Anthropogenic impacts on Late Holocene land-cover change and floristic biodiversity loss in tropical southeastern Asia

Significance

Palaeoecological analysis reveals that the expansion of rice agriculture in southern China and Southeast Asia around 2,000 y ago caused widespread deforestation and biodiversity changes in tropical and subtropical forests. Tropical forests, with the highest level of plant diversity and concentration of endemic species, suffered greater decline of arboreal richness than forests in subtropical and temperate areas. In subtropical ecosystems, total plant richness increased, despite arboreal decline, possibly thanks to the flourishing of herbs in an opening landscape. The disappearance of Glyptostrobus in southeastern Asia provides a case study as to how early rice agriculture endangered an endemic species by causing losses of its natural habitats, with prehistoric land-use changes leaving a clear legacy for today’s landscapes and species compositions.

Abstract

Southern China and Southeast Asia witnessed some of their most significant economic and social changes relevant to human land use during the Late Holocene, including the intensification and spread of rice agriculture. Despite rice growth being associated with a number of earth systems impacts, how these changes transformed tropical vegetation in this region of immense endemic biodiversity remains poorly understood. Here, we compile a pollen dataset incorporating ∼150,000 identifications and 233 pollen taxa to examine past changes in floral biodiversity, together with a compilation of records of forest decline across the region using 14 pollen records spanning lowland to mountain sites. Our results demonstrate that the rise of intensive rice agriculture from approximately 2,000 y ago led not only to extensive deforestation but also to remarkable changes of vegetation composition and a reduction in arboreal diversity. Focusing specifically on the Tertiary relic tree species, the freshwater wetland conifer Glyptostrobus (Glyptostrobus pensilis), we demonstrate how key species that had survived changing environmental conditions across millions of years shrank in the face of paddy rice farming and human disturbance.

The tropical forests of southern China and Southeast Asia (which we term together here as southeastern Asia) are renowned for their rich and unique endemic flora that, in many cases, can trace its origins back to the end of the Mesozoic approximately 65 million y ago (1). The relatively stable geological conditions of the region since the Late Cretaceous period have meant that it has frequently been seen as a long-term refugium for relict plant species, and is home to one of the highest concentrations of endemism around the globe (2, 3) and at least 4 of the 25 globally important biodiversity hotspots (4). Today, however, the tropical forests of southeastern Asia are among the most threatened land-based ecosystems on the planet, witnessing some of the fastest rates of deforestation in the face of urban population growth, changes in land-use, and human-induced climate change since the 20th century (2, 5). Agricultural expansion is seen as one of the major drivers of forest loss in southeastern Asia, with the conversion of tropical forest ecosystems to plantations or subsistence fields leading to drastic reductions in biodiversity (6–9).

While the anthropogenically induced changes of these supposedly “ancient” ecosystems are often considered recent, there is a growing consideration that humans have interacted with tropical vegetation, including in southeastern Asia, across millennia, leaving legacies of biodiversity change, habitat manipulation, and forest cover that have lasting implications for economies, environments, and conservation policies today (10, 11). The arrival of rice cultivation in southern China and Southeast Asia took place around 4,000 y ago (ka), and the expansion of large-scale intensive rice agriculture in this area can be traced back to as early as 2.5 to 2 ka (12, 13), as the formation of coastal plains under freshwater conditions created expansive areas suitable for rice cultivation, together with a massive influx of rice farming immigrants and advances in agricultural technology, with potentially early impacts on tropical forest cover in the region. There have also been suggestions that the expansion of rice paddy fields, alongside domesticated water buffalo, in southeastern Asia may have resulted in a rise in methane emissions that fundamentally altered regional, and perhaps global, climate (14). Nevertheless, to test the degree to which these early agricultural expansions impacted tropical forest cover and biodiversity, and their relevance to ecosystem management in the 21st century, we require direct evidence for paleoecological change at the local and regional scale.

