For each record we compiled information about latitude, longitude, site name or location, country, sample depth, number of samples, depositional environment, current climate based on the climatic classification of Beck et al. (2018) (see above and Figure 1), and relevant publications.

We used chronology-control points for the construction of an age-depth model for each record. We selected only the chronology control-point types appropriate for estimating robust age-depth models (see Supplementary Table 2). We only used records with more than two chronology-control points of the selected chronology-control types including core-top age.

We preferred fossil pollen records with raw counts for further data processing and analysis. However, 48 records with pollen percentages originating from data-sparse regions were also considered. We only included records collected from selected sediment depositional environments (Supplementary Table 3), and pollen counts or percentages for terrestrial trees, shrubs, herbs, palms, succulents, and mangrove taxa. We subjected the records to taxonomic harmonisation (see below). We then processed the taxonomically harmonised pollen records based on age criteria. For this, we set the calibrated age of 12,000 cal year BP (calibrated years before present, where 0 year BP = 1950 common era CE) as the oldest age-limit (lower age-limit) for each record and discarded samples older than the lower age-limit. We also discarded records in which the calibrated age of the youngest sample (upper age-limit) was >1,600 cal year BP or the age of the oldest sample was <5,800 cal year BP. Thus, the lower age-limit of the selected records varies from 5,800 to 12,000 cal year BP, whereas the upper age-limit varies from 0 to 1,600 cal year BP. We then processed the record based on the number of taxa and samples. We only selected records with five or more samples and with five of more harmonised taxa. We then filtered the records based on the number of pollen grains in each sample of a record, where we discarded samples containing <25 grains. We also discarded samples with 25–150 grains if 50% of the samples in a record have more than 150 grains. This was done to avoid losing too many pollen records with only a few samples that consistently have a low pollen count. For a statistical summary of the pollen data, see Supplementary Table 4.

In each record we analysed the spatial patterns of the total pollen-compositional variation and temporal pattern of each PAP (except MRT partitions) along the temporal gradient of 12,000 years.

We fitted a generalised additive model (GAM) regression using “mgcv 1.8-40” package in R (Wood, 2022) for analysing latitudinal and longitudinal patterns of total compositional turnover and compositional change (MRT partitions), where we treated these estimates as a response variable against latitude and longitude of the records as predictor variables. We assumed a “Tweedie” error distribution (p = 1.1) for the models of compositional turnover and a “Poisson” error distribution for the MRT partition models.

For cross spatial-scale comparison of the temporal patterns of PAPs during the last 12,000 years, we analysed the patterns of each PAP at three spatial scales (grains): “Pollen record level” (patterns for individual records), “Climate-zone level” (assembled the records into the five climate-zones and summarised the patterns of each PAPs within each climate-zone), and “Continent level” (summarised the temporal patterns of each PAPs for the entire continent).

For “record level,” we developed a single generalised linear model (GAM) for each record using “mgcv 1.8-40” package in R (Wood, 2022), in which we treated estimates of each PAP as a response variable and time (ka) as a predictor variable. However, for the “climate-zone level,” we developed one hierarchical generalised additive model (hGAM) for each climate-zone, and for the “continent level,” we developed a single hGAM for the whole continent using “mgcv 1.8-40” package in R (Pedersen et al., 2019).

We treated location (“dataset_id”) of each record as a random effect variable with a separate smoother and penalty for each record within the climate-zone, and a shared smoother and penalty for the overall (global) model of the climate-zone (similar to Model GI in Pedersen et al., 2019). For the temporal patterns at the “continent level,” we fitted a single hierarchical GAM model with a separate smoother and penalty for each “dataset_id” and a shared smoother and penalty for the overall (global) model for the whole continent (similar to Model GI in Pedersen et al., 2019). We assumed a “Tweedie” error distribution (p = 1.1) for models of N0, N1, N2, DCCA1, and RoC, and a “beta” error distribution for models of evenness measures (N2/N1 and N1/N0). For all models we predicted values for each 100 years but limited the prediction to only span the time period with data (i.e., we did not fore- or back-cast more than 100 years from the first and last data point, respectively). For “records level” patterns, we summarised the similarity between patterns of records using the root-mean-square error (RMSE), calculated for each 100 years. All figures of the spatial and temporal patterns are plotted using “ggplot2” (Wickham, 2016).

