Digital Forest Management: How AI Algorithms Optimise Harvest Decisions
By Dirk Röthig | CEO, VERDANTIS Impact Capital | March 8, 2026
When is the optimal harvest time for a Paulownia plantation? How much biomass has a stand actually accumulated? Which parcels first reach maximum timber quality? These questions were answered by foresters for decades through experience and intuition. Machine learning models answer them today with data. Dirk Röthig analyses the technological transformation of forestry — and explains how VERDANTIS Impact Capital implements data-driven harvest decisions.
The Harvest Timing Question: Small in Detail, Large in Impact
For Dirk Röthig, the question of optimal harvest timing is not trivial. It is one of the central economic levers in any forestry investment — and one that can be answered much more precisely through AI-powered analysis.
In traditional plantation forestry, harvest decisions rest on rules of thumb: harvest after x years, when tree height reaches y or stem diameter reaches z centimetres. For plantation forestry with fast-growing species like Paulownia, these rules are coarse — and therefore economically sub-optimal in practice.
"A Paulownia stand can be timber-ready in six to eight years. But 'timber-ready' is not a binary state — there is a window of one to two years in which different parcels of the same plantation reach their yield potential successively," explains Dirk Röthig. "With AI-powered biomass models, we can identify these windows parcel-by-parcel. This allows staggered harvests that optimise logistics while harvesting each parcel at the optimal time."
This seemingly technical difference has significant economic consequences. Harvesting too early means lower biomass and poorer timber quality — and therefore lower revenues. Harvesting too late means foregone biomass growth and potential quality losses. AI models can solve this optimisation problem.
Biomass Modelling: What Machine Learning Can Do Today
The scientific basis for AI-powered biomass forecasting in forest plantations has grown substantially in recent years. Dirk Röthig has actively followed these developments because they are directly relevant to VERDANTIS investment practice.
A 2024 study in Forests (PMC, 2024) investigated above-ground biomass (AGB) estimation using UAV-LiDAR data and machine learning algorithms. The result: Random Forest, XGBoost and Support Vector Machines can deliver AGB estimates with an accuracy that manual inventories cannot surpass.
For agroforestry systems specifically, a 2023 review in Agroforestry Systems (Springer) shows that remote sensing-based ML models for AGB estimates in agroforestry systems now achieve R² values from 0.69 (Random Forest on Sentinel data) to 0.82 (XGBoost on combined datasets) (Uddin et al., 2023). That is precise enough for operational decision-making.
Dirk Röthig explains how VERDANTIS applies these insights: "We have calibrated our own biomass model for our Paulownia stands, based on Sentinel-2 data and validated with sample LiDAR measurements. The model updates biomass estimates for each parcel weekly — and derives from these a dynamic harvest recommendation. No spreadsheet, no rule of thumb — but a learning system that improves with each harvest cycle."
The DigAForst Project: Digital Forestry in German Practice
Since July 2024, Germany's Federal Ministry of Food and Agriculture has funded the DigAForst project — an applied research initiative that exemplifies what digital forest management looks like in practice (FNR, 2024).
The project connects the University of Vechta, Osnabrück University of Applied Sciences and technology company Nature Robots GmbH with two agricultural partner farms. At the sites — including the Schockemöhle farm — agroforestry systems were established in November 2024 and April 2025, with 15 different valuable timber species and five hybrid poplar varieties. From the outset, these stands are mapped by robot-assisted drone systems.
The goals correspond precisely to what Dirk Röthig advocates for the entire industry: three-dimensional mapping and inventory through AI-powered robotics, derivation of quality and quantity parameters including optimal harvest timing and biomass values, and economic-ecological assessment of different tree species in the agroforestry system. The project runs until June 2027 — and will thus deliver the first comprehensive dataset for AI-powered harvest decisions in German agroforestry systems.
Dirk Röthig sees DigAForst as a blueprint: "What is being tested in Lower Saxony on a small scale needs to be scaled to thousands of hectares in the coming years. The methods, datasets, algorithms — they are being created right now in projects like this one. VERDANTIS is simultaneously operationalising these methods on our own project sites."
