#Assessing the associations between renewable energy generation and other human-induced threats to terrestrial biodiversity #Authors: Antonella Gorosábel*, Harith Farooq, Mark Mulligan, Lars L. Iversen, Jonas Geldmann # CONSERVATION APPROACH # setup ------------------------------------------------------------------- library(tidyverse) library(terra) library(sf) library(readxl) library(sp) library(gridExtra) library(tidyr) library(boot) library(writexl) #As an example the script is written with the information of Birds in Europe. For the rest of the tables, change the "sheet name" and follow the same script table <- terra::vect("data/Birds_EU.shp") %>% st_as_sf() %>% setNames(gsub("B_", "", colnames(.))) %>% st_drop_geometry() %>% as_tibble() %>% select("layer", "SolC", "SolF", "WFC", "WFF", "HydroC", "HydroF", "Agr25", "Hun25", "Inv25", "Log25", "Pol25", "Urb25", "VU_spp", "EN_spp", "CR_spp") %>% rename( DamC = HydroC, DamF = HydroF) %>% rename_with(~ sub("25$", "", .), everything()) #Tables for each situation: present and future #PRESENT threat_table <- table[, c("SolC", "WFC", "DamC", "Agr", "Hun", "Inv", "Log", "Pol", "Urb")] #FUTURE threat_table <- table[, c("SolF", "WFF", "DamF", "Agr", "Hun", "Inv", "Log", "Pol", "Urb")] #Threatened species data for present and future species_info <- table[, c("VU_spp", "EN_spp", "CR_spp")] #################################################### # Mann-Whitney Test ------------------------------------------------------- #Repeat the same analysis for the future #Identify all possible overlaps between all threats in a new table combinations <- combn(colnames(threat_table), 2) overlap_matrix <- matrix(0, nrow = nrow(threat_table), ncol = ncol(combinations)) colnames(overlap_matrix) <- apply(combinations, 2, paste, collapse = "-") for (i in seq_along(colnames(overlap_matrix))) { # obtain the indices of the two columns (threat_table) en the combination idx1 <- which(colnames(threat_table) == combinations[1, i]) idx2 <- which(colnames(threat_table) == combinations[2, i]) # verify if there is an overlap for the combination of threats overlap_matrix[, i] <- as.integer(threat_table[, idx1] == 1 & threat_table[, idx2] == 1)} #Define as a data.frame for the following calculations table_comb <- as.data.frame(overlap_matrix) #Create a new data.frame to put the results results <- data.frame(threat = character(), variable = character(), U_statistic = numeric(), p_value = numeric(), mean_overlap = numeric(), mean_no_overlap = numeric(), max_mean = character(), significance = character(), mean_boost_overlap = numeric(), mean_boost_no_overlap = numeric(), max_mean_boost = character(), ci_overlap_low = numeric(), ci_overlap_high = numeric(), ci_no_overlap_low = numeric(), ci_no_overlap_high = numeric(), stringsAsFactors = FALSE) # Define the function for bootstrapping the mean bootstrap_mean <- function(data, indices) { sampled_data <- data[indices] return(mean(sampled_data, na.rm = TRUE))} #Define the information in the new table of results and run the Mann-Whitney U for (col in colnames(table_comb)) { #Indicate the cells with overlaps overlap <- which(table_comb[[col]] == 1) #Indicate the cells without overlaps no_overlap <- which(table_comb[[col]] == 0) for (var in colnames(species_info)) { # Calculate the mean number of species in each group mean_overlap <- mean(species_info[[var]][overlap]) mean_no_overlap <- mean(species_info[[var]][no_overlap]) #Determine which group has a higher mean value if (mean_overlap > mean_no_overlap) { max_mean <- "overlap"} else {max_mean <- "no_overlap"} #Calculate the Mann-Whitney U mw_test <- wilcox.test(species_info[[var]][overlap], species_info[[var]][no_overlap]) #Determine the significance