Fig. 4

Species identification of immature mosquito larvae. Immature mosquitoes were obtained by filtering water collected from different pools, followed by 2–3 rinsing steps before killing and storing them at − 20 °C. Samples, mostly 3rd instar larvae were analysed with the same REIMS settings used for adult specimens. Their species was confirmed by sampling the same pool of larvae and raising them to adults before, during and after taking samples to be used for this model. The differences detected by PC-LDA are visualised in the form of an OMB model (a) and kernel density and scatterplots produced in R (b). The separation was put to test via cross-validation in OMB (c) and random forest analysis with training/test dataset split of 70%/30% (d). The separation of Ae. detritus (n = 125) and Ae. punctor (n = 125) larvae required little information/variance, reaching very distinct separation with only 20 principal components for PC-LDA. When testing the random forest model, samples were identified correctly most of the time leading to correct identification rates of 100% (Ae. detritus) and 99% (Ae. punctor)