Tree diversity enhances tree transpiration in a Panamanian forest plantation



1. Tree plantations play an important role in meeting the growing demand for wood, but there is concern about their high rates of water use. Recent approaches to reforestation in the tropics involve the establishment of multispecies plantations, but few studies have compared water use in mixed vs. monospecific stands.

2. We hypothesized that tree species diversity enhances stand transpiration. Tree water use rates were estimated in monocultures (n = 5), two-species mixtures (n = 3), three-species mixtures (n = 3) and five-species mixtures (n = 4). Sap flux densities were monitored with thermal dissipation probes in 60 trees for 1 year in a 7-year-old native tree plantation in Panama. We also estimated changes in the amount of wood produced per unit water transpired (i.e. water use efficiency, WUEwood).

3. Annual stand transpiration rates in two-/three-species mixtures (464 ± 271 mm year−1) and five-species mixtures (900 ± 76 mm year−1) were 14% and 56% higher than those of monocultures (398 ± 293 mm year−1), respectively. Trees growing in mixtures had larger diameters, conductive sapwood and basal area than those in monocultures, which partly explained the enhanced stand transpiration in mixtures.

4. The five-species mixtures maintained equally high stand transpiration rates during wet (2·64 ± 0·30 mm day−1) and dry seasons (2·51 ± 0·21 mm day−1), whereas monocultures and two-species mixtures had significantly lower transpiration rates during the dry season, because of the presence of dry season deciduous species.

5. The WUEwood of the five-species mixtures (2·1 g DM kg−1 H2O) was about half that of either monocultures, two- or three-species mixtures.

6. The comparably high stand transpiration rates in the five-species plots may arise from enhanced vegetation-atmosphere-energy exchange through higher canopy roughness and/or complementary use of soil water.

7.Synthesis and applications. Stand transpiration increased linearly with tree species richness and basal area in monocultures, two- and three-species mixtures, but the ratio of stand transpiration to basal area was larger for five-species mixtures. In conclusion, species selection and consideration of species richness and composition is crucial in the design of plantations to maximize wood production while conserving water resources.


The world-wide increase in the number and area of tree plantations is especially pronounced in the tropics. The main factor driving the expansion of plantation forestry is the growing regional and global demand for wood. Currently, most plantations in the tropics are monocultures of pine, eucalyptus, teak and acacia species (FAO 2005; Lamb, Erskine & Parrotta 2005). Such monospecific and intensively managed plantations produce high yields, but have made only minor contributions to the restoration of ecosystem health (Lamb, Erskine & Parrotta 2005). To improve ecosystem function and biodiversity, several authors advocate the establishment of mixed-species plantations (e.g. Montagnini & Jordan 2005; Paquette & Messier 2010). In addition to their improved ecological value, there are some indications that mixed-species plantations are more productive than monospecific stands (Forrester, Bauhus & Khanna 2004; Bristow et al. 2006).

Enhanced productivity is often associated with increased tree water use (Law et al. 2002). For example, significantly higher transpiration rates have been measured in a very productive eucalyptus plantation in comparison to a natural forest in Ethiopia (Fritzsche et al. 2006). However, whether the increase in water consumption is proportional to biomass increase and whether this relationship changes with increasing tree diversity is unclear (Forrester et al. 2010). The measure linking the amount of stem biomass (or carbon) produced with the amount of water transpired to the atmosphere is defined as water use efficiency per amount wood production (WUEwood, Hubbard et al. 2010). WUE of tree plantations is relevant for water management but to date has mainly been studied in eucalyptus plantations (Whitehead & Beadle 2004; Forrester et al. 2010).

Both carbon gain and water use are controlled by a variety of physical, biological and chemical factors. Carbon uptake and tree growth are mainly influenced by nutrient and water availability (Kozlowski, Kramer & Pallardy 1991; Rennenberg et al. 2009), light conditions (Binkley et al. 2010) and silvicultural treatments (Hubbard et al. 2010). Higher productivity of species-rich plant communities compared with monospecific communities is often explained by two mechanisms: complementarity and selection (Hooper et al. 2005). The complementarity mechanism implies that mixtures are able to access and use resources more efficiently because they consist of species with a variety of functional traits that may complement each other (Tilman, Lehman & Thomson 1997; Firn, Erskine & Lamb 2007). The selection (or sampling) effect suggests that with increasing number of species, it is more likely to have one or several highly productive species present in the mixture (Tilman, Lehman & Thomson 1997).

