example for RM evaluation and subsequent analysis

Shows the evaluation of RMs at two resolutions, provides an example for mapping and data extraction and an analysis of differences in performance related to spatial resolution.

inst/examples/evaluateRM_help.R

@@ -0,0 +1,73 @@
+# EcoRM evaluation at different resolutions (>4min runtime)
+root.dir  <- list.files(system.file(package = "gbif.range"), pattern = "extdata", full.names = TRUE)
+
+res1km <- evaluateRM(root.dir = root.dir, 
+           valData.dir = "SDM", 
+           ecoRM.dir = "EcoRM",
+           verbose = TRUE, 
+           print.map = TRUE,
+           valData.type = "TIFF", 
+           mask = NULL, 
+           res.fact = NULL)
+
+res10km <- evaluateRM(root.dir = root.dir, 
+                   valData.dir = "SDM", 
+                   ecoRM.dir = "EcoRM",
+                   verbose = TRUE, 
+                   print.map = TRUE,
+                   valData.type = "TIFF", 
+                   mask = NULL, 
+                   res.fact = 10)
+
+
+# Extract and plot a specific overlay map
+terra::plot(
+  res10km$overlay_list[[1]],
+  col = c("gray", "red", "blue", "purple"),
+  breaks = c(-0.5, 0.5, 1.5, 2.5, 3.5),
+  legend = FALSE,
+  main = paste0("Species: ", res10km$df_eval[1,1],"\n", 
+                "Sensitivity = ", round(res10km$df_eval[1,6], digits = 2)," & ", 
+                "Precision = ", round(res10km$df_eval[1,7], digits = 2)),
+  las = 1
+)
+
+legend("bottomright",
+       legend = c(
+         "Abs in both",
+         "Pres in ecoRM only (FP)",
+         "Pres in valRM only (FA)",
+         "Pres in both (TP)"
+       ),
+       fill = c("gray", "red", "blue", "purple"),
+       bg = NA,
+       box.col = NA,
+       inset = c(0,0.2)
+)
+
+## Compare sensitivity and precision outputs ##
+# Calculate overall performance (e.g. mean sensitivity & precision)
+res1km$df_eval$Mean_SenPrec <- (res1km$df_eval$Sen_ecoRM+res1km$df_eval$Prec_ecoRM)/2 
+res10km$df_eval$Mean_SenPrec <- (res10km$df_eval$Sen_ecoRM+res10km$df_eval$Prec_ecoRM)/2 
+
+# Combine the data frames and add a resolution column
+combined_df <- bind_rows(
+  res1km$df_eval %>% mutate(Resolution = "1km"),
+  res10km$df_eval %>% mutate(Resolution = "10km")
+
+)
+
+# Gather data into long format
+long_df <- combined_df %>%
+  select(Sen_ecoRM, Prec_ecoRM, Mean_SenPrec, Resolution) %>%
+  pivot_longer(cols = c(Sen_ecoRM, Prec_ecoRM, Mean_SenPrec), 
+               names_to = "Variable", values_to = "Value")
+
+## Plot results for different range data resolutions ##
+# Sensitivity increases and precision decreases with increasing resolution
+# Overall performance is slightly higher at more coarse grain
+ggplot(long_df, aes(x = Variable, y = Value, fill = Resolution)) +
+  geom_boxplot() +
+  labs(x = "Variable", y = "Value", title = "Boxplot of Sen_ecoRM and Prec_ecoRM") +
+  scale_fill_manual(values = c("10km" = "#619CFF", "1km" = "#FFC300")) +
+  theme_minimal()