experiment are presented first, followed by  42  ensemble average plots of advective fluxes and their effect on the carbon balance of the site during the 23 day period. 3.2.1 Temperature and radiation climatology Air temperature during the period August 2-25, 2009 (Figure 15) spanned a range between 10 to 35 °C, with strong diurnal variation at the beginning and near the end of the period. Radiation (Figure 16), measured above the canopy with a net radiometer (model CNR1, CSI), showed a similar trend to Figure 15, with a cloudy period between August 8 and August 12. Rain occurred at the site during this time.  Figure 15: Air temperature. Figure 17 shows an ensemble average of potential air temperature, describing the diurnal pattern at the 2-, 27- and 44-m heights on the main flux tower. The canopy height is around 35 m, so the 27-m height represents a location high in the crown space, the 44-m height is in the roughness sublayer above the canopy, and the 2-m height is above any understory vegetation. During the night (on average), the crown space was cooler than the air above and below, due to radiative energy loss to the sky above. The 44-m height was the warmest during the night, meaning that the entire forest was capped by a statically stable air layer. However, the forest floor was 43  Figure 16: Above-canopy net radiation. slightly warmer than the crown space, meaning that the trunk space air was neutral to unstable throughout the night. During the day, the crown space was warmed by the sun, creating unstable air above and stable air in the trunk space. There were also two brief transition periods during the morning (6:30 – 8:30 PST) and afternoon (16:30 to 19:30 PST) when both the 2- and 27-m height temperatures were colder than the 44-m level, as well as being stratified with height. Figure 18 and Figure 19 present ensemble average vertical buoyancy fluxes above and within the canopy as calculated from sonic temperature and vertical velocity fluctuations. The typical diurnal course was observed above the canopy, where the peak average heat flux (upwards) was around 300 W m-2 during the middle of the day. A downwards buoyancy flux was seen during the night. However, the opposite was observed within the canopy, especially at the 20-m height. At 8 m, there was a net upwards transport of buoyancy during the night, and an average near zero during the day. Further down in the subcanopy, a lag was observed; the average minimum buoyancy flux at 1 and 2.6 m occurred in the evening at around 18:00 PST. This time coincided with a minimum in the ensemble average standard deviation of wind direction 44  and a maximum in the ensemble average wind speed at all subcanopy heights (Figure 21 and Figure 23), as well as a peak in subcanopy CO2 concentration (Figure 27). It would make sense that the strongest, most coherent katabatic flow close to the forest floor occurred at the time when the subcanopy air at that location was most strongly stable. CO2 may also have accumulated due to this reason.  Figure 17: Ensemble average of the diurnal variation in potential temperature. 3.2.1  Wind direction and speed  Figure 20 and Figure 21 show ensemble average vector mean wind direction and its standard deviation4. Flow was on average downslope during the night (270 o) and upslope during the day (50o) at all heights. Individual half-hour averages were  Calculated as 81 1  U/ S , where U is a vector mean wind speed and S is an arithmetic mean wind speed (CR 21X manual, CSI). 4  45  Figure 18: Vertical profile of buoyancy flux.  Figure 19: Vertical profile of buoyancy flux, showing the subcanopy only.  46  Figure 20: Ensemble average vector mean wind direction. occasionally upslope at night and downslope during the day, especially at the lower measurement heights. The transition between katabatic and anabatic flow took around an hour, and did not follow any remarkable order. However, the afternoon transition took several hours (4 on average); the katabatic flow layer began close to the forest floor and gradually built in thickness until it reached the 42-m height. After the transition to katabatic flow, direction at the 1- through 20-m heights was similar for the first half of the night, around 260 to 270o. This was especially true for the 1- and 2.6-m heights between 16:00 and 19:00 PST. However, ensemble average wind directions through the early morning became somewhat separated; the 42-, 20- and 8-m heights experienced lower average wind directions and the 1- and 2.6-m heights had a higher average wind direction.  47  Figure 21 demonstrates a clear diurnal pattern in the standard deviation of wind direction (σWD), with a maximum during the middle of the day and a minimum (as previously mentioned) in the evening. Around noon, the highest σ was at the 1-m height, and the lowest at the 20 m height (the 42 m height is not shown).  Figure 21: Subcanopy ensemble average of the standard deviation in wind direction. During the evening, the trend was reversed, with the steadiest flow close to the forest floor. Morning σWD was intermediate for all levels. Figure 22 and Figure 23 show ensemble average three-dimensional (3D) wind speed5; Figure 23 without the 42-m trace. It is immediately obvious from Figure 22 that the above-canopy wind speeds were much greater than those below. Additionally, the  5  Calculated as  u2  v2  w 2  48  Figure 22: Ensemble average of wind speed.  Figure 23: Ensemble average of wind speed within the subcanopy.  49  Figure 24: Standard deviation of wind speed in the subcanopy.  Figure 25: Ensemble average friction velocity profile.  50  1- through 20-m height wind speeds showed little differentiation with height in comparison with the 42-m height. Barring instrument or logger malfunction (as the subcanopy data was collected with a separate logging system), this constancy with height throughout the lower canopy may be attributable to the density of the stand. Figure 23 shows the differences between speeds at the 1- through 20-m heights; speeds were similar during the night and midday for the majority of heights (excepting the 1-m level). During the evening, the highest speeds throughout the entire day were seen at the 1 and 2.6-m heights, with speeds at the 2.6-m height (both up and down slope) being the highest. This represents a typical S-shaped vertical profile seen during katabatic flow at many sites, especially if the trunk space region is open (Queck and Bernhofer, 2010). This maximum speed at the 2.6-m height continued throughout the remainder of the night, though the magnitude was reduced as the katabatic flow broke down and became more erratic. Figure 24 shows the standard deviation in wind speed within the subcanopy. Fairly constant values during the night were replaced by a midday peak; the diurnal course was clearer than that of the wind speed itself. The standard deviation during the night was around half of the magnitude of the wind speed, while during the day the magnitude of the wind speed and its standard deviation were