Tropical rainforest mesocosm (TRF)
The tropical rainforest mesocosm at the Biosphere 2 covers ca. 1950 m2 and is representative of a managed tropical rainforest ecosystem similar to those found in South America. It is contained under a 26700 m3 glass ziggurat enclosure25 (Fig. 1). Within the enclosure, there are 95 species of tropical plants, including 23 species of trees and 67 species of understory plants. The most dominant species are Clitoria fairchildiana, Phytolacca dioica, Arenga pinnata, Ficus benjamina, Syngonium podophyllum, Piper sp., Musa sp. and Pachira aquatica. To recreate real rainforest conditions, such as low light intensity beneath the canopy, larger trees and understory plants (Musa sp., Piper sp., Hibiscus rosa-sinensis) were planted around the edges of the glass enclosure to block out direct sunlight. The soil profile consists of an up to 4 m deep subsoil layer and a topsoil layer of variable thickness (70-100 cm)26. Located at the centre of the rainforest is a small artificial “mountain” within which a laboratory was constructed to house the on-line GC-MS and where the ambient temperature and humidity inside could be controlled.
Drought experiment
This experiment was conducted during the Water, Atmosphere, and Life Dynamics (WALD) campaign. Prior to the commencement of the drought period, the tropical rainforest was wetted from above with a sprinkler system to simulate rainfall, using ~20,000 liters of water 3 times a week. After watering on 7th October 2019 the rainforest biome was left to dry. During some stages air handler units were employed for the removal of humidity by condensation, otherwise the rainforest biome was left to dry naturally, until 3rd December 2019. The first water to the rainforest was introduced at the bottom through a network of drainage pipes under the soil on top of the concrete and steel structure underlying Biosphere 2. The rainforest was again watered from above using the sprinkler system, on 12th December 2019, 19th December 2019, and every 2 days afterwards. The rainforest temperature was controlled throughout the experiment and the temperature at 13 m was on average between 28 and 32 °C during daylight hours and between 21 and 24 °C during nighttime hours.
The photosynthetically active radiation (PAR), temperature and relative humidity (RH) were recorded every 15 minutes using sensors connected to a datalogger (PAR sensors (Apogee SQ110, Campbell Scientific, Logan, UT, USA), temperature and RH with Vaisala HMP 45c sensors (Vaisala Oyi., Vantaa, Finland; purchased through Campbell Scientific). Sensors reported every 15 minutes to a centralized CR1000 datalogger (Campbell Scientific, Logan, UT, USA) with an AM16/32B multiplexer (Campbell Scientific, Logan, UT, USA). The dataloggers were connected to a centralized database with NL100 communications modules (Campbell Scientific, Logan, UT, USA) and the data is available through the Biosphere 2 website (www.biosphere2.org/data-models/rainforest-data). The PAR sensor was located at 13 m on the central measurement tower and the humidity and temperature sensors were located at a height of 13 m on the north-eastern measurement tower together with the sampling inlet. The soil moisture data presented in Fig. 2d is an average of measurements that were recorded every 15 minutes from four different soil pits (soil moisture and temperature sensors (Truebner SMT100, Truebner Gmbh, Neustadt, Germany) and water potential sensors (TEROS 21, Meter Group, Pullman, WA, USA) in all four pits at 5 cm depth and at the soil-concrete interface (subsoil bottom)). Since the sensors are 30 mm wide and inserted vertically into the soil with the soil depth indicated at the midpoint, each depth is ± 1.5 cm.
The rainforest enclosure acted as a semi-enclosed system where there was constant air exchange with the outside environment. The air-exchange rate from the tropical rainforest enclosure was measured using sulfur hexafluoride (SF6) at low ppb levels as a tracer gas, as it is completely anthropogenic and its concentration is < 10 ppt in background air and therefore it is only be affected by leakage and flushing. More information on using SF6 as a tracer gas in the tropical rainforest mesocosm at the Biosphere 2 can be found elsewhere25. Once the percentage exchange rate of SF6 was obtained and interpolated, the measured data was corrected by using the equation
![](https://myfiles.space/user_files/44073_bbc9ffc5e562ebe7/44073_custom_files/img1628870152.png)
Where the VMRu was the uncorrected data, ER was the exchange rate percentage and VMRc was the corrected data. The incoming VMR of all VOC’s were assumed negligible.