Fossil pollen data from sediment cores can provide continuous records of past vegetation change and are widely used as a reliable proxy for forest-cover change (15, 16). Moreover, pollen data contain information on plant diversity and can thus also be used as a measure for past taxonomic richness. Although pollen records can struggle to capture the real plant diversity in tropical areas, where entomophilous taxa produce very few pollen grains and are thus usually underestimated in pollen assemblages, it is still possible to get meaningful results from pollen data. Supporting evidence comes in the form of the distinct latitudinal trend of palynological richness seen among woody taxa (17), indicating that the high diversity of vegetation characteristic of tropical regions can be revealed by pollen records. In addition, modern pollen studies undertaken in eastern Asia, including within both subtropical and tropical areas, suggest that modern pollen “rain” can successfully predict the natural biome, revealing a good relationship between palynological data and plant composition (18). The latest pollen productivities study in Hainan Island (19), which presents by far the most detailed pollen–vegetation survey from the northern part of tropical southeastern Asia, indicates that, at a regional scale—although there were differences between the modern pollen assemblages and related vegetation composition—the pollen types identified in surface soil samples captured the majority of investigated plant types (∼65%) at the family level. Moreover, this modern survey suggests that pollen assemblages of cultivated lands can be distinguished from recovered secondary woodlands, indicating that information of agriculture-induced vegetation change can be captured in pollen records in this part of the tropics (19).

Here, we synthetically analyze pollen records in coastal areas of southeastern Asia that host northern tropical and southern subtropical moist broadleaf forests today, examining land-cover and floristic biodiversity change during the past 6,000 y. Our study time periods, spanning the last 6,000 y, are designed so as to focus on the known archaeological periods when the Neolithic subsistence economy mainly relied on hunting and gathering (before 3 ka) and when large-scale intensive rice agriculture began to emerge and expand (around 2.5 ka). First, 14 pollen records spanning lowland to mountain sites were systematically interrogated in order to reveal a comprehensive deforestation history for the study region. Then, we compiled a pollen dataset incorporating ∼150,000 identifications and 233 harmonized pollen morphological types from 9 pollen records to examine past changes in floral biodiversity. Furthermore, we examined how early paddy rice farming interacted with the now-endangered swamp conifer, Glyptostrobus pensilis, that had stood tall in the tropical forests of the region since the Tertiary period. Our data provide an example as to how regional paleoecological analysis can provide broader insights into the degree to which human economies and societies altered past landscapes and biota in the tropics, with potential ramifications for understanding the remnant vegetation and habitat distributions of the 21st century and the ongoing threats facing tropical forests in the “Anthropocene” (20).

Results

Landover Change from Pollen Records.

A total of 14 pollen records, extending from lowland to mountain sites, were examined for changes in forest cover during the past 6,000 y in this study (Fig. 1). The pollen spectra reveal heavily forested landscapes before 4 ka, characterized by a high proportion of arboreal pollen (Fig. 1 E and F), indicating that the overall anthropogenic impact on vegetation cover was negligible. During this period, the regional Neolithic subsistence economy mainly relied on fishing and hunting (21, 22). After 4 ka, fragmentation of forests began to take place in some lowland areas (Fig. 1 C and D). The low-lying freshwater flatlands may have encouraged the earlier local development of Neolithic rice farming (12), which is consistent with archaeological evidence suggesting that rice appears as part of the broad-spectrum Neolithic subsistence economy that remained reliant on wild resources (e.g., refs. 21 and 22). Widespread fragmentation of forest was recorded in our study area from 2 ka (Fig. 1B). This timing coincides with the establishment of regionally extensive paddy rice agriculture, following the formation of expansive coastal plains that are suitable for paddy rice agriculture (12) and a massive influx of rice farming communities (23).

Fig. 1.

(A–F) Change in southeastern Asian forest cover during the past 6,000 y revealed by pollen data. See SI Appendix for pollen site information and natural vegetation in the study region.

