Dinoflagellate cyst based modern analogue technique at test — A 300 year record from the Gulf of Taranto (Eastern Mediterranean)

•Regional reference data based MAT reconstructions are better than global based.•Excluding degradable species leads to loss of variability in the reconstructions.•MAT and qualitative reconstructions co-vary to instrumental data for industrial times.•MAT reconstructions do not co-vary to instrumental data for pre-industrial times.Coastal reference dataset samples do not represent pre-industrial conditions.

To test the performance of dinoflagellate cyst based palaeoenvironmental reconstructions derived via the modern analogue technique (MAT) and a non-quantitative method, we have compared, reconstructions of mixed layer temperature, sea surface temperature and mixed layer salinity of a well dated site from the Gulf of Taranto, to instrumental data of air temperature, precipitation and river discharge. Two hypotheses are tested: the reconstructions of dinoflagellate cyst based MAT improve when a) a local reference dataset rather than an extra-regional reference dataset is used, b) the reference and downcore datasets are corrected for species specific preservation prior to analysis. This is achieved by executing four experiments based on different reference datasets 1) a North Atlantic dataset including all species, 2) a North Atlantic dataset including degradation resistant species, 3) a Mediterranean dataset including all species, 4) a Mediterranean dataset including degradation resistant species only.We show that MAT based reconstructions improve when a local rather than an extra-regional reference dataset is used. Exclusively including species resistant to degradation improves the reconstruction if an extra-regional reference dataset is used, but leads to considerable loss of variability if the local reference dataset is used.Both MAT-based and qualitative reconstructions correlate to instrumental data for industrial times. For pre-industrial times, MAT reconstructions of variability in temperature and salinity do not co-vary with variability in the instrumental data, whereas qualitative reconstructions provide a relatively good fit. We suggest this to be a result of the anthropogenic influence on coastal marine ecosystems existing today at the majority of the coastal sample sites in the reference dataset. As a result, conditions at these sites do not correspond to pre-industrial conditions. We advise users of dinoflagellate cyst transfer functions to pay more attention to this aspect.