publications

2025

1. Best Practices in Software Development for Robust and Reproducible Geoscientific Models Based on Insights from the Global Carbon Project Models (PREPRINT)

   Konstantin Gregor, Benjamin F. Meyer , Tillmann Gaida , and 29 more authors

   Jun 2025

   Abstract Link

   Computational models play an increasingly vital role in scientific research, by numerically simulating processes that cannot be solved analytically. Such models are fundamental in geosciences and offer critical insights into the impacts of global change on the Earth system today and in the future. Beyond their value as research tools, models are also software products and should therefore adhere to certain established software engineering standards. However, scientists are rarely trained as 5 software developers, which can lead to potential deficiencies in software quality like unreadable, inefficient, or erroneous code. The complexity of these models, coupled with their integration into broader workflows, also often makes reproducing results, evaluating processes, and building upon them highly challenging.

2. Simulating Lightning-Induced Tree Mortality in the Dynamic Global Vegetation Model \textscLPJ-\textscGUESS

   Andreas Krause , Konstantin Gregor, Benjamin F. Meyer , and 1 more author

   Global Change Biology, Jun 2025

   Abstract Link

   Lightning is an important yet often overlooked disturbance agent in forest ecosystems. Recent research conducted in Panama suggests that lightning is a major cause of large tree mortality in tropical forests. However, lightning-\-induced tree mortality is not included in state-\-of-\-the-\-art ecosystem models. Here, we implement a general lightning mortality module in the dynamic global vegetation model LPJ-\-GUESS to explore the impacts of lightning on forests at local and global scales. Lightning mortality was implemented stochastically in dependency of local cloud-\-to-\-ground lightning density and simulated forest structure based on findings from the Panamanian forest. For this site, LPJ-\-GUESS adequately simulates the average number of trees of different size classes killed per lightning strike, with a total of 2.9 simulated versus 3.2 observed. The model also captures the estimated contribution of lightning to the overall mortality of large trees (21% simulated vs. 24% observed). Applying the new model version to other tropical and temperate forests for which observation-\-based estimates on lightning mortality exist, LPJ-\-GUESS reproduces estimated impacts in some forests but simulates substantially lower impacts for others. Global simulations driven by two alternative products of cloud-\-to-\-ground lightning densities suggest that lightning kills 301–340 million trees annually, thereby causing 0.21–0.30 GtC yr.-1 of dead biomass (2.1%–2.9% of total killed biomass). The simulations also reveal that the global biomass would be 1.3%–1.7% higher in a world without lightning. Spatially, simulated lightning mortality is largest in the tropical forests of Africa. Although our simulations suggest an important role of lightning in forest ecosystems on a global scale, more data on lightning-\-induced tree mortality across different forest types would be desirable for more accurate model calibration and evaluation. Given the anticipated increase in future lightning activity, incorporating lightning mortality into ecosystem models is needed to obtain more reliable projections of terrestrial vegetation dynamics and carbon cycling.

3. Disentangling Future Effects of Climate Change and Forest Disturbance on Vegetation Composition and Land Surface Properties of the Boreal Forest

   Lucia S. Layritz , Konstantin Gregor, Andreas Krause , and 4 more authors

   Biogeosciences, Jul 2025

   Abstract Link

   Abstract. Forest disturbances can cause shifts in boreal vegetation cover from predominantly evergreen to deciduous trees or non-forest dominance. This, in turn, impacts land surface properties and, potentially, regional climate. Accurately considering such shifts in future projections of vegetation dynamics under climate change is crucial but hindered (e.g., uncertainties in future disturbance regimes). In this study, we investigate how sensitive future projections of boreal forest dynamics are to additional changes in disturbance regimes. We use the dynamic vegetation model LPJ-GUESS to investigate and disentangle the impacts of climate change and intensifying disturbance regimes in future projections of boreal vegetation cover as well as changes in land surface properties such as albedo and evapotranspiration. Our simulations find that (1) warming alone drives shifts towards more densely forested landscapes, (2) more intense disturbances reduce tree cover in favor of shrubs and grasses, and (3) the interaction between climate and disturbances leads to an expansion of deciduous trees. Our results additionally indicate that warming decreases albedo and increases evapotranspiration, while more intense disturbances have the opposite effect, potentially offsetting climate impacts. Warming and disturbances are thus comparably important agents of change in boreal forests. Our findings highlight future disturbance regimes as a key source of model uncertainty and underscore the necessity of accounting for disturbances-induced effects on vegetation composition and land surface–atmosphere feedback.