Tree water use and transpiration are controlled by stomatal and boundary layer conductance (Jarvis & McNaughton 1986). Tree water use is also influenced by canopy structure, conductive sap wood area and diameter (Meinzer, Goldstein & Andrade 2001). Mixed-species stands differ from monocultures in their structural characteristics, such as canopy architecture and leaf traits (Menalled, Kelty & Ewel 1998; Bauhus, van Winden & Nicotra 2004) as well as in the spatial and temporal stratification of roots (da Silva et al. 2009). Mixtures of plants with diverse stature and traits may use the available resources more efficiently and thus enhance plant growth and most probably water consumption (Law et al. 2002). Comparing water use in mixed eucalyptus–acacia plantations with monocultures revealed higher rates of stand transpiration in the mixed plantation (Forrester et al. 2010). Enhanced rates of water use in mixed stands may involve complementarity or selection mechanisms. Distinguishing between these two mechanisms has implications for the selection of species mixtures to maximize plantation productivity given a certain level of resource availability.

The focus of our study was to (i) estimate transpiration rates and WUE in monocultural and mixed-species plots, (ii) assess the effect of tree diversity on tree water use rates and (iii) assess the complementarity and selection effect upon basal area and stand transpiration using the additive partitioning approach developed by Loreau & Hector (2001). We hypothesized that higher tree diversity enhances tree water use and stand transpiration rates.

Materials and methods

Study site

The study was conducted in an experimental plantation close to the village of Sardinilla, Central Panama (9°19′N, 79°38′W), 50 km north of Panama City. The elevation of the site is approximately 70 m above sea level. Mean annual precipitation of the nearby meteorological station in Salamanca is 2300 mm (1977–2007), with a pronounced dry season from January to March (STRI 2010). The mean annual temperature in the region is 26·2 °C. The clay-rich soils in Sardinilla are classified as Typic and Aquic Tropudalfs and are derived from limestone (Potvin, Whidden & Moore 2004). Most of the area was clear-cut in the 1950s, but the original vegetation in the area around Sardinilla was most probably a tropical moist forest similar to that on Barro Colorado Island (Holdridge & Budowski 1956).

Experimental design of the plantation

The experimental tree plantation was set up with plots of varying tree species richness and species combinations (Fig. 1). In total, 24 plots were established between June and July 2001. Each plot was 45 × 45 m and was further divided into four subplots of 22·5 × 22·5 m. Six native tree species were planted based on their range of relative growth rates (Scherer-Lorenzen et al. 2005): the fast-growing species Luehea seemannii (Triana & Planch, Tiliaceae) and Cordia alliodora [(Ruiz & Pavon) Oken, Boraginaceae]; the intermediate species Anacardium excelsum [(Bertero & Balb. ex Kunth) Skeels, Anacardiaceae] and Hura crepitans (Linné, Euphorbiaceae) and the slow growing species Tabebuia rosea [(Bertol.) DC, Bignoniaceae] and Cedrela odorata (Liné, Meliaceae). The relative growth rates measured in the 50-ha permanent plot on Barro Colorado Island are 9·1%, 7·0%, 5·9%, 4·9%, 3·4% and 2·3% per year, respectively.

Figure 1.

 Plot layout and description of plot species composition, Sardinilla, Panama. Quadrats highlighted in grey indicate subplots where selected trees were located.

