
==== Front
Philos Trans R Soc Lond B Biol Sci
Philos Trans R Soc Lond B Biol Sci
RSTB
royptb
Philosophical Transactions of the Royal Society B: Biological Sciences
0962-8436
1471-2970
The Royal Society

10.1098/rstb.2023.0017
rstb20230017
100160207197129Articles
Opinion Piece
Using ancient sedimentary DNA to forecast ecosystem trajectories under climate change
Using ancient sedimentary DNA to forecast ecosystem trajectories under climate change
http://orcid.org/0000-0002-8610-1085
Alsos Inger Greve Conceptualization Funding acquisition Methodology Visualization Writing – original draft Writing – review & editing inger.g.alsos@uit.no
1
Boussange Victor Conceptualization Methodology Writing – original draft Writing – review & editing 2 3
http://orcid.org/0000-0002-6194-4414
Rijal Dilli Prasad Data curation Formal analysis Writing – original draft Writing – review & editing 1
Beaulieu Marieke Visualization Writing – review & editing 1
http://orcid.org/0000-0002-1990-4654
Brown Antony Gavin Visualization Writing – review & editing 1
http://orcid.org/0000-0003-0999-1261
Herzschuh Ulrike Writing – review & editing 4 5
http://orcid.org/0000-0002-3415-0862
Svenning Jens-Christian Writing – review & editing 6
http://orcid.org/0000-0002-2289-8259
Pellissier Loïc Conceptualization Methodology Writing – original draft Writing – review & editing 2 3
1 The Arctic University Museum of Norway, UiT The Arctic University of Norway, 9037 Tromsø, Norway
2 Department of Environmental System Science, ETH Zürich, Universitätstrasse 16, 8092 Zürich, Switzerland
3 Swiss Federal Research Institute WSL, Zürcherstrasse 111, 8903 Birmensdorf, Switzerland
4 Alfred Wegener Institute for Polar and Marine Research, Telegraphenberg A43, 14473 Potsdam, Germany
5 Institute of Environmental Sciences and Geography, Potsdam University, 14479 Potsdam, Germany
6 Center for Ecological Dynamics in a Novel Biosphere (ECONOVO) & Center for Biodiversity Dynamics in a Changing World (BIOCHANGE), Department of Biology, Aarhus University, Ny Munkegade 114, 8000 Aarhus C, Denmark
One contribution of 16 to a theme issue ‘Ecological novelty and planetary stewardship: biodiversity dynamics in a transforming biosphere’.

Electronic supplementary material is available online at https://doi.org/10.6084/m9.figshare.c.7093279.

27 5 2024 May 27, 2024
8 4 2024 April 8, 2024
8 4 2024 April 8, 2024
379 1902 Theme issue ‘Ecological novelty and planetary stewardship: biodiversity dynamics in a transforming biosphere’ compiled and edited by Jens-Christian Svenning, Melodie A. McGeoch, Signe Normand, Alejandro Ordonez and Felix Riede 202300171 7 2023 July 1, 2023
22 10 2023 October 22, 2023
© 2024 The Authors.
2024
https://creativecommons.org/licenses/by/4.0/ Published by the Royal Society under the terms of the Creative Commons Attribution License http://creativecommons.org/licenses/by/4.0/, which permits unrestricted use, provided the original author and source are credited.

Ecosystem response to climate change is complex. In order to forecast ecosystem dynamics, we need high-quality data on changes in past species abundance that can inform process-based models. Sedimentary ancient DNA (sedaDNA) has revolutionised our ability to document past ecosystems' dynamics. It provides time series of increased taxonomic resolution compared to microfossils (pollen, spores), and can often give species-level information, especially for past vascular plant and mammal abundances. Time series are much richer in information than contemporary spatial distribution information, which have been traditionally used to train models for predicting biodiversity and ecosystem responses to climate change. Here, we outline the potential contribution of sedaDNA to forecast ecosystem changes. We showcase how species-level time series may allow quantification of the effect of biotic interactions in ecosystem dynamics, and be used to estimate dispersal rates when a dense network of sites is available. By combining palaeo-time series, process-based models, and inverse modelling, we can recover the biotic and abiotic processes underlying ecosystem dynamics, which are traditionally very challenging to characterise. Dynamic models informed by sedaDNA can further be used to extrapolate beyond current dynamics and provide robust forecasts of ecosystem responses to future climate change.

This article is part of the theme issue ‘Ecological novelty and planetary stewardship: biodiversity dynamics in a transforming biosphere’.

modelling
, sedimentary ancient DNA
, forecast
, ecosystem
, biodiversity
, time series
Norges Forskningsråd http://dx.doi.org/10.13039/501100005416 250963/F20 Schweizerischer Nationalfonds zur Förderung der Wissenschaftlichen Forschung http://dx.doi.org/10.13039/501100001711 310030E_205556 310030_188550 FeedBaCks, 193907 Danmarks Grundforskningsfond http://dx.doi.org/10.13039/501100001732 DNRF173 Villum Fonden http://dx.doi.org/10.13039/100008398 16549 HORIZON EUROPE European Research Council http://dx.doi.org/10.13039/100019180 No 819192 Biodiversa+ http://dx.doi.org/10.13039/100019184 BiodivClim ERA-Net COFUND cover-dateMay 27, 2024
==== Body
pmc1. Introduction

Ecosystem change results from species response to abiotic drivers and dynamic species interactions. Ecological processes such as dispersal, growth and reproduction occur on a range of timescales, ultimately resulting in complex ecosystem dynamics [1]. While process-based ecosystem models are well-developed [2,3], most forecasts of biodiversity responses to global changes are conducted using species distribution modelling which is based on correlations between current species’ ranges and climate [4]. While these models assume that species demography is at equilibrium dictated by their environmental niche, there are increasing observations that species only partially track their predicted suitable habitat due to transient demographic mechanisms [5,6]. Furthermore, global warming can reshape the structure of communities, notably through the emergence of novel biotic interactions and temperature-dependent competition [3,7]. Forecasts of the future distribution of species and ecosystems are limited by processes that modify rates of change and thus the timescale of the dynamics of the system [8]. Yet, correlation-based, static models are often preferred over their process-based, dynamic counterparts, because of the difficulty of assimilating this data in process-based models, due to limited knowledge of biological processes and rates, and a lack of comprehensive temporal data on whole ecosystems, including multi-species assemblages or communities across trophic levels. We argue that a leap forward in our ability to model the dynamics of whole ecosystems, including the testing of apparent species interactions, is offered by sedimentary ancient DNA (sedaDNA).

Palaeo-records such as macro- and microfossils have allowed the documentation of past changes in ecosystems [9–11]. For example, dynamic population models have been fitted to pollen data of four tree taxa to identify the relative importance of temperature change, nitrogen availability and species interaction in determining population dynamics [12]. Similarly, process-based community models with pollen data have been used to study how competition among tree species, density-dependent survival and dispersal rate affect tree abundance [13]. These studies have mainly focused on a handful of species or on shorter time frames, mainly due to the limited taxonomic resolution of pollen data and the scarcity of macrofossils of most species.

SedaDNA records have the potential to overcome the limitations of traditional micro- and macrofossils data [14]. In particular, sedaDNA data provide better taxonomic resolution and can recover the abundance of a broader range of species. These include, for example, insect-pollinated forbs, which are generally underestimated in pollen records [15,16], and higher trophic levels like mammals [17–19]. Since sedaDNA data contains richer information on past dynamics through time, it has great potential for an improved understanding of the drivers and processes of change. Combining it with process-based models may allow us to improve our forecasts of the effects of ongoing climate change on future biodiversity.