Determination of isoprene and monoterpene ambient mixing ratios
From 9th September to 23rd December 2019, the ambient air from 13 m high within Biosphere 2 rainforest was continuously drawn at a flow of approximately 800 ml min-1 through a main Teflon inlet line which comprised of 37 m of 0.625 cm (¼ “) Teflon tubing. 13 m was chosen as the sampling height as this was the height within the enclosure that had the greatest leaf area index. The main inlet line was fitted with a (Cole Palmer, EW-02915-31) filter. After approximately 26 m, a t-piece was connected to the main inlet line which was connected to a thermal desorption unit (TD) (TT247-xr, MARKES International Ltd., U.K.) using 7 m of = 3.175 mm (1/8 “) Teflon tubing. All sampling lines were insulated and heated to 50 °C to avoid water condensation within the lines. The line to the TD was continuously purged to avoid the sampling of a dead volume with a pump situated the behind the TD in the flow path. During sampling, the pump behind the TD drew air from the main inlet line at flows ranging between 70 and 200 ml min-1 for 10 minutes. The collected air was sampled first through a water condenser (kori-xr, MARKES International Ltd., U.K.). This allowed for the removal of water whilst leaving the target VOCs unchanged. The dehumidified sample was then pre-concentrated onto a cold injection trap at 30 °C (Material emissions, MARKES International Ltd., U.K.). After sampling, the injection trap was purged for 1 minute with helium with a flow of 50 ml min-1 before being rapidly heated to 300 °C and desorbed for 3 minutes. The sample was removed from the cold trap with a Helium flow of 3 ml min-1 including a split flow of 2 ml min-1 and injected into the separating column.
The rainforest ambient air was analysed using a Gas Chromatograph (GC) (6890A, Agilent Technologies, U.K.). The carrier gas used was research 6.0 grade helium (Airgas®, USA.) Separation of the sampled compounds was achieved using a 30 m β-DEX™ 120 column (Sigma-Aldrich GmbH, Germany) with 0.25 mm internal diameter and a 0.25 μm film thickness. The temperature program used was as follows, 40 °C for 5 minutes then 40 °C to 150 °C at 4 °C min-1 and 150 °C to 200 °C at 30 °C min-1. The column flow was set to 1 ml min -1.
The GC was coupled with a quadrupole mass spectrometer (MS) (5973N, Agilent Technologies, U.K.), operated in selected ion mode for the identification of mass ions 68, 69, 93, 94, 119, 120, 136, 137 each with a dwell time of 60 ms.
The identification of the target compounds was achieved by first operating the MS in scan mode to obtain full mass spectra to be able to compare with the NIST 70 eV electron ionization library. For further confirmation, a gas standard mixture (Apel-Riemer Environmental Inc., 2019) containing the target compounds was injected into the GC-MS system. The same gas standard mixture was also injected onto sorbent cartridges and subsequently desorbed into a GC - time of flight - mass spectrometer (GC-TOF-MS) operated with identical conditions to the online GC-MS. Using liquid standards, the headspace of the individual compounds was taken onto sorbent cartridges and also desorbed into the GC-TOF-MS. The retention times from the chromatograms of the individual compounds was then crosschecked with the chromatogram of the gas standard mixture.
The MS was tuned on a weekly basis and the linearity was checked throughout the campaign. The gas standard mixture was injected into the system after each tuning and after 10 samples were analysed. Routine calibrations were performed by initially flushing the TD system with the gas standard mixture at a flow of 20 ml min-1 for 2 minutes in order to remove the dead volume. The calibration gas was then injected with a flow of 20 ml min-1 for 5 minutes directly onto the cold injection trap within the TD. The calibration gas sample was then treated with the same TD-GC-MS parameters as the routine sampling. This step was repeated three times before sampling continued. The MS responses to the injected calibration gas samples were then plotted against the time since the last MS tune to track the MS sensitivity drop, which allowed for the correction and calibration of the raw data. To check the linearity, the same procedure was used with the calibration gas injection time being increased in stepwise intervals of 2.5 minutes from 0.5 to 12.5 minutes.
