Biologists have long been fascinated by the processes that give rise to the phenotypic complexity of organisms, yet whether there exist geographic hotspots of phenotypic complexity remains poorly explored. Phenotypic complexity can be readily observed in ant colonies, which are superorganisms with morphologically differentiated queen and worker castes analogous to the germline and soma of multicellular organisms. Several ant species have evolved ‘worker polymorphism’, where workers in a single colony show quantifiable differences in size and head-to-body scaling. Here, we use 256,754 occurrence points from 8,990 ant species to investigate the geography of worker polymorphism. We show that arid regions of the world are the hotspots of superorganism complexity. Tropical savannas and deserts, which are typically species-poor relative to tropical or even temperate forests, harbor the highest densities of polymorphic ants. We discuss the possible adaptive advantages that worker polymorphism provides in arid environments. Our work may provide a window into the environmental conditions that promote the emergence of highly complex phenotypes.

We reviewed the literature and classified 15,518 species of ants as either 'polymorphic’ according to a broad definition of worker polymorphism or as ‘monomorphic’ (Table S1, S2). Species lacking significant size or allometric variation in the worker caste were classified as monomorphic whereas all others were classified as polymorphic (see Fig. S1 for more details). To classify ant species as polymorphic or monomorphic, we first systematically searched AntWiki [34] and AntWeb [35], two open-access repositories curated by myrmecologists that collate information and list studies on each ant species, for relevant literature. These databases were chosen because resources are easily accessible and the taxonomic upkeep of the species pages is consistent [34, 35]. Using the primary literature cited on these pages, we searched for explicit mention of the terms monomorphic, polymorphic, or any of the categories of polymorphism defined by Wilson [21]. When multiple sources were available for a given species, we used the most recent article.

We used the species occurrence data from the Global Ant Biodiversity Informatics(hereafter referred to as GABI) database [43]. This occurrence database is a synthesis of data from published literature (~10,000 publications), as well as from online, open-source databases, museum records (including 87 online open-source databases, such as  Antweb), and personal collection databases. The taxonomy of legacy records is placed in a common framework and continuously updated with new data. It also includes records extracted from 87 online, open-source databases, such as Antweb, as well as museum records. Species name validity is being checked for each record and follows the Bolton Catalogue available from AntCat.org. Species identification errors are corrected on the basis of taxonomic revisions and biodiversity literature (e.g. species checklist),  contact with experts, and direct decisions based on biogeographic knowledge.  extracted all spatially validated occurrence points from the database (accessed in May  2018), and after excluding occurrence data from oceanic islands, we retained 256,754points matching 8,990 ant species for our analyses.