Transient social–ecological dynamics reveal signals of decoupling in a highly disturbed Anthropocene landscape

Significance In the Anthropocene, critical social–ecological transitions hold profound significance for both environmental stability and human well-being. Our study advances a dynamic-based approach to assess the adaptability or transformative potential of social–ecological systems (SES) through time, by integrating a social–ecological coupling index based on rate of change. Drawing on empirical evidence from sedimentary ancient DNA, biophysical, and socioeconomic records, we explore the evolutionary dynamics of interconnected SES in China’s Lake Taihu watershed over the last century. Notably, we present compelling evidence of unprecedented decoupling signals between socioeconomic growth and ecoenvironmental degradation, particularly in the last two decades. These decoupling trends, crucial for sustainability assessments, may remain cryptic if relying solely on the magnitude of change derived from short-term change metrics.

Nutrient and trace metal.Total organic carbon (TOC) and total nitrogen (TN) content in the samples, ground and pre-treated with 1 mol L -1 HCl at room temperature to remove carbonate, was measured using a Euro 3000 elemental analyzer.The results are expressed in % of dry weight sediment.Standard reference materials GSS-9, GSS-13 and GSS-16 (lacustrine sediment and soil from National Institute of Metrology of China) were used as a quality control, and the measured values were all within the ranges of certified values.After sequential digestion of the freeze-dried, ground samples (4), concentrations of sediment total phosphorus (TP) were determined by inductively coupled plasma-atomic emission spectrometry (ICP-AES; Leeman Labs, Profile DV), and concentrations of trace metals (such as Cd, Pb, Cd) were determined by inductively coupled plasma mass spectrometry (ICP-MS; Agilent 7700x).These results are expressed in mg kg -1 dry weight sediment.The data quality was ensured by measuring blanks and standard reference materials (GBW07358; Chinese geological reference materials) after every tenth sample.
Contamination assessment.Pollution load index (PLI) is widely applied to assess total level of trace metal contamination (4).Mathematically, PLI is calculated as PLI ＝ (CF1×CF2×CF3×•••×CFn) 1/n ; CF＝[M]sample／[M]background.Where CF is the ratio obtained by dividing the concentration of each metal in the sediment by the background value.The bottom sediments (n=2) of cores from Taihu Lake, which were deposited in pre-industrial periods (i.e.before the 1920s), could be considered as reference background (3,4).Pollution load index (PLI) of Cd, Pb, and Sb was thus determined to reflect anthropogenic pollution particularly from atmospheric deposition (3,4).When PLI>1, it means that trace metal contamination exists.Subfossil pollen.Plant pollen analysis was conducted on ~2 g of dried sub-samples following ref.(5).One tablet of Lycopodium spores (27,637 ± 563 grains/tablet) was added to each sample as a tracer for calculating the pollen concentration.All samples were treated with 10% HCl, 10% NaOH, and 2.1 g cm −3 heavy liquid flotation (the conventional HF method was not used due to the lack of siliceous materials in the sediment) before being sieved through a 7 μm mesh and acetolyzed.Finally, the samples were preserved in glycerine.All pollen and spores were identified under ×400 magnification using an Olympus BX-53 microscope to a sum of over 500 grains.Aquatic taxa were removed, and the remaining terrestrial taxa were expressed as percentages of the sum of the terrestrial pollen assemblages.The pollen concentrations of major vegetation taxa were calculated as grains g −1 .
Pigment biomakers.Sedimentary chlorophyll and carotenoid pigments were analyzed using an Agilent 1200 series HPLC (Agilent Technologies, Mississauga, ON, Canada) with quaternary pump, autosampler, ODS Hypersil column (250×4.6 mm; 5 µm particle size), and photo-diode array detector as described in ref. (6) with the extraction and separation conditions modified from ref. (7).Pigments were calibrated using authentic standards (DHI, Denmark) and are expressed in nmol pigment g -1 TOC to compensate for diagenetic degradation (8).The ultraviolet radiation (UVR)-absorbing pigment (scytonemin derivative) identified in ref. (9) was divided by the sum of key carotenoids (diatoxanthin, lutein-zeaxanthin and alloxanthin) and multiplied by 100 to derive a UVR index.Calibration in whole-lake experiments revealed that this index increases as a linear function of the depth of UVR penetration, such that higher index values indicate greater exposure to potentially damaging UVR (9), namely better light conditions and indicate higher water clarity (10).