Cedrela odorata is deciduous, whereas the other species were classified as semi-deciduous. Seedlings were planted with 3-m spacing that is the standard commercial planting density in Central America, equating to 1111 stems per ha. The planting design included two monoculture plots for each species (12 monoculture plots in total), six-three-species mixture plots with different species combinations and six-six-plots containing all six species (Fig. 2). However, all C. alliodora trees grown in monocultures and the majority of C. alliodora in the three-species and six-species mixtures died within the first year probably due to undrained and compacted soil (Potvin & Gotelli 2008). Consequently, we did not include C. alliodora in this study. Measurements and analyses were carried out using the actual diversity at the time of study, consisting of three-two-species mixtures, three-three-species mixtures and four-five-species mixtures.

Figure 2.

 Left: Normalized daily sap flux density shown as annual course (weekly means for each species, n = 4). Right: Scatter plots show the relationships between sap flux in mixtures and monoculture. Straight line represents the 1 : 1 relation. Photosynthetic photon flux density (PPFD), vapour pressure deficit (VPD), weekly rainfall (annual total 2260 mm) and soil moisture (mean of n = 4 sensors per soil depth ± SD) over the course of the study period. No VPD data were available for June 2008 because of sensor failure. Mean PPFD and VPD were significantly higher during the dry season than during the wet season (< 0·001).

Micrometeorological and soil moisture measurements

Micrometeorological data including photosynthetic photon flux density (PPFD, mol m−2 s−1), air temperature (°C), relative humidity (%) and precipitation (mm) were provided by ETH Zürich. Vapour pressure deficit (VPD, kPa) was calculated from air temperature and relative humidity. Soil moisture content was measured with time domain reflectometry sensors (TDR, CS610; Campbell Scientific Inc., Logan, UT, USA) at four locations within the plantation at three depths (10, 35 and 60 cm) and stored on data loggers (CR800 and CR1000; Campbell Scientific Inc.).

Sap flux density, tree water use and stand transpiration

Sap flux measurements were conducted in the five surviving tree species. Sample trees were chosen in five monoculture plots (one plot for each of the five investigated species, with four sample trees per plot); three plots of two-species mixtures (four sample trees per species), three plots of three-species mixtures (four sample trees per species) and four replicate plots of the five-species mixture (one sample tree of each species in every five-species mixture plot, Fig. 1). Sap flux density was measured with thermal dissipation sensors (Granier 1985). Every tree was equipped with two sensors, placed on the southern and northern sides of the trunk at 130 cm above the ground. Probe output voltage was recorded every 30 s, and the average value stored every 15 min (CR800 and CR1000 datalogger; AM16/32 and AM416 multiplexer). Sap flux density (Js in g cm−2 h−1) was calculated from differences in voltage using the calibration equation determined by Granier (1987) and averaged for the two sensors per tree prior to further calculations. Changes of sap flux density with increasing xylem depth were assessed for each tree in the last 2 weeks of the experiment. Js was measured at 20–40 mm depth below the cambium and was also determined at a third depth (40–60 mm) in trees with a diameter >8 cm. Sap flux was measured from July 2007 to June 2008 (365 days). Sap flux sensors were changed when necessary to ensure that sap flux was measured in the outermost part of the sapwood.

The estimation of water use (Q) for individual trees was based upon the sapwood area of the tree and the radial changes in Js present in the sapwood. Species-specific xylem depth was estimated by dye injection for three trees per species, showing that sapwood depth of all species was greater than sensor length. Sap flux density was extrapolated to tree water use (Q, kg day−1) by summing up the water flow in a given number of ring-shaped stem cross sections corresponding to the respective installation depth, Js as measured at the reference depth and the normalized profile of Js for the species considered (Edwards, Becker & Cermák 1996):


where Qi is the water flow through ring i, Jsci (g cm−2 day−1) is the cumulative sap flux density and Ai (cm2) is the ring-shaped area of sapwood that extends between the tip and the end of each probe for a given depth interval i.

Tree water use was up scaled to stand transpiration rate (Tstand mm day−1) as follows:


where QDBH is the water use rate of a given tree and Aplot (m2) is plot area. QDBH was estimated by deriving relationships between measured tree water use rates and tree diameters for each day and species. We excluded the outer tree row of each plot from the analysis to reduce edge effects on tree development.