Data management
The highest individual VMRs measured were in excess of 3 ppb for (-)-alpha-pinene and 400 ppt for (+)-alpha-pinene were measured during ED (Ext. Data Fig. 5 and 6). However, to evaluate the general trends of all measured compounds through each stage, the hourly total monoterpene and isoprene data was smoothed by applying a Savitzky-Golay filter to keep long-term trends whilst removing short-term fluctuations. The dataset was further split into daytime and nighttime, using photosynthetically active radiation (PAR) data collected at the point of measurement. The smooth function (MATLAB) was then used to suppress noise in the trend line for each compound and the uncertainties were propagated using the same functions. To obtain average diel cycles for each compound, and the average composition all data in each of the five stages of the campaign were taken. A moving median calculation with a window length of 5 data points was applied to each group. The diel cycles were averaged over 435, 526, 349, 136 and 193 data points for PD, ED, SD, DRW and RRW respectively. β-Myrcene, (+)-β-pinene, α -terpinene and terpinolene, were also observed but not included as they amounted to an average of less than 5 % of the average total monoterpene for the entire measurement period. Ocimene was also observed but not calibrated with the on-line GC-MS system.
The atmospheric water potential was calculated from the Vaisala temperature and relative humidity measurements according to:
![](https://myfiles.space/user_files/44073_bbc9ffc5e562ebe7/44073_custom_files/img1628870192.png)
where Ψ is the water potential (in Pa), R the gas constant (8.3144 J mol-1 K-1), T the temperature (oC), Vw0 the molar volume of water at 293 K, and %RH is the relative humidity (%)33. Zone 3 and 4 are the two height zones that contain the majority of the canopy in the Biosphere 2 rainforest (Ext. Data Fig. 3). The environmental conditions were averaged over all the sensors that were located within these zones. NEE was calculated every 15 minutes based on the change in moles of CO2 within the rainforest ecosystem and the amount of CO2 lost or gained with the air-exchange with the outside:
![](https://myfiles.space/user_files/44073_bbc9ffc5e562ebe7/44073_custom_files/img1628870228.png)
where CO2t, CO2t-1, CO227m, and CO2outside are the moles of CO2 at the time calculated, previous time-step, 27m or top of the rainforest where the air flows out, and the outside air coming into the rainforest. Area stands for the soil surface area of the rainforest. The moles of CO2 were calculated based on the ideal gas law and the CO2 concentration:
![](https://myfiles.space/user_files/44073_bbc9ffc5e562ebe7/44073_custom_files/img1628870252.png)
where V denotes the representative volume for the CO2 measurement (either the rainforest volume fraction or the volume of air exchanged), P denotes the pressure (measured inside and outside the Biosphere 2 rainforest using WeatherHawk WXT 530, Vaisala Oyi., Vantaa, Finland), TAve denotes the average air temperature for the measurement zone or the outside air temperature (measured using HMP45c temperature and humidity sensors (Vaisala Oyi., Vantuu, Finland), R denotes the gas constant, and [CO2] denotes the CO2 concentration measured inside Biosphere 2 with GMP343 CO2 sensors (Vaisala Oyi., Vantuu, Finland) and outside Biosphere 2 with GMP220 sensors (Vaisala Oyi., Vantuu, Finland) and inside the air inflow with an Aerodyne Dual QCL (Aerodyne Research Inc., Billerica, MA, USA). The ecosystem assimilation (A, mmoles m-2 s-1) was calculated from the Net Ecosystem Exchange (NEE) and assuming that the nighttime respiration (R) was representative for the daytime respiration (a reasonable assumption in tropical forest ecosystems34,35):