Here, we assess the opportunities of combining sedaDNA data with process-based models to study complex ecosystem dynamics under past and future climate changes. We discuss the main advantages of sedaDNA data over previous techniques for measuring community change. Aiming at studying biodiversity as a whole, we focus on multicellular organisms, lake sediments and entire communities. Extending beyond plants, our framework can be used to model complete food webs through animal sedaDNA. We discuss how complementary data sources can be used, such as climate reconstructions, as well as nitrogen deposition and other global change factors. Finally, we provide a roadmap for the combination of this data with process-based models in order to better quantify the relative importance and rate of biological changes.

2. The advantage of sedaDNA data

(a) The nature of sedaDNA data

Ancient DNA can be found in a range of substrates like ice cores, permafrost, soils, lake sediments, archaeological artefacts and caves [20]. For time series, it is essential to have a good age–depth model. In general, the most continuous and reliable age–depth models are obtained from lake sediments [21], and therefore we mainly focus on lake sediments here.

Lake sedaDNA can be detected in the absence of plant macroscopic remains and is generally thought to derive from DNA binding to clay or other fine particles transported to the lake [14,22]. Whilst some lakes produce poor results due to their sediment composition (acidic bedrocks or algal-dominated gyttja) there is no evidence that changes in natural sediment type has differential effects on the diversity of DNA from plant or animals species, although additions to the sediment, such as bone, may enhance certain taxa [23–25]. It is possible to obtain DNA from ancient pollen using dedicated protocols. However, pollen is in general not the source of sedaDNA due to relative low biomass contribution and low chloroplast DNA content [14,26], thus the uncertainties in distinguishing long-distance from local source is avoided. Similar to macrofossil, the sedaDNA source area is almost exclusively from within the hydrological catchment and particularly close to the lake, which gives a spatially defined source region [25,26]. This is important for modelling as it means that the sedaDNA from even closely located lakes is truly independent, which is not the case with pollen.

By sampling lake sediment cores, one can extract DNA and then: (i) amplify the DNA of interest using a primer (a short sequence DNA sequence that serves as a starting point for PCR synthesis) dedicated to the organism group of interest for example mammals or vascular plants (metabarcoding), (ii) ‘capture’ the DNA fragment of interest using capture-probes before sequencing, or (iii) go directly to the sequencing step (shotgun sequencing) which would yield DNA of all organisms in the samples. While the shotgun sequencing has the advantage that it works well even for very old and fragmented DNA, avoids potential bias caused by PCR (see below), and gives simultaneous data on plants and animals, the total gain of target DNA fragments is low and the taxonomic resolution obtainable from metabarcoding studies is not (yet) reached. Because of the higher costs, the time resolution of the few studies undertaken to date is low [17,18,27]. Similarly, only two studies have used broad capture-probes so far, and the taxonomic resolution is low for plants [28,29]. The most commonly used method is metabarcoding, which gives the highest taxonomic resolution for plants [30], and with recent method improvement, also detects mammals well [19,24]. Early studies showed low overlap and poor richness in sedaDNA compared to pollen and macrofossil [31,32], but as DNA reference libraries have expanded and molecular methods have improved, the overall richness in sedaDNA studies is typically 2–4 times higher than pollen [16,33,34]. Thus, sedaDNA may give more complete information about past vegetation than has been possible before, and may also add information about animals in the absence of macrofossils (figure 1). Figure 1. A reconstruction of plants and animals at Nordvivatnet, north Norway, through the last 12 000 years shows examples of species that can be detected in sedaDNA studies (drawn after Brown et al. [24]). Illustration by Siw Utheim Pedersen.

To identify the DNA fragments obtained, a series of bioinformatic steps are involved, of which the most crucial is matching sequences to a DNA reference library [35]. Both false positives and false negatives occur in sedaDNA data, and stringent processes from field to laboratory and bioinformatics are therefore needed [26,36,37]. Conservative data filtering may minimize false positives, but this will always be at the cost of losing true positives [25], so one needs to find a balance. Common reasons for false identification are that closely related taxa may have identical sequences for the marker used, and/or errors in the reference library. The availability of highly curated local DNA reference libraries, such as PhyloAlps or PhyloNorway, reduces the chance of false identification, and may give 40–50% identification to species level [15,19]. As the availability of highly curated reference libraries increases [38], older datasets may be re-analysed to improve identification.

Metabarcoding data provides presence/absence data or quantitative data, and both measures may be biased by the transport and deposition of DNA, the quality of DNA obtained, and technical issues during DNA amplification and bioinformatic analyses [22,25]. For example, the PCR procedure may cause amplification bias due to sequence length and composition, as well as possible mismatch in primer binding sites [25,39]. Repeating the PCR analyses may provide a more robust detection of species [40], and a conservative quantification using these repeats may be advantageous [15,41]. Nevertheless, some taxa such as willows and aquatic macrophytes are commonly assumed to be overrepresented in both metabarcoding and shotgun studies [15,17]. This is likely because their habitat is in/along streams and lakes, and therefore more DNA enters the lake [15,17,25]. However, estimates of ecosystem changes can be quite robust to the quantitative measurement of PCR repeats or proportion of reads, probably because this bias is not changing over time [15].

Metabarcoding is now approaching standardization across different laboratories, which allows dataset pooling [41,42]. There are high-resolution metabarcoding data available from mainly arctic and alpine sites, where also most ancient DNA studies in general have been done (figure 2). However, studies from some warmer regions show promising results: for example, in Italy [44] where Holocene sediments generally provide good DNA quality and some DNA was obtained from up to 31 000 years ago; an African savanna site with good DNA over the last 170 years [45]; and an early study on an African high altitude site with up to 5000 year old DNA [46]. There is a large metabarcoding circum-Arctic study of 21 sites and 242 samples [47], but this is based on permafrost. The largest lake sediment metabarcoding study is one of 10 northern Fennoscandian lakes (387/355 samples x 8 PCR replicates, [15,41]), and a study of 705 samples from 14 lakes in the European Alps will soon be available (figure 2; electronic supplementary material, table S1). There is also an increasing number of lakes analysed from Siberia and the Tibetan Plateau [34,48] that generally cover long time spans (up to about 50 kyr), and many shorter records from the British Isles [49,50] (figure 2, electronic supplementary material, table S1). By using quality measures and standardization [41], the dataset may be combined into even larger datasets. Figure 2. Map of lakes where sedaDNA metabarcoding has been analysed for mammals and/or vascular plant. The data is based on von Eggers et al. [43] but limited to original metabarcoding studies of lake sediments and updated with relevant papers from the sedaDNA society track list (https://sedadna.github.io/ October 2023) and unpublished study sites from our own research groups (see electronic supplementary material, table S1). Square denotes lakes used in figure 3. To show the broader potential to obtain ancient DNA data from other regions, all sites from [43] are shown as crosses except when sediment type was cave, lake surface, marine surface, midden, sediment trap or soil, and except sites that lacked age or had age = 0 (except for permafrost sediments and sites in Wang et al. [17] which we assume are ancient).

In association with sedaDNA data, complementary abiotic and biotic information can be obtained from the sediment core. It is common to make other simultaneous measurements such as X-ray fluorescence (XRF), magnetic susceptibility, stable isotopes (N and C) and biomarkers [51–53]. In addition, independent climate reconstructions such as CHELSA [54], NGRIP [55], oxygen isotope [56,57] or chironomids may be used [19], although, as with all palaeoproxies, one has to account for uncertainties due to different spatial and temporal resolution [58]. These provide important information about changes in the environment. Thus, by combining sedaDNA data with other proxies, we are moving towards the reconstruction of both abiotic changes and species composition which allows studies of interactions, food webs and, ultimately, whole-ecosystem changes over time.

