How tropical forest leaves respond to climate change has important implications for the global carbon cycle and biodiversity.  Climate change could impact the energy balance properties of tropical forest canopies through 1) long-term trait changes and 2) abrupt disruptions/damage to leaf/photosynthetic machinery.  We assessed the radiative and evaporative impacts of two recently proposed impacts of climate change on tropical forest canopies: 1) long-term leaf darkening and 2) leaf death through high temperature extremes.  We darkened leaves to absorb 138 Wm^-2 more energy in the upper canopy of a seasonally-dry tropical moist forest in Panama.  20% of this energy went towards heating leaves by ~4°C, 3% went towards warming the air, and 77% went towards evaporative cooling.  This leaf warming led to the appearance of necrosis across 9±5 % of the leaf area on certain species.   In contrast, brightening leaves decreased energy absorbed by an average of 58 Wm^-2, which mainly reduced evaporation (88%) with only 12% reducing leaf temperatures (and no sensible heat flux).  This asymmetrical result suggests leaves may be close to hydraulic limitations towards the end of the dry season.  Similar albedo increases in a model (CLM 4.0) did not diverge between brightening and darkening leaves and generally showed sensible heat flux to dominate although there were strong geographic trends.  Heat death in leaves generally heated nearby leaves (by an average of ~1.35°C) and air temperature (by 0.5°C), but less than hypothesized because leaf albedo increased.  Overall, our canopy top experiments question important potential climate feedbacks, but need further study.

Site location –We used a 60-m tall canopy crane managed by the Smithsonian Tropical Research Institute (STRI) in Parque Natural Metropolitano (8.994410, -79.543000) near Panama City, Panama, to access canopy top leaves (Fig 1).  We focused on five distinct canopy level trees of five species. The species we used were: Anacardium excelsum (Bertero & Balb. ex Kunth) Skeels (Anacardiaceae), Castilla elastica Cerv. (Moraceae), Aiouea montana (Sw.) R.Rohde (Lauraceae), Spondias mombin L. (Anacardiaceae) and Luehea seemannii Triana & Planch. (Malvaceae).  A meteorological station installed at 25 m height on the tower of the crane shows that this area has a mean annual temperature of 26.2°C (average day/night: 28.0/24.5°C), and receives ~1900 mm rain per year, with a 4-month dry season from late December to late April (Paton, 2020).  We accessed living canopy top leaves on March 22, 25, 26 and 27 of 2024, which is towards the end of the dry season.  Land surface temperatures (LST) for the canopy crane area for diurnal and annual timescales derived from ECOSTRESS (ECO2LSTE.001) data (id 4a36c5d2-54c6-4d81-8679-7ed5f0182c53) (Fig 1) show that measurements were collected during warm, but not unusually warm periods.

Leaf Manipulations

Albedo - We added a thin coat of Viva Doria Virgin Activated Charcoal Powder from hardwood tree to darken 3–5 leaves per branch, and white kaolin clay powder (Al₂Si₂O₅(OH)₄) to brighten 3–5 leaves per branch, each on several (3–5) branches per tree.  We put the powders in a small plastic bag and dipped the leaf (still attached to the tree) in the bag trying to evenly coat the top of the leaf with a thin layer without getting powder on the bottom.

Dead leaves – We heat-killed leaves by dipping attached leaves into boiled water (~100°C), submerging most of the leaf, while keeping the petiole dry and unaffected by the heat treatment.  We would dip ~20–30 canopy top leaves over a period of ~15 minutes for single big leaves or branches of small leaves and hold the leaves in the water for ~10 seconds per leaf/branch.  Leaves would often start to show signs of necrosis within minutes of being heated.  After boiling the leaf, we measured temperatures and reflectance properties (see below for details) between ~5 hours and 5 days after leaf death.  All leaves that were killed remained on the branches.  To estimate heat transfer from dead to live leaves, we put a dead leaf next to (with sides of leaves barely touching) a live leaf.  This live leaf near the dead leaf is called the treatment and another leaf generally more than 20cm away from the dead leaf was the control.

Leaf Microclimate - We measured leaf temperature adaxially using a handheld IR temperature gun (brand) held a few centimeters away from the measured leaf.  Then we put three Ecomatik broadleaf temperature sensors (LAT-B3) connected to a CS1000X datalogger logging at 1 Hz that measures leaf surface temperature and air temperatures 3.5cm above the leaf for a period of between 2–5 minutes.  With the leaf thermocouples on the leaves, we further took a thermal and RGB image of the leaves with a Parrot ANAFI thermal drone.  We ensured that the compared leaves were maintained at the same orientation (often, but not always flat).  We measured PAR at the time of the measurements with an LI-190R quantum sensor (LI-COR, Lincoln, Nebraska, USA). 

Spectroscopy - We used an ASD field spectrometer 4 with a fiber optic cable, contact probe and a leaf clip (Analytical Spectral Devices, Boulder, Colorado, USA) which measures from 325-2500 nm wavelength to estimate the change in leaf albedo from our manipulations.  We randomly selected three leaves of each branch and measured hemispherical reflectance near the mid-point between the main vein and the leaf edge (Asner and Martin, 2008). Measurements were collected with 136-ms integration time per spectrum (Asner and Martin, 2008; Doughty, Asner and Martin, 2011). We calibrated for dark current and stray light, and white-referenced to a calibration panel (Spectralon, Labsphere, Durham, New Hampshire, USA) after every branch. For each measurement, 25 spectra were averaged together to increase the signal-to-noise ratio of the data.