![](https://myfiles.space/user_files/44073_bbc9ffc5e562ebe7/44073_custom_files/img1628870276.png)
13CO2 pulse labelling experiment
The 13CO2 pulses were carried out on 5th October at 8:00 am (MST) and 23rd November 2019 at 9:00 am (MST) to coincide with peak photosynthetic activity36. A deliberate effort was made to proceed with the pulse experiments on days with high amounts of direct sunlight, when there would be a high rate of photosynthesis, to maximize CO2 uptake. During the first pulse, the rainforest was fumigated with 10 lpm of 99% 13CO2 (Millipore Sigma, Burlington, MA, USA) for 15 minutes. In order to balance reduced carbon assimilation rates during the drought, 20 lpm of 99% 13CO2 was released over 15 minutes during the second pulse. The δ13C value of atmospheric CO2 within the rainforest was monitored throughout each pulse using a Tunable Infrared Laser Direct Absorption Spectrometer (TILDAS, Aerodyne Research, Billerica, MA, U.S.A). After 4 hours during the first pulse, and 5.2 hours during the second pulse, the flow of air through the rainforest was increased and, at midday, windows were temporarily removed from the enclosure of the B2-TRF and excess 13CO2 was ventilated to the outside air, so that the entry of 13C into the mesocosm could be more accurately traced back to a fixed point in time.
Determining enantiomer 13C isotope ratios
Pairs of monoterpene enantiomers abundant in the air were analysed in the TRF ecosystem in a height of 13m at the atmosphere tower. 5, 16, 36, 11 and 14 glass cartridges were sampled during pre-pulse, first pulse, first post-pulse, second pulse and second post pulse respectively. For terpene accumulation, ambient air was drawn through glass cartridges filled with about 100 mg Tenax (Sigma, Germany) as an adsorbent at a controlled flow rate of 200 ml min-1 for 90 min using a handheld pump (SKC ltd., Dorset, U.K). Glass cartridges were kept at 4 °C until analysis. The samples were analyzed at the university of Freiburg on a system consisting of a gas chromatograph (GC 7980, Agilent Technologies, Germany) coupled to a mass selective detector (MSD 5975C, Agilent Technologies, Germany) and equipped with a thermodesorption unit (TDU, Gerstel, Germany) and a cold injection system (CIS, Gerstel, Germany). For analysis of 13C isotope ratios, this system was coupled to an isotope ratio mass spectrometer (IRMS, Isoprime precisION, Elementar Analysesysteme GmbH, Langenselbold, Germany) via a combustion furnace (GC 5 interface, Elementar Analysesysteme GmbH, Langenselbold, Germany). For analysis, air sampling glass cartridges were heated to 220 °C for 5 min to thermodesorb terpenes and channel them into the CIS which was kept at -70 °C. By heating the CIS to 240 °C for 3 min, terpenes were directed onto the GC separation column (BetaDex 120 Chirality, 60 m x 250 μm x 0.25 μm, Supelco, USA) with a He stream of 1 ml min-1. The oven program started at 45 °C which was kept for 1 min, temperature was then stepwise increased to 60 °C, 150 °C and 210 °C at rates of 2 °C min-1, 1 °C min-1 and 3.5 °C min-1, respectively. The eluate was split and ca. 10 % was directed into the MSD for terpene identification and quantification. For this purpose, the MSD was run in SIM mode detecting m/z 68, 93, 119 and 136. The remainder eluate passed the combustion furnace where at a temperature of 850 °C the terpenes were oxidized to form CO2 and H2O. After elimination the H2O by a Nafion water trap, the 13C/12C ratios of the CO2 were measured by the IRMS.
Statistical information
The t-tests used on the data presented in Fig. 3, were one-tailed, two sample unequal variance t-tests that were performed using the MATLAB R2017B software (Ext. Data Table 1).
