Periodic Reporting for period 1 - BEAST (Biodiversity dynamics across a continuum of space, time, and their scales)
Berichtszeitraum: 2023-01-01 bis 2025-06-30
We face an unprecedented threat from global alteration of nature and biodiversity, but we still lack rigorous estimates of how fast, where, and at which scales biodiversity changes. Studies report fragmented and seemingly contradictory results, suffer from mismatches in biodiversity metrics, mismatches in temporal and spatial grains, and are constrained by huge data gaps. Moreover, local loss and gain of biodiversity is decoupled from changes in countries or continents, with opposing directions at different scales being plausible. A quantitative synthesis that connects all this, and bridges the gaps, is needed.
The objective of BEAST is to map and interpolate temporal biodiversity change in Europe, the US, and the world, across continuous space, time, and their grains, from locations as small as 1 m, to countries and continents, over the last ca 40 years, for birds, plants, and butterflies. To do this we will combine data from local time series with high-quality gridded atlas data from countries and continents. We will use a new cross-scale model to interpolate biodiversity change jointly across space and time, and across the data gaps. We will test if temporal change of diversity, distributions, and turnover can be estimated from: (i) static patterns of diversity and distributions, (ii) from data lacking temporal replication, (iii) from space-for-time substitution of spatial vs temporal species turnover, (iii) from spaceborne remotely sensed spectral diversity and turnover.
These methods will enable integration of heterogeneous and messy biodiversity data, and they will improve estimates of change in data-poor regions of the global South. BEAST will deliver the first integrative statistical model revealing, for the first time, how multiple facets of biodiversity change across scales. It will show which regions, habitats, and biomes undergo the most pronounced change, which is critical for informed large-scale conservation policy.
The objective of BEAST is to map and interpolate temporal biodiversity change in Europe, the US, and the world, across continuous space, time, and their grains, from locations as small as 1 m, to countries and continents, over the last ca 40 years, for birds, plants, and butterflies. To do this we will combine data from local time series with high-quality gridded atlas data from countries and continents. We will use a new cross-scale model to interpolate biodiversity change jointly across space and time, and across the data gaps. We will test if temporal change of diversity, distributions, and turnover can be estimated from: (i) static patterns of diversity and distributions, (ii) from data lacking temporal replication, (iii) from space-for-time substitution of spatial vs temporal species turnover, (iii) from spaceborne remotely sensed spectral diversity and turnover.
These methods will enable integration of heterogeneous and messy biodiversity data, and they will improve estimates of change in data-poor regions of the global South. BEAST will deliver the first integrative statistical model revealing, for the first time, how multiple facets of biodiversity change across scales. It will show which regions, habitats, and biomes undergo the most pronounced change, which is critical for informed large-scale conservation policy.
These are our achievements during the first two years of BEAST:
*Biodiversity data*
- We have amassed a new global database comprising of temporally replicated gridded distributional atlases. These are the reference high-quality data and a cornerstone of BEAST. We have managed to fit all of them in a single database framework so that they can be analysed jointly. Most of them cover the period of past 50 years, and focus on birds, mammals, and butterflies. The data are now ready.
- We initiated work on a database that can store local and regional species inventories. A first outline of the database structure, together with example regional data, has been released. We are now testing if the use of such data in macroecological analyses is on par with the high-quality gridded data.
- We started collecting data on locally and regionally threatened and extinct species.
*Spatio-temporal cross-scale interpolation*
- An important assumption of the proposed interpolation method is the clumping of species distributions (and diversity) in space and time. We tested this assumption by quantifying the magnitude of autocorrelation in empirical data (i.e. the gridded atlases) across scales, and we examined how this autocorrelation changes in time. We found that the autocorrelation is generally robust to variation in the grain of the data, and we found a systematic relationship between dynamics of the autocorrelation and dynamics of occupancy.
- We tested a machine-learning version of the interpolation approach to account for varying sampling effort and sample area in a country-wide dataset (Czech birds), and to examine patterns of biodiversity change across scales. An important achievement of this analysis is a proposed mechanisms for how different modes of colonization and extinction affect biodiversity change across spatial scales.
*Linking biodiversity dynamics to static patterns*
- We have performed a first large-scale test of the proposition that specific temporal processes leave characteristic imprints in static snapshot of species distributions. We found that static characteristics of species distributions do not predict change of occupancy well, but they strongly predict turnover of sites (i.e. overall magnitude of species dynamics).
- We described the first mechanistic link between rarity (a static property) or a species and its probability of extinction at multiple spatial scales.
*Remotely-sensed biodiversity change*
- We collaborated on a study of the relationship between spectral diversity of landscapes and bird diversity.
- We have been reviewing the literature on temporal dynamics of diversity in remotely sensed metrics (such as spectral diversity), and on metrics from landscape ecology.
*Synthesis and integrated cross-scale predictions*
- Using a new class of Bayesian occupancy models, we integrated multiple lines of evidence (3 types of data) to assess continental range dynamics and biodiversity dynamics across the entire Latin America.
*Biodiversity data*
- We have amassed a new global database comprising of temporally replicated gridded distributional atlases. These are the reference high-quality data and a cornerstone of BEAST. We have managed to fit all of them in a single database framework so that they can be analysed jointly. Most of them cover the period of past 50 years, and focus on birds, mammals, and butterflies. The data are now ready.