Around 2 ka, most of our pollen records from river deltas and coastal plains reveal that the arboreal pollen percentages had decreased to less than 50% of the total terrestrial pollen. Mountainous areas also witnessed a decline in forest cover by this time, though not as serious as at lower altitudes (Fig. 1B). These pollen spectra generally reveal semiopen forest landscapes. This feature of the pollen assemblages in our study region may indicate either strong human impact on vegetation (i.e., deforestation) or a substantially drier climate. Although stalagmite δ¹⁸O records from Chinese caves suggest a weakening of the Asian summer monsoon during the Late Holocene, an increasing number of studies have discussed the complexity of north–south dipolar rainfall patterns in Asian summer monsoon regions (e.g., refs. 24 and 25), with a weakening Asian summer monsoon causing drought in northern China and flooding in southern China. In our study region, the lack of unequivocal proxy records for local rainfall inhibits us from fully discussing the potential impacts of climate change. Nevertheless, according to the results of the TraCE-21 ka transient simulation that simulated continuous climate evolution of the last 21 ka (26), precipitation remained relatively stable in our study site during 6 to 1 ka (SI Appendix, Fig. S5), suggesting that climate change had little impact on the forest fragmentation seen during this period.

However, historical records document major southward migrations of the Han Chinese into southern China and northern Vietnam during the period of Warring States (475 to 221 BCE) and subsequent Qin (221 to 207 BCE) and Western Han (206 BCE to 8 CE) periods. This large-scale incoming wave of farming immigrants is also revealed by genetic research (27). Additionally, agricultural technological advances, including ox-drawn plows and iron farm implements, spurred the intensification of rice agriculture (23). Taken together, these support the potential of human impacts to have severe consequences for regional forests by 2 ka.

The pollen data demonstrate increasingly severe impacts on forests from human activities thereafter. By 1 ka, most of the sites in our study region, including mountainous areas, had experienced a remarkable decrease in forest cover (Fig. 1A). This environmental and ecosystem transformation coincides with the timing of an additional wave of immigrations of war refugees from the north during the Chinese Tang (618 to 907 CE) and Song (960 to 1279 CE) Dynasties. Sediment cores that have analyzed charcoal particles deposited together with pollen suggest an abrupt increase of charcoal concentrations alongside declines in arboreal pollen percentages, suggesting that frequent fire activity occurred synchronously with forest-cover change (28). Together, these data likely support an intensification of slash-and-burn deforestation, which happened earlier at some lowland sites around 3.5 ka BP, and gradually extended into mountainous regions during the next two millennia (28).

Biodiversity Change.

In this study, 9 pollen records, including ∼150,000 identifications, were used to test floristic biodiversity change caused by human-induced land-cover change. We compared pollen richness for different time periods based on the rarefaction analysis. This method enables estimation of pollen richness in an exactly equal-sum count for samples (29), thus removing the potential bias of comparing different sample sizes (i.e., that the number of pollen types increase as the sample size increases). The results of rarefied richness are normalized to 0-to-1 values as we aimed to reveal the trends, rather than the absolute values, of pollen richness. Moreover, the rarefaction curves for the time windows are also presented in our study in order to show the trends of pollen richness as the sample size increased. As well as the site-based analyses, samples were pooled for two large regions, Tropical and Subtropical, which enabled us to capture the signals from taxa that have low abundances and pollen productivities. A regional rarefaction curve for the different time periods was calculated based on a sample size of more than 12,000 identifications for each time window.

We tested the results of plant-richness change in two aspects. First, pollen identification usually reaches the taxonomic levels of plant genus and family, but rarely species. Pollen-richness analysis provides a way of using higher-taxon surrogacy, a method that was shown to have good performance in tropical areas (30, 31). In this study, we present site-based rarefaction curves in two taxonomic hierarchical levels (H0 and H1). For each site, in both the tropical and subtropical regions, the results of richness change in different hierarchies show similar patterns and trends (SI Appendix, Figs. S6–S9), suggesting that plant-richness patterns are little affected by pollen taxonomic biases. Second, pooling regional samples can amplify the signal from taxa that have lower abundances and pollen productivities (32), although the attainment of pollen data from different sedimentary archives could impact richness patterns. In order to evaluate the potential bias introduced by compiling pollen-richness data from different types of records, we conducted the rarefaction curves using combined pollen data from peat/lake cores alone, and delta boreholes alone, respectively. The results revealed the same patterns of plant-richness change among different sedimentary archives (SI Appendix, Fig. S10), suggesting that our results are little influenced by the nature of sedimentary archives.

