Agricultural and Forest Meteorology 152 (2012) 178–188
Contents lists available at SciVerse ScienceDirect
Agricultural and Forest Meteorology journal homepage: www.elsevier.com/locate/agrformet
Application of eddy covariance to determine ecosystem-scale carbon balance and evapotranspiration in an agroforestry system P.R. Ward a,b,∗ , S.F. Micin a,b , I.R.P. Fillery a,b a b
CSIRO Plant Industry, Private Bag No 5, Wembley, WA 6913, Australia CRC for Plant-based Management of Dryland Salinity, University of Western Australia, 35 Stirling Highway, Nedlands, WA 6009, Australia
a r t i c l e
i n f o
Article history: Received 29 June 2010 Received in revised form 22 September 2011 Accepted 22 September 2011 Keywords: Alley farming Carbon sequestration Dryland salinity Eddy correlation Eucalyptus Tree belt
a b s t r a c t The inclusion of belts of trees in the agricultural areas of south-western Australia is gaining popularity, through perceived benefits in water use, biodiversity and carbon sequestration. However, water use and carbon assimilation are difficult to quantify at the ecosystem scale. In this research, we investigate the application of eddy covariance in a ‘belt and alley’ system. Footprint modelling indicated that the proportion of the signal from the tree belts was similar to the proportion of tree belts in the ecosystem. Eddy covariance units were installed in two adjacent fields: one containing belts of four-year-old oil mallee Trees 10 m wide planted 60 m apart; and one containing agricultural crops. Tree belts had little impact on wind direction or its standard deviation for any measurement height, and stationarity, integral turbulence and energy balance closure were similar for the two fields. Persistent upward wind flows were observed for wind directions aligned with the tree belts, suggesting the possibility of advection. For our experimental site, removal of data for winds aligned with the tree belts had a negligible impact on fluxes of carbon dioxide and water. During summer and autumn, the field containing oil mallees used 27 mm more water, and assimilated 0.93 Mg CO2 /ha more than the field without trees. Both daytime and night time CO2 fluxes were greater in magnitude for the oil mallee field than the control field during summer. Water use by the trees was lower than other fields in the region where herbaceous perennials had been grown, and also lower than other estimates based on sap flow measurements. Further research will be necessary to determine the impact of spatial variability of water use on groundwater recharge at the catchment scale. Extra carbon storage associated with tree belts may increase their economic attractiveness to farmers in the region, but the impact of respiration after summer rainfall requires further investigation. © 2011 Elsevier B.V. All rights reserved.
1. Introduction South-western Australia experiences a Mediterranean-style climate, and rainfed cropping predominates in the areas receiving between 300 mm and 500 mm of annual rainfall. In recent years, environmental concerns over groundwater recharge and associated dryland salinity (McFarlane et al., 2004), and loss of biodiversity (Lefroy et al., 2004), have affected farming systems within the region. In particular, there has been a focus on increasing the proportion of perennial vegetation in the landscape. The impact of herbaceous perennials such as lucerne on the water balance has been extensively studied (as summarised by Ward et al., 2006), and attention is now turning to woody perennials. Competition for light, water and nutrients between tree belts and adjacent crops, and techniques to manage it, has been well
∗ Corresponding author at: CSIRO Plant Industry Private Bag No. 5, Wembley, WA 6913, Australia. Tel.: +61 8 9333 6616; fax: +61 8 9387 8991. E-mail address:
[email protected] (P.R. Ward). 0168-1923/$ – see front matter © 2011 Elsevier B.V. All rights reserved. doi:10.1016/j.agrformet.2011.09.016
researched in recent years (e.g. Ong et al., 2002). From a more hydrological perspective, the inclusion of belts of trees with the aim of controlling groundwater recharge in the traditional agricultural areas of southern Australia has been theoretically studied by Stirzaker et al. (1999, 2002). In particular, the inclusion of oil mallees (various Eucalyptus species) is gaining popularity, through their potential for bio-energy, activated charcoal, and solvent oil production (Wu et al., 2007). Wildy et al. (2004), Ellis et al. (2005) and Oliver et al. (2005) investigated the impact of belts of oil mallees on water balance and crop performance. Tree belts were found to be effective in reducing groundwater recharge and the subsequent threat of dryland salinity, but require specific site conditions to be satisfied for their inclusion in a farming system to be economically attractive (Lefroy et al., 2004; Stirzaker et al., 2002). In order to calculate belt or ecosystem water use from sap flow data from individual trees (e.g. Carter et al., 2004; Ellis et al., 2005; Wildy et al., 2004), an assumption must be made about the hydrological area ‘occupied’ by the tree. A common measurement used for this calculation is the projected crown area, but recent analysis by Crosbie et al. (2008) suggests that the water use by belts of trees might be
P.R. Ward et al. / Agricultural and Forest Meteorology 152 (2012) 178–188
overestimated by as much as 100% if projected crown area is used in up-scaling calculations. Another factor influencing the adoption of tree belts is their potential for carbon sequestration (Nair et al., 2009), and the inclusion of such systems under the Kyoto Protocol. Recent estimates of carbon sequestration for agroforestry systems in various ecosystems range from 0.3 Mg C/ha/year in the West African Sahel (Takimoto et al., 2008) to more than 15 Mg C/ha/year in Puerto Rico (Parrotta, 1999). Takimoto et al. (2009) and Haile et al. (2008) demonstrated that alley farming led to changes in soil C and N quantities, with the stable carbon in the soil derived mainly from the tree components. However, as discussed by Nair et al. (2009), quantification of above-ground, and more particularly below-ground biomass is difficult in agroforestry systems. Furthermore, the capacity for various Eucalyptus species in general, and oil mallee types in particular, to sequester carbon does not appear to have been scientifically evaluated and quantified under alley farming conditions. The technique of eddy covariance appears well-suited to quantify the carbon and water dynamics of alley farming systems, as it measures vertical fluxes of both carbon and water at the ecosystem scale. This technique has been used for many years to assess the carbon balance of many ecosystems, as reviewed by Baldocchi (2008). However, its application to alley farming systems can be difficult. The structured differences in canopy heights and roughness across the alley farming area can lead to complex air flow patterns, which could affect fluxes measured at a point (Mao et al., 2008; Sogachev et al., 2005a). Furthermore, advection associated with temperature differences between the two canopy types (Zhang et al., 2007) could result in consistent upward or downward air flows at the point of measurement, which will also impact on the fluxes measured at that point. In this research, we aimed to determine whether the eddy covariance technique could be applied to a ‘belt and alley’ ecosystem in order to quantify carbon dioxide (Net Ecosystem exchange, NEE) and water fluxes. We performed a footprint analysis based on the SCADIS model (Sogachev and Lloyd, 2004; Sogachev et al., 2005b) in a simulated landscape based on a field in south-western Australia, to determine the impact of tree belts on air flow patterns in the field. We then measured patterns in horizontal and vertical wind speed associated with wind direction in a field containing belts of oil mallee trees, to determine (a) whether the presence of parallel tree belts caused changes in horizontal wind direction, and (b) whether the structured vegetation pattern caused advection at a point between the tree belts. Finally, we used the eddy covariance technique to quantify the carbon capture, and water balance, of a field containing belts of trees, compared with an adjacent field with no tree belts. These results were used to estimate the effectiveness of tree belts in overcoming dryland salinity, and as a sink for carbon sequestration.
