Computational protocol: Recent Invasion of the Symbiont-Bearing Foraminifera Pararotalia into the Eastern Mediterranean Facilitated by the Ongoing Warming Trend

Similar protocols

Protocol publication

[…] The obtained 19 sequences of P. calcariformata were analysed together with 24 sequences of benthic foraminifera belonging to the lineage of Globothalamea [] downloaded from GenBank (see ). The sequences were automatically aligned with MAFFT v.7 [] with default options. Only the fragment covered by the obtained sequences of Pararotalia was retained for further analyses (see ). The model of evolution (GTR+I+G) was selected using jModeltest v. 2.1.4 [] under Akaike Information Criterion (AIC). Using this model of evolution, the most likely tree topology was inferred from the alignment using a Maximum Likelihood Approach implemented in PhyML 3.0 software [], using NNI+SPR tree improvement and non-parametric bootstrapping (1000 pseudo replicates). The resulting tree was visualized with iTOL v 2.1 () []. The two symbiont sequences were compared to the SILVA database [] on the 21/08/2014 in order to determine their most probable taxonomic assignation. The SINA 1.2.11 alignment tool [] has been used with default options. [...] Occurrence records of P. calcariformata in the Mediterranean were obtained by literature search and combined with new observations during this study (see ). For the calibration of the species distribution model (SDM), occurrences were converted to presence records on a grid used by the modeling software. Environmental data for these grid cells were obtained from the BIO-Oracle database, which provides oceanographic variables with a grid-cell resolution of 5 arc minutes []. BIO-Oracle also includes gridded data from climate model projections that are based on SRES climate-change scenarios [] and for our model we used the intermediate scenario A1B for the a projection to year 2100. We based the SDM for P. calcariformata mainly on temperature (annual minimum SST) and added annual mean diffuse attenuation (mean DA) and annual mean photosynthetically available radiation (mean PAR). The latter variables provided the possibility to incorporate the effects of terrestrial and trophic influences, as well as solar radiation on the potential distribution. These variables have been proven useful in previous modeling calibrations from other symbiont-bearing foraminifera []. The resulting SDM was refined in a two-step clipping process in order to avoid a biased relation between the variables, an approach that has been successfully used in other models on foraminiferal distributions []. First, we used only minimum SST, which was subsequently projected on the future climate scenario. Second, we built a model on mean DA and mean PAR. The final SDM for both current and future conditions () thus comprises a climate-model based on temperature (including the projection on the A1B scenario), which was overlain and clipped by a habitat-model based on the other variables. The editing of the climate model was performed with the software DIVA-GIS.We used Maxent 3.3.3k [] to model the potential distribution of P. calcariformata in the eastern Mediterranean and to project it onto future climate conditions. The program uses a grid-based machine-learning algorithm following the principles of maximum entropy []. In the course of the modeling process, Maxent begins with a uniform distribution and successively fits it to the data (occurrence records and environmental variables). For an overview on the operating mode of Maxent and the interpretation of its output see []. Note that Maxent does not predict the actual distribution of the taxon, but rather the relative suitability of the habitat, which is interpreted as the potential distribution of the taxon under study. A total of 10,000 random background points were automatically selected by Maxent within the eastern Mediterranean. The logistic output format with suitability values ranging from 0 (unsuitable) to 1 (optimal) was used [], where the probability of presence at sites with "typical" conditions is set to 0.5 by default []. The modeling process was performed with 50 replicates and the average predictions across all replicates were used for further processing. The continuous probability surfaces of the SDMs were subsequently converted into presence/absence maps using the “Equal training sensitivity and specificity logistic threshold” as recommended by [], which has also been used in previous foraminiferal models [].Projecting a model on future climate scenarios may result in an extrapolation or “clamping” of the probability values [] especially in regions where the environmental predictors are outside the training range, which could lead to an over- or underfitting of the model. In Maxent, a multivariate similarity surface (MESS) analysis is implemented, which shows how similar predictor variables within future climate scenarios are seen during model training []. We added the result of the MESS analysis to our future model, highlighting areas of possible extrapolation of the model due to minimum temperature values of the future scenario being outside the training range. […]

Pipeline specifications

Software tools MAFFT, jModelTest, PhyML, iTOL, DIVA-GIS
Application Phylogenetics