Drought stress and tree size determine stem CO 2 efflux in a tropical forest

Summary CO 2 efflux from stems (CO 2_stem) accounts for a substantial fraction of tropical forest gross primary productivity, but the climate sensitivity of this flux remains poorly understood. We present a study of tropical forest CO 2_stem from 215 trees across wet and dry seasons, at the world's longest running tropical forest drought experiment site. We show a 27% increase in wet season CO 2_stem in the droughted forest relative to a control forest. This was driven by increasing CO 2_stem in trees 10–40 cm diameter. Furthermore, we show that drought increases the proportion of maintenance to growth respiration in trees > 20 cm diameter, including large increases in maintenance respiration in the largest droughted trees, > 40 cm diameter. However, we found no clear taxonomic influence on CO 2_stem and were unable to accurately predict how drought sensitivity altered ecosystem scale CO 2_stem, due to substantial uncertainty introduced by contrasting methods previously employed to scale CO 2_stem fluxes. Our findings indicate that under future scenarios of elevated drought, increases in CO 2_stem may augment carbon losses, weakening or potentially reversing the tropical forest carbon sink. However, due to substantial uncertainties in scaling CO 2_stem fluxes, stand‐scale future estimates of changes in stem CO 2 emissions remain highly uncertain.


Diurnal measurements of CO2_stem
During October 2013, 20 trees of the dominant genera (Trees shown in Table 1 in Rowland et al 2015) across both plots were fitted with automated open path stem respiration system. A large or small transparent acrylic chamber (large: volume (with foam and tubing) = 676 cm 3 , surface area =126 cm 2 ; small: volume (with foam and tubing) = 420 cm 3 , surface area = 90 cm 2 ) was attached to each tree, using a system of straps, as described in the methods section.
The chamber was connected to a large storage box which acted as a 145 L buffer volume to minimise fluctuations in CO2 concentration in the delivery air. Air was pumped from the buffer volume to the chamber and from the chamber using two 1.8 L min -1 micro diaphragm air pumps, these pumps were connected to manual flow regulators to regulate the flow to between 1-1.3 L min -1 . A CIRAS1 IRGA (PP Systems, Hitchen, UK), was then used in an open path system to sample the coming from the buffer and going into the chamber (reference), and the air coming out of the chamber (sample, Hutchinson & Livingston 1993). The IRGA was connected to a CR1000 data logger (Campbell Scientific, Logan, USA) which recorded data from the reference and the sample tubes every 15 seconds, alongside air temperature and atmospheric pressure. Respiration was calculated as the difference in CO2 between the reference and the sample lines (moles) multiplied by the moles of gas through the chamber per second, calculated using temperature, pressure and the universal gas constant. The respiration flux was then divided by chamber surface area and converted to μmol m -2 s -1 . The CR1000 also monitored wood temperature from a type T thermocouple inserted into the sapwood of each tree. Two respiration systems were created and left on each tree for a 24 hour period. After installation each system was tested for leaks following the same method listed above. All data was averaged into hourly intervals for each plot. Hours with data for <16 trees were removed, these hours tended to be between 0900 hrs and 1300 hrs when the chambers were changed over. Due to problems with high humidity during wet season, we were unable to successfully use this equipment in the wet season, hence our estimates are based off dry season only measurements. However during the dry season we found limited diurnal variation in CO2_stem.
On the control plot we found a slight decline (mean -3.5±10.7%) in the measurements between 9 am and 7 pm (Fig. S1a). However on the TFE plot we found very limited diurnal variation in CO2_stem (Fig. S1b), suggesting there was very limited influence of time of day on our measurements during the dry season. Due to very high humidity and water in equipment, data quality was not high enough to use the wet season data, so we assume the same diurnal stability during wet and dry season.

Measurements of sapwood depth
In October 2013 tree cores between 15-20 cm in length were extracted at 1.3 m from 21 trees on the TFE and control plot (trees shown in Table 1 in Rowland et al 2015), using an increment borer. In 18 of these tree cores there was a visible change in colour between the sapwood and the heartwood. The sapwood depth in these trees were measured and the (linear) relationships between sapwood depth and diameter, and sapwood area and basal area were plotted (see Fig.   S2).

Scaling CO2_stem
Method 1: Scaling by surface area The average wet and dry season CO2_stem values were multiplied by the total stem surface area per plot, calculated using the following equation S1 from Chambers et al., 2004. 10^ 0.105 0.686 * 10 2.208 * 10 0.627 * 10 Equations S1 Where SA is surface area in m 2 and dbh is diameter at breast height (1.3 m). These scaling equations are based on highly simplified tree forms, and may not accurately represent the diversity of branching structures which exists in tropical forests.
Method 2: Scaling by sapwood volume Using a relationship between sapwood depth and diameter (Fig. S1, see above for methods) sapwood area was calculated using basal area, calculated from dbh measured on our sample trees in December 2015. Sapwood area was used to convert CO2_stem to μmol m -3 s -1 . Using the same basal area-sapwood area relationship applied to all trees we calculated that sapwood area comprised on average 34% of basal area at breast height (1.3 m). Firstly, we assumed that this was fixed across the whole tree volume, which was calculated using the tropical moist forest equations from Chave et al. (2005). The small diameter parts of the tree, particularly branches, are however, likely to comprise greater amounts of sapwood and therefore 34% is likely to be a substantial underestimate. Given that there are no estimates of the proportion of total tree volume that is sapwood within tropical trees, we tested how sensitive our flux values were to an increase of sapwood volume to 50% and 80% of total volume. than we estimate here (Fig. 1b), and this may be because of a limited sample size (only 21 trees per plot). However we use these data to support our finding that differences in CO2_stem between the control and the TFE plots are consistently elevated with respect to the control plot during the wet season, and more equal during the dry season (Fig. 1b). Figure S1: