Fluxes of Fine Particles Over a Semi-Arid Pine Forest: Possible Effects of a Complex Terrain

Abstract

Semi-arid forests are of growing importance due to expected ecosystem transformations following climatic changes. Dry deposition of atmospheric aerosols was measured for the first time in such an ecosystem, the Yatir forest in southern Israel. Size-segregated flux measurements for particles ranging between 0.25 μm and 0.65 μm were taken with an optical particle counter (OPC) using eddy covariance methodology. The averaged deposition velocity (V_d ) at this site was 3.8 ± 4.5 mm s⁻¹ for 0.25–0.28 μm particles, which is in agreement with deposition velocities measured in mid and northern latitude coniferous forests, and is most heavily influenced by the atmospheric stability and turbulence conditions, and to a lesser degree by the particle size. Both downward and upward fluxes were observed. Upward fluxes were not associated with a local particle source. The flux direction correlated strongly with wind direction, suggesting topographical effects. We hypothesize that a complex terrain and a patchy fetch affected the expected dependence of V_d on particle size and caused the observed upward fluxes of particles. The effect of topography on the deposition velocity grows greater as particle size increases, as has been shown in modeling and laboratory studies but had not been demonstrated yet in field studies. This hypothesis is consistent with the observed relationship between V_d and the friction velocity, the topography in the area of the flux tower, and the observed correlation of flux direction with wind direction.

[Supplementary materials are available for this article. Go to the publisher's online edition of Aerosol Science and Technology to view the free supplementary files.]

Copyright 2013 American Association for Aerosol Research

1. INTRODUCTION

Aerosol exchange between the atmosphere and forests affects both ecosystem nutrient loading (Bruijnzeel 1991) and the atmosphere's aerosol budget, which influence both air quality and climate (Fowler et al. 2009; Isaksen et al. 2009). Exchange processes between forests and the atmosphere include both particle emission and deposition. Primary bioaerosols, (e.g., pollen, fungi) and secondary organic aerosols can originate from within a forest, while aerosol such as soot, sulfates, and mineral dust are transported from local, regional, and global sources and deposited to a forest. Particles deposit in ecosystems through both wet and dry deposition. Although aerosol wet deposition dominates globally, water is a severe limiting factor in arid and semi-arid ecosystems meaning that dry deposition is an important and likely dominant surface exchange mechanism. To accurately model aerosol budgets, it is necessary to understand particle exchange processes and thus particle fluxes and deposition.

One metric for the deposition rate is the deposition velocity (V_d ), which is typically considered to be a function of particle diameter (d_p ), and to vary with vegetation type and height. Particle flux measurements are the most direct way to measure V_d . While there is some information on dry deposition of particles in the accumulation mode for temperate, boreal, and tropical forests (Rannik et al. 2003; Petroff et al. 2008; Vong et al. 2010), there is a lack of measurements and empirical information from semi-arid environments, despite their large global extent and the projections for their extension due to climate change (Dai 2011). Moreover, a recent model study suggests that by the year 2040, extended droughts and heavy photochemical smog episodes will prevail as a consequence of higher average summer temperatures in temperate regions (Sitch et al. 2007). These stress conditions may result in changes of forest ecosystems and subsequently may change the local and global atmosphere-biosphere interaction and energy balance in these areas. It is also expected that as droughts become more frequent, the contribution of dry deposition to the total atmospheric deposited mass will increase (Erisman et al. 1997)

This article presents measurements of dry deposition of accumulation mode aerosols over a planted pine forest in southern Israel. This 30–40 years old forest is located in a semi-arid climate zone and experiences low annual precipitation during a brief winter season and high summer temperatures, as well as additional stressors including anthropogenic pollution and frequent dust storms. Shifts in nutrient inputs through particle deposition may occur as a result of changes in these stressors (Lovett 1994; Magill et al. 2004). Despite the local extreme conditions, the nitrogen use efficiency and carbon sequestration rates in the Yatir forest are high and comparable to boreal forests, but occur mostly during the short winter time, December to March (Rotenberg and Yakir 2010). Semi-arid zones are particularly sensitive to erratic short-term variations in ecosystem conditions and long-term climatic changes (Challinor et al. 2007; IPCC 2007). Dry deposition measurements in this site therefore provide insights into deposition mechanisms relevant to other pine forests, particularly those that may shift to semi-arid ecosystems as a result of global climate change.

