Long-term resource addition to a detrital food web yields a pattern of responses more complex than pervasive bottom-up control

Background Theory predicts strong bottom-up control in detritus-based food webs, yet field experiments with detritus-based terrestrial systems have uncovered contradictory evidence regarding the strength and pervasiveness of bottom-up control processes. Two factors likely leading to contradictory results are experiment duration, which influences exposure to temporal variation in abiotic factors such as rainfall and affects the likelihood of detecting approach to a new equilibrium; and openness of the experimental units to immigration and emigration. To investigate the contribution of these two factors, we conducted a long-term experiment with open and fenced plots in the forest that was the site of an earlier, short-term experiment (3.5 months) with open plots (Chen & Wise, 1999) that produced evidence of strong bottom-up control for 14 taxonomic groupings of primary consumers of fungi and detritus (microbi-detritivores) and their predators. Methods We added artificial high-quality detritus to ten 2 × 2-m forest-floor plots at bi-weekly intervals from April through September in three consecutive years (Supplemented treatment). Ten comparable Ambient plots were controls. Half of the Supplemented and Ambient plots were enclosed by metal fencing. Results Arthropod community structure (based upon 18 response variables) diverged over time between Supplemented and Ambient treatments, with no effect of Fencing on the multivariate response pattern. Fencing possibly influenced only ca. 30% of the subsequent univariate analyses. Multi- and univariate analyses revealed bottom-up control during Year 1 of some, but not all, microbi-detritivores and predators. During the following two years the pattern of responses became more complex than that observed by Chen & Wise (1999). Some taxa showed consistent bottom-up control whereas others did not. Variation across years could not be explained completely by differences in rainfall because some taxa exhibited negative, not positive, responses to detrital supplementation. Discussion Our 3-year experiment did not confirm the conclusion of strong, pervasive bottom-up control of both microbi-detritivores and predators reported by Chen & Wise (1999). Our longer-term experiment revealed a more complex pattern of responses, a pattern much closer to the range of outcomes reported in the literature for many short-term experiments. Much of the variation in responses across studies likely reflects variation in abiotic and biotic factors and the quality of added detritus. Nevertheless, it is also possible that long-term resource enhancement can drive a community towards a new equilibrium state that differs from what would have been predicted from the initial short-term responses exhibited by primary and secondary consumers.

The most direct way to uncover the strength of bottom-up control and its pervasiveness 24 across trophic levels is to observe how adding energy-and/or nutrient-rich detritus to replicated 25 plots in a field experiment alters densities of major taxa of primary and secondary consumers.  Long-term experiments are more likely to capture the influence of variation in abiotic factors 41 and also are more likely to reveal indirect effects that propagate at different rates through a complex 42 food web. Among experiments conducted to date, densities of some trophic groups occasionally 43 responded negatively to addition of detritus. Do these negative effects reflect differential responses    Here we report results of a 3-year detrital-addition experiment in a secondary oak-maple-84 hickory forest, with a few scattered pine trees, in Madison Co., Kentucky, USA. Our experiment 85 was conducted within ~0.5 km of the sites of a previous similar, but short-term (3.5-months), The experiment started two years after CW99 and ran from 1997 through 1999 (hereon 108 designated Years 1, 2, and 3). Each experimental unit (20 in total) was a 2 x 2-m area of forest were enclosed with 35-cm aluminum flashing inserted 8 cm into the ground (Fenced). The fence 115 was topped with a 15-cm horizontal strip of flashing that formed two lips coated on the underside 116 with a tree-banding compound (Tanglefoot, Grand Rapids, Michigan) to further retard 117 movement of epigeic (ground-active) arthropods across the barrier. This design is more complex 118 than that of CW99, which had no fenced plots but employed the same total number of 119 experimental units: twenty 2 x 5-m open plots, half of which received a detrital supplement. 120 We employed the detritus-supplementation protocol of CW99, which has also been used  We decided initially to supplement at a rate approximately 1/3 that of CW99 because we 131 hypothesized that the strong responses exhibited in the earlier experiment were due to a high 132 level of detrital enhancement. We planned to continue this rate of supplementation in the 133 following years, but decided to increase the rate because the increase in densities of most taxa in 134 response to the detrital enhancement in Year 1 was much less (including no responses) than that 135 observed by CW99. Therefore, in Years 2 and 3 we increased the rate of supplementation to a 136 level similar to CW99. In Year 1 each Supplemented plot received 195 g (dry wt.) m -2 of detritus 137 7 (26 g m -2 , 79 g m -2 and 90 g m -2 of mushrooms, potatoes, and Drosophila medium, respectively).