Text S3. Sedimentology and geochemical characteristics
Major physicochemical characteristics of the core sediments were analyzed, and the results are presented in Fig. 2 and SI Appendix, Figs.S3 and S7.Color of the sediments varied from gray to dark-brown from the bottom to the upper core, and then appeared grayish yellow in the top centimeters.Water content of the sediments was 44.4±4.2%.The sediments were mainly composed of silt (4-63 µm) and clay (<4 µm), accounting for 81.6±1.7% and 17.4±2.0%,respectively.The median grain size was 11.7±1.4µm, and showed an overall decreasing trend in contrast to increasing fine silt and clay contents from the bottom to the top of the core.The value of mass magnetic susceptibility (MS) ranged between 16×10 -8 -27×10 -8 m 3 kg -1 , showing a sharp increase in the 1960s and maintaining high values afterward.Total organic carbon content in the sediments was 0.87±0.21%with an increasing trend, whereas C/N ratios showed a decreasing trend that is generally less than 10.Sediment TP increased significantly from the middle of the core (about the 1980s), and showed a decrease in the top centimeters.Based on the XRF scanning results, zirconium and zirconium-to-rubidium (Zr/Rb) of the core sediments showed an sustained increase during 1900-1950 CE and abnormally peaks during 1950-1980 CE, followed by a significant decrease.

Text S4. Plant sedaDNA records
Fluorescent quantification results showed that total DNA concentration in the 23 slices of sediments fluctuated largely between 1.09-98.11ng g -1 with an average of 23.88±29.91ng g -1 (Fig. 3).After high-throughput sequencing and standard filtering procedure, we obtained 1,920,292 usable merged reads, corresponding to 98 terrestrial sequences (including herb, shrub and tree taxa).Comparing them with the "Flora of China" (http://iplant.cn/foc)as listed in Jiangsu and Zhejiang provinces around the lake region, 57 well-identified taxa among 23 samples (from 3 to 55 taxa per sample) were retained and used to investigate watershed vegetation dynamics over the past hundred years (SI Appendix, Fig. S8).Temporally, total sedaDNA concentration changed similarly with the plant DNA content and richness as indicated by the numbers of taxa and DNA reads (Fig. 3).Four phases of changes were recorded as follows: low level before the 1950s; slight increase during the 1950s-1980s; sharp increase until high level during the 1980s-2000s; and gradual decline after the 2000s.Nonmetric multidimensional scaling (NMDS) of plant DNA chronosequences (SI Appendix, Fig. S9), using the number of reads in four PCR replicates, showed higher replicate reproducibility during the 1980s-2000s when DNA quantity and number of taxa were high.This analysis of similarities confirms the significant difference of plant communities observed during the 1980s-2000s, and also proves the reliability of the sedaDNA records (11).
Comparing the plant DNA results with modern vegetation compositions in the watershed (12), we found that the plant diversity recovered through sedaDNA in the core clearly provides a relatively local signal (Fig. 3 and SI Appendix, Fig. S8).Percent results of plant DNA reads showed that trees accounted for 18.0±7.7%with an obvious decrease during the 1950s-1960s in contrast to herbs, and crops accounted for 34.7±10.9%with several peaks in the 1920s, 1950s, 1970s-1990s and 2010s.We then focused on sedaDNA of major anthropochore taxa (13) to track the floristic dynamics associated with human activities, such as agriculture and urbanization.
Temporal consistencies are recorded among Brassicaceae, Poaceae, PACMAD clade, BOP clade, and Hordeinae, probably indicating historical crop cultivation such as rapeseed and/or cabbage, and cereal taxa.Indeed, PACMAD and BOP clade reflect specific DNA information from most cereal crops including maize, millet, sorghum (panicoideae subfamily) and rice (oryzoideae subfamily), wheat, barley, oat and rye (pooideae subfamily), respectively.The temporal evolution of these taxa is in good agreement with the grain production data collected and showing a maximum between the 1960s and 2000s (Fig. 4).While the exact composition of the cereal crops is difficult to establish due to the low taxonomic resolution, the presence of rice is supported by the detection of other herbs related to paddy fields (SI Appendix, Fig. S8b-c).
Taxa from the Rosaceae family, other than Potentilla sp. and Colurieae, may represent the cultivation of fruit trees.This cultivation increased sharply from the 1990s onwards, which is consistent with the high detection of DNA from Rosaceae while that from Potentilla sp.dropped.
Changes in Pinidae and other taxa suggested that trees were dominated by coniferous speciesthe fast-growing pioneer taxa until the 1960s.Subsequently, increases in Betulaceae (including Alnus sp.), Taxus sp., Broussonetia Papyrifera mainly reflected anthropogenic plantation for dike protection and erosion control along the lakeshore and riverside, and possible timber supply for building and paper industry (14).Some arboreal taxa that were widely cultivated in gardens and streets as ornament, such as Acer sp., Caprifoliaceae (including Lonicera sp.) and Spermatophyta (including Ginkgo) (13), first increased in the 1960s and then enriched around the 1980s, following the human population growth and development of urban areas.Herbaceous plants, like Galium sp., Scorzoneroides sp., Ranunculus sp., Maianthemu sp., Ononis sp., Teucrium sp., Polygonoideae and others were also detected from the 1960s, and become abundant during the 1980s-2000s (SI Appendix, Fig. S8d).Over the last two decades, however, sedaDNA of most plant taxa was at a low level or even undetectable.

