**4. Discussion**

Our results show that changes in the forest-steppe ecotone followed the Holocene climate change in the Tien Mountains. As the degree of temperature and moisture changed, the forest belt changed, which further determines the position migration of the forest-steppe ecotone, as supported by the changes in the pollen taxa diversity observed by the deposition sites in the Tien Mountains (Figures 5 and 6). Under the co-influence of solar insolation and westerlies, climate change exhibits a unique pattern in arid central Asia, characterized by warm and dry climates during the early Holocene, warm and wet climates during the middle Holocene, and moderately cold and wet climates during the late Holocene (Figure 5a,b) [16,18,19,29,31]. The forest belt with *Picea* pollen dominance expanded and widened during the middle Holocene, leading to a decrease in the degree of species evenness of the conifer forest belt, as the observed pollen taxa diversity around the two lakes was very low. Specifically, on the northern slope of the Tien Mountains, the mid-elevation humid environment favors the existence of *Picea schrenkiana*, which largely a ffects mountainous diversity, as supported by our surface pollen results (Figure 3; Table 2). Indeed, this kind of mid-elevation forest distribution is popular in arid and semi-arid mountainous areas in China, which has gained wide attention [5,12,19,32]. Moreover, both the forest belt and the forest-steppe ecotone moved upward during the early Holocene and downward during the late Holocene, leading to a reduced forest belt influence on the sites, as the observed pollen taxa diversities were high (Figure 3; Table 2).

Indeed, Holocene temperature and moisture changes in arid central Asia have a ffected the vertical migration of vegetation belts and their species composition in mountainous areas [18,19]. The upper limit of the forest belt is controlled by temperature, while the lower boundary is limited by moisture. When the climate is warm and dry, the forest belt moves up along the elevation to acquire more moisture, and vice versa [11,33]. The species composition changes from wet-preferring to drought-tolerant species when climate drying exceeds its physiological limits [5]. Besides this, we found that the forest belt responded to the warm and wet climate through the width of the forest belt, which further made the forest-steppe ecotone move downward accordingly. This corresponding mode was also found in the Holocene change of vegetation belts in the Manas River drainage on the northern slope of the Tien Mountains [14]. Therefore, we emphasize that the migration of the forest-steppe ecotone responded

to the Holocene climate change as the forest belt changed, as supported by the change in pollen taxa diversity at the deposition sites.

Although the pollen taxa diversity in arid mountain areas can reflect vegetation belt change (Figures 5 and 6), does a narrowed forest belt lead to a decline in mountain pollen taxa diversity, and further a ffect the estimation of the forest-steppe ecotone migration? If the width or area of the forest belt becomes narrower, then the pollen taxa diversity observed at the deposition sites is considered to decline. However, we emphasize that if only the width of the forest belt is reduced, while forest species richness and evenness do not change, then the pollen taxa diversity may not necessarily decline. Furthermore, it is worth noting that some taxa in the forest-steppe ecotone may disappear when climate drying exceeds its threshold in arid mountainous areas. Moisture loss may pose the greatest threat to a relatively narrow forest-steppe ecotone, considering the climatic niche space of humidity-sensitive taxa and their limited living space. For example, as one of the diversity hot spots in the world, the narrow vegetation belts of the eastern Andes show grea<sup>t</sup> vulnerability to climate warming, as the expected rates of climate change may move the narrowly distributed taxa out of the climate niche space in hundreds of years [34]. More seriously, as global warming intensifies, regional aridity is expected to become more severe and last longer than in the present dry climate in arid regions [3]. Therefore, it seems that future climate warming with the associated lower moisture will probably cause unexpected threats to the taxa with narrow elevational distributions, especially for the taxa in the forest-steppe ecotone in the Tien Mountains. We thus propose that for the protection and maintenance of the current forest-steppe ecotone in mountainous areas, much attention should be paid to humidity-sensitive taxa under the drying climate.

Besides this, pollen taxa diversity may serve as an important proxy for reflecting the migration of the forest-steppe ecotone [4]. However, there are uncertainties, mainly due to the following aspects. Firstly, the limitation of pollen indicators—potential uncertainties come from di fferences in the pollen productivity and dispersal ability among various species [4,35–37], which may not present a one-to-one dynamic of vegetation belts and richness of fossil pollen. Secondly, the impact of vegetation structure—for wind-transported pollens, their ability to spread in the woodland is much worse than that in open land. This not only a ffects the amount of pollen that can be observed at the deposition site, but also a ffects our ability to recover vegetation using pollen data [38]. Thirdly, di fferent pollen types have di fferent representativeness—if a community with the same species composition has di fferent coverage, the yield of pollen will also be di fferent. When the number of statistical grains is constant, the low-representative species may not be counted, which will a ffect the estimation of pollen taxa diversity and in future will a ffect the estimation of the migration of the forest-steppe ecotone [38]. Because this study is mainly based on the Shannon-Wiener index and the Simpson index for calculating the pollen taxa diversity, which takes into account both richness and uniformity, taxa with very low pollen numbers or percentages have little e ffect on the results.