Holocene-wide patterns of Hill’s (1973) diversity estimates are rather heterogeneous across spatial scales. Pollen richness (N0) at the record level is highly heterogeneous and shows no general temporal pattern with predicted values ranging between 2.4 and 48.7 pollen-taxa (CI 1.4–53.5; Figure 4, “N0: Pollen record”). However, when N0 estimates are summarised at the climate-zone level (Figure 4, “N0: Climate-zone”), there is less heterogeneity in the patterns across the climate-zones. All the climate-zones except the “Temperate” zone either reveal a gradual or no temporal variation in pollen richness (predicted values 11.5–12.7, CI range 9.3–5.0). Pollen richness in the “Temperate zone” shows an increasing trend during the early-Holocene and then it decreases gradually until the present (predicted values 13.6–16.4, CI 10.5–21.1; Figure 4, “N0: Climate-zone”). At the continental-scale, a gradually declining pattern from the early- to mid-Holocene and then a gradual increase in N0 from the mid-Holocene to the present is evident (predicted values 12.2–12.6, CI 11.4–13.5; Figure 4, “N0: Continent”).

Temporal patterns of Hill’s (1973) N1 and N2 measures also vary across spatial scales. No general temporal patterns of N1 and N2 are observed at the record level (Figure 4, “N1, N2 Pollen record”). At the climate-zone level, both PAPs in the “Temperate” climate-zone reveal a unique pattern, where both N1 and N2 reveal a gradual increase during the early-Holocene and then decrease linearly from the mid-Holocene to the present (predicted values 5.5–6.6 and 3.9–4.2; CI 4.5–7.9 and 3.3–5, respectively for N1 and N2; Figure 4, “N1, N2: Climate-zone”). However, in the climate-zone of “Cold without dry season,” both diversity estimates decrease gradually during the early-Holocene and then increase from the late early- to mid-Holocene, followed by a slight decrease during the most recent period of the late-Holocene (predicted values 5.0–5.6 and 3.4–4; CI 4.6–6 and 3.2–4.3, respectively, for N1 and N2; Figure 4, “N1, N2: Climate-zone”). There is a gradual linear increasing or no temporal pattern of N1 and N2 in other climate-zones. At the continental scale, a gradually increasing pattern in N1 and N2 from the early- to late-Holocene is evident (predicted values 5.2–5.4 and 3.7–3.8; CI 4.9–5.7 and 3.5–4, respectively, for N1 and N2; Figure 4, “N1, N2: Continent”).

No general temporal patterns of Hill’s evenness ratio estimates (N2/N1 and N1/N0) are evident at the record level (Figure 4, “Pollen record”). At the scale of climate-zones, estimates of both evenness ratios in the majority of the climate-zones either decrease gradually or remain unchanged from the early- to late-Holocene (Figure 4, “N2/N1 and N1/N0: climate-zone”). However, a gradual increase in the evenness ratios from the early- to mid-Holocene, followed by a decline in the estimates from the mid- to late-Holocene, is evident in the “Arid” zone (predicted values 0.70–0.75 and 0.43–0.47; CI 0.67–0.77 and 0.38–0.42, respectively for N2/N1 and N1/N0; Figure 4, “N2/N1 and N1/N0: Climate-zone”). At the continental scale, both the evenness estimates reveal a gradual declining trend from the early-Holocene to the present (predicted values 0.70–0.72 and 0.44–0.46; CI 0.69–0.73 and 0.42–0.48, respectively, for N2/N1 and N1/N0; Figure 4, “N2/N1 and N1/N0: Climate-zone”).

Temporal pattern of compositional turnover (CaseR scores of DCCA axis 1) is relatively more consistent across the spatial scales of individual records, climate-zones, and the whole continent, which declines linearly from the onset of the Holocene to present (Figure 4, “DCCA1”). However, turnover in the “Temperate” zone does not reveal a clear temporal pattern from the early- to mid-Holocene, whereas it declines linearly from the mid-Holocene to the present (predicted values 0.14–0.81; CI 0.11–1; Figure 4, “DCCA1: climate-zone”). At the continental scale, a linear declining pattern in DCCA1 from the early-Holocene to the present is evident (predicted values 0.18–0.97; CI 0.16–1.1; Figure 4, “DCCA1: continent”).