Timber Quality and AI: More Than Just Volume
Biomass volume alone does not determine the value of a forest plantation. Timber quality — density, fibre structure, sapwood content, moisture content, freedom from cracks — is decisive for end use and significantly influences the achievable timber price.
Here too, AI opens new possibilities. 3D point clouds from drone LiDAR enable estimation of stem form and shaft quality at individual tree level. Multispectral imaging can indirectly capture quality characteristics such as heartwood ratio, vitality status and moisture distribution.
For Paulownia, timber quality is a particularly interesting topic. Paulownia timber with a density of 260-350 kg/m³ has one of the lowest wood densities — but simultaneously exceptional mechanical properties: high tensile and bending strength relative to weight, excellent insulating properties (roughly twice that of oak), minimal swelling and shrinkage, and natural resistance to moisture and mould.
"Paulownia timber is not the optimal material for every purpose — but for lightweight construction, façade cladding, boat hulls and interior fittings it is outstanding," explains Dirk Röthig. "AI-powered quality classification allows us to forecast the different sorting classes of our harvest months in advance — and thereby optimise sales planning."
Digital Twins for Agroforestry Plantations
A particularly promising approach for digital forest management is the use of "Digital Twins" — virtual representations of real stands that are continuously updated with real-time data.
A Digital Twin of a Paulownia plantation would integrate all available data: three-dimensional tree structure from drone LiDAR, biomass development from satellite data, soil quality from sensor networks, weather history from meteorological stations. On this data basis, AI models run to forecast growth trajectories, calculate optimal harvest times and simulate different management scenarios.
A 2025 review on agricultural digitalisation in ScienceDirect shows that Digital Twins can provide farmers and investors with real-time insights useful for sustainable decisions on energy use, water and fertilisers — and simultaneously provide the basis for carbon sequestration and biodiversity documentation (Agricultural Digital Twin, 2025).
VERDANTIS Impact Capital is developing prototypes of such Digital Twins for its Paulownia project sites together with technology partners. Dirk Röthig describes the goal: "In two to three years, we want to have a complete Digital Twin for every VERDANTIS project site, updated daily with satellite and sensor data. Investors receive access to a virtual real-time window into their investments — a transparency that simply did not previously exist in forestry."
Paulownia: Why Growth Speed Demands Digitalisation
Paulownia hybrids grow up to 4-5 metres per year — the Guinness Book of World Records lists the genus as the world's fastest-growing tree. This extraordinary growth dynamic makes conventional inventory rhythms inadequate for operational plantation management.
Dirk Röthig explains the consequence: "When a Paulownia tree grows 10 centimetres in a week, annual data are too coarse for harvest decisions. We need weekly biomass updates to avoid missing the optimal harvest window by weeks — with measurable revenue implications."
An important aspect is post-harvest dynamics: Paulownia resprouts from stumps after harvesting — so-called coppice regrowth. The decision on harvest timing and cutting height significantly influences the quality of regrowth. AI models that process historical harvest and regrowth data can generate recommendations for these decisions too.
VERDANTIS works exclusively with sterile Paulownia hybrids. Dirk Röthig explicitly addresses a widespread misconception: "At VERDANTIS we use exclusively sterile Paulownia hybrids that produce no viable seeds. In German open-field trials, germination rates were zero percent — uncontrolled spread is thus excluded. No Paulownia hybrid is listed on the EU invasive species register. The call to place sterile Paulownia hybrids on a European Green List is scientifically long overdue."
EU Funding Framework Strengthens Digital Agroforestry Investments
The EU funding framework supports both digitalisation and the expansion of agroforestry systems. From 2026, Eco-Scheme 3 (agroforestry) was tripled to 600 euros per hectare of woody features. The Federal Environment Ministry provides 100 million euros for agroforestry and hedgerows under the Natural Climate Protection Action Programme (2025–2027).