Tree and stand structure, leaf area index, biomass increment and water use efficiency

Diameter at breast height (DBH) of all sample trees was measured with a girth tape. Tree height and crown base height of each sample tree were determined with a hypsometer (Vertex III; Haglöf, Lensele, Sweden). The crown extension in each of eight cardinal directions was estimated by projecting the edges of the crown vertically to ground using a 5-m-long plastic tube. Crown projection area was calculated as the sum of eight pitch circles.

Hemispherical photographs were taken vertically with a digital camera (Minolta Dimage Xt, Chuo-Ku, Osaka, Japan). The camera was equipped with a 185° fish-eye lens and was placed in a levelling device (Regent Instruments, Sainte-Foy, Quebec, Canada). Five pictures were taken per plot, and measurements were repeated five times during the study period. Images were analysed for leaf area index with Gap Light Analyzer Version 2.0 (GLA, Simon Fraser University, Burnaby, BC, Canada).

An inventory of all plots was conducted every year at the onset of the dry season, and individual tree height and DBH were measured as explained above. Above-ground biomass (kg) was calculated using species-specific allometric equations derived in 2006/07 by harvesting 10 trees per species per diversity treatment (Oelmann et al. 2010). Above-ground biomass increment for the subplots was estimated by subtracting the above-ground biomass calculated from inventory data for 2007 from the 2008 biomass. Water use efficiency of wood production [WUEwood, g dry matter (DM) kg−1 H2O] was calculated as the ratio between annual increment in above-ground biomass and the annual water use of all trees of a given plot.

Additive partitioning of biodiversity effects

We used the approach developed by Loreau & Hector (2001) to measure the net biodiversity effect (ΔY) on stand transpiration and basal area by additive partitioning of the selection vs. the complementarity effect. The net biodiversity effect has the dimension of yield, where yield stands for any measurable variable in an ecosystem (stand transpiration and basal area in this case) and is expected to be zero under the null hypothesis of no biodiversity effects. These various effects can be related by additive partitioning as follows:


where N is the number of species in the mixture, Mi is the yield of species i in monoculture, Y0,i is the observed yield of species i in mixture, Y0 = ∑iY0,i is the total observed yield of the mixture and YE = ∑iYE,i is the total expected yield of the mixture.

Accordingly, the complementarity and selection effects are expressed as a function of the deviation from the expected relative yield in the mixture (ΔRY). The yield in a mixture is influenced by a complementarity effect (inline image), if it is on average higher than the expected yield calculated from the weighted average of the monoculture yield. The covariance between the monoculture yield of species and the change in relative yield in the mixture measures the selection effect (Ncov(ΔRY,M)). The grand mean across all mixtures was tested against monocultures using a one-sample t-test.

Data analyses

Normalized daily Js was calculated by dividing daily integrated Js by the highest observed daily integrated Js during the study period (percentage of maximum). Gap filling for missing sap flux measurements (35 days in H. crepitans, all other species <14 days) was performed based on the species-specific relationships between measured sap flux densities and PPFD (with r2 values ranging between 0·67 and 0·83, < 0·05). We set the starting date of the dry season 3 weeks after the last major rain event, so that it lasted from 15 January 2008 to 15 April 2008 (90 days). The dry season was defined as the period when <100-mm rainfall was measured during 3 months (Dietrich et al. 1996). The periods from 1 July 2007 to 14 December 2007 and 1 June 2008 to 30 June 2008 were characterized as wet season (195 days). All annual values and means are based on 365 days.

Possible controls on tree water use were tested using linear regression analysis. Differences in sap flux density, basal area and stand transpiration among plots were assessed by applying analysis of variance (anova, followed by a post hoc Tukey HSD test). Tree water use rates among mixtures were compared using analysis of covariance with DBH as a covariate (ancova, followed by a post hoc Tukey HSD test). Differences in stand transpiration rates between seasons were determined with a paired Student’s t-test. All statistical analyses were performed using the SPSS 13.0 software (SPSS Inc., Chicago, IL, USA).


Sap flux density, seasonality and tree water use rates

Mean daily integrated sap flux densities (Jdaily) ranged from 72 to 202 g cm−2 day−1 (Table 1). No statistically significant difference in Jdaily was found between monocultures and mixtures for a given species (Table 1). The seasonal patterns of normalized Jdaily varied strongly among species and also among diversity levels for a given species (Fig. 2, left panels).