4. Simulating the Drought Response of European Tree Species with the Dynamic Vegetation Model LPJ-GUESS (v4.1, 97c552c5)

   Benjamin F. Meyer , João P. Darela-Filho , Konstantin Gregor, and 9 more authors

   Geoscientific Model Development, Jul 2025

   Abstract Link

   Abstract. Due to climate change, severe-drought events have become increasingly commonplace across Europe in recent decades, with future projections indicating that this trend will likely continue, posing questions about the continued viability of European forests. Observations from the most recent pan-European droughts suggest that these types of “hotter droughts” may acutely alter the carbon balance of European forest ecosystems. However, substantial uncertainty remains regarding the possible future impacts of severe drought on the European forest carbon sink. Dynamic vegetation models can help to shed light on such uncertainties; however, the inclusion of dedicated plant hydraulic architecture modules in these has only recently become more widespread. Such developments intended to improve model performance also tend to add substantial complexity, yet the sensitivity of the models to newly introduced processes is often left undetermined. Here, we describe and evaluate the recently developed mechanistic plant hydraulic architecture version of LPJ-GUESS and provide a parameterization for 12 common European forest tree species. We quantify the uncertainty introduced by the new processes using a variance-based global sensitivity analysis. Additionally, we evaluate the model against water and carbon fluxes from a network of eddy covariance flux sites across Europe. Our results indicate that the new model is able to capture drought-induced patterns of evapotranspiration along an isohydric gradient and manages to reproduce flux observations during drought better than standard LPJ-GUESS does. Further, the sensitivity analysis suggests that hydraulic process related to hydraulic failure and stomatal regulation play the largest roles in shaping the model response to drought.

5. Stakeholder Engagement for Inclusive Climate Impact Attribution Studies

   Albert Nkwasa , Inga Menke , Lisa Murken , and 18 more authors

   Environmental Research: Climate, Jan 2025

   Abstract Link

   Abstract Since its emergence in the 1990s, the science of attributing observed phenomena to human-induced and natural climate drivers has made remarkable progress. To ensure the relevance and uptake of climate impact attribution studies, scientists must effectively engage with stakeholders. This engagement allows stakeholders to pose key questions, which scientists can then substantiate with evidence evaluating the existence of causal links. Although significant advancements have been made in climate impact attribution science, much work remains to understand the varied requirements of different stakeholders for impact attribution findings. This perspective explores the usefulness of stakeholder engagement in climate impact attribution, the challenges it presents, and how it can be made more relevant for addressing societal questions. It advocates for prioritizing stakeholder involvement to achieve greater transparency, legitimacy, and practical application of findings. Such involvement can enhance the societal impact of attribution studies and support informed decision-making in the face of climate change.

2024

1. Saturating Response of Photosynthesis to Increasing Leaf Area Index Allows Selective Harvest of Trees without Affecting Forest Productivity

   Olivier Bouriaud , Ernst-Detlef Schulze , Konstantin Gregor, and 8 more authors

   Oct 2024

   Abstract Link

   Abstract. Maintaining or increasing forest carbon sinks is considered essential to mitigate the rise of atmospheric CO2 concentrations. Harvesting trees is perceived as having negative consequences on both the standing biomass stocks and the carbon sink strength. However, harvesting needs to be examined from a forest stand canopy perspective since carbon assimilation occurs in the canopy. Here we show that a threshold of leaf area exists beyond which additional leaves do not contribute to ecosystem fluxes. The associated biomass can be harvested without affecting the forest carbon fluxes. Based on eddy covariance measurements we show that CO2 uptake (GPP) and net ecosystem exchange (NEP) in temperate forests are of similar magnitude in both unmanaged and sustainably managed forests, in the order of 1500–1600 gC m-2 y-1 for GPP and 542 – 483 gC m-2 y-1 for NEP. A threshold of about 4 m2 m-2 LAI (leaf area index) can be used as a definition of sustainable harvesting with regard to CO2 uptake. Simulations based on the LPJ-GUESS model reproduce the saturation of GPP and NEP and convergence on the LAI threshold range. Accordingly, in managed forests, trees can be harvested while maintaining a high tree biomass and carbon sink of the remaining stand. In this case competition between neighbour trees in unmanaged forests is replaced by harvest management. In unmanaged forests, competition for light, nutrient and water cause self-thinning, thereby limiting the carbon sink strength.

