Larval chironomids (Diptera: Chironomidae) were investigated to provide climate reconstruction tools using transfer functions and based on 52&nsbp;lake samples from the Tibetan Plateau. Based on measured environmental variables an obvious electrical conductivity/salinity gradient from 0.015 to 130.0 mS · cm-1 was detected as the most influencing factor which can be used to reconstruct chironomid inferred salinities from lake sediment cores. Tested model types were: artificial neural networks (ANN), Bayesian, weighted averaging (WA), partial least squares (PLS), weighted averaging partial least squares (WAPLS), maximum likelihood (ML) and modern analogue technique (MAT). Performances of transfer models, tested by leave one out cross-validation, yield a maximum correlation value of r2LOOCV 0.762/0.764 with a root mean squared error of prediction (RMSEP) of 0.475/0.473 mS · cm-1log 10 for ANN models with three or four hidden neurons and a learning rate of 0.01. For apparent models r2app. varies from 0.958 to 0.664 with a root mean squared error (RMSE) of 0.200 up to 0.610 mS · cm-1log 10. Summarising all calculated transfer models with their summed error values and whether they yield balanced infer ed values of electrical conductivity a ranking can be stated as follows: Bayesian ≈ ANNneu=3/4-0.1/0.01 (with three or four hidden neurons and learning rates of 0.1 or 0.01) ≈ WAPLS-3 (with 3 components) > PLS-5 (5 components) ≳ WAcla+inv (classical/inverse) ≳ WA… tol (with tolerance downweighting) > W/MAT (weighted or unweighted) > ML. Sampling depth as a second important influencing factor detected by Canonical Correspondence Analysis (CCA) yields only week and unreliable transfer models with r2LOOCV = 0.475 and a RMSEP of 7.2 m. Furthermore the following measured environmental variables showed a statistical significant relationship in CCA for the benthic chironomid community: electrical conductivity, sampling depth, mean air temperature of October, mean precipitation of December, pH value and finally water area respectively in descending order of significance.
To enhance the part of determination the Chironomidae Identification Program CHIP was developed that provides scientists a convenient way to organise all literature, references, images and descriptions of scientific publications related to Chironomidae. It uses interactive, flexible local websites and the free programming language PHP with MySQL as database engine with the possibility of further open-source development. Developed for larvae primarily, the program can be used to work with pupae and adults as well. Including also a tool that provides data for normalised elliptical Fourier outline analysis of black/white scanned images this analysis was tested on related taxa to Psectrocladius in separating outlines of menta. Thereby it can give a great advantage for determination as decisions can be made more objectively and it should also be able to detect halves of menta.