Effects of varied modern pollen-climate calibration-set establishment approaches on quantitative climate reconstructions

A representative modern calibration dataset is an essential prerequisite for a reliable quantitative reconstruction of paleoclimate. However, there has been a notable lack of systematic research on the optimal filtering of calibration sets. Four representative high-resolution pollen records were selected to repeatedly reconstruct paleoclimate using multiple approaches such as geographical distance, analogues and climate gradient. Comparisons of pollen-based reconstructions indicate that these varied approaches exhibit significant discrepancy in model performance and reconstruction accuracy under different geographical circumstances and time scales. Generally, in terms of calibration-set selection, the geographical distance method demonstrates greater efficiency for the plain region, but exhibit limitations in areas characterized by significant topographic changes. Analogue-calibration has been found to be more suitable for the edge of the database and also has been proven superior to other two methods for reconstructions exceed Holocene. Selecting calibration sets based only on climate gradients is inappropriate for independent applications, but may act as a limitation for other methods. This preliminary research can contribute to a more precise projection of potential future climate changes by providing essential data input or validating hindcasts for Earth system models. •The approach of the calibration-set selection does matter much to the accuracy of quantitative climate reconstruction.•The distance-calibration is suitable for plain areas, but not for areas with significant topographic changes or atthe edge of the database.•Analog-calibration is aneffective method for the area with complex terrain, as well as for long-time series.•Gradient-calibration could not be used alone, but shouldbetaken into the other two methodsin the meantime.