138
In Years 2 and 3 the rate of detritus supplementation was increased ca. 4x (to 770 g m -2 and 874 139 g m -2 total dry wt., respectively). The slightly larger amount of detritus added during Year 3 140 reflects a slightly longer period of detrital addition than in Year 2. Biweekly rates were the same 141 in Years 2 and 3.

142
Adding artificial detritus can influence the structure of the leaf-litter layer, so it is  Due to our decision to increase the supplementation rate in Years 2 and 3 to ~4x the rate 151 of Year 1, differences in response patterns between Year 1 and the following two years can be 152 attributed not only to differences in abiotic factors between years and time lags in the appearance 153 of direct and indirect effects, but also to markedly different rates of detrital supplementation. In 154 contrast, differences between Years 2 and 3 must have been due primarily to factors other than a 155 difference in the rate of resource addition, since biweekly rates of supplementation in Years 2 156 and 3 were the same.
8 Differences in fungal abundance between Supplemented and Ambient plots at the end of the 161 experiment were estimated by assaying leaf litter for ergosterol (Appendix S1), a common sterol 162 in fungal hyphae that is nearly absent from plants (Weete & Weber 1980). The amount of 163 ergosterol is correlated with both total hyphal mass and membrane content and likely assayed        patterns. 238 We postulated that the system would respond differently in summer and fall because 239 summer samples had been exposed to detrital enhancement for fewer months than fall samples

261
We implemented S15 because this version of Gower's measure is a symmetrical index that gives 262 equal weight to double zeroes (absence/absence) and ++ (presence/presence), which is the type   Interpretation of how the system responded was based upon evaluating the results of multivariate 294 and univariate analyses together as a whole. Emphasis was placed upon using the statistical 295 analyses to aid in interpreting the overall patterns of the ordinations and univariate plots. We 296 avoided a completely NHST (Null Hypothesis Significance Testing) approach as much as 297 possible. Thus, we made no corrections for "multiple comparisons" in the univariate analyses, 298 primarily because multivariate effects were clear; and we did not conclude that P values close to 299 .05, but slightly greater, indicated the absence of a response. Instead, we relied on P values as a   The most abundant taxa were six families of microbi-detritivores (Collembola) collected by 318 Kempson extraction, with the Hypogastruridae, Onychiuridae, and Entomobryidae being the 319 three most numerous (Fig. 1A). Remaining response variables were similar in value to each 320 other, differing by less than 50%. Centipedes (Chilopoda) were the exception. This predatory 321 taxon had the fewest individuals sampled of all response variables (Fig. 1B). Results of multivariate analyses are presented first because they constituted the critical first step 326 in the analysis. If detrital supplementation had no effect upon overall community structure,   Divergence between Ambient and Supplemented plots was greater in fall than summer ( Fig. 2; 347 Table S3.3 in Appendix S3). In summer, percentages of total variation explained by the Resource 348 treatment for Years 1-3 were 12%, 26%, and 28%, respectively. Percentages for fall samples 349 were 24%, 41%, and 32%. Divergence in community structure over time was due primarily to 350 increasing separation of centroid locations (Fig. 2). However, on the last sampling date of the   Fall --The pattern for fall samples is more complex (Fig. 3A, Fall). Years 1 and 2 377 exhibited a pattern broadly similar to that of summer samples for Years 2 and 3, although the 378 total number of responding taxa was greater in fall than summer (16 vectors versus 10, 379 respectively). All vectors were positively correlated with CAP Axis 1 except for Thysanoptera, a

