Biogeosciences, 16, 785–796, 2019 https://doi.org/10.5194/bg-16-785-2019 © Author(s) 2019. This work is distributed under the Creative Commons Attribution 4.0 License. Automatic high-frequency measurements of full soil greenhouse gas fluxes in a tropical forest Elodie Alice Courtois1,2 , Clément Stahl3 , Benoit Burban3 , Joke Van den Berge1 , Daniel Berveiller4 , Laëtitia Bréchet1,3 , Jennifer Larned Soong1,5 , Nicola Arriga1 , Josep Peñuelas6,7 , and Ivan August Janssens1 1 Department of Biology University of Antwerp, Centers of Excellence Global Change Ecology and PLECO (Plants and Ecosystems), Universiteitsplein 1, 2610 Wilrijk, Belgium 2 Laboratoire Ecologie, évolution, interactions des systèmes amazoniens (LEEISA), Université de Guyane, CNRS, IFREMER, 97300 Cayenne, French Guiana 3 INRA, UMR EcoFoG, CNRS, Cirad, AgroParisTech, Université des Antilles, Université de Guyane, 97310 Kourou, France 4 CNRS/Université Paris-sud, 362 rue du doyen André Guinier, 91405 Orsay CEDEX, France 5 Climate and Ecosystem Sciences Division, Lawrence Berkeley National Laboratory, 94720, Berkeley, USA 6 CSIC, Global Ecology Unit CREAF-CSIC-UAB, Bellaterra, 08193 Catalonia, Spain 7 CREAF, Cerdanyola del Vallès, 08193 Catalonia, Spain Correspondence: Elodie Alice Courtois (elodie.courtois@cnrs.fr) Received: 16 July 2018 – Discussion started: 15 August 2018 Revised: 23 January 2019 – Accepted: 25 January 2019 – Published: 12 February 2019 Abstract. Measuring in situ soil fluxes of carbon dioxide pogenic global warming. Understanding the magnitude of (CO2 ), methane (CH4 ), and nitrous oxide (N2 O) continu- GHG fluxes in natural ecosystems has recently become a pri- ously at high frequency requires appropriate technology. We ority in the study of GHG balances (Merbold et al., 2015). tested the combination of a commercial automated soil CO2 Tropical intact forests cover 1392 Mha globally and repre- flux chamber system (LI-8100A) with a CH4 and N2 O ana- sent about 70 % of the total tropical forest area (1949 Mha), lyzer (Picarro G2308) in a tropical rainforest for 4 months. which accounts for the largest area of global forest biomes A chamber closure time of 2 min was sufficient for a reli- (∼ 50 %). Very few reliable long-term datasets on full GHG able estimation of CO2 and CH4 fluxes (100 % and 98.5 % balances are available from tropical ecosystems, despite their of fluxes were above minimum detectable flux – MDF, re- known importance for the global cycles of these three GHGs spectively). This closure time was generally not suitable for (Dutaur and Verchot, 2007). This is in part due to the chal- a reliable estimation of the low N2 O fluxes in this ecosystem lenges of designing and operating continuous, multi-gas flux but was sufficient for detecting rare major peak events. A analysis systems in tropical forests. Soil processes in partic- closure time of 25 min was more appropriate for reliable es- ular are responsible for an important part of GHGs that are timation of most N2 O fluxes (85.6 % of measured fluxes are produced or consumed in tropical ecosystems (Oertel et al., above MDF ± 0.002 nmol m−2 s−1 ). Our study highlights the 2016). Soil physical, chemical, and biological characteristics importance of adjusted closure time for each gas. are linked to variation in GHG emissions from soils, which in turn can display very high spatial and temporal variability (Arias-Navarro et al., 2017; Silver et al., 1999). Historically, soil GHG fluxes (emission or consumption) 1 Introduction have been measured using the static chamber method. This involves closing chambers manually for a known period of After water vapor, carbon dioxide (CO2 ), methane (CH4 ), time, usually 30–60 min, and repeated collection of air sam- and nitrous oxide (N2 O) are the three main greenhouse gases ples for further analysis via gas chromatography (Verchot et (GHGs) in terms of radiative forcing. Increases in these al., 1999, 2000). Fluxes are then computed from the change GHG concentrations in the atmosphere are driving anthro- Published by Copernicus Publications on behalf of the European Geosciences Union. 786 E. A. Courtois et al.: Automatic high-frequency measurements of full soil greenhouse gas fluxes in gas concentration per unit of time, per surface area en- 2 Methods closed by the chamber, and corrected by the volume of the chamber. While these labor-intensive and time-consuming 2.1 Measurement site manual measurements are well-adapted to capture high spa- tial flux variability (Arias-Navarro et al., 2017; Pumpanen et This study was conducted at the Paracou research station al., 2004), they do not capture high temporal variation, which (5◦ 150 N, 52◦ 550 W), located in the coastal area of French is necessary for the accurate estimation of annual GHG bud- Guiana, South America. The automated soil GHG flux sys- gets. Moreover, short-term transient spikes in the emission tem was deployed in the footprint of the Guyaflux site, which or consumption of these GHGs likely remains undetected holds a 55 m tall tower upon which canopy CO2 , H2 O, and with static chamber methods, imposing a lost opportunity to energy fluxes have been monitored since 2004 using the fully understand the production or consumption processes of eddy covariance technique (Aguilos et al., 2018; Bonal et al., GHGs and their response to rapidly changing environmental 2008). The site is covered with pristine tropical forest and conditions. One of the key challenges of contemporary GHG located in the northernmost part of the Guiana shield. It is flux research is to close these knowledge gaps in order to im- characterized by a succession of small elliptical hills rising prove the quantitative prediction of GHG fluxes (Merbold et to 10–40 m a.s.l., sometimes associated with plateaus of sim- al., 2015). ilar altitude. The use of automatic chambers is one approach to obtain The soils are mostly nutrient-poor Acrisols continuous estimation of soil GHG flux data at high