2. SITE DESCRIPTION

Flux measurements were conducted at the Yatir forest research station (31°20′42.62″ N, 35° 2′59.70″E) for a 45-day-period between 6 October and 28 November 2010. Aleppo pine trees dominate the forest. The research station and tower are located at the center of the Yatir forest, 630 m above sea level, and have been used for forestry research for over a decade, including measurements of ecosystem functioning under dry conditions and eddy covariance flux measurements of CO₂ and H₂O for the FluxNet project (Grunzweig et al. 2003; Rebmann et al. 2005; Rotenberg and Yakir 2010). The forest is located in a semi-arid area in which the dry season is usually between May and October. During 2010 specifically, the dry season was even longer, since there was no recorded rainfall above 5 mm between March and December. To avoid the dry and hot hours and to minimize water loss, the forest has assimilated its activity (Rotenberg and Yakir 2010). During October, leaf transpiration and flux measurements show that the photosynthetic activity and CO₂ intake starts at sunrise (6–7 AM), peaks between 8 AM and 9 AM and this activity is then reduced to a third of its peak by mid-day. It remains at this minimal level of CO₂ intake until sunset (5–6 PM) (Maseyk et al. 2008; Klein 2012). The forest peak activity (in terms of carbon flux) during October is not more than 10% of the activity during February in which the forest is in its maximum annual activity.

The sonic height is 18.7 m. The average tree height around the tower is 10.5 m and the stem density is 160 ± 40 trees per acre. The forest canopy is open and the land cover is ∼60%. The sky is always visible through the canopy and the leaf area index is 1–1.5 (Sprintsin et al. 2009). The vegetation in the area surrounding the research tower is patchy. This is partially the result of uneven topography and tree clearing south of the tower causing inhomogeneous tree density. Three cultivated vineyards are located 200–300 meters southeast of the station. An orthophoto picture of the study site is shown in Figure S1. Yatir forest is located about 20 km from three urban and industrial centers—Hebron, Arad, and Beer-Sheva—and 55 km from the Ashkelon coal power plant.

3. METHODS

Eddy covariance fluxes of aerosols were calculated as the covariance between the measured vertical wind velocities and particle concentrations (Foken 2008). Details of the instrumentation, sampling, data processing, spectral, and error analysis are given in the online supplementary materials.

FIG. 1 The diurnal cycle (30 min averaged bins) of the particle size distribution for a week during autumn days measured by the GRIMM OPC at the Yatir forest. The color scale depicts log (dN/dlogD_p) for particle concentrations in particles L ⁻¹. (Color figure available online.)

TABLE 1 Average values of critical parameters for day hours (7:00–15:00) at the Yatir station. The bracketed values are the standard deviation from the mean, or the interquartile range (IQR) of the median. A total of 901 half-hour measurements were collected during the campaign

Size (nm)	250–280	280–300	300–350	350–400	400–450	450–500	500–580	580–650
% of data points of Vd > 0	44	46	45	41	38	37	31	23
Mean and std of Vd > 0 (mm s^−1)	3.8 (4.5)	4.8 (5.8)	5.7 (7.1)	7.0 (8.5)	6.8 (8.2)	6.4 (7.3)	5.2 (5.6)	5.1 (5.1)
Mean and std of Vd < 0 (mm s^−1)	−2.1 (2.4)	−2.9 (3.1)	−3.1 (3.1)	−4.2 (4.8)	−6.2 (7.6)	−7.9 (9.4)	−8.3 (10.9)	−11.5 (16.0)
Median concentration (cm ^−3)	48	25	21	10	3	1	1	0.7
Median u * and IQR (m s^−1)	0.53 (0.30)
Median and IQR of wind speed (m s^−1)					−51 (94)
Median and IQR of L (m)					2.4 (1.5)
Mean sensible heat and std (Wm^−2)					281 (133)
T° (K)					301 (4)

4. RESULTS AND DISCUSSION

4.1. Particle Size Distributions

Figure 1 shows the diurnal particle number size distribution pattern over a week of measurements. The 250–450 nm particles (5 smallest size bins) dominated the total aerosol number concentration. The diurnal cycle of particle concentration was consistent throughout the measurement period. The number concentrations in all size bins were highest during the night, decreasing to a minimum at noon. The similar concentration trends between all particle sizes is observed also in the time series analysis of the entire measurement period, suggesting that the concentration changes of all particle sizes are dictated by changes in the boundary layer height. It appears that there are no additional local sources of either small or large particles, which would lead to varying concentration ratios between the various size bins over the diurnal cycle.