2. Reconciling the EU Forest, Biodiversity, and Climate Strategies

   Konstantin Gregor, Christopher P. O. Reyer , Thomas A. Nagel , and 4 more authors

   Global Change Biology, Aug 2024

   Abstract Link

   Forests provide important ecosystem services (ESs), including climate change mitigation, local climate regulation, habitat for biodiversity, wood and non-\-wood products, energy, and recreation. Simultaneously, forests are increasingly affected by climate change and need to be adapted to future environmental conditions. Current legislation, including the European Union (EU) Biodiversity Strategy, EU Forest Strategy, and national laws, aims to protect forest landscapes, enhance ESs, adapt forests to climate change, and leverage forest products for climate change mitigation and the bioeconomy. However, reconciling all these competing demands poses a tremendous task for policymakers, forest managers, conservation agencies, and other stakeholders, especially given the uncertainty associated with future climate impacts. Here, we used process-b\-ased ecosystem modeling and robust multi-\-criteria optimization to develop forest management portfolios that provide multiple ESs across a wide range of climate scenarios. We included constraints to strictly protect 10% of Europe’s land area and to provide stable harvest levels under every climate scenario. The optimization showed only limited options to improve ES provision within these constraints. Consequently, management portfolios suffered from low diversity, which contradicts the goal of multi-\-functionality and exposes regions to significant risk due to a lack of risk diversification. Additionally, certain regions, especially those in the north, would need to prioritize timber provision to compensate for reduced harvests elsewhere. This conflicts with EU LULUCF targets for increased forest carbon sinks in all member states and prevents an equal distribution of strictly protected areas, introducing a bias as to which forest ecosystems are more protected than others. Thus, coordinated strategies at the European level are imperative to address these challenges effectively. We suggest that the implementation of the EU Biodiversity Strategy, EU Forest Strategy, and targets for forest carbon sinks require complementary measures to alleviate the conflicting demands on forests.

3. Quantifying the Impact of Key Factors on the Carbon Mitigation Potential of Managed Temperate Forests

   Konstantin Gregor, Andreas Krause , Christopher P. O. Reyer , and 4 more authors

   Carbon Balance and Management, Mar 2024

   Abstract Link

   Background\enspace Forests mitigate climate change by reducing atmospheric CO2-concentrations through the carbon sink in the forest and in wood products, and substitution effects when wood products replace carbon-intensive materials and fuels. Quantifying the carbon mitigation potential of forests is highly challenging due to the influence of multiple important factors such as forest age and type, climate change and associated natural disturbances, harvest intensities, wood usage patterns, salvage logging practices, and the carbon-intensity of substituted products. Here, we developed a framework to quantify the impact of these factors through factorial simulation experiments with an ecosystem model at the example of central European (Bavarian) forests. Results\enspace Our simulations showed higher mitigation potentials of young forests compared to mature forests, and similar ones in broad-leaved and needle-leaved forests. Long-lived wood products significantly contributed to mitigation, particularly in needle-leaved forests due to their wood product portfolio, and increased material usage of wood showed considerable climate benefits. Consequently, the ongoing conversion of needle-leaved to more broad-leaved forests should be accompanied by the promotion of long-lived products from broad-leaved species to maintain the product sink. Climate change (especially increasing disturbances) and decarbonization were among the most critical factors influencing mitigation potentials and introduced substantial uncertainty. Nevertheless, until 2050 this uncertainty was narrow enough to derive robust findings. For instance, reducing harvest intensities enhanced the carbon sink in our simulations, but diminished substitution effects, leading to a decreased total mitigation potential until 2050. However, when considering longer time horizons (i.e. until 2100), substitution effects became low enough in our simulations due to expected decarbonization such that decreasing harvests often seemed the more favorable solution. Conclusion\enspace Our results underscore the need to tailor mitigation strategies to the specific conditions of different forest sites. Furthermore, considering substitution effects, and thoroughly assessing the amount of avoided emissions by using wood products, is critical to determine mitigation potentials. While short-term recommendations are possible, we suggest risk diversification and methodologies like robust optimization to address increasing uncertainties from climate change and decarbonization paces past 2050. Finally, curbing emissions reduces the threat of climate change on forests, safeguarding their carbon sink and ecosystem services.