Text S5. Implications of multi-proxy paleoenvironmental records
Lake act as a sink for material influx from the watershed, and therefore sediment records, are invaluable archives to evaluate past natural and anthropogenic modifications of terrestrial landscapes and aquatic ecosystems, in particular combining classical paleo-proxies and emerging ancient DNA indicators (15)(16)(17).For example, rubidium (Rb) and titanium (Ti) are mostly found in the fine-particle fraction and are geochemically conservative (18); zirconium (Zr) is enriched mainly in the silty-sandy fraction that contain heavy minerals such as zircon (19).
Increases in Zr/Rb point thus to more detrital inputs from coarse-grained sources (20).Sediment magnetic susceptibility (MS) depend on the abundance of magnetic minerals and also trace metals (21), which are typically supplied to lakes via watershed soil erosion and anthropogenic emissions (1,22).Lake sediments can also provide a long-term history of changes in natural and human pollutant input (e.g., toxic metals), and the possibility to evaluate temporal patterns of anthropogenic pollution (2).
Total extracellular DNA preserved in sediments is closely related to both lake and watershed biomass from all functional groups and the supply of fine particles (23)(24)(25).In particular, extracellular plant DNA becomes protected against microbial degradation when adsorbed on clay particles, and transported to lakes by surface runoff and erosion processes (25)(26)(27).In Lake Taihu, DNA was present in all samples through the sediment core, and showed a decrease in the upper part of the core against the expected degradation pattern of DNA decrease with depth/time.However, in 5 samples corresponding to CE 1915CE , 1930CE , 1950CE , 2011 and 2018, both the number of DNA reads and of taxa are very low, which has to be considered with caution.Our comprehensive analyses of plant sedaDNA, pollen, and geochemical records, which were subdivided into four phases of changes (Figs. 3 and 4), can be used to unveil century-scale dynamic changes of surface erosion, pollution, vegetation, and their relationships with human landscape shaping in the watershed.
Meanwhile, autotrophic communities are key components in determining ecosystem state, and they are strong indicators of water ecological dynamics in lakes (28)(29)(30)(31).Paleolimnological analysis of fossil photosynthetic pigments is an effective method to determine past ecological responses from nutrient enrichment and climate change, with different algal groups responding to different environmental drivers (4,32).Sedimentary chlorophylls (e.g., chlorophyll-a and its chemically-stable production of degradation -pheophytin-a) represents the pigments from total (ubiquitous) algae, thus can indicate lake primary production.Considering potential degradation effects with depth/age (30)(31)(32), chlorophylls in the topmost core sediment are abnormally high thus often be excluded from the variability analysis.The UVR index inferred from pigments has been proven to be a good indicator for past water clarity (7,8).Taken together, our analysis of sediment organic matter (TOC and C/N ratios), nutrients (TP), and particularly pigments can track past lake ecosystem dynamics.