Rates of pollen-compositional change per 500-years vary along the temporal gradient of Holocene as well as across the spatial scales of study (Figure 4, “RoC”). No general temporal pattern in RoC is observed at the record level, whereas it decreases gradually from the early-Holocene to the present in most of the climate-zones. Notably, an increase in RoC during the early-Holocene, followed by a gradual decrease until the present time, is evident in the “Temperate zone” (predicted values 0.16–0.20; CI 0.12–0.25), whereas there is a linear increase in RoC from the early- to late-Holocene in the “Arid zone” (predicted values 0.14–0.20; CI 0.10–0.26) (Figure 4, “RoC: Climate-zone”). At the continental scale, a very gradual linearly declining pattern of RoC from the early-Holocene to the present is evident (predicted values 0.167–0.168; CI 0.15–0.18; Figure 4, “RoC: Continent”).

There is considerable debate about the fine-scale patterns and timing of Holocene climate change in Asia. However, there is a general consensus that humid and warmer conditions gradually increased in eastern and central Asia during the early-Holocene due to an increase in temperature and East Asian summer monsoon activity resulting in relatively wet conditions during the mid-Holocene, and a subsequent decline in precipitation (Chen F. et al., 2015, 2019). Trends in temperature follow the general trend of precipitation but a steep increase in temperature trend has been observed over recent centuries (Chen F. et al., 2015). Arid central Asia experienced a cold dry climate during the early-Holocene, and possibly warmer climate conditions during the mid- to late-Holocene (Wang and Feng, 2013; Chen F. et al., 2019). Pollen-based climatic reconstructions for the Tibetan Plateau indicate progressively warmer and wetter conditions in the early- to mid-Holocene and a general trend of cooling and drying toward the present (Herzschuh et al., 2006; Chen et al., 2020; Sun and Feng, 2021). The Indian subcontinent experienced a rising trend in temperature and Indian summer monsoon rainfall during the early- to mid-Holocene. However, this trend was punctuated by a global cooling event around 8,000 cal year BP. Climate started to shift from warm humid to cool arid from the mid-Holocene and reached an arid phase during the late-Holocene (Laskar and Bohra, 2021). Inferences on palaeoclimate from various proxies (such as fossil pollen, diatom, and bio-chemical data) suggest that the early-Holocene in south-west Asia was characterised by dry and warm conditions, whereas the mid-Holocene was wetter and colder, and the late-Holocene was drier and warmer (Litt et al., 2012; Cheng et al., 2015). However, some regional palynological studies infer relatively humid conditions in the eastern Mediterranean region during and Levant the early-Holocene (e.g., Cheddadi and Rossignol-Strick, 1995; Kadosh et al., 2004; Langgut et al., 2011). This may be due to an extension of the monsoonal climate system combining with the Mediterranean climate system (Rossignol-Strick, 1985; Rossignol-strick, 1999; Langgut, 2018), where as only the latter now influences the present climate. Historically, Siberian climate varied broadly, being both warmer and colder than the present. However, during the past century, it became much warmer and winter precipitation north of 55° N increased (Groisman et al., 2013). The early-Holocene in Siberia was characterised by a warmer and wetter climate resulting in a change from severe continental climatic conditions to a milder climate. However, during this period, cooling was noted in southern Siberia. The thermal maximum was most strongly expressed in Siberia during the mid-Holocene, when conditions were warmer and wetter than the twentieth century climate (Monserud et al., 1995). The late-Holocene was characterised by dry and cool conditions, especially in western Siberia (Karpenko, 2006). Most of the present-day Siberian vegetation and ecosystems started to develop after ca. 3,000 cal year BP, when climatic conditions started to become cooler and drier.

In this section we summarise the modern vegetation and the Holocene vegetation development of the major areas in Asia from where pollen-stratigraphical data are used in our analyses. Southern Asia is not discussed because of a lack of any available palynological data. A summary of the area’s vegetation and its vegetation history is given in the Supplementary Material 1.