comparable. The 2.6-m height showed the largest standard deviation at all times, followed by the 8-, 20- and 1-m heights during the day and the 1-, 8- and 20-m heights at night. Finally, Figure 25 shows ensemble averaged friction velocities (u*) throughout the canopy and at the EC measurement height. Similar to wind speed, values at 42 m are much larger than within the canopy, with a clear maximum during the day and relatively 51  constant value at night. However, u* in the subcanopy has its maximum at the same time as that at the 42-m height, unlike in Figure 23 for average wind speed (though similar to maximum half-hourly wind speed, not shown). Curiously, Figure 25 shows similar friction velocities throughout the lowest 20 m of the canopy at night, suggesting the existence of a “constant momentum flux layer” within the canopy or an input of momentum from the subcanopy katabatic flow. 3.2.2 CO2 concentration Figure 26 presents half-hour averaged CO2 concentration (mixing ratio) for five levels on the main flux tower. The range was between 370 µmol mol-1 at the 20- and 42-m heights at midday to 470 µmol mol-1 close to the forest floor in the early morning. Higher concentrations of over 500 µmol mol-1 (not shown) were observed before August 2 during a heat wave (Figure 15). Figure 27 shows the ensemble average of Figure 26; a clear, though muted, diurnal pattern can be observed. During the night, subcanopy CO 2 concentrations built slowly, with values below the 42-m height similar in magnitude. The transition between katabatic and anabatic flow promoted significant CO2 accumulation especially at the 1and 2.6-m heights at around 6:00 PST, while occasional sweeps high up in the canopy began to reduce the 20- and 42-m height CO2 concentrations around an hour later (see Appendix 4). During the day, photosynthesis and higher wind speeds reduced the CO2 concentration at all levels, with the 20- and 42-m heights well coupled with the atmosphere and the lower levels still at a higher concentration. The afternoon transition  52  Figure 26: Half-hour average CO2 concentration.  Figure 27: Ensemble average CO2 concentration.  53  to katabatic flow (close to the forest floor), occurring when the subcanopy air was at its most stable (Figure 17 and Figure 19), also produced a buildup of CO2. This was seen especially at the 1- and 2.6-m levels. Figure 28 presents an ensemble average of the standard deviation in CO 2 concentration (σCO2, 1 or 2 Hz) throughout the day at DF49 for the subcanopy levels. The 20-m height experienced the greatest variability around 7:00 PST (again, see Appendix 4), with little variability otherwise. Lower levels experienced two peaks in σCO2 during the morning and evening transition periods. The evening peak variability in σCO2 was more prolonged at the 1-m than the 2.6-m or 8-m levels, though the 1- and 2.6-m levels experienced remarkably similar average σCO2 around 17:00 PST. This was also  Figure 28: Ensemble average of the standard deviation of CO2 concentration.  54  true during the middle of the day for the 1- and 2.6-m heights, and during the night for the 8- and 20-m heights. A similar σCO2 between two levels may indicate that they were well coupled during that time. 3.2.3 CO2 concentration differences Horizontal and vertical CO2 concentration differences for the second experiment are presented in Figure 29 through Figure 33. Differences are positive for a higher downslope CO2 concentration. Figure 29 and Figure 30 show half-hour average horizontal CO2 concentration differences for individual 24-hour periods during clear-sky and cloudy conditions (respectively). Half-hour average differences were rarely above 15 µmol mol-1 over the 73.5-m long transect, with 5 µmol mol-1 being more common. Negative differences were rarely observed. The largest differences were seen at the 1and 2.6-m heights (the ground being the largest source of CO2), especially during the morning and evening. This pattern can be more clearly observed in Figure 31, which shows ensemble average horizontal CO2 concentration differences. The differences were clearly greater close to the forest floor during the day. Remarkably, there was a lack of large differences during most of the night. There was also a negative average difference for the 1-m height during this time. Figure 32 shows the standard deviation of these averages; a similar minimum occurs during the middle of the day; separation of the traces with height is still visible, though less clear during the night. Figure 33 shows the ensemble averaged vertical CO2 concentration differences (C42m – Ccanopy) throughout the day; it is immediately obvious that the difference was  55  Figure 29: Subcanopy horizontal CO2 concentration differences on August 4th, 2009.  Figure 30: Subcanopy horizontal CO2 concentration differences on August 9th, 2009.  56  Figure 31: Ensemble average subcanopy horizontal CO2 concentration.  Figure 32: Standard deviation in horizontal CO2 concentration.  57  Figure 33: Ensemble average vertical CO2 concentration difference between the EC height and the air column below (error bars are one standard deviation). always negative, due to the source of CO2 below the 42-m height. The largest difference was seen during the morning, with intermediate differences in the evening and during the night, and a minimum at midday. The difference between 42 m and the average subcanopy concentration most likely varied for the same reasons as the CO2 concentration itself; i.e. the presence or absence of photosynthesis and the turbulent state of the atmosphere. 3.2.4 Advective CO2 fluxes Both FHA and FVA are presented in Figure 34, again as ensemble averages over the entire second experiment. NEE, as estimated by the storage-corrected vertical turbulent flux, is presented for reference. FVA appeared to be more variable than FHA, at least as shown, with an average of -1 µmol m-2 s-1 over the entire day. FVA showed little 58  Figure 34: Ensemble average FHA and FVA (PFM) as well as storage-corrected NEE. tendency for either positive or negative values during the night, but was positive during the morning transition time and negative during the day when vertical wind speeds are positive. FHA (integrated from 0 to 30 m) shows a near-constant 5 µmol m-2 s-1 flux throughout the night, with a peak in the late afternoon of closer to 8 µmol m-2 s-1 and a slight minimum of -2 µmol m-2 s-1 during the morning. Figure 35 and Figure 36 show the same curves as in Figure 34 but with error bars showing a single standard deviation for each half-hour average. For both figures, the standard deviation between the 1081 half-hours is smallest during the middle of the day and largest at night. Especially for FVA, statements about the magnitude or even the sign of the flux (as in the previous paragraph) are difficult to make; the error bars span 0 µmol m-2 s-1 at all times.  59  Figure 35: Ensemble average FHA shown with error bars (one standard deviation).  Figure 36: Ensemble average FVA (PFM) shown with error bars (one standard deviation). 