Text S6. Research limitations and prospects
Although this study provides new methodological insights and inspiring findings, there are potential limitations and prospects that deserve increasing research investment.First, the metabarcoding approach developed here led to limited taxonomic resolution of plant taxa (especially cereal crops) we identified, which reduces the interpretative capacity.The development of a complete DNA reference library for local/regional species and the use of more specific primers (e.g., for Poaceae) (24,33) is a key priority to fully exploit the power of metabarcoding for past, longer time-scale community reconstruction and paleoecological evolution.Second, taphonomic (e.g., poor DNA transfer or preservation) and other analytical (e.g., PCR inhibition) issues can also appear and led to misleading results (17,25).However, in sedimentary systems highly controlled by detrital inputs, as in the case of Lake Taihu, we assume that taphonomic issues are reduced, although changes in the type of erosion (e.g., soil surface vs deep soil horizons) may affect the quality of the reconstructions (25,26).Third, the identified critical transitions of SES were based on empirical records with entropy-weighted integration (see the Materials and Methods), with associated caveats.Despite the enormous advantage of time series that sedaDNA and other paleoenvironmental data provided, their availability, accuracy, and precision can always be improved (34), particularly through more robust age models, multi-proxy and/or multi-source data comparison and validation (e.g., monitored and surveyed data) (35).The weights of the indicators/variables for the SES synthesis were only calculated using the classical numerical method, while not considering their specific ecological or socioeconomic roles in the system, primarily because most of the indicated processes are either not necessarily mutually exclusive, or are difficult to quantify (15).Nevertheless, we believe the integration of sedaDNA and other paleoenvironmental records in long-term social-ecological coupling research will lead to a more thorough understanding of Anthropocene environmental evolution in the Earth surface system and provide a more detailed picture of the complex mechanisms.As increasing data and modelling parameters become available in the future, the suggested approach can readily be applied to generate new knowledge for sustainable management of ecosystems, landscapes, and resources.S1.

Fig. S1 .
Fig. S1.Location of the Yangtze River Delta and Lake Taihu watershed in China.

Fig. S2 .
Fig. S2.Changes in land-use cover in the Lake Taihu watershed from the 1980s to 2010.Landuse data is interpreted from Landsat (TM/ETM+) images at the scale of 1:100,000, and can be accessed through the Lake and Watershed Data Center (https://lwdc.niglas.cas.cn).

Fig. S3 .
Fig.S3.Photograph of sediment core (TAI-18-2) from Lake Taihu, and vertical variations of sediment water content, grain size compositions, median grain size (Md), mass magnetic susceptibility (MS), total organic carbon (TOC), TOC/TN molar ratios (C/N), and total phosphorus (TP).Four general phases can be defined by the horizontal dashed lines.

Fig. S4 .
Fig. S4.Historical profiles of total terrestrial pollen concentrations, different pollen assemblages, and key pollen indicators related to major land cover change inferred from Lake Taihu sediments (TH1 core).The AP/NAP value (ratio of arboreal taxa proportion to non-arboreal taxa proportion) reflects forest variation.The A/W value(ratio of anthropochore taxa proportion to wetland taxa proportion) probably indicates the intensity of agricultural activities.Anthropochore taxa include plants which are spread by human activity and are intensively distributed in anthropogenic environments, such as Xanthium, Humulus, Chenopodiaceae, and other Compositae taxa.The wetland taxa in our study mainly include Cyperaceae and Polygonum.The sediment core is divided into two major pollen assemblage zones by CONISS (constrained incremental sum) clustering embedded in the Tilia program (36), with zone 1 containing two sub-zones (zone 1a and 1b).

Fig. S5 .
Fig. S5.Historical changes in algal chlorophyll and carotenoid pigments inferred from Lake Taihu sediments (TH1 core) (left), and nonmetric multidimensional scaling analysis (NMDS) of the pigment assemblage compositions (right).The dominant NMDS score (i.e., NMDS1 that explains 95% of variance) reflects algal community turnover.Dotted circles indicate the clustering groups as determined by the CONISS analysis.