tem- (FAO/ISRIC/ISSS, 1998) with pockets of sandy Ulti- poral frequency (several measurements per days) at various sols developed over a Precambrian metamorphic formation sampling points. Since the 1970s (Denmead, 1979), a vari- called the “Bonidoro series” and composed of schist and ety of technical solutions for automated flux sampling have sandstone, sporadically traversed by veins of pegmatite, been developed (Ambus et al., 2010; Breuer et al., 2000; aplite, and quartz (Bonal et al., 2008). The forest around the Görres et al., 2016; Kostyanovsky et al., 2018; O’Connell tower is characteristic of a pristine tropical forest with both et al., 2018; Petrakis et al., 2017a; Savage et al., 2014), par- high tree density (∼ 620 trees with a dbh > 10 cm ha−1 ) ticularly for soil CO2 fluxes. However, accurate detection of and species richness (∼ 140 species ha−1 ). The climate CH4 and N2 O fluxes from soils via flow through systems is highly seasonal due to the north–south movement of is more difficult than CO2 due to significantly lower back- the Inter-Tropical Convergence Zone. The wet season, ground concentrations and lower flux rates (Kostyanovsky et characterized by heavy rain events, lasts for 8 months al., 2018). The budgetary requirements for large infrastruc- (December–July) and alternates with a 4-month dry period ture and intensive maintenance compared to manual cham- (August–November) during which precipitation is typically ber measurements have prevented the widespread application lower than 100 mm month−1 . For the period 2004–2015, of automated systems. The use of automated and continuous annual rainfall quantities were on average 3103 mm yr−1 , methods to estimate full GHG budgets in situ remains scarce, relative extractable water (an index of soil water availability; especially in complex biomes with extreme climate such as Wagner et al., 2011) varied from 0.93 in the wet season to tropical forests. Therefore, only a few studies actually ad- 0.46 in the dry season, and soil temperature was on average dress the difficulties and challenges associated with operat- 25.1 ◦ C with little seasonal nor diurnal variation (Aguilos et ing these systems under field conditions (Görres et al., 2016; al., 2018). Koskinen et al., 2014). Recent technological advances have now made more au- 2.2 Automated sampling system tomated chamber systems commercially available, and an in- creasing number of custom-made systems are being designed A schematic view of the automatic sampling system is shown and deployed for soil GHG flux measurements (De Klein in Fig. 1a. The system consisted of four main components: and Harvey, 2012). Here, we present a detailed field deploy- 16 automated long-term chambers (8100-104, LI-COR Bio- ment of a custom-built automated soil GHG flux system – the sciences), a multiplexer to link one chamber at a time to the LI-8100A Soil CO2 Flux System (LI-COR Biosciences Inc., gas analyzers (LI-8150, LI-COR Biosciences), an infrared Lincoln, NE, USA) running in line with a Picarro G2308 (Pi- gas analyzer (IRGA) to measure CO2 concentrations (LI- carro Inc., Santa Clara, CA, USA). Using a 4-month dataset 8100A, LI-COR Biosciences), and a cavity ring-down spec- of continuous measurements of CO2 , CH4 , and N2 O fluxes troscopy (CRDS) instrument to measure CH4 and N2 O con- simultaneously under tropical forest conditions, we present centrations (G2308, Picarro) that was fitted with an exter- an optimized sampling protocol for the estimation of the full nal recirculation pump (A0702, Picarro). Both the IRGA and GHG budget in this ecosystem. CRDS systems were necessary to measure all three GHG concentrations due to the different abundances and flux rates of CO2 , CH4 , and N2 O. The IRGA methodology is accu- rate and precise enough to detect small CO2 concentration changes at high background concentrations (approximately Biogeosciences, 16, 785–796, 2019 www.biogeosciences.net/16/785/2019/ E. A. Courtois et al.: Automatic high-frequency measurements of full soil greenhouse gas fluxes 787 400 ppmv; parts per million in volume units). However, the concentration increase in the chamber, which was used to detection of small changes in CH4 and N2 O concentrations, quantify the flux of CO2 from the soil surface into the at- even at their low background atmospheric concentrations on mosphere (taking into account the enclosed soil surface area the order of 2000 ppbv (ppbv; parts per billion in volume and the total system volume). A subsampling loop was in- units) and 300 ppbv, respectively, requires higher accuracy serted after the analyzer (LI-8100A) and before the multi- and precision levels that can be detected with the CRDS. plexer (LI-8150) to pull the air sample through the Picarro Power supply was delivered through a 12 kVa generator G2308 CRDS analyzer for the determination of CH4 and (Perkins STORM15) fitted with batteries located 400 m away N2 O concentrations and flux estimations, before going back from the instruments. Both the CO2 analyzer control unit and to the chamber (Fig. 1a). All three gas concentrations were the multiplexer (LI-COR) had their own weather-proof cas- recorded every second over the sampling periods. ing, requiring no additional protection in the field. Nonethe- less, in consideration of the high precipitation at the site, 2.3 Flux calculations these devices were placed under a wooden shelter for added protection. The CH4 and N2 O analyzer (Picarro), its exter- All flux estimations were carried out using commercially nal pump, and a computer monitor were housed in a wa- available SoilFluxPro software (LI-COR Biosciences). An R terproof shelter that was specifically designed to host them script (Supplement) was created to merge all the Picarro files (Fig. 1c). The LI-8100 and the G2308 computers were con- from a given week in order to import them into the SoilFlux- nected through an ethernet connection to ensure time syn- Pro software. The Picarro creates one file per hour and