60  Figure 37: Histogram of FHA.  Figure 38: Histogram of FVA (PFM).  61  Figure 37 and Figure 38 show 100-bin histograms of FHA and FVA (respectively). Both figures have the same scale, but the FHA distribution does not contain any instances below -40 µmol m-2 s-1. The size of both datasets is again 1081 half-hour averages. The mean of both distributions is small, with that of FHA slightly positive, and the mean of FVA slightly negative (Table 3). The FVA distribution appears to have larger tails (i.e. more values of greater magnitude), and is skewed towards negative values, while the FHA distribution is skewed towards positive values. The median values are smaller than the means in both cases. Table 3: FHA and FVA (PFM) statistics for the second experiment (µmol m-2 s-1). Mean Median Standard deviation Skewness Kurtosis FHA  2.31  1.06  8.44  1.62  12.29  FVA  -1.05  0.08  12.8  -1.16  10.82  Figure 39 shows cumulative FHA fluxes for each subcanopy level throughout the second experiment. Each trace represents the flux occurring through a 1-m thick layer in the forest surrounding each sampling point. The 1- and 2.6-m height traces follow each other closely, while the 8-m height is similar to the 20-m height (though reduced in magnitude). Individual events are occasionally of opposite sign between the lower and upper two levels, and significant flux over a few half-hour periods may be followed by several days where little cumulative flux occurs.  62  Figure 39: Cumulative FHA fluxes by subcanopy level. 3.2.5 Cumulative carbon balance This section presents the NEE for the second DF49 experiment as the vertical turbulent flux corrected for storage and either filtered using the u* method or corrected with advective fluxes. Figure 41 is an ensemble average of these fluxes, while Figure 42 and Figure 43 are cumulative (integral) plots of CO2 flux over the course of the second DF49 experiment. Figure 41 shows that the u*- and advection flux-corrected NEE were similar throughout the middle of the night, with a respiration flux between 6-10 µmol m-2 s-1. The uncorrected NEE was slightly smaller at 3 – 5 µmol m-2 s-1. During the morning, all three NEE traces show the transition to net uptake at around the same time, with the advection flux-corrected NEE crossing zero slightly after the other two. There was a  63  slight peak in the advection flux-corrected NEE during the early morning hours, due to a similar peak in FVA at that time. This peak was not noticeable in the NEE or u*-corrected NEE traces, though cleaning or despiking associated with routine data quality control procedures may have affected this result to some degree. The slight difference in uptake between the NEE and u*-corrected NEE traces during the morning is due partly to a different storage flux correction, with the u*-filtered NEE corrected with a storage flux calculated at the EC height only and the NEE trace corrected with a storage flux calculated from CO 2 concentrations throughout the subcanopy (Figure 40).  Figure 40: Cumulative subcanopy and EC storage fluxes. The u*-filtered trace also uses the natural coordinate system, where individual half-hourly fluxes are rotated to ensure the vertical velocity is zero, while the NEE trace 64  is rotated using the planar fit method. The CO2 storage trace shown in Figure 41 considers subcanopy CO2 concentration. During the middle of the day and the early evening, the NEE and u*-corrected NEE were similar, while the advection flux-corrected NEE showed stronger uptake. This change is attributable to the fact that both FHA and FVA were negative on average during this time. The peak in advection-corrected NEE occurred just before midday. During the evening (17:00 – 20:00 PST), the advection-flux corrected NEE was slightly more positive than the u*-corrected NEE and NEE traces (the forest becoming a CO2 source earlier) due to a peak in FHA.  Figure 41: Ensemble average storage flux, advection corrected- and u*-corrected NEE. The cumulative CO2 flux traces in Figure 42 and Figure 43 give an idea of the total effect of each correction over a 23-day (summer) period. Figure 42 shows that the 65  Figure 42: Cumulative carbon balance over the 23-day main experiment, with corrections.  Figure 43: Cumulative carbon balance over the 23-day main experiment, showing various wind velocity tilt corrections.  66  net effect of FHA was positive, or a loss of CO2 to the atmosphere, while FVA (as shown here) added to the uptake of the forest in the latter half of the period. Adding both FHA and FVA to NEE gave a result similar to the u*-corrected NEE, while either adding a cosine correction to FHA or performing the planar fit using data points from neutral conditions had less of an effect on the cumulative carbon balance. Figure 43 shows three of the same traces as Figure 42 (NEE, u*-corrected NEE and planar-fit method advection flux-corrected NEE), as well as the effect of different tilt angle methods on the cumulative carbon balance of DF49. The “no tilt” trace shows the effect of neglecting tilt corrections and simply using the instrument coordinate system, normally aligned with gravity. In this case, large positive vertical velocity values during the day led to a strongly negative cumulative FVA. The tilt angle method (“TAM”), in contrast, removed most of the positive vertical velocity fluctuation during the day6, changing the cumulative FVA to a net source of CO2 (Figure 44). Hence, while the planar-fit method gave a cumulative CO2 flux that agrees with the u*-filter method, changing the rotation method gave a significantly different answer (Table 4). Obviously, the “no tilt” trace presents an extreme example, as failing to rotate the instrument‟s coordinate system essentially leads to the inclusion of horizontal velocity in the vertical velocity trace, significantly magnifying the vertical velocity. However, applying the planar fit or tilt angle method led to a difference in uptake of 39 g C m-2 over the entire period, or 1.73 g C m-2 day-1.  6  Using an averaging interval over the entire experiment, as with the PFM.  67  Figure 44: Cumulative FVA using the PFM and TAM. Table 4: Cumulative carbon balance values, with various corrections (g C m-2). NEE  NEE, u*  FHA  FVA, PFM  FVA, TAM  NEE + FHA  -86  -60  54  -25  15  -32  NEE + FVA, PFM  NEE + FHA + FVA, PFM  NEE + FHA, cos + FVA, PFM  NEE + FHA + FVA, neut.  