“…For comparison, the proportion of variance explained by climate variables in CCAs or RDAs is 3.9-8.1% in northern Europe (Salonen et al, 2014), 25% in southern America (Schäbitz et al, 2013), 18.6% in the Tibetan Plateau (Lu et al, 2011), 11.4-15.6% in Siberia (Klemm et al, 2013), 5.6-19.8% in Mongolia (Tian et al, 2014, and 3.3-14.3% in southern Europe (Finsinger et al, 2007). The relatively low proportion of variance explained by climate variables in CCA or RDA can be possibly attributed to many factors, such as the potential source area of pollen from soil samples reflecting local rather than regional vegetation (Li et al, 2005;Zhao et al, 2009;Pan et al, 2010), the effects of taphonomic processes related to the soil oxidation that can remove some pollen types (Li et al, 2005;Phuphumirat et al, 2011;Gil-Romera et al, 2014), the relatively large size of these datasets with a high number of pollen types and many zero values (ter Braak, 1986;Guisan et al, 1999;Birks et al, 2010), and the uneven spatial distribution of surface samples in different geographical regions (Salonen et al, 2014). Such datasets are unable to cover completely and evenly the climatic gradients, which likely cause hidden nuisance gradients in the datasets (Salonen et al, 2014).…”