2. Materials and methods 2.1. Site details The site was located on a farm in south-western Australia at 32◦ 53 S, 117◦ 47 E, near Tincurrin, 200 km south-east of Perth. Two fields were selected on shallow gravelly duplex soils, with a loamy sand A horizon of 10–15 cm depth overlying a gravelly sandy clay. The first field (the ‘oil mallee’ field) was about 1000 m by 750 m, and contained nine belts of oil mallees (Eucalyptus polybractea) planted in 2003 in an E–W orientation, with a distance of approximately 66 m between belts (Fig. 1). Each belt consisted of four rows of trees, with a total belt width of approximately 7 m. The alleys between the tree belts were planted with legume pasture in
179
N
Oil mallees EC locatio location n Ma May y 06 – Oct 06 EC location Apr 07 – May 08 Field boundary 250 m
Fig. 1. Layout of the oil mallee field.
2006 (Ornithopus sativus) and 2007 (Trifolium subterraneum). The second field (the ‘control’ field, about 1 km east of the oil mallee field) of similar dimensions contained no trees until seedlings at the same spacing as the oil mallee field were planted in August 2007. The remainder of the area was planted to wheat (Triticum aestivum) in 2007, and legume pasture (O. sativus) was sown over the whole field in 2006. Rainfall during the trial was obtained from an automatic weather station located in the oil mallee field. Longterm average annual rainfall at Dudinin, 12 km east of the site, is 345 mm, of which 244 mm falls in the May–October period. 2.2. Eddy covariance measurements One eddy covariance unit was installed in each field. Each unit consisted of a R3-50 sonic anemometer (Gill Instruments, UK), and a LICOR LI-7500 (LiCor, USA) open-path infra-red gas analyser. Raw data was collected at a frequency of 20 Hz and stored in a purpose-built logger, and subsequently analysed in 1-h time periods with EdiRe 4.2 software (University of Edinburgh). Prior to flux calculations, coordinate rotation was performed using the Planar Fit method (Wilczak et al., 2001), with coefficients determined for each period of measurement and sensor location. Fluxes were corrected for density effects (Webb et al., 1980) and frequency response by comparison with standard modelled spectra using routines available in EdiRe 4.2. Stationarity was determined after planar fit coordinate rotation for horizontal wind speed u, vertical wind speed w, sonic temperature T, water vapour concentration q, and CO2 concentration C, and for the quantities uw, wT, wq and wC. Each 1-h time period was divided into six 10-min segments, and data was considered stationary if the variance or covariance was less than 30% (e.g. Foken and Wichura, 1996; Rebmann et al., 2005). Integral turbulence characteristics were also calculated for the wind components u and w under unstable atmospheric conditions (z/L < −0.2), and a deviation of less than 30% from theoretical values was again considered acceptable (Rebmann et al., 2005). Eddy covariance measurements were conducted for periods: May 25–July 3, 2006; July 13–August 29, 2006; September 5–October 19, 2006; and April 4, 2007 to May 7, 2008. These periods covered two ‘growing seasons’ (i.e. May to October) and one summer–autumn period. A plot of hourly values of net CO2 flux against u* (where u* is the friction velocity, and is an indicator of atmospheric mixing) identified that a threshold value of u* = 0.1 m/s was an appropriate minimum value for both fields, and data where u* was less than 0.1 m/s were discarded (Goulden et al., 1996). Following analysis of average wind direction and average vertical wind speed prior to axis rotation (W ) (Section 3), data for wind directions within 23◦ of east or west were also discarded. Gaps in the carbon and water flux data were filled by calculating the average monthly values for each hour of measurement, in a manner similar to the ‘Mean Diurnal Variation’ method described by Falge et al. (2001). We used a longer time period (30 days) for averaging than that used
180
P.R. Ward et al. / Agricultural and Forest Meteorology 152 (2012) 178–188
by Falge et al. (2001) (7–14 days) to ensure accurate average calculations. There was a gap of 31 days for data from the oil mallee field in August and September 2007, and this gap for both water and CO2 flux was filled by linear regression using data from the control paddock. Regression equations were developed for 14-day periods either side of the gap. Average tree height was approximately 2 m in November 2005, 3 m in October 2006, and 4 m in November 2007. Measurement height in the oil mallee field was 3.7 m until July 3, 2006, 5.4 m from July 3, 2006 until October 18, 2006, and 6.0 m for subsequent measurements. Measurement location in the field was changed in April 2007 as indicated in Fig. 1 to allow installation of a walk-up tower enabling higher measurements. Measurement height was 1.7 m in the control field.
2.3. Energy balance Net radiation (Rn) (REBS Q*7.1) and ground heat flux (G) (REBS HFT-3.1) were measured for the periods April 27, 2007–May 31, 2007, and September 27, 2007–October 23, 2007. These periods approximately correspond with periods of minimum evapotranspiration and maximum evapotranspiration, respectively. A correlation between Rn-G and solar radiation (Rs) measured at the automatic weather station was established for these periods: in units of w/m2 , R-G = 0.590Rs–36.8 (R2 = 0.96). This equation was used to estimate Rn-G for other periods of eddy covariance operation. Estimates of Rn-G were then compared with the sum of sensible energy (H) and latent energy (LE) measured with the eddy covariance units to determine the degree of energy balance closure provided by the eddy covariance technique for both fields.
2.4. Footprint calculations The footprint for measurements taken in the oil mallee field was analysed using SCADIS software as described by Sogachev and Lloyd (2004) and Sogachev et al. (2005b). This software uses numerical solutions for ensemble-averaged Navier–Stokes equations to calculate flux footprints for inhomogeneous vegetation. The model allows the user to specify vegetation height, leaf area index (LAI) and leaf distribution with height for various distances from the point of measurement. A repeating vegetation pattern of a 10 m belt of trees followed by a 60 m alley of short crop or pasture was set up for a distance of 500 m, with the point of measurement located 30 m from a tree row (that is, close to the mid-point between tree rows). This approximates the situation experienced in the oil mallee field for a north or south wind. Winds at 45◦ angles were simulated by changing the horizontal grid parameter from 10 m to 14 m, resulting in a tree belt of width 14 m and an alley width of 84 m. For all model calculations, the alley vegetation was given a height of 0.5 m and a LAI of 1.0, broadly in line with expectations for a growing wheat crop. The distribution of leaf area within the canopy was specified with the shape parameter ␣ (Markkanen et al., 2003), which was set at 3.0, representing a normal curve with maximum LAI at a height of half the vegetation height (0.25 m in this case). Tree height was set at either 2 m, 3 m, or 4 m, and measurement height was set at 3.7 m, 5.4 m or 6.0 m, respectively, reflecting experimental procedures. The shape parameter ␣ for the tree belts was set at 5, giving maximum LAI at 0.75 times the tree height. For all tree heights, calculations were performed for tree LAI of 1.0, 2.0 and 3.0, and for geostrophic wind velocities of 3 m/s and 10 m/s. (Note that footprints for geostrophic wind velocities less than 3 m/s could not be calculated due to numerical instabilities in the model.) Surface roughness for the soil was set at 0.01 m for all calculations.