Both regional and site-based analyses in this study reveal a clear decrease of pollen richness in northern tropical areas after 2 ka (Fig. 2 and SI Appendix, Figs. S6, S7, and S12). This decline in richness corresponds to the reduction of arboreal pollen content interpreted as a product of severe deforestation. According to the pollen assemblages of the Tropical zone, the dominant taxa of evergreen broadleaved forests, such as Castanopsis-Lithocarpus, Cyclobalanopsis, Flacourtiaceae, Mallotus, and so forth, decrease sharply (SI Appendix, Fig. S12). However, the decrease of the dominant trees probably contributes little to overall changes in pollen richness, as these tree taxa are always present in the pollen assemblages, although, during certain periods, in relatively low frequency. On the other hand, other arboreal pollen taxa that are frequently found in the natural tropical forests of East Asia, including Randia, Aphanamixis, Sterculia, Symplocos, Antidesma, Theaceae, and so forth, become quite rare in the pollen spectra since 2 ka. Pollen assemblages of modern surface soil samples suggest that these important taxa, characteristic of tropical forests, are rare or even absent in both cultivated lands and recovered secondary woodlands (19). Our results indicate that this character of floral biodiversity decline probably traces back to 2 ka.

Fig. 2.

Tropical and Subtropical regional plant biodiversity change revealed by pollen-richness analyses. (A) Comparison of rarefaction curves for different time periods and (B) time series change of regionally combined samples on DCA axis 1 scores, pollen richness, the count of arboreal pollen and nonarboreal. Regional sample sizes are also presented using histograms at the bottom. See Materials and Methods for additional information.

Mangrove plants recorded in coastal sites also declined sharply after 2 ka, which may have been caused by both intensified human activities and changes in coastal sedimentary environments from brackish to freshwater conditions (33, 34). However, we suggest that the contribution of mangrove decline to regional richness change is small, as only a few mangrove types are recorded in sites VN and GZ-2. Rarefaction curves of sites VN and GZ-2 that exclude the taxa of mangrove plants reveal little difference to their counterparts that include them (SI Appendix, Fig. S11). Moreover, the fact that the same patterns of richness change seen in other tropical sites with a constant sediment environment (sites GY1, SH-1, and SC-1) are witnessed in sites VN and GZ-2 indicate that the biodiversity decline observed in these records do not mainly result from deltaic depositional environment change. Indeed, increasingly frequent burning in delta regions after 3 ka, suggested by charcoal records (12), indicates severe human impacts on vegetation change in these areas.

We also used detrended correspondence analysis (DCA), alongside rarefaction analyses, in order to evaluate changes in floristic composition. The differences in the scores of the first DCA axis are interpreted as reflecting a change in pollen composition and thus in vegetation composition. The DCA results indicate a major floral composition change at 2.5 to 2 ka in both Tropical and Subtropical regions (Fig. 2B), corresponding to the start of widespread deforestation in our study areas. Site-based DCA results, as well as cluster analyses, also indicate obvious vegetation composition change when arboreal pollen decreased (SI Appendix, Figs. S12 and S13), suggesting the severe impacts of human-induced deforestation on floristic diversity.

