Dataset: Fertilizer impacts on damselfish behavior in Mo'orea

This dataset presents results examining the effects of fertilizer addition on damselfish behavior from a before-after control impact (BACI) experiment. We identified 15 spatially blocked pairs of Stegastes nigricans (damselfish) territories and into one randomly determined territory of each pair we attached a nylon mesh pouch filled with 50 grams of Osmocote slow-release fertilizer (14:14:14 NPK) directly to the substrate using u-nails. In the other plot of that pair, we attached a nylon pouch filled with 50 grams of sediment collected nearby as a physical control. All territories were at a depth of < 3 m. Pairs were ~20 meters from each other, with blocked pairs separated by > 20 meters. We measured territory size following Blanchette et al. (2019) by measuring the length, width, and height of the territory. The number and size (to the nearest centimeter) of all damselfish within each territory were estimated visually.

The treatments' effects on behavior were assessed by video recording each territory for 20­–30 minutes immediately before applying treatments (“Day 0”) and 14 days post-treatment (“Day 14”). Small video-cameras (Go-Pro Hero 7 Black) were mounted on dive weights that could be placed unobtrusively on nearby substrates. We ignored the first 5 minutes of each video to minimize the impact of camera placement on behavioral observations. Because damselfish tend to chase other herbivores from their territories, we assessed: 1) grazing by damselfish in their own territories, 2) grazing by roving herbivores in damselfish territories, 3) aggression towards conspecifics, and 4) aggression towards heterospecifics in each video. Behaviors were quantified for all resident damselfish that were visible for at least 50% of the video for each territory (i.e. not concealed in crevices or behind reef structure) and all behaviors (i.e. the number of bites or chases made by an individual fish) were standardized by the length of time that any damselfish was visible and the area of the territory at the timepoint a video was taken (i.e., Day 0 or Day 14). All behaviors were compared using generalized linear mixed effects models (GLMER) with timepoint, treatment, and their interaction as fixed effects, as is standard in BACI designs. Territory ID was nested within spatial blocking as a random effect to account for pseudoreplication from both observing multiple fish within a territory and observing the same territory over two timepoints. Within factor pairwise comparisons were conducted using the emmeans package in R (Lenth, 2023). Preliminary data visualization suggested that behavior data could be zero-inflated or overdispersed, so all behaviors were modeled using i) a Poisson family, ii) a Poisson family with a zero-inflated intercept, and iii) a negative binomial. Model assumptions for all families were assessed using DHARMa (Hartig 2022), and models were compared using the Second-order Akaike’s Information Criterion (AICc) using the MuMIn package in R (Bartoń 2024). The top model that satisfied model assumptions was then used for each behavior. Damselfish bite rate data were overdispersed, so were modeled using a negative binomial distribution. Bite rates from non-damselfish herbivores and heterospecific chase rates were both zero-inflated, so were modeled as a zero-inflated Poisson regression. Conspecific chases were modeled using a standard Poisson regression.

Changes in traits of damselfish turfs and biomass were assessed by collecting six samples of turf that were about two centimeters square (~2 cm2) using a hammer and chisel from each damselfish territory both pre- and 14 days post-treatment. Weights, ash-free dry weights, and isotopic compositions were determined to assess and describe the turf biomass. Full details can be found in BCO-DMO dataset 950508 "Damselfish Turf Biomass". (See Related Datasets section below).