when chronization. The 16 automated soil chambers (8100-104, Picarro files are not merged, SoilFluxPro software is not able LI-COR Biosciences) were installed in a grid in the for- to deal with measurements overlapping between two distinct est (Fig. 1b) covering in total an area of approximately Picarro files (e.g., when a single measurement is performed 300 m2 (15 m × 20 m). Each chamber was only closed dur- from 09:50 to 10:15 UTC−3), leading to incorrect estimation ing individual chamber measurement periods and was fully of CH4 and N2 O fluxes. To avoid underestimation of fluxes open when not sampling. The PVC collars that were pro- (Fig. S1 in the Supplement), CO2 , CH4 , and N2 O fluxes were vided with the 8100-104 automatic chambers were inserted measured as exponential fit of gas concentration with time in the soil 1 month prior to the first measurement (20.3 cm using SoilFluxPro software and include a 60 s dead band to inner diameter and 21.3 cm outer diameter; enclosed soil account for soil surface pressure disturbances due to the clos- area ∼ 318 cm2 ; insertion depth ∼ 7 cm; offset ∼ 4 cm; green ing of the chamber. PVC). When the chambers close, they are automatically low- 2.4 Minimum detectable fluxes ered so that they cover each soil collar and ensure a fully sealed chamber. The chamber lid does not directly rest on The minimum detectable flux (MDF) for each gas was esti- the collar rim, but on a metal plate surrounding the collar, mated by using a metric originally developed by Christiansen leaving the collar undisturbed and minimizing lateral leaks et al. (2015), which was modified by Nickerson (2016) (Hupp et al., 2009). to make it more suitable for high-frequency measurements The 16 chambers were connected via 15 m Bev-A-Line (Christiansen et al., 2015; Nickerson, 2016): tubing (8 mm inner diameter) with the multiplexer (LI-8150),    which allows for switching among each of the 16 chambers Aa VP in any given sequence. Soil temperature (0–10 cm) was mon- MDF = √ , (1) tc n SRT itored with 8100-201  thermistor probes (Omega Engineer- ing Inc., Stamford, CT, USA), and soil volumetric water con- where Aa is the analytical accuracy of the analyzer (25 ppb tent (0–10 cm) was monitored with 8100-202 ECH2 O model for N2 O and 10 ppb for CH4 with the Picarro G2308 and EC-5 soil moisture sensors (Decagon Devices Inc., Pullman, 600 ppb for CO2 with the LI-8100, recorded from the tech- WA, USA). Soil temperature and soil volumetric water con- nical data sheets of the analyzers), tc is the closure time of tent sensors were directly connected to the chambers and the chamber in seconds, n is the number of points that are recorded by the LI-COR system using the same time step. available to compute the flux (i.e., tc divided by the sampling Each chamber was purged for 15 s prior to each mea- periodicity, every 1 s in this study), V is the chamber volume surement and 45 s after each measurement in order to flush (0.0040761 m3 ), P is the atmospheric pressure (101 325 Pa), the lines and restore background gas levels in the system. S is the chamber surface area (0.03178 m2 ), R is the ideal gas The flow rate during the purging and the measurements was constant (8.314 m3 Pa K−1 mol−1 ), and T is the ambient tem- ∼ 2.8 L min−1 between the LI-8150 and the chambers, which perature (298.15 K). We computed the MDF of each gas for ensures sufficient air mixing in the chamber headspace dur- closure times from 2 to 30 min in order to select the optimal ing the measurements (Görres et al., 2016). Flow rates in the chamber closure time for each gas in our integrated system subsampling lines (LI-8100 and Picarro) were lower and set (Table 1). between 1.5 and 1.7 L min−1 as recommended by the man- ufacturers. The LI-8100 software provided the rate of CO2 www.biogeosciences.net/16/785/2019/ Biogeosciences, 16, 785–796, 2019 788 E. A. Courtois et al.: Automatic high-frequency measurements of full soil greenhouse gas fluxes Figure 1. Experimental design. (a) Schematic view of the installation composed of four main components: 16 automated long-term chambers (8100-104, LI-COR Biosciences), a multiplexer to link one of these chambers to the gas analyzers (LI-8150, LI-COR Biosciences), an in- frared gas analyzer (IRGA) to measure CO2 concentrations (LI-8100A, LI-COR Biosciences), and a cavity ring-down spectroscopy (CRDS) instrument to measure CH4 and N2 O concentrations (G2308, Picarro) that was fitted with an external pump. (b) Schematic representation of the grid with the shelter housing the equipment in the middle and the 16 chambers (grey dots) linked to the LI-8150 multiplexer with 15 m cables (black lines). (c) Picture of the instruments in the field. Table 1. Minimum detectable fluxes (MDFs) for each gas and for 2.5 Closure time closure times from 2 to 30 min. The two closure times that were used in this study (2 and 25 min) are highlighted in bold. Selecting the best length of time for soil GHG measurements Closure N2 O CH4 CO2 and accurate flux calculation in an integrated CO2 , CH4 , and time (nmol m−2 s−1 ) (nmol m−2 s−1 ) (nmol m−2 s−1 ) N2 O automated measurement system requires careful con- (minutes) sideration. At low fluxes, longer measurement periods are needed to reach reliable measurements of real concentration 2 0.100 0.040 2.393 5 0.025 0.010 0.605 changes, while at high fluxes possible storage and saturation 10 0.009 0.004 0.214 effects in the chamber headspace might result in nonlinear 15 0.005 0.002 0.117 concentration increases and thereby underestimated fluxes if 20 0.003 0.001 0.076 fluxes are calculated linearly. In order to maximize the de- 25 0.002 0.001 0.054 tectable percentage of fluxes for N2 O and CH4 without im- 30 0.002 0.001 0.041 peding spatial coverage and temporal resolution, we built a combined program with two different closure times. Each week, 4 out of 16 chambers were programmed to stay closed Biogeosciences, 16, 785–796, 2019 www.biogeosciences.net/16/785/2019/ E. A. Courtois