Skip to content

Latest commit

 

History

History
261 lines (227 loc) · 8.47 KB

sensitivity_analysis_results.md

File metadata and controls

261 lines (227 loc) · 8.47 KB

Sensitivity Analysis Results from MANTRIX

Beau Larkin Last updated: 27 October, 2025

Description

This script summarizes sensitivity analysis results for MANTRIX using output from the MPG Matrix tool. Figures 4 and 5 visualize predicted percent changes in mantid and trophic guild abundance under alternative scenarios, which considered mantids, their natural enemies, and abiotic conditions. Results are based on paired manipulations of Mantis religiosa abundance, cold snap frequencies, and parasitoid wasp presence/absence.

Packages and Libraries

packages_needed <- c("tidyverse", "colorspace", "knitr", "ggpubr", "Cairo", "rprojroot")
packages_installed <- packages_needed %in% rownames(installed.packages())
if (any(!packages_installed)) install.packages(packages_needed[!packages_installed])
for (pkg in packages_needed) library(pkg, character.only = TRUE)
root_path <- function(...) rprojroot::find_rstudio_root_file(...)

Plot Style

source(root_path("supplement", "styles.R"))
fig_pt_size <- 3
fig_dodgewidth <- 0.7
guild_labs <- c(
  "Invertebrate\nPredators",
  "Invertebrate\nHerbivores",
  "Pollinators",
  "Songbirds",
  "Small\nMammals"
)

Sensitivity Analysis Results

Data ETL

Extract, transform, and load data for later use

sens_path <- root_path("data", "sensitivity_analysis_data")
sens_files <- list.files(sens_path, full.names = TRUE, pattern = "matrix_export")
sens_list <- map(sens_files, read_csv, show_col_types = FALSE)
sens_scenarios <- regmatches(sens_files, regexpr("S\\d+", sens_files))
names(sens_list) <- sens_scenarios
df_names <- c(
  "name", "rel_abund", "rel_abund_lc", "rel_abund_uc",
  "abund", "abund_lc", "abund_uc",
  "pct_change", "pct_change_lc", "pct_change_uc"
)
sens_list <- map(sens_list, function(df) {
  colnames(df) <- df_names
  df %>%
    mutate(
      across(starts_with("pct"), ~ round(.x * 100, 1)),
      name = case_match(
        name,
        "Inv. Herbivores" ~ "Invertebrate Herbivores",
        "Inv. Predators"  ~ "Invertebrate Predators",
        "Sm. Mammals"     ~ "Small Mammals",
        .default          = name
      )
    ) %>%
    filter(name %in% c("Mantids", "Invertebrate Predators", "Songbirds",
                       "Small Mammals", "Invertebrate Herbivores", "Pollinators")) %>%
    mutate(name = factor(name, levels = c(
      "Mantids", "Invertebrate Predators", "Invertebrate Herbivores",
      "Pollinators", "Songbirds", "Small Mammals"
    ), ordered = TRUE)) %>%
    select(name, starts_with("pct")) %>%
    arrange(name)
})

Scenario Labels

scenario_names <- data.frame(
  scenario = sens_scenarios,
  scenario_labels = c(
    "Observed-Low", "Predicted Equilibrium", "Observed-Mid", "Observed-High",
    "No Cold Snaps, No Wasps", "Four Cold Snaps, No Wasps",
    "No Cold Snaps, 6X Wasps", "Two Cold Snaps, 6X Wasps",
    "Four Cold Snaps, 6X Wasps"
  )
) %>%
  separate_wider_position(scenario, widths = c(1, s_num = 1), cols_remove = FALSE) %>%
  mutate(
    s_num = as.numeric(s_num),
    scenario_labels = fct_reorder(scenario_labels, s_num)
  )

Figure 4 – Manipulated mantid abundance

fig4_data <- bind_rows(
  list(S1 = sens_list$S1, S2 = sens_list$S2, S3 = sens_list$S3, S4 = sens_list$S4),
  .id = "scenario"
) %>%
  left_join(scenario_names, by = join_by(scenario)) %>%
  filter(name != "Mantids")

fig4_pal <- sequential_hcl(4, h = c(109, 252), c = c(100, 150, 7), l = c(86, 13), power = c(0.5, 0.9))
fig4 <- ggplot(fig4_data, aes(x = name, y = pct_change)) +
  geom_hline(yintercept = 0, linewidth = 0.2, linetype = "longdash") +
  geom_linerange(
    aes(ymin = pct_change_lc, ymax = pct_change_uc, group = scenario_labels),
    position = position_dodge(width = fig_dodgewidth), linewidth = 0.3
  ) +
  geom_point(
    aes(fill = scenario_labels), shape = 21, size = fig_pt_size,
    position = position_dodge(width = fig_dodgewidth)
  ) +
  scale_x_discrete(labels = guild_labs) +
  scale_fill_manual(name = "Scenarios 1−4:\nMantid Abundance", values = rev(fig4_pal)) +
  labs(x = NULL, y = "Abundance Change (Percent)") +
  theme_bgl_s +
  guides(fill = guide_legend(position = "inside")) +
  theme(legend.position.inside = c(0.86, 0.30), legend.title = element_text(hjust = 0))
fig4

