Results
We measured a total of 147 pairs of cooling curves (49 in OLEU, 50 in MAGR, 15 in PLOC, 15 in SOLY and 18 in POTO) for bald and hairy leaf patches. Figure 2 shows a typical experiment for an OLEU leaf. In that experiment, after transient radiative heating, the temperature of each patch declined from approximately 30 K above air temperature to a final value just below air temperature, in each case closely following an exponential model (Eqn 3) fitted to the data. In this example, the bald patch was initially warmed to a slightly greater temperature than the hairy patch, presumably due to the greater albedo contributed by the hairs (albedo was not measured); thus, the bald patch cooled more rapidly in absolute terms (i.e., with a greater rate of change of temperature). Yet, the intrinsic kinetics of cooling were also faster in the bald patch – that is, the time constant for cooling was shorter (τ = 9.79 s vs 10.94 s for the hairy patch); indeed, despite having begun at a higher temperature, after about 10 seconds the temperature of the bald patch dropped below that of the hairy patch. In 141 of 147 experiments (96%), τ was smaller (cooling was faster) in the bald patch than in the hairy patch (Fig 3). The mean ratio ofτ between bald and hairy patches within species ranged from 0.578 ± 0.050 (mean ± SE; SOLY) to 0.957 ± 0.007 (MAGR).
We inferred leaf boundary layer conductance to heat (g bh) by applying measured values of τ , air temperature, leaf surface conductance to water vapor (g sw), and leaf heat capacity (k ) to Eqn 4. The relationship between g bh and τ was fairly consistent within most species (Fig 4), because the two factors that could create divergence in that relationship –g sw and k – were fairly similar among leaves for a given species (Tables 2 & 3) (though one PLOC leaf had a much lower water content, and hence lower heat capacity, than the other leaves of that species; that leaf corresponds to the group of blue symbols closest to the origin in Fig 4). Because the leaves had been acclimated in darkness for at least an hour before measurement,g sw (measured by porometry before each cooling experiment) was generally low (Table 3) and did not increase after trichome removal by shaving for the two species subjected to this treatment (OLEU and MAGR) (Table 3). For the two species for which we compared different leaves with and without hairs (SOLY and POTO),g sw was marginally greater in hairy leaves than in bald leaves (Table 3). In the fifth species, PLOC,g sw was on average greater in bald than in hairy patches (0.072 ± 0.005 mol m-2 s-1[bald] vs 0.043 ± 0.012 mol m-2s-1 [hairy]). Across all species and experiments,g sw was slightly but significantly lower in bald patches than in hairy patches (p < 0.0001; linear mixed model of g sw with species and hair condition as fixed effects and leaf as a random effect); Tukey’s post hoc tests also found this result also held within each individual species except PLOC (p < 0.0001). Our estimation ofg bh from τ accounted for observed differences in g sw between patches (discussed below), and our finding of faster cooling in bald patches could thus not be attributed to g sw differences.
Boundary layer conductance varied among leaves and experiments, from 2.33 mol m-2 s-1 (for a hairy patch in a POTO leaf) to 8.12 mol m-2 s-1(for a bald patch in a PLOC leaf) (Fig 5). Boundary layer conductance was greater in the bald patch than in the hairy patch in nearly all experiments (138 of 147 cases; 94%) (Fig 5). Excluding a single outlier leaf of POTO (the three green points farthest from the 1:1 line in Fig 5), the increase of boundary layer resistance (decrease ing bh) due to trichomes ranged from 2.4 ± 1.4% (mean ± SE; PLOC) to 38.7 ± 6.1% (POTO) (Fig 6).
Hair layer depth varied 6-fold across species (Table 2). The predicted resistance contributed by hairs (r h, calculated from hair layer depth by assuming hairs entrained a layer of still air equal in thickness to the hair depth) was unrelated to the value ofr h estimated from experimentally observed differences in g bh between bald and hairy patches (Fig 7). The latter estimates, however, depended sensitively on the assumed ratio (β ) of boundary layer resistance between the adaxial and abaxial surfaces in the absence of hairs (Fig S4); the values shown in Fig 7 assumed a nominal value of β = 1.
Our modelling predicted that reduction of g bh by leaf hairs in most cases reduced gas exchange rates, though to a greater extent for transpiration than for CO2 assimilation (cf. Figs 8a,b). In most cases, trichomes caused predicted declines of approximately 0.25 – 0.50% in assimilation rate, but 1 – 2% in transpiration rate, leading to increases of 0 – 1% in WUE. The main exception was at low temperature, when hairs were predicted to increase gas exchange rates and decrease WUE. The effect of trichomes on WUE was predicted to be greater (1) when evaporative demand (VPD) is high, whether due to low vapor pressure or high air temperature, (2) when stomatal conductance is large compared to boundary layer conductance, (3) when PPFD is subsaturating, and (4) in small leaves, provided air temperature is high (Fig 8).