Natural vegetation here today changes from a moist coastal-forest zone to steppe/desert, following the south-east – north-west precipitation gradient. Main vegetation types are broad-leaved, deciduous, and boreal coniferous forests. Alpine vegetation on the Tibetan Plateau consisting of montane forest, high meadow, steppe, semi-desert, and desert follows the south-east to north-west gradient of decreasing precipitation (Hou, 2001; Cao et al., 2015). During the early-Holocene, tropical and subtropical regions were colonised by thermophilous broad-leaved forest (dominant taxa Castanea, Castanopsis, Cyclobalanopsis, Fagus, Pterocarya, etc.). Forest expanded westward and achieved its maximum distribution and abundance during the early- or mid-Holocene, reflecting optimal regional conditions (Cao et al., 2015). Forest declined from the mid- to late-Holocene, probably in response to changes in regional climate and land-use (Gong et al., 2007; Fuller and Qin, 2010). Mixed broad-leaved forest (dominant taxa Juglans, Quercus, Tilia, Ulmus, etc.) was confined to the southern parts of eastern Asia and extended northward and increased in abundance during the early- to mid-Holocene. However, during the mid- to late-Holocene, mixed vegetation declined, and conifer taxa such as Abies, Picea, and Pinus increased in response to changes in regional climate, especially the weakening of the summer monsoon.

On the north-eastern border of the Tibetan Plateau, Picea together with Betula flourished during the early- to mid-Holocene, in response to increasingly humid conditions. However, Betula and Picea forests decreased regionally as conditions became drier during the mid- to late-Holocene and forest was replaced by alpine steppe vegetation (Herzschuh et al., 2010). Alpine steppe on the central and eastern Tibetan Plateau was relatively stable throughout the Holocene, with a gradual replacement of temperate steppe by high-alpine meadows during the mid- to late-Holocene. The northern and western parts were dominated by temperate desert through the Holocene (Herzschuh et al., 2010).

Vegetation here consists of Irano-Turanian and Central-Asiatic floristic elements. The predominant ecosystems are desert, semi-desert, and steppe (Zhang Y. et al., 2020). The gravel-desert flora typically includes taxa in the genera Anabasis, Atraphaxis, Caragana, Halolachne, Salsola, and Seriphidium. Arid ecosystems prevail at low elevations and in the foothills. Grassland, shrub, and forest are widespread at mid-altitudes on mountain slopes. Meadow and tundra-like ecosystems are found at high elevations. Abies and Betula forests occur mainly in the Tian Shan Mountains, whereas old-growth Juniperus forests are more common in the Pamir-Altai Mountains. Patches of riverine woodland with Elaeagnus, Halostachys, Populus euphratica, Tamarix, and Ulmus occur along rivers and lake shores. Most of the area in the south and south-west is sandy desert, changing to steppe vegetation in the east. In the north, forests form a transition zone between Siberian taiga forest and central Asian steppe. Major elements of the taiga forest are Betula, Larix sibirica, Pinus sibirica, and Pinus sylvestris (Zhang Y. et al., 2020).

In lowland regions during the early-Holocene, there was low forest cover and a high representation of drought-tolerant species compared to the mid-Holocene. Forest in general increased after the mid-Holocene in the low-elevation basins and plains due to increased humidity (Huang et al., 2018; Zhang Y. et al., 2020). High-elevation valleys in the Altai were dominated by meadow-steppe vegetation during the early-Holocene, which was gradually replaced by taiga (dominated by Abies, Juniperus, Picea, and Pinus) from the early- to mid-Holocene. Trees decreased after the mid-Holocene (Blyakharchuk et al., 2007; Huang et al., 2018), and steppe-meadow expanded since the mid-Holocene (Blyakharchuk et al., 2007; Zhang Y. et al., 2020).

Vegetation here can be broadly grouped into five phytogeographic divisions: Mediterranean, Irano-Turanian, Saharo-Arabian, Sudano-Zambesian, and Euro-Siberian (Schiebel, 2013). Mediterranean vegetation, mainly distributed around the Mediterranean Sea is dominated by oaks (e.g., Quercus calliprinos and Q. ithaburensis), as well as olive (Olea europaea) and pistachio (Pistacia lentiscus) (Schiebel, 2013). Irano-Turanian vegetation in the Asian steppes of the Syrian Desert, Iran, Anatolia in Turkey, and Gobi Desert supports Achillea santolina, Artemisia herba-alba, Thymelaea hirsuta, and several Poaceae and Amaranthaceae taxa. Sudano-Arabian vegetation is distributed in the Sahara, Sinai, and Arabian deserts and is characterised by many Amaranthaceae and Tamarix taxa. Sudano-Zambesian vegetation is mainly distributed in the subtropical savannahs of Africa and in oases along the Jordan Valley. Major taxa include Acacia, Balanites aegyptiaca, and Phoenix dactylifera (Zohary, 1982; Schiebel, 2013). Euro-Siberian vegetation grows mainly in damp habitats, along the Mediterranean coasts, and to the north of Mount Hermon (south-eastern Turkey). This vegetation consists of broad-leaved (Carpinus betulus, Cornus mas, Fagus orientalis, Ilex aquifolium, Quercus, etc.), beech (Fagus, Alnus, Ulmus, etc.), chestnut (Castanea sativa), alder (Alnus barbata), coniferous (Picea orientalis, P. sylvestris, etc.), and dry (Juniperus oxycedrus) forests (Atalay, 1986).