2.5. Wind direction analysis Wind direction was stratified into 20◦ classes for both the oil mallee and control fields, for periods associated with each of the different measurement heights in the oil mallee field. The distribution of wind direction was compared between the two fields to determine the impact of the parallel tree belts on wind direction. The standard deviation of wind direction was also assessed in 20◦ classes for the oil mallee field for each measurement height.
2.6. Average hourly vertical wind speed (W ) analysis Because of the different vegetation characteristics within the oil mallee field, air temperature over each of the components could also be different, which could lead to consistent patterns of vertical air movement in the field. To identify if this was occurring, W was analysed as a function of horizontal wind direction for each measurement height. The sonic anemometer was not installed perfectly vertically, and so there was a sinusoidal artefact in the graph of W against wind direction. This artefact was removed prior to analysis using a function of the form: √ 2 √ 2 2 2 W cor = ((W / (U + V )) − (a sin(wind dir − b))) × (U + V ), (1) where W cor is the tilt-corrected vertical wind speed, U and V are the horizontal wind vectors (prior to planar fit axis rotation), a = tan Ф, the angle of tilt, and b is the direction of tilt plus 90◦ . The parame√ 2 2 ters a and b were estimated from a graph of W /( U + V ) against wind direction. We anticipate that under conditions of persistent vertical air movement due to differences in temperature between the vegetation types, W cor at the point of measurement in the centre of the alley should be biased towards either positive (upward) or negative values when the wind is from the east or west, blowing along the alleys. Biases in W cor for winds from the north or south would indicate direct influence of the tree belt on wind patterns.
2.7. Error analysis for eddy covariance measurements Following the analysis of Moncrieff et al. (1996), errors in eddy covariance measurements were separated into systematic and random errors. According to Loescher et al. (2005), different sonic anemometers (and presumably infra-red gas analysers) can have impacts on the measured fluxes of CO2 and water. In order to compare the eddy covariance units for systematic errors associated with instrument bias, both units were installed above short, wellwatered grass for several days during March 2007. Measurement height was 1.8 m, and the fetch was at least 40 m in all directions. Other systematic errors were minimised (or at least, applied in the same way to both fields) by comparing vertical and horizontal patterns of air movement, energy balance, and stationarity and integral turbulence statistics, as described above. The impact of random errors on the comparison of total carbon dioxide and water fluxes for the two fields was assessed by a conventional statistical approach. Initially, data was separated into periods of 1–5 months based on responses to seasonal conditions, so that carbon dioxide and water fluxes within a field were similar throughout the 1–5 month period. Within each 1–5 month period, the average flux was calculated for each 1-h interval (Moncrieff et al., 1996; Aubinet et al., 2000). Each day within the 1–5 month period was then treated as a ‘replicate’, and fluxes for the two fields were compared using Student’s t-test. Differences between the fields were considered significant at the 1% level.
P.R. Ward et al. / Agricultural and Forest Meteorology 152 (2012) 178–188
(a)
181
(b) Veg Ht Tree LAI 1.0 Tree LAI 2.0 Tree LAI 3.0
15
0.02
10
0.01
5
0.00
0 (d)
15
0.02
10
0.01
5
0.00
Contribution function (m -1 )
Vegetation height (m)
(c)
0 (e)
(f)
15
0.02
10
0.01
5
0.00
0 -400
-300
-200
-100
0
-400
-300
-200
-100
0
Distance from sensor (m) Fig. 2. Calculated footprint contribution functions for north or south winds for tree heights of 2 m (a and b), 3 m (c and d), and 4 m (e and f), for geostrophic wind speeds of 3 m/s (a, c and e) and 10 m/s (b, d and f). The maximum value for the contribution function in (b) is 0.047 m−1 .
3. Results and discussion 3.1. Comparison of eddy covariance units For the period in March 2007 when both units were installed over short, well-watered grass for a total of 68 h, regression of hourly averaged latent energy flux yielded a slope of 1.016, an intercept of 2.53 w/m2 , and r2 of 0.98. Values for LE varied between 0 w/m2 and 300 w/m2 . Corresponding values for the regression of CO2 flux were 0.994 (slope), 0.020 mg/m2 /s (intercept), and 0.96 (r2 ), with a range of −0.4 to +0.5 mg/m2 /s. These results indicate that systematic errors between the two eddy covariance units due to instrumentation are likely to be minimal. 3.2. Seasonal conditions Rainfall during the traditional winter and spring growing season (May–October) was 117 mm in 2006, and 251 mm in 2007. The long-term average is 244 mm, so 2007 was close to average
(46th percentile), but 2006 was much drier than average (lowest on record). Summer and autumn (November–April) rainfall was 182 mm for 2005/2006, 78 mm for 2006/2007, and 124 mm for 2007/2008. The long-term average for November to April is 101 mm, so the summer and autumn of 2005/2006 (leading into the first period of measurement) was considerably wetter than average (95th percentile), but the next two summer and autumn periods experienced close-to-average rainfall (31st and 67th percentiles, respectively). 3.3. Footprint analysis The eddy covariance technique is usually applied to landscapes with relatively uniform vegetation. However, for a ‘belt and alley’ field, vegetation is clearly not uniform, and the presence of regular belts of trees could induce complex flow patterns, particularly for wind directions across the belts. For this reason, we examined footprint patterns with the SCADIS tool (Sogachev and Lloyd, 2004; Sogachev et al., 2005b). For all modelled scenarios, more than 80%
182
P.R. Ward et al. / Agricultural and Forest Meteorology 152 (2012) 178–188
Table 1 Percentage of footprint signal sourced from within four tree belts (300 m for N or S winds, 420 m for NE, NW, SE, or SW winds) of the point of measurement. Wind
Tree
Sensor
Geostrophic wind 3 m/s
Direction
Height (m)
Height (m)
LAI 1.0
LAI 2.0
LAI 3.0
LAI 1.0
LAI 2.0
LAI 3.0
N, S (across tree belts)
2 3 4 2 3 4
3.7 5.4 6.0 3.7 5.4 6.0
96 93 92 97 95 94
94 90 89 96 93 92
93 89 88 95 92 91
97 93 91 96 94 94
na 88 85 na 92 92
na 87 83 na 91 91
NE, NW, SE, SW (45◦ to tree belts)
Geostrophic wind 10 m/s
Table 2 Percentage of footprint attributable to trees where trees were planted in belts of 10 m width with 60 m between belts (i.e. total tree area = 14.3%).