FIG. 2 Diurnal cycle of the half hour mean values of (a) friction velocity and (b) Obukhov length averaged over the full autumn day's data set. (Color figure available online.)

4.2. Diurnal Cycles and Average Conditions of the Meteorological Parameters

The diurnal friction velocity (Figure 2a) and boundary layer stability described by the Obukhov length, L (Figure 2b), show typical unstable conditions during the daytime with slightly stable conditions developing during the evening and night. The shift from unstable to stable conditions usually occurred in the late afternoon.

The average meteorological parameters and size resolved particle concentrations and deposition velocities (V_d ) in the Yatir forest are presented in Table 1. The data are constrained to the day time (7:00–15:00, local time [GMT+2] in which turbulent conditions dominated [u _* >0.3 m s⁻¹ and L < 0]). We calculated the mean V_d separately for the downward and upward fluxes as described by Nemitz et al. (2002) to separate apparent emission (upward flux) and deposition processes. Less than 50% of the daytime data points showed positive V_d (downward flux) for all size bins.

The fraction of downward (deposition) flux data points during the day hours does not change significantly between the first and the sixth size bin (44% and 37%, respectively) and the deposition velocity values are on the same order of magnitude across these bins (3.8–6.4 mm s⁻¹). However, the mean V_d for the apparent upward fluxes increases significantly with particle size also for these six size bins. In this analysis we do not calculate V_d for particles larger than 650 nm because of the larger errors caused by higher particle losses in the sampling tube (>13%) and the increased statistical error due to counting statistics.

4.3. Dependence of Deposition Velocity on Wind Direction

The size-segregated deposition velocity shows a strong dependence on the wind direction (Figure 3). For particles with d_p < 500 nm, deposition fluxes dominate for wind directions between 0° and 225° (N to SE), with the strongest deposition occurring between 30–110°. For wind directions between 225° (SW) and 360° (NW), we observed mostly upward fluxes. The fraction of upward flux measurements increased with particle size and for particles larger than 500 nm is observed for winds from all directions. Its magnitude, however, is more pronounced for the wind sector 225–300°. Thus, in further analysis, we calculated V_d after filtering out measurements from wind directions between 225° and 360° with the predominantly upward fluxes. 21% of the data points fell into this westerly wind direction criterion. A detailed discussion on the dependence of the flux on wind direction and on vertical rotation angle (tilt) angle is provided in the online supplementary materials.

4.4. Diurnal Cycle of the Deposition Velocity—Upward Fluxes

Following flux corrections (section S2, in the online supplementary materials) and filtering of data to include only daytime hours and exclude wind directions for which upward flux dominates (225–360°), at least 50% of the remaining data points still show an upward flux (apparent emission). The diurnal pattern of particle deposition and the apparent emission (Figure 4) correlate with the behavior of the friction velocity, with a mid-day maximum (compare with Figure 2). The magnitude and relative frequency of upward fluxes increase with particle size (Figure 4, Table 1). We note that the separation between positive and negative fluxes may cause a systematic error in our estimates of deposition velocity, potentially leading to an overestimate of the deposition velocity. However, this separation is needed in order to understand the deposition and emission processes, since by this separation, different processes may be revealed and analyzed allowing for a detailed understanding of parameters that control dry deposition (Pryor et al. 2008a,b).

FIG. 3 Dependence of daytime deposition velocity on wind direction (15° sections) at the Yatir station. Data points and bars are the mean deposition velocities and standard deviations per 15° section over the full autumn days data set. The horizontal black (green) line marks zero deposition velocity. (Color figure available online.)

FIG. 4 Diurnal cycle of the half hour averaged deposition velocity during autumn days. Data points are displayed for downward fluxes (V_d > 0) by a solid (blue) line, upward fluxes (V_d < 0) by a dotted (red) line, and their sum by a dashed line (green) and exclude the westerly wind directions (225°–360°). Note the change in Y-scale between the upper and lower panels. (Color figure available online.)

Bi-directional particle fluxes over forests have been observed also in previous studies (Gallagher et al. 1997; Buzorius et al. 1998; Pryor et al. 2008a, 2011). However, these studies focused on smaller particles than the 250–650 nm size range that were studied here. The fraction of data points showing upward flux in this study is higher than what was found for particles <150 nm in coniferous forests in Denmark and Finland (Gronholm et al. 2007; Pryor et al. 2008a), yet they are comparable to the predominant upward fluxes observed for particles <450 nm in a mixed coniferous and hardwood forest in Canada (Gordon et al. 2011). While two of these studies (Gronholm et al. 2007; Gordon et al. 2011) did not observe size dependence of the upward flux, Pryor et al. (2008a) noted that smaller particles (10 nm) experience more upward fluxes than larger particles (up to 110 nm). The observations described herein show a different trend, namely that smaller particles have less upward flux events than larger particles, although the two studies measure different particle sizes.

