loading page

Climate, human population density, and land cover link the distributions of two globally important dengue vectors from local to continental scales
  • +2
  • Christopher Anderson,
  • Erin Mordecai,
  • Lisa Mandle,
  • Jeffrey Smith,
  • Gretchen Daily
Christopher Anderson
Stanford University

Corresponding Author:cbanders@stanford.edu

Author Profile
Erin Mordecai
Stanford University
Author Profile
Lisa Mandle
Natural Capital Project, Stanford University
Author Profile
Jeffrey Smith
Stanford University
Author Profile
Gretchen Daily
Stanford University
Author Profile

Abstract

The distributions of mosquito vectors are expected to shift with rising temperatures due to climate change. But other global change patterns, like land cover change and human population growth, are simultaneously occurring. How will these changes interact to shift the future distributions of these vectors? Here, we analyzed how climate, land cover, and human population density regulate and predict habitat for two mosquito species, Aedes aegypti and Ae. albopictus, the primary vectors of dengue, Zika and chikungunya. We asked the following questions: How do environmental response curves derived from vector occurrence data compare to lab-derived responses? Based on these environmental response curves, which environmental drivers best predict the spatial distribution of each vector? Are environmental responses derived from large-scale (continental) occurrence data consistent at fine spatial scales? To answer these questions, we analyzed 6,317 Ae. aegypti occurrence records, 3,629 Ae. albopictus records, 10 satellite-derived environmental covariates, and two independent field surveys cover 134 sites. We found close agreement in the range of lab and environmental temperature responses, though the mean of observed temperatures was higher in the environment (31.0 °C for Ae. aegypti, 29.1 °C for Ae. albopictus) than lab predictions of the thermal optimum for transmission (29.1 °C for Ae. aegypti, 26.4 °C for Ae. albopictus). Using presence-only species distribution modeling approaches, we found that human population density was the best predictor for each vector’s spatial distribution (explaining 68.4% of model performance for Ae. aegypti, 48.7% for Ae. albopictus). These patterns were consistent in the field for presence/absence Ae. aegypti data (0.71 AUC, 0.80 recall), but failed to predict Ae. albopictus distributions in the sites we surveyed (0.53 AUC, 0.20 recall). In this session, we will explore these results and discuss the potential to predict and monitor Aedes habitat using satellite data.