Assessing decoupling of above and below canopy air masses at a Norway spruce stand in complex terrain

https://doi.org/10.1016/j.agrformet.2020.108149Get rights and content

Highlights

  • Decoupling and sub-canopy horizontal flow can bias forest NEE estimates.

  • The common u* filtering is not applicable at the study site.

  • Two new approaches for two-level filtering are introduced.

  • These two new approaches are applicable on short time scales.

  • Sonic anemometers above and below canopy are needed to reveal decoupling.

Abstract

Concurrent below and above canopy sonic anemometer vertical velocity (w) measurements reveal frequent decoupling events between the air masses below and above the canopy at a dense spruce forest stand in mountainous terrain. Decoupling events occurred predominantly during nighttime but not exclusively. Several single-level approaches based on steady state and integral turbulence characteristic tests as well as friction velocity (u*) filtering and two-level CO2 flux filtering methods are tested. These tests aimed at evaluating the filtering schemes to address decoupling and its effect on above canopy derived eddy covariance net ecosystem CO2 exchange (NEE). In addition to the already existing two-level filtering approach based on the correlation of σw above and below canopy, two new filtering methods are introduced based on w raw data below and above the canopy. One is a telegraphic approximation agreement, which assumes coupling when w both above and below canopy are pointing in the same direction. Another one evaluates the cross-correlation maximum between below and above canopy w data. This study suggests that none of the single-level approaches can detect decoupling when compared to two-level filtering approaches. It further suggests that the newly introduced two-level approaches based on w raw data may have advantages in comparison to the conventional σw approach regarding their flexibility on shorter time scales than one year. We tested the correlation of the newly introduced filtering approaches with the parameters u*, global radiation, buoyancy forcing across the canopy and wind shear across the canopy. In any case, this correlation was not existing or weakly positive, suggesting that concurrent below and above canopy measurements are necessary for addressing decoupling. Sonic anemometer measurements near the forest floor and above the canopy are adequate to apply the new procedures and can be implemented in a routine manner at any forest site globally.

Introduction

Few dispute the claim that forests play a major role in ameliorating the effects of rising atmospheric CO2 on climate (Carvalhais et al., 2014; Le Quere et al., 2018). Long-term measurements of the net removal of atmospheric CO2 molecules by forests yield knowledge about how forests uptake CO2 and to what extent this CO2 remains in carbon stocks. Over the last several decades, the eddy covariance (EC) method has become an accepted standard method for determining the forest net ecosystem exchange (NEE) of CO2 (e.g. Baldocchi, 2003; Luyssaert et al., 2007; Schwalm et al., 2010).

Because such measurements are required over averaging intervals that are less than one hour, 24 hours a day, 365 days a year and over the lifecycle of a forest, a major question arises as to how representative are the above canopy derived NEE values for the probed ecosystem. For a stationary and planar homogeneous flow in the absence of subsidence and at high Reynolds number, the space and time-averaged mean scalar continuity equation (for CO2) is set asNEE=0hS(z)dz+Fg,where S are within-canopy net sources and sinks of CO2, h is the mean canopy height, z is the vertical distance from the forest floor (assumed to be at z = 0), and Fg is the ground or forest floor CO2 efflux at z = 0. In practice, when applying the EC method for a specific time-averaging interval dt, NEE sums up via the turbulent canopy-atmosphere CO2 exchange and changes in CO2 storage as well as potential advection processes within the chosen time interval with varying relevance of these processes depending on the environmental conditions.

The air masses near the forest floor and within the canopy can be decoupled from the air masses above the canopy due to canopy elements blocking eddy penetration and divergence in atmospheric stratification (e.g. Cionco, 1983; Pinker, 1983). Complex topography surrounding the tower in combination with the aforementioned decoupling may then induce below canopy advection exchanging respired CO2 with areas outside of the EC footprint (e.g. Aubinet et al., 2005; Belcher et al., 2012; Butler et al., 2015; Feigenwinter et al., 2004, 2008; Jocher et al., 2017, 2018; Kutsch & Kolari, 2015; Staebler & Fitzjarrald, 2004; Wang et al., 2017). As one and most common consequence, the above canopy EC system measures only a part of the total NEE and overestimates the net CO2 uptake from the atmosphere primarily due to some missing respiration components (e.g. Fg) contributing to the above canopy measurement (e.g. Jocher et al., 2017; 2018).