Table 1.  Structural characteristics of the study trees (mean values ± SD) and mean daily integrated sap flux densities (Jdaily) and tree water use rates (Q)
Species name and abbreviationNDBH (cm)Tree height (m)Crown area (m2) J daily (g cm−2day−1) Q (kg day−1)
  1. Small letters indicate significant differences between monocultures and mixtures for a given species.

Anacardium excelsum Aemonoculture mixtures410·10·66·40·410·63·8202a3510·9a3·4
Cedrela odorata Comonoculture mixtures412·00·611·71·19·44·673a245·8a1·7
Hura crepitans Hcmonoculture mixtures418·01·95·41·021·54·586a3012·4a5·5
Luehea seemannii Lsmonoculture mixtures411·81·68·71·011·81·7183a2912·5a2·4
Tabebuia rosea Trmonoculture mixtures411·51·37·40·318·64·693a307·5a1·9

Plotting weekly averages of normalized Jdaily in monocultures vs. mixtures revealed that differences between monocultures and mixtures were most obvious during the dry season (Fig. 2, right panels). This was most pronounced for T. rosea. Throughout the whole dry season (January–April 2008), T. rosea trees grown in mixed stands appeared to the left of the 1 : 1 line indicating higher normalized Jdaily compared with monocultures. Anacardium excelsum trees grown in mixtures sustained higher sap flux density rates towards the end of the dry season (April, May 2008) compared with monocultures. Luehea seemannii in mixtures had lower sap flux density rates at the onset of the dry season (January 2008) but outperformed monocultures towards the end of the dry season (March, April 2008).

Regression lines of crown projection area and diameter differed between monocultures and mixtures with trees in mixtures having a larger crown projection area than trees in monocultures (Fig. 3a). In contrast, no differences were found between monocultures and mixtures when considering conductive sapwood area and crown projection area (Fig. 3b).

Figure 3.

 Crown projection area in relation to (a) diameter at breast height (DBH) (monocultures: r2 = 0·655, < 0·001; mixtures: r2 = 0·715, < 0·001) and (b) conductive sap wood area (monocultures: r2 = 0·511 <0·001; mixtures: r2 = 0·687, < 0·001). Maximum daily water use in relation to (c) DBH (monocultures: r2 = 0·410, = 0·005; mixtures: r2 = 0·678, < 0·001) and (d) conductive sap wood area (monocultures: r2 = 0·601, < 0·001; mixtures: r2 = 0·638, <0·001). Functions derived for monocultures and mixtures are indicated by dotted and solid lines, respectively.

Maximum water use rates for individual trees were highly variable ranging from 18·9 kg day−1 for T. rosea (DBH 8·3 cm) to 101·1 kg day−1 for a H. crepitans (DBH 28·2 cm). Linear regressions of maximum water use rates of all 60 sample trees against DBH and conductive sapwood area explained 54% and 61% of the observed variation, respectively (Fig. 3c,d).

Stand transpiration and water use efficiency

All five-species mixtures consistently maintained high daily stand transpiration rates across both wet (2·64 ± 0·30 mm day−1, mean ± SD) and dry seasons (2·51 ± 0·21 mm day−1, mean ± SD), whereas monocultures and two-species mixtures containing C. odorata and L. seemannii had significantly lower transpiration rates during the dry season (Table 2).

Table 2.  Characteristics of study plots (mean values ± SD, n = 225 trees per plot) and stand transpiration rates (Tstand) for different periods. Values for basal area and annual stand transpiration are absolute values. Species contributions to stand transpiration are given on an annual basis. Water use efficiency (WUE) is reported as g dry mass (DM) per kg H2O
PlotSpecies numberDBH (cm)Tree height (m)LAI (m2 m−2)Basal area (m2 ha−1)Species contribution to basal area (%)Tstand (mm d−1)Tstand (mm yr−1)Species contribution to Tstand (%)WUEwood (g DM kg−1 H2O)
MeanSDMeanSDMeanSDTotalAeCoHcLsTrWet seasonDry seasonAnnualAeCoHcLsTrAnnual plot level
  1. Small letters indicate significant differences between wet and dry season for a given plot. Capital letters indicate significant differences among plots.