N, S
NE, NW, SE, SW
Tree height (m)
2 3 4 2 3 4
Sensor height (m)
3.7 5.4 6.0 3.7 5.4 6.0
(average 92%) of the total flux signal at the point of measurement was sourced from within four tree belts (Table 1), which is the minimum number of tree belts for any wind direction in the experimental oil mallee field. Taller trees with higher LAI resulted in a higher proportion of the signal coming from further than four belts, particularly for the higher geostrophic wind velocity, but this was always less than 20% of the total signal. In all calculations for winds at right angles to the tree belts, the presence of tree belts had a noticeable impact on the contribution function (which describes the contribution of the various ecosystem elements to the total flux measurement), particularly immediately upwind of the belt (Fig. 2). Similar results were calculated for winds at 45◦ to the tree belts (data not shown). In the calculations, tree belts occupied 14.3% of the field, and their contribution to the total footprint varied from 6.4% (for windy conditions at a measurement height of 3.7 m over 2 m trees with a LAI of 1.0) to 20.2% (for calm conditions at a measurement height of 6.0 m over 4 m trees with LAI of 1.0), and decreased with increasing LAI (Table 2). Oil mallee tree belts in south-western Australia have average LAI of between 2 and 3 (Oliver et al., 2005). For these values, a measurement height of 3.7 m, with a tree height of 2 m (as used for our measurements between May 2006 and July 2006), could result in under-estimation of the tree belt contribution to the total flux. However, when the measurement height was increased to 5.4 m (July 2006–October 2006) or 6.0 m (April 2007–May 2008), the tree belt (LAI 2 or 3) contribution to the calculated footprint varied between 10.2% and 15.6% for winds directly across the tree belts, which was close to the theoretical tree area of 14.3%. These calculations suggest that in theory, the eddy covariance technique can be applied to measurements from a ‘belt and alley’ field in order to determine relative contributions from the tree belts and intervening alley vegetation.
Geostrophic wind 3 m/s
Geostrophic wind 10 m/s
LAI 1.0
LAI 2.0
LAI 3.0
LAI 1.0
LAI 2.0
LAI 3.0
14.8 18.6 20.2 12.9 17.4 19.3
13.0 15.2 16.6 11.6 14.6 16.0
12.6 14.1 15.2 11.4 13.6 15.0
6.4 11.5 14.0 12.7 17.2 18.9
na 10.9 12.7 na 15.1 16.7
na 10.2 12.2 na 14.5 16.1
and south, blowing across the belts (Fig. 4). Standard deviation of wind direction for northerly or southerly winds was not affected by the tree belts for higher measurement heights, again suggesting that higher measurements are likely to give a better indication of ecosystem measurements in a ‘belt and alley’ system. However, given the previously studied effects of tree belts on wind patterns (e.g. Tuzet and Wilson, 2007), these impacts should be investigated for each application of eddy covariance to non-uniform vegetation.
(a)
Oil mallee field Control field
0.10
0.05
0.00 (b)
Frequency
Wind direction
0.10
0.05
0.00 (c) 0.10
0.05
3.4. Wind direction The presence of tree belts had little impact on wind direction for any of the measurement heights (Fig. 3), suggesting that ‘funnelling’ of wind along the tree belts did not happen to any significant extent during our measurements. However, for measurement height of 3.7 m, the standard deviation of wind direction was increased by the tree belts (relative to the control field) for winds from the north
0.00 0
90
180
270
360
Wind direction (degrees) Fig. 3. Frequency of wind direction (in 20◦ classes) for the oil mallee and control fields for periods when measurement height in the oil mallee field was 3.7 m (a), 5.4 m (b) and 6.0 m (c).
P.R. Ward et al. / Agricultural and Forest Meteorology 152 (2012) 178–188
183
Table 3 Total water and carbon dioxide fluxes from the oil mallee field for May–November 2006, and May 2007–April 2008, before and after removal of data associated with winds blowing along the direction of the tree belts. May–November 2006
H2 O (mm) CO2 (Mg/ha)
April 2007–May 2008
All wind directions
E and W winds excluded
All wind directions
E and W winds excluded
158 −3.42
159 −3.40
399 −15.06
398 −15.48
3.5. Vertical wind speed Vertical wind speed at the point of measurement was analysed to determine whether the regular tree belts were having a consistent impact on vertical air movement. W cor was generally close to zero for winds from the north or south (across the tree belts and alleys), but was biased towards positive values for wind directions within 22◦ of east and west (along the alleys) for all measurement heights (Fig. 5). These peaks are consistent with persistent patterns of vertical air movement caused by the repeating spatial patterns of vegetation type, and provide evidence for advective conditions. Data for winds from these directions were excluded from subsequent analysis, and in this instance, exclusion of this data had limited impact on total water and carbon dioxide fluxes (Table 3). Once again though, this should be investigated for each application of eddy covariance to non-uniform vegetation. 3.6. Stationarity and integral turbulence Stationarity was generally very good for u and w, with more than 95% of values having variances of less than 30% for both fields (Table 4). Stationarity for the quantities T, q and C was generally lower (22–54% of values lower than 30%), but the cross-products uw, wT, wq and wC were more stationary (50–99% of values <30%). In the oil mallee field, stationarity was generally similar to values measured in the control field. There was no consistent effect of
(a)
Average standard deviation of wind direction (degrees)
50
Oil mallee field Control field
40 30 20 Fig. 5. Vertical wind speed corrected for tilt angle for measurements taken at heights of 3.7 m (a), 5.4 m (b) and 6.0 m (c).
(b)
50 40 30
Table 4 Stationarity and integral turbulence (% of data points with variance or covariance <30%) for various heights of measurement (Zm ) in the control and oil mallee fields.
20
(c)
50 40 30 20 0
90
180
270
360
Wind direction (degrees) Fig. 4. Average standard deviation of wind direction (in 20◦ classes) for the oil mallee and control fields for periods when measurement height in the oil mallee field was 3.7 m (a), 5.4 m (b) and 6.0 m (c).
u w T q C uw wT wq wC ua wa a
Control field
Oil mallee field
Zm = 1.7 m
Zm = 3.7 m
96 100 52 32 54 98 86 64 81 51 94
96 100 45 32 49 99 86 73 80 88 100
Integral turbulence measurements.
