Methods
Literature search
To construct the network of proposed drivers of insect decline from the non-primary literature, we performed the following literature on ISI Web of Science:
TS=(("insect" OR "insects" OR "insecta" OR "Megaloptera" OR "Alderflies" OR"Dobsonfl*" OR "Fishfl*" OR "Zoraptera" OR "Angel Insect*" OR "Neuroptera" OR "Antlion*" OR "Owlfl*" OR "Lacewing*" OR "Mantidfl*" OR "Hymenoptera" OR "Ants" OR "ant" OR "Bee" OR "bees" OR "Wasp" OR "Sawfl*" OR "Psocodea" OR "Barklice" OR "Booklice" OR "Parasitic Lice" OR "Coleoptera" OR "Beetle*" OR "Microcoryphia" OR "Bristletail*" OR "Lepidoptera" OR "Butterfl*" OR "Moths" OR "moth" OR "Trichoptera" OR "Caddisfl*" OR "Blattodea" OR "Cockroach*" OR "Termite*" OR "Odonat*" OR "Dragonfl*" OR "Damselfl*" OR"Dermaptera" OR "Earwig*" OR "Siphonaptera" OR "Flea*" OR "Diptera" OR "Flies" OR "fly" OR "Orthoptera" OR "Grasshopper*" OR "Cricket*" OR "Katydid*" OR "Notoptera" OR "Ice Crawler*" OR "Rock Crawler*" OR "Mantodea" OR "Mantid*" OR "Ephemeroptera" OR "Mayfl*" OR "Protorthoptera" OR "Mecoptera" OR "Scorpionfl*" OR "Hangingfl*" OR "Zygentoma" OR "Silverfish" OR "Raphidioptera" OR "Snakefl*" OR "Phasmid*" OR "Plecoptera" OR "Stonefl*" OR "Thysanoptera" OR "Thrip*" OR "Hemiptera*" OR "True Bug*" OR "Cicada*" OR "Hopper*" OR "Aphid*" OR "Strepsiptera" OR "Embiidina" OR "Webspinner*" OR hoverfl* OR bumblebee* OR pollinator*) NEAR/10 ((declin* OR extinction OR "insect apocalypse" OR "insectageddon" OR "biodiversity loss" OR "biodiversity crisis" OR "population decline" OR "pollinator decline")))
This initial search identified 3,387 research articles that have been published since 2017. We then screen for reviews, meta-analyses, or perspectives that specifically proposed causes of insect declines. This final set of papers was then read in full and all hypothesized causal pathways between a driver and an outcome for insects were extracted.
Network creation
Each pathway was turned into an edge, with the linked ideas forming the nodes (drivers, mediators, outcomes). Nodes were categorized at different levels of resolution. Each edge was tagged with taxa it referred to, with varying specificity. For instance, the pathway “light pollution -> moth abundance” would be associated with both “Lepidoptera” and “insects” for taxa. Finally, a variable was created for each edge to indicate what the root cause was when it was proposed. For example, the root cause of the pathway “nitrification -> floral resources -> larval mortality” is nitrification. The network was built using the igraph package and is presented in an interactive form using the shinyR package in R.