The dynamics of ecosystems are influenced by natural and anthropogenic environmental changes, which can include a variety of abiotic factors such as temperature, precipitation and soil characteristics. Biotic interactions regulate the coexistence between species and play a critical role in shaping ecosystem changes [3,59]. The interactions between species can take many forms, including competition for resources such as food, water and shelter, predation, facilitation, inhibition, mutualism and parasitism [60,61]. Understanding how these interactions affect the changes in species abundance and distribution, and how they are themselves influenced by abiotic factors, is essential for predicting the response of ecosystems to all environmental changes [3].

Direct quantification of biotic interactions can be a challenging task, but inverse modelling techniques can be used to extract the signatures left by biotic interactions on past ecosystem dynamics. SedaDNA time series are highly relevant in this task. The majority of metabarcoding studies have focused on plants and microbes, but an increasing number of studies also include other organisms such as mammals [22,43]. Early metabarcoding methods gave scattered records of mammals [47] or provided good detection only of domestic mammals, which are commonly at high densities [49,62]. It has generally been challenging to obtain mammal DNA from sediments [63] for two reasons: (i) the abundance of mammal DNA is lower and more patchy than for plants, and (ii) co-amplification of human DNA during PCR [63,64]. Co-amplification of human DNA was partly overcome by adding small fragments of DNA that bind to human DNA, so-called blocking oligos, on both forward [62] and, more recently, reverse primers [19]. Using both blocking oligos increases detection of mammals and also co-amplifies birds, fish, amphibians and worms (figure 1). Additional organisms, such as diatoms, fungi and lichens, can be amplified through the use of dedicated primers [65–67], giving a broader spectrum of biodiversity. Having multiple trophic levels allows a more comprehensive tracing of ecosystem dynamics.

The high taxonomic resolution obtained in recent sedaDNA studies of plants allows the assigning of traits that provide a window on ecosystem characteristics and the function of species. Species-level information allows linking of pollinator dependence, nutrient demands, ability to compete with established vegetation, and dispersal mode [15]. For example, certain plants may inform us about past human land-use and its effects on biodiversity [19,50,62]. Further, soil disturbance, temperature optimum and moisture values can be estimated from plant traits, and be shown to be correlated with local glacial activity [68]. The high taxonomic resolution obtained from northern Fennoscandia allowed the combining of 227 of 238 vascular plant taxa with the Swedish plant trait database of 30 traits [15,69]. Also, species-level information on mammals can be used to reconstruct trophic interactions linking species through prey–predator dynamics or food webs. Thus, by linking species to traits, one can reconstruct past external drivers such as human or glacial activity. Furthermore, our understanding of how traits relate to one another can be increased and we can gain functional understanding of internal ecosystem dynamics. Traits can also be used to reduce the dimensions in ecological equations and thereby reduce ecosystem model complexity [70].

(b) Temporal resolution and processes

The processes that shape ecosystem dynamics occur over much longer time scales than the few decades over which we have been monitoring ecosystems. In the most extreme cases, legacies of the last glaciation are still found in current species distribution patterns [71–73]. As regards modern time series, most span only a few decades and even the best resolved time series are usually yearly and focus on only one taxonomic group [74]. Additionally, the temperature changes that are predicted to occur over the next few decades have only occurred over millennial timescales in the past. Therefore, it is crucial to use long-term ecosystem time series for the calibration of models. It is important to recognise that all ecosystem processes occur at differential rates, and some may occur quite rapidly. For example, competitive exclusion can have a significant effect on an ecosystem and may happen over just a few decades. Therefore, high-resolution time series are essential for accurately capturing the dynamics of these processes. The use of sedaDNA data is a promising approach for obtaining longer-term ecosystem time series. However, it remains to be seen how far we can extend these records and how well they can capture the full range of ecosystem processes. High temporal resolution (on average every 1.3–4 years) is available from short-term studies covering a few decades up to three centuries [26,45,67]. For millennial-scale studies, the highest temporal resolutions we are aware of are on average every 53 [23] and 64 years [15,41], but there are large numbers of sites published or in preparation with time resolution of decades to a few hundred years (figure 2; electronic supplementary material, table S1). Dense sampling across rapid climatic changes such as the Bølling/Allerød-Younger Dryas-Holocene boundaries at every 87 years, are particularly relevant for ongoing studies of rapid climate changes [75].

To explore the minimum density of samples needed to obtain a representative temporal resolution, we sub-setted one to three samples from each 500 years bin from Alsos et al. [15]. We used lake Gauptjern which had mean temporal resolution of 153 years over the last 8.5 kyr. When 18, 33 and 44 samples (resulting in mean temporal resolution 475, 264 and 198 years, respectively) were re-analysed, both the taxonomic richness and compositional turnover of all sampling frequencies showed similar and comparable patterns to the full core (figure 3a,b). However, although not significant, we do note that both parameters may be underestimated for some periods, e.g. around 5.5 and 4.5 ka. We also note that this is a boreal region with less fluctuating pattern than for example seen in the Alps [19]. Thus, we advocate a minimum temporal resolution of every 150–260 years. Figure 3. Impact of temporal and spatial resolution on richness and compositional turnover detected. Effect of temporal sample resolution at one lake (lake Gauptjern, 8.5–0 cal BP, figure 2) on (a) taxonomic richness and (b) compositional turnover. Effect of spatial resolution using 1 to 10 lakes on (c) taxonomic richness and (d) compositional turnover. Data from Alsos et al. [15]. Coloured shadings indicate 95% confidence interval of the fitted values of the generalised additive models (GAM).

(c) Toward a denser network of sediment records for spatial processes

The temporal dynamics of species within an ecosystem may switch between internal processes and external factors, such as dispersal from other ecosystems [76]. Local ecological systems are connected to form meta-communities that exchange species. The addition of new species through dispersal can bring new ecological traits that can alter the abundance of species and the overall ecological trajectory of the ecosystem [15]. For example (i) the arrival of tree species generally causes a major change to arctic vegetation [77], (ii) the reintroduction of wolves has had a significant impact on the ecological dynamics of Yellowstone [78], and (iii) beaver, which is well recorded in sedaDNA [24], has profound effects on temperate and boreal ecosystems [79]. This highlights the importance of studying both the internal dynamics of an ecosystem and its interactions with other ecosystems in order to fully understand its functioning. In ecological research, there has traditionally been a trade-off between spatial and temporal resolution, with studies often focusing on one or the other. However, increased availability of high taxonomic and temporal resolution sedaDNA data makes it possible to simultaneously study both spatial and temporal dynamics of ecosystems through time. By analysing sediment samples from lakes, researchers can obtain millennia-long time series of species composition, providing a unique window into the spatial dynamics of these ecosystems. While some lakes have been studied using this technique, the vast majority remain uninvestigated and represent a wealth of untapped information about the history of ecosystems. Expanding the use of sedaDNA analysis to more sites, especially to regions currently underrepresented (figure 2), will generate a better understanding of the ecosystem change at a range of temporal and spatial scales.

To determine past dispersal rates, spatially and temporally high resolution data are required. The spatial resolution of sedaDNA studies is far lower than currently available for pollen, with only about 100 lakes analysed for plant sedaDNA [42,43] (figure 2). However, even a scattered network of sites allows the production of past distribution maps and the calculation of minimum dispersal rates, as has been done for tree species based on pollen [80,81]. As there is currently no or only scattered fossil information on the majority of species, this already has provided a major increase in knowledge that can inform models.