It was shown that stochastic effects in the data could result in upward fluxes (Gaman et al. 2004). In order to determine whether it reflects physical processes within the forest, we evaluated the reliability of the observed upward flux. First, we examine the number of points for which the absolute value of the flux (F) is larger than its error (σ_F ).

We evaluated the uncertainty in the calculated flux (σ _F ) by two independent methods: (1) following error propagation on the flux and (2) estimating the error as the variance of the tail of the correlation function, between the fluctuations of the (rotated) vertical wind velocity and those of the particle concentration, for times much longer than the integral correlation time (Wienhold et al. 1995). The second method showed larger uncertainties and we hence used the second approach as a more conservative estimate.

We found that at least 70% of the upward flux data points of the first eight size bins (250–650 nm) differed from zero by ±1σ_F , and about 50% differed by ±2σ_F . This suggests that these data points represent statistically reliable upward fluxes.

It was previously suggested that a possible explanation for observed upward fluxes is a high angle of attack (i.e., >20°) induced by shading of the anemometer by the tower equipment, which may lead to inaccurate determination of the flux (Nakai et al. 2006). While the angle of attack at the Yatir tower does not exceed the recommended limit of 20° (Figure S3, in the online supplementary materials), it displays a greater scattering compared to results from Hyytiala and other pine forests (Pryor et al. 2008a). No characteristic angle of attack is observed for upward or downward fluxes though, implying that the upward flux is not a result of shading effects. The large variation in angle of attack at the Yatir tower may result from the high surface roughness and relatively low mast height that have been shown to cause similar scattering at other sites (Gash and Dolman 2003).

As noted, the upward flux, even following the filtering of a specific wind sector, is observed from all wind directions. Upward fluxes observed in the absence of an obvious surface based particle source have been attributed to local emissions (Buzorius et al. 2001), sources of particles within or close to the canopy top (Buzorius et al. 1998), particle re-suspension (Hicks et al. 1989b), and entrainment of clean, particle-depleted air from aloft during the growth of the mixing layer that causes either changes in particle growth or evaporation due to chemical partitioning (Nilsson et al. 2001), or dilution of particles stored within the canopy (Pryor et al. 2008a; Gordon et al. 2011). In the Yatir forest, local anthropogenic sources are unlikely responsible for the observed upward flux: the site is isolated and far from sources such as roads with high traffic or industrial activity.

The previous observations of upward flux have been associated with smaller (<120 nm) particles (Gallagher et al. 1997; Nilsson et al. 2001; Nemitz and Sutton 2004) and have been attributed to gas-to-particle conversion and aerosol evaporation induced by mixing of air masses, heat, and moisture fluxes. However, growth and evaporation have a more pronounced effect on the radius of smaller particles (Nemitz and Sutton 2004; Gordon et al. 2011); neither the diameters monitored in Yatir are consistent with sizes relevant to nucleation events and growth, nor is the trend of greater upward flux with particle size consistent with such processes. Moreover, the average relative humidity and the Bowen ratio measured at the canopy height during the campaign at mid-day were ∼25% and higher than 10, respectively. The sensible heat increases from sunrise to its maximal value (380 Wm⁻²) at noon and decreases back to its minimal value by 4–5 PM. The latent heat is uncorrelated with the diurnal aerosol flux as it increases after sunrise from ∼0 Wm⁻² to ∼40 Wm⁻² at 8–9 AM and remains almost constant until 1 PM when it decreases back to ∼0 Wm⁻² before sunset (4–5 PM). Considering the water stress, the diurnal cycle of CO₂ uptake and the extremely high Bowen ratio during mid-day, we maintain that the aerosol at this site is dry and far from conditions that will lead to deliquescence or to sizeable hygroscopic size growth. Moreover, the diurnal particle number concentrations for all size bins show identical patterns of high and low concentrations throughout the day. Thus, we eliminate in-canopy chemistry or particle growth as the source of these observations. Entrainment of clean, particle-depleted air into the boundary layer (Pryor et al. 2008a) is also an unlikely cause of the observed upward particle fluxes because it was previously shown that the free tropospheric air above the boundary layer in our region contains small anthropogenic particles from Europe and from the Israeli coastal plain (Andreae et al. 2002). Further, the upward fluxes are independent of boundary layer stratification and observed throughout the day.