et al.: Automatic high-frequency measurements of full soil greenhouse gas fluxes 789 for a longer measurement period to ensure a reliable esti- served only a slight saturation effect when using the LONG mation of low fluxes while the other 12 chambers were pro- closure time (Fig. 2). Variation in the flux calculations did not grammed to stay closed for a shorter period to capture diel differ between the SHORT and LONG chamber closure mea- variation and detect high fluxes. For the short closure time surements. N2 O flux calculations were much more variable (SHORT hereafter), we used a 2 min measurement period be- when measuring with the SHORT closure time compared to cause (1) this is a standard closure time for soil CO2 flux cal- the LONG closure time (Fig. 2). Even if fluxes were above culations (Janssens et al., 2000), (2) MDF for CO2 flux is typ- the detection limit, the low fluxes estimated with the SHORT ically low (Epron et al., 2006; Bonal et al., 2008; Bréchet et closure time were not reliable as shown by the low correla- al., 2009; Courtois et al., 2018), and (3) corresponding MDFs tion in Fig. 2. For both CH4 and N2 O, we therefore decided of CH4 (0.04 nmol m−2 s−1 ) and N2 O (0.1 nmol m−2 s−1 ) to apply the following quality check procedure and to discard are compatible with the detection of emission or consump- (1) all fluxes that were not complying with MDF criterion, tion peaks of these two gases in this region (Courtois et (2) all fluxes estimated with the SHORT closure time with a al., 2018; Petitjean et al., 2015). For the long closure time R 2 lower than 0.8 (Savage et al., 2014), and (3) all anomalous (LONG hereafter), we decided to use a 25 min measurement values (difference greater than 5 nmol m−2 s−1 with adjacent period in order to optimize the trade-off between a reliable measurements). estimation of low N2 O fluxes (Table 1) and a program length that allows for a sufficient number of flux measurements per chamber and per day. 3 Results and discussions We therefore programmed the multiplexer for 2.5 h cy- cles (9–10 measurements per chamber per day), which in- A cleaning frequency of once a week was necessary and suf- cluded four chambers with LONG measurements and 12 ficient to remove falling leaves and branches from the auto- chambers with SHORT measurements. Each week, the pro- matic chamber system, prevent leaks, and generate a continu- gram was modified manually so that the four LONG mea- ous dataset of soil GHG fluxes from this tropical forest. Tem- surements were rotated across the chambers. Each chamber perature variations are typically small below the canopy due was therefore measured with the LONG closure time for one to the shadowing by dense canopy crown and microclimatic 7-consecutive-day period per month (4 weeks). conditions. During the study period, temperature at 2 m in height varied from 22 ◦ C in the night to 28 ◦ C during the day. 2.6 System maintenance and data processing The presence of water condensation inside the tubing lines was carefully checked every week and never occurred dur- The automated sampling system was installed on 1 June 2016 ing the study period. The automatic chamber system worked and operated until 29 September 2016 (4 months), total- well most of the time, but some data gaps did exist. Over the ing 17 592 individual measurements for each gas (4098 with 17 592 individual flux estimations, 343 (1.9 %) had to be dis- LONG closure time and 13 494 with SHORT closure time). carded because of (1) problems in the connection between Coarse wood debris were removed weekly but small litter, the chamber and the multiplexer (154 measurements, 0.9 % such as leaves, fruits, and twigs, was left in the collar area. of data points) and (2) imperfect chamber closing, which was Every week, living plants growing inside the collars, and detected by an insufficient increase in CO2 (189 measure- the dead leaves on the chambers, were carefully removed by ments, 1 % of data points). hand. The R 2 value of the exponential increase in CO2 over 2 min was used as an indicator that the system was function- 3.1 CO2 fluxes ing correctly and not impeded by debris (Görres et al., 2016; Savage et al., 2014). When the R 2 of the regression between In addition to the 343 fluxes that were removed after time and CO2 concentration was lower than 0.9, we consid- the first steps of the quality check procedure, 758 CO2 ered this to be an indication that there may have been an issue flux estimations were also considered to be anomalous be- with the chamber closing and sealing correctly and removed cause the difference with adjacent measurements were ei- the flux measurement for all three gases from our analysis. ther greater than 5 µmol m−2 s−1 (758 measurements, i.e., For CO2 , we observed a strong concentration saturation 4.3 %) or lower than 0 µmol m−2 s−1 (14 measurements). effect when using the LONG closure time (25 min), leading In total, 16 477 CO2 fluxes over 17 592 (93.6 %) could be to an underestimation of fluxes (Fig. 2). All CO2 flux esti- used over the 4-month period. CO2 fluxes were on aver- mates were therefore based on 2 min regressions only, using age 8.1 ± 1.6 µmol m−2 s−1 (Table 2), which would corre- either full concentration measurements of the SHORT clo- spond to a mean annual soil CO2 efflux of 3050 gC m−2 yr−1 , sure time or the first 2 min of the LONG closure time. Fol- which falls into the upper range of the extensive review of lowing recommendations (Rubio and Detto, 2017), we re- mean annual soil CO2 efflux estimations in tropical forest moved anomalous values, i.e., CO2 fluxes estimation with a provided recently by Rubio and Detto (2017). Nonetheless, difference greater than 5 µmol m−2 s−1 with adjacent mea- our study period (June–September) only covered the end surements or lower than 0 µmol m−2 s−1 . For CH4 , we ob- of the wet season and more data are needed to make this www.biogeosciences.net/16/785/2019/ Biogeosciences, 16, 785–796, 2019 