AE218·14·85·81·82·440·655·8A100    0·76a0·98b298AB100    3·6B
CO217·93·89·23·30·800·607·4A 100   0·79a0·36b220AB 100   4·2B
HC1;215·44·43·61·71·790·4611·4A  100  0·69a0·45b204AB  100  3·1B
LS1116·37·79·31·73·220·6032·4A   100 2·62a2·74a911AB   100 3·8B
TR2110·45·16·21·71·280·5310·3A    1001·06a1·16a360AB    1004·1B
T1211·47·16·93·81·910·466·7A 5842  0·55a0·20b152A 6832  6·0C
T2313·28·69·23·12·740·5822·3A28  60122·15a3·04b862BC29  58134·0B
T3311·88·47·63·42·900·4816·8A247 69 1·50a1·94b581BC278 64 2·8B
T4312·87·39·34·52·571·1319·4A 41950 1·88a1·68a629BC 30861 3·5B
T5210·55·06·72·02·680·548·4A  51 500·66a0·74a242A  40 606·3C
T6210·06·35·01·61·640·597·4A46   540·74a1·17b316A59   414·9C

Annual stand transpiration ranged from 204 (H. crepitans) to 911 mm year−1 (L. seemannii) in the monocultures. In mixtures, annual stand transpiration varied between 152 and 945 mm year−1 (Table 2). Annual stand transpiration rates differed significantly among diversity levels with monocultures and two-species plots having significantly lower transpiration rates than three- and five-species plots.

Though not statistically significant, average basal area tended to be highest in three-species plots. We found a positive relationship between annual stand transpiration and basal area (Fig. 4a). A positive relationship between annual water use per plot and annual increment of above-ground biomass was also present (Fig. 4b).

Figure 4.

 (a) Annual stand transpiration in relation to basal area (r2 = 0·629, <0·001) and (b) annual water use per plot in relation to annual increment of above-ground biomass (r2 = 0·431, = 0·003).

Trees growing in five-species mixtures transpired a significantly higher amount of water to produce the same amount of biomass, resulting in a lower water use efficiency (WUEwood = 2·1 g DM kg−1 H2O) in comparison to monocultures (WUEwood = 3·7 g DM kg−1 H2O), two-species mixtures (WUEwood = 5·7 g DM kg−1 H2O) and three-species mixtures (WUEwood 3·4 g DM kg−1 H2O) (Table 2).

Net, selection and complementarity effect

The additive partitioning method revealed a positive, significant net effect of species richness on annual stand transpiration and basal area (Fig. 5a,b). The two components of the net effect, selection and complementarity had a different pattern for stand transpiration and basal area. No selection effect was found for stand transpiration (the grand mean was not significantly different from monocultures). In contrast, the selection effect was significant for basal area but did not increase with species richness (Fig. 5c,d). The grand means of stand transpiration and basal area were significantly different from the monoculture mean indicating a significantly positive complementarity effect (Fig. 5e,f). The complementarity effect increased significantly with species richness for basal area and especially for stand transpiration.

Figure 5.

 Additive partitioning of biodiversity effects on annual stand transpiration and basal area. (a, b) Net biodiversity effect; (c, d) selection effect; and (e, f) complementarity effect as a function of species richness (single asterisk, < 0·05; double asterisk, < 0·01; triple asterisk, < 0·001; ns: non-significant).


Tree diversity effects on basal area and stand transpiration

We found a significantly positive complementarity effect of tree diversity on basal area. Complementarity is often cited as one the fundamental principles to explain the higher productivity of mixed stands compared with single species stands (e.g. Haggar & Ewel 1997). Further, we found a significant selection effect upon basal area indicating the dominance of one species in the mixtures characterized by a high basal area. This observation is not unexpected as species with different growth rates were selected for this experiment (Huston 1997).