The number of sedaDNA studies is increasing rapidly [22,82], and for some regions, average distance among sites is low, allowing the estimation of regional post-glacial arrival patterns of species as well as relative importance of abiotic and biotic factors in determining arrival times [15]. Furthermore, such data can be used to calculate the time from arrival to local spread, allowing the possibility of incorporating a colonisation term into the models. The dataset also allows us to test how many study sites within a region are needed to detect consistent spatial patterns. Based on our estimates of the effect of spatial resolution on species richness and turnover (figure 3c,d), sedaDNA is starting to provide consistent temporal patterns if a broader time period (e.g. 500-year bin) is represented in two lakes. For example, temporal trends of both taxonomic richness and compositional turnover remain comparable (figure 3c,d) at the regional scale in northern Fennoscandia, when a 500-year bin is represented by one random sample each from two randomly chosen lakes, indicating that as few as two lakes can be sufficient to capture regional trends of diversity. Thus, for some regions, the available data already provides sufficient coverage to infer both spatial and temporal variation using a meta-community framework, while for more heterogenous regions, more sites may be required.

3. Confronting dynamic models with sedaDNA data

(a) Modelling ecosystem trajectories

Development and validation of models are necessary to obtain a more comprehensive understanding of the ecosystem changes, and a good model will eventually permit robust predictions [83]. Two modelling paradigms are commonly used to model biodiversity. On the one hand, correlative models, such as species distribution models (SDMs, [4,84]), are derived from statistical patterns extracted from the data and deriving relationships between species presence (or abundance) and environmental covariates (e.g. temperature and precipitation). Predictions with correlative models assume that potentially very complex [85] patterns contained in the data will repeat [86]. On the other hand, process-based models explicitly account for the processes underlying ecosystem dynamics [87]. Process-based models rely on data for their calibration and the estimation of the initial state of the system but are built together with a priori scientific knowledge about the system [88]. The process knowledge embedded in their structure renders them more robust for predicting the future trajectories of ecosystems under novel ecological conditions [87]. Process-based models are theoretically superior to correlative models since statistical patterns observed at time t may not hold at time t + 1 because of ecological feedbacks, which trigger large shifts in ecosystem states [59,86]. Moreover, process-based models can account for both abiotic factors and interactions between species, which can provide a more accurate representation of how biodiversity will respond to changes in their environment [86].

Correlative models are nonetheless the most used approach due to their sound statistical basis and practicality [89]. SedaDNA data may be used in combination with palaeoclimate models to validate species presence back in time, as has also been done with macrofossil and pollen [90,91]. Modelling efforts with sedaDNA data have included individual-based models and simulations under climate warming [92,93]. To our knowledge, sedaDNA has yet to be used for the purpose of differential equation-based modelling, which has the advantage of being much more scalable than other process-based modelling approaches [94].

A process-based dynamical model describes how the population of a particular species changes over time in an ecosystem (figure 4). Denoting the abundance of a species i at site k by Nik, a dynamical community model generally takes the form:3.1 1NikddtNik⏞p.c. growth rate=bi(uk(t))⏟basal growth rate−∑ j=1S[αi,jNjk⏞competition or facilitation+ε[F(N)]i,jNjk−[F(Nk)i,jNjk]⏟trophic interactions]+∑l=1Lmik,lNil−Nik∑l=1Lmil,k⏟dispersal,

where NK=(Nik,…,NSk) is a vector containing species abundance at site k. Equation (3.1) relates the per capita growth rate of species, 1/Nik dNik/dt, to the biotic and abiotic processes that affect them. These processes generally comprise reproduction, captured by the basal growth rate bi which may depend on time-dependent environmental drivers uk(t). They further include biotic interactions that depend on the abundance of other species Njk, which comprise competition and facilitation, captured by the coefficient αi,j measuring the competition or facilitation strength between species i and species j, and comprises trophic interactions, captured by the functional response F. Lastly, they can include spatial processes, i.e. dispersal across locations, where mik,l is the dispersal rate for species i from site k to site l. By explicitly accounting for biological processes, equation (3.1) constrains the model predictions. For instance, equation (3.1) predicts that the growth rate of a species decreases with increasing competition from other species, as captured by the competition coefficient αi,j. Also, with spatial dispersal (mik,l>0), the model allows some metapopulation dynamics, as the abundance of a species at one location can influence its abundance at other locations. Dynamical models are simulated over time by numerically integrating equation (3.1). This requires an initial condition estimate of the system, which together with the model parameter values will determine the ecological trajectory. Figure 4. Fitting a general dynamical community model of three plant species using sedaDNA data. Here, the dynamics of species (blue, red, black) at two sites depend on the species' basal growth rates, which are differentially affected by abiotic environmental factors such as temperature, as well as interactions between species and dispersal of species between sites. Climate models can provide estimates of past temperature while the sedaDNA time series data itself (here illustrated as change in abundance [proportion of repeats]), supported by external input (expert knowledge, experimental data, etc.), can be used to fit the parameters of complex and competing models in order to best describe the past dynamics of biotic communities as well as project future community dynamics under novel environmental conditions.

(b) Inverse modelling

The parameters and processes within process-based models have traditionally been determined independently of the model, with empirical data only used for model validation and comparison [95]. Independent estimation of parameters and processes rapidly becomes impossible as the number of species modelled increases, especially when considering species interactions, dispersal, etc. Inverse modelling, which consists in using observation data to recover the parameters of a model that can best explain the data, allows the bridging of this gap [95,96]. In particular, the sedaDNA observations contain signatures of underlying biological processes that have shaped the dynamics of ecosystems, and can therefore be used to parameterize the model. These signatures may be tapped by inverse modelling methods to recover the processes by matching model predictions and data [97]. Different approaches may be used for this process, such as Bayesian methods or variational methods, all of which rely upon the estimation of the likelihood of the model parameters given the data [98,99]. The complexity of the models, which mirrors the intricacies of biotic processes, should be thoughtfully selected in alignment with the quality and quantity of sedaDNA data—a principle reminiscent of Ockham's razor [83]. In reducing model complexity, trait-based transfer functions emerge as a valuable tool, streamlining the modelling process [70].

The interpretation of the calibrated model parameters can provide information on the ecological processes taking place and advance ecological theory. For instance, the sign and absolute value of the competition/facilitation interaction term (α) in equation (3.1) informs about the interactions between species. To effectively interpret inferred parameters, it is essential to account for uncertainty estimates. Bayesian methods excel in explicitly handling uncertainties when compared to variational methods, but they do come with a notable computational overhead. A prudent approach involves generating synthetic data by simulating information using a model with known parameters [100,101]. This step serves as a crucial validation process to confirm the model's ability to successfully extract and reflect biotic processes present in the real data. Moreover, prior information on biotic processes, whether derived from expert opinion [102,103], experimental data or other data sources [2] can be useful to help characterise biotic processes with weak or complex signatures. For communities with no extant analogs, the repetition of inferred biotic interactions across independent sites can provide greater support for causal inference. Inverse modelling may be further used for hypothesis testing. In such cases, alternative models corresponding to the different hypotheses are formulated, and inverse modelling is used to evaluate the likelihood of each model [104].

The calibrated model may eventually be used to make predictions about future ecosystem dynamics under different environmental change scenarios. Future trajectories of the ecosystem may be simulated, for instance to predict how the abundance of different species might change under climate change [105]. By comparing the predicted dynamics under different scenarios, ecologists can gain insights into how ecosystems might respond to different environmental stressors and inform conservation and management strategies.