790 E. A. Courtois et al.: Automatic high-frequency measurements of full soil greenhouse gas fluxes Figure 2. Comparison of 2 and 25 min estimations for (a) CO2 , (b) CH4 , and (c) N2 O fluxes. For this, we used measurements made over 25 min and recomputed the flux with the first 2 min for 2 weeks (from 2 to 9 August in black and from 16 to 25 August in grey) covering the whole range of fluxes during the study period. All fluxes were computed using exponential fit. The dashed line represents the 1 : 1 line while the solid grey line represents the linear regression between 2 and 25 min estimations (R 2 values of these regressions are indicated in each panel). Table 2. Mean, standard deviation (SD), minimum (Min), and maximum (Max) values of each gas and each chamber over the study period. These values are computed using all flux estimations (with either SHORT or LONG closure times) remaining after quality check. The number (N) of fluxes that were used is also indicated for each chamber. The last line of the table is the mean of all fluxes by chamber by gas and the min and max for all chambers by gas. CO2 (µmol m−2 s−1 ) CH4 (nmol m−2 s−1 ) N2 O (nmol m−2 s−1 ) Mean SD Min Max N Mean SD Min Max N Mean SD Min Max N Chamber 1 7.19 0.93 2.14 10.81 940 10.97 7.73 −2.08 28.79 840 0.10 0.12 −0.48 0.70 284 Chamber 2 7.60 1.11 4.00 12.21 1166 −1.62 1.75 −4.09 11.68 899 0.00 0.14 −1.03 0.75 285 Chamber 3 5.58 0.99 2.11 11.12 1135 0.35 2.95 −2.48 22.94 745 0.03 0.23 −0.61 2.85 208 Chamber 4 7.94 1.37 4.36 12.13 1154 −1.85 1.23 −3.63 6.09 1105 0.04 0.10 −0.66 0.60 224 Chamber 5 4.14 0.92 0.53 10.05 1139 1.37 3.26 −2.20 12.61 752 0.15 0.33 −1.04 3.23 382 Chamber 6 8.87 1.70 3.36 17.68 1070 −1.38 1.78 −3.20 8.04 801 −0.02 0.12 −1.04 0.63 272 Chamber 7 13.47 2.78 0.89 22.12 988 1.37 3.60 −2.63 19.56 749 0.64 1.37 −0.85 7.93 216 Chamber 8 7.44 1.19 2.03 11.02 1099 0.03 2.96 −3.37 18.47 785 0.02 0.15 −1.36 0.84 202 Chamber 9 4.25 1.20 0.44 11.37 1002 2.06 3.13 −2.14 11.53 879 0.02 0.11 −0.62 0.58 332 Chamber 10 5.60 1.30 0.69 13.13 1037 1.21 2.46 −1.91 10.34 657 0.04 0.13 −0.64 0.77 252 Chamber 11 11.97 2.19 6.84 18.78 1004 6.72 7.61 −1.06 41.49 855 0.03 0.17 −1.01 1.04 199 Chamber 12 9.42 2.70 3.45 21.54 968 1.40 6.68 −3.29 41.94 891 0.02 0.09 −0.75 0.30 204 Chamber 13 5.85 1.34 0.42 8.49 944 5.29 5.92 −4.60 26.64 654 0.10 0.19 −0.84 1.71 335 Chamber 14 5.66 1.15 0.72 10.72 987 2.78 6.22 −2.48 35.15 691 0.09 0.17 −0.63 0.93 231 Chamber 15 16.63 3.27 9.42 29.64 850 −0.46 2.05 −3.25 8.26 839 −0.02 0.16 −0.96 0.72 185 Chamber 16 7.35 1.13 3.98 11.37 994 −1.34 1.48 −3.60 6.11 843 0.00 0.11 −1.00 0.83 187 8.06 1.58 0.42 29.64 16 477 1.68 3.80 −4.60 41.94 12 985 0.08 0.23 −1.36 7.93 3998 Biogeosciences, 16, 785–796, 2019 www.biogeosciences.net/16/785/2019/ E. A. Courtois et al.: Automatic high-frequency measurements of full soil greenhouse gas fluxes 791 Figure 3. Distribution of fluxes: histogram of (a) CO2 , (b) CH4 , and (c) N2 O fluxes over the study period. For (b) and (c), the dotted line represents null fluxes. estimation more precise. All 2 min measurements of CO2 clearly demonstrated a strong spatial variability in soil respi- fluxes from the 4-month study period were above the MDF ration, even at a low spatial scale (Fig. 5, Table 2), with some of 2.39 nmol m−2 s−1 for the LI-8100 analyzer (Table 1). local spots clearly displaying stronger values of soil respira- No saturation effect was detected using the SHORT closure tion during the study period. time, and estimation of CO2 over a shorter time period is not recommended (Davidson et al., 2002). CO2 fluxes using 3.2 CH4 fluxes the LONG closure time would be underestimated due to the buildup of high CO2 concentrations due to large fluxes over In addition to the 343 fluxes that were removed after the first this long time period (Fig. 2), and are not recommended. For steps of the quality check procedure, CH4 flux estimations small chambers like the one used in this study, we therefore were also discarded because of (1) problems with Picarro conclude that a 2 min sampling time including a dead band of files (12 measurements), (2) application of the MDF crite- 60 s should be used for CO2 flux calculations since the MDF rion (137 measurements), (3) application of the R 2 criterion of this short measurement period allowed for the retention for SHORT closure time (3751 measurements, i.e., 28 % of of 100 % of the data. When the chambers stay closed longer the SHORT measurements), and (4) detection of anomalous for accurate detection of N2 O and CH4 fluxes, only the first values (364 measurements). In total, 12 985 CH4 fluxes over 2 min of data should be used for CO2 flux calculations. 17 592 (73.8 %) could be used over the 4-month period. No The use of 16 automated flux chambers allowed for the saturation effect was detected using the LONG closure time capture of spatial and temporal variability in soil respira- and fluxes estimated with the SHORT closure time were very tion. Over this 4-month period, corresponding to the end well-correlated to fluxes using the LONG closure time, even of the wet season in French Guiana, temporal variability for small fluxes (Fig. 2). Totals of 68.4 % and 98.2 % of remained low (Fig. 4). This dataset is therefore not long fluxes measured with the SHORT and LONG closure times, enough to detect the seasonal variation in soil respiration respectively, were retained in our quality control data pro- that was highlighted in previous study (Rowland et al., 2014; cessing over the 4-month study period. These measurement Rubio and Detto, 2017). We did found that soil respiration periods, therefore, allowed for the retention of a large major- tended to decrease in very humid soils (Fig. S2 in the Sup- ity of CH4 emission or consumption fluxes in our data anal- plement) as highlighted previously at the same site (Rowland ysis. et al., 2014) but more data are needed to precisely disentan- CH4 fluxes were on average 1.7±3.8 nmol m−2 s−1 with a gle