The work, published in Environmental and Sustainability Indicators, is deliberately framed as a methodological proof of concept rather than a ready-to-use planning system. Even so, its findings are striking. When the algorithm was unleashed on representative model stands drawn from the university’s forest enterprise — more than 9,200 hectares divided among the Šášov, Budča and Zvolen forest districts — it consistently converged on a narrow set of management strategies, and it translated the priorities of very different stakeholders into measurably different harvesting patterns.

The technical heart of the framework is a two-way conversation between optimization and simulation. The researchers generated 120 model stands of 0.25 hectares each, spanning four geobiotope types that together cover roughly 80 percent of the enterprise: extremely dry oak–hornbeam forests, fresh oak–hornbeam forests, and two variants of floristically rich beech forests. Each stand was characterized by site productivity class, age and stocking density. The BPSO algorithm then assigned one of three fates to every tree in a stand: keep it alive, harvest and remove it, or harvest it and leave the trunk behind as deadwood. A simple logical rule ensured that no living tree could be declared deadwood without first being selected for harvest.

For every candidate solution, the framework calculated standardized scores for four groups of values: production, regulation, cultural ecosystem services and biodiversity. Indicators included harvested volume and timber revenues, water quantity and quality, stand density, aesthetic value, visibility, longevity, seasonality and species diversity. These group scores were combined into a single multicriteria utility function whose weights came not from the researchers’ imaginations but from real stakeholder interviews previously conducted in Slovak Paradise National Park. Six profiles were tested, ranging from a production-minded economist, who assigned 36.4 percent weight to production, to a “do nothing” conservationist, who gave biodiversity a dominant 64.1 percent.

The algorithm then searched the vast decision space using swarm intelligence. Each tree was treated as a particle moving through a binary hyperspace, guided by its own best past position and by the best solution found by the whole swarm. The configuration used an inertia of 0.9, a cognitive factor of 0.8 and a social factor of 1.6, with runs capped at 100 iterations and an early stopping rule after 50 fruitless rounds. Once an optimal decision vector was found, the updated stand was handed back to Sibyla, which simulated five years of growth, mortality and competition. The cycle repeated across a 20-year planning horizon divided into four decision periods, and the whole process was run for nine predefined management schedules, from near-continuous intervention to complete non-intervention.

The results revealed a strong asymmetry in the optimizer’s preferences. Schedule V, meaning no intervention at all, and Schedule III, a single intervention in the first period, together accounted for roughly three-quarters of all selected schedules across geobiotopes, stakeholder profiles and stand ages. Schedules involving repeated or delayed interventions were rarely or never chosen. Some solutions also accumulated implausibly large volumes of deadwood, because the unconstrained algorithm was free to maximize utility without any operational guardrails. The authors are candid about what this means: maximizing multicriteria utility alone is insufficient when ecological and economic feasibility are not written into the problem as explicit constraints.

Yet the framework also produced ecologically coherent and genuinely interesting patterns. In the oak–hornbeam geobiotopes, oak declined while beech and hornbeam increased, mirroring the natural competitive dynamics of Central European mixed forests where shade-tolerant species replace oak under reduced disturbance. In the beech forests, spruce consistently declined while broadleaves and sometimes fir increased — a trajectory that echoes field observations of Norway spruce struggling outside its optimal range under warming temperatures and prolonged drought. Crucially, these successional directions were driven mainly by site conditions and initial stand composition rather than by stakeholder preferences, suggesting that the growth simulator kept the optimization anchored in ecological reality.

Where stakeholder preferences did matter was in the intensity and distribution of management. Conservation-oriented profiles retained a greater share of dominant trees as deadwood, concentrating large retained trunks in the canopy class that matters most for saproxylic beetles, cavity-nesting birds and nutrient cycling. Economically oriented profiles directed more harvesting toward those same dominant trees. The framework was therefore sensitive enough to reflect contrasting societal priorities without prescribing a single universal solution — a property the authors see as essential for participatory forest planning.

The study also exposes the hard problems that stand between prototype and practice. The multicriteria utility depends on a two-stage standardization procedure whose scaling choices shape the outcome, yet no formal sensitivity analysis was performed. The deadwood indicator carried a disproportionately high inherent weight, and the stand-volume indicator inflated the appeal of doing nothing. The regulation indicators for water rely on expert-assigned values available only for common species, masking differences among rarer ones. The authors recommend converting volume and deadwood into constraining conditions rather than freely optimized objectives, extending the regulation group with erosion control, noise reduction and carbon sequestration, and refining the deadwood mechanism by linking it to Sibyla’s natural-mortality module with explicit retention bounds and decomposition dynamics.

The researchers also sketch the road ahead: integrating an economic-evaluation module, embedding climate-change scenarios into the growth simulations, testing modern BPSO variants for more reliable convergence, and validating the framework against operational management strategies. Planning horizons of 10 to 50 years appear most appropriate, balancing predictive accuracy against the rising risk of windthrow, insect outbreaks and climatic extremes. If those challenges can be met, the payoff could extend beyond better harvest plans. Because the infrastructure is flexible in scale and transparent in its ecosystem-service accounting, the authors suggest it could one day support payment schemes for ecosystem-service providers and give communities a structured, quantified voice in how their forests are managed. For now, the message is clear: the mathematics exists to plan a forest for everything it provides, not just its timber — the task now is to teach the algorithm the rules that foresters, and forests themselves, cannot break.

Subject of Research: Simulation–optimization framework for ecosystem-service-oriented forest management planning

Article Title: New approach in forest management planning

Article References: Kúdela, P., & Sedmák, R. (2026). New approach in forest management planning. Environmental and Sustainability Indicators, 32, Article 101564. https://doi.org/10.1016/j.indic.2026.101564

Image Credits: AI Generated

DOI: 10.1016/j.indic.2026.101564

Keywords: forest management, ecosystem services, particle swarm optimization, forest growth simulation, biodiversity, deadwood, Sibyla, Slovakia, multicriteria optimization, stakeholder preferences, decision support systems, silviculture