Temperature slope changes with inserted snow zeros depended on the model in Swedish insect data from 2019

We tested how inserted zero catches changed temperature slopes in the Swedish 2019 Malaise-trap archive. We split sites by latitude and records by half-year. Models compared the same traps with and without zero catches that had no sample ID. The archive has 276 such rows. All have at least 0.1 m mean snow depth. Source methods describe a snow-based zero insertion rule. In the northern early sample, the slope change was negative for log(1 + catch rate), but positive for log(catch rate + 0.01). A short-duration filter brought the first change close to zero. These results show sensitivity to the response and record selection. They do not show which model is correct or prove bias. One year, selected sites, common weather and changed seasonal support limit inference. The package gives code, source hashes, aggregate estimates and a figure. Raw data remain outside the publication.

insect biomass · observation bias · snow · imputation · temperature · seasonality · model sensitivity