the importance of seasonal and diurnal variability from high variability among chambers (Table 2) but the frequency the responses to environmental triggers of soil respiration. of negative CH4 fluxes (consumption, 59 % of fluxes) was Nonetheless, even during this relatively short period, our data greater than for positive fluxes (emission, 41 % of fluxes) www.biogeosciences.net/16/785/2019/ Biogeosciences, 16, 785–796, 2019 792 E. A. Courtois et al.: Automatic high-frequency measurements of full soil greenhouse gas fluxes Figure 4. CO2 fluxes through time: CO2 fluxes for each chamber (1 to 16) over the study period with fluxes estimated with a SHORT (2 min) closure time in black and fluxes estimated with the first 2 min of the LONG (25 min) closure time in grey. All panels have the same limits on the y axis (from 0 to 25 µmol m−2 s−1 ). during this period (Fig. 3). Most of the time, soils were ei- ther consuming or emitting small amounts of CH4 , but large transient emission peaks were periodically detected at indi- vidual chamber locations during the study period (Fig. 6). Tropical soils are generally considered to be a sink at a yearly basis (Dutaur and Verchot, 2007) but it is known that these soils can shift from a source in the wet season to a sink in the dry season (Courtois et al., 2018; Davidson et al., 2008; Teh et al., 2014). No clear temporal trend could be detected during the study period and there was a slight correlation of CH4 fluxes with surface soil humidity (higher fluxes at in- termediate soil humidity, Fig. S2). Longer time series cov- ering at least a full year are needed to explore the seasonal and diurnal variability in fluxes. As highlighted previously in French Guiana (Courtois et al., 2018), spatial variabil- ity in CH4 emission was high, even at a small spatial scale (Figs. 5, 6). Interestingly, some spots clearly displayed high CH4 emissions during the whole study period (Figs. 5, 6). 3.3 N2 O fluxes In addition to the 343 fluxes that were removed after the first steps of the quality check procedure, N2 O flux estimations were also discarded because of (1) problems with Picarro files (12 measurements), (2) application of the MDF crite- rion (1594 measurements), (3) application of the R 2 criterion for SHORT closure time (11 643 measurements, i.e., 28 % Figure 5. Mean values per day for (a) CO2 , (b) CH4 , and (c) N2 O of the SHORT measurements), and (4) detection of anoma- fluxes over the study period. Each chamber is represented by a dis- lous values (364 measurements). In total, 3998 N2 O fluxes tinct color. over 17 592 measurements (22.7 %, 140 measurements with Biogeosciences, 16, 785–796, 2019 www.biogeosciences.net/16/785/2019/ E. A. Courtois et al.: Automatic high-frequency measurements of full soil greenhouse gas fluxes 793 Figure 6. CH4 fluxes through time: CH4 fluxes for each chamber (1 to 16) over the study period with fluxes estimated with a SHORT (2 min) closure time in black and fluxes estimated with a LONG (25 min) closure time in grey. The dotted line displays the zero flux line. All panels have the same limits on the y axis (from −5 to 30 nmol m−2 s−1 ). Figure 7. N2 O fluxes through time: N2 O fluxes for each chamber (1 to 16) over the study period with fluxes estimated with the SHORT (2 min) closure time in black and fluxes estimated with the LONG (25 min) closure time in grey. The dotted line displays the zero flux line. Due to the high differences among chambers, each panel has a specific limit on the y axis. www.biogeosciences.net/16/785/2019/ Biogeosciences, 16, 785–796, 2019 794 E. A. Courtois et al.: Automatic high-frequency measurements of full soil greenhouse gas fluxes the SHORT and 3858 measurements with the LONG clo- are sufficiently high (9 to 10 measurements per day) to cap- sure time) could be used over the 4-month period. A total ture potential diurnal variation (Nicolini et al., 2013; Rubio of 94.1 % of fluxes measured with the LONG closure times and Detto, 2017) in the three gases with good spatial replica- were retained after our quality control data processing over tion (16 chambers); (3) periodic extreme events of high N2 O the 4-month study period. When measured over 25 min, N2 O fluxes can still be detected with the SHORT closure time pe- fluxes at our site could therefore be considered reliable. Us- riod, which occurs at a higher frequency than the LONG clo- ing the SHORT closure time, most flux estimations had to sure measurements. Our study underlines the importance of be discarded because they led to unreliable flux estimations appropriate closure time for each GHG gas for accurate esti- (Fig. 2). Nonetheless, the SHORT closure time still allowed mation of GHG budgets. This information is crucial for the the detection of high N2 O emission or consumption events calculation of accurate soil fluxes at diurnal time steps and than were detected during the study period (Figs. 5 and 7). for the estimation of annual GHG budgets. This combination N2 O fluxes were on average 0.1 ± 0.2 nmol m−2 s−1 with of automated closed dynamic chambers and advanced GHG a high variability among chambers (Table 2). At the same analyzers makes it possible to (1) account for short-term chamber, N2 O flux can shift from consumption to emission variability in GHG fluxes while taking into account spatial with 28 % of fluxes indicating a sink and 72 % a source for variability, (2) estimate annual GHG budgets at these loca- N2 O (Fig. 3). The high variability in N2 O fluxes that we de- tions, (3) tracking the variability in GHG fluxes along hours, tected over 4 months with our automated system is in agree- days, seasons, and years, and (4) study the impact of climatic ment with the typical high variability in N2 O fluxes mea- change on soil GHG budgets. sured from tropical soils over space and time using static chambers (Arias-Navarro et al., 2017; Courtois et al., 2018). Moreover, N2 O fluxes did not show any relationship with Data availability. Data