It suffices to state that the conventional approach to address insufficient mixing between below and above canopy air masses is to use only above canopy CO2 fluxes above a certain friction velocity (u*) threshold for further analysis (Papale et al., 2006). A high u* is presumed to lead to large mechanical production of turbulent kinetic energy just above the canopy thereby producing energetic eddies that can penetrate the entire canopy depth, breaking-up any strongly stratified layers within canopy that promote drainage of CO2 near the ground, and flushing out much of the respired CO2 to the canopy-free zone within a narrow footprint of the EC system. This u* value is determined by evaluating the relation between u* and nighttime above canopy CO2 fluxes, normalized for temperature dependency. This so-called u* filtering assumes that nighttime above canopy CO2 fluxes become ideally independent of u* at a certain u* threshold. Above this experimentally determined site specific u* threshold, one assumes that the above canopy measurements represent the entire ecosystem CO2 exchange (e.g. Goulden et al., 1996; Gu et al., 2005). The original ecological justification of the u* approach was that NEE can be estimated as a function of the physical environment during periods with low turbulence as respiration rates change on much slower time scales than turbulence does (Goulden et al., 1996). There is some appeal to the simplicity of a u* filtering as the implementation of alternative approaches is usually more complicated. However, it is not possible at all sites to determine a reliable u* threshold as above canopy nighttime CO2 fluxes may fall in clearly indefinable regimes. Furthermore, it might occur that the filtering at a certain u* threshold is not sufficient to ensure full mixing between below and above canopy air masses (e.g. Aubinet et al., 2012; Jocher et al., 2017; Speckmann et al., 2015; Thomas et al., 2013), especially over tall and dense canopies.

Prior to the application of the u* filtering, data are commonly evaluated and filtered for their quality by a quality flagging scheme that tests the EC data for stationarity and development of turbulence. These are two major preconditions for the usage of EC methods to determine NEE as noted earlier (Foken et al., 2004). This flagging scheme addresses decoupling issues only indirectly by testing the degree of turbulence activity relative to the low frequency variation in the mean state. Ruppert et al. (2006) used this quality filtering scheme also as an independent filtering approach without any subsequent filtering steps. Only those data with best quality were used then for further analysis of NEE. In the current study, both the quality filtering by Foken et al. (2004) before further filtering steps and the evaluation of this quality filtering as an independent filtering approach are considered.

Recently, increased concern has been raised that single-level CO2 flux filtering approaches based on above canopy data alone might not be sufficient to correct for the biasing influence of decoupling events (Alekseychik et al., 2013; Jocher et al., 2017; Thomas et al., 2013). To address this issue, Thomas et al. (2013) proposed a two-level filtering based on the relation between the vertical wind velocity standard deviations (σw) from above and below the main canopy measurements. Hereafter, below canopy measurements refer to an EC sensor situated below the main crown but above the forest floor. The mentioned relation is linear if the above and below canopy air masses are fully coupled and vanishes for decoupled conditions. The approach by Thomas et al. (2013) identifies a threshold for both σw above and below the canopy above which full coupling is assumed. Commonly, these thresholds are identified using a longer period of measurements to have a robust data basis for the threshold deduction and represent fixed values specific for the given site. However, the transition from decoupled to coupled conditions is dynamic and one pair of ‘hard’ thresholds for e.g. one year may overrate decoupled periods throughout the year as thresholds may vary slightly in the course of the seasons (Thomas et al., 2013; Jocher et al., 2018).

With the aim to overcome this potential shortcoming of the approach by Thomas et al. (2013), two alternative approaches are proposed and explored here. They are based on the same measurement setup, i.e. above and below the main canopy w measurements, yet founded on analyses of the instantaneous vertical velocities instead of the half-hourly statistics to identify the level of coupling between above and below canopy. The first of these two approaches uses an index of agreement in telegraphic approximation of w (TAa; e.g. Cava & Katul, 2009) that identifies portion of records within each flux averaging interval for which both the direction of the turbulent w above and below the canopy is the same. A high fraction of TAa within a flux averaging interval suggests that air masses above and below the canopy are expected to be coupled while low TAa points to decoupling. The second approach uses the cross-correlation maximum (CCFmax) between above and below canopy w within each flux averaging interval as a measure of the coupling degree between above and below canopy air masses (cf. e.g. Foken (2017) where the cross-correlation was used as an indicator for horizontal coupling of air masses). While TAa is considering only the temporal agreement in w, CCFmax takes also the magnitude of w into account.

With this background, the aims here are: i) the evaluation of the applicability of different kinds of single- and two-level CO2 flux filtering approaches at a spruce forest site that do not conform to the niceties used in deriving eq. (1), ii) the quantification of the effect of these CO2 flux filtering approaches on the above canopy derived forest carbon exchange, iii) the derivation of recommendations on how to treat carbon exchange EC data at forested sites in complex terrain so as to minimize potential biasing effects introduced by decoupling and advection on the above canopy derived carbon fluxes.

Section snippets

Site description and characteristics

The experimental ecological study site Bílý Kříž (49° 30’ 17’’ N, 18° 32’ 28’’ E, 800-900 m a.s.l.) is located in the Moravian-Silesian Beskydy Mountains, the Czech Republic. It is situated on a SSW-oriented planar slope next to a WE-oriented mountain crest (Fig. 1; Sedlák et al. (2010)). The experimental forest is a Norway spruce (Picea abies (L.) H. Karst.) monoculture with only 1 % of silver fir (Abies alba Mill.), planted in 1981 with four years old seedlings on an area of 6.5 ha (

filtering based on quality flags

This filtering left 45 % of the initial unfiltered data set for further analysis. The filtered fraction of data clearly differed between nighttime and daytime periods. While 53 % of the initial unfiltered daytime data passed the quality flag filtering, during nighttime it was only 35 % of the initial unfiltered data. The quality checking procedure tests the data for stationarity and development of turbulence (Foken et al., 2004). Daytime periods were defined as periods with Rg > 0 W m-2.