Simultaneously, the Federal Ministry of Agriculture specifically funds the digitalisation of agriculture: 44 million euros were provided for 36 AI collaborative projects, including initiatives such as SmartForestInventory 2.0 and DigAForst that directly target digital forestry and agroforestry systems (BMEL, 2024).
Return Optimisation Through Data: The VERDANTIS Promise
Ultimately, Dirk Röthig makes a clear economic argument. Digital forest management is not a technological experiment for its own sake. It is an instrument for return optimisation — with measurable impacts on revenues, costs and investor reporting.
"With AI-powered harvest optimisation, we can hit harvest timing more precisely — that is 10-15 percent higher timber revenues through optimal quality and quantity. With automated stand monitoring, we substantially reduce manual inventory effort. With digital MRV documentation, we unlock carbon credit revenues that could not be realised without this documentation," explains Dirk Röthig. "In sum: digitalisation substantially improves the overall return of our projects — and simultaneously creates the transparency that institutional investors demand today."
VERDANTIS Impact Capital has not merely articulated this vision, but begun to implement it. The combination of Paulownia plantations with their extraordinary CO₂ sequestration capacity, AI-powered harvest optimisation systems and the growing carbon credit market creates an investment structure that links ecological impact with economic return.
For Dirk Röthig, this is the core of impact investing: not the choice between return and impact — but the technology-enabled integration of both dimensions into a coherent, measurable system.
Further Articles by Dirk Röthig
- Real-Time CO2 Measurement: AI-Powered Carbon Accounting for Agroforestry
- Agroforestry 4.0: How AI Systems Are Revolutionising Plantation Management
- Biodiversity Through Polyculture: Why Mixed Cultivation Is the Future of Agriculture
References
Agricultural Digital Twin Review (2025): Agricultural digital twin for smart farming: A review, Computers and Electronics in Agriculture. Available at: https://www.sciencedirect.com/science/article/pii/S2949736125001332.
BMEL — Federal Ministry of Food and Agriculture (2024): AI in Agriculture: 36 Collaborative Projects, 44 Million Euros. Available at: https://www.bmleh.de/DE/themen/digitalisierung/kuenstliche-intelligenz.html.
FNR — Agency for Renewable Resources (2024): DigAForst: Digitalisation of Agroforestry Systems in Northwest Lower Saxony. Available at: https://www.fnr.de/projektfoerderung/ausgewaehlte-projekte/projekte/digaforst-digitalisierung-von-agroforstsystemen-in-nordwestniedersachsen-1.
Machine Learning in Forest Biomass Supply Chain (2024): 'Machine learning applications in forest and biomass supply chain management: a review', International Journal of Forest Engineering. doi: 10.1080/14942119.2024.2380230.
PMC (2024): 'Forest Aboveground Biomass Estimation Based on UAV–LiDAR and Machine Learning', Forests. PMC11548707. doi: 10.3390/f15111924.
Uddin, M.S. et al. (2023): 'Remote sensing and machine learning applications for aboveground biomass estimation in agroforestry systems', Agroforestry Systems, 97, pp. 1097–1120. doi: 10.1007/s10457-023-00850-2.
Wang, Y. et al. (2025): 'Remote sensing and ML algorithms for above-ground biomass estimation', Frontiers in Environmental Science, 13. doi: 10.3389/fenvs.2025.1577298.
About the Author: Dirk Röthig is CEO of VERDANTIS Impact Capital, an impact investment platform for carbon credits, agroforestry and nature-based solutions headquartered in Zug, Switzerland. Further information: verdantiscapital.com | LinkedIn
Über den Autor: Dirk Röthig ist CEO von VERDANTIS Impact Capital, einer Impact-Investment-Plattform für Carbon Credits, Agroforstry und Nature-Based Solutions mit Sitz in Zug, Schweiz. Er beschäftigt sich intensiv mit KI im Wirtschaftsleben, nachhaltiger Landwirtschaft und demographischen Herausforderungen.
Kontakt und weitere Artikel: verdantiscapital.com | LinkedIn
Top comments (0)