Neogene and early Quaternary pollen-based probabilistic climate reconstructions in the Netherlands

Abstract

Both Miocene and Pliocene can be used as analogues for future climate scenarios. Both time periods showed higher global temperatures and precipitation levels than today. Fossil pollen can be used to reconstruct past vegetation and climate changes. There is a long tradition of quantitative climate reconstructions for the Netherlands for the Neogene and Quaternary. These previous studies also defined stratigraphical zones that are important for the stratigraphy for northwestern Europe. Climate reconstructions showed warmer Neogene conditions and alternating glacial and interglacial periods during the Early Pleistocene. However, these reconstructions need better validation and reproducibility. The Climate REconstruction SofTware (CREST) method is a new statistical model that uses probability density functions (pdfs) to quantify climate reconstructions based on modern day plant occurrences and can be used to validate previous reconstructions. In this study, the CREST method has been used on new and legacy pollen data from the Netherlands, covering intervals of time from the Miocene (~17 Ma) to the Early Pleistocene (~2.4 Ma). Two different models are used, one including all taxa that could be linked to modern day taxa and the second model excluding taxa based on four different criteria: aquatic taxa, bisaccate pollen, widespread taxa and possible reworked pollen. These two models are used to reconstruct four different climate variables; Mean Annual Temperature, Mean Temperature of the Warmest Quarter, Mean Temperature of the Coldest Quarter and the Mean Annual Precipitation. Excluding taxa from the model showed better reconstructions for extreme high and low values. Furthermore, excluding taxa showed in general higher reconstructed temperatures and precipitation levels. The Leave One Out analyses showed a relatively large influence of rare taxa on the climate reconstruction. Trends observed in the Miocene to Pleistocene climate using this model align fairly well with previous studies. CREST reconstructions show in general lower temperatures compared to previous studies using different type of proxies (e.g. brGDGT, megafloras or TEX86) in or close to the North Sea basin. These differences might be due to biases due to dispersal when comparing marine and terrestrial deposits. Furthermore, it might be due to the inclusion of too many moderate informative taxa in the CREST models, that could level climate reconstructions to average value. In conclusion, these results suggest that the CREST method can be used to capture overall trends in the past climate, but better calibration could improve the reconstructions.