Leaf temperatures in glasshouses and open-top chambers

Authors

  • Hans J. De Boeck,

    1. Research Group of Plant and Vegetation Ecology, Department of Biology, Universiteit Antwerpen (Campus Drie Eiken), Universiteitsplein 1, B-2610 Wilrijk, Belgium
    Search for more papers by this author
  • Toon De Groote,

    1. Research Group of Plant and Vegetation Ecology, Department of Biology, Universiteit Antwerpen (Campus Drie Eiken), Universiteitsplein 1, B-2610 Wilrijk, Belgium
    Search for more papers by this author
  • Ivan Nijs

    1. Research Group of Plant and Vegetation Ecology, Department of Biology, Universiteit Antwerpen (Campus Drie Eiken), Universiteitsplein 1, B-2610 Wilrijk, Belgium
    Search for more papers by this author

Author for correspondence:
Hans J. De Boeck
Tel: +32 3 265 22 82
Email: hans.deboeck@ua.ac.be

Summary

  • Climate manipulation experiments are of key importance in identifying possible responses of plant communities and ecosystems to climate change. Experiments for warming the air under sunlit conditions are carried out in (partial) enclosures. These inevitably alter the energy balance inside, potentially altering tissue temperatures which affect metabolism and growth.
  • Using an empirically validated energy balance model, we investigate effects of two widely used warming methods, climate-controlled glasshouses and passively warmed open-top chambers (OTCs), on leaf temperatures. The model applies standard energy balance formulas, supplemented with data on optical properties of glasshouse materials and wind conditions inside OTCs.
  • Results show that the different radiation environment inside glasshouses did not produce large leaf temperature deviations compared with outside. Poor glasshouse design with significant radiation blockage by the structure or with insufficient ventilation did affect tissue temperatures more significantly. The drastic wind speed reduction inside OTCs approximately doubled the actual (canopy) warming compared with earlier reported increases in air temperature provided by this technique – an effect that was inflated if the plants’ stomates closed.
  • These results demonstrate that leaf temperatures were higher than previously considered in OTCs but not in climate-controlled glasshouses.

Introduction

In the context of a warming world (IPCC, 2007), the researchers studying effects of higher temperatures on plant and ecosystem responses have been steadily gaining ground. Methods of experimental warming are varied, including exclusive heating of the soil (Bergh & Linder, 1999), growth chambers with an artificial light environment (Swindell et al., 2007), infrared heating of canopy and soil (Marchand et al., 2004), climate-controlled glasshouses (Gielen et al., 2005), and passive warming in open-top chambers (OTCs; Henry & Molau, 1997). Comparison of the different techniques and their drawbacks has been, and continues to be, done elsewhere (Beier et al., 2004; Aronson & McNulty, 2009).

In this study, we attempt to look into the effects of two of these warming techniques, namely climate-controlled glasshouses and passively warmed OTCs, on the leaf temperature of plants. Indeed, while most researchers only consider air temperature, it is the tissue temperature that is of fundamental importance for the metabolic processes within the plant. It is well known that species growing in cold environments have adapted to be highly efficient at trapping heat, so that their leaves are significantly warmer than the surrounding (suboptimally cold) air (Körner, 2003). In hot climates, plants have adapted to minimize the warming of their leaves, for example by making them very small (to ensure temperatures close to that of the air) or by tilting the leaves to an angle that traps less radiation at midday (to avoid excess heating by the sun). Given the importance of canopy temperatures, we believe that the many warming studies that report exclusively on air temperatures may in some cases under- or overestimate the degree of warming. This was already suggested in the infrared imagery study by Scherrer & Körner (2010), in which they demonstrated that mountain plants can be nearly 10°C cooler or warmer than the surrounding air, implying a strong decoupling from atmospheric conditions. The authors therefore claim that simple projections of local or regional species losses based on rising air temperatures are misleading (Scherrer & Körner, 2010).

In this study, we specifically look into the effects of glasshouses and open-top chambers on leaf temperatures. In glasshouses, the radiative environment is different from outside. The temperature of clear skies is often well below 0°C (Nobel, 2005), while the ‘sky’ inside the glasshouse consists of the cover materials which, in most circumstances, will be warmer than the outside sky. Sky temperatures determine the downward longwave radiation and therefore directly affect the energy balance and thus canopy temperatures. Other properties of the glasshouse such as the total light transmission and the reflectance of longwave radiation could also affect the leaf temperatures. In OTCs, a substantial proportion of the sky is not artificial, while the absence of a roof permits free convective heat exchange with the outside (the prevention of which is the most important warming mechanism in closed chambers). The side panels of the OTC inevitably affect wind speed, which in turn influences energy exchange as calmer conditions reduce heat dissipation. No journal study has formerly quantified wind speeds inside and outside OTCs, such as the standard hexagonal chambers used in the International Tundra Experiment (ITEX), although this subject has been reported upon in a PhD thesis (Dalen, 2004). These measurements, made both in a forest and on a mountaintop, demonstrate that wind speeds inside OTCs were reduced markedly.

