Agricultural and Forest Meteorology 91 (1998) 39±49 On micrometeorological observations of surface-air exchange over tall vegetation Xuhui Lee* School of Forestry and Environmental Studies, Yale University, New Haven, Connecticut 06511, USA Received 14 October 1997; received in revised form 18 February 1998; accepted 20 February 1998 Abstract It is shown from the mass conservation and the continuity equations that the net ecosystem exchange (NEE) of a scalar constituent with the atmosphere should be Z Z zr @c 1 zr 0 0 NEE dz wcr wr cr cdz 0 @t zr 0 where the ®rst term on RHS is the storage below the height of observation (zr), the second term is the eddy ¯ux, and the third term is a mass ¯ow component arising from horizontal ¯ow convergence/divergence or a non-zero mean vertical velocity (wr) at height zr. The last term, unaccounted for in previous studies of surface-air exchanges,R becomes important over tall zr vegetation and at times when the vertical gradient of the atmospheric constituent cr 1=zr 0 cdzis large, as is the case with CO2 in forests at night. Experimental evidence is presented to support the postulation that the mass ¯ow component is in large part responsible for the large run-to-run variations in eddy ¯uxes, the lack of energy balance closure and the apparent low eddy ¯uxes at night under stable strati®cations. Three mechanisms causing the non-zero mean vertical velocity are discussed. Of these, drainage ¯ow on undulating terrain is the most important one for long-term ¯ux observations because only a small terrain slope is needed to trigger its occurrence. It is suggested from the data obtained at a boreal deciduous forest that without proper account of the mass ¯ow component, the assessment of annual uptake of CO2 could be biased signi®cantly towards higher values. It is argued that quantifying the mass ¯ow component is a major challenge facing the micrometeorological community. # 1998 Elsevier Science B.V. All rights reserved. Keywords: Net ecosystem exchange; Drainage; CO2; Forest 1. Introduction CO2 uptake by terrestrial ecosystems (Goulden et al., 1996a; Black et al., 1996; Grace et al., 1995), In recent years micrometeorological techniques air-vegetation exchanges of pollutants (Fuentes et al., have been widely used in observational studies of 1992; Guenther et al., 1996) and trace metals (Lindberg et al., 1997), and ecosystem processes *Corresponding author. Tel.: (203)432-6271; fax: (203)432- (Baldocchi and Harley, 1995; Hollinger et al., 1994; 3929; e-mail: [email protected] Verma et al., 1986). One major concern with the 0168-1923/98/$19.00 # 1998 Elsevier Science B.V. All rights reserved. P I I S 0 1 6 8 - 1 9 2 3 ( 9 8 ) 0 0 0 7 1 - 9 40 X. Lee / Agricultural and Forest Meteorology 91 (1998) 39±49 observations is the lack of energy balance closure and 2. Theory the large run-to-run variations in scalar ¯uxes, a problem that is particularly serious over tall vegeta- Almost all micrometeorological studies of surface± tion. Mechanisms such as advection due to inadequate air exchange assume a zero mean vertical velocity or fetch (Moore and Hicks, 1973; Fitzjarrald and Moore, zero dry air ¯ux. While this may be a good assumption 1994), a mismatch of net radiometer and eddy corre- very close to the ground, there are many reasons to lation (EC) footprints (Moore et al., 1996), instrument question its validity over tall vegetation. For example, uncertainties (Shuttleworth et al., 1988; Thompson, the horizontal divergence rate frequently used for 1979), and descending motion associated with scale analysis of synoptic weather systems is stationary convective cells (Lee and Black, 1993b) 105 s1 (Wallace and Hobbs, 1977), which amounts have been proposed as possible causes of the problem. to a vertical velocity of 0.04 cm s1 at the height of A thorough survey of the problem can be found in 40 m above the ground, or roughly two tree heights. Mahrt (1998). (In the following, this height is taken as a typical Another problem in the forest environment, fre- height for observations of surface-air exchange over quently encountered at nights of low wind speed, is the forests.) The vertical velocity associated with local apparent low EC ¯ux. The best example for this is thermal circulations (Segal et al., 1988; Mahrt and Ek, provided by CO2 ¯ux observations (Goulden et al., 1993; Sun et al., 1997), topographically modi®ed ¯ow 1996b; Jarvis et al., 1997). Possible causes of the (Finnigan and Brunet, 1995), and divergence in con- problem include drainage ¯ow (Grace et al., 1996), vective cell-like structures will be at least an order of advection (Sun et al., 1997), line averaging and tube magnitude