Fig. 4 Percentage abundance change for target nodes in Scenarios 1-4, each corresponding to incrementally increased M. religiosa abundance based on the pre-invasion baseline FIW. Points show median changes and vertical lines encompass 95% confidence intervals; these are derived from a sensitivity analysis (10,000 permutations). Point color indicates scenario; the color gradient reflects increasing mantid abundance

Figure 5 – Manipulating abiotic conditions and natural enemies

Cold snaps and parasitoid wasps effects on mantids. A two-panel plot is produced.

fig5_data <- bind_rows(
  list(S5 = sens_list$S5, S6 = sens_list$S6, S7 = sens_list$S7,
       S8 = sens_list$S8, S9 = sens_list$S9),
  .id = "scenario"
) %>%
  left_join(scenario_names, by = join_by(scenario)) %>%
  separate_wider_delim(scenario_labels, delim = ", ", names = c("winter", "wasps"), cols_remove = FALSE)

fig5_pal <- sequential_hcl(5, h = c(109, 252), c = c(100, 150, 7), l = c(86, 13), power = c(0.5, 0.9))[
  order(c(1, 3, 2, 4, 5))
]

Panel 1: mantids

fig5_mantis <- fig5_data %>%
  filter(name == "Mantids") %>%
  ggplot(aes(x = name, y = pct_change)) +
  geom_hline(yintercept = 0, linewidth = 0.2, linetype = "longdash") +
  geom_linerange(
    aes(ymin = pct_change_lc, ymax = pct_change_uc, group = scenario_labels),
    position = position_dodge(width = fig_dodgewidth), linewidth = 0.3
  ) +
  geom_point(
    aes(fill = scenario_labels), shape = 21, size = fig_pt_size,
    position = position_dodge(width = fig_dodgewidth)
  ) +
  scale_fill_manual(name = "Scenario", values = fig5_pal) +
  scale_y_continuous(breaks = seq(-100, 25, 25)) +
  labs(x = NULL, y = "Abundance Change (Percent)") +
  theme_bgl_s

Panel 2: target nodes

fig5_targets <- fig5_data %>%
  filter(name != "Mantids") %>%
  ggplot(aes(x = name, y = pct_change)) +
  geom_hline(yintercept = 0, linewidth = 0.2, linetype = "longdash") +
  geom_linerange(
    aes(ymin = pct_change_lc, ymax = pct_change_uc, group = scenario_labels),
    position = position_dodge(width = fig_dodgewidth), linewidth = 0.3
  ) +
  geom_point(
    aes(fill = scenario_labels), shape = 21, size = fig_pt_size,
    position = position_dodge(width = fig_dodgewidth)
  ) +
  scale_x_discrete(labels = guild_labs) +
  scale_fill_manual(name = "Scenarios 5−9: Winter Severity\nand Parasitoid Wasps", values = fig5_pal) +
  labs(x = NULL, y = "Abundance Change (Percent)") +
  theme_bgl_s +
  guides(fill = guide_legend(position = "inside", ncol = 1)) +
  theme(legend.position.inside = c(0.58, 0.16), legend.title = element_text(hjust = 1))

Arrange panels

fig5 <- ggarrange(
  fig5_mantis + theme(legend.position = "none"),
  fig5_targets +
    scale_y_continuous(
      limits = c(-3.4, 5.4),
      breaks = c(-3, -1.5, 0, 1.5, 3, 4.5),
      labels = c("-3.0", "-1.5", "0", "1.5", "3.0", "4.5")
    ) +
    theme(
      axis.title.y = element_blank(),
      legend.position.inside = c(0.80, 0.80)
    ),
  nrow = 1,
  ncol = 2,
  labels = c("a", "b"),
  hjust = c(-6.2, -5),
  vjust = 2,
  font.label = list(size = 12, face = "plain"),
  widths = c(0.25, 0.75),
  align = "h"
)
fig5

Fig. 5 Percentage abundance change for a M. religiosa and b target nodes in Scenarios 5-9. Scenarios correspond to altered winter severity and presence/absence of parasitoid wasps and are based on the post-invasion baseline FIW. Points show median changes and vertical lines encompass 95% confidence intervals; these are derived from a sensitivity analysis (10,000 permutations). Point color indicates scenario; the color gradient reflects increasing mantid abundance