Zm = 5.4 m 96 100 38 38 45 98 83 74 78 84 99
Zm = 6.0 m 96 99 41 22 53 98 80 50 78 83 99
P.R. Ward et al. / Agricultural and Forest Meteorology 152 (2012) 178–188
3.7. Energy balance With the eddy covariance technique, energy balance closure is usually around 80%, as discussed in a review of 22 sites by Wilson et al. (2002). Closure at our sites was similar to these values, varying between 67% and 80% (Table 5). Under the rainfall conditions experienced during the experiment, runoff and drainage losses were expected to be small (most likely, zero), and so according to the water balance equation, rainfall should be similar in magnitude to evapotranspiration. This was most clearly observed for eddy covariance data from the control field from December 2007 to January 2008 (Fig. 8a), with increases in cumulative ET matching increases in cumulative rainfall under both wet and dry conditions. The close alignment of rainfall and ET through this period gives confidence that the eddy covariance technique is providing a robust estimate of the water balance under the experimental conditions. It also suggests that the imbalance in energy is dominated by sensible rather than latent energy losses, or by other forms of energy not accounted for. For example, energy storage in the canopy was not estimated as part of our energy balance calculations. Carbon flux data also matches closely with patterns of vegetation growth. For these reasons, we are confident in concluding that the eddy covariance technique can successfully be applied to our ‘belt and alley’ field. 3.8. Evapotranspiration During the drier than average growing season of 2006, there was little difference in cumulative evapotranspiration (ET) measured from the two fields (both sown to legume pasture) between May and September (Fig. 6a). However, ET from the oil mallee field started to increase relative to the control field in September and October. On this basis, and also considering cumulative carbon dioxide flux (Fig. 6b), the data set was split into three periods, comprising May 25–July 31, 2006, August 1–August 31, 2006, and September 1–October 19, 2006 (Fig. 7). For the first two periods there was no significant difference in ET between the two fields, but ET rate in the second period (peaking at 0.18 mm/h) was considerably greater than in the first period (peaking at 0.10 mm/h). Total ET for these periods was about 45 and 42 mm for the first (67 days) and second (31 days) periods, respectively. In the third period, ET rate was maintained in the oil mallee paddock, but was significantly (p < 0.001) decreased from mid-morning to late afternoon in the control field. Total ET for the period (49 days) was 71 mm and 45 mm for the oil mallee and control fields, respectively. It is impossible to attribute this difference in ET rate to the presence of the oil mallees because the two pastures would have also been actively growing at this time.
250
Cumulative ET or rainfall (mm)
measurement height in the oil mallee field on stationarity of any of the quantities. A higher proportion of data for u and w met the integral turbulence benchmark of <30% variance from theoretical values in the oil mallee field compared with the control field (Table 4). With the exception of u in the control field (51%), satisfactory integral turbulence was achieved for u and w 83–100% of the time, and was slightly higher for values measured at 3.7 m in the oil mallee field. In the international eddy covariance research community, stationarity and integral turbulence are commonly used for quality control of eddy covariance measurements (e.g. Aubinet et al., 2000; Rebmann et al., 2005). In our results, stationarity and integral turbulence were similar in both fields, and as the eddy covariance technique is well suited to measurements over short vegetation (e.g. Baldocchi, 2003), this gives confidence in our measurements in the oil mallee field.
(a)
200
Oil mallee Control ET 0
150
Rainfall
100
50
0 1 May 06
Cumulative flux (Mg CO 2/ha)
184
1 Jul 06
1 Sep 06
(b)
0
-1
-2
-3
-4 1 May 06
1 Jul 06
1 Sep 06
Fig. 6. Cumulative evapotranspiration (actual and potential), and cumulative rainfall (a) and cumulative carbon dioxide flux (b) from the oil mallee field and the control field for the period May 25 2006 to October 19 2006.
Similar results were observed for the period between April and October 2007 (Fig. 8a). This block of data was split into four periods of interest comprising: April 3–July 31, 2007; August 1–October 1, 2007; November 1, 2007–March 30, 2008; and April 1–May 7, 2008 (Fig. 9). For the first, second and fourth periods, differences in evapotranspiration rate were not significant. In the third period, ET rate for both fields was much lower than for other periods, but despite this, ET rate from the oil mallee field was significantly (p < 0.001) greater than from the control field throughout the daylight hours. For this period, total ET was 62 mm from the oil mallee field and 35 mm from the control field. As expected, day time (6 am–6 pm) ET was much greater than night time ET (Figs. 10a and 11a) throughout all measurement periods. The best estimate of the difference in water use between the two fields attributable to the presence of the belts of oil mallees comes from the summer and autumn measurements, when the trees were the only living plants (Fig. 9c). For this period, the difference in water use between the two fields was 27 mm. This is important because extra water use by perennial vegetation relative to annual crops and pastures has been linked to reduced risk of dryland salinity (Ward, 2006). For comparison, previously published values of extra water use for lucerne (relative to annual crops and pastures) in a similar environment varied between 60 mm and 200 mm (summarised by Ward et al., 2006). Values quoted in these studies assumed that lucerne was grown over the whole landscape, but in the current study, oil mallees were only planted on about 14% of the landscape. Therefore, the extra water use on the 14% of the landscape occupied by trees was of the order of 210 mm. This water was likely drawn from deep in the soil below the tree belts, and also
P.R. Ward et al. / Agricultural and Forest Meteorology 152 (2012) 178–188
185
Table 5 Slope of linear regression between H + LE and R-G for the control and oil mallee fields, at the indicated height of measurement. Period
Measurement height (m)
May 25, 2006–July 3, 2006 July 13, 2006–Oct 19, 2006 April 4, 2007–May 7, 2008
Slope of regression
Control
Oil mallee
Control
Oil mallee
1.7 1.7 1.7
3.7 5.4 6.0
0.74 0.67 0.72
0.74 0.76 0.80
from significant distances laterally from the tree belt, as observed by Ellis et al. (2005). Extra water use provided by tree belts relative to a field without trees is likely to help restrict the spread of dryland salinity. Calculations using LeBuM, which calculates long-term average leakage below the root zone based on soil water deficits (Ward, 2006), suggest that average annual groundwater recharge for the region could be reduced from 6 mm to less than 2 mm by incorporating belts of oil mallees into the farming system. However, because the spatial distribution of the extra water use is likely to be non-uniform, the actual impact on groundwater recharge difficult to estimate (Walker et al., 2003), and further research with spatial groundwater models will be necessary to determine the likely outcome. Carter et al. (2004) and Wildy et al. (2004) used sap flow techniques in similar soil types and climatic zones to measure oil mallee ET of 2–3 mm/day (based on tree crown area) in December, in the absence of groundwater. Assuming coverage of 14% of the landscape, this equates to 0.3–0.4 mm/day at the landscape scale. In our measurements, the amount of ET attributable to trees during the dry conditions experienced in January 2008 was 0.21 mm/day, which is substantially less than the rate reported by Carter et al. (2004) and Wildy et al. (2004). (December values were not used because of rainfall during the month). As noted by Crosbie
et al. (2008), scaled-up sap flow calculations might over-estimate ecosystem-scale ET from tree belts by as much as 100% due to difficulties in estimating the area actually occupied by the belts, and our results provide further support for this. Nevertheless, the extra water use provided by the oil mallees is likely to assist in restricting the spread of dryland salinity. 3.9. Carbon balance Based on graphs of cumulative NEE (Figs. 6b and 8b), average hourly rates of NEE were calculated for the same periods as used for analysis of the ET results (Figs. 12 and 13). For the first period in the 2006 data set (autumn and early winter), NEE for the control field was significantly (p < 0.001) greater during daylight hours than for the oil mallee field, but similar at night. Total NEE for this period was −1.03 Mg CO2 /ha in the control field and −0.13 Mg CO2 /ha in the oil mallee field. Note that random errors of 20% (e.g. Moncrieff et al., 1996) could change the sign of the NEE value in the oil mallee field, so this field was not necessarily a CO2 sink during the early winter period. During the late winter and early spring period, NEE was similar (p = 0.262) for both fields, with greater (higher in magnitude) fluxes during both daylight and night hours. Total NEE for period 2 was −0.86 Mg CO2 /ha and −0.74 Mg CO2 /ha in the control and
(a)
500 Cumulative ET or rainfall (mm)
Oil mallee Control
0.15 0.10 0.05 0.00
0.10
400 300 200
Oil mallee Control Potential ET Rainfall
100 0 1 Apr 07
0.05 0.00
Cumulative flux (Mg CO 2 /ha)
ET rate (mm/hr)
(b) 0.15
(a)
(c) 0.15 Hour vs +H2O Aug 1 - Aug 31 Hour vs -H2O Aug 1 - Aug 31
0.10 0.05 0.00 0
6
12
18
1 Jul 07
1 Oct 07
1 Jan 08
1 Apr 08
1 Jul 07
1 Oct 07
1 Jan 08
1 Apr 08
0 -4 -8 -12 -16
(b)
24
Hour Fig. 7. Average daily patterns of evapotranspiration for the oil mallee and control fields for the periods: (a) May 25–July 31, 2006; (b) August 1–August 31 2006; and (c) September 1–October 19 2006. Error bars represent ±1 standard error.