If glasshouses and OTCs do indeed distort leaf temperatures significantly, this would imply that the amount, and possibly the variability, of true (canopy) heating achieved differs from previous estimates of attained warming. This in turn suggests that extrapolations from glasshouse and OTC studies may have to be reconsidered. Specifically, we hypothesize that the important reductions in wind speed inside OTCs would cause substantial canopy warming on top of the raised air temperatures. In the case of actively regulated glasshouses, a warming of the canopy caused by higher sky temperatures may be offset or reversed by blocking of incoming radiation by the glasshouse structure and the altered radiation environment within. Any discrepancies between leaf temperatures inside and outside the glasshouses could be buffered or mitigated by an altered heating control. To investigate the possible effects of glasshouses and OTCs on leaf temperatures, we use an energy balance model that was validated empirically. We then discuss the implications for ecosystem warming experimentation.

Materials and Methods

We opted for a straightforward model that calculates leaf temperatures based on the environmental conditions and robust physical relationships. Although actual leaf temperatures may deviate slightly as approximations and biological variability are inevitable, the model gives a clear indication of existing trends. The energy balance equation is central in resolving the leaf temperature.

image(Eqn 1)

This basic formula states that the sum of incoming energy (shortwave radiation Rs,in and longwave radiation Rl,in) and outgoing energy (outgoing longwave radiation Rl,out, and latent heat λE) has to be zero for the leaf to be energetically at equilibrium under specific environmental conditions (the flux of sensible heat H can be either incoming or outgoing). The incoming shortwave radiation is calculated from

image(Eqn 2)

  This formula quantifies how much shortwave energy is absorbed by the leaf, based on the total shortwave energy (Rs, W m−2), the shortwave absorbtivity of the leaf (αs,leaf) and the shortwave reflectivity from the soil (ρs,soil). The two parts of the equation represent the two sides of the leaf and therefore need to be divided by two in order to account for the leaf’s view factors (see Campbell & Norman, 1998). From the same authors, we assume αs,leaf to be 0.5, while ρs,soil (the albedo) is set at 0.1. The radiation is assumed to be perpendicular to the leaf. The incoming longwave radiation is

image(Eqn 3)

Again, the two parts of the equation represent incident radiation on either side of the leaf: from above, for which we need the longwave sky emissivity (εl,sky), the air temperature (Tair, °C), the Stefan Boltzman constant (σ = 5.67e−8 W m−2 K−4) and the longwave absorbtivity of the leaf (αl,leaf = 0.97); from below, for which we need σ, αl,leaf, the longwave soil emissivity (εl,soil = 0.945) and the soil temperature (Tsoil). Furthermore, εl,sky is derived from

image(Eqn 4)

With the cloud cover ranging from 0 (bright) to 1 (overcast), and

image(Eqn 5)

where ξ is the water vapour path length

image(Eqn 6)

and ea is the vapour pressure (Pa),

image(Eqn 7)

where RH is the relative humidity.

The outgoing longwave radiation depends on the longwave emissivity of the leaf (εl,leaf = 0.97), σ and the leaf temperature (Tleaf, °C):

image(Eqn 8)

The sensible heat loss or gain of the leaves (Eqn 1) is dependent on the conductivity for heat (gHa, mol m−2 s−1), the specific heat content of air (cp, J mol−1 °C−1) and the temperature gradient between leaf and air according to

image(Eqn 9)

The calculation of the conductivity for heat depends on the prevailing conditions which control whether free or forced convection is dominant. This can be derived from the Reynolds (Re) and the Grashof (Gr) numbers, with

image(Eqn 10)

and

image(Eqn 11)

where u is the wind speed (m s−1), d is the characteristic dimension of the leaf (m), ν is the kinematic viscosity (9.08e−8 Tair + 1.3267e−5, m2 s−1), and g is the gravitational constant (9.81 m s−2). The calculation schedule for gHa can be seen in Fig. 1.

Figure 1.

