How populations of ecosystem engineers are both driven by and drive biodiversity is poorly known, even less so in detrital subsystems. Deadwood plays a key role in maintaining biodiversity and ecosystem functions. Wood economics spectrum (WES), which represents the initial wood quality through a cluster of wood traits, via afterlife effects might affect termite populations and deadwood invertebrate community structure. Termites, as ecosystem engineers, exert a significant impact on other invertebrate diversity. However, how the WES modulates the effects of termite foraging intensity on deadwood invertebrates is unclear. We hypothesized that the WES had significant effects on termite foraging intensity and deadwood invertebrate abundance and richness. Moreover, the WES was hypothesized to modulate the effects of termite foraging intensity on deadwood invertebrate diversity. We conducted a wood decomposition experiment to test our hypotheses in two subtropical forests in China. Logs of 22 tree species with distinct functional traits were incubated for 30 months to measure termite foraging intensity (relative termite feeding area and the mass of soil materials imported by termites) and deadwood invertebrate abundance and richness. We found that from the conservative (slow-growing species with low wood quality) to the acquisitive (fast-growing species with high wood quality) ends of the wood economics spectrum, termite foraging intensity increased. As termite foraging activity intensified, the deadwood invertebrate abundance and richness increased correspondingly. Moreover, there were significant positive relationships between termite foraging intensity and important detritivore (Acari, Collembola, and Opisthopora) abundance. In contrast, the position along the WES had no direct effect on the abundance and richness of deadwood invertebrates besides via termite foraging intensity.

Synthesis. Our findings showed the pathway through which the WES affects deadwood invertebrate diversity. It supported the hypothesis that the WES plays a crucial role in shaping the effects of termites as ecosystem engineers on the broader invertebrate community. Future studies should focus more broadly on whether and how plant traits, via afterlife effects on ecosystem engineers, influence invertebrate community composition and structure. Such studies will promote our understanding of the importance of both plant traits and ecosystem engineer traits for ecosystem carbon and nutrient cycling and biodiversity.

Data collected

Invertebrate sampling, identification and counting: In July 2020, 147 logs (22 species) of 35 cm length were quickly and carefully put into plastic zip-lock bags in filed and taken into the laboratory at Tiantong and Putuo Ecological Station, where the samples were stored at 15 °C for subsequent processing. First, we put the wood sample in a deep plastic tray and collected larger, visibly active invertebrates by hands to prevent escape before processing. Then, we sawed the log into two halves, and then we sawed a 2.5 cm thick disk at each end of each half log. This way, we collected two disks from the middle and end positions. We used one middle disk and one ends disk to measure termite foraging intensity, and the other half for mass loss and traits analysis (not included in this study). The remaining logs (25-cm long) were used to analyze the composition and diversity of the other wood-inhabiting invertebrate community. 

To collect the macroinvertebrates (body width > 2 mm) of logs, firstly we put the residual wood sample in a deep plastic tray and separated the bark from the wood using chisels or manually. After the bark had been peeled off and fragmented into small pieces, we extracted the invertebrates living in the bark. The wood remaining after peeling off the bark was split into small pieces by an axe, after which we collected the macroinvertebrates by hand. The soft xylem with holes from termite consumption was carefully cut away with chisels until all soft wood was removed to make sure that all invertebrates were collected and transferred them to 30 ml sealed plastic bottles with 70% ethanol for identification and counting. If wood under the bark was intact and did not show any signs of xylophagy, we would not break it for invertebrate searching. The fragmented and soft xylem that we cut away were collected to extract the mesoinvertebrates by using a modified Tullgren funnel at 35 ℃ for 24 h and collected in 50 ml sealed plastic bottles with 70% ethanol solution. All the samples were processed within 20 days after collection from the field, i.e. when the invertebrates were generally still alive inside the logs.

Then an electron optical type microscope (DVM6, Leica Germany) at a magnification of 35~500× was used for mesoinvertebrate identification and counting. All extracted Acari were grouped into Mesostigmata, Oribatida and Prostigmata (also including any Astigmata) following Yin (1992). All Collembola and adults of other dominant groups were identified to family level following Yin (1992, 2000). Juveniles and some rare taxa were only identified to the order level due to difficulty of identification. Richness was calculated as the number of all other taxa (including ants) present per log and abundance was calculated as the total number of individuals per log, excluding ants (but conducted a separate analysis of termite-ant relationships).