NEE + FHA + FVA,No tilt  NEE + FHA + FVA, TAM  -111  -57  -63  -68  -226  -18  3.3  Third experiment  The third observational period at DF49 (Figure 5) examined CO2 concentration differences along the 2D advection transect. Four of the five along-slope sampling points were equally spaced, and one was placed between the tower and the next upslope sampling point at half the separation. All sampling points were placed 2.6 m 68  (±0.2 m) above local ground level. The experiment ran for 9 days (August 26 – September 3). Figure 45 shows the results of the experiment as ensemble averaged differences along the slope at different points, with all differences relative to the upslope sampling point. Therefore, the 21-m difference is between 73.5 m and 52.5 m on the advection transect. This sign convention recognizes the theory that CO2 concentration should build up along the wind direction during stable conditions (i.e. downslope katabatic flow). On average, the difference between the first and second sampling points was small, regardless of the time of day. However, the subsequent three downslope sampling points had sequentially increasing CO 2 differences (on average), especially during the day. The pattern was most pronounced (as with many other plots) during the morning and evening transition periods. Additionally, the 60-m sampling point, situated halfway between the 49-m and 74-m points, showed an appropriately small increase in CO2 difference over that at 49 m – as might be expected for its location. It should be noted that during individual half-hour periods, the 60-m difference was occasionally greater than the 74-m difference. Figure 46 shows the standard deviation of the differences shown in Figure 45; a similar pattern is observed, though the evening peaks do not occur at the same time. Figure 47 shows the standard deviation in CO2 concentration during the same period, for the same five sampling points as in Figure 45. A peak in variability was seen during the morning from 8:00 until 11:00 PST, coinciding with a maximum in CO2 differences at the same time, but a similar peak in σCO2 was absent for the afternoon peak in CO2 differences. This may be an indicator that slightly different mechanisms 69  Figure 45: Ensemble average horizontal CO2 differences.  Figure 46: Standard deviation of horizontal CO2 concentration differences.  70  Figure 47: Standard deviation of CO2 concentration. generated CO2 differences throughout the day during this particular observational period. In other words, the morning differences may have been driven by variations in mixing strength and above-to-below canopy coupling at different locations throughout the trunk space, while the afternoon difference could have been generated by the common horizontal advective flux theory of a stably stratified atmosphere hampering vertical CO2 transport. Comparison of this figure with Figure 28 shows that such a clear pattern was not present earlier in the summer at DF49. It is also possible that a relatively small number of events may have biased the ensemble average during the morning hours. It is interesting to note, however, that the ensemble average σCO2 was similar at all 5 sampling points at the 2.6-m height along the 73.5 m advection transect. The standard deviation of CO2 concentration also had a different diurnal pattern when compared to the standard deviation of horizontal CO 2 concentration differences. 71  3.4  Soil CO2 efflux  Figure 48 shows soil effluxes measured with a manual chamber method over the course of the entire summer along the advection transect. The majority of points fall between 5 and 10 µmol m-2 s-1 along the measured transect, with individual points occasionally showing large variability between measurement days. Notable characteristics of the graph are a (possible) minimum around the base of the tower (0 m), where traffic has created an area of bare soil, as well as the somewhat higher average soil efflux between 50 and 80 m upslope of the tower near the top of the advection transect (73.5 m). (Negative distances refer to locations downslope of the tower.) The measured flux gradient was somewhat contrary to expectations, as atmospheric CO2 concentration differences measured along the transect showed CO2 concentrations increasing down the slope towards the tower. This pattern could suggest either that there is a simple sign error in calculations of the atmospheric CO 2 difference presented in earlier sections, or that the atmospheric CO 2 difference close to the forest floor was not entirely controlled by a source gradient in soil efflux along the advection transect. Figure 49 and Figure 50 are companion traces to Figure 48; they show soil efflux measured with an auto-chamber at the base of the main flux tower during the second DF49 experiment. Some diurnal variation is observed, with brief peaks during the afternoon and several days of somewhat elevated efflux during the second warm period of the experiment (Figure 15). While the variation between day and night is clearly evident in Figure 49, it becomes less noticeable in the cumulative trace of Figure 50, which is a contrast to Figure 51 (cumulative FHA). 72  Figure 48: Soil CO2 efflux during the three experiments.  Figure 49: Auto-chamber soil CO2 efflux at the base of the flux tower during the second experiment. 73  Figure 50: Cumulative soil CO2 efflux over the second experiment.  Figure 51: Cumulative trace of FHA over the course of the second experiment.  74  Figure 50 and Figure 51 show that, while FHA may be driven in part by an input of CO2 from the soil, FHA is somewhat event-based and is obviously controlled by additional factors such as atmospheric mixing, stability and wind speed. Some explanatory power may also come from foliar respiration, both in the understory (Appendix 5) and crown space.  75  4 Conclusions During an eight-week period in the summer of 2009, an observational study was conducted to measure advective carbon dioxide fluxes in a Douglas-fir stand situated on a mild slope. Temperatures were 20-30 °C during the day and 10-20 °C at night, and a mixture of clear and overcast skies gave a wide range of meteorological conditions. A typical slope flow system was observed at the site during a majority of days, with upslope winds during the day and downslope flow at night. Wind speeds above the canopy were 1-5 m s-1, while within the subcanopy speeds were typically 0-1 m s-1. During the day, the wind direction was variable, while the direction of evening katabatic flow was more stable. During the afternoon, the subcanopy became decoupled from the atmosphere above due to earlier initiation of drainage flow. CO2 concentration ranged between 380 and 480 µmol mol-1 throughout the month of August, with typical subcanopy concentrations at night reaching 440-460 µmol mol-1. CO2 concentration was highest during transition periods between the anabatic and katabatic flow, especially close to the forest floor; spatial differences in CO 2 concentration were also largest during these times. The CO2 concentration measurement system used during the study followed designs by Black and McNaughton (1971) and Burns et al. (2009) in order to remove IRGA offset and gain error (respectively). Advective fluxes generated from this system, along with a single vertical profile of sonic anemometers, were similar in magnitude to the vertical turbulent flux measured above the canopy. However, larger outliers were 76  observed for the advective fluxes, with magnitudes up to 90 µmol m-2 s-1. Fluxes were smaller during the middle of the day and larger both at night and in the early morning. The distribution of FHA had a positive mean and a slightly positive skew, while the distribution of FVA had a negative mean closer to zero and a slight negative skew. Both distributions were Laplacian. The cumulative carbon balance of the Douglas-fir forest was calculated using several different corrections to the vertical turbulent flux, including the storage flux, the u*- filter, and the advective fluxes. Additionally, FVA was generated using two separate tilt correction algorithms to estimate the mean vertical velocity above the canopy. The vertical turbulent flux corrected using the subcanopy storage flux (NEE), was -86 g C m-2 over the 23-day main experiment. The u*-filter method predicted an uptake of –60 g C m-2, slightly less than NEE alone. The cumulative horizontal advective flux (cosine-corrected) over the same period was positive and 60% (55%) of NEE. Cumulative PFM (TAM) FVA was negative (positive) at 30% (15%) of NEE. Combined, the two advective fluxes generated a positive correction to NEE, regardless of coordinate rotation method; the range of carbon balance estimates given by the advective flux corrections was between -18 and -68 g C m. The two most typical methods, the u*-filter method and advective flux corrections using the PFM, gave similar estimates at -60 and -57 g C m-2, respectively. Similar to cumulative values, ensemble averaged u*-corrected and advective fluxcorrected NEE followed the same pattern over the course of a day. However, the advective flux-corrected NEE had a slightly greater average CO2 loss from the forest  77  during the morning and evening, and a slightly higher average uptake during the middle of the day. Measurement of the horizontal CO2 difference at several points along the advection transect for 9 days after the main experiment showed that the difference increased monotonically down slope (on average), especially during the day. In other words, when compared to the top of the transect, CO 2 concentration was usually higher at downslope locations. The differences were largest, as during the main experiment, during the morning and evening transition times between anabatic and katabatic flow. This result was in opposition to manual measurements of soil efflux conducted along the advection transect, which showed that the efflux was generally higher at the upper end of the transect. This apparent contradiction may validate horizontal advection theory, which states that CO2 concentration should increase along the horizontal wind direction if vertical turbulent transport is limited (Lee and Hu, 2002). A persistent relative difference in vertical turbulent exchange or even photosynthesis between the upslope and downslope locations may also have generated this pattern. Regardless of the generating mechanism, CO2 differences were measured horizontally within the canopy as well as vertically between the canopy and atmosphere above. This result indicates that advection of CO2 occurs at a horizontal scale of 75 m and a vertical scale of 40 m at this site, unless winds are calm.  78  Works Cited Aubinet, M., Heinesch, B. and Yernaux, M. (2003) Horizontal and vertical CO 2 advection in a sloping forest. Boundary-Layer Meteorology. 108: 397-417. Aubinet, M., Berbigier, P., Bernhofer, C. et al. (2005) Comparing CO2 storage and advection conditions at night at different Carboeuroflux sites. Boundary-Layer Meteorology. 116: 63-94. Aubinet, M. (2008) Eddy covariance CO2 flux measurements in nocturnal conditions: an analysis of the problem. Ecological Applications. 18: 1368-1378. Aubinet, M., Feigenwinter, C., Heinesch, B., Bernhofer, C., Canepa, E., Lindroth, A., Montagnani, L., Rebmann, C., Sedlak, P. and Van Gorsel, E. (2010) Direct advection measurements do not help to solve the night-time CO2 closure problem: Evidence from three different forests. Agricultural and Forest Meteorology. 150: 655-664. Baldocchi, D., Falge, E., Gu, L. et al. (2001). FLUXNET: A new tool to study the temporal and spatial variability of ecosystem-scale carbon dioxide, water vapor, and energy flux densities. Bulletin of the American Meteorological Society. 82: 2415-2434. Baldocchi, D. (2008) „Breathing‟ of the terrestrial biosphere: lessons learned from a global network of carbon dioxide flux measurement systems. Australian Journal of Botany. 56: 1-26.  79  Baldocchi, D., Finnigan, J., Wilson, K., Paw U, K. and Falge, E. (2000) On measuring net ecosystem carbon exchange over tall vegetation on complex terrain. Boundary-Layer Meteorology. 96: 257-291. Baldocchi, D., Hincks, B. and Meyers, T. (1988) Measuring biosphere-atmosphere exchanges of biologically related gases with micrometeorological methods. Ecology. 69: 1331-1340. Black, T. A., Den Hartog, G., Neumann, H. et al. (1996). Annual cycles of water vapour and carbon dioxide fluxes in and above a boreal aspen forest. Global Change Biology. 2: 219-229. Black, A. and McNaughton, K. (1971) Psychrometric apparatus for Bowen-ratio determination over forests. Boundary-Layer Meteorology. 2: 246-254. Burns, S., Delany, A., Sun, J., Stephens, B., Oncley, S., Maclean, G., Semmer, S., Schroter, J., and Ruppert, J. (2009) An evaluation of calibration techniques for in situ carbon dioxide measurements using a programmable portable trace-gas measuring system. Journal of Atmospheric and Oceanic Technology. 26: 291316. Christen A., Grimmond S., Roth M. and Pardyjak E. (2009). The IAUC Urban Flux Network - An international network of micrometeorological flux towers in urban ecosystems. Eos Trans. AGU. 90(52), Fall Meet. Suppl., Abstract B31C-06. de Araujo, A., Dolman, A., Waterloo, M., Gash, J., Kruijt, B., Zanchi, F., de Lange, J., Stoevelaar, R., Manzi, A., Nobre, A., Lootens, R., and Backer. J. (2010) The spatial variability of CO2 storage and the interpretation of eddy covariance fluxes in central Amazonia. Agricultural and Forest Meteorology. 150: 226-237.  80  Dellwik, E., Mann, J., Bingol, F. and Larsen, K. (2009) Mean vertical velocities and flow tilt angles at a fetch-limited forest site in the context of carbon dioxide vertical advection. Biogeosciences Discuss. 6: 8167-8213. Drewitt, G. (2002) Carbon dioxide flux measurements from a coastal Douglas-fir forest floor. Ph.D. Thesis, University of British Columbia. Etzold, S., Buchmann, N. and Eugster, W. (2010) Contribution of advection to the carbon budget measured by eddy covariance at a steep mountain slope forest in Switzerland. Biogeosciences Discuss. 7: 1633-1673. Falge, E., Baldocchi, D., Olson, R., Anthoni, P., Aubinet, M., Bernhofer, C., Burba, G., Ceulemans, R., Clement, R., Dolman, H., Granier, A., Gross, P., Grunwald, T., Hollinger, D., Jensen, N.-O., Katul, G., Keronen, P., Kowalski, A., Lai, C., Law, B., Meyers, T., Moncrieff, J., Moors, E., Munger, J., Pilegaard, K., Rannik, U., Rebmann, C., Suyker, A., Tenhunen, J., Tu, K., Verma, S., Vesala, T., Wilson, K. and Wofsy, S. (2001) Gap filling strategies for defensible annual sums of net ecosystem exchange. Agricultural and Forest Meteorology. 