of the test if (is.na(mw_test$p.value)) {significance <- "NA"} else if (mw_test$p.value > 0.05) { significance <- "Non significant"} else {significance <- "Significant"} #Bootstrap of the mean overlap_data <- species_info[[var]][overlap] no_overlap_data <- species_info[[var]][no_overlap] #Bootstrap for overlap data boot_overlap <- boot(data = overlap_data, statistic = bootstrap_mean, R = 1000) mean_boost_overlap <- boot_overlap$t %>% mean() #Obtain the CI and its limits for overlap data ci_overlap <- boot.ci(boot_overlap, type = "perc", conf = 0.95) ci_overlap_low <- if (!is.null(ci_overlap$percent)) ci_overlap$percent[4] else NA ci_overlap_high <- if (!is.null(ci_overlap$percent)) ci_overlap$percent[5] else NA # Bootstrap for no overlap data boot_no_overlap <- boot(data = no_overlap_data, statistic = bootstrap_mean, R = 1000) mean_boost_no_overlap <- boot_no_overlap$t %>% mean() #Obtain the CI and its limits for no_overlap data ci_no_overlap <- boot.ci(boot_no_overlap, type = "perc", conf = 0.95) ci_no_overlap_low <- if (!is.null(ci_no_overlap$percent)) ci_no_overlap$percent[4] else NA ci_no_overlap_high <- if (!is.null(ci_no_overlap$percent)) ci_no_overlap$percent[5] else NA #Determine which group has a higher bootstrapping mean value if (mean_boost_overlap > mean_boost_no_overlap) { max_mean_boost <- "overlap"} else {max_mean_boost <- "no_overlap"} # Save the results in the table results <- rbind( results, data.frame(threat = col, variable = var, U_statistic = mw_test$statistic, p_value = mw_test$p.value, mean_overlap = mean_overlap, mean_no_overlap = mean_no_overlap, max_mean = max_mean, significance = significance, mean_boost_overlap = mean_boost_overlap, mean_boost_no_overlap = mean_boost_no_overlap, max_mean_boost = max_mean_boost, ci_overlap_low = ci_overlap_low, ci_overlap_high = ci_overlap_high, ci_no_overlap_low = ci_no_overlap_low, ci_no_overlap_high = ci_no_overlap_high, stringsAsFactors = FALSE))} } #Present results_final <- results %>% separate(threat, into = c("Threat1", "Threat2"), sep = "-") %>% filter(Threat1 %in% c("WFC", "SolC", "DamC") | Threat2 %in% c("WFC", "SolC", "DamC")) #Future results_final <- results %>% separate(threat, into = c("Threat1", "Threat2"), sep = "-") %>% filter(Threat1 %in% c("WFF", "SolF", "DamF") | Threat2 %in% c("WFF", "SolF", "DamF")) write_xlsx(results_final, "results_final.xlsx") ##################################################### # Supplementary Material 2 - Table S2.1 ----------------------------------- # Overlap of threats #Select the variables of interest overlap <- table[, c("SolC", "SolF", "WFC", "WFF", "DamC", "DamF", "Agr", "Hun", "Inv", "Log", "Pol", "Urb")] #Create the pair combination of variables combinations_variables <- combn(names(overlap), 2) #Obtain the percentage of cells with overlaps results <- data.frame(Combination = character(), Cells_Presence = numeric(), Percenatge_Cells = numeric(), stringsAsFactors = FALSE) for (i in 1:ncol(combinations_variables)) { #Obtain each combination vars <- combinations_variables[, i] #Filter the rows that have the presence of both variable rows_with_presence <- overlap[overlap[, vars[1]] == 1 & overlap[, vars[2]] == 1, ] #Count the number of rows with the presence of both variables cells_presence <- nrow(rows_with_presence) #Calculate the percentage of cells percentage_cells <- (cells_presence / nrow(overlap)) * 100 #Save the results in the new table results <- rbind(results, list(paste(vars[1], vars[2], sep = "-"), cells_presence, percentage_cells)) } # Rename the columns of the result table colnames(results) <- c("Combination", "Cells_Presence", "Percenatge_Cells") print(results) #####################################################