1 Apr 07
Fig. 8. Cumulative evapotranspiration (actual and potential), and cumulative rainfall (a) and cumulative carbon dioxide flux (b) from the oil mallee field and the control field for the period April 4 2007–May 8 2008.
186
P.R. Ward et al. / Agricultural and Forest Meteorology 152 (2012) 178–188 3
Evapotranspiration (mm/day)
(a) Oil mallee Control
0.2 0.1 0.0
(b)
(a)
1 night time 0
CO 2 flux (Mg/ha/day)
ET rate (mm/hr)
night time
(b)
0.2 0.1 0.0 (c)
Oil mallee field Control field
day time 2
0.0
-0.1
day time
-0.2
-0.3
0.2
May
Jun
Jul
Aug
Sep
Oct
Nov
Dec
Jan
Feb Mar
Apr
May
Date
0.1 Fig. 11. Seven-day moving average net day time and night time water flux (a) and NEE (b) for the period April 4 2007–May 8 2008.
0.0 (d) 0.2 0.1 0.0 0
6
12
18
24
oil mallee fields, respectively. Later in spring (period 3), NEE was maintained at the higher level in the oil mallee field, but was significantly (p < 0.001) lower during daylight hours in the control field. Total NEE for period 3 was −1.18 Mg CO2 /ha and −2.48 Mg CO2 /ha in the control and oil mallee fields, respectively. As discussed above, differences between the fields in any period could be due to differences in patterns of growth of the two pastures in the fields, and cannot be ascribed to the presence of oil mallee belts.
Hour 0.004
Fig. 9. Average daily patterns of evapotranspiration for the oil mallee and control fields for the periods: (a) April 4–July 31 2007; (b) August 1–October 1 2007; (c) November 1 2007–March 30 2008; and (d) April 1–May 7 2008. Error bars represent ±1 standard error.
0.000 -0.004
Net Ecosystem Exchange (t CO 2/ha/hr)
Evapotranspiration (mm/day)
2.0 Oil mallee field Control field
1.5 1.0
day time
0.5 night time 0.0 (a) 0.05
CO 2 flux (Mg/ha/day)
night time 0.00
Oil mallee Control
-0.008 (a) 0.004 0.000 -0.004 -0.008 (b) 0.004
0
6
12
18
24
18
24
X Data
0.000
-0.05
day time
-0.004
-0.10
-0.008
-0.15
(c)
(b) -0.20 Jun
Jul
Aug
Sep
Oct
Date Fig. 10. Seven-day moving average net day time and night time water fluxes (a) and NEE (b) for the period May 25 2006–October 19 2006.
0
6
12
Hour Fig. 12. Average daily patterns of NEE for the oil mallee and control fields for the periods: (a) May 25–July 31, 2006; (b) August 1–August 31 2006; and (c) September 1–October 19 2006. Error bars represent ±1 standard error.
P.R. Ward et al. / Agricultural and Forest Meteorology 152 (2012) 178–188
0.00 -0.01
Net ecosystem exchange (t CO2 /ha/hr)
-0.02
Oil mallee Control
(a)
0.00 -0.01
(b)
-0.02
0.00 -0.01
(c)
-0.02
0.00 -0.01
(d)
-0.02 0
6
12
18
24
Hour Fig. 13. Average daily patterns of NEE for the oil mallee and control fields for the periods: (a) April 4–July 31 2007; (b) August 1–October 1 2007; (c) November 1 2007–March 30 2008; and (d) April 1–May 7 2008. Error bars represent ±1 standard error.
Nevertheless, they clearly show different seasonal patterns of NEE in the two fields. Under the close-to-average rainfall conditions experienced in 2007, NEE was much greater than in 2006 (Figs. 6b and 8b). NEE was analysed in the same periods as used for evapotranspiration data (Fig. 13). NEE was initially significantly (p < 0.001) greater in the oil mallee (legume pasture) field, with total NEE for the autumn and winter period of −4.29 Mg CO2 /ha, compared with −1.51 Mg CO2 /ha in the control (wheat) field. During period 2 (late winter and early spring), NEE for both fields was similar during the day, but a slightly higher respiration rate in the oil mallee field resulted in a significant difference (p < 0.001), between the control field (−8.86 Mg CO2 /ha) and the oil mallee field (−8.03 Mg CO2 /ha). Over the summer and autumn of 2007/2008 (periods 3 and 4; Fig. 13c and d), net carbon uptake continued in the oil mallee field but fluxes were much lower (close to 0) in the control field. This pattern continued even after rainfall in late March 2008 (Fig. 13d). Day time net CO2 flux peaked at about −0.15 Mg/ha/day in September 2006, and at about −0.25 Mg/ha/day during the wetter 2007 growing season (Figs. 10b and 11b). Night time CO2 flux in both fields tended to follow patterns of water availability and temperature. Maximum rates were observed in August, when soils are typically moist and temperatures are starting to rise. Rapid responses in night time fluxes were also observed in response to substantial summer rainfall events (e.g. 18.2 mm on December 18, and 16.2 mm on March 31), but not for smaller rainfall events (e.g. 5.4 mm on February 8). Similar results for grassland in a
187
Mediterranean-style climate were reported by Xu and Baldocchi (2004). Frequent rainfall events of more than 10–15 mm could have large impacts on annual net carbon exchange in Mediterraneanstyle environments such as southern Australia. The cumulative results presented in Figs. 6b and 8b do not represent net sequestration for the system, due to export of carbon in the form of grain or livestock. Nevertheless, differences in NEE measured during summer and autumn (November to April, when export from the system was minimal) do represent carbon uptake by the oil mallees relative to the control field, and can be used to estimate the difference in sequestration between the two fields. Measurements between November 2007 and March 2008 indicated that the NEE in the oil mallee field was −1.49 Mg CO2 /ha, and the corresponding number for the control field was −0.56 Mg CO2 /ha. This difference between the two fields (0.93 Mg CO2 /ha) is far greater than the 1–2% difference measured during the comparison of ET units over the same vegetation type. Assuming that the difference of between the two fields is solely attributable to the presence of the oil mallees, this is equivalent to a dry matter (DM) gain of around 0.47 Mg DM/ha at the ecosystem scale, which will include both above- and below-ground growth. Wildy and Pate (2002) measured above- and below-ground growth of oil mallee trees growing in belts of around 10 kg/tree/year and 3 kg/tree/year, respectively, which with their particular configuration of tree belts, equated to 1.4 Mg DM/ha/year. Carbon capture by an agroforestry system will depend on seasonal conditions, as well as the spacing and arrangement of trees within the system. For this reason, it is difficult to compare results from one study with those of another. Nevertheless, the results presented here confirm that the incorporation of belts of trees into a landscape dominated by annual agricultural communities will lead to increased carbon sequestration. Further research will be necessary to quantify net carbon sequestration associated with spatial arrangements of trees, and this could affect the adoption of ‘belt and alley’ farming systems. Carbon sequestration of around 1 Mg CO2 /ha, as measured for the 2007/2008 summer, could be enough to make oil mallee alley farming more attractive in the south-west of Western Australia. With the potential for carbon trading, a price of $23/Mg (as recently introduced by the Australian Government) provides the oil mallees with extra value of $23/ha. Although this on its own is not enough to offset the value of the loss of production from the annual crops displaced by the tree belts, when combined with other tree products (solvent oil, energy, activated charcoal) and environmental benefits (salinity, biodiversity), belts of oil mallees might become a more attractive option than they currently are (Roberts et al., 2009). 