Fig. S6 .
Fig. S6.Multivariate geochemical elements from XRF scanning of THM-2 sediment core and individual factor maps for the principal component analysis (PCA).The data can be transformed into three end-members (detritus, metallic pollution & organic matter, and coarse particles), according to their PC loading coefficients and geochemical properties.

Fig. S7 .
Fig. S7.Historical changes in typical geochemical elements from XRF results and three PC scores from the PCA results, together with their sedimentological and paleoenvironmental implications.Shaded zones shows periods of the relative high values over the past hundred years.

Fig. S9 .
Fig. S9.Nonmetric multidimensional scaling (NMDS) analysis of plant sedaDNA chronosequences based on the number of reads in four PCR replicates per sample.

Fig. S10 .
Fig. S10.Tipping points and potential regime shifts of the century-long biophysical processes in the complex system, as identified by sequential T-test algorithm based on segmental mean and calculation of cumulative sum of differences (CUSUM).The detected sedimentary paleo-variables include mass magnetic susceptibility (MS), Zr/Rb ratio, pollution load index of atmospheric trace metal (PLI), chlorophylls (sum of chlorophyll-a and pheophytin-a), UVR index, total sedaDNA concentration, and the numbers of taxa and DNA reads of terrestrial plants.The Regime Shift Index (RSI) and CUSUM peaks indicate major shifting timings during the 1950s-1980s.

Fig. S11 .
Fig. S11.GAM-based trends fitted to key socioeconomic (a-d) and biophysical (e-i) variables (z-score normalized) of Lake Taihu social-ecological system through time, and estimated first derivatives (black lines) with 95% simultaneous confidence intervals.The bold blue and red represents the significant periods of increase and decrease in the rates of change (RoCs) in the SES variables, respectively.Statistical summary for the GAMs is shown in SI Appendix, TableS1.
Fig. S14.Ecological footprint per capita and its components during 1980-2020 in the study region.Changes in ecological footprint related to human land-use in Jiangsu Province (a) and freshwater resource consumption in Lake Taihu watershed (b) over time have been shown.
Fig. S15.National ecological footprint (EF) per capita of China and its relationships with economic growth during 1980-2020.(a) Changes in national EF and its components over time (37).(b) Relationships between national ecological footprint (EF) per capita and GDP per capita of China with LOESS fit.The size of each data point represents the corresponding date (year), which is labeled around the data point.This graph generally shows synchronous increases in per capita GDP and ecological footprint at the national level, which shows limited decoupling trends of ecological footprint from economic growth in recent years.

Table S1 .
Statistical summary of GAM-based trends fitted to key socioeconomic, biophysical and climate variables of Lake Taihu SES through time.This table is a supplement of SI Appendix, Figs.S11 and S12.Biophysical dynamics (ecological degradation) regarding to anthropogenic soil erosion, primary production, and community turnover, are indicated by paleoecological time series of sediment mass magnetic susceptibility, total sedaDNA concentration, chlorophylls concentration, and NMDS scores of terrestrial plant community (sedaDNA) and aquatic algal community (pigments), respectively.The time-series of these variables are z-score normalized before model analysis.Model performance is indicated by the R-sq.(adj) and deviance explained (dev.expl.) of the respective models.

Table S2 .
GAM analysis summary for the relationships between socioeconomic and ecological RoCs in SI Appendix, Fig.S13.The summary for the additive model fitted to the ecological variable, with the probability levels obtained for two major socioeconomic drivers is presented.For all variables in the models, the estimated degrees of freedom (edf) and F values with corresponding significance levels (***p<0.001) are shown.

Table S3 .
Paleoenvironmental indicator records used for determining the dominant ecoenvironmental variability of Taihu lake-watershed system.

Table S5 .
(38)meters for fossil energy footprint account.The electricity consumption was converted into built-up land area, the other energy consumption types were converted into fossil production land area, and all the consumed energy was converted into fossil land area according to the heat conversion coefficient used in global energy statistics(38)and the FOA statistical databases.(ii) The unit of the conversion coefficient of electricity is GJ (kW•h) -1 .