107: 43-69. Feigenwinter, C., Bernhofer, C., and Vogt, R. (2004) The influence of advection on the short term CO2-budget in and above a forest canopy. Boudary-Layer Meteorology. 113: 201-224. Feigenwinter, C., Bernhofer, C., Eichelmann, U. et al. (2008). Comparison of horizontal and vertical advective CO2 fluxes at three forest sites. Agricultural and Forest Meteorology. 148: 12-24.  81  Feigenwinter, C., Montagnani, L. and Aubinet, M. (2010a) Plot-scale vertical and horizontal transport of CO2 modified by a persistent slope wind system in and above an alpine forest. Agricultural and Forest Meteorology. 150: 665-673. Feigenwinter, C., Molder, M., Lindroth, A. and Aubinet, M. (2010b) Spatiotemporal evolution of CO2 concentration, temperature, and wind field during stable nights at the Norunda forest site. Agricultural and Forest Meteorology. 150: 692-701. Finnigan, J. (1999) A comment on the paper by Lee (1998): “On micrometeorological observations of surface-air exchange over tall vegetation”. Agricultural and Forest Meteorology. 97: 55-64. Finnigan, J. and Belcher, S. (2004) Flow over a hill covered with a plant canopy. Quarterly Journal of the Royal Meteorological Society. 130: 1-29. Finnigan, J. (2006) The storage term in eddy flux calculations. Agricultural and Forest Meteorology. 136: 108-113. Fitzjarrald, D. (1984) Katabatic wind in opposing flow. Journal of the Atmospheric Sciences. 41: 1143-1158. Fitzjarrald, D. (1986) Slope winds in Veracruz. Journal of Climate and Applied Meteorology. 25: 133-144. Goulden, M., Miller, S. and da Rocha, H. (2006) Nocturnal cold air drainage and pooling in a tropical forest. Journal of Geophysical Research. 111 (D8): D08S04. Goulden, M., Munger, J., Fan, S. et al. (1996). Measurements of carbon sequestration by long-term eddy covariance: methods and a critical evaluation of accuracy. Global Change Biology. 2: 169-182.  82  Gudiksen, P., Leone, J., King, C., Ruffieux, D. and Neff, W. (1992) Measurements and modeling of the effects of ambient meteorology on nocturnal drainage flows. Journal of Applied Meteorology. 31: 1023-1032. Heinesch, B., Yernaux, M. and Aubinet, M. (2007) Some methodological questions concerning advection measurements: a case study. Boundary-Layer Meteorology. 122: 457-478. Heinesch, B., Yernaux, Y. and Aubinet, M. (2008) Dependence of CO2 advection patterns on wind direction on a gentle forested slope. Biogeosciences. 5: 657668. Humphreys, E., Black, A., Morgenstern, K., Cai, T., Drewitt, G., Nesic, Z and Trofymow, J. (2006) Carbon dioxide fluxes in coastal Douglas-fir stands at different stages of development after clearcut harvesting. Agricultural and Forest Meteorology. 140: 6-22. Hunner, M., Siebicke, L. and Foken, T. (2009) Influence of coordinate rotation on calculation of vertical advection. Atmospheric Transport and Chemistry in Forest Ecosystems, Castle of Thurnau, Germany, Oct. 5-8, 2009. Johnson, M., Lehmann, J., Selva, E., Abdo, M., Riha, S. and Couto, E. (2006) Organic carbon fluxes within and streamwater exports from headwater catchments in the southern Amazon. Hydrological Processes. 20: 2599-2614. King, C. (1989) Representativeness of single vertical wind profiles for determining volume flux in valleys. Journal of Applied Meteorology. 28: 463-466. Kljun, N., Calanca, P., Rotach, M. and Schmid, H. (2004) A simple parameterization for flux footprint predictions. Boundary-Layer Meteorology. 112: 503-523.  83  Kondo, J. and Akashi, S. (1976) Numerical studies on the two-dimensional flow in horizontally homogeneous canopy layers. Boundary-Layer Meteorology. 10: 255272. Lee, X. (1998) On micrometeorological observations of surface-air exchange over tall vegetation. Agricultural and Forest Meteorology. 91: 39-49. Lee, X. and Hu, X. (2002) Forest-air fluxes of carbon, water and energy over non-flat terrain. Boundary-Layer Meteorology. 103: 277-301. Lee, X., Shaw, R. and Black, A. (1994) Modelling the effect of mean pressure gradient on the mean flow within forests. Agricultural and Forest Meteorology. 68: 201212. Leuning, R., Zegelin, S., Jones, K., Keith, H., and Hughes, D. (2008) Measurement of horizontal and vertical advection of CO2 within a forest canopy. Agricultural and Forest Meteorology. 148: 1777-1797. Loescher, H., Law, B., Mahrt, L., Hollinger, D., Campbell, J. and Wofsy, S. (2006) Uncertainties in, and interpretation of, carbon flux estimates using the eddy covariance technique. J. Geophys. Res. 111, D21890. Mahrt, L. (1982) Momentum balance of gravity flows. Journal of the Atmospheric Sciences. 39: 2701-2711. Mahrt, L., Vickers, D., Nakamura, R., Soler, M., Sun, J., Burns, S. and Lenschow, D. (2001) Shallow drainage flows. Boundary-Layer Meteorology. 101: 243-260. Mammarella, I., Kolari, P., Rinne, J., Keronen, P., Pumpanen, J. and Vesala, T. (2007) Determining the contribution of vertical advection to the net ecosystem exchange at Hyytiala forest, Finland. Tellus B. 59: 900-909.  84  Marcolla, B., Cescatti, A., Montagnani, L., Manca, G., Kerschbaumer, G. and Minerbi, S. (2005) Importance of advection in the atmospheric CO 2 exchanges of an alpine forest. Agricultural and Forest Meteorology. 130: 193-206. Massman, W. and Lee, X. (2002) Eddy covariance flux corrections and uncertainties in long-term studies of carbon and energy exchanges. Agricultural and Forest Meteorology. 113: 121-144. McMillen, R. (1988). An eddy correlation technique with extended applicability to nonsimple terrain. Boundary-Layer Meteorology. 43: 231-245. Moderow, U., Feigenwinter, C. and Bernhofer, C. (2007). Estimating the components of the sensible heat budget of a tall forest canopy in complex terrain. Boundary Layer Meteorology. 123: 99-120. Montagnani, L., Manca, G., Canepa, E., Georgieva, E., Acosta, M., Feigenwinter, C., Janous, D., Kerschbaumer, G., Lindroth, A., Minach, L., Minerbi, S., Molder, M., Pavelka, M. ,Seufert, G., Zeri, M. and Ziegler, W. (2009) A new mass conservation approach to the study of CO2 advection in an alpine forest. Journal of Geophysical Research. 114: 25 pp. Morgenstern, K., Black, A., Humphreys, E., Griffis, T., Drewitt, G., Cai, T., Nesic, Z., Spittlehouse, D. and Livingston, N. (2004) Sensitivity and uncertainty of the carbon balance of a Pacific Northwest Douglas-fir forest during an El Niňo/La Niňa cycle. Agricultural and Forest Meteorology. 123: 201-219. Papadopoulos, K. and Helmis, C. (1999) Evening and morning transition of katabatic flows. Boundary-Layer Meteorology. 92: 195-227.  85  Papale, D., Reichstein, M., Aubinet, M., Canfora, E., Bernhofer, C., Kutsch, W., Longdoz, B., Rambal, S., Valentini, R., Vesala, T. and Yakir, D. (2006) Towards a standardized processing of Net Ecosystem Exchange measured with eddy covariance technique: algorithms and uncertainty estimation. Biogeosciences. 3: 571-583. Paw U, K. T., Baldocchi, D., Meyers,T. and Wilson, K. (2000) Correction of eddycovariance measurements incorporating both advective effects and density fluxes. Boundary-Layer Meteorology. 