4. Future perspectives

The combination of sedaDNA data with process-based models offers a unique approach to understanding past ecosystem dynamics and how they may respond to global changes. The reliable detection of taxa still represents a challenge for sedaDNA studies. To obtain better data, especially for rare species, one could add more technical replicates as well as increasing sampling density both in time and space, and so reducing the rate of false negatives [25]. This would better inform process-based models and make the use of sedaDNA even more relevant to species of conservation concern. Also, the current paper focuses mainly on plants and mammals from metabarcoding data, as these are the most readily available now and best studied in terms of sedaDNA, present ecology and modeling. For other organism groups, it would be valuable to compare sedaDNA data with an independent record of species occurrence, to know how well the species are represented in the sedaDNA data [25]. Further improvement is offered by (i) improved DNA reference libraries to ensure correct identification, (ii) multiplexing and other method development to detect more organism groups from the same samples, and (iii) more sites investigated, especially for ecosystems currently underrepresented in sedaDNA studies. SedaDNA data are generally available in raw format in data repositories such as Dryad. An important step will be to make the sedaDNA data available in an easily accessible format such as the final species list, for example in Neotoma [42], which can be combined with other databases of for example traits. Thus, we expect that in the near future, sedaDNA will be commonly used in combination with process-based models to forecast future biodiversity. With sedaDNA that contains rich information on past dynamics, modellers can at last test and refine their biodiversity models and theory at the species, community or ecosystem level. This is a major step forward, given that ecological modellers have until now barely confronted model predictions against data, and were largely confined to theoretical predictions.

An important postscript is that our approach outlined above relies on the assumption that traits observed in the current time were also operating in the past. This might not be the case if species currently are restrained by interactions that did not take place in the past. However, using independent proxies, niche stability through time may be explored. As more metagenomic data becomes available, one can imagine obtaining functional genes directly [106] and thereby omitting the step of species and traits. However, this would also require knowledge on how genes relate to the environment, which is an exciting emerging avenue.

Acknowledgments

We thank Siw Utheim Pedersen for the illustration in figure 1, Signe Normand and two anonymous reviewers for constructive comments on the draft, and our research groups for fruitful discussion about the prospects of using sedaDNA for future forecasts.

Data accessibility

Data used in figure 3 are available in Alsos et al. [15].

The data used in figure 2 are provided in the electronic supplementary material [107].

Declaration of AI use

We have not used AI-assisted technologies in creating this article.

Authors' contributions

I.G.A.: conceptualization, funding acquisition, methodology, visualization, writing—original draft, writing—review and editing; VLJ.B.: conceptualization, methodology, writing—original draft, writing—review and editing; D.P.R.: data curation, formal analysis, writing—original draft, writing—review and editing; M.B.: visualization, writing—review and editing; A.G.B.: visualization, writing—review and editing; U.H.: writing—review and editing; J.S.: writing—review and editing; L.P.: conceptualization, methodology, writing—original draft, writing—review and editing.

All authors gave final approval for publication and agreed to be held accountable for the work performed therein.

Conflict of interest declaration

We declare we have no competing interests.

Funding

I.G.A. was supported by the European Research Council (ERC) under the European Union's Horizon 2020 research and innovation programme (grant no. 819192) and the ECOGEN project ‘Ecosystem change and species persistence over time: A genome-based approach,’ financed by Research Council of Norway (grant no. 250963/F20). J.C.S. was supported by the VILLUM Investigator project ‘Biodiversity dynamics in a changing world’, funded by VILLUM FONDEN (grant no. 16549), and Center for Ecological Dynamics in a Novel Biosphere (ECONOVO), funded by Danish National Research Foundation (grant no. DNRF173). V.B. was supported by the 2019–2020 BiodivERsA joint call for research proposals, under the BiodivClim ERA-Net COFUND programme, and with the funding organisation Swiss National Science Foundation SNF (project: FeedBaCks, grant no. 193907). L.P. was supported by the Swiss National Science Foundation (grant nos 310030_188550 and 310030E_205556).
==== Refs
References

