This study was carried out within an area of lowland, old growth dipterocarp rainforest in the Maliau Basin Conservation Area, Sabah, Malaysia (4° 44' 35" to 55" N and 116° 58' 10" to 30" E; mean annual rainfall 2838 mm ± 93 mm). On the 20^th of July 2017, there was a storm at the study site, which generated winds speeds of 8.4 m/s (Fig. S1). These were among the strongest winds normally experienced in inland forests of the region, which placed extreme sheer stress on trees. Consequently, a large number of trees fell within the same 24-hour period in the study location. Ten tree-fall gaps (mean length: 32 m ± 2.8, mean width: 24.5 m ± 3; see table S1 for gap characteristics) created during this event were selected for use in this investigation, along with ten adjacent closed canopy sites, located 20 m from the edge of each gap. We took 10 hemispherical photos in each gap and closed canopy sites to quantity canopy openness at each location.

Decomposition assay

In October 2017, we established a wood decomposition assay. Using a termite suppression experiment combined with macroinvertebrate accessible and inaccessible mesh bags, Griffiths et al. (2019) demonstrated that non-termite macroinvertebrates did not contribute significantly to the decomposition of a standardised wood substrate, Pinus radiata blocks, at this site. Therefore, to assess the rate of decomposition within these paired gap and closed canopy sites and determine the relative contributions of termites versus microbes to the process, we used the same assay of mass loss from untreated P. radiata wood within macroinvertebrate accessible and inaccessible bags. Wood blocks (9 x 9 x 5 cm, 161.2 ± 1.3 g; wood density of 0.40 g cm^-3 [Zanne et al., 2009]; wood C:N ratio of 462 ± 48 [Ganjegunte, Condron, Clinton, Davis, & Mahieu, 2004]) were dried at 60 ºC until they reached a constant weight and placed inside “open” (accessible to macroinvertebrates, principally termites, and microbes), or “closed” (accessible to microbes only) bags, which were all made with 300 micron nylon mesh (Plastok™, Merseyside, UK). The open woodblocks had ten 1 cm holes cut into the top and bottom of the bags to avoid confounding effects of using mesh of different sizes in decomposition assays (Stoklosa et al., 2016). The edges of the closed bags were folded several times and sealed with staples to prevent access by invertebrates. In each gap and closed canopy site, we ran a 50-m transect and randomly placed 5 open and 5 closed wood blocks 5 m apart along the transect (n = 10 per site; n = 200 woodblocks in total: 10 x forest sites x 2 canopy treatments [closed canopy or gap] x 2 mesh treatments [open or closed] x 5 replicates). Because the gaps were irregular in shape (Appendix table S1), we placed the transects along the longest axis of each gap. In all but one of the gap sites, we were unable to establish a 50 m transect, therefore, we placed an additional line perpendicular to the first, ensuring that each block was always at least 5 m apart from its nearest neighbouring block (Fig. 1).

A hemispherical photograph was taken by placing an iPhone 6 with a fisheye lens attachment directly on top of each wood block. Photographs were analysed using the function Hemiplot in R to calculate canopy openness, which was twice as high within the gaps compared with closed canopy sites (t = 9.67, P < 0.001, mean openness in gap sites = 0.24 ± 0.03; mean openness in closed canopy sites = 0.12 ± 0.02; Fig. S2). When placing the woodblocks, the top layer of leaf litter was removed, and the blocks were put directly on the humus layer. Wood blocks were left on the forest floor for 12 months (October 2017 to October 2018), after which they were collected and dried at 60 ºC until they reached a constant weight. Once dried, wood material was separated from termite soil. The remaining deadwood and termite material (carton and soil) was then re-weighed separately to calculate the proportion of mass loss from each block and the mass of soil brought into the mesh bags by termites. Given that termites are the only invertebrates known to translocate soil into deadwood (Oberst, Lai, & Evans, 2016), the mass of soil moved into the experimental woodblocks provides additional information on the termite activity compared with non-termite wood-feeding invertebrates.

Soil conditions and termite communities

Every month for the 12-month duration of the study, soil moisture percentage and soil temperature were measured within 5 cm of each wood block using a Delta-T Devices HH2 moisture metre (precise to 0.01 %) and a digital soil thermometer. Measurements were taken in dry conditions, between 8 AM and 10 AM.  To assess termite communities located within the gap and closed canopy sites, we carried out termite transects in September 2018 using the Jones and Eggleton transect method (Jones & Eggleton, 2000). This method uses a 100 m x 2 m belt transect which is divided into twenty 5 m x 1 m sections. Each section is sampled for 30 minutes by two trained collectors searching for termites in twelve 12 cm x 12 cm x 10 cm soil pits and examining all dead wood, leaf litter and trees for the presence of termites. When encountered, termite specimens were collected in 70% ethanol and taken to the laboratory for identification. Because our gap sites were not big enough to place a 100 m transect, we carried out the same method but two using smaller transects to equal a 50 m transect combined. Therefore the sampling effort was half that of the Jones & Eggleton (2000) method.

Dead wood surveys

To estimate the volume of deadwood found on the forest floor in areas affected by tree-fall, compared with undisturbed areas, we carried out deadwood surveys in December 2017. To avoid disturbing our decomposition assays, these surveys were carried out in areas within the forest surrounding experimental plots. We established eight 50 m transects, four of which were within 5 m of a tree that had fallen during the storm in July 2017 and four that were in areas of forest at least 20 m from the nearest tree fall. Along each transect, we recorded the diameter of each piece of deadwood that intersected with the line, and these values were used to calculate the volume of dead wood using the following equation (Van Wagner 1968):

V= π²8Ld²

Where V is the volume of deadwood (cm³/50 m), d is the diameter of the deadwood item at the intersection and L is the length of the sample line.

1. Proportion mass loss from pinus radiata blocks that were accessible to invertebrates and micobes or microbles only in canopy gaps and intact forest in an old growth rainforest. 

2. Monthly soil mositure and temperature beneath tree fall canopy gaps and closed canopy sites.

3. Termite communities sampled within tree fall canopy gaps and closed canopy sites.

4. The volume of deadwood on the forest floor close to or more than 20 m from a tree fall gap.