97: 487-511. Poggi, D., Katul, G., Finnigan, J. and Belcher, S. (2008) Analytical models for the mean flow inside dense canopies on gentle hilly terrain. Quarterly Journal of the Royal Meteorological Society. 134: 1095-1112. Pyles, R., Paw U, K., and Falk, M. (2004) Directional wind shear within an old-growth temperate rainforest: observations and model results. Agricultural and Forest Meteorology. 125: 19-31. Pypker, T., Unsworth, M., Mix, A., Rugh, W., Ocheltree, T., Alstad, K. and Bond, B. (2007) Using nocturnal cold air drainage flow to monitor ecosystem processes in complex terrain. Ecological Applications. 17: 702-714. Queck, R. and Bernhofer, C. (2010) Constructing wind profiles in forests from limited measurements of wind and vegetation structure. Agricultural and Forest Meteorology. 150: 724-735. Staebler, R. and Fitzjarrald, D. (2004) Observing subcanopy CO 2 advection. Agricultural and Forest Meteorology. 122: 139-156.  86  Staebler, R. and Fitzjarrald, D. (2005) Measuring canopy structure and the kinematics of subcanopy flows in two forests. Journal of Applied Meteorology. 44: 1161-1179. Stull, R. (1988) An introduction to boundary layer meteorology. Kluwer Academic Publishers. Dordrecht, The Netherlands. Sun, H., Clark, T., Stull, R. and Black, A. (2006a) Two-dimensional simulation of airflow and carbon dioxide transport over a forested mountain. Part I: Interactions between thermally-forced circulations. Agricultural and Forest Meteorology. 140: 338-351. Sun, H., Clark, T., Stull, R. and Black, A. (2006b) Two-dimensional simulation of airflow and carbon dioxide transport over a forested mountain. Part II: Carbon dioxide budget analysis and advection effects. Agricultural and Forest Meteorology. 140: 352-364. Sun, J., Desjardins, R., Mahrt, L. and MacPherson, I. (1998) Transport of carbon dioxide, water vapour, and ozone by turbulence and local circulations. Journal of Geophysical Research. 103: 25873-25885. Sun, J., Burns, S., Delany, A., Oncley, S., Turnipseed, A., Stephens, B., Lenschow, D., LeMone, M., Monson, R. and Anderson, D. (2007) CO 2 transport over complex terrain. Agricultural and Forest Meteorology. 145: 1-21. Tanner, C. and Thurtell, G. (1969) Anemoclinometer measurements of reynolds stress and heat transport in the atmospheric surface layer. Research and Development Technical Report, ECOM 66-G22-F, Final Report. Vickers, D. and Mahrt, L. (2006) Contrasting mean vertical motion from tilt correction methods and mass continuity. Agricultural and Forest Meteorology. 138: 93-103.  87  Wang, W., Davis, K., Cook, B., Bakwin, P., Yi, C., Butler, M. and Ricciuto, D. (2005) Surface layer CO2 budget and advective contributions to measurements of net ecosystem-atmosphere exchange of CO2. Agricultural and Forest Meteorology. 135: 202-214. Wang, W. (2009) The influence of topography on single-tower-based carbon flux measurements under unstable conditions: a modeling perspective. Theoretical and Applied Climatology. DOI 10.1007/s00704-009-0130-0 Wilczak, J., Oncley, S. and Stage, S. (2001) Sonic anemometer tilt correction algorithms. Boundary-Layer Meteorology. 99: 127-150. Wofsy, S., Goulden, M., Munger, J. et al. (1993) Net exchange of CO2 in a mid-latitude forest. Science. 260: 1314-1317. Yang, P., Black, A., Neumann, H., Novak, M. and Blanken, P. (1999) Spatial and temporal variability of CO2 concentration and flux in a boreal aspen forest. Journal of Geophysical Research. 104: 27653-27661. Yi, C., Anderson, D., Turnipseed, A., Burns, S. et al. (2008) The contribution of advective fluxes to net ecosystem exchange in a high-elevation, subalpine forest. Ecological Applications. 18: 1379-1390.  88  Appendices Appendix 1: Switching system mathematics The solenoid valve switching system was designed to remove individual instrument offset error while still maintaining near-complete coverage at all sampling points in the experimental design (Black and McNaughton, 1971). Assuming no gain error (at least partially removed by the twice daily CO 2 gas calibration), offset error from two paired instruments is removed as follows: Assume one switching cycle of six minutes, during which instrument 1 measures first at point A (3 min, period-1) and then subsequently at point B (3 min, period-2). Instrument 2 measures at point B followed by A during the same cycle (Figure 52). Each instrument reports 180 seconds of CO2 concentration data for each point, which is used to generate a three-minute average, C. Additionally, assume that each instrument has a constant (but unique) offset error, ε. Then, in order to generate an average difference measurement of CO2 concentration between point A and point B over six minutes (DA-B), one may calculate: DAB, period1  (C1A   1 )  (C2B   2 )  DAB, period2  (C2 A   2 )  (C1B   1 )  DA  B   DA  B , period 1  DA  B , period  2 2    (C1A   1 )  (C2 B   2 )  (C2 A   2 )  (C1B   1 ) 2  89  D A B   C1A   1  C 2B   2  C 2 A   2  C1B   1 2  D A B   C1A  C 2B  C 2 A  C1B 2  Throughout the 6 minutes, instrument drift (e.g. caused by temperature change) must be small to maintain a constant offset ε between period 1 and 2. However, C and hence D can change between period 1 and 2 with no effect. In other words, stationarity in CO2 concentration between the two periods is not a requirement.  Figure 52: Example of CO2 concentration traces from individual IRGAs while switching.  90  An important question is whether the gain in accuracy given by the switching method is worth doubling the number of instruments. A 6-min difference can also be generated by a single instrument switching between upslope and downslope sampling points (with 50% coverage at each point). This difference can be compared to the switching difference to generate an idea of any possible improvement. Difference scatter plots (Figure 53 and Figure 54) and ensemble average differences (Figure 55 and Figure 56) comparing the two methods for the 1-m level during the second experiment are shown below. Both a 6-min and a 30-min difference timescale are considered; 6 minutes is the minimum timescale possible with the data, and 30 minutes is the current averaging period of turbulent flux computations at DF49. Symmetry of the difference estimates is visible in the scatter plots – if one instrument generates a difference larger than the switching difference estimate, the other instrument will generate a smaller difference with the same error. The 6-minute differences also show larger magnitudes and greater error between estimates when compared to the 30-minute differences. The ensemble averages show a similar trend to the scatter plots, with greater variability at the 6-minute timescale. However, the average difference generated with a single instrument (over the 23-day main experiment) is quite close to that generated by the switching procedure. It is probable that errors generated by using a single instrument are random and would average to zero over a long measurement campaign.  91  Figure 53: Switching difference comparison, 6 minute timescale.  Figure 54: Switching difference comparison, 30 minute timescale.  92  Figure 55: Ensemble average difference comparison, 6 minute timescale.  Figure 56: Ensemble average difference comparison, 30 minute timescale.  93  Appendix 2: Westham Island sonic anemometer comparison  Figure 57: Westham Island sonic anemometer calibration setup. The Westham Island long-term monitoring tower is visible against the sky to the left of the array. Zero offsets for the CSAT3 sonic anemometers were also determined from the Westham Island intercomparison (Figure 57). Offsets were calculated by regressing average u, v, and w values of SN 0126, 1341, 1393 and 1396 against a reference sonic (SN 1389) and correcting for the average zero offset of the reference (Table 5). During each measured half-hour, individual 10 Hz data points were filtered if the instantaneous wind direction was from behind the sonic anemometers or along the mounting bar (Figure 57), translating in the CSAT3 coordinate system to acceptance angles between 125°-245° („Restr.‟). If 80% or greater of the 18000 points in each half-hour were accepted, the half-hour was included in the zero offset regression. Offset values were  94  different from those determined in the laboratory, and varied depending on whether or not the wind direction criterion was applied. Table 5: Sonic anemometer zero offsets calculated from Westham Island, spring of 2009. (Against SN 1389) Coordinate  Restr.  All  Restr.  All  Restr.  All  Restr.  All  u (cm s-1)  6.58  -4.39  -5.29  -1.17  0.89  -0.46  -4.00  -0.45  v (cm s )  -0.36  5.14  4.73  2.31  3.69  -3.51  0.51  0.12  w (cm s-1)  3.36  3.41  -2.86  -2.98  -2.59  -1.49  -0.76  -0.95  -1  SN 0126  SN 1341  SN 1393  SN 1396  Zero offsets were additionally determined from the Westham Island comparison by Kate Liss and Andreas Christen (Table 6) using a ±45° and 90% acceptance criterion (also corrected for the reference sonic zero offsets). Offsets differed from both of the previously reported analyses. Table 6: Sonic anemometer zero offsets calculated from Westham Island, spring of 2009. Analysis performed by Kate Liss and Andreas Christen. (Against SN 1389) u (cm s-1) -1  v (cm s ) -1  w (cm s )  SN 0126  SN 1341  SN 1393  SN 1396  -3.35  -3.77  -0.55  -0.36  -4.97  3.93  3.41  1.20  1.93  -2.99  -2.52  -3.71  95  Appendix 3: Post-experiment IRGA calibration All eight IRGAs were calibrated in the laboratory before and after the advection experiments. Assuming perfect calibration before they left the laboratory (a rough assumption as not all IRGAs were calibrated exactly at the same time), information from the post-experiment calibrations gives information about instrument drift over the 6week field depoloyment. A perfect instrument with no drift would have had a row of values similar to the following: [ 0.0 | 0.0 | 420.94 | 14.0 / 14.0 ]. Table 7: Post-experiment IRGA calibration results.  -0.67  Span CO2 (420.94 µmol mol-1 goal) 419.2  Span H2O (mmol mol-1 before and after) 13.88 / 13.98  -0.81  -0.056  418.0  13.82 / 13.90  SN 160 (COM 7)  12.7  -1.11  416.9  14.11 / 13.99  SN 708 (COM 8)  -9.6  0.33  418.45  14.00 / 14.05  SN 234 (COM 9)  3.0  0.11  420.0  13.65 / 14.09  SN 483 (COM 10)  2  -0.5  418.9  14.09 / 13.93  SN 174 (COM 11)  -0.7  0.05  421.05  14.21 / 14.30  SN 321 (COM 12)  -2.1  -0.03  422.5  13.80 / 13.88  IRGA  Zero CO2 (µmol mol-1)  Zero H2O (mmol mol-1)  SN 253 (COM 5)  5.2  SN 254 (COM 6)  The table shows values before correction for zero and CO 2 span. H2O is spanned using a dewpoint value; both before and after mmol mol -1 values are therefore shown. CO2 concentrations are mole fraction (uncorrected for water vapour), though calibration gases are dry. COM 5-8 were LI-7000s; COM 9-12 were LI-6262s. COM 5 & 6 measured at 1 m during the second DF49 experiment; 7 & 8 at 2.6 m, 9 & 10 at 8 m and 11 & 12 at 20 m.  96  Appendix 4: High-frequency time evolution of morning subcanopy CO2 concentration. The largest vertical CO2 differences observed at DF49 occurred during the morning transition between katabatic and anabatic flow (Figure 33). The figures shown here are an example time series of the subcanopy CO 2 concentration profile for the morning of August 9th, 2009. The higher subcanopy heights (8 and 20 m) become increasingly coupled with the atmosphere above, while the lower levels remain at a relatively high CO2 concentration. Lower levels are continuous lines and higher levels are dashed; -u denotes the upslope profile and -d denotes the downslope or flux tower profile. CO2 concentrations are mixing ratios.  Figure 58: Subcanopy CO2 concentration profile, 5:30 – 6:00 PST  97  Figure 59: Subcanopy CO2 concentration profile, 6:00 – 6:30 PST  Figure 60: Subcanopy CO2 concentration profile, 6:30 – 7:00 PST  98  Figure 61: Subcanopy CO2 concentration profile, 7:00 – 7:30 PST  Figure 62: Subcanopy CO2 concentration profile, 7:30 – 8:00 PST  99  Figure 63: Subcanopy CO2 concentration profile, 8:00 – 8:30 PST  Figure 64: Subcanopy CO2 concentration profile, 8:30 – 9:00 PST  100  Figure 65: Subcanopy CO2 concentration profile, 9:00 – 9:30 PST  101  Appendix 5: Understory leaf CO2 exchange Soil respiration measured along the advection transect, presumed to effect the measured horizontal CO2 differences, will be modified by the understory plants. Specifically, CO2 will be added to the trunk space air by leaves respiring at night, while CO2 will be removed during the day as the leaves photosynthesize. An estimate of this flux was measured using a portable photosynthesis system (model LI-6400, LI-COR Biosciences Inc.) along the advection transect at midday on August 25, 2009. Leaves were darkened (0 µmol m-2 s-1 PAR) to generate an estimate of respiration, and net exchange was also measured at a range of light intensities (Figure 66). Leaf exchange is normalized per unit leaf area (different from per unit ground area).  Figure 66: Understory net leaf exchange.  102  Appendix 6: Site pictures  Figure 67: Instrumentation table with calibration gas tanks.  Figure 68: Flow control and electronics for the IRGA system. 103  Figure 69: DF49 canopy with upslope 1-m and 2.6-m sampling points.  Figure 70: Upslope co-located sampling points during the third experiment. The next sampling point is visible in the background.  104  Figure 71: Calibration gas distribution box.  105  Appendix 7: IRGA calibration period Between the second and third experiments, short pieces of Synflex tubing were attached to the IRGA system in place of the 60-m long sampling tubes. All four differencing pairs were set to measure the same environmental difference. Though differences in IRGA cell pressure from the normal sampling setup existed, and the IRGAs were not calibrated during this period, CO2 differences give some idea as to the comparability of the four systems.  Figure 72: Calibration period IRGA CO2 differences and error between the four systems.  106  Appendix 8: Transect map  Figure 73: Plan view of the advection transect during the third experiment at DF49, summer of 2009.  107  

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.24.1-0052565/manifest

Comment

Related Items