Forest water use often increases with productivity (Law et al. 2002), and thus, the average 14% higher stand transpiration in two- and three-species plots compared with the monocultures at Sardinilla can be explained by a 27% higher biomass increment. Forrester et al. (2010) also found a significantly higher annual biomass increment and annual transpiration in a mixed 15-year-old Eucalyptus globulus and Acacia mearnsii plantation compared with E. globulus and A. mearnsii grown in monocultures. In contrast, the 56% higher stand transpiration in the five-species mixtures compared with monocultures at Sardinilla cannot be explained by higher basal area alone. The dominance of a single tree species does not account for the high transpiration, either, as no species contributed a proportionally higher water use to the total amount of transpiration (Table 2). Further, the additive partitioning approach did not reveal a selection effect. In contrast, additive partitioning suggests that the enhanced stand transpiration was because of a significantly positive complementarity effect. The strong complementarity effect in the five-species mixtures suggests lower competition for resources. Potvin & Gotelli (2008) explained the enhanced growth in 5-year-old mixed-species plots at Sardinilla as a consequence of reduced competition for light. Further, light interception increases in mixtures because of canopy stratification and the selection of trees with different crown structures (Menalled, Kelty & Ewel 1998; Laclau et al. 2008). Although not statistically significant, trees growing in three- and five-species mixtures had larger crown projection areas and larger leaf area indices compared with monocultures and two-species mixtures. Larger crown area, foliage cover and/or LAI lead to increased tree water use and stand transpiration (Motzer et al. 2005; Kunert, Schwendenmann & Hölscher 2010). Furthermore, the observed higher heterogeneity in tree height and also mean tree height may be an indicator of higher canopy roughness (Kelliher, Leuning & Schulze 1993) leading to higher stand transpiration rates (Baldocchi 2005). Therefore, the aerodynamic roughness length of the tree canopy in mixed stands is expected to be larger, which in turn increases the degree of turbulent energy exchange with the atmosphere, resulting in higher transpiration rates. This may partly explain the proportionally higher stand transpiration rates in the five-species plots.

Because of the strong seasonality at Sardinilla, major resources such as light and water are scarce to different degrees throughout the year. With more than 200 mm of precipitation per month during the wet season, soil water content at all depths remained high (>0·45 m3 m−3) and was not considered to have any effect on tree growth and transpiration (given that trees were fully foliated). However, soil water content decreased considerably during the dry season, whereas PPFD and VPD were highest during the 3-month dry period. Anacardium excelsum, which maintained a high leaf cover throughout the dry season, had significantly higher stand transpirations rates during the dry season. Higher tree transpiration rates during the dry season were also reported for an open eucalypt forest in Northern Australia (O’Grady, Eamus & Hutley 1999). This was explained by an increased evaporative demand and no water limitation because of access to groundwater, which may also apply for the trees maintaining a high leaf cover during the dry season in our study.

Significantly lower daily transpiration rates during the dry season were measured for C. odorata, H. crepitans monocultures. Both species were characterized by a comparatively low leaf cover during the dry season (Kunert, Schwendenmann & Hölscher 2010). In contrast, no differences in daily stand transpiration rates between dry and wet seasons were found for five-species mixtures, indicating that transpiration was maintained at a high level throughout the wet and dry seasons. This can be explained by the wide range of species-specific traits such as leaf phenology, which act to balance the reduction in transpiration of deciduous species by enhanced transpiration of species sustaining their leaves during the dry season. Using stable isotope techniques (δ2H) showed that tree species at Sardinilla differed in their soil water uptake pattern. During the dry season, C. odorata and H. crepitans took up most of their water from the top 30 cm, whereas T. rosea, A. excelsum and L. seemannii used soil water from deeper layers (>30 cm) (L. Schwendenmann & R. Sánchez-Bragado, unpublished data). Access to water from different depths in the soil profile makes available a larger total volume of soil water, which in turn may explain the higher leaf cover and enhanced transpiration in mixed-species plots during the dry season.