A flowchart to determine which equations need to be used for the calculation of abaxial (ab) and adaxial (ad) boundary layer conductance for heat (gHa) and water (gvb), based on the Reynolds (Re) and Grashof (Gr) numbers and temperatures of leaves (Tleaf) and air (Tair). Formulas furthermore feature wind speed (u) and leaf dimension (d).

Finally, the latent heat loss can be calculated from

image(Eqn 12)

where λ is the latent heat of vaporization (−42.575 Tair + 44 994 J mol−1), gv is the total conductivity to water vapour (mol m−2 s−1), es (Tleaf) is the saturation vapour pressure at leaf temperature (Pa) and pa is the atmospheric pressure (Pa). Here, gv is calculated from the abaxial (gvs,ab) and adaxial (gvs,ad) stomatal conductances to water vapour and the boundary layer conductance to water vapour (gvb) via the pathway described in Fig. 1.

The leaf temperature is calculated in an iterative manner: upon running the model, it will assume that Tleaf = Tair. If this is the case, the energy budget will equal zero. If this is not the case, then the model will assume Tleaf to be lower than Tair if the energy budget is negative, and it will assume Tleaf to be higher than Tair if the energy budget is positive. The iteration proceeds in a stepwise manner, until a precision of 0.01°C is achieved.

We believe this method of calculation is a fair approximation of conditions inside OTCs. The fact that there is some light attenuation would yield only limited cooling effects (see results for glasshouses) which could be counteracted or reversed by small warming effects caused by reflection of outgoing longwave radiation. In climate-controlled glasshouses, the net effect of the altered radiative environment on leaf temperatures is studied by adapting Eqns 2 and 3 into Eqns 13 and 14, respectively:

image(Eqn 13)

where τs,gh is the shortwave transmissivity of the glasshouse, because only a portion of the outside incoming shortwave radiation is transmitted into the glasshouse.

image(Eqn 14)

where τl,gh, εl,gh and ρl,gh are, respectively, the longwave transmissivity, emissivity and reflectivity of the glasshouse material and Tgh is the glasshouse temperature (°C), which is assumed to be equal to Ta. The two new terms in the equation account for the incoming longwave radiation emitted by the glasshouse material and the portion of the longwave outgoing radiation of the leaf that is reflected back into the glasshouse.

The choice of material greatly determines transmissivity, emissivity and reflectivity, and hence we opted to test three different cover materials: glass (4 mm), polyvinyl chloride (PVC, 0.6 mm) and low-density polyethylene (LDPE, 0.15 mm), the optical properties of which were derived from Harvie (1996) and Papadakis et al. (2000). Note that these properties are for radiation perpendicular to the surface. Unless the radiation is under very shallow angles, the optical properties are robust (Harvie, 1996). An overview of energy fluxes in the three different glasshouses can be seen in Fig. 2. Furthermore, we assumed that the glasshouse’s structural components would block a proportion of the incident radiation, here 5%. The inputs are therefore Tair (°C), Rs (W m−2), RH, u (m s−1), d (m), gvs,ad (mol m−2 s−1), gvs,ab (mol m−2 s−1), cloud cover and cover material (OTCs are considered to have no cover; see earlier discussion).

Figure 2.

The radiation diagrams for glasshouses coated with glass, PVC and low-density polyethylene (LDPE). The longwave (l) and shortwave (s) radiation is shown with arrows. The values for emissivity (ε), transmissivity (τ), reflectivity (ρ) and absorptivity (α) for the longwave and shortwave radiation used in the model are depicted where they affect the radiation.