Thus, we have eliminated large random errors, high angles of attack, local sources, in-canopy growth, and entrainment of clean air as the source of the observed upward fluxes. We instead attribute the upward fluxes observed in the Yatir forest to topographic effects.

4.5. Deposition Velocity Dependencies: Friction Velocity, Size, and Stability

The daytime hours are dominated by turbulent conditions (Figure 2b). Under these conditions, dry deposition is controlled by the settling velocity, aerodynamic resistance, turbulent diffusion of the particles (Brownian motion), and impaction and interception of the particles within the forest canopy (Slinn 1982). The contribution of Brownian motion and settling velocity of accumulation mode particles to the total dry deposition is negligible (Seinfeld and Pandis 2006). The contribution of aerodynamic resistance to particle deposition at Yatir was calculated following Gallagher et al. (1997), who related the surface deposition velocity (V_ds ) to the measured deposition velocity (V_d ) ratio by the contribution of the aerodynamic resistance, r_a , above the canopy,

where the aerodynamic resistance r_a between the measurement height and the surface is defined as (Hicks et al. 1989a):

where z is the measurement height, d is the displacement height, z ₀ is the roughness length, L is the Obukhov length, and Ψ _M (z/L) is the empirical stability correction function. The aerodynamic resistance may also be written as (Gallagher et al. 1997):

where the vertical wind velocity U and the friction velocity u _* are calculated directly from the anemometric measurements (see the online supplementary materials). Under turbulent conditions, the aerodynamic resistance accounts for <6% of the total deposition velocity and therefore V_ds ≅V_d . The near equality between surface and total deposition velocity indicates that the deposition in this forest is dominated by inertial projection of the particles through the viscous air layer of the pine needles and the consequent impaction and interception.

Current theoretical frameworks for particle fluxes predict that particle V_d and friction velocity should depend on particle size (Pryor et al. 2008b). We analyzed the inertial processes (impaction and interception) within the canopy that contribute to particle deposition using the observed dependence of V_d on friction velocity. For all particle sizes, the deposition velocity increases almost linearly with increased turbulence, up to a friction velocity of u _* = 0.8 m s⁻¹ (Figure 5a). The linear dependence of deposition velocity on friction velocity is stronger for the smaller particles than the larger particles and for d_p > 400 nm, the slope of V_d /u _* decreases as particle diameter increases (Figures 5b and S3). This stems from the size dependent behavior of V_d , which increased with particle size for the same smaller size bins, but decreased beyond d_p > 400 nm as the particle diameter increases (Figure 5c).

FIG. 5 (a) V_d as a function of the friction velocity for size bin 250–280 nm. The open (red) circles represent the mean and standard deviation values (binned in 0.2 increments). V_d/ u _* (b) and V_d (c) as a function of particle diameter between 250 nm and 650 nm during autumn days. Data points are averaged for day hours (07:00–15:00), V_d > 0 and exclude the westerly wind directions (225–360°). (Color figure available online.)

Although the deposition velocity clearly depends on the friction velocity, the dependence deviates from that expected if deposition is governed only by impaction. In cases where impaction is the only mechanism responsible for deposition to the forest, V_d should be a power function of u _* (Slinn 1982; van Aalst 1986; Nemitz 1998; Zhang et al. 2001). Other studies in forests with comparable V_d /u _* values also displayed a linear or close to linear dependence of V_d on u _* for particles in the accumulation mode (Lamaud et al. 1994; Gallagher et al. 1997; Wyers and Duyzer 1997; Nemitz et al. 2002; Gronholm et al. 2007). These studies suggested that the linear dependence results from a combination of deposition mechanisms other than impaction (Wyers and Duyzer 1997), a burst of eddy turbulence (Feng 2008), or the presence of very rough surfaces within the fetch (Nho-Kim et al. 2004). Surface roughness is an expression of surface cover, vegetation type, and surface topography (Seinfeld and Pandis 2006). The fetch area surrounding the tower in Yatir forest has open trunk spaces as a result of years of drought and poor forest management (Figure S1). The prolonged droughts led to extensive tree mortality, resulting in an inhomogeneous and “patchy” land cover around the tower. Further, the topography within the fetch is uneven with inclines to the West and South West of the tower with slopes of 4% and 8% at 270° and 225°, respectively, from the tower. At a distance of 100 m from the tower, the gradient in the opposite wind sector between 45° and 90° is 0, i.e., the area is flatter. This combination of land features produces a particularly high surface roughness, which we hypothesize is the source of not only the broad dispersion of angle of attack (Section 4.4), but also of the linear dependence of V_d on u _*.