4. Frost Matters: Incorporating Late-Spring Frost into a Dynamic Vegetation Model Regulates Regional Productivity Dynamics in European Beech Forests

   Benjamin F. Meyer , Allan Buras , Konstantin Gregor, and 5 more authors

   Biogeosciences, Mar 2024

   Abstract Link

   Abstract. Late-spring frost (LSF) is a critical factor influencing the functioning of temperate forest ecosystems. Frost damage in the form of canopy defoliation impedes the ability of trees to effectively photosynthesize, thereby reducing tree productivity. In recent decades, LSF frequency has increased across Europe, likely intensified by the effects of climate change. With increasing warming, many deciduous tree species have shifted towards earlier budburst and leaf development. The earlier start of the growing season not only facilitates forest productivity but also lengthens the period during which trees are most susceptible to LSF. Moreover, recent forest transformation efforts in Europe intended to increase forest resilience to climate change have focused on increasing the share of deciduous species in forests. To assess the ability of forests to remain productive under climate change, dynamic vegetation models (DVMs) have proven to be useful tools. Currently, however, most state-of-the-art DVMs do not model processes related to LSF and the associated impacts. Here, we present a novel LSF module for integration with the dynamic vegetation model Lund–Potsdam–Jena General Ecosystem Simulator (LPJ-GUESS). This new model implementation, termed LPJ-GUESS-FROST, provides the ability to directly attribute simulated impacts on forest productivity dynamics to LSF. We use the example of European beech, one of the dominant deciduous species in central Europe, to demonstrate the functioning of our novel LSF module. Using a network of tree-ring observations from past frost events, we show that LPJ-GUESS-FROST can reproduce productivity reductions caused by LSF. Further, to exemplify the effects of including LSF dynamics in DVMs, we run LPJ-GUESS-FROST for a study region in southern Germany for which high-resolution climate observations are available. Here, we show that modeled LSF plays a substantial role in regulating regional net primary production (NPP) and biomass dynamics, emphasizing the need for LSF to be more widely accounted for in DVMs.

5. Can Triad Forestry Reconcile Europe’s Biodiversity and Forestry Strategies? A Critical Evaluation of Forest Zoning

   Thomas A. Nagel , Mariano Rodríguez-Recio , Tuomas Aakala , and 40 more authors

   Ambio, Dec 2024

   Abstract Link

   Balancing increasing demand for wood products while also maintaining forest biodiversity is a paramount challenge. Europe’s Biodiversity and Forest Strategies for 2030 attempt to address this challenge. Together, they call for strict protection of 10% of land area, including all primary and old growth forests, increasing use of ecological forestry, and less reliance on monocultural plantations. Using data on country wide silvicultural practices and a new database on strict forest reserves across Europe, we assess how triad forest zoning could help meet these goals. Our analysis reveals that zoning in Europe is overwhelmingly focused on wood production, while there has been little concomitant protection of forests in strict reserves. Moreover, most strict forest reserves are } 50 ha in size, likely too small to capture the minimum dynamic area necessary to sustain many taxa. We outline research priorities to meet future demands for timber while minimizing the impact on native biodiversity.

2023

1. Modeling the Carbon Costs of Plant Phosphorus Acquisition in Amazonian Forests

   Tatiana Reichert , Anja Rammig , Phillip Papastefanou , and 6 more authors

   Ecological Modelling, Nov 2023

   Abstract Link

   Plants growing in low phosphorus (P) soils, such as in the predominant soils of Amazonia, are believed to devote more energy to acquiring P through absorptive root production, symbionts, and root exudates than plants in more fertile soils. Accounting for these energy costs in vegetation models is essential, as underestimating carbon (C) allocation to nutrient acquisition may lead to overestimating plant biomass growth. We developed a quantitative model to test a theoretical framework of C costs of P acquisition across soil P gradients. The model considers four strategies: P foraging via absorptive roots and arbuscular mycorrhizal fungi and P mining via root exudation of phosphatases and organic acids. We used field observations (i.e., soil data, plant biomass production, and stoichiometry of different organs) from ten sites across Amazonia to calibrate the model and explore different scenarios of (i) experimental soil P addition and (ii) elevated atmospheric CO2 concentrations (eCO2). Our model reproduced expected trends in P-acquisition strategies, with plants increasingly investing in foraging strategies as soil soluble inorganic P (Pi) increases and increasingly investing in mining strategies as total P and less available P forms decrease. Relative investment in P acquisition was within observed ranges. Plants, on average and across all sites, invested the equivalent of 20.5% of their estimated total net primary production (NPP) in P acquisition. On average, plants allocated 15.3% of their NPP to P acquisition in the three most fertile sites, compared to 29.0% in the least fertile sites. C allocation to arbuscular mycorrhizas, phosphatases, and organic acids, which are not commonly measured components of total NPP, was up to 25.8% (16.9% on average) of the total NPP. We highlight the need for quantitative data on plant C allocation to P acquisition from the soil to strengthen further model development and future model projections.

