Florian Hartig

← Publications

Productivity of Fagus sylvatica under climate change – A Bayesian analysis of risk and uncertainty using the model 3-PG

Andrey L.D. Augustynczik, Florian Hartig, Francesco Minunno, Hans-Peter Kahle, Daniela Diaconu, Marc Hanewinkel & Rasoul Yousefpour

Forest Ecology and Management, 401, 192–206 (2017)
Cite this
@article{augustynczik2017productivity,
  author = {Andrey L.D. Augustynczik and Florian Hartig and Francesco Minunno and Hans-Peter Kahle and Daniela Diaconu and Marc Hanewinkel and Rasoul Yousefpour},
  title = {Productivity of Fagus sylvatica under climate change – A Bayesian analysis of risk and uncertainty using the model 3-PG},
  journal = {Forest Ecology and Management},
  volume = {401},
  pages = {192--206},
  year = {2017},
  doi = {10.1016/j.foreco.2017.06.061},
}

DOI: 10.1016/j.foreco.2017.06.061
Citation statistics: 58 (Google Scholar) · 60 (OpenAlex), as of 29 October 2026

Google Scholar citations per year, as of 21 October 2026 (yearly counts are fetched separately from the total)

Funding

This study was supported through funding by German Research Foundation; ConFoBi project (GRK 2123); BMEL; BMUB; BuTaKli project (28WC4069).

View article (DOI)

Abstract

To assess the long-term impacts of forest management interventions under climate change, process-based models, which allow to predict transient dynamics under environmental change, are arguably the most suitable tools available. A challenge for using these models for management decisions, however, is their higher parametric uncertainty, which propagates to predictions and thus into the decision-making process. Here, we demonstrate how this problem can be addressed through Bayesian inference. We first conduct a Bayesian calibration to generate an estimate of posterior parametric uncertainty for the process-based forest growth model 3-PG for Fagus sylvatica. The calibration uses data from twelve sites in Germany, together with a robust (Student’s t) error model. We then propagate the estimated uncertainty together with economic uncertainty to forest productivity and Land Expectation Value (LEV), allowing us to evaluate alternative management regimes under climate change. Our results demonstrate that parametric and economic uncertainty have strong impacts on the variation of predicted forest productivity and profitability. Management regimes with increased thinning intensity were overall most robust to economic, climate change and parametric model uncertainty. We conclude that estimating and propagating economic and model uncertainty is crucial for developing robust adaptive management strategies for forests under climate change.

What the paper shows and why it matters

(AI-generated)

Process-based forest models, which simulate tree growth from physiology and climate, are arguably the most suitable tools for judging how forest management will fare under climate change, because they can follow change over time. Many of their numerous parameters, however, are not well known, and this parametric uncertainty carries through into predictions and decisions. According to the paper, this uncertainty is usually neglected in forest planning, as are uncertainty in future climate and in timber prices and interest rates. The problem the paper takes on is whether the best result of a single simulation with fixed (median) parameter values is a sound basis for choosing management when so much remains uncertain.

The study argues that Bayesian inference offers a coherent way to estimate parametric uncertainty and carry it, together with climate and economic uncertainty, into a risk analysis that identifies management regimes that remain robust. As a demonstration, the growth model 3-PG (Landsberg and Waring 1997) is calibrated for European beech (Fagus sylvatica) with data from twelve German plots, using an error model that limits the influence of outliers. The calibrated uncertainty is then run through 20 management regimes under 12 climate trajectories for a site in Baden-Württemberg, with uncertain timber prices and interest rates. Regimes are judged by Land Expectation Value, the economic value of the forest land, and by its Value-at-Risk, the value exceeded in 95% of simulated cases. At interest rates of 1–3%, the most robust regime combined increased thinning with a 160-year rotation. Ignoring parametric and economic uncertainty would instead have picked regimes with shorter rotations (120–140 years), whose Value-at-Risk was 29–82% lower, meaning a higher risk of poor returns. With economic uncertainty alone, a different regime would have counted as robust; adding parametric uncertainty changed the choice.

The paper reads these results as showing that neglecting parametric, climate and economic uncertainty can lead to risky management choices, and presents the approach as a blueprint for including such uncertainty in forest planning. It supplies a thinned sample of the calibrated parameters so that later studies can reuse it as a starting point when new data arrive. Limits remain: regimes were judged by profit alone, disturbances such as pests were not included, and wide priors were used because the authors knew of no earlier calibration of 3-PG for beech.

References

  • Landsberg & Waring (1997). A generalised model of forest productivity using simplified concepts of radiation-use efficiency, carbon balance and partitioning. Forest Ecology and Management 95, 209–228. doi:10.1016/s0378-1127(97)00026-1