The observed dampened dependence of V_d on particle diameter (Figures 5b and c) is noted also when analyzing the dependence of the deposition velocity normalized to the friction velocity that theoretical frameworks for particle fluxes predict should depend on atmospheric stability (Wesely et al. 1985). Based on this framework, Gallagher et al. (1997) described the size and stability-dependence of V_d//u _* for particles <500 nm by the following equation:

where k ₁ = 0.0135. We investigated the relative importance of atmospheric stability (L) and particle size in determining V_d for the Yatir dataset. The deposition-dominated V_d values were normalized to the friction velocity (30 min means), and binned with respect to the Obukhov length (Figure 6).

FIG. 6 Mean values of half hourly V_d/ u _* vs. L ⁻¹ for size bins 250–280 nm, 350–400 nm, and 580–650 nm (Binned in 0.01 increments). Data points are for day hours, V_d > 0 and exclude the westerly wind directions (225–360°). The solid (red) lines show the modified parameterization equation (Equation (5)) for these three size bins following Gallagher et al. (1997). The equation coefficient (k₁) for each size bin is 0.005 as determined by the match of the parameterization (Equation (5)) to the Yatir data for the smallest size bin 250–280 nm. (Color figure available online.)

The transition between stable and unstable conditions is clear with V_d/ u _* decreasing sharply as stability is approached and has a near constant value at stable conditions. Although the standard deviation is high (from 20% to >100%), the trend of increasing normalized deposition velocity with increasing instability is apparent, though little difference is seen for the different particle sizes. We used the Yatir data to fit an empirical equation with the same form as Gallagher et al. (1997) and found that the best fit for the smallest particle size (250–280 nm) is matched as k ₁, reduced from 0.0135 to 0.005. However, applying this fit parameter to larger particle sizes overestimates the observed V_d/ u _* dependencies on L (Figure 6).

Gallagher et al. (2002) extended the above parameterization to include the effect of surface morphology and roughness on measured deposition velocities and showed that for near neutral conditions, k ₁ is empirically correlated with the surface roughness: k ₁ = 0.001222 log(Z ₀) + 0.003906 for small particles (100–200 nm), where Z ₀ is the surface roughness length (m). The surface roughness of Yatir forest is about 1.3 (Yakir and Rotenberg 2007) and the expected k ₁ following this correlation, which is independent of the particle diameter 0.004, consistent with the derived k ₁ (0.005) for the smaller particle sizes at Yatir.

Although the observations from Yatir forest do not show a clear size-dependence as the large uncertainties cause substantial overlap across different particle sizes, the observations suggest that atmospheric stability is the most important parameter in determining particle deposition rates at this site. Previous studies also showed that quantifying the influence of stability with a high degree of statistical certainty is challenging (Pryor et al. 2008a). While our results suggest that the dependence on stability and particle size is significantly weaker than what was found by Gallagher et al. (1997), our measurements match the extended parameterization connecting between roughness length and deposition velocity for small particles (Gallagher et al. 2002).

4.6. The Effect of Topography on Deposition Velocity Measurements

The above discussion demonstrates that particle deposition depends on friction velocity, stability, and particle diameter, but suggests that roughness affects the impaction mechanism controlling particle fluxes over Yatir.

The measurement and interpretation of V_d from eddy covariance measurements relies on assumptions of stationarity and planar homogeneous flow that are appropriate only for flat-terrain conditions. However, these assumptions seem to break down in sites with complex terrain (Belcher et al. 2012), such as the Yatir forest, likely resulting in micrometeorological behavior not captured by the typical roughness length schemes, such as the multiple-resistance approach to dry deposition used in most models (Hicks 2008). For example, hilly terrain may result in substantial changes in the spatial friction velocity compared to a “flat-world” presentation, i.e., the calculated friction velocity does not represent well the “real” u _* in an hilly environment (Nho-Kim et al. 2004).

Modeling studies show that a hilly terrain may cause V_d to be underestimated and that the topography-induced spatial differences in V_d are highly sensitive to the particle size (d_p ). As the particle diameter increases, the underestimation of V_d values increases, leading to a different particle size-dependence than that of a “flat world,” as assumed in current models. The variability is most extreme for particles with deposition velocity that is regulated by inertial impaction, i.e., between the 100 nm and 1 μm size range. Topography with the same height scale as the forest canopy, a situation pertinent to the Yatir landscape where the relief is on the scale of 10's of meters and the canopy height is about 10 m, displays the most interesting dynamical effects of how topography alters the spatial structure of V_d and particle fluxes (Katul and Poggi 2010, 2011).