used in this study can be freely accessed surface soil humidity (Fig. S2), which underlines the com- through the Zenodo repository: https://doi.org/10.5281/zenodo. plexity of the biological process underlying these fluxes. In 2555299 (Courtois and Stahl, 2019). a previous study in the same environment (Courtois et al., 2018), we estimated that the minimum detectable flux us- ing gas chromatography analysis of four discrete gas sam- Supplement. The supplement related to this article is available online at: https://doi.org/10.5194/bg-16-785-2019-supplement. ples over 30 min for N2 O was ±8.3 µg N m−2 h−1 . MDF es- timated in the present study using high-frequency measure- ment was 0.002 nmol m−2 s−1 or 0.2 µg N m−2 h−1 for N2 O Author contributions. JVB and NA designed the experiment and which is therefore ∼ 40 times lower. Such result indicates EAC, CS, BB, and DB carried it out. EAC and CS prepared the pa- that this long-term system is well-adapted to capture and es- per with contributions from all co-authors. EAC and CS contributed timate the low N2 O fluxes occurring in this ecosystem. equally to the paper. 4 Conclusions Competing interests. The authors declare that they have no conflict of interest. We demonstrated here that the combination of a commer- cial soil GHG chamber system – the LI-8100A Automated Soil CO2 Flux System – running in parallel with a Picarro Acknowledgements. This research was supported by the European G2308, enables the continuous long-term measurement of Research Council Synergy grant ERC-2013-SyG 610028- CO2 , CH4 , and N2 O simultaneously under tropical con- IMBALANCE-P. We thank Jan Segers for help in the initial setting ditions. Similar configurations have been recently imple- of the system and Renato Winkler from Picarro and Rod Madsen mented in temperate climate (Petrakis et al., 2017a, b), but and Jason Hupp from LI-COR for their help in combining the systems. We thank the staff of Paracou station, managed by UMR to our knowledge, this is the first time that this experimental Ecofog (CIRAD, INRA; Kourou), which received support from setup is fully described and tested under tropical field con- “Investissement d’Avenir” grants managed by Agence Nationale ditions for the measurement of the three soil GHG fluxes si- de la Recherche (CEBA: ANR-10-LABX-25-01, ANAEE-France: multaneously. Additionally, our study determined the opti- ANR-11-INBS-0001). This study was conducted in collaboration mal chamber closure time for each GHG. The sampling sys- with the Guyaflux program belonging to SOERE F-ORE-T, tem of SHORT and LONG closure times with a weekly ro- which is supported annually by Ecofor, Allenvi, and the French tation presented here has three major advantages, which ulti- national research infrastructure, ANAEE-F. This program also mately can provide high confidence in the annual estimation received support from an “investissement d’avenir” grant from of the full GHG budgets of tropical soils: (1) the LONG clo- the Agence Nationale de la Recherche (CEBA, ref ANR-10- sure time allows a reliable estimation of the low N2 O fluxes LABX-25-01). Ivan August Janssens acknowledges support from in this ecosystem, which was clearly not achieved using a Antwerp University (Methusalem funding), Nicola Arriga from ICOS-Belgium and Fonds Wetenschappelijk Onderzoek (FWO), shorter closure time; (2) the number of data points per day Biogeosciences, 16, 785–796, 2019 www.biogeosciences.net/16/785/2019/ E. A. Courtois et al.: Automatic high-frequency measurements of full soil greenhouse gas fluxes 795 and Jennifer Larned Soong from the U.S. Department of Energy Denmead, O.: Chamber systems for measuring nitrous oxide emis- under contract DE-AC02-05CH11231. sion from soils in the field, Soil Sci. Soc. Am. J., 43, 89–95, 1979. Edited by: Kees Jan van Groenigen Dutaur, L. and Verchot, L. V.: A global inventory Reviewed by: three anonymous referees of the soil CH4 sink, Glob. Biogeochem. Cy., 21, https://doi.org/10.1029/2006GB002734, 2007. Epron, D., Bosc, A., Bonal, D., and Freycon, V.: Spatial variation of soil respiration across a topographic gradient in a tropical rain References forest in French Guiana, J. Trop. Ecol., 22, 565–574, 2006. FAO/ISRIC/ISSS: World Reference Base for Soil Resources.FAO, Aguilos, M., Hérault, B., Burban, B., Wagner, F., and Bonal, D.: ISRIC, ISSS, World Soil Resources Reports 84, Rome, 1998. What drives long-term variations in carbon flux and balance in Görres, C.-M., Kammann, C., and Ceulemans, R.: Automation of a tropical rainforest in French Guiana?, Agr. Forest Meteorol., soil flux chamber measurements: potentials and pitfalls, Biogeo- 253, 114–123, 2018. sciences, 13, 1949–1966, https://doi.org/10.5194/bg-13-1949- Ambus, P., Skiba, U., Drewer, J., Jones, S., Carter, M. S., Albert, 2016, 2016. K. R., and Sutton, M.: Development of an accumulation-based Hupp, J. R., Garcia, R. L., Madsen, R., and McDermitt, D. K.: Mea- system for cost-effective chamber measurements of inert trace surement of CO2 evolution in a multiplexed flask system, Amer. gas fluxes, Eur. J. Soil Sci., 61, 785–792, 2010. Soc. Horticultural Science, Alexandria, USA, Vol. 44, 1143– Arias-Navarro, C., Díaz-Pinés, E., Klatt, S., Brandt, P., Rufino, M. 1143, 2009. C., Butterbach-Bahl, K., and Verchot, L.: Spatial variability of Janssens, I. A., Kowalski, A. S., Longdoz, B., and Ceulemans, R.: soil N2 O and CO2 fluxes in different topographic positions in a Assessing forest soil CO2 efflux: an in-situ comparison of four tropical montane forest in Kenya, J. Geophys. Res.-Biogeo., 122, techniques, Tree Physiol., 20, 23–32, 2000. 514–527, 2017. Koskinen, M., Minkkinen, K., Ojanen, P., Kämäräinen, M., Lau- Bonal, D., Bosc, A., Ponton, S., Goret, J. Y., Burban, B., Gross, P., rila, T., and Lohila, A.: Measurements of CO2 exchange with Bonnefond, J., Elbers, J., Longdoz, B., and Epron, D.: Impact of an automated chamber system throughout the year: challenges severe dry season on net ecosystem exchange in the Neotropical in measuring night-time respiration on porous peat soil, Biogeo- rainforest of French Guiana, Glob. Change Biol., 14, 1917–1933, sciences, 11, 347–363, https://doi.org/10.5194/bg-11-347-2014, 2008. 