Palynological studies suggest that at the onset of the Holocene, vegetation was dominated by steppe and grasslands with Artemisia and members of the Amaranthaceae, Apiaceae, and Poaceae. This was gradually replaced by open oak-pistachio forest-steppe vegetation and then by forest from the early- to mid-Holocene. Forest vegetation consisted of oaks (Quercus spp.), conifers (Pinus and Abies), and broad-leaved species such as Acer, Alnus, Carpinus, Corylus, Fagus, Pistacia, Tilia, and Ulmus. This forest started to decline from the mid-Holocene and modern vegetation developed during the mid- to late-Holocene (e.g., Van Zeist and Wright, 1963; Wright, 1976; Van Zeist and Bottema, 1991).

Siberian vegetation can be summarised in latitudinal vegetation zones (Safronova and Yurkovsksya, 2019). The tundra zone occupies the northern part. High Arctic tundra consists of herb-lichen-moss associations, with Alopecurus magellanicus, Papaver polare, and Saxifraga spp., whereas the central and southern tundra is dominated by low-shrub species such as Arctous alpina, Betula nana, and Cassiope tetragona. The taiga zone extends between ca. 52° N and 72° N latitude, mainly in western and central Siberia. It is dominated by conifers such as Larix gmelinii, L. sibirica, Picea obovata, Pinus sibirica, and P. sylvestris, and with scrub and woodlands of Betula pendula, B. pubescens, Populus tremula, and Salix spp. Almost half of this zone is occupied by mires. The forest-steppe zone occurs south of ca. 56° N in western Siberia and is dominated by birch-aspen (B. pendula and P. tremula) forest. The zone also contains meadow steppes, which are fragmentary and occur only in the southern region of western Siberia.

Pollen studies show that at the onset of the Holocene north-eastern Siberia supported cold and drought-resistant tundra and steppe plant communities with Artemisia, Betula, and members of the Poaceae and Amaranthaceae. Betula and Larix forests occupied flood plains. Warmer, humid climate during the early-Holocene favoured the replacement of tundra–steppe vegetation by spruce and larch forest-tundra with a shrub understory. In north-eastern Siberia, warmer and drier conditions during the early-Holocene also resulted in the spread of Betula nana – Duschekia fruticosa shrub-tundra. In southern Siberia, the Holocene onset is marked by an increase in woody cover in response to significant climate amelioration. In the southern taiga of Siberia, forests of Abies sibirica, Picea obovata, Pinus sibirica, and P. sylvestris spread. In the northern Arctic belt, Betula exilis, Duschekia fruticosa, and Larix dahurica forests dominated the area during the early- to mid-Holocene. After the mid-Holocene, when climate became cooler and drier, afforestation of forest-steppe by cold- and drought-tolerant birch forest, freezing of wet mires in the taiga, and the formation of palsa bogs occurred. During this period, relatively thermophilous Abies forests retreated southward to the position of the present southern taiga, widespread expansion of Pinus sibirica occurred, and forest vegetation retracted from the high mountain areas of the Altai. The late-Holocene is characterised by a continuous cold and dry phase, resulting in a southward retreat of the northern boundary of the forest zone to reach its present position (Blyakharchuk et al., 2007; Blyakharchuk, 2009; Groisman et al., 2013).

Rather heterogeneous temporal variation in the observed record-scale patterns of Hill’s indices (richness, diversity, and evenness) suggests a lack of synchronicity in the patterns (see above) (Figure 4, “N0, N1, N2, N2/N1, N1/N0: record”). However, when we aggregate the estimates of records into different climate-zones and the whole continent, record-specific spatio-temporal variations are averaged or smoothed slightly, and only the minimal general temporal pattern exhibited by the majority of the records in each aggregation are observed.