Open-top chambers can affect many environmental conditions (Kimball et al., 1997), but the parameter most affected in OTCs is the wind speed: data for wind profiles inside and outside OTCs in a forest and on a mountaintop location gathered for 6 d each (Dalen, 2004) reveal that the wind speed at 20 cm height (OTC height was 35 cm) was reduced 3.3-fold on average in the forest (0.16 vs 0.53 m s−1), and 8.4-fold on the mountain top (0.23 vs 1.92 m s−1). We use an average of these two ratios for our OTC simulations, with wind speed assumed as the only variable that differs significantly between OTCs and natural, outside conditions (see earlier discussion). The sensitivity of leaf temperatures to wind speed is shown in Table 1. Note that in the wind database recorded by Dalen (2004), wind speeds at 20 cm height inside OTCs never surpassed 0.35 m s−1. Compared with the inputs needed for calculation of leaf temperatures inside OTCs, the choice of cover material is the only other input that needs to be defined when running the model for climate-controlled glasshouses, although we also perform a run in which the wind speed inside is reduced. For the glasshouse runs, we model daily (24 h) courses of leaf temperature based on data from a meteorological database in Brasschaat (part of Fluxnet, Gielen et al., 2010). We pick one bright day per season (DOY 18, 102, 200, 283) that is close to the long-term average for air temperature (mean, minimum and maximum). Cloudy conditions are simulated by replacing the radiation data from the bright days with radiation from overcast days that occurred around the same dates. This allows us to compare the effects of radiation on leaf temperatures without changes in other variables. The half-hourly values of air temperature, shortwave radiation and relative humidity from the Brasschaat database are supplemented with leaf dimensions (0.02 m), wind speed (the average for that season in Belgium, Royal Meteorological Institute of Belgium), cloud cover (0 or 1 as we simulate fully bright and fully overcast days) and stomatal conductance. This conductance is set to vary between 0.05 mol m−2 s−1 at night and 0.3 mol m−2 s−1 during the day, with a 3 h linear-transition phase in the morning and the evening, the timing of which depends on sunset and sunrise, and thus on the season. This is an approximation based on Körner (1994), and although true conductance courses are probably less linear (Nijs et al., 1997), this simplified approach is adequate to observe the major trends. For the simulation of summer drought, the maximum gvs was decreased to 0.2 mol m−2 s−1 and was maintained for merely 4 h in the morning. It should be noted that the model does not account for any longer-term plant responses to changing temperatures which could, in turn, affect tissue temperatures. These include effects of growth changes in response to altered nutrient mineralization (via soil temperature and moisture), stomatal acclimation, changes in species composition, etc. (e.g. Beerling, 1999; Rinnan et al., 2007).

Table 1.   Modelled effects of wind speed (u) on leaf temperature (Tleaf)
u (m s−1)Tleaf (°C)
  1. Air and soil temperature (10°C), radiation (750 W m−2), relative humidity (60%), leaf width (0.02 m), stomatal conductance (0.3 mol m−2 s−1) and cloud cover (0%) were all kept constant.

0.115.16
0.214.03
0.512.84
112.15
1.511.81
211.60
2.511.46
311.35
411.19
511.07
610.99
710.92
810.87
910.82
1010.78

In order to test whether our energy balance model is sufficiently accurate in its predictions of leaf temperature, we took a series of measurements under outside conditions, in open-top chambers and in climate-controlled glasshouses. Data were gathered throughout November 2011, which was considerably sunnier (115 vs 66 h of sunshine) and warmer (8.6 vs 6.8°C) than an average November month in Belgium (data from the Belgian Royal Meteorological Institute), meaning grasses had not yet entered winter dormancy. We installed a small, hexagonal OTC (100 cm base diameter, 35 cm tall) with 4-mm-thick Lexan (General Electric, Fairfield, CT, USA) side panels and manufactured according to ITEX standards on a homogeneous grass lawn (dominated by Lolium perenne). Measurements on nonenclosed plants were recorded on the same lawn. Finally, to test conditions in glasshouses, we installed a tray (0.5 × 1 m and 5 cm deep) filled with Deschampsia cespitosa grass plants of equal age (c. 3 months old) inside a small glasshouse. Details on the glasshouse setup can be found in Van den Berge et al. (2011). The material of the glasshouse walls (polyethylene, 0.2 mm) and roof (PVC plate, 4 mm) differed from the three types discussed here. In our model runs, simulating it as glass, PVC or LDPE revealed little differences. All required meteorological and plant data were gathered on the spot.

Air temperature, RH and average 1 min wind speed at canopy height were recorded with a pocket weather meter (Kestrel 3000, Nielsen Kellerman, PA, USA). Outside wind speed data were divided by 5.8 (the average used in the model; see earlier) as a proxy of wind speed inside the OTCs, since air movement was very limited inside the chambers and the measurement device was unable to record such very low wind speeds because of its dependence on moving parts. Stomatal conductance was recorded with a porometer (AP4, Delta-T Devices Ltd, Burwell, UK), and leaf width was set at 3 mm. Surface temperature was measured with an infrared camera (FLIR i3, FLIR Systems Inc., Wilsonville, OR, USA). For each surface temperature data point in time, we took the average of six separate spot measurements made perpendicular to the surface. Cloud cover was estimated visually, always by the same person. Air pressure was set at 101.2 kPa, which was the average recorded during this month, dominated by anticyclonical conditions (Royal Meteorological Institute). Soil temperature was set as 5°C whenever Tair was < 5°C, 10°C when 5°C < Tair <15°C and 15°C whenever Tair was > 15°C.

