The application of salt-marsh foraminifera to reconstruct historical sea-level trends was investigated for the Croatian coast of the Adriatic Sea using a transfer function approach. This technique, whilst well practised from north Atlantic sites along the shores of America and UK, has previously evaded any published study in the Mediterranean region. A total of 70 surface samples were collected across separate transects established at two micro-tidal salt-marsh sites from the central Croatian coastline to establish a modern dataset of foraminifera. In addition, environmental variables were also investigated including salinity, pH, organic matter, particle size, distance (from open water) and altitude, relative to the Croatian national datum. Three sediment cores were sampled for fossil foraminifera and composite chronologies involving short-lived radionuclides, radiocarbon dating and pollution indicators from XRF. Age-depth models were created using classical and Bayesian approaches. Quantitative analysis of the foraminiferal assemblages revealed on average three faunal zones in which characteristic species occurred. This comprised a faunal zone composed almost exclusively of agglutinated species; J. macrescens, T. inflata and M. fusca extending between mean tidal level (MTL) and beyond MHWST (higher altitude). A second faunal zone was more variable and comprised of a mixed assemblage of agglutinated species described above in addition to calcareous species; Ammonia spp., Elphidium spp., Haynesina germanica and Quinqueloculina spp. This zone spanned a large vertical range above and below MTL. In a third faunal zone calcareous species dominated and was restricted to the lower altitudes of the salt-marsh environments. Further quantitative measures were employed to test the hypothesis that foraminiferal distributions were controlled by altitude. Partial ordination techniques revealed altitude as a statistically significant control confirming their suitability as proxies for sea-level in transfer function reconstructions. An analysis of species environment relationships revealed strong linear response suggesting the use of PLS regression models. Transfer functions were then developed for both site specific and a total combined dataset, where small r2 jack values largely reflected the short environmental gradients despite relatively low predictions errors (RMSEP jack =
展开▼