- We initiated work on a database that can store local and regional species inventories. A first outline of the database structure, together with example regional data, has been released. We are now testing if the use of such data in macroecological analyses is on par with the high-quality gridded data.
- We started collecting data on locally and regionally threatened and extinct species.
*Spatio-temporal cross-scale interpolation*
- An important assumption of the proposed interpolation method is the clumping of species distributions (and diversity) in space and time. We tested this assumption by quantifying the magnitude of autocorrelation in empirical data (i.e. the gridded atlases) across scales, and we examined how this autocorrelation changes in time. We found that the autocorrelation is generally robust to variation in the grain of the data, and we found a systematic relationship between dynamics of the autocorrelation and dynamics of occupancy.
- We tested a machine-learning version of the interpolation approach to account for varying sampling effort and sample area in a country-wide dataset (Czech birds), and to examine patterns of biodiversity change across scales. An important achievement of this analysis is a proposed mechanisms for how different modes of colonization and extinction affect biodiversity change across spatial scales.
*Linking biodiversity dynamics to static patterns*
- We have performed a first large-scale test of the proposition that specific temporal processes leave characteristic imprints in static snapshot of species distributions. We found that static characteristics of species distributions do not predict change of occupancy well, but they strongly predict turnover of sites (i.e. overall magnitude of species dynamics).
- We described the first mechanistic link between rarity (a static property) or a species and its probability of extinction at multiple spatial scales.
*Remotely-sensed biodiversity change*
- We collaborated on a study of the relationship between spectral diversity of landscapes and bird diversity.
- We have been reviewing the literature on temporal dynamics of diversity in remotely sensed metrics (such as spectral diversity), and on metrics from landscape ecology.
*Synthesis and integrated cross-scale predictions*
- Using a new class of Bayesian occupancy models, we integrated multiple lines of evidence (3 types of data) to assess continental range dynamics and biodiversity dynamics across the entire Latin America.
So far, we have finished 2 years of BEAST. Tentatively, our first results have these implications for the assessment of the ongoing global biodiversity change:
(1) It is necessary to monitor temporal dynamics of multiple facets of biodiversity change jointly. We demonstrated (on an example of species occupancy and clumping) that a single biodiversity variable can, on average, show zero or no temporal trend, but when analyzed jointly with other variables there can be important systematic temporal shifts.
(2) It is necessary to consider spatial scale of an analysis. We showed that there can be zero average change of biodiversity at one spatial resolution, but an increase at other resolution even within the same region. This has rarely been done. Further, we have identified mechanisms behind this phenomenon, namely the density-dependent per-capita death rate, and the specific models of colonization and extinction.
(3) There are more aspects of biodiversity that we need to monitor, but we don’t. For instance, we showed that not only species range (or occupancy), but also its spatial clumping, can show conspicuous temporal trends.
(4) We need to combine multiple lines of evidence to show trends. We showed that this is particularly useful when any single line of evidence is insufficient; for instance, in the case of temporal dynamics of diversity of Latin-American carnivores, we showed that meaningful trends can be estimated from a combination of data from camera traps, opportunistic citizen science records, and expert range maps. We now provide new statistical methods to do this.
(5) So far, our results tentatively indicate that biodiversity can be interpolated jointly in space and time. If this is indeed the case (and we will have a better idea in the later stages of BEAST) then we will have a powerful tool to monitor biodiversity change in regions without temporally replicated surveys.
(6) Our results indicate that we may not be able to predict if species are going to expand, or decline, just from static data. We may, however, be able to predict some unexpected aspects of range dynamics, namely the overall turnover of lost and gained sites. More research is needed here.
(1) It is necessary to monitor temporal dynamics of multiple facets of biodiversity change jointly. We demonstrated (on an example of species occupancy and clumping) that a single biodiversity variable can, on average, show zero or no temporal trend, but when analyzed jointly with other variables there can be important systematic temporal shifts.
(2) It is necessary to consider spatial scale of an analysis. We showed that there can be zero average change of biodiversity at one spatial resolution, but an increase at other resolution even within the same region. This has rarely been done. Further, we have identified mechanisms behind this phenomenon, namely the density-dependent per-capita death rate, and the specific models of colonization and extinction.
(3) There are more aspects of biodiversity that we need to monitor, but we don’t. For instance, we showed that not only species range (or occupancy), but also its spatial clumping, can show conspicuous temporal trends.
(4) We need to combine multiple lines of evidence to show trends. We showed that this is particularly useful when any single line of evidence is insufficient; for instance, in the case of temporal dynamics of diversity of Latin-American carnivores, we showed that meaningful trends can be estimated from a combination of data from camera traps, opportunistic citizen science records, and expert range maps. We now provide new statistical methods to do this.
(5) So far, our results tentatively indicate that biodiversity can be interpolated jointly in space and time. If this is indeed the case (and we will have a better idea in the later stages of BEAST) then we will have a powerful tool to monitor biodiversity change in regions without temporally replicated surveys.
(6) Our results indicate that we may not be able to predict if species are going to expand, or decline, just from static data. We may, however, be able to predict some unexpected aspects of range dynamics, namely the overall turnover of lost and gained sites. More research is needed here.