Radiation was recorded with a pyranometer (LI-200SA, Li-Cor, Lincoln, NE, USA) perpendicular to the (horizontal) soil. We corrected the radiation measured during sunny moments for the difference in measurement angle between the radiation sensor (Shorizontal) and the leaves (Sleaf). The grass leaves were found to be at an angle of 77° (average of 78 measurements) to the surface, while the solar altitude at 51°N was 20° on average during our measurements. This implies a mean angle of 70° for incoming radiation on the leaves, whereas it was 20° on the pyranometer. We correct for this by calculating the difference in radiation loads between these two angles, with the formulas

image(Eqn 15)

where α is the solar altitude (20°) and β is the average leaf angle towards the 20° sun (90°, or 20° + 70°). The radiation load on sun-exposed leaves is thus, on average, 177% higher than on the pyranometer, and we adjust the recorded radiation values on sunny moments likewise. Next, we have to correct for the fact that the surface temperatures were measured perpendicular to the surface, meaning a higher fraction of shaded vegetation in the infrared camera’s view (measuring in an angle of 20° from the direction of the sun was impossible because of the small size of the OTCs). The shaded surface apparent with a 90° view angle was calculated as the fraction of black pixels in an image of a patch within our OTC. The original colour image (made with a Nikon D80 camera) was therefore bimodally transformed to a pure black-and-white image using the ‘threshold’ tool in GIMP 2.6 (freeware, General Public License), with a cut-off of 60/255, yielding 22.2% black and 77.8% white pixels (Supporting Information, Fig. S1). The angle-corrected radiation value was therefore multiplied by 0.78, as only this fraction of the incident radiation directly hit the leaves.

Measurements inside OTCs and under outside conditions were carried out in pairs, which enabled us to directly compare the values. This was not possible in the glasshouses as the cooling system was temporarily out of order. Data were gathered between 09:00  and 20:00 h, under a range of conditions and temperatures: Tair ranged between 0.8 and 20.0°C (average (± SD) = 10.6 ± 4.3°C, = 88), RH between 0.53 and 1.0 (average = 0.82 ± 0.12, = 88), wind speed at canopy height between 0.03 and 1.0 m s−1 (average = 0.29 ± 0.27 m s−1, = 88), and horizontal plane radiation between 0 and 370 W m−2 (average = 79 ± 111 W m−2, = 88).

Results

Model validation

The ensemble of all 88 Tleaf values (from 528 measurements) compared with the temperatures modelled had a Pearson correlation coefficient of 0.98, and the linear regression (R2 = 0.97) deviated only marginally from the 1 : 1 line (slope 0.99, intercept 0.19). Also when assessing the three ‘locations’ (outside, OTC, glasshouse) separately, the model results seem to match the measured data well (Fig. 3a), especially in the temperature range most relevant in this study (5–20°C). Because these correlations can be (in part) driven by variation in air temperature, we also compare the signal relative to Tair (Fig 3b). Correlation coefficients were 0.81, 0.88 and 0.42 for outside, OTC and glasshouse, respectively and 0.76 for all data pooled together. The lower value for glasshouses partly reflects the narrow range on the x-axis (merely 1.5°C). Overall, the difference between modelled and measured Tleaf values was smaller than 1°C in 77% and smaller than 1.5°C in 91% of cases (with a maximum deviation of 2.7°C).

Figure 3.

Comparison of measured and modelled canopy temperatures. Data were recorded in homogeneous grass patches: nonenclosed, inside a small open-top chamber (OTC) and inside a glasshouse (results for PVC cover shown). See text for details.

Taking paired readings inside OTCs and outside enabled us also to directly compare the results. On clear days (< 50% cloud cover), air temperature was 0.4 ± 0.2°C and leaf temperature 1.6 ± 0.2°C higher inside than outside OTCs (= 15). On cloudy days, air inside OTCs was 0.3 ± 0.1°C warmer than those outside, while the leaf temperature difference was approx. zero (0.0 ± 0.2, = 14).