1. Tilman D. 1994 Competition and biodiversity in spatially structured habitats. Ecology 75 , 2-16. (10.2307/1939377)
2. Urban MC et al. 2016 Improving the forecast for biodiversity under climate change. Science 353 , aad8466. (10.1126/science.aad8466)27609898
3. Åkesson A, Curtsdotter A, Eklöf A, Ebenman B, Norberg J, Barabás G. 2021 The importance of species interactions in eco-evolutionary community dynamics under climate change. Nat. Commun. 12 , 1-12. (10.1038/s41467-020-20314-w)33397941
4. Guisan A, Thuiller W. 2005 Predicting species distribution: offering more than simple habitat models. Ecol. Lett. 8 , 993-1009. (10.1111/j.1461-0248.2005.00792.x)34517687
5. Harrison S. 2020 Plant community diversity will decline more than increase under climatic warming. Phil. Trans. R. Soc. B 375 , 20190106. (10.1098/rstb.2019.0106)31983333
6. Svenning J-C, Sandel B. 2013 Disequilibrium vegetation dynamics under future climate change. Am. J. Bot. 100 , 1266-1286. (10.3732/ajb.1200469)23757445
7. Descombes P, Pitteloud C, Glauser G, Defossez E, Kergunteuil A, Allard P-M, Rasmann S, Pellissier L. 2020 Novel trophic interactions under climate change promote alpine plant coexistence. Science 370 , 1469-1473. (10.1126/science.abd7015)33335062
8. Alexander JM et al. 2018 Lags in the response of mountain plant communities to climate change. Glob. Chang. Biol. 24 , 563-579. (10.1111/gcb.13976)29112781
9. Huntley B. 1990 Studying global change: the contribution of Quaternary palynology. Glob. Planet. Change 2 , 53-61. (10.1016/0921-8181(90)90033-9)
10. Birks HJB. 2019 Contributions of Quaternary botany to modern ecology and biogeography. Plant Ecol. Divers. 12 , 189-385. (10.1080/17550874.2019.1646831)
11. Milligan G, Bradshaw RHW, Clancy D, Żychaluk K, Spencer M. 2020 Effects of human land use and temperature on community dynamics in European forests. Quat. Sci. Rev. 247 , 106458. (10.1016/j.quascirev.2020.106458)
12. Jeffers ES, Bonsall MB, Froyd CA, Brooks SJ, Willis KJ. 2015 The relative importance of biotic and abiotic processes for structuring plant communities through time. J. Ecol. 103 , 459-472. (10.1111/1365-2745.12365)
13. Ramiadantsoa T, Stegner MA, Williams JW, Ives AR. 2019 The potential role of intrinsic processes in generating abrupt and quasi-synchronous tree declines during the Holocene. Ecology 100 , e02579. (10.1002/ecy.2579)30707453
14. Parducci L, Bennett KD, Ficetola GF, Alsos IG, Suyama Y, Wood JR, Pedersen MW. 2017 Transley Reviews: Ancient plant DNA from lake sediments. New Phytol. 214 , 924-942. (10.1111/nph.14470)28370025
15. Alsos IG et al. 2022 Postglacial species arrival and diversity buildup of northern ecosystems took millennia. Sci. Adv. 8 , eabo7434. (10.1126/sciadv.abo7434)36170372
16. Clarke CL et al. 2020 A 24,000-year ancient DNA and pollen record from the Polar Urals reveals temporal dynamics of arctic and boreal plant communities. Quat. Sci. Rev. 247 , 106564. (10.1016/j.quascirev.2020.106564)
17. Wang Y et al. 2021 Late Quaternary dynamics of Arctic biota from ancient environmental genomics. Nature 600 , 86-92. (10.1038/s41586-021-04016-x)34671161
18. Kjær KH et al. 2022 A 2-million-year-old ecosystem in Greenland uncovered by environmental DNA. Nature 612 , 283-291. (10.1038/s41586-022-05453-y)36477129
19. Garcés-Pastor S et al. 2022 High resolution ancient sedimentary DNA shows that alpine plant diversity is associated with human land use and climate change. Nat. Commun. 13 , 1-16. (10.1038/s41467-022-34010-4)34983933
20. Rawlence NJ, Lowe DJ, Wood JR, Young JM, Churchman GJ, Huang Y-T, Cooper A. 2014 Using palaeoenvironmental DNA to reconstruct past environments: progress and prospects. J. Quat. Sci. 29 , 610-626. (10.1002/jqs.2740)
21. Lang G, Ammann B, Behre K-E, Tinner W. 2023 Quaternary vegetation dynamics of Europe. Bern, Switzerland: Haupt Verlag.
22. Capo E et al. 2021 Lake sedimentary DNA research on past terrestrial and aquatic biodiversity: overview and recommendations. Quaternary 4 , 6. (10.3390/quat4010006)
23. Giguet-Covex C et al. 2019 New insights on lake sediment DNA from the catchment: importance of taphonomic and analytical issues on the record quality. Sci. Rep. 9 , 14676. (10.1038/s41598-019-50339-1)31604959
24. Brown T et al. 2022 Paleoeconomy more than demography determined prehistoric human impact in Arctic Norway. PNAS Nexus 1 , gac209. (10.1093/pnasnexus/pgac209)
25. Alsos IG, Lammers Y, Yoccoz NG, Jørgensen T, Sjögren P, Gielly L, Edwards ME. 2018 Plant DNA metabarcoding of lake sediments: how does it represent the contemporary vegetation. PLoS ONE 13 , e0195403. (10.1371/journal.pone.0195403)29664954
26. Sjögren P, Edwards ME, Gielly L, Langdon CT, Croudace IW, Merkel MKF, Fonville T, Alsos IG. 2017 Lake sedimentary DNA accurately records 20th century introductions of exotic conifers in Scotland. New Phytol. 213 , 929-941. (10.1111/nph.14199)27678125
27. Courtin J, Perfumo A, Andreev AA, Opel T, Stoof-Leichsenring KR, Edwards ME, Murton JB, Herzschuh U. 2022 Pleistocene glacial and interglacial ecosystems inferred from ancient DNA analyses of permafrost sediments from Batagay megaslump, East Siberia. Environ. DNA 4 , 1265-1283. (10.1002/edn3.336)
28. Murchie TJ et al. 2021 Optimizing extraction and targeted capture of ancient environmental DNA for reconstructing past environments using the PalaeoChip Arctic-1.0 bait-set. Quat. Res. 99 , 305-328. (10.1017/qua.2020.59)
29. Murchie TJ, Karpinski E, Eaton K, Duggan AT, Baleka S, Zazula G, Macphee RDE, Froese D, Poinar HN. 2022 Pleistocene mitogenomes reconstructed from the environmental DNA of permafrost sediments. Curr. Biol. 32 , 851-860. (10.1016/j.cub.2021.12.023)35016010
30. Revéret A, Rijal DP, Heintzman PD, Brown AG, Stoof-Leichsenring KR, Alsos IG. 2023 Environmental DNA of aquatic macrophytes: the potential for reconstructing past and present vegetation and environments. Freshw. Biol 68 , 1929-1950. (10.1111/fwb.14158)
31. Pedersen MW, Ginolhac A, Orlando L, Olsen J, Andersen K, Holm J, Funder S, Willerslev E, Kjær KH. 2013 A comparative study of ancient environmental DNA to pollen and macrofossils from lake sediments reveals taxonomic overlap and additional plant taxa. Quat. Sci. Rev. 75 , 161-168. (10.1016/j.quascirev.2013.06.006)
32. Parducci L, Väliranta M, Salonen JS, Ronkainen T, Matetovici I, Fontana SL, Eskola T, Sarala P, Suyama Y. 2015 Proxy comparison in ancient peat sediments: pollen, macrofossil and plant DNA. Phil. Trans. R. Soc. B 370 , 20130382. (10.1098/rstb.2013.0382)25487333
33. Garcés-Pastor S et al. 2021 High resolution ancient sedimentary DNA shows that regional alpine plant diversity is a result of human land use. Res. Square 13 , 6559. (10.21203/rs.3.rs-908129/v1)
34. Liu S, Stoof-Leichsenring KR, Kruse S, Pestryakova LA, Herzschuh U. 2020 Holocene vegetation and plant diversity changes in the north-eastern Siberian treeline region from pollen and sedimentary ancient DNA. Front. Ecol. Evol. 8 , 560243. (10.3389/fevo.2020.560243)
35. Boyer F, Mercier C, Bonin A, Le Bras Y, Taberlet P, Coissac E. 2016 OBITOOLS: a unix-inspired software package for DNA metabarcoding. Mol. Ecol. Resour. 16 , 176-182. (10.1111/1755-0998.12428)25959493
36. Zinger L et al. 2019 DNA metabarcoding—need for robust experimental designs to draw sound ecological conclusions. Mol. Ecol. 28 , 1857-1862. (10.1111/mec.15060)31033079
