Aim: Evaluating the relative impacts of land use and climate change on community change is challenging, and their impact may be context-dependent. Here, we use long-term moth community data to evaluate the relative impacts of changing habitats vs. changing climates on community composition and diversity of moths in different landscape settings and for moth species associated with different traits.

Location:  Finland, Northern Europe.

Time period: 1998-2020

Major taxa studied: Nocturnal macro-moths

Methods: We used Hierarchical Modelling of Species Communities to pinpoint moth species’ responses to climate and habitat composition in 109 sites across Finland. To characterise context-dependence, we extended this framework with conditional variance partitioning analysis. We used the model predictions to evaluate the relative effect of drivers on community diversity across Finland.

Results: The landscape context (i.e. the habitat composition around the site and its changes) emerged as the dominant driver of macro-moth communities. At the site level, where forests or shrub-like vegetation dominates, variation in species occurrence was mostly explained by local habitat conditions. In heterogeneous and water-dominated habitats, both habitat and climate variability contributed equally to patterns in species occurrence. At the species level, macro-moth responses to drivers of change varied according to their host plant affinity but independently of their wingspan. Climate and habitat changes can thus contribute congruently or unequally to community change, depending on the habitat. At the community level, traits also give insights into trends in and temporal variability of biogeographic patterns. 
Main conclusions: Our results underpin the importance of land-use change as a key driver of community change – even among heat-sensitive ectotherms. We also demonstrate that the sensitivity of local communities to climate and land use change varies among habitat profiles. Overall, our results highlight the importance of accounting for local conditions to understand and predict community patterns under global change.

Data: See manuscript for details.

• Moth occurrences: We used occurrence and abundance observations from the Finnish moth monitoring scheme (Nocturna). This scheme has been running across Finland since 1993 (Leinonen et al., 2016). Our data comprises 109 sites sampled every night from early spring to late autumn using ‘Jalas’ light traps (160 W mixed light or 125 W mercury (Hg) vapour bulbs) (Jalas, 1960). We focused our study on the main flight period of moths, as lasting from 1 April to 15 October. Our analyses were focused on years 1998-2020, which is the period for which both habitat and climate data were available. Since species of very low incidence will contribute little information on the drivers of distributions, we considered only macro-moth species with a prevalence higher than 10%. This resulted in a total of 56 966 observations (26.8% of all observations) and a total abundance of about 3.5 million of individuals (68% of the total abundance) for 1196 trap-years across 78 species.
• Moth traits: All information on wing size and host plant use were gathered from Yazdanian et al. (2023) and M. Hällfors et al. (2021).
• Climatic information: All climatic covariates were derived from 10x10 km gridded data provided by the Finnish Meteorological Institute (FMI; (Aalto, Pirinen, & Jylhä, 2016). Growing degree-days were calculated as the cumulative sum of temperatures over 5˚C during the sampling season (from 1^st     of April till mid-October). The cumulative summer-time precipitation was calculated for the same period of time, as based on daily precipitation values. In calculating winter chilling degree-days, we followed (Delgado et al., 2020) in summing negative deviations from 5 ˚C (as the reference temperature) during the winter preceding the sampling (from 15^th October till 1^st of April). Average winter-time snow depth for the winter preceding sampling was extracted from daily measurements of snow depth.
• Habitat information: We characterised the landscape around trapping sites by the overall level of fragmentation, and the proportions and diversity of four habitat types: broad-leaved forest, coniferous forest, mixed forest, and semi-natural or herbaceous type of habitat based on information from the CORINE land cover (CLC) database (Feranec, 2016), as available for years 2000, 2006, 2012 and 2018. We converted CLC data to a pixel resolution of 20x20 meters using the R package terra (Hijmans, 2021) and classified each pixel to one of the habitat categories.

Methods:

• Scripts: We provide the R scripts for the following:
  • Data preparation
  • Environmental analyses
  • Hmsc model run
  • Post analyses (Variance partitioning)
  • Predictions and community analyses