GC-IRMS data processing
13C isotopologues elute slightly faster from the GC than their 12C counterparts, meaning that δ13C values are not homogenous across the peak37. For chromatographically unresolved compounds such as (-)-α-pinene (extended data figure 6), integrating from the beginning of the peak to the trough between it and the subsequent unidentified co-eluting peak results in δ13C values that appear artificially enriched in 13C. Absolute δ13C values for such compounds cannot be reported. We therefore report relative offsets between measured δ13C values during the 13CO2 pulses and ambient conditions
![](https://myfiles.space/user_files/44073_bbc9ffc5e562ebe7/44073_custom_files/img1628870311.png)
While the relative abundance of (-)-α-pinene to its coeluter changes its apparent δ13C value, we are confident that these chromatographic effects cannot account for the relative 13C-enrichment of this compound during the 13CO2 pulses for two reasons. First, the 13C-enrichment of the samples from the 13CO2 pulses are significantly enriched relative to the variability in ambient δ13C values, which are subject to a greater range of relative peak heights. Second, 13C enrichment is only apparent during the 13CO2 pulses when (-)-α-pinene is integrated in various combinations with the other three peaks in the co-eluting hump, but not when any of these other peaks are individually integrated trough-to-trough (Ext. Data Fig. 6).
There is no indication that (+)-α-pinene becomes enriched in 13C during either 13CO2 pulse. However, given its small size and poor resolution from the preceding unlabeled peak, we cannot definitively rule out slight 13C-enrichment for (+)-α-pinene. Myrcene, trans-β-ocimene, and (+)-limonene all have well resolved fronts and poorly resolved tails. These compounds display either high 13C-enrichment during the 13CO2 pulses (trans-β-ocimene during both pulses, myrcene during the second) or no enrichment at all (myrcene during the first pulse, (+)-limonene during both). Since these chromatographically similar peaks only display large 13C-enrichment in some compounds and not others, we consider apparent presence or absence of label uptake to be robust results for these compounds (Ext. Data Fig. 8 & 9).
Soil uptake and emission of monoterpenes experiment
For the investigation of how monoterpenes are taken up and emitted by the soil, 3 soil chambers made from polyvinyl chloride were placed on pre-installed soil collars around the B2-TRF (Ext. Data Fig. 2). Using sorbent cartridges, samples were taken from the atmosphere, located at the inlet (MRatm) of the soil chamber, and at the same time from the outlet (MRsoil). Samples were collected at 200 ml min-1 for ~ 10 minutes using a handheld pump (SKC ltd., Dorset, U.K). The sorbent cartridges were made from inert coated stainless steel (SilcoNert 2000 (SilcoTek™, Germany)). The sorbent consisted of 150 mg of Tenax® TA followed by 150 mg of Carbograph™ 5 TD (560 m2/g).The size of the Carbograph™ particles was in the range of 20-40 mesh. The Carbograph™ 5 was supplied by L.A.R.A s.r.l. (Rome, Italy) and Buchem BV (Apeldoorn, The Netherlands) supplied the Tenax®.
33 Lambers, H. & Oliveira, R. S. in Plant Physiological Ecology 187-263 (Springer International Publishing, 2019).
34 Restrepo-Coupe, N. et al. What drives the seasonality of photosynthesis across the Amazon basin? A cross-site analysis of eddy flux tower measurements from the Brasil flux network. Agricultural and Forest Meteorology 182-183, 128-144, doi:https://doi.org/10.1016/j.agrformet.2013.04.031 (2013).
35 Hutyra, L. R. et al. Seasonal controls on the exchange of carbon and water in an Amazonian rain forest. Journal of Geophysical Research: Biogeosciences 112, doi:https://doi.org/10.1029/2006JG000365 (2007).
36 RASCHER, U. et al. Functional diversity of photosynthesis during drought in a model tropical rainforest – the contributions of leaf area, photosynthetic electron transport and stomatal conductance to reduction in net ecosystem carbon exchange. Plant, Cell & Environment 27, 1239-1256, doi:https://doi.org/10.1111/j.1365-3040.2004.01231.x (2004).
37 Ricci, M. P., Merritt, D. A., Freeman, K. H. & Hayes, J. M. Acquisition and processing of data for isotope-ratio-monitoring mass spectrometry. Organic geochemistry 21, 561-571, doi:10.1016/0146-6380(94)90002-7 (1994).