37. Ficetola GF, Taberlet P, Coissac E. 2016 How to limit false positives in environmental DNA and metabarcoding? Mol. Ecol. Resour. 16 , 604-607. (10.1111/1755-0998.12508)27062589
38. Alsos IG et al. 2020 The treasure vault can be opened: large-scale genome skimming works well using herbarium and silica gel dried material. Plants 9 , 432. (10.3390/plants9040432)32244605
39. Nichols RV, Vollmers C, Newsom LA, Wang Y, Heintzman PD, Leighton M, Green RE, Shapiro B. 2018 Minimizing polymerase biases in metabarcoding. Mol. Ecol. Resour. 18 , 927-939. (10.1111/1755-0998.12895)
40. Ficetola GF et al. 2015 Replication levels, false presences and the estimation of the presence/absence from eDNA metabarcoding data. Mol. Ecol. Resour. 15 , 543-556. (10.1111/1755-0998.12338)25327646
41. Rijal DP et al. 2021 Sedimentary ancient DNA shows terrestrial plant richness continuously increased over the Holocene in northern Fennoscandia. Sci. Adv. 7 , eabf9557. (10.1126/sciadv.abf9557)34330702
42. Williams JW et al. 2023 Strengthening global-change science by integrating aeDNA with paleoecoinformatics. Trends Ecol. Evol. 38 , 946-960. (10.1016/j.tree.2023.04.016)37230884
43. Von Eggers J, Monchamp M-E, Capo E, Giguet-Covex C, Spanbauer T, Heintzman PD. 2022 Inventory of ancient environmental DNA from sedimentary archives: locations, methods, and target taxa. Zenodo. (10.5281/zenodo.6847522)
44. Nota K. 2022 Palaeoenvironmental reconstruction from modern and ancient plant DNA. PhD thesis, University of Uppsala.
45. Tabares X et al. 2020 Vegetation state changes in the course of shrub encroachment in an African savanna since about 1850 CE and their potential drivers. Ecol. Evol. 10 , 962-979. (10.1002/ece3.5955)32015858
46. Boessenkool S, Mcglynn G, Epp LS, Taylor D, Pimentel M, Gizaw A, Nemomissa S, Brochmann C, Popp M. 2014 Use of ancient sedimentary DNA as a novel conservation tool for high-altitude tropical biodiversity. Conserv. Biol. 28 , 446-455. (10.1111/cobi.12195)24372820
47. Willerslev E et al. 2014 Fifty thousand years of Arctic vegetation and megafaunal diet. Nature 506 , 47-51. (10.1038/nature12921)24499916
48. Jia W et al. 2022 Sedimentary ancient DNA reveals past ecosystem and biodiversity changes on the Tibetan Plateau: overview and prospects. Quat. Sci. Rev. 293 , 107703. (10.1016/j.quascirev.2022.107703)
49. Brown AG et al. 2022 New integrated molecular approaches for investigating lake settlements in north-western Europe. Antiquity 96 , 1179-1199. (10.15184/aqy.2022.70)
50. Brown AG et al. 2021 Ancient DNA, lipid biomarkers and palaeoecological evidence reveals construction and life on early medieval lake settlements. Sci. Rep. 11 , 1-13. (10.1038/s41598-020-79139-8)33414495
51. Alsos IG et al. 2021 Ancient sedimentary DNA shows rapid post-glacial colonisation of Iceland followed by relatively stable vegetation until the Norse settlement (Landnám) AD 870. Quat. Sci. Rev. 259 , 106903. (10.1016/j.quascirev.2021.106903)
52. Alsos IG et al. 2020 Last Glacial Maximum environmental conditions at Andøya, northern Norway; evidence for a northern ice-edge ecological ‘hotspot’. Quat. Sci. Rev. 239 , 106364. (10.1016/j.quascirev.2020.106364)
53. Crump SE, Miller GH, Power M, Sepúlveda J, Dildar N, Coghlan M, Bunce M. 2019 Arctic shrub colonization lagged peak postglacial warmth: molecular evidence in lake sediment from Arctic Canada. Glob. Chang. Biol. 25 , 4244-4256. (10.1111/gcb.14836)31603617
54. Karger DN, Nobis MP, Normand S, Graham CH, Zimmermann NE. 2021 CHELSA-TraCE21k v1.0. Downscaled transient temperature and precipitation data since the last glacial maximum. Clim. Past Discuss. 19 , 1-27. (10.5194/cp-2021-30)
55. Andersen KK et al. 2004 High-resolution record of Northern Hemisphere climate extending into the last interglacial period. Nature 431 , 147-151. (10.1038/nature02805)15356621
56. Meyer H, Kostrova SS, Meister P, Lenz MM, Kuhn G, Nazarova L, Syrykh LS, Dvornikov Y. 2022 Lacustrine diatom oxygen isotopes as palaeo precipitation proxy—Holocene environmental and snowmelt variations recorded at Lake Bolshoye Shchuchye, Polar Urals, Russia. Quat. Sci. Rev. 290 , 107620. (10.1016/j.quascirev.2022.107620)
57. Vystavna Y, Matiatos I, Wassenaar LI. 2021 Temperature and precipitation effects on the isotopic composition of global precipitation reveal long-term climate dynamics. Sci. Rep. 11 , 1-9. (10.1038/s41598-021-98094-6)33414495
58. Brown KA, Bunting MJ, Carvalho F, De Bello F, Mander L, Marcisz K, Mottl O, Reitalu T, Svenning J-C. 2023 Trait-based approaches as ecological time machines: developing tools for reconstructing long-term variation in ecosystems. Funct. Ecol. 37 , 2552-2569. (10.1111/1365-2435.14415)
59. Scheffer M et al. 2009 Early-warning signals for critical transitions. Nature 461 , 53-59. (10.1038/nature08227)19727193
60. Bascompte J, Jordano P. 2007 Plant-animal mutualistic networks: the architecture of biodiversity. Annu. Rev. Ecol. Evol. Syst. 38 , 567-593. (10.1146/annurev.ecolsys.38.091206.095818)
61. Michalet R, Brooker RW, Cavieres LA, Kikvidze Z, Lortie CJ, Pugnaire FI, Valiente-Banuet A, Callaway RM. 2006 Do biotic interactions shape both sides of the humped-back model of species richness in plant communities? Ecol. Lett. 9 , 767-773. (10.1111/j.1461-0248.2006.00935.x)16796565
62. Giguet-Covex C et al. 2014 Long livestock farming history and human landscape shaping revealed by lake sediment DNA. Nat. Commun. 5 , 3211. (10.1038/ncomms4211)24487920
63. Lammers Y et al. 2019 Clitellate worms (Annelida) in lateglacial and Holocene sedimentary DNA records from the Polar Urals and northern Norway. Boreas 48 , 317-329. (10.1111/bor.12363)
64. Seeber PA, Epp LS. 2022 Environmental DNA and metagenomics of terrestrial mammals as keystone taxa of recent and past ecosystems. Mamm. Rev. 52 , 538-553. (10.1111/mam.12302)
65. Seeber PA, Von Hippel B, Kauserud H, Löber U, Stoof-Leichsenring KR, Herzschuh U, Epp LS. 2022 Evaluation of lake sedimentary ancient DNA metabarcoding to assess fungal biodiversity in Arctic paleoecosystems. Environ. DNA 4 , 1150-1163. (10.1002/edn3.315)
66. Stoof-Leichsenring KR, Pestryakova LA, Epp LS, Herzschuh U. 2020 Phylogenetic diversity and environment form assembly rules for Arctic diatom genera—a study on recent and ancient sedimentary DNA. J. Biogeogr 47 , 1166-1179. (10.1111/jbi.13786)
67. Ibrahim A, Höckendorff S, Schleheck D, Epp L, Van Kleunen M, Meyer A. 2022 Vegetation changes over the last centuries in the Lower Lake Constance region reconstructed from sediment-core environmental DNA. Environ. DNA 4 , 830-845. (10.1002/edn3.292)
68. Elliott LD, Rijal DP, Brown AG, Bakke J, Topstad L, Heintzman PD, Alsos IG. 2023 Sedimentary ancient DNA reveals local vegetation changes driven by glacial activity and climate. Quaternary 6 , 7. (10.3390/quat6010007)
69. Tyler T, Herbertsson L, Olofsson J, Olsson PA. 2021 Ecological indicator and traits values for Swedish vascular plants. Ecol. Indic. 120 , 106923. (10.1016/j.ecolind.2020.106923)
70. Chalmandrier L, Hartig F, Laughlin DC, Lischke H, Pichler M, Stouffer DB, Pellissier L. 2021 Linking functional traits and demography to model species-rich communities. Nat. Commun. 12 , 1-9. (10.1038/s41467-021-22630-1)33397941
71. Herzschuh U. 2020 Legacy of the Last Glacial on the present-day distribution of deciduous versus evergreen boreal forests. Glob. Ecol. Biogeogr. 29 , 198-206. (10.1111/geb.13018)