Conversion to albedo – We averaged reflectance between 400–700nm to calculate visible albedo and between 701 and 2500nm for NIR+SWIR albedo, and then averaged those to get total albedo. We did not measure leaf transmittance, so we use a general value of 0.4 for the NIR and 0.03 for the VIS (Doughty, Asner and Martin, 2011).  However, some of the NIR transmitted will be reabsorbed from below depending on LAI and other variables, so we use a value of 0.2 for the NIR to account for reabsorbed upwelling shortwave energy.  LAI below impacts upwelling shortwave energy because higher LAI will reflect more upwelling energy. 

Transmittance - We did not measure how our manipulations modified leaf transmittance in the field but did measure transmittance later in the lab on sycamore (Platanus occidentalis) and aspen (Populus tremuloides) (N=3 for each) using a Lambda 750S UV/VIS/NIR spectrophotometer (PerkinElmer Life and Analytical Sciences, Shelton, CT, USA).  On average, the Kalonite reduced transmittance by 0.05 in the NIR and 0.02 in the VIS and the charcoal reduced transmittance by 0.11 in the NIR and 0.02 in the VIS.  This is slightly less than a prior study showed that the Kalonite reduced transmittance by 0.15 in the NIR and 0.05 in the VIS (ABOUKHALED, Antoine, 1966).  Wiebe et al. (in prep) shows transmittance goes down as reflectance goes up in oven-dehydrated leaves, resulting in only minor changes to absorption below ~1300nm.  To account for this, we estimate that heat killed leaves have reduced transmittance by 0.05 in the NIR and 0.02 in the VIS.

Leaf energy balance modelling – Leaf energy balance is explained by eq 1:

Eq 1: ΔRabs = (ΔSr + ΔH + ΔL)

where ΔRabs is the change in energy absorbed from the albedo or leaf death manipulations. ΔH is the change in sensible heat (eq 3) and ΔL is the change in latent heat (eq 4).  ΔSr is the thermal radiation change in W m^-2 and a function of leaf temperature solved using the Stefan-Boltzmann blackbody equation (assuming heat storage in the leaf is negligible) as follows:

Eq 2:  ΔSr = (2*ε *σTcon^{^4}) – (2*ε *σTman^{^4})

Where σ =  5.67e^–8, ε = 0.98, Tcon  = leaf temperature of control leaves, Tman = leaf temperature of manipulated leaves and the 2 accounts for longwave radiation emitted from both sides of a leaf.  Leaves also absorb longwave from both the understory and the sky, but we do not consider this when calculating LE and H because understory and sky temps are the same across treatments.

We calculate sensible heat flux ΔH by calculating the energy needed in W m^-2 to achieve the measured change in air temperature.  We use the following equation:

Eq 3:     ΔH = (𝑐*𝑇air_con * m) –(𝑐*𝑇air_man *m)

where 𝑐 is the specific heat of air (1.005 J /g∘C at constant pressure), 𝑇air_con is the air temperature 3.5 cm above the control leaves, 𝑇air_man is the air temperature 3.5cm above the manipulated leaves.   The mass of air heated every second (m) is the volume of air cleared over a 1 m² area multiplied by the density of air (0.0012 g cm^-3 at sea level).  For a 1 m² area heating 5 cm of air, the volume is 50000 cm³ and under windy conditions (5 m s^-1), we assume that this mass of air (60g) would clear every 0.2 second, or 12g s^-1 m^-2.  Therefore, in our example, to heat 5 cm of air over a 1 m² area by 1°C, it would take 12 W m^-2.  We vary this number in a sensitivity study by testing values between 6 and 24 W m^-2.

We calculate latent heat flux ΔL, as the remainder according to eq 4.

Eq 4:  ΔL  = ΔRabs - (ΔSr + ΔH)

Percent necrosis - To determine percent necrosis on the darkened leaves we used Matlab’s image segmenter where we created ROIs (Regions of Interest) for the leaf and ROIs for the necrosis regions to get percent necrosis for a subsample of 9 leaves.

Earth System Modelling - We simulated biophysical feedbacks of a change in tropical leaf albedo using NCAR’s Community Atmosphere Model (CAM-4.0), coupled with the Community Land Model (CLM 4.0) with prescribed surface ocean temperatures, a river transport model and the Los Alamos Sea Ice Model (compset F_2000_CN). We ran the model with a resolution of 2° by 2.5° at the equator at a 20-min time step for 100 years following (Doughty et al., 2018). We ran the model with no dynamic vegetation response and atmospheric CO₂ was held constant at 367 ppm. We simulated tropical evergreen broadleaved plant functional types where NIR leaf-level reflectance was increased by 0.05 and 0.10, and decreased by 0.05 and 0.10 from a control NIR albedo of 0.45.  We averaged the final 50 years of the following variables (collected monthly) from CLM 4.0: surface albedo (W m⁻²); latent heat flux (W m⁻²); sensible heat flux (W m⁻²); rainfall (mm s⁻¹); and cloud cover (%).

Statistics – We used a simple t-test for each wavelength to see which wavelengths were statistically different between treatments.