2. Large Variability in Simulated Response of Vegetation Composition and Carbon Dynamics to Variations in Drought-Heat Occurrence

   Elisabeth Tschumi , Sebastian Lienert , Ana Bastos , and 11 more authors

   Journal of Geophysical Research: Biogeosciences, Apr 2023

   Abstract Link

   The frequency of heatwaves, droughts and their co-occurrence vary greatly in simulations of different climate models. Since these extremes are expected to become more frequent with climate change, it is important to understand how vegetation models respond to different climatologies in heatwave and drought occurrence. In previous work, six climate scenarios featuring different drought-heat signatures have been developed to investigate how single versus compound extremes affect vegetation and carbon dynamics. Here, we use these scenarios to force six dynamic global vegetation models to investigate model agreement in vegetation and carbon cycle response to these scenarios. We find that global responses to different drought-heat signatures vary considerably across models. Models agree that frequent compound hot-dry events lead to a reduction in tree cover and vegetation carbon stocks. However, models show opposite responses in vegetation changes for the scenario with no extremes. We find a strong relationship between the frequency of concurrent hot-dry conditions and the total carbon pool, suggesting a reduction of the natural land carbon sink for increasing occurrence of hot-dry events. The effect of frequent compound hot and dry extremes is larger than the sum of the effects when only one extreme occurs, highlighting the importance of studying compound events. Our results demonstrate that uncertainties in the representation of compound hot-dry event occurrence in climate models propagate to uncertainties in the simulation of vegetation distribution and carbon pools. Therefore, to reduce uncertainties in future carbon cycle projections, the representation of compound events in climate models needs to be improved.

2022

1. Trade-Offs for Climate-Smart Forestry in Europe Under Uncertain Future Climate

   Konstantin Gregor, Thomas Knoke , Andreas Krause , and 6 more authors

   Earth’s Future, Sep 2022

   Link
2. Quantifying the Impacts of Land Cover Change on Gross Primary Productivity Globally

   Andreas Krause , Phillip Papastefanou , Konstantin Gregor, and 6 more authors

   Scientific Reports, Nov 2022

   Abstract Link

   Historically, humans have cleared many forests for agriculture. While this substantially reduced ecosystem carbon storage, the impacts of these land cover changes on terrestrial gross primary productivity (GPP) have not been adequately resolved yet. Here, we combine high-resolution datasets of satellite-derived GPP and environmental predictor variables to estimate the potential GPP of forests, grasslands, and croplands around the globe. With a mean GPP of 2.0 kg C m -2 yr -1 forests represent the most productive land cover on two thirds of the total area suitable for any of these land cover types, while grasslands and croplands on average reach 1.5 and 1.8 kg C m -2 yr -1 , respectively. Combining our potential GPP maps with a historical land-use reconstruction indicates a 4.4% reduction in global GPP from agricultural expansion. This land-use-induced GPP reduction is amplified in some future scenarios as a result of ongoing deforestation (e.g., the large-scale bioenergy scenario SSP4-3.4) but partly reversed in other scenarios (e.g., the sustainability scenario SSP1-1.9) due to agricultural abandonment. Comparing our results to simulations from state-of-the-art Earth System Models, we find that all investigated models deviate substantially from our estimates and from each other. Our maps could be used as a benchmark to reduce this inconsistency, thereby improving projections of land-based climate mitigation potentials.

2014

1. Tracking Indistinguishable Translucent Objects over Time Using Weakly Supervised Structured Learning

   Luca Fiaschi , Ferran Diego , Konstantin Gregor, and 4 more authors

   In 2014 IEEE Conference on Computer Vision and Pattern Recognition , Jun 2014

   Abstract Link

   We use weakly supervised structured learning to track and disambiguate the identity of multiple indistinguishable, translucent and deformable objects that can overlap for many frames. For this challenging problem, we propose a novel model which handles occlusions, complex motions and non-rigid deformations by jointly optimizing the flows of multiple latent intensities across frames. These flows are latent variables for which the user cannot directly provide labels. Instead, we leverage a structured learning formulation that uses weak user annotations to find the best hyperparameters of this model.

2013

1. Keeping Count: Leveraging Temporal Context to Count Heavily Overlapping Objects

   Luca Fiaschi , Gregor Konstantin , Bruno Afonso , and 2 more authors

   In 2013 IEEE 10th International Symposium on Biomedical Imaging , Apr 2013

   Abstract Link

   When tracking and segmenting multiple objects under heavy occlusion, a large class of algorithms can greatly benefit from a preprocessing that reliably assesses the number of individuals in each cluster. This is a difficult task when relying on local information only, due to scarcity of training examples and lack of strongly predictive features. In this paper, we develop a deterministic graphical model to address the problem of counting the number of objects in each foreground region as global inference across the entire video sequence. We show that global inference improves over local predictions, and is able to produce an accurate and coherent output within an useful runtime.