388
The strongest correlation in Year 3 was that for web-building spiders. However, the relationship 389 with detrital addition was negative, not positive as was the case in Year 1 (Fig. 3A, Fall).  Univariate analyses 404 We analyzed each of the 18 response variables separately for summer and fall samples. Plots  (Figs. 2, 3). First, effects of detrital 442 supplementation tended to be stronger and more widespread across taxa in fall samples, 443 particularly in Year 1 (Fig. 4A). Secondly, the pattern of response changed over the three years.     with CAP Axis 1 (Fig 4A compared with Fig. 3). All univariate variables that exhibited one to 488 three instances of a negative response to detrital addition among the permANOVAs also negative in the fall of Years 1-3, respectively (Fig. 3A, Fall). The multiple correlation coefficient 495 in Year 3 (Fig. 3B, Fall) was also negative.    per Resource level, the same as CW99. Most critical is the need to explain why our study did not 587 reveal continued bottom-up control over three years by the same taxonomic groups that showed 588 such a clear response after 3.5 months of detrital addition in CW99 (Fig. 4A, B). Below we 589 evaluate the possible influence of several factors. noise might make the prospects of a "long-enough" field experiment seem hopeless. However, 705 after reviewing short-and long-term experiments (2 to 31 mos.) in the intertidal, Menge (1997) 706 concluded that ". . . community dynamics may be more predictable than expected . . ." because 707 most indirect effects had appeared half-way through the experiment. "Long enough" is at least 708 the number of generations sufficient to reveal indirect effects. An even longer time would reveal 709 how close the community may be to a new equilibrium. This longer time will never be achieved 710 with certainty, but snapshots yielded by short-term experiments cannot answer the question.    (Table A3.2 in Appendix S3). Open and Fenced plots have been pooled for perMANOVA because there were no interactions involving (Resource x Year) and Fencing (Table S3.1 in Appendix S3). Because each season exhibited a (Resource x Year) interaction (Table S3.3 in Appendix S3), P-values for the Resource treatment are given for each Year. . Except for Summer of Year 1, separation of communities along axis 1 (CAP1) is strongly related to the impact of the Supplemented resource treatment on community structure. Thus, the extent to which a vector is parallel with CAP1 reflects the extent of the negative (to the left) or positive (to the right) correlation of densities of that taxon with the resource treatment. The length of each vector represents the joint correlation of the response variable with both axes of the ordination, with the circle representing a correlation of 1. Vectors shown have Spearman coefficients with CAP1 ≥ .50 or ≤ -.50. To prevent clutter on the graph, arrow heads of the vectors are not drawn. Key to abbreviations is in Fig. 1. (B) Constrained ordination (CAP) with vector overlays representing multiple correlation coefficients (analogous to univariate partial correlation coefficients).  Figure. 4 (cont'd). . . . Arrow width reflects an Adjusted Effect Size: the simple effect size, based upon changes over time in the difference between means of Supplemented and Ambient treatments, modified by the P value (a measure of strength of evidence) of the Resource x Year and/or Resource pseudo-F's from the permANOVA. This was done by inspection, not by a single, simple rule. Thus, widths and placement of these arrows do not precisely reflect statistical tests for each sampling date, but are meant to summarize the overall patterns. For example, if the simple effect size is ~2x but the P value of the pseudo-F statistic is between .10 and .05 for both the interaction and overall resource effect, the smallest-width arrow is given for the Adjusted Effect Size (i.e. Adjusted Effect Size of 1.5x instead of 2x). Furthermore, if there is no Resource x Time interaction but there is an overall Resource effect, an arrow is present for each date even though the P value of a test on that date might be higher than the overall P value. Empty cells for all three years in a season means there was no Resource x Year interaction, nor was there a simple Resource effect over the three years (for that season). One or two blank cells across a season indicates the shape of the Resource x Year interaction for that season. A "zero" indicates that densities in both Supplemented and Resource treatments were ≈ 0 that sampling period, although there was an overall Resource x Year interaction across the entire experiment. A simple arrow, i.e. one without a letter next to it, depicts analyses based upon pooled Open and Fenced plots (N = 10 / Resource treatment). If there was an interaction with Fencing, the arrow describes the fencing treatment that displayed a Resource x Year interaction or simple Resource effect, indicated by "O" or "F" to the right of the arrow for Open or Fenced plots, respectively. (B) Results for the study by CW99, which are based upon 3.5 months of adding detritus to open (unfenced) 2 x 5-m plots (N = 10 / Resource treatment). P values for all test statistics were < .05, with most being < .01 or < .001 . N/A = response variable not sampled.