In the Subtropical region, however, the clear decline of pollen richness after 2 ka observed in the tropical zone is not so visible (Fig. 2 and SI Appendix, Figs. S8, S9, and S13). Instead, the regional pollen richness witnessed a slight increase, although there still seems to have been a loss of arboreal pollen taxa caused by anthropogenic deforestation (Fig. 2B and SI Appendix, Fig. S13). We suggest that the opening of the landscape, as well as the regrowth of secondary shrub forest following human disturbance, provided space for light-demanding plants and the flourishing of herbaceous flora, thus increasing pollen richness in these subtropical settings. Similar phenomena are also commonly recorded in Europe for the Late Holocene (32). However, in tropical areas, this phenomenon seems to be masked by the decline of arboreal pollen richness. This highlights the need to look at habitat-specific subtropical and tropical forest responses to different human activities, with ecological resilience and dynamics meaning that human behaviors translate differently into land-cover and biodiversity changes in different parts of the equatorial realm.

Disappearance of Glyptostrobus in Southeastern Asia.

Glyptostrobus (a endemic wetland conifer) fossil pollen was identified in the sediment cores through both light microscopy and scanning electron microscopy (SI Appendix, SI Text), allowing us to specifically study the changes in the abundance of this species in southeastern Asia during major periods of land-use change documented above. Glyptostrobus (G. pensilis), commonly known as the Chinese water-pine, is a typical Tertiary relict species and a so-called “living fossil.” It is the sole living species of the genus Glyptostrobus, which was formerly widespread in the Northern Hemisphere from the Cretaceous to the early Pleistocene (35), with its range being periodically reduced to certain refugia in southeastern Asia during Quaternary glaciations. Today, this species is on the verge of full extinction in the wild, with only a few fragmented native stands found in South China (Fig. 3A), Vietnam, and Laos, and it is categorized as a “critically endangered” species by International Union for Conservation of Nature (36). What caused the disappearance of Glyptostrobus in southeastern Asia, and when the greatest population reduction of this species took place, remain long-standing questions.

Fig. 3.

(A) Extant stands of Glyptostrobus in the Youxi Glyptostrobus Protected Area, Fujian Province, China (photo was taken by T.M.). (B) Large accumulation of buried Glyptostrobus residues at the head of the Pearl River Delta (photos were taken by Z.Z.). (C) Distribution of Glyptostrobus/Taxodiaceae forests in South China during the Middle to Late Holocene revealed by pollen records.

Fossil evidence from both plant residues and the pollen records studied here indicate that Glyptostrobus was still widely distributed in South China during the Middle to Late Holocene. The buried wetland forests dominated by Glyptostrobus were found in Holocene sediments at the lower reaches of the Pearl River at the head of Pearl River Delta (37, 38). Furthermore, high abundances of Glyptostrobus/Taxodiaceae fossil pollen are documented at peat bogs in Daiyun, Jiufeng, Donggong, and the Xianxia mountains, indicating the existence of coniferous forests of this species around the wetlands between approximately 6 and 1 ka (Figs. 3C and 4). Glyptostrobus is a hydrophilous plant, which commonly grows on flooded or waterlogged soils. The downstream floodlands of the Pearl River, as well as the mountain swamps, would thus have provided important habitats for Glyptostrobus in the Holocene. However, these natural wetlands were also the ideal places for early paddy field rice farming.

Fig. 4.

Pollen records reveal the disappearance of Glyptostrobus forests in the Pearl River Delta, and in the Daiyun, Jiufeng, Donggong, and Xianxia mountains.

The pollen data from the Pearl River flood plain and subtropical mountainous wetlands record a period of high levels of Glyptostrobus/Taxodiaceae pollen, indicating that these areas used to be covered by dense water-pine forests between 6 and 2 ka (Fig. 4). All of the pollen records show a sharp decrease of Glyptostrobus/Taxodiaceae toward the Late Holocene (approximately 2 to 1 ka). Climatic amelioration after the deglaciation may have facilitated the expansion of Glyptostrobus during the Holocene. However, the major decline of this taxon witnessed during the Late Holocene is unlikely to be a result of climate change, as both temperature and precipitation in our study regions witnessed little change during the Late Holocene based on available climate records and models (39) (SI Appendix, Fig. S5). According to the extant distribution of this taxon, Glyptostrobus trees now grow in areas with a mean annual temperature range from 13 to 25 °C, and a mean annual precipitation range from 1,500 to 2,000 mm (SI Appendix, Table S2). The pollen sites that revealed a decline in Glyptostrobus during the Late Holocene have climatic conditions within this climate range; for example, in the Pearl River Delta, the mean annual temperature is 22 °C and mean annual precipitation is about 1,700 mm. Nevertheless, there are no longer any native stands of Glyptostrobus. Placed against the backdrop of millennia of retraction and recovery of Glyptostrobus populations in the face of climate change, the present-day challenges facing this species are thus most likely linked to the intensified anthropogenic pressures witnessed over the course of the Holocene.

