Catch contrasts for day-to-day temperature changes had a negative adjusted interval only in later Sweden in four 2019-2020 samples

We tested mean absolute changes between consecutive daily mean temperatures against insect catch in Sweden and Madagascar. Four country and period models used 5,804 measured catches in 2019-2020. We compared current and future weather slopes while controlling mean weather, temperature spread, day order, trap, season and duration. Only later Sweden had a negative four-test adjusted interval: -0.0595 log(1 + grams/day) per C of mean day-to-day change, with an interval from -0.0744 to -0.0445. A post-fit error envelope widened it to -0.1108 to -0.0082. Seven-day blocks and a different catch transformation gave intervals across zero. The other three samples had intervals across zero in all planned fits. The evidence shows a model-sensitive association in one sample, not a common population response. Each country has one survey year. Catch combines abundance, activity and capture. Selected sites, incomplete catches, coarse weather, few time clusters and remaining seasonal bias limit inference. Code, aggregate results and source hashes support reproduction. Raw data and catch-level tables remain outside the package.

insect biomass · temperature change · observation bias · weather · temporal specificity · regional replication