We investigated the legacy effects of drying and refilling of wetlands via mesocosms on growth and survival of recolonizing aquatic vertebrates and their subsequent effects on ecological function. Specifically, we tested the effects of prior drying on anuran and fish growth and survival (using Gompertz growth curves), and joint effects of vertebrates and drying on periphyton biomass, and whole-system gross primary productivity. Using 64 1,000-L hard plastic cattle tanks to create replicate aquatic mesocosms, four trophic treatments (no vertebrate control, larval anurans, fish, larval anurans and fish) were crossed with two drying treatments (undried and dried-refilled) to simulate different outcomes of colonization by two common vertebrate guilds and drying/refilling cycles. Mesocosms with anuran treatments received 50 Lithobates blairi tadpoles (1,600 total) and mesocosms with fish treatments received one Lepomis cyanellus juvenile (32 total). We recorded anuran growth rates, survival, and time/size at metamorphosis, fish growth, gross primary productivity, and periphyton biomass over 12 weeks from 9 May 2021 to 1 August 2021.

This dataset contains 4 R script files (one for each of the 4 components of this study: Amphibians, fish, gross primary productivity, and periphyton) along with their corresponding Excel files. Survival for each tank was calculated as the total number of emerged frogs from weeks 1–12 divided by 50 (initial number of larvae per tank). Size at metamorphosis was determined by dry mass (g) measured after 72 hours in a drying oven at 60° C. Metamorph dry mass and wet mass were strongly correlated (grams dry mass = 0.1957*grams wet mass – 0.0398; R² = 0.9733) and we chose dry mass for size at metamorphosis to reduce variability caused by water weight. We measured the dry mass of metamorphosed frogs to the nearest 0.01g. Time to metamorphosis was measured as the days past the first day of the experiment (9 May 2021) to the day the frog reached Gosner stage 46 (Gosner, 1960). We calculated relative fish growth as the change in standard length (SL) over the duration of the experiment divided by the initial SL for each fish. Individual fish growth calculations used each of the single fish per tank as independent replicates. Relative fish growth measurements for fish within the same treatments (n = 8) were combined and analyzed together. Gross primary productivity values were collected by measuring dissolved oxygen (DO) values after sunset for maximum DO and before sunrise for minimum DO. DO measurements were collected using a YSI meter (ProQuatro XA00088 – 02; YSI Inc., Yellow Springs, OH, USA) as a percentage out of 500.  

Gompertz growth curves provided estimates of asymptotic larval body mass (W_max), maximum growth rate (K), and the time at which growth rates reach K (xm). However, only W_max and K were included in statistical analyses as they were the primary parameters of interest for our study. All statistical analyses were performed in RStudio version 4.1.1 (R Core Team, 2021). We constructed generalized linear models (GLMs; ‘stats’ package in R; R Core Team, 2021) for each of Wmax, K, number of emerged frogs (survival), size at metamorphosis, time to metamorphosis, range of metamorph sizes, and range of metamorphosis days as response variables with drying treatment as the predictor variable. We also constructed GLMs with fish growth as the response variable and a combination of drying treatment and anuran presence as predictor variables with interaction terms. We then used type II ANOVAs for models without significant interaction terms and type III ANOVAs for models with interaction terms via the ‘car’ package in R (Fox & Weisberg, 2019) to determine the treatment effects on above response variables. GLM model fit and assumptions were checked by a visual assessment of the plotted residuals of each GLM model.

Since GPP varied nonlinearly over the duration of the experiment (Fig. 4), we investigated the effects of drying and refilling and trophic treatment on GPP using generalized additive mixed models (GAMMs) with the ‘gam’ function (‘mgcv’ package; Wood, 2015). The GAMM model included trophic treatment and drying treatment as parametric effects, trophic treatment over time, drying treatment over time, and time as cubic regression smoother terms, and individual mesocosm and individual mesocosm over time as random effects. The number of knots was set to -1 to utilize the generalized cross-validation method to automatically choose the number of knots for the model. GAMMs allow us to test hypotheses related to main effects (e.g., trophic treatment, drying treatment) within non-linear time series while accounting for random effects related to individual mesocosms and changes over time due to seasonality. We then used a subsequent ANOVA via the ‘anova.gam’ function within the ‘gam’ package to determine treatments effects. 

We investigated the effects of drying and refilling and trophic treatment on periphyton biomass using GAMMs constructed and analyzed identically to GPP except that they are only for experimental days 19–80. This is because periphyton sampling techniques were not standardized until after the second week of data collection. Because data for periphyton biomass was right-skewed and included some 0’s that were not true zeros (periphyton was present but below measurement limits), we added 0.5 to each measurement (untransformed mean was 1.838 g) and then log-transformed them. Models run with the transformed data met the assumptions of GAMMs and subsequent ANOVAs.

All data were collected in the field at the location of the experimental study. All field data were entered into Excel sheet format and subsequently analyzed via the R Studio software program. The only modification to the collected data is the following: data from tank #1 were omitted from periphyton and gross primary productivity analyses due to a separate amphibian species ovipositing within that tank. 

All statistical analyses were performed in RStudio version 4.1.1 (R Core Team, 2021). All packages required for the statistical analyses within the R program are listed at the top of each R script file. All files submitted are either .xslx or .Rmd file types and both Microsoft Excel and R Studio will be required to open the data files.