The decrease of Glyptostrobus/Taxodiaceae recorded in the pollen spectra is associated with rises in Poaceae which, in our study area, reflect the expansion of grasses and suggest anthropogenic impacts on natural forest and the presence of cultivated rice fields where a great number of weeds grew inside paddy fields due to low-level planting techniques (40). The frequent counting of rice-type Poaceae (although not all the pollen assemblages separate rice-type Poaceae from other Poaceae) (Fig. 4) and the increase of certain taxa (including Dicranopteris, Artemisia, and so forth) (SI Appendix, Figs. S12 and S13), that are revealed to be closely associated with rice agriculture by modern pollen rain studies in double-cropping rice agricultural systems (40), confirm rice cultivation in the region. Charcoal analysis, conducted together with pollen analyses at some sites, suggest that this process was accompanied or immediately preceded by regional fire activities (Fig. 4), indicating the development of slash-and-burn agriculture.

In the Pearl River Delta, the large accumulation of exhumed Glyptostrobus fossil wood date mostly to 5 to 2 ka, a time when there was a prevalence of Neolithic communities focusing on fishing and hunting groups in coastal South China (41, 42). Over the last three millennia, the freshwater wetland habitats of Glyptostrobus forest expanded seawards, from the apex area of the Pearl River Delta to the deltaic basin, as the deltaic shoreline advanced (43). However, this process coincided with significant human population increase and development of intensive paddy field farming from around 2.5 to 2 ka (12). Most communities of Glyptostrobus disappeared from the Pearl River Delta after 2 ka. Pollen records indicate that the destruction of natural Glyptostrobus wetlands in the mountains occurred later (at around 1 ka) (Fig. 4), likely due to the fact that the expansion of agriculture-induced environmental impacts from lowland into mountainous regions occurred later. Today, in southern China, a few fragmented stands of Glyptostrobus found in high mountains areas, where small swamps, which have not been exploited by human agriculture, provide the last habitats for Glyptostrobus in China.

Discussion and Conclusions

Our data indicate that widespread human-induced deforestation and obvious floristic biodiversity change in our study region, including the northern tropical and southern subtropical forests of coastal southeastern Asia, can trace their origins back to as early as 2 ka and the extensive development of paddy rice agriculture. The forest transformations that occurred at this time involved a simplifying of vegetation structure, loss of arboreal diversity, and an opening of forests that were colonized by grassland and shrubs. Pollen compositional changes provide evidence for vegetation turnover in tropical and subtropical areas related to the development of agriculture. According to our results, deforestation and the development of agriculture resulted in different patterns of pollen-richness change being seen in tropical areas, subtropical areas, and temperate areas, something that has been documented elsewhere (32) and likely relates to the different responses of particular forest types to human intervention. As tropical forests have the highest level of plant diversity and concentration of endemic species, human-induced deforestation is likely to cause a greater decline of arboreal plant richness. Meanwhile, in the temperate regions of the middle and high latitudes, the biodiversity patterns following deforestation are different: closed climax temperate forests are generally characterized by low taxonomic richness due to the prevalence of a few dominant trees, while the subsequent open landscape provide habitats for the light-demanding plants and herbaceous flora that can even have a higher plant richness (32). In the subtropical zone, the present study shows a pattern of pollen richness after the expansion of agriculture, which combines responses of both a decrease in tree taxa and increase in herbs.