2014. Bréchet, L., Ponton, S., Roy, J., Freycon, V., Coûteaux, M.-M., Kostyanovsky, K., Huggins, D., Stockle, C., Waldo, S., and Lamb, Bonal, D., and Epron, D.: Do tree species characteristics influ- B.: Developing a flow through chamber system for automated ence soil respiration in tropical forests? A test based on 16 tree measurements of soil N2 O and CO2 emissions, Measurement, species planted in monospecific plots, Plant Soil, 319, 235–246, 113, 172–180, 2018. 2009. Merbold, L., Wohlfahrt, G., Butterbach-Bahl, K., Pilegaard, K., Breuer, L., Papen, H., and Butterbach-Bahl, K.: N2 O emission from DelSontro, T., Stoy, P., and Zona, D.: Preface: Towards a full tropical forest soils of Australia, J. Geophys. Res.-Atmos., 105, greenhouse gas balance of the biosphere, Biogeosciences, 12, 26353–26367, 2000. 453–456, https://doi.org/10.5194/bg-12-453-2015, 2015. Christiansen, J. R., Outhwaite, J., and Smukler, S. M.: Compari- Nickerson, N.: Evaluating gas emission measurements using Min- son of CO2 , CH4 and N2 O soil-atmosphere exchange measured imum Detectable Flux (MDF), Eosense Inc., Dartmouth, Nova in static chambers with cavity ring-down spectroscopy and gas Scotia, Canada, 2016. chromatography, Agr. Forest Meteorol., 211, 48–57, 2015. Nicolini, G., Castaldi, S., Fratini, G., and Valentini, R.: A literature Courtois, E. A. and Stahl, C.: Dataset from: Automatic overview of micrometeorological CH4 and N2 O flux measure- high-frequency measurements of full soil greenhouse ments in terrestrial ecosystems, Atmos. Environ., 81, 311–319, gas fluxes in a tropical forest Biogeosciences 2019, 2013. https://doi.org/10.5281/zenodo.2555299, 2019. O’Connell, C. S., Ruan, L., and Silver, W. L.: Drought Courtois, E. A., Stahl, C., Van den Berge, J., Bréchet, L., Van Lan- drives rapid shifts in tropical rainforest soil biogeochem- genhove, L., Richter, A., Urbina, I., Soong, J. L., Peñuelas, J., istry and greenhouse gas emissions, Nat. Commun., 9, 1348, and Janssens, I. A.: Spatial variation of soil CO2 , CH4 and N2 O https://doi.org/10.1038/s41467-018-03352, 2018. fluxes across topographical positions in tropical forests of the Oertel, C., Matschullat, J., Zurba, K., Zimmermann, F., and Erasmi, Guiana Shield, Ecosystems, 21, 1445–1458, 2018. S.: Greenhouse gas emissions from soils-A review, Chem. Erde- Davidson, E., Savage, K., Verchot, L., and Navarro, R.: Minimiz- Geochem., 76, 327–352, 2016. ing artifacts and biases in chamber-based measurements of soil Petitjean, C., Hénault, C., Perrin, A.-S., Pontet, C., Metay, A., respiration, Agr. Forest Meteorol., 113, 21–37, 2002. Bernoux, M., Jehanno, T., Viard, A., and Roggy, J.-C.: Soil N2 O Davidson, E. A., Nepstad, D. C., Ishida, F. Y., and Brando, P. M.: emissions in French Guiana after the conversion of tropical for- Effects of an experimental drought and recovery on soil emis- est to agriculture with the chop-and-mulch method, Agr. Ecosyst. sions of carbon dioxide, methane, nitrous oxide, and nitric oxide Environ., 208, 64–74, 2015. in a moist tropical forest, Glob. Change Biol., 14, 2582–2590, Petrakis, S., Seyfferth, A., Kan, J., Inamdar, S., and Vargas, R.: In- 2008. fluence of experimental extreme water pulses on greenhouse gas De Klein, C. and Harvey, M.: Nitrous oxide chamber methodology emissions from soils, Biogeochemistry, 133, 147–164, 2017a. guidelines, Ministry for Primary Industries: Wellington, New Zealand, 2012. www.biogeosciences.net/16/785/2019/ Biogeosciences, 16, 785–796, 2019 796 E. A. Courtois et al.: Automatic high-frequency measurements of full soil greenhouse gas fluxes Petrakis, S., Barba, J., Bond-Lamberty, B., and Vargas, R.: Using Silver, W. L., Lugo, A., and Keller, M.: Soil oxygen availability greenhouse gas fluxes to define soil functional types, Plant Soil, and biogeochemistry along rainfall and topographic gradients in 1–10, 2017b. upland wet tropical forest soils, Biogeochemistry, 44, 301–328, Pumpanen, J., Kolari, P., Ilvesniemi, H., Minkkinen, K., Vesala, T., 1999. Niinistö, S., Lohila, A., Larmola, T., Morero, M., and Pihlatie, Teh, Y. A., Diem, T., Jones, S., Huaraca Quispe, L. P., Baggs, M.: Comparison of different chamber techniques for measuring E., Morley, N., Richards, M., Smith, P., and Meir, P.: Methane soil CO2 efflux, Agr. Forest Meteorol., 123, 159–176, 2004. and nitrous oxide fluxes across an elevation gradient in Rowland, L., Hill, T. C., Stahl, C., Siebicke, L., Burban, B., the tropical Peruvian Andes, Biogeosciences, 11, 2325–2339, Zaragoza-Castells, J., Ponton, S., Bonal, D., Meir, P., and https://doi.org/10.5194/bg-11-2325-2014, 2014. Williams, M.: Evidence for strong seasonality in the carbon stor- Verchot, L. V., Davidson, E. A., Cattânio, H., Ackerman, I. L., Er- age and carbon use efficiency of an Amazonian forest, Glob. ickson, H. E., and Keller, M.: Land use change and biogeochem- Change Biol., 20, 979–991, 2014. ical controls of nitrogen oxide emissions from soils in eastern Rubio, V. E. and Detto, M.: Spatiotemporal variability of soil res- Amazonia, Global Biogeochem. Cy., 13, 31–46, 1999. piration in a seasonal tropical forest, Ecol. Evol., 7, 7104–7116, Verchot, L. V., Davidson, E. A., Cattânio, J. H., and Ackerman, I. L.: 2017. Land-use change and biogeochemical controls of methane fluxes Savage, K., Phillips, R., and Davidson, E.: High temporal fre- in soils of eastern Amazonia, Ecosystems, 3, 41–56, 2000. quency measurements of greenhouse gas emissions from soils, Wagner, F., Hérault, B., Stahl, C., Bonal, D., and Rossi, V.: Mod- Biogeosciences, 11, 2709–2720, https://doi.org/10.5194/bg-11- eling water availability for trees in tropical forests, Agr. Forest 2709-2014, 2014. Meteorol., 151, 1202–1213, 2011. Biogeosciences, 16, 785–796, 2019 www.biogeosciences.net/16/785/2019/