Periodic Reporting for period 2 - HOPE (Humans On Planet Earth - Long-term impacts on biosphere dynamics)
Reporting period: 2019-07-01 to 2020-12-31
HOPE has two main objectives.
1. Test the recently proposed hypothesis that prehistoric human impact about 6000 years ago altered the fundamental ecological processes that determine the assembly, ecological structure, and composition and dynamics of terrestrial plant communities. Its proponents argue the use of past ‘natural experiments’ and of the past as an analogue to our uncertain future are a “flawed strategy” and the basic concept behind Earth system science and all historical sciences, namely uniformitarianism (‘the present is the key to the past’) should be discarded. It is clearly imperative to test this hypothesis critically as it has far-reaching implications.
2. Test the associated hypothesis that inter-relationships (correlations) between estimates of ecosystem properties such as turnover, diversity, composition and rate-of-change (RoC) during the Holocene (last 11700 years) changed in response to prehistoric human activities.
Testing these two hypotheses is important for the modus operandi of Earth system science, global change biology and ecological and societal predictions as a whole. If the hypotheses are not falsified, the question immediately arises can the past can be used reliably as an analogue to the uncertain future of Planet Earth and its inhabitants in the so-called ‘Anthropocene’. If the hypotheses are not falsified, Earth today has no analogue in the past 11700 years and the widely employed research approach of using ‘natural experiments’ in the past as a guide to predict what might happen in the future is a “flawed strategy” and should be discontinued. As almost all predictive models in global-change science rely on this approach, it is urgent to test these hypotheses. This is the primary aim of HOPE.
HOPE’s research relies on dated pollen-stratigraphical data from lakes or mires in Europe, Asia, North & South America, Africa, Australasia, and oceanic islands. These data are explored in a consistent manner using state-of-the-art numerical techniques to discern patterns in over 20 ecosystem properties. Temporal, spatial and correlation patterns in these properties are compared statistically within each pollen sequence, between sequences within ecoregions (≡ biomes), between ecoregions within continents, and between ecoregions on different continents using statistical modelling techniques with human-impact events as predictors along with model-simulated climate variables.
(1) Development of a rigorous workflow in R to capture, evaluate, format, filter, and check all possible data-sets available in databases such as Neotoma and Pangea. So far about 500 pollen-sequences have been processed from Europe, about 800 from North America and about 150 from Asia. Work is in progress on expanding Asian coverage, in developing a Latin American pollen database (ca. 150 sequences), and in acquiring African and Australasian pollen data. Robust age-depth models using chronological controls for each sequence have been made using Bayesian modelling (Bchron) with 1000 runs to derive model uncertainties.
(2) Ecosystem properties such as diversity, compositional change and turnover have been estimated for the sequences. Properties still to be estimated are taxon-pair occurrences, functional and phylogenetic diversity and spatial beta-diversity. It is impossible to estimate biomass due to lack of enough data. Dark diversity is more difficult to estimate than originally envisaged and may be discarded. A new approach to estimate RoC (=temporal beta-diversity) from pollen data has been developed.
(3) Hypothesis testing has begun for Europe and Asia. Multivariate regression has been used to test the statistical significance of relationships between temporal changes in ecosystem properties (responses) and the human-impact events (predictors). Although preliminary, results suggest that hypothesis 1 is falsified for Europe and Asia. Work is currently refining our hypothesis-testing procedures, and on improving procedures for testing hypothesis 2.
1. Development of a complete R workflow for inputting basic pollen-stratigraphical data and associated meta-data from various sources, re-formatting and tabulating the data, applying quality-control criteria, developing Bayesian age-depth models, coding of human impact events, harmonisation of nomenclature between sequences within an area and final quality-control filtering (see figure) to produce the data used in numerical analysis.
2. An important ecosystem property is rate-of-change (RoC) analysis. This estimates the amount of pollen compositional change per unit time. This analysis, first presented in 1986, has a major limitation, namely are RoC peaks statistically significant or are they chance? HOPE has developed a statistically rigorous analysis that identifies peaks are not a result of chance, can find a lack of any significant peaks, and can be used to estimate and compare RoC at many sites in the same ecoregion or between ecoregions. This method will soon submitted for publication.
3. Within archaeology, a recent development has been mining large databases of radiocarbon dates associated with archaeological material to create summed probability distributions (SPD) as a proxy for human population density within an area of interest. HOPE has compiled a geo-referenced global database of over 500000 dates. Quality-control criteria are applied to filter out dates with large errors and the dates are calibrated into calibrated years before present (= 1950 CE). HOPE has developed a means of deriving an estimated human-population density time-series within a given radius of a particular pollen-sequence, within a given ecoregion, and between sequences and ecoregions. A computer-intensive numerical procedure has been developed to establish if peaks in the density time-series are statistically significant and not the result of chance. We generate by repeated bootstrapping a theoretical 95% confidence interval for the null model based the simulated time-series. SPD values outside the confidence interval are significant.
HOPE has published six papers in international scientific journals directly related to HOPE in terms of concepts, methods, or basic evaluation and given eight presentations about HOPE and its research at international or national symposia. In addition, the Bergen HOPE team has published 52 other international papers and given 29 presentations at conferences.