the two vignettes are hidden in the tar.gz that you can download from CRAN. Then go to TTR.PGM/inst/doc 3/3

if you are interested in using the process model for attribution or species distribution modelling, check out our R package: besjournals.onlinelibrary.wiley.com/doi/10.1111/... 2/3

Congratulations Jens! Truly deserved!

another very useful finding of this study imo is that process models parametrized inversely from species distribution data performed better than forward-parametrized process models, at least in this set of species with reasonable distribution data.

Okay ;) but if there IS a ‚correct‘ state and we fail to recognize it, as has been the case in the forest-savanna-bistability assertion, then we make badly informed management decisions.

forgot this review: nph.onlinelibrary.wiley.com/doi/10.1111/...

References (first 4 forest-savanna, last one lakes):
Higgins 2023: www.science.org/doi/10.1126/...
Mattos 2023: www.pnas.org/doi/abs/10.1...
Williamson 2024: iopscience.iop.org/article/10.1...
Zwaan 2024: iopscience.iop.org/article/10.1...
Davidson 2023 Lakes: www.nature.com/articles/s41...

whereas shorter time series suggest alternative biomass states are possible. I agree that the definition of desired ecosystem states is often debatable, but the use of alternative stable states theory to justify goals is problematic when alternative states are only apparent but not true. 7/7

A similar problem occurs when the observation time span in experiments is too short to allow deterministic end points to be reached. For example, when the monitoring time series of phytoplankton biomass is extended, it turns out that phytoplankton is linearly related to nutrient concentrations 6/7

A string of recent works shows that when these factors are considered and better data is used, the remaining overlap between forest and savannas is minimal, refuting the hypothesis of widespread bistability and pointing to a predictable endpoint in this system. I include some references below. 5/7

such as growing season length and intensity or topography-driven variation in soil hydrology. 4/7

The graph from the Pausas and Bond paper you posted is an example: It shows overlap along one variable (annual rainfall), but most people would agree that the distribution of forest and savannas is also influenced by other environmental factors 3/7

The problem is that we often misinterpret prediction uncertainty in ecological models as evidence for overlapping environmental conditions between two or more ecosystem states, when in fact this uncertainty can be greatly reduced by better models and data. 2/7

I would partly disagree. The evidence for multiple stable states is often not robust enough to use this framework for prescribing management actions. 1/7

Thank you!