Phantom midge-based models for inferring past fish abundances

We sampled living and subfossil phantom midge (Diptera: Chaoboridae) larvae from surface sediments of 21 small lakes in Southern Sweden to examine the influence of fish and selected abiotic variables on the abundance and species composition of chaoborid assemblages. We expected total Chaoborus abundance to be inversely correlated with fish abundance and Chaoborus species most sensitive to fish predation to be found only in fishless lakes. We aimed to use the observed relationships to develop models to reconstruct past fish abundances from chaoborid remains and the abiotic environment. C. flavicans occurred in almost every lake, whereas subfossil C. obscuripes were found in the surface sediments of only one fishless lake. The density of living C. flavicans larvae correlated negatively with fish abundance, lake order and size. The concentration of C. flavicans subfossils was negatively associated with pH, lake size, water transparency and fish abundance. Regression models that included lake morphometry and landscape position as additional predictors of fish abundance performed better than models that used only Chaoborus predictors. The explained variance in fish abundance varied from 52 to 86%. Leave-one-out cross-validation indicated moderate performance of the two best models. These models explained 51 and 56% of the observed untransformed fish density and biomass, respectively. In addition, all Chaoborus models were unbiased in closely following the 1:1 reference line in plots of observed versus predicted values. These results are a promising step in developing midge-based paleolimnological reconstructions of past fish abundance, and the approach might be improved by including chironomid remains in the models.