It should be pointed out that information relating to past plant richness revealed by pollen data are biased and incomplete. This is mainly a product of the fact that different plant taxa have different pollen productivities (44, 45). In particular, entomophilous taxa are important elements in tropical forests. However, these taxa are rarely captured by pollen deposition in the sediment records as a result of their low pollen productivities (46). Although the inherent biases of pollen analyses cannot be eliminated, pollen records, by far, still represent one of our best data sources for past land-cover change and plant diversity. In this study, the pollen records provide a glimpse of the severe deforestation, floral compositional change, and arboreal richness decline in northern tropical forests that occurred at around 2 ka, as paddy field rice agriculture started to expand across southeastern Asia.

The disappearance of Glyptostrobus in southeastern Asia provides a particularly intriguing example as to how early rice agriculture could impact certain endemic species. This species had survived in changing environmental conditions for millions of years since the Cretaceous (35). Yet, Glyptostrobus started to disappear from records in the lowlands, and later the highlands, between 2 and 1 ka, in the face of a conversion of its usual swamp habitats into paddy fields. Prior to this point, Neolithic groups focused on fishing and foraging had lived in these wetland areas, seemingly leaving a more limited footprint on the surrounding forested landscapes and species, such as Glyptostrobus. Today, this species is one of the most critically endangered tropical trees species in southeastern Asia as land-use change continues apace. Nevertheless, it is clear that the threats facing this taxon may stem back to as early as the first rise of extensive wetland rice agriculture in the region, highlighting the role of agricultural conversion as a major conservation threat to certain taxa, past and present.

Our study feeds into a growing body of work that demonstrates that southeastern Asia is a crucial region for determining the degree to which past human land-use and economy changes (e.g., the arrival of rice agriculture, the spread of domesticated animals such as pigs and water buffalo and, later, the extension of metallurgy) had impacts on tropical forest vegetation cover (14, 47). Today, human impacts on tropical forests are seen as a key part of the Anthropocene (10, 20), with local and regional changes in forest cover impacting soils, carbon storage, temperature, and rainfall through earth systems feedback mechanisms. The development of regional-scale paleoecological studies offers the opportunity to study the changing tempo of human influences on tropical forests, and how these impacts may have left economic, landscape, and earth systems legacies for human populations and conservation priorities in the equatorial regions today. We have highlighted how it may even be used to highlight the longevity of threats to specific species in endemic-rich areas. The development of further forest-specific, high-resolution, and multiproxy records promises to further help us use the past to address the challenges of the present (48) in some of the most rapidly disappearing environments anywhere on Earth.

Materials and Methods

Pollen Data.

Fourteen pollen records were reviewed for land-cover change from a collection of both raw and digitized pollen spectra. These records are located in the coastal areas of tropical and south-subtropical southeastern Asia ranging from 19 to 28°N latitude, and covered both high and low altitudes, representing deposits under various geographical conditions including delta and mountainous areas (see SI Appendix for detailed pollen site information). We divided these sites into two regions, Tropical and Subtropical, in order to differentiate between broad regions with different natural floristic composition. Sites south of 23.5°N were assigned to tropical, including northern part of tropical moist broadleaf forests and the ecotones of tropical–subtropical moist broadleaf forest. Sites at latitudes higher than 23.5°N were assigned to Subtropical, where the natural vegetation is dominated by subtropical moist broadleaf forest, albeit with tropical elements also appearing frequently.

Age-Depth Models.

Chronologies for the pollen records were established by radiocarbon dating (SI Appendix, Table S3). ¹⁴C ages were calibrated or recalibrated to calendar years before the present (cal yr BP, where 0 y BP = 1950 CE) using the mixed IntCal20 and SHCal20 calibration curve or Marine20 calibration curve (49, 50). For each pollen sequence, we constructed an age-depth model (see SI Appendix for details) using the Bchron R package (51).

Pollen Taxonomic Standardization.