Climate-controlled glasshouses

The differences between the cover materials are modest, especially under cloudy conditions, with leaf temperatures underneath the LDPE foil generally being closest to those outside (Table 2). On average, only small deviations of leaf temperatures inside and outside glasshouses were found. On clear days, leaves inside glasshouses were slightly warmer than those outside (< 0.5°C), especially in autumn and winter. A summer drought did not affect this inside–outside difference. On cloudy days, the leaf temperature inside glasshouses was marginally lower than that outside (< 0.2°C), with the biggest differences found in the summer. We also ran a simulation in which the wind speed was reduced to 1 m s−1, compared with 2.9–4.1 m s−1 outside (depending on the season), and this resulted in more pronounced deviations in leaf temperatures (exceeding 1°C in certain cases), which is unsurprising given the sensitivity of surface temperatures to wind speed (Table 1). Finally, we also ran simulations in which the percentage of radiation blocked by the glasshouse structural components (5% in the other model runs) was changed to 0 or 25%, respectively. In the case of no radiation blocking, the leaf temperatures were raised by c. 0.15°C in comparison to 5% blocking, leading to slightly increased (decreased) differences with outside leaf temperatures on sunny (cloudy) days. When the glasshouse structure blocked 25% of incoming radiation, the leaf temperature was decreased by c. 0.5°C compared with a 5% blocking on both bright and overcast days. This led to leaf temperatures slightly below those outside if sunny (−0.3°C), and somewhat more pronounced if cloudy (−0.6°C). The results were comparable for all cover materials. Daily courses of leaf temperatures reveal no strong or evident trends (Fig. 4). The courses pronounce the merely minor differences in leaf temperature effects of the different covering materials.

Table 2.   Leaf temperatures (Tleaf) determined for different cover materials of climate-controlled glasshouses (glass, PVC, low-density polyethylene (LDPE))
 Cloud coverTair,av (°C)u (m s−1)Tleaf,av (°C)
OutsideGlassPVCLDPE
  1. Averages (av) of half-hourly values based on meteorological data from the Brasschaat FluxNet site for one typical (i.e. close to the season average) day in each of the four seasons, except for wind speed (u), which was set as the seasonal average (data from the Royal Meteorological Institute). Radiation data for a cloudy day were superimposed on the meteorological data of a sunny day to ensure that only radiation differed.

Winter02.704.12.753.153.082.79
12.704.13.033.002.962.95
Spring08.453.48.468.748.728.40
18.453.48.278.198.158.14
Summer017.972.917.2517.4617.4217.15
117.972.917.4417.2317.2217.23
Summer drought017.972.917.9318.1418.1317.82
Autumn09.723.49.469.829.769.47
19.723.49.709.629.579.57
Figure 4.

Air and leaf temperatures during an entire summer’s day, for three different situations: (a) a bright day; (b) a cloudy day; and (c) a bright day with drought-stressed plants. Leaf temperatures outside and inside climate-controlled glasshouses (at outside temperature) were modelled using half-hourly data on shortwave radiation, air temperature and relative humidity from the Fluxnet site in Brasschaat, Belgium, with wind speed set at 2.9 m s−1, the long-term average during summer in Belgium. See text for details on cover material properties and other inputs. LDPE, low-density polyethylene.

Open-top chambers

We ran a number of simulations under varying conditions (Table 3). The daytime shortwave radiation on clear days was set at 750 W m−2, a maximum daily value that is exceeded during > 4 months in Barrow (radiation data retrieved from the National Oceanic and Atmospheric Administration), located at 71°17’ N. The 750 W m−2 input is therefore fairly conservative for maximum irradiation during the growing season as most OTC experiments are conducted at more southern latitudes (Henry & Molau, 1997). If air temperatures are assumed not to be increased inside OTCs (i.e. to compare net effects with climate-controlled glasshouses), our analysis shows that leaf temperatures on clear days can be expected to be 2.5°C higher inside than outside OTCs for the conditions in Table 2. If we consider OTCs to increase air temperatures by 1.5°C (a realistic value; see Marion et al., 1997; Hollister & Webber, 2000), the same conditions as before would lead to increases in leaf temperatures of 3.6°C compared with outside. During clear nights, the calmer conditions inside OTCs would stimulate leaf cooling, resulting in leaf temperatures c. 0.5°C below those found outside. Note that we considered nights to be dark (shortwave radiation set at zero), which is not necessarily the case during Arctic summers, and that we did not take into account trapping of longwave radiation inside the OTCs. Night-time air warming was not taken into account as studies documented that this is mostly absent in OTCs (Marion et al., 1997; Sierra-Almeida & Cavieres, 2010), which could indicate that heat trapping in these chambers is generally limited. Cloudy conditions result in only minor daytime leaf temperature increases within OTCs (assuming equal Tair), while leaves inside OTCs would no longer become cooler than those outside during the night. Under a scenario of drought stress (where stomatal conductance is decreased from 0.3 to 0.05 mol m−2 s−1), the Tleaf difference is increased by a further 1–1.5°C during the day (Table 3). Leaf temperatures are generally increased if the RH is higher as this suppresses latent energy loss. Finally, the smaller the leaves, the less their temperature differs inside and outside OTCs because of their stronger coupling to the atmosphere.