Temporal patterns of Hill’s indices at the level of climate-zone (Figure 4) are likely to reflect the broad-scale vegetation successional development in response to climatic (temperature and precipitation) and human land-use drivers during the Holocene (see above). Mostly, major expansion in forest biomes at the onset of the Holocene (Cao et al., 2019a,b) driven by climate change (progressively developed warmer and humid conditions) (see the Holocene climate and vegetation development sections for eastern and central Asia) may have resulted in a reduced taxon (∼ pollen) richness (N0) from the early- to mid-Holocene. A progressive cooling and drying in most parts of the continent during the mid-Holocene resulted in a marked decline in forests with a southward retreat of the forests at higher latitudes, retreat of scrub from arid central Asia, and a decline of total vegetation in subtropical Asia during the period (Cao et al., 2019a,b; Dallmeyer et al., 2021). Monodominant coniferous forests that developed during the early- to mid-Holocene in most of the climate-zones might also have negatively altered habitat conditions (increased aridity, formation of acidic soil, etc.) for the development of understorey (shrub and herb) plant diversity. Extensive anthropogenic land-use change in, for example, eastern (Ren, 2000; Li et al., 2009; Cheng et al., 2018) and south-west (e.g., Bakker et al., 2012) Asia (see the Holocene climate and vegetation development sections for eastern, central, and south-west Asia) might also have amplified the decrease in vegetation cover as well as taxon richness (Hill’s N0) during the early- to mid-Holocene. However, the decrease in forest cover and subsequent increase in scrub and grasslands in some of the climate-zones (e.g., “Cold without dry season,” “Arid” zones) during the mid- to late-Holocene, may be a reflection of habitat diversification via changes in regional climate and anthropogenic disturbance regimes (grazing, fire, logging, agricultural activities, etc.). These resulted in a greater variety of habitats, including meadow development, thus increasing pollen diversity. This continent-scale temporal pattern of Hill’s N0 is almost similar to the temporal pattern of plant taxa richness documented by a recent sedimentary-DNA based study from the Tibetan Plateau (Liu et al., 2021), which assigns the observed pattern to changes in regional climatic conditions and anthropogenic disturbances, especially livestock grazing. Consequently, the number of equally abundant taxa in the vegetation and thus in the pollen rain may have increased slightly from the early-Holocene to the present, resulting in a gradual temporal increase in Hill’s diversity estimates in the majority of climate-zones (Figure 4, “N1 and N2: climate-zone”). Abundance, and hence evenness, of forest or other major woody taxa was possibly maximal during the early- to mid-Holocene in response to progressively warmer and humid climatic conditions (see the Holocene climate and vegetation development section). However, as the regional climate started to become cooler and/or drier, and human pressure on the landscapes increased, dominant forest taxa in many of the climate-zones decreased resulting in a gradual decrease in evenness estimates from the mid- to late-Holocene (Figure 4, N2 divided by N1 and N1 divided by N0: “climate-zone”).

Pollen richness (N0) and diversity (N1 and N2) in the “Temperate zone” (Figure 4, “N0, N1, and N2: climate-zone”) show small but an opposite temporal trends. Optimal climate conditions for forest development in the present “Temperate” climate-zone (Asian monsoon climate-zone) that developed during the early- to mid-Holocene (see the Holocene climate and vegetation development section for the eastern Asia) favoured the expansion of broad-leaved forests and resulted in a significant reduction of arid vegetation (Yu et al., 2000; Ren, 2007; Zhao et al., 2009; Cao et al., 2015; Tian et al., 2016). A gradual replacement of steppe and steppe-savanna in the current Temperate zone by broad-leaved forests might have enhanced taxa richness as well as diversity initially. However, as the dense broad-leaved forests reached their mature stage, canopy closure may have resulted in unfavourable conditions for shade-intolerant understorey vegetation. Subsequently, changes in regional climatic conditions and increasing human land-use during the mid- to late-Holocene caused a decline in forests (Gong et al., 2007; Fuller and Qin, 2010), and may have negatively impacted richness and diversity of major forest taxa in this climate-zone.