We use two levels of taxonomic hierarchy for analyzing the pollen data. In the first level (H0), we harmonized names for taxonomic synonyms, and combined morphologically similar types. This pollen taxonomical level (H0) often corresponds to plant genus or closely related genera, and represents the generally accepted detailed level of pollen identification according to the literature and expert knowledge (52). However, for the original pollen data of some pollen records, taxa that are separated in H0 but present with low frequencies may be grouped together with similar taxa from the same family or related genera, and these pollen records represents lower taxonomic detail at the level of H0. This problem cannot be eliminated in H0, which results from the different traditions of pollen identification of different investigators. Therefore, we also constructed a higher hierarchical level (H1). In this hierarchy level (H1) pollen taxa represent morphologically distinct and identifiable features and often correspond to plant genus or family. Pollen data in H1 represent the same level of taxonomic precision among the sites, and can be used for regional analyses. The taxonomic harmonization of our pollen data yielded 233 pollen types at level H0 and 158 types at H1. See details of H0 and H1 in Dataset 1.

Plant Biodiversity.

Nine pollen records that are available with original counting data were used for the pollen-richness analysis. Among them, sites VN, SC-1, GY1, GZ-2, and SH-1 belong to the Tropical region, and sites FZ4, FZ5, LTY and THD belong to the Subtropical region. All terrestrial pollen types identified in these pollen records were first matched with the taxonomic hierarchies of H0 and H1, and then used for richness analysis. Since H0 represents more detailed taxonomic precision, it is used for site-based analyses, while H1 is used for regional pollen-richness analyses as it enables the same level of taxonomic precision to be used across the sites. An exception is that we present site-based rarefaction curves in both H1 and H0 (SI Appendix, Figs. S6–S9) in order to reveal whether the different hierarchical levels impacted the changes observed in plant richness.

Comparisons of pollen richness between different samples requires an equal-sum count for all samples (29). In this study, we use rarefaction analysis to standardize for sample size. This method estimates how many taxa would have been found if all the pollen counts within a sequence had been of the same size. For each sequence, the actual minimum count among the sequence is used as the base value. The rarefaction analyses were carried out in R using the vegan package (53). The results of rarefied richness are normalized to 0-to-1 values.

We synthesized the time-series changes of regional richness. For each pollen time series (site), we first excluded the samples that contained fewer than 100 pollen counts, and then the pollen data were interpolated at 400-y intervals using the program AnalySeries 2.0.4 (54) based on the Bayesian age-depth model. For the newly constructed “regional” samples, we then excluded those that contained fewer than 800 pollen counts. This resulted in the exclusion of the tropical samples before 5,000 y, as these samples contained fewer than 800 pollen counts. Rarefied richness at 400-y resolution was carried out for the last 5,000 y in the Tropical region and for the last 6,000 y in the Subtropical region (Fig. 2B).

The rarefaction curves for time-window changes (with each time window representing a 2,000-y age span, e.g., 2 to 4 ka) are also presented in our study to show the trends in richness as the sample size increased. For each sequence, the pollen counts within each time window were summed into one single pollen sample. Regional samples were obtained by summing pollen identifications from samples for a given region and time. Pollen samples that were excluded from the analyses of time-series change are included in the time window pool. For the regional analyses, each time window contains more than 12,000 identifications.

Regional and site-based DCA were conducted using Canoco 5, in order to reveal how palynological compositional changed over time. For each pollen time series (site), stratigraphically constrained cluster analysis was also conducted using CONISS in the TILIA program (55).

Pollen Taxa Percentage.

Percentages of pollen taxa and ferns (Dicranopteris) were calculated based on the sum of terrestrial pollen (including arboreal pollen and nonarboreal pollen). In Fig. 1, percentage changes of arboreal pollen and nonarboreal pollen for each sequence are presented as the average values for every 1,000 y.

Data Availability

All of the previously published data derive from the sources cited. All data generated during this study are included in the main text and Supporting Information.

Change History

October 6, 2021: The author line has been updated.