Tree diversity effect on water use efficiency

The WUEwood values estimated at Sardinilla are within the range of values (3·3–9·4 g DM kg−1 H2O) reported from a 15-year mixed and monospecific acacia–eucalypt plantation in Australia (Forrester et al. 2010). The same study revealed that the acacia–eucalypt mixtures had higher WUE than monocultures because of a significant increase in the WUE of eucalypts in the mixed stand (Forrester et al. 2010). At Sardinilla, the average WUEwood of two-species mixtures was significantly higher compared with all other diversity levels, which is in accordance with the findings of Forrester et al. (2010). It should be noted that the high WUEwood in the two-species plots might also be an artefact of the die-off of C. alliodora, which resulted in a 30% reduction in biomass of these plots. In contrast to the results of Forrester et al. (2010), the five-species plots had a much lower WUEwood compared with the monocultures, which can be explained by the considerably higher transpiration rates; still, the reasons for the lack of a proportional increase in productivity are discussed below.

There are four factors that potentially limit productivity in the five-species plots: silvicultural treatments, water availability, light conditions and nutrients. Silvicultural treatments were consistent amongst plots, and water was most likely not limiting (as explained above); thus, these two factors can be omitted for the comparison between monocultures and five-species mixtures.

The third factor, light conditions were seasonally variable because of cloud cover during the wet season, indicating that the influence of diffuse solar radiation may have a role in differences between WUEwood of monocultures and mixtures. During the dry season, increased evaporative demand enhances transpiration rates (O’Grady, Eamus & Hutley 1999). However, in the five-species plots, the more complex crown structure of the mixtures may lead to more self-shading, thereby preventing an associated increase in productivity during this time. Furthermore, light use efficiencies tend to be lower in direct solar radiation in comparison to diffuse solar radiation (Alton, North & Los 2007), and thus, the dry season was the period of highest water use but also potentially the period of lowest productivity, resulting in lower WUEwood at the annual time-scale. Unfortunately, this assumption of lower productivity during the dry season could not be confirmed by biomass measurements because inventory was only carried out at an annual intervals.

The fourth factor that governs productivity is nutrient availability. Limitation by N and/or P as well as species- and diversity-related differences in N and P nutrient use efficiency may limit increases in productivity in the five-species mixtures (Richards et al. 2010). Hura crepitans and T. rosea had a lower N- and P-nutrient use efficiency compared with A. excelsum and L. seemannii (Zeugin et al. 2010). Further, tree species at Sardinilla tended to have lower P use efficiency in two- and three-species mixtures compared with monocultures and five-species mixtures (Zeugin et al. 2010). Regardless of the mechanism of limitation of productivity in five-species mixtures that changed the proportionality between water use and growth, mixed stands are clearly not able to make the most efficient use of water resources. Low resource-use efficiency is also indicated by highly diverse old-growth stands in the Amazon forest respiring 70% of the assimilated carbon (Chambers et al. 2004).

Management considerations and conclusions

In tropical tree plantations, site-specific management plans need to define the goal of the enterprise (e.g. high yields, land cover protection, water resource restoration or a combination thereof). Our study showed that water use in mixtures was enhanced, and stand transpiration was highest in the five-species plots while WUEwood was lowest in the species-richest plots. Enhanced water use may not be problematic when water is plentiful (e.g. in non-seasonal climates or during the wet season). However, managers need to be aware of the potential impact of higher transpiration rates on water resources in drier climates or during periods of limited water availability. We recommend the establishment of species mixtures containing a low number of species. Such plantations should achieve high growth rates through complementary resource use while keeping water use rates at a modest level.


We thank Sebastian Wolf for providing the micrometeorological data and Jose Monteza for collecting data on tree growth. We thank Klaus Winter for advice and Milton Garcia for technical support. We would also like to thank Cate Macinnis-Ng and three reviewers for their valuable comments on an earlier version of the manuscript. This study benefited from the logistical support provided by Smithsonian Tropical Research Institute and Panama’s National Authority for the Environment (ANAM) and was funded by the German Research Foundation (DFG, Ho-2119/3) and a Natural Sciences and Engineering Research Council of Canada (NSERC) operating grant.