4. Conclusions The eddy covariance technique can be successfully applied to a belt and alley farming system, although account must be made of possible influences on wind direction and patterns of advective air movement. Measurements of water use indicated that belts of oil mallees occupying 14% of the field used an extra 28 mm relative to an adjacent field without oil mallees. This extra water use is lower than other published measurements of water use for herbaceous perennial vegetation occupying 100% of the landscape, but is still likely to reduce the risk of dryland salinity. CO2 measurements indicate that carbon sequestration of 0.93 Mg CO2 /ha was achieved during the 2007/2008 summer, which, with adequate economic incentives, could help to increase adoption of oil mallees. Carbon sequestration, NEE and ET are likely to vary considerably depending on seasonal conditions (in particular, summer rainfall quantity and distribution), and further research will be necessary to determine values over the longer term.
188
P.R. Ward et al. / Agricultural and Forest Meteorology 152 (2012) 178–188
Acknowledgements Thanks to Neil Ballard and family for their enthusiastic adoption of oil mallees, and for allowing us to measure their water use and carbon sequestration potential. Thanks also to Ian Foster of DAFWA for access to their weather station data from the oil mallee paddock. Ray Leuning provided very helpful advice on the analysis of eddy covariance data, and the potential impact of tree belts on advective patterns of air movement. Rob Clement assisted with aspects of eddy covariance data analysis. References Aubinet, M., Grelle, A., Ibrom, A., Rannik, Ü., Moncrieff, J., Foken, T., Kowalski, A.S., Martin, P.H., Berbigier, P., Bernhofer, C., Clement, R., Elbers, J., Granier, A., Grünwald, T., Morgenstern, K., Pilegaard, K., Rebmann, C., Snijders, W., Valentini, R., Vesala, T., 2000. Estimates of the annual net carbon and water exchange of forests: the EUROFLUX methodology. In: Fitter, A.H., Raffaelli, D.G. (Eds.), Advances in Ecological Research. Academic Press, New York, pp. 113–175. Baldocchi, D.D., 2003. Assessing the eddy covariance technique for evaluating carbon dioxide exchange rates of ecosystems: past, present and future. Global Change Biology 9, 479–492. Baldocchi, D., 2008. TURNER REVIEW No. 15. ‘Breathing’ of the terrestrial biosphere: lessons learned from a global network of carbon dioxide flux measurement systems. Australian Journal of Botany 56, 1–26. Carter, J., Ward, P.R., White, D.A., Brooksbank, K., Rance, S., Walker, S., 2004. Water use and growth of oil-mallees with variation in groundwater availability. In: Ridley, A., Feikema, P., Bennett, S., Rogers, M.J., Wilkinson, R., Hirth, J. (Eds.), Salinity Solutions: Working with Science and Society. CRC for Plant-Based Management of Dryland Salinity, Bendigo, Australia. Crosbie, R., Wilson, B., Hughes, J., McCulloch, C., 2008. The upscaling of transpiration from individual trees to areal transpiration in tree belts. Plant and Soil 305, 25–34. Ellis, T., Hatton, T., Nuberg, I., 2005. An ecological optimality approach for predicting deep drainage from tree belts of alley farms in water-limited environments. Agricultural Water Management 75, 92–116. Falge, E., Baldocchi, D., Olson, R., Anthoni, P., Aubinet, M., Bernhofer, C., Burba, G., Ceulemans, R., Clement, R., Dolman, H., Granier, A., Gross, P., Grünwald, T., Hollinger, D., Jensen, N.-O., Katul, G., Keronen, P., Kowalski, A., Lai, C.T., Law, B.E., Meyers, T., Moncrieff, J., Moors, E., Munger, J.W., Pilegaard, K., Rannik, Ü., Rebmann, C., Suyker, A., Tenhunen, J., Tu, K., Verma, S., Vesala, T., Wilson, K., Wofsy, S., 2001. Gap filling strategies for defensible annual sums of net ecosystem exchange. Agricultural and Forest Meteorology 107, 43–69. Foken, T., Wichura, B., 1996. Tools for quality assessment of surface-based flux measurements. Agricultural and Forest Meteorology 78, 83–105. Goulden, M.L., Munger, J.W., Fan, S.M., Daube, B.C., Wofsy, S.C., 1996. Measurements of carbon sequestration by long-term eddy covariance: methods and a critical evaluation of accuracy. Global Change Biology 2, 169–182. Haile, S.G., Nair, P.K.R., Nair, V.D., 2008. Carbon storage of different soil-size fractions in Florida silvopastoral systems. Journal of Environmental Quality 37, 1789–1797. Lefroy, E.C., Flugge, F., Avery, A., Hume, I., 2004. Potential of current perennial plantbased farming systems to deliver salinity management outcomes and improve prospects for native biodiversity: a review. In: Ridley, A., Feikema, P., Bennett, S., Rogers, M.J., Wilkinson, R., Hirth, J. (Eds.), Salinity Solutions: Working with Science and Society. CRC for Plant-Based Management of Dryland Salinity, Bendigo, Australia. Loescher, H.W., Ocheltree, T., Tanner, B., Swiatek, E., Dano, B., Wong, J., Zimmerman, G., Campbell, J., Stock, C., Jacobsen, L., Shiga, Y., Kollas, J., Liburdy, J., Law, B.E., 2005. Comparison of temperature and wind statistics in contrasting environments among different sonic anemometer-thermometers. Agricultural and Forest Meteorology 133, 119–139. Mao, S.L, Leclerc, M.Y., Michaelides, E.E., 2008. Passive scalar flux footprint analysis over horizontally inhomogeneous plant canopy using large-eddy simulation. Atmospheric Environment 42, 5446–5458. Markkanen, T., Rannik, U., Marcolla, B., Cescatti, A., Vesala, T., 2003. Footprints and fetches for fluxes over forest canopies with varying structure and density. Boundary-Layer Meteorology 106, 437–459. McFarlane, D.J., George, R.J., Caccetta, P.A., 2004. The extent and potential area of saltaffected land in Western Australia estimated using remote sensing and digital terrain models. In: 1st National Salinity Engineering Conference, Perth, Western Australia. Moncrieff, J.B., Malhi, Y., Leuning, R., 1996. The propagation of errors in long-term measurements of land-atmosphere fluxes of carbon and water. Global Change Biology 2, 231–240.