72. Stewart L et al. 2016 The regional species richness and genetic diversity of Arctic vegetation reflect both past glaciations and current climate. Glob. Ecol. Biogeogr. 25 , 430-442. (10.1111/geb.12424)
73. Douda J, Havrdová A, Doudová J, Mandák B. 2018 Legacy of post-glacial colonization affects β-diversity: insights into local community assembly processes. J. Biogeogr. 45 , 1604-1615. (10.1111/jbi.13235)
74. Shipley B, Vile D, Garnier É. 2006 From plant traits to plant communities: a statistical mechanistic approach to biodiversity. Science 314 , 812-814. (10.1126/science.1131344)17023613
75. Bjune AE et al. 2021 Rapid climate changes during the Lateglacial and the early Holocene as seen from plant community dynamics in the Polar Urals, Russia. J. Quat. Sci. 37 , 805-817. (10.1002/jqs.3352)
76. Doncaster CP, Edwards ME, Clarke CL, Alsos IG. 2023 The drivers of plant community composition have shifted from external to internal processes over the past 20,000 years. Commun. Earth Environ. 4 , 1-12. (10.1038/s43247-023-00834-1)37325084
77. Clarke CL et al. 2019 Persistence of arctic-alpine flora during 24,000 years of environmental change in the Polar Urals. Sci. Rep. 9 , 19613. (10.1038/s41598-019-55989-9)31873100
78. Ripple WJ, Beschta RL. 2012 Trophic cascades in Yellowstone: the first 15 years after wolf reintroduction. Biol. Conserv. 145 , 205-213. (10.1016/j.biocon.2011.11.005)
79. Grudzinski BP et al. 2022 A global review of beaver dam impacts: stream conservation implications across biomes. Glob. Ecol. Conserv. 37 , e02163. (10.1016/j.gecco.2022.e02163)
80. Huntley B, Birks HJB. 1983 Pollen maps for Europe. Cambridge, UK: Cambridge University Press.
81. Douda J, Doudová J, Drašnarová A, Kuneš P, Hadincová V, Krak K, Zákravský P, Mandák B. 2014 Migration patterns of subgenus Alnus in Europe since the Last Glacial Maximum: a systematic review. PLoS ONE 9 , e88709. (10.1371/journal.pone.0088709)24586374
82. Revéret A, Rijal DP, Heintzman PD, Brown AG, Stoof-Leichsenring KR, Alsos IG. 2023 Environmental DNA of aquatic macrophytes: the potential for reconstructing past and present vegetation and environments. bioRxiv 533457. (10.1101/2023.03.27.533457)
83. Coelho MTP, Diniz-Filho JA, Rangel TF. 2019 A parsimonious view of the parsimony principle in ecology and evolution. Ecography 42 , 968-976. (10.1111/ecog.04228)
84. Elith J et al. 2006 Novel methods improve prediction of species' distributions from occurrence data. Ecography 29 , 129-151. (10.1111/j.2006.0906-7590.04596.x)
85. Davis AJ, Jenkinson LS, Lawton JH, Shorrocks B, Wood S. 1998 Making mistakes when predicting shifts in species range in response to global warming. Nature 391 , 783-786. (10.1038/35842)9486646
86. Dormann CF. 2007 Promising the future? Global change projections of species distributions. Basic Appl. Ecol. 8 , 387-397. (10.1016/j.baae.2006.11.001)
87. Pilowsky JA, Colwell RK, Rahbek C, Fordham DA. 2022 Process-explicit models reveal the structure and dynamics of biodiversity patterns. Sci. Adv. 8 , eabj2271. (10.1126/sciadv.abj2271)35930641
88. Rangel TF, Edwards NR, Holden PB, Diniz-Filho JAF, Gosling WD, Coelho MTP, Cassemiro FAS, Rahbek C, Colwell RK. 2018 Modeling the ecology and evolution of biodiversity: biogeographical cradles, museums, and graves. Science 361 , eaar5452. (10.1126/science.aar5452)30026200
89. Urban MC. 2015 Accelerating extinction risk from climate change. Science 348 , 571-573. (10.1126/science.aaa4984)25931559
90. Alsos IG, Alm T, Normand S, Brochmann C. 2009 Past and future range shift and loss of genetic diversity in dwarf willow (Salix herbacea L.) inferred from genetics, fossils, and modelling. Glob. Ecol. Biogeogr. 18 , 223-239. (10.1111/j.1466-8238.2008.00439.x)
91. Veloz SD, Williams JW, Blois JL, He F, Otto-Bliesner B, Liu Z. 2012 No-analog climates and shifting realized niches during the Late Quaternary: implications for 21st-century predictions by species distribution models. Glob. Chang. Biol. 18 , 1698-1713. (10.1111/j.1365-2486.2011.02635.x)
92. Epp LS, Kruse S, Kath NJ, Stoof-Leichsenring KR, Tiedemann R, Pestryakova LA, Herzschuh U. 2018 Temporal and spatial patterns of mitochondrial haplotype and species distributions in Siberian larches inferred from ancient environmental DNA and modeling. Sci. Rep. 8 , 1-9. (10.1038/s41598-018-35550-w)29311619
93. Liu S, Kruse S, Scherler D, Ree RH, Zimmermann HH, Stoof-Leichsenring KR, Epp LS, Mischke S, Herzschuh U. 2021 Sedimentary ancient DNA reveals a threat of warming-induced alpine habitat loss to Tibetan Plateau plant diversity. Nat. Commun. 12 , 2995. (10.1038/s41467-021-22986-4)34016962
94. Champagnat N, Ferrière R, Méléard S. 2006 Unifying evolutionary dynamics: from individual stochastic processes to macroscopic models. Theor. Popul. Biol. 69 , 297-321. (10.1016/j.tpb.2005.10.004)16460772
95. Hartig F, Dyke J, Hickler T, Higgins SI, O'Hara RB, Scheiter S, Huth A. 2012 Connecting dynamic vegetation models to data—an inverse perspective. J. Biogeogr. 39 , 2240-2252. (10.1111/j.1365-2699.2012.02745.x)
96. Boussange V. 2022 Forward and inverse modelling of eco-evolutionary dynamics in ecological and economic systems. PhD thesis, ETH Zurich.
97. Skeels A, Bach W, Hagen O, Jetz W, Pellissier L. 2022 Temperature-dependent evolutionary speed shapes the evolution of biodiversity patterns across tetrapod radiations. Syst. Biol. 72 , 341-356. (10.1093/sysbio/syac048)
98. Boussange V, Aceituno PV, Pellissier L. 2022 Mini-batching ecological data to improve ecosystem models with machine learning. bioRxiv 501365. (10.1101/2022.07.25.501365)
99. Blei DM, Kucukelbir A, Mcauliffe JD. 2017 Variational inference: a review for statisticians. J. Am. Stat. Assoc. 112 , 859-877. (10.1080/01621459.2017.1285773)
100. Borisov I, Metelkin E. 2020 Confidence intervals by constrained optimization—an algorithm and software package for practical identifiability analysis in systems biology. PLoS Comput. Biol. 16 , e1008495. (10.1371/journal.pcbi.1008495)33347435
101. Remien CH, Eckwright MJ, Ridenhour BJ. 2021 Structural identifiability of the generalized Lotka–Volterra model for microbiome studies. R. Soc. Open Sci. 8 , 201378. (10.1098/rsos.201378)34295510
102. Choy SL, O'Leary R, Mengersen K. 2009 Elicitation by design in ecology: using expert opinion to inform priors for Bayesian statistical models. Ecology 90 , 265-277. (10.1890/07-1886.1)19294931
103. Lemoine NP. 2019 Moving beyond noninformative priors: why and how to choose weakly informative priors in Bayesian analyses. Oikos 128 , 912-928. (10.1111/oik.05985)
104. Johnson JB, Omland KS. 2004 Model selection in ecology and evolution. Trends Ecol. Evol. 19 , 101-108. (10.1016/j.tree.2003.10.013)16701236
105. Franklin J, Serra-Diaz JM, Syphard AD, Regan HM. 2016 Global change and terrestrial plant community dynamics. Proc. Natl Acad. Sci. USA 113 , 3725-3734. (10.1073/pnas.1519911113)26929338
106. Perfumo A, Çabuk U, Schulte L, Courtin J, Harms L, Stoof-Leichsenring KR, Herzschuh U. 2023 Paleometagenomics reveals environmental microbiome response to vegetation changes in northern Siberia over the millennia. Environ. DNA 5 , 1252-1264. (10.1002/edn3.446)
107. Alsos IG, Boussange V, Rijal DP, Beaulieu M, Brown AG, Herzschuh U, Svenning J-C, Pellissier L. 2024 Using ancient sedimentary DNA to forecast ecosystem trajectories under climate change. Figshare. (10.6084/m9.figshare.c.7093279)
