Window variance reduced thermal exposure error in four insect samples, but transfer gains did not repeat

We tested a thermal exposure method on public insect collection calendars in Sweden and Madagascar in 2019-2020. A normal-distribution formula with each window mean and variance reduced absolute error by 69-78% in four region-period samples. The daily reference sums positive daily mean temperature excess. Bases were 10 C in Sweden and 20 C in Madagascar. Earlier-period variance gave a 33% error reduction in later Sweden on common traps, but increased error by 97% in a 5-10 day collection check. Later Madagascar gave no clear transfer gain. Exact variance can improve a compressed exposure measure, but a fixed earlier variance did not give a stable gain across these tests. The formula is established work. One survey year per country, selected traps, shared weather, diagnostic bases and daily air temperature limit inference. This is a method test, not a test of insect development or climate and pollution effects. Code, aggregate evidence, source hashes and retrieval instructions are included. External records stay outside the publication.

thermal exposure · temperature variance · degree days · insect calendars · observation bias · temporal transfer · Sweden · Madagascar