Nair, P.K.R., Kumar, B.M., Nair, V.D., 2009. Agroforestry as a strategy for carbon sequestration. Journal of Plant Nutrition and Soil Science-Zeitschrift Fur Pflanzenernahrung Und Bodenkunde 172, 10–23. Oliver, Y.M., Lefroy, E.C., Stirzaker, R., Davies, C.L., 2005. Deep-drainage control and yield: the trade-off between trees and crops in agroforestry systems in the medium to low rainfall areas of Australia. Australian Journal of Agricultural Research 56, 1011–1026. Ong, C.K., Wilson, J., Deans, J.D., Mulayta, J., Raussen, T., Wajja-Musukwe, N., 2002. Tree–crop interactions: manipulation of water use and root function. Agricultural Water Management 53, 171–186. Parrotta, J.A., 1999. Productivity, nutrient cycling, and succession in single- and mixed-species plantations of Casuarina equisetifolia, Eucalyptus robusta, and Leucaena leucocephala in Puerto Rico. Forest Ecology and Management 124, 45–77. Rebmann, C., Göckede, M., Foken, T., Aubinet, M., Aurela, M., Berbigier, P., Bernhofer, C., Buchmann, N., Carrara, A., Cescatti, A., Ceulemans, R., Clement, R., Elbers, J.A., Granier, A., Grünwald, T., Guyon, D., Havránková, K., Heinesch, B., Knohl, A., Laurila, T., Longdoz, B., Marcolla, B., Markkanen, T., Miglietta, F., Moncrieff, J., Montagnani, L., Moors, E., Nardino, M., Ourcival, J.M., Rambal, S., Rannik, Ü., Rotenberg, E., Sedlak, P., Unterhuber, G., Vesala, T., Yakir, D., 2005. Quality analysis applied on eddy covariance measurements at complex forest sites using footprint modelling. Theoretical and Applied Climatology 80, 121–141. Roberts, A.M, Helmers, M.J., Fillery, I.R.P., 2009. The adoptability of perennial-based farming systems for hydrologic and salinity control in dryland farming systems in Australia and the United States of America. Crop and Pasture Science 60, 83–99. Sogachev, A., Lloyd, J., 2004. Using a one-and-a-half order closure model of the atmospheric boundary layer for surface flux footprint estimation. Boundary-Layer Meteorology 112, 467–502. Sogachev, A., Leclerc, M.Y., Karipot, A., Zhang, G., Vesala, T., 2005a. Effect of clearcuts on footprints and flux measurements above a forest canopy. Agricultural and Forest Meteorology 133, 182–196. Sogachev, A., Panferov, O., Gravenhorst, G., Vesala, T., 2005b. Numerical analysis of flux footprints for different landscapes. Theoretical and Applied Climatology 80, 169–185. Stirzaker, R.J, Cook, F.J., Knight, J.H., 1999. Where to plant trees on cropping land for control of dryland salinity: some approximate solutions. Agricultural Water Management 39, 115–133. Stirzaker, R.J., Lefroy, E.C., Ellis, T.W., 2002. An index for quantifying the trade-off between drainage and productivity in tree–crop mixtures. Agricultural Water Management 53, 187–199. Takimoto, A., Nair, P.K.R., Nair, V.D., 2008. Carbon stock and sequestration potential of traditional and improved agroforestry systems in the West African Sahel. Agriculture Ecosystems & Environment 125, 159–166. Takimoto, A., Nair, V., Nair, P., 2009. Contribution of trees to soil carbon sequestration under agroforestry systems in the West African Sahel. Agroforestry Systems 76, 11–25. Tuzet, A., Wilson, J.D., 2007. Measured winds about a thick hedge. Agricultural and Forest Meteorology 145, 195–205. Walker, G., Bartel, B., Bourne, J., Hamilton, A., Young, D., 2003. Low rainfall alley farming: potential for reducing regional salinity in the Murray Basin. Natural Resource Management 6, 13–23. Ward, P.R., 2006. Predicting the impact of perennial phases on average leakage from farming systems in south-western Australia. Australian Journal of Agricultural Research 57, 269–280. Ward, P.R., Micin, S.F., Dunin, F.X., 2006. Using soil, climate, and agronomy to predict soil water use by lucerne compared with soil water use by annual crops or pastures. Australian Journal of Agricultural Research 57, 347–354. Webb, E.K., Pearman, G.I., Leuning, R., 1980. Correction of flux measurements for density effects due to heat and water vapour transfer. Quarterly Journal of the Royal Meteorological Society 106, 85–100. Wildy, D.T, Pate, J.S., 2002. Quantifying above- and below-ground growth responses of the Western Australian oil mallee, Eucalyptus kochii subsp. plenissima, to contrasting decapitation regimes. Annals of Botany 90, 185–197. Wildy, D., Pate, J., Bartle, J., 2004. Budgets of water use by Eucalyptus kochii tree belts in the semi-arid wheatbelt of Western Australia. Plant and Soil 262, 129–149. Wilczak, J., Oncley, S., Stage, S., 2001. Sonic anemometer tilt correction algorithms. Boundary-Layer Meteorology 99, 127–150. Wilson, K., Goldstein, A., Falge, E., Aubinet, M., Baldocchi, D., Berbigier, P., Bernhofer, C., Ceulemans, R., Dolman, H., Field, C., Grelle, A., Ibrom, A., Law, B.E., Kowalski, A., Meyers, T., Moncrieff, J., Monson, R., Oechel, W., Tenhunen, J., Valentini, R., Verma, S., 2002. Energy balance closure at FLUXNET sites. Agricultural and Forest Meteorology 113, 223–243. Wu, H.W, Fu, Q., Giles, R., Bartle, J., 2007. Production of mallee biomass in Western Australia: Energy balance analysis. In: International Conference on Bioenergy Outlook, SINGAPORE, American Chemical Society, United States, pp. 190–198. Xu, L.K., Baldocchi, D.D., 2004. Seasonal variation in carbon dioxide exchange over a Mediterranean annual grassland in California. Agricultural and Forest Meteorology 123, 79–96. Zhang, G., Thomas, C., Leclerc, M.Y., Karipot, A., Gholz, H.L., Binford, M., Foken, T., 2007. On the effect of clearcuts on turbulence structure above a forest canopy. Theoretical and Applied Climatology 88, 133–137.