Species distribution models applied to mosquitoes: Use, quality assessment, and recommendations for best practice

Abstract

Mosquito borne diseases (MBD) are a major global health concern. To aid MBD management efforts, the distribution of mosquito species is frequently investigated through species distribution models (SDMs). However, the quality these SDMs for management use has not been examined. We evaluated 127 publications of mosquito SDMs published between 1998 and 2020 and assessed each against a set of recently-developed, best-practice standards pertaining to quality of the response variable, predictor variables, model building, and model evaluation aspects. Mosquito SDMs were predominantly trained with presence-background response variables (77% of studies), bioclimatic predictor variables (39-63%), maximum entropy algorithm (54%), and evaluated by area under the receiver operating curve (36%) or confusion matrix metrics (34%). Aedes were the best-studied genus (70 studies). Pan-African (20%) and global (16%) distribution studies dominated. All published studies had one or more unacceptable standards within considered aspects, but no aspect observed unacceptable standards in all publications. The highest proportion of unacceptable standards were observed within predictor variables (60%), followed by model building (53%), model evaluation (34%), and response variable (17%). Response variable and model building demonstrated 8% and 0.2% increases in quality over time, but predictor variables and model evaluation exhibited 6% and 2% decreases in quality, respectively. Quality of mosquito SDMs has not changed since introduction of best practice standards. Quality of mosquito SDMs can be improved by ensuring known species temperature and precipitation thresholds are represented within the response variable. Resolution of predictor variables must be justified from ecological knowledge or statistically approximated. SDMs of mosquitoes require improved evaluation against independent data or creation of geographically-structured data. We encourage future mosquito SDM applications to utilize the most recent SDM standards and recommendations to improve applicability.