Dataset DOI: 10.5061/dryad.pg4f4qs1r

Description of the data and file structure

README — Unimodal, JBE, Covariance Analyses (Figures 3–5)

This repository contains scripts and data required to reproduce Figures 3–5 from the manuscript. The analyses cover:

1. Unimodal and linear regression of fish community metrics versus environmental and productivity predictors (Figure 3).
2. Nested regression modeling of productivity–richness relationships (Figure 4).
3. Covariance tests across biomass gradients for different fish functional groups (Figure 5).

1. Figure 3 — Unimodal and Linear Regression

Script: unimodal_and_linear_regression_Figure3.R

Description:
Generates Figure 3 and supporting appendix figures, exploring linear vs. quadratic (unimodal) relationships between fish community metrics and predictors.

Input:

• unimodal_and_linear_regression_Figure3.csv — dataset containing lake-level observations.

Dataset columns:

How to use:

1. Load the R script into R.
2. Run the script from top to bottom.
3. The script fits linear and quadratic models, calculates unimodal peaks, compares models (AIC, ANOVA), and generates scatter plots with regression curves.
4. Produces Figure 3 and appendix plots.

Output:

Plots are generated for the following relationships:

Each model comparison outputs:

• AIC values for linear vs. quadratic models
• ANOVA test between models
• Plots showing best-fitting curves and peak indicators

2. Figure 4 — JBE Models

Script: JBEscript-Figure4.R

Description:
Reproduces Figure 4 by comparing observed productivity–richness relationships with predictions from nested regression models (JBE, JB, JE, J, B, E). Computes RMSE for quantitative comparison and generates publication-ready figures.

Input:

This script requires three data files in CSV format:

• abun — fish abundance matrix
  • Row names: lake codes.
  • First column No: lake/sample identifier.
  • Remaining columns: species abbreviations.
  • Units: individuals (ind).
• bio — fish biomass matrix
  • Row names: lake codes.
  • First column No: lake/sample identifier.
  • Remaining columns: species Latin names.
  • Units: grams (g).
• var — environmental & community summary variables (data frame with lake rows). Important columns:
  • No — lake/sample identifier (numeric)
  • SRic — species richness (mean species count per lake/station)
  • LogP — proxy for productivity, calculated as
    log(Productivity) = 0.91 * log(BPUE) - 0.38 * log(W)
    where W = BPUE / NPUE
  • TP — total phosphorus (mg L⁻¹)
  • TN — total nitrogen (mg L⁻¹)
  • Chla — chlorophyll a (µg L⁻¹)
  • COD — chemical oxygen demand (permanganate-oxalate titration; mg L⁻¹)
  • SD — Secchi depth (m)
  • depth — maximum water depth (m)
  • sup — lake area (km²)

Script overview:

1. Import the three CSV files (abun.csv, bio.csv, var.csv)..
2. Build full (lmnbe) and selected (bmnbe) regression models.
3. Generate partial predictions under six scenarios (JBE, JB, JE, J, B, E).
4. Calculate RMSE for model evaluation.
5. Produce Figure 4, showing observed vs. predicted relationships.

How to run:

source("JBEscript-Figure4.R")

Set mod <- lmnbe or mod <- bmnbe in the script to reproduce full or selected model.

Output:

• Figure 4 (observed vs. predicted productivity–richness curves)
• RMSE values.

3. Figure 5 — Covariance Test across Biomass Gradients

Script: Covscript-Figure5.R

Description:
Calculates the covariance test (U, W, p) for different fish biomass groups (all species, zooplanktivores, piscivores, stocked species) across low, medium, and high lake biomass groups. Visualizes the results using ggplot2.

Input:

• The same three CSV data files (abun.csv, bio.csv, var.csv) as used in Figure 4.

Script overview:

1. Define biomass groups and compute total biomass per functional group.
2. Define the tcov function to compute U, W, and p for covariance tests.
3. Rank lakes by biomass and split into low, medium, high groups.
4. Compute covariance statistics for all functional groups and biomass classes.
5. Prepare results for plotting (covtest) with significance labels.
6. Generate Figure 5 using ggplot2.

Output:

• Figure 5: U-values of covariance test across biomass gradients, with significance annotations.

This unified README provides a concise description of the three main scripts, their inputs, outputs, and usage instructions, maintaining consistent style across all analyses.