u* filtering

The

Conclusions

The current study evaluates different types of single- and two-level EC flux filtering approaches in terms of canopy decoupling events and its potentially biasing influence on above canopy derived CO2 fluxes at long time scales. Decoupling is a common issue at the study site not only during nighttime when turbulence is dampened due to missing radiative input but might be also relevant during daytime. Two new filtering approaches based on below and above canopy w data were proposed. Their

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgments

This work was supported by the Ministry of Education, Youth and Sports of CR within the CzeCOS program, grant number LM2018123. MF, MP and LŠ were supported by the project SustES - Adaptation strategies for sustainable ecosystem services and food security under adverse environmental conditions (CZ.02.1.01/0.0/0.0/16_019/0000797).

References (55)

  • R.M. Staebler et al.

    Observing subcanopy CO2 advection

    Agric. For. Meteorol.

    (2004)
  • C. Thomas et al.

    Toward biologically meaningful net carbon exchange estimates for tall, dense canopies: Multi-level eddy covariance observations and canopy coupling regimes in a mature Douglas-fir forest in Oregon

    Agric. For. Meteorol.

    (2013)
  • X. Wang et al.

    Quantifying and reducing the differences in forest CO2-fluxes estimated by eddy covariance, biometric and chamber methods: a global synthesis

    Agric. For. Meteorol.

    (2017)
  • M. Aubinet et al.

    Comparing CO2 storage and advection conditions at night at different carboeuroflux sites

    Bound. Layer Meteorol.

    (2005)
  • D.D. Baldocchi

    Assessing the eddy covariance technique for evaluating carbon dioxide exchange rates of ecosystems: Past, present and future

    Global Change Biology

    (2003)
  • S.E. Belcher et al.

    The wind in the willows: Flows in forest canopies in complex terrain

    Annu. Rev. Fluid Mech.

    (2012)
  • B.W. Butler et al.

    High-resolution observations of the near-surface wind field over an isolated mountain and in a steep river canyon

    Atmos. Chem. Phys.

    (2015)
  • N. Carvalhais et al.

    Global covariation of carbon turnover times with climate in terrestrial ecosystems

    Nature

    (2014)
  • D. Cava et al.

    The effects of thermal stratification on clustering properties of canopy turbulence

    Bound. Layer Meteorol.

    (2009)
  • R.M. Cionco

    On the coupling of canopy flow to ambient flow for a variety of vegetation types and densities

    Bound. Layer Meteorol.

    (1983)
  • C. Feigenwinter et al.

    The influence of advection on the short term CO2-budget in and above a forest canopy

    Bound. Layer Meteorol.

    (2004)
  • Foken, T. (Ed.), 2017. Energy and matter fluxes of a spruce forest ecosystem. Springer International Publishing, Basel,...
  • T. Foken et al.

    Post-field data quality control

  • G. Fratini et al.

    Towards a consistent eddy-covariance processing: An intercomparison of EddyPro and TK3

    Atmos. Meas. Techn.

    (2014)
  • A. Freundorfer et al.

    Forest wind regimes and their implications on cross-canopy coupling

    Agric. For. Meteorol.

    (2019)
  • M.L. Goulden et al.

    Measurements of carbon sequestration by long-term eddy covariance: Methods and a critical evaluation of accuracy

    Global Change Biol.

    (1996)
  • Cited by (10)

    • Detecting nighttime inversions in the interior of a Douglas fir canopy

      2022, Agricultural and Forest Meteorology
      Citation Excerpt :

      So-called u* (friction velocity) filtering is often used to filter flux data for periods with low wind shear (Goulden et al., 1996; Papale et al., 2006; Barr et al., 2013). Other methods use the relationship between the standard deviation of vertical wind speed (σw), both above and below the canopy (Thomas et al., 2013), or look at (cross-)correlation between the measurements of vertical wind speeds (Cava and Katul, 2009; Jocher et al., 2020). However, the above methods are only proxies for the momentum exchange, and do not actually detect the density stratification over the full canopy-atmosphere continuum.

    • Direct partitioning of eddy-covariance water and carbon dioxide fluxes into ground and plant components

      2022, Agricultural and Forest Meteorology
      Citation Excerpt :

      Under such circumstances, any partitioning method that uses EC fluxes (independent of the formulation) will incur uncertainties. To overcome this limitation, previous studies (Paul-Limoges et al., 2017; Thomas et al., 2013) used below and above canopy EC data to investigate turbulent mixing and to identify periods of decoupling, a topic that continues to spawn research interest for ideal and non-ideal conditions alike (Jocher et al., 2020). More broadly, the investigation of CEC, MREA, and FVS methods over four sites elucidated the main weaknesses and strengths of these approaches, and yielded findings on the conditions under which they (dis)agree.

    View all citing articles on Scopus
    View full text