Chironomid-climate continentality conundrum

PLOS One | https://doi.org/10.1371/journal.pone.0327780 August 5, 2025 13 / 18 Creating a kerner oceanity index-based model Two continentality indices (ATR, KOI) were tested in this study. ATR is calculated as the difference between the cold- est and the warmest months, while the KOI calculation uses the ATR value and incorporates spring (April) and autumn (October) temperatures. A previous chironomid-continentality study [24] used the Gorzynski continentality index, which is also ATR-based, but includes latitude. However, as our dataset is constrained to a narrow latitudinal band (Fig 1, Table 1), it is assumed that any impact of latitude would not be detectable. Furthermore, the Gorczynski continentality index is not applicable to oceanic sites [102] and cannot therefore be used for coastal areas of the eastern Baltic and southern Scandinavia. The RMSEP of the chironomid-inferred WA-PLS KOI model is 5.1. Considering that the RMSEP covers about 15% of KOI gradient length in the dataset (−4.5–27.5; Table 1), this seems to be a very promising variable. However, no KOI-based model has been published so far. In Self et al. [24], where the Gorzynski continentality index was used, the reported R2 value was 0.73, which is the same as in our model. The prediction errors (Fig 6) tend to increase from the continental to oceanic part of the dataset; a pattern also seen in Self et al. [24]. The issue could be because our dataset is relatively small but covers a large geographical area, which is reflected in the species occurrence pattern, with only 66%

of the species occurring in at least five sediment samples (10% of the investigated dataset). Also, the higher proportion of continental than transitional and oceanic sites, increases the robustness of the continental part of the model. An increase in the number of sites, especially transitional and oceanic ones, and in the density of the training set may improve the model’s accuracy. July air temperature has previously been considered as the main driver to explain chironomid assemblage distribution [21,39,42,85]. The high amount of variation explained by July air temperature in our dataset, where the summer tempera- ture gradient was intentionally reduced by the sampling design, highlights its importance for chironomid assemblages in the selected study area. However, the performance of the current dataset in the statistical tests indicates that chironomids may be used as a continentality proxy. Development of specialised training sets dedicated to specific parameters are an essential prerequisite for successful reconstructions of different palaeoenvironmental variables. Conclusions Summer temperatures are commonly considered to be a key driver of chironomid assemblage patterns. However, in our dataset, collected along a transect from the oceanic Atlantic coast to the continental central Russia, July air temperature is the second strongest explanatory variable and a part of a general longitude-related group of variables, which also includes annual temperature range, ice-cover duration, and growing degree days (GDD5). The Kerner Oceanity Index (KOI) rep-

resents a distinct gradient in the dataset. It explains the highest variation in the chironomid assemblages and is indepen- dent of the other tested variables. Therefore, KOI is a comprehensive continentality metric in our dataset, accounting for both the annual temperature range and the spring and autumn air temperatures. Despite the dataset being relatively small for creating a robust palaeoclimate continentality reconstruction, the WA-PLS model performance for KOI shows promising results with an R2 = 0.73 and RMSEP of 5.1. We conclude, therefore, that further investigation of chironomid-continentality relationships and the creation of a larger continentality-based chironomid training set are justified. Supporting information S1 File. Chironomidae assemblages composition of the continentality dataset. (CSV) S2 File. Environmental data. (CSV)