Initial wood traits measurement: After we cut all trees into logs, we took the adjacent wood disks to the lab for analyses of initial wood traits. We measured the wood density (WD), nutrient concentration (N, P concentration), wood dry mass content (WDMC), cellulose and lignin content. For measurement of initial wood traits (without bark) considered relevant to the WES and to deadwood invertebrate preference, each 5 cm disk was stored in a zip-lock plastic bag immediately after collection in the field, and kept cool until processing. Within 12 h, after the bark had been removed, a subsample was cut from each disk to obtain fresh mass and initial volume using Archimedes' principle of water displacement. Then wood subsamples that had been measured for fresh mass and volume were dried at 75 °C for 72 h to determine dry mass. Initial wood density (g·cm^-3) was calculated as dry mass per (fresh) volume. Initial wood dry matter content was calculated as dry mass/fresh mass. Initial dry subsamples were ground in a laboratory mill and passed through a 0.15 mm sieve. Thereafter, 0.2 g subsamples were digested using concentrated H₂SO₄ to determine N and P concentrations (mg·g^-1) on an infrared spectrophotometer (Smartchem 200, Alliance, France). Cellulose content was determined by anthrone-sulphuric acid colorimetry and lignin content was determined by acid detergent fiber (ADF).

Measurement of termite feeding area and imported soil mass: After the harvested disks were cleaned thoroughly and dried, we first measured the thickness of the disk (for the subsequent calculation of disk volume), then digital images of the top and bottom surfaces of each disk were taken. The image resolution was enhanced (e.g., backgrounds made whiter and brighter, Figure S2a) and areas of the remaining wood surface after termite consumption were delineated (e.g., colored white, Figure S2b) using Adobe Photoshop CC19.0. Thereafter Image J 1.53c was used to measure the surface area of intact wood. Each wood sample had two disks and was measured four times. The average of the percentage of (%) termite feeding area and intact wood area of the four measurements were used as termite feeding extent due to the relatively short length of the wood samples. The damage could still be discerned rather clearly in the 30-month samples, but soon after that it became too difficult to accurately estimate the damage.

We carried out a fire experiment to measure the termite imported soil mass as another index of termite foraging intensity. The measured soil mass cannot represent the activity of all termites, but represent the activity of Odontotermes and some of Reticulitermes (the dominant species in the study area). According to Ulyshen & Wagner (2013), completely burning the wood away to leave only the mineral soil behind as ash is a more accurate method than washing to isolate soil from decomposed wood and collect the soil for quantifying the termite foraging intensity. The fire experiment was carried out in the Fire Laboratory located at Hangzhou Normal University, Zhejiang province. From September 2020 to January 2021, we did a complete combustion of all the wood disks of all 22 species. All burns were conducted under a fume hood on a solid fire-resistant plate. Before the fire experiment started, the fume hood was turned on and ventilated at a constant speed. Prior to each burn, the lab door was closed and the fume hood was turned on and the experiment started when the room temperature was 21 ± 4 °C. For each burn, we put the wood disk on a stainless-steel plate and then put the plate on an electronic heating plate. We set the temperature of the electronic heating plate at 300 °C. After about 30 min preheating, we ignited the wood by a butane gas torch. After ignition, we let the wood burn and smolder naturally until completely burning. For wood with low flammability that cannot completely burn by itself after ignition, we used a butane gas torch to burn the remaining mass completely.

Wood economics spectrum (PC1) construction : We used a principal component analysis (PCA) to construct the wood economics spectrum. Traits in including Carbon, Nitrogen, Phosphorus, Lignin, Cellulose, and wood density. Because of the relatively high proportion of species variance explained by the first PCA axis, we used the PC1 scores to represent a synthetic variable incorporating multiple trait variables as an index of the wood economics spectrum in the subsequent analyses.

Dataset  processed: Deadwood invertebrate abundance (the number of individuals) data were log-10 transformed and imported soil mass data and termite feeding area data were Arcsine-transformed before analysis to better achieve normality and linearity.