The theoretical size-related underestimation in V_d values could explain the Yatir forest observations. The weak dependence of V_d on particle size and the breakdown of the near linear dependence for the larger particle sizes, the dampened dependence of V_d on u _* for the larger particles, and the lack of a clear particle size dependence of V_d/ u _* on the Obukhov length at near neutral conditions (|L > 20|) are all consistent with local topography affecting the derived deposition velocity values. Since the difference between the terrain evaluated and “flat world deposition” increases with particle size (Katul and Poggi 2010, 2011), V_d of the smallest size bins is the most accurate in this study. Quantifying the combined effect of hill geometry and canopy morphology on flow variables is complex and beyond the scope of this work.

4.7. Topography and Upward Flux

We have shown that upward fluxes were prevalent from all wind directions but especially dominated the wind sector 225–300°. The increasing fraction of upward fluxes with increase in particle size (Table 1) is best exemplified within this wind sector in which the phenomena is amplified and for which the upward flux fraction increases from 52% for particles within the diameter range 250–280 nm to 100% for 580–650 nm particles. Figure S5 shows the surface cover of the fetch in a 100 m radius around the tower and shows that large open patches are common in all directions. The surface inclines however are greater in this upward flux sector, with slopes to the West and South West of the tower of 4% and 8% at 270° and 225°, respectively. The steepness of the gradients arises from an erosion channel (with low surface coverage) that runs from the tower within this South-West direction for a distance of more than 100 m. While patchiness and surface inhomogeneities are prevalent in all directions, the contribution of terrain relief to the roughness in the wind direction of the upward fluxes is greater than that in the other wind directions where the gradient is nearly flat. This strongly supports that topography enhances the upward fluxes.

Modeling studies of flow fields over forested topography (Finnigan and Belcher 2004; Ross and Vosper 2005) and experimental simulations of fields over hilly terrain (Poggi and Katul 2007) demonstrate that a recirculation region exists close to the ground on the lee side of the hilltop; see Figure 6 in Poggi and Katul (2007). This recirculation region traps scalars (in our case particle concentration), promotes strong horizontal gradients of scalar quantities, and leads to these scalars being ejected from the canopy on the lee side of the hill (Poggi and Katul 2007, 2010, 2011). Based on this explanation, we suggest that the trapping of particles in a recirculation region close to the Yatir tower leads to a vertical gradient in particle concentration. Vertical concentration gradients and intermittent changes in the direction of vertical velocity may lead to the prominent upward fluxes observed at the Yatir site (Katul and Poggi 2010). The size dependence of the upward fluxes in Yatir, increasing both in frequency and magnitude with increase in particle diameter (d_p ) supports the upward fluxes being topographically induced as the model shows a similar increasing effect of topography with increasing particle size.

TABLE 2 Selected averaged deposition velocity values of particles in the accumulation mode over coniferous forests. Upward fluxes were (a) included, (b) not included, or (c) unknown (or not applicable) in the averaging of deposition velocities

Canopy (height in m)	dp (μm)	u * (m·s^−1)	Vd (mm·s^−1)	Reference	LAI
Pine (10.5)	0.25–0.28	0.65	3.8 ± 4.5 (b)	This Study	1–1.5
Pine (9)	0.5–1	(–)	4.3 (c)	(Lorenz and Murphy 1989)	6–12
Pine (15)	0.05–0.5	0.60	7.0 (c)	(Lamaud et al. 1994)	3
Pine (8)	0.44–0.54	(–)	6.1 (a)	(Vong et al. 2010)	5.1
Spruce (0.45)	0.82	0.45	5.0 (c)	(Ould-Dada et al. 2002)	3.2
Spruce (30)	0.5–1	(–)	15 ± 40 (c)	(Gravenhorst 1989)	(–)
Spruce (17)	0.3–0.5	0.15–0.75	0.3–16.6 (a)	(Gallagher et al. 1997)	10.3
Hardwood and coniferous (–)	0.23–0.40	0.2–0.80	0. 5–1 (a)	(Gordon et al. 2011)	(–)

4.8. Comparison of Derived Deposition Parameters to Other Sites

The deposition velocities observed at Yatir forest can be compared with other measurement sites, and are found to be similar to values measured in other forests (Table 2). The comparison further is limited to 0.25–0.28 μm since this size bin is the least affected by topography; it is in the middle range of most of the previously published results. The calculated averaged deposition velocity in Yatir is 3.8 ± 4.5 mm s⁻¹. This V_d value represents the maximal estimate of the deposition due to the separation of upward and downward fluxes prior to calculation. Since the observations were taken only during one autumn season, they may not be representative of the entire year. However, the average deposition velocity is comparable to previous measurements over coniferous forests in Europe (Gravenhorst 1989; Lamaud et al. 1994; Gallagher et al. 1997; Ould-Dada et al. 2002; Nemitz and Sutton 2004) and the United States and Canada (Lorenz and Murphy 1989; Vong et al. 2010; Gordon et al. 2011; Table 2).

Recently, deposition models have been extended to resolve detailed morphological properties of the vegetation, such as the leaf size, shape, and two-side leaf area index as well as the height of the vegetation canopy and the profile of the canopy crown (Petroff et al. 2009; Petroff and Zhang 2010), resulting in a greater sensitivity to changes of the land cover compared to earlier models (Slinn 1982; Zhang et al. 2001). The semi-arid Yatir forest is expected to differ from the European and American coniferous forests in terms of tree stand density and height, open-trunk space, and leaf area index (LAI data are shown in the Table 2). However, although these highly resolved models display the strong impact of canopy morphology on deposition velocity, it is evident from Table 2 that despite the low stand density (350 trees/ha) and low LAI (1–1.5) in Yatir (Sprintsin et al. 2009), the average deposition velocity values do not differ significantly from those of other forests in less arid regions.

5. CONCLUSIONS

Yatir forest is the southern-most pine forest in the Northern Hemisphere. It is located in a semi-arid environment and exposed to harsh, dry, and hot conditions with frequent droughts. Although it is not as dense or homogenous as other pine or spruce forests in boreal or Mediterranean areas, the deposition velocity of 250–280 nm aerosols during the autumn daytime hours is 3.8 ± 4.5 mm s⁻¹, which is very similar to previous measurements in coniferous forests in more temperate climates.

The observed deposition velocities depend on the turbulence conditions and, to a lesser extent, on the particle size. The deposition velocity in Yatir forest displays a linear dependence of V_d on u _* rather than the expected power-dependence, and a size dependence that holds only for smaller particles. V_d /u _* has a clear dependence on turbulent conditions, yet its dependence on particle size and L⁻¹, as parameterized according to the Gallagher equation, is weak.

We hypothesize that the difference between the expected dependencies of deposition velocity on the micrometeorological conditions and particle size and those observed result from the hilly terrain and patchy, inhomogeneous fetch around the Yatir tower station.

Both upward and downward particle fluxes were measured in all size ranges (250–650 nm), although upward fluxes increased in both magnitude and frequency with particle size. In particular, fluxes influenced by wind directions from SW to NW (225–300°) sector were dominated by upward fluxes. This wind direction corresponded to an inclined terrain within which lies a deep erosion channel and large patchy openings in the forest. We suggest that the frequently observed upward fluxes result from advective spatial gradients and trapping of particles induced by the inhomogeneous vegetation cover and gentle relief, significantly affecting micrometeorology and particle flux observations at this site.

These observations are consistent with laboratory and modeling studies in which the effect of topography on deposition velocity is greater as particle size increases, most dominantly for particles with diameters of 100 nm–1 μm, for which the dominant removal mechanism is inertial impaction. Therefore, the particle size dependencies in Yatir forest for this particle size range deviate from those expected by “flat world” models.

This study is particularly relevant to the application of eddy covariance methodology for particle fluxes in complex terrains, and to dry deposition estimates for emerging semi-arid environments. Ecosystems that will be affected by climate changes will possibly entail nonuniform transitions and oscillations in local conditions, inevitably leading to some of the complexities of patchiness and inhomogeneous vegetation cover encountered at the Yatir forest. The need to further develop models for the exchange between the atmosphere and a natural landscapes containing vegetation inhomogeneities under changing climate and precipitation will be all the more pertinent.

Acknowledgments

This research was supported by a grant (#2008146) from the United States—Israel Binational Science Foundation (BSF), Jerusalem, Israel. Yinon Rudich acknowledges support by the Helen and Martin Kimmel Award for Innovative Investigation. The financial support of The Dr. Scholl Center for Water and Climate Research is gratefully acknowledged.