The “Cold without dry season” climate-zone is represented by boreal vegetation, the temporal dynamics of which were likely driven by climatic changes. Pollen studies show that during the onset of the Holocene much of this zone was covered by cold, drought-resistant tundra and steppe plant communities with dwarf Betula, Artemisia, and members of the Poaceae and Amaranthaceae (Groisman et al., 2013). A warmer and humid climate during the early-Holocene favoured the replacement of tundra–steppe vegetation by forest mainly dominated by Abies, Larix, Picea, Pinus, etc., which negatively impacted the diversity of the less dominant herbaceous and shrub taxa. The climate progressively became cooler and drier after the mid-Holocene which caused the southward retreat of the relatively thermophilous coniferous forests (see the Holocene climate and vegetation development section for Siberia, and central and eastern Asia). This would have favoured the development of steppe, tundra, and meadow vegetation with a higher diversity of herbaceous and shrub taxa, thus explaining the climate-zone level temporal pattern of N1 and N2 observed in Figure 4 (“N1 and N2: climate-zone”). Continental-scale temporal patterns in Hill’s richness, diversity, and evenness estimates generally reflect the trends revealed in majority of the climate-zones (Figure 4, “N0, N1, N2, N2 divided by N1, and N1 divided by N0: continent”).

The higher heterogeneity in the temporal patterns of turnover at the record level (Figure 4, “DCCA1: record”), but slightly smoothed and more pronounced DCCA1 temporal patterns at the climate-zone- and continent-scales (Figure 4, “DCCA1: record and continent”), are likely a response to major underlying drivers (see above for possible explanations). However, of the temporal patterns analysed, those of temporal compositional turnover are relatively more consistent across the spatial scales of analysis. This indicates that pollen-compositional turnover has changed substantially and consistently from the beginning of the Holocene to the present throughout the whole of Asia. Based on studies on vegetation and climate history in Asia (see the Holocene climate and vegetation development section), this pattern suggests that temporal changes in turnover followed the major temporal climatic gradient of the Holocene at the continental scale, i.e., becoming progressively cooler and drier from the early- to late-Holocene (see the Holocene climate and vegetation development section for details). Progressively increasing environmental stress from the early- to late-Holocene probably increased environmental filtering in the regional assemblages, resulting in compositionally homogenous assemblages temporally. As a result, sample-to-sample compositional dissimilarity (turnover) decreased temporally throughout the analysed records. Broadly consistent with this trend, distributional as well as compositional changes in Holocene vegetation throughout the continent in response to temporal changes in climate have also been documented by recent broad-scale studies (e.g., Cao et al., 2019a,b; Herzschuh et al., 2019; Dallmeyer et al., 2021).

Rates of pollen-compositional change vary along the temporal gradient of Holocene as well as across the spatial scales (Figure 4, “RoC”). Lack of a general temporal pattern in the RoC estimates at the record level is likely due to the heterogeneous record-specific patterns (see above for possible explanations). However, at the climate-zone scale (Figure 4, “RoC: climate-zone”), a general pattern of slightly gradually declining RoC from the early-Holocene to the present is detectable. The Holocene-wide pattern of RoC in Asia is partly consistent with the global as well as continental RoC patterns presented by Mottl et al. (2021a), despite the fact that the spatial density of the analysed records and the length of temporal gradient in our study differ from those used by Mottl et al. (2021a). However, in contrast to the increasing late-Holocene RoC pattern in Asia revealed by Mottl et al. (2021a), our study reveals a gradually linear-decreasing pattern of RoC in most of the climate-zones during the mid-Holocene to the present time. A pronounced rise in RoC in the “Temperate” climate-zone during the early- to mid-Holocene may be linked to climate change-driven vegetation changes in the region during the period (see the Holocene climate and vegetation development section for eastern Asia). Overall, a linear decline in RoC at the “climate-zone” scale is possibly driven by temporal variation in major regional climate and land-use changes during the Holocene, while an influence of shorter-term climatic fluctuations remains undetected in the RoC estimates at coarser spatial scales (see above for possible explanations). However, progressively increased aridity accompanied by divergent temporal temperature and land-use conditions in the different parts of “Arid” climate-zone in Asia, especially from the mid- to late-Holocene (see the Holocene climate and vegetation development section for south-west and central Asia), might have contributed to the temporal increase in RoC in the region (Figure 4, “RoC: climate-zone”).