Rainfall, not temperature, explained year-to-year change in female mosquito catches at 47 NEON sites in 2016–2024; the data did not resolve lab thermal-curve predictions

Question: do female mosquito catches at NEON sites follow the year-to-year temperature sensitivity that unimodal laboratory thermal performance curves predict (more mosquitoes in warm years at cool sites, fewer at warm sites)? And which weather variable explains year-to-year change? Approach: we used NEON CO2 light-trap data (DP1.10043.001, RELEASE-2026, through neonDivData) and Daymet daily weather. We made quasi-Poisson annual indices with month correction for 377 site-years at 47 sites (2016–2024; 49,858 night samples; about 9.6 million females). We wrote the analysis plan before we fitted any model. In the primary test, a random-effects meta-regression of the per-site temperature sensitivity on site mean temperature gave a slope of −0.012 per °C per °C (SE 0.018). The pooled model gave +0.001 (SE 0.018; site-permutation p = 0.96). The sensitivity at a 15 °C site was 0.00 per °C (SE 0.09). Temperature explained 1.6% of within-site variance. The values that the published curves (Shocket et al. 2020) predict, −0.023 to −0.027 and +0.08 to +0.09, are inside these intervals. In a direct test of observed anomalies against curve-predicted log-density anomalies, the slope was 0.64 (95% CI 0.05–1.23) with all sites. Without the one Arctic site, BARR, the slope was 0.20 (−0.05 to 0.44). The random-effects mean of the per-site slopes was 0.19 (SE 0.50). Thus the transfer of lab curves to field abundance is not resolved. Rainfall in the sampling window had a robust positive effect: elasticity 1.17 (95% CI 0.58–1.75), which explained 10.4% of within-site variance. The elasticity stayed between 0.87 and 1.44 in all robustness checks (year jackknife, year effects, other indices, other pseudo-counts, without remote sites). It was positive in all five genera, largest for floodwater Psorophora (3.4) and smallest for Coquillettidia (0.6). Rain in the next year (placebo) had no effect (−0.25, CI −0.79 to 0.28). Main limitations: only 5–9 years per site with small temperature anomalies, so the temperature null is a precision limit and not evidence of no effect. The indices are trap catches of host-seeking females. The weather is from a 1 km grid. We added a pseudo-count for zero-catch years after the plan; the results did not change with other pseudo-count values.

mosquito · Culicidae · NEON · thermal performance curve · precipitation · temperature sensitivity · interannual variability · Psorophora · Culex · Aedes · Daymet · vector ecology · climate change