larger. An extended discussion on this attenuation (Moncrieff et al., 1996; Leuning and subject is given in Section 4. Judd, 1996) To quantify the effect of the non-zero vertical The objective of this study is to examine the velocity on NEE, we begin with the conservation effect of a non-zero mean vertical velocity or hori- equation of a scalar c in the x±z plane, ignoring the zontal ¯ow divergence/convergence on the atmo- molecular term spheric exchange over forests. Central to the study @c @ uc @ wc is a formulation for the net ecosystem exchange (NEE) s (1) @t @x @z of a scalar constituent with the atmosphere derived from the mass conservation equation. It is shown where axis x is aligned with the mean wind direction in Section 2 that a non-zero mean vertical velocity and axis z is perpendicular to the local terrain surface, at the height of ¯ux observations will lead to a s is a source term, and u and w are the velocity signi®cant additional mass ¯ow component of NEE components in the x and z directions, respectively. unaccounted for by previous studies. Data from two After Reynolds decomposition and averaging, Eq. (1) forest turbulence experiments, one at Browns River leads to (BR, Lee and Black, 1993a, b) and the other at Prince @c @u0c0 @c @u @w0c0 @ wc Albert (PA, Black et al., 1996; Lee et al., 1997; u c s (2) Blanken et al., 1997), are analyzed in Section 3 to @t @x @x @x @z @z support this new formulation. BR is a case dominated where primes denote departures from the mean, and by highly convective motions, while PA data suggests overbars denote time averaging. A more rigorous the existence of a negative vertical velocity at night, treatment of the averaging procedure can be found indicative of drainage ¯ow, even though the site was in Finnigan (1985). The air is assumed to be incom- thought to be suf®ciently ¯at and uniform to meet pressible so that micrometeorological ¯ux observation criteria. Finally, @u @w w Section 4 presents a discussion of several important ' r (3) @x @z z mechanisms responsible for the non-zero mean r vertical velocity and explores their implications for where wr is the mean vertical velocity at the height of quantifying NEE of CO2 between the atmosphere and ¯ux observation (zr). Assuming no divergence of the forest vegetation. horizontal eddy ¯ux (term 2 on LHS of Eq. (2) (0) X. Lee / Agricultural and Forest Meteorology 91 (1998) 39±49 41 and no horizontal advection (term 30) and using observation over tall vegetation will be much more Eq. (3), the integration of Eq. (2) with respect to z pronounced than over short vegetation. yields Moncrieff et al. (1996) and Sun et al. (1997) treat Z zr the problem in a slightly different manner. They 0 0 NEE sdz w c jz0 consider the term @ uc=@x instead of u @c=@x as Z0 (4) in this study to be the advection term. It is believed that zr @c 0 0 the advection effect is con®ned to small areas down- dz w c r wr cr hci 0 @t wind of the transitional boundary of the surface source and can be minimized by choosing the site following where subscript r denotes a quantity at height zr and hci is the averaged concentration between the ground the usual fetch and footprint criteria. However, the and this height. mass ¯ow component can be signi®cant even at an Eq. (4) illustrates that NEE should consist of three extensive and uniform site. An example is given in components, storage below the height of observation Section 4 that compares the relative importance of the (term 1 on RHS), eddy ¯ux (term 2) and a mass ¯ow advection and mass ¯ow terms for the HAPEX- component arising from the horizontal ¯ow diver- MOBILHY forest. gence/convergence or a non-zero mean vertical velo- city (term 3). The last term, somewhat analogous to a 3. Energy imbalance bulk heat de®cit term due to nocturnal subsidence (Carlson and Stull, 1986), can be understood by con- 3.1. Correcting mean vertical velocity of sonic sidering a volume of air over the EC footprint as anemometers shown in Fig. 1. Suppose that the scalar concentration It is not appropriate to directly use the mean vertical decreases with increasing height, typical for CO2 in velocity measured by 3D sonic anemometers in eval- forests at night, and that wr is negative. To satisfy the uating the mass ¯ow component because the signal continuity requirement, the negative wr must be balanced by a horizontal ¯ow divergence. Air entering level is very low and can be contaminated by even a slight zero offset in the electronics, the sensor tilt the volume from above carries less CO2 than air moving out from the sides by divergence, causing a relative to the terrain surface and the aerodynamic shadow of the sensor structure.