Burrowing detritivores regulate nutrient cycling in a desert ecosystem

Nutrient cycling in most terrestrial ecosystems is controlled by moisture-dependent decomposer activity. In arid ecosystems, plant litter cycling exceeds rates predicted based on precipitation amounts, suggesting that additional factors are involved. Attempts to reveal these factors have focused on abiotic degradation, soil–litter mixing and alternative moisture sources. Our aim was to explore an additional hypothesis that macro-detritivores control litter cycling in deserts. We quantified the role different organisms play in clearing plant detritus from the desert surface, using litter baskets with different mesh sizes that allow selective entry of micro-, meso- or macrofauna. We also measured soil nutrient concentrations in increasing distances from the burrows of a highly abundant macro-detritivore, the desert isopod Hemilepistus reaumuri. Macro-detritivores controlled the clearing of plant litter in our field site. The highest rates of litter removal were measured during the hot and dry summer when isopod activity peaks and microbial activity is minimal. We also found substantial enrichment of inorganic nitrogen and phosphorous near isopod burrows. We conclude that burrowing macro-detritivores are important regulators of litter cycling in this arid ecosystem, providing a plausible general mechanism that explains the unexpectedly high rates of plant litter cycling in deserts.

Appendix S1 -Climate conditions at the research site Figure S1-1. Annual precipitation in Even-Ari research station  [1]. Mean annual precipitation is 92.19 mm, and the mean absolute difference between two consecutive years is 45.43 mm.

Possible association between wind characteristics and litter mass loss
Wind data at our field site was collected in a temporal resolution of 10 minutes, during the entire experimental period [1]. We used this information to explore whether accidental litter loss by wind may contribute to the elevated rates of litter mass loss from the macrobaskets. The data clearly show that average and maximal wind speeds or the number of extreme wind events (above 10 m/s) cannot explain our findings (Table S1-1; Fig. S1-3). For example, both the maximal wind speed and the number and magnitude of strong gusts (above 10 m/s) were highest during the winter when litter mass loss in the macro-baskets was the lowest. The pattern was opposite during the late summer in which we observed the largest rates of litter mass loss, but the average and maximal wind speeds were the lowest and there were no extreme wind gusts that exceed 10 m/s. Thus, even if wind can cause litter loss from our baskets, then the effect size during the summer should be smaller than the one observed during the winter (when winds are strongest) and not much larger as clearly shown by our results.

Appendix S2 -Materials and Methodsadditional information
The contribution of macro-detritivores to surface litter removal We used square plastic baskets with the following dimensions: Bottom -120 X 130 mm We monitored all baskets every two weeks to check for litter accumulation near the outer meso-nets, prevent litter spills (from the macro-baskets) and test the effectiveness of our design (observing macro-arthropods behavior). We have never seen accumulation let alone spill, suggesting the 2mm mesh of the litter-baskets serves as a wind breaker and that accidental litter loss by wind, if at all exist, was minimal. Isopods were observed consuming litter within the baskets.
We transported each pre-weighed basket to the field in a new individual Ziploc bag that was placed within a sturdy container. After positioning the basket in a preassigned location, we checked the Ziploc bags for remaining litter and added the leftover litter, if found, to the corresponding basket. After returning to the lab, we rechecked the empty Ziploc bags above a tray covered with white paper and subtracted the weight of the tiny litter remains (rarely found) from the initial litter weight.
At the end of each trial, we carefully removed the arched metal stakes that tightly anchored the baskets to the soil and carefully inserted the baskets into new Ziploc bags for transportation to the lab. In the lab, we removed the baskets and all the litter from the bags above a tray covered with white paper in order to minimize accidental losses. We collected the small litter fragments manually with the aid of a large magnifying glass. In six cases (1 macro, 1 micro and 4 meso baskets) during the first trial, litter was accidently spilled out of the tray. As a precaution, we excluded these measurements from the analysis. We also excluded one meso-basket during the second trial because of possible mass loss that happened while positioning the baskets in the field.

Soil characteristics around Hemilepistus reaumuri burrows
We first sampled the feces mounds around isopod burrows, as well as the soil crust at a distance of 5 cm from the outer boundaries of the mound (figure S2-4).
After removing the mound and the soil crust, we took 2 cm diameter samples of the upper 30 cm of the soil profile, at different distances (0, 10, and 20 cm) from the burrow opening (figure S2-5).

Microcosms experiment (Isopolis)
We constructed five custom-made Plexiglas microcosms that allow detailed monitoring of isopod activity and nutrient distribution belowground. Isopolis was made of an aboveground chamber (dimensions 60*30*20cm) connected to a narrow (60*40*0.8cm) belowground transparent chamber (Fig. S2-6A). We filled the aboveground chamber with 5 cm deep homogenized field collected soil and paved it with field collected BSC. The belowground chamber was compacted with the same homogenized soil, to which we translocated one H. reaumuri mating pair (Fig. S2-6B). The transparent belowground chambers were covered by opaque screens (Fig.   S2-7A). Twice a week throughout the experiment we temporarily removed the opaque screens to record the distribution of isopods and feces within the burrows (Fig. S2-7B). We used this information to determine the isopods and feces frequencies of occurrence in different depths. At the end of experiment, we carefully removed the front wall of the belowground chamber and sampled the soil at 1 cm intervals from the burrow wall to a distance of 10 cm. We repeated this sampling protocol from the surface to the burrow maximal depth in 5 cm intervals (Table S2-1).

The contribution of macro-detritivores to surface litter removal
We used mean-centered dummy coding for contrasts of both basket type and trial.
This definition of contrasts allowed us to compare between levels of the factors, while avoiding artificial collinearity between the two fixed effects. The models that were used in this analysis are depicted in table S3-1.
We used a series of LMMs to analyze whether the rates of litter removal differ between the three basket types in each trial separately. We divided the data to four subsets, one for each trial. For each subset we ran an LMM with basket type as fixed effect and with by-bush random intercept. We then tested the significance of the fixed effect using LRT, comparing models that include and do not include the fixed effect.
The models that were used in this analysis are depicted in table S3-2.  Inclusion of by-bush random slopes in these models was not applicable using the 'lme4' package, due to over-parameterization. We performed an additional analysis incorporating by-bush random slopes using R package 'nlme' [7]. The results of this analysis are depicted in table S3-3.

Table S3-3 -Results of statistical analysis using R package 'nlme': (A) Likelihood ratio tests (LRTs) testing the effects of litter basket type, trial and their interaction on litter removal rates. (B) LRTs testing the effect of litter basket type on litter removal rates in each trial separately. (C) Pairwise comparisons between litter removal rates in different litter basket types.
A. LRTs for all data 1 II. df = 10,7 in all trials.

Soil characteristics around Hemilepistus reaumuri's burrows
Data of moisture content and electrical conductivity (EC) in mounds and soil crust were log-transformed, in order to meet the assumptions of paired t-tests. Data of moisture and NO3 content in soil were log-transformed in order to meet the assumptions of LMM. One measurement of soil PO4 content was marked as an outlier and was removed from data before analysis. N content data were log-transformed in order to meet the assumptions of LMM.