Table 3.   Comparison of leaf temperatures (Tleaf) inside and outside open-top chambers (OTCs) under varying conditions of air temperature (Tair), radiation (Rs), relative humidity (RH), leaf dimension (d), adaxial and abaxial stomatal conductance (gvs), and cloud cover
LocationTair (°C)Rs (W m−2)RHu (m s−1)d (m)gvs,ad (mol m−2 s−1)gvs,ab (mol m−2 s−1)Cloud coverTleaf (°C)ΔT (OTC – outside) (°C)
  1. Reduction of wind speed inside OTCs is based on data from Dalen (2004).

Outside107500.61.220.020.30.3012.43 
OTC107500.60.190.020.30.3014.922.49
Outside500.61.220.020.050.0504.44 
OTC500.60.190.020.050.0503.84−0.60
Outside102000.61.220.020.30.3110.19 
OTC102000.60.190.020.30.3110.800.61
Outside500.61.220.020.050.0515.30 
OTC500.60.190.020.050.0515.650.35
Outside107500.61.220.020.050.05013.85 
OTC107500.60.190.020.050.05017.753.90
Outside107500.61.220.020.30.3012.43 
OTC11.57500.60.190.020.30.3016.033.59
Outside107500.61.220.020.050.05013.85 
OTC11.57500.60.190.020.050.05018.965.11
Outside107500.91.220.020.30.3013.31 
OTC107500.90.190.020.30.3016.192.87
Outside500.91.220.020.050.0504.67 
OTC500.90.190.020.050.0504.31−0.36
Outside107500.61.220.0050.30.3011.35 
OTC107500.60.190.0050.30.3012.941.59
Outside500.61.220.0050.050.0504.69 
OTC500.60.190.0050.050.0504.31−0.38

Discussion

The comparison with actually recorded leaf temperatures shows that our energy balance model, although a simplification of reality, is capable of a fair approximation of real-life values. To our surprise, the warmer ‘sky’ inside glasshouses in combination with the optical properties of the cover materials produced only small deviations in leaf temperatures compared with outside. We suggest that these differences are so small compared with the inevitable increased variability in air temperatures within a climate-controlled glasshouse vs outside (caused by lag times in heating/cooling cycles) that devising a correction for the temperature control system to negate leaf temperature deviations is unnecessary. It is more important to ensure that the glasshouse structure is as insubstantial as stability requirements allow, as glasshouses that block a substantial proportion of the incoming radiation (e.g. Svejcar et al., 1999) would lead to lower-than-normal canopy temperatures, apart from the direct and unwanted effect on photosynthesis of reduced light. More crucial still with regard to maintaining leaf temperatures close to natural values is ensuring that the wind speed inside glasshouses does not deviate too much from that outside. The sensitivity analysis indicates that leaf temperature deviations from outside (assuming equal air temperature) could be kept small if the fans produce a wind speed of c. 2 m s−1 (unless the outside wind speed is lower). Also of note is the fact that the absence of a humidity control in many glasshouses will also affect leaf temperatures through direct RH effects on transpiration and indirect effects via stomatal conductance. In contrast to wind, our simulations suggest that the choice of covering material of climate-controlled glasshouses has fairly little effect on leaf temperatures. The thinnest cover (LDPE) led to the smallest temperature differences, which is relatively unsurprising as a thinner foil reduces decoupling from outside conditions. Finally, it is worth noting that glasshouses do not modulate stomatal effects on leaf temperatures, which is relevant when simulating droughts.

More consequential were the results from our simulations with open-top chambers. They demonstrate that the warming effect generated by these OTCs is likely to have been underestimated. While OTCs have been praised for their practicability and low manufacturing cost, an important criticism is that the temperature increases they provide are small in comparison to year-on-year variability (Hollister & Webber, 2000), requiring OTC experiments to run for many years before being able to assess temperature effects (Hollister et al., 2005; Hudson & Henry, 2010). For the widely used ITEX chambers, reported increases in air temperature range from less than 0.5ºC in forests (De Frenne et al., 2010) to c. 1.5ºC in tundra (Marion et al., 1997). Our results suggest that the effective warming for plants could well be twice that in many ecosystems, caused by the much reduced wind speed inside OTCs. Note that the trend of significantly higher increases in leaf than in air temperatures inside OTCs was also apparent from our actual measurements. This would counter the criticism that OTC-generated temperature rises are only minor. It also means that OTC studies that have extrapolated their findings on warming responses should be treated with caution. For example, if the authors were under the impression that their technique provided merely 1°C of warming, whereas in reality it was 2°C at tissue level, then they may have overstated the possible responses to a future, 4°C warmer climate by extrapolating their findings. In other words, our results suggest that the sensitivity of plants to climate warming may have been overestimated if studied in OTCs. A related potential pitfall is the fact that temperature variability inside OTCs could be higher than previously thought, as the phenomenon of higher-than-expected leaf : air temperature ratios is more apparent during the day than during the night. Moreover, lower wind speeds will affect shaded and sunlit foliage differently, also increasing variability within the canopy. In a recent study, Godfree et al. (2011) proposed an OTC with increased thermal mass (through water stored in pipes) to top off extreme temperatures. This technique indeed decreased differences between day and night, but the authors unfortunately did not report on canopy temperatures.

It is well worth noting that wind-related increases in canopy temperature inside OTCs are likely to be exacerbated if plants are drought-stressed. Stomatal closure is a common response to dry conditions (of soil and/or air) and has been shown to increase leaf temperatures substantially (De Boeck et al., 2011) through the reduction in transpirational cooling. Our model runs demonstrate that stomatal closure indeed led to rises in leaf temperature relative to air temperature under outside conditions, but the increases inside OTCs were much larger (Table 2). Drought-stressed leaves inside OTCs ended up almost 8°C warmer than air, compared with nearly 4°C outside. Researchers using OTCs in a drought experiment or in a region with common droughts should be well aware of this effect. Also, increased water losses caused by a larger moisture gradient between leaf and air may be larger than previously acknowledged because of the higher than expected canopy temperatures. This could increase the drying effect of open-top chambers. Note that the drastic decreases in wind speeds inside the chambers could also affect pollination. A recent study by Klady et al. (2011) found that insect pollination was increased inside OTCs in several Arctic locations. We would argue that this effect may have been caused or propagated by an attraction of insects towards the warm low-wind islands that are the OTCs and that, at the same time, wind pollination could have been significantly hampered by the drastic decreases in wind speed inside the chambers (a concern already raised by Molau & Shaver, 1997). Poor wind dispersal may also lead to increased self-pollination, which would affect self-incompatible plant species negatively in OTC studies.

The model used to simulate conditions inside OTCs only considered the documented reduction in wind speed inside the chambers. The comparison of modelled with measured leaf temperatures indicates that other effects, such as reductions in incoming shortwave radiation and entrapment of outgoing longwave radiation were not important enough to distort the modelled trends. For a broad study on temperature trends, this removes the need for complex calculations of the proportion of the incoming radiation that passes through the OTC cover material, which varies with solar elevation during the day and the season, as well as with the geometry of the specific type of OTC used. For a more precise and specific evaluation, which is not within the scope of this study, such effects could be more relevant as they are likely to cause heterogeneity of warming inside the OTC. For example, longwave radiation is reflected more towards the edges (as determined by the view factor) while other variables affecting leaf temperature, such as wind speed and RH, may also differ within the chamber.

In conclusion, glasshouses tracking field temperature seem to affect leaf temperatures only marginally, provided that the structure blocks as little light as possible and the fans produce wind speeds comparable to outside. The sensitivity of surface temperatures to wind speed is reflected in substantial leaf temperature increases within the wind-blocking OTCs. This would in many cases double the actual warming provided by this passive technique, while the temperature variability is also likely to be higher than previously thought. The deviation from outside leaf temperatures is further increased under drought conditions. These findings should allow a more precise evaluation of the degree of warming provided by the two widely used methods discussed here. Our results also illustrate the importance of wind, which tends to be substantially reduced in enclosures. Although often neglected, we advise that wind speeds should always be measured in climate-warming experiments in order to have a more accurate understanding of the induced temperature increases.

Acknowledgements

H.J.D.B. is a postdoctoral research associate of the Fund for Scientific Research – Flanders. T.D.G. is a PhD fellow of the Fund for Scientific Research – Flanders (FWO) and the Flemish Institute for Technological Research (VITO). We thank Marc Wellens, Maya Verlinden and Freja Dreesen for assistance and three anonymous referees for helpful suggestions. The research leading to these results has received funding from the European community’s Seventh Framework programme (FP2007-2013) under grant agreement no. 262060 (ExpeER).

Ancillary