Editor's note
This Deep Dive look at an emerging shift in precision forestry: moving from canopy-level analytics to true seedling-level intelligence. New tools, from drones to tool-mounted loggers and autonomous planters are starting to illuminate the part of the regeneration cycle that has historically been invisible.
Together they point toward a data foundation where every seedling can be traced, monitored and modelled, forming the groundwork for digital twins, next level tree breeding and future automation across the sector.
Would love to hear your thoughts on this,
Axel
PS. Should you find a moment, I would be grateful for your thoughts via the brief BTB reader survey. In less than two minutes, you can help shape the future of this newsletter. A handful of participants will be rewarded with BTB prizes. The opportunity to share your voice will close in a week's time.
Three Paths to Precision Reforestation
Most “precision forestry” starts when trees are big enough to show up in satellite or airborne data. Seedlings, the start of every stand, are effectively invisible at those scales, and manual inventories are too slow and patchy to close the gap. That missing layer weakens provenance tracking, survival analytics, and the future of automated pre-commercial thinning and harvesting.
A new generation of tools now makes seedling-level mapping realistic at scale. This article looks at three complementary approaches:
Drone-based seedling inventory (Skyforest)
Manual tool-mounted logging (STA system)
Robotic planting and mapping (Södra BraSatt)
All three shift precision forestry from “canopy-first” to “seedling-first”.
Why Seedling Logging Matters
When each seedling is mapped, reforestation changes from a bulk operation to a traceable, data-rich process.
Provenance and genetics. Modern nurseries track seed lots, clones, and climate-adapted material through cold storage and logistics. If that information is linked to the exact planting location, forest owners can later see which provenances actually perform on specific soils, aspects, and climates.
Survival and growth analytics. Seedling coordinates, planting date, and method can be joined with weather, soil, and competition data to explain why some cohorts thrive and others fail. Instead of “survival is low here,” managers can pinpoint “spring drought on unscarified mineral soil is killing this batch.”
Operational risk and lost years. Regeneration failures are not only biological problems but economic ones. For every year a sparse or missed planting area goes undetected, forest owners lose growth, value and future options. What appears as a small gap in year one can translate into measurable volume and revenue losses decades later. Seedling-level logging systems make these issues visible early. Areas that are under-planted, unevenly stocked or missed entirely can be identified while corrective planting is still feasible, turning regeneration follow-up from a delayed audit into an active risk-management tool.
Scenario analysis and spacing-aware growth modelling. When each seedling has an estimated growth rate and precise coordinates, managers can run realistic “what-if” scenarios at tree level, testing different thinning timings, density targets, or species mixes, and seeing how changes in spacing and competition are likely to affect stand volume and value over time.
Future automation. AI-assisted thinning and harvesting already rely on tree-level data. If those systems also know each tree’s planting history and genetic background, they can make finer decisions; favoring robust, well-placed individuals and protecting desired genetic diversity.
Seedling logging is not just monitoring; it is the ground truth that future automation and climate-smart breeding will depend on.
Approach 1: Drone Seedling Inventory
Traditional plant inventories are slow, subjective and sparse. A 5-hectare stand may take close to an hour on foot and still only sample a fraction of the area, which is one reason only around 40% of new plantations are inventoried at all in some regions. Many failed regeneration patches are never detected in time for corrective planting.
Skyforest replaces this with a high-throughput drone workflow built around photo plots rather than full orthomosaics.
Technique
Flight design. Skyforest uses custom drones, on-board laser range finding and its own flight-planning software to position a dense grid of circular photo plots across each stand.
Photo plots. Each plot image covers roughly 9–13 m in diameter, calculated from altitude so that every frame represents a consistent ground area. At each point, the drone captures multiple high-resolution images to ensure clean data.
Speed. A 5-hectare stand with around 60 plots can be flown in about 15 minutes, compared to nearly an hour on foot with fewer plots. This makes it economical to inventory much larger shares of the planting program.
Back in the office, Skyforest’s pipeline detects seedlings, gaps and vegetation competition in each plot and turns them into stand-level statistics and gap maps. Results flow through the SkyPort portal into customers’ forest information systems, so plant inventory becomes part of regular planning and follow-up rather than a standalone PDF.
Strengths and Limits
Strengths: Fast, objective, repeatable; much higher plot density than manual methods; strong at identifying failed patches early. A further advantage is the ability to detect and plot naturally regenerated or naturally seeded seedlings with main-stem potential, so future crop trees from natural regeneration can be mapped alongside planted stock and explicitly included in planning for spacing, tending and future yield. Other providers are emerging with similar AI-based detection, but Skyforest’s edge is tight integration with operational planning and a proven method for consistent image quality.
Limits: Currently provides dense sampling rather than exact coordinates for every individual. As AI seedling detection matures, drone workflows are moving toward explicit, tree-by-tree mapping, but today they are best seen as “statistical x-ray” rather than per-seedling provenance logs.
Approach 2: Manual Tool-Mounted Logging
Where Skyforest looks down after planting, the STA system turns the workers themselves into mappers while they plant, spray, or tend seedlings. The tool is also widely used for targeted spraying against invasive species, ensuring precise logging of treatment areas without interrupting workflow. Developed by TerraLab, STA is a rugged logger that mounts on tools like planting tubes, backpack sprayers or boom sprayers. Its core idea: no apps, no interaction, no friction, just automatic logging as people work.
Technique
Hardware. The device integrates GPS, accelerometers, storage and battery in a compact form factor that can be quickly mounted and moved between tools.
Event detection. For planting with a planting tube, the accelerometer recognizes the motion signature of inserting and withdrawing the tube, triggering a log entry with time and coordinates.
Data flow. At day’s end the unit uploads automatically (via base station or network) into a web portal and directly into the organization’s GIS as feature layers—no manual file handling.
STA describes four guiding principles that drive adoption:
Versatility. One device supports planting, weed control, boom spraying and more, building a multi-layer dataset around each seedling.
Simplicity. No on/off button, no Bluetooth, no daily setup; once fitted, it just logs.
Efficiency. It never asks the worker to stop what they’re doing, so output per hour stays the same.
Integration. Data appears in a web portal but also slots directly into existing GIS and reporting workflows.
Seedling Data in Practice
A planting crew using STA with planting tubes generates one point per seedling with timestamp and operator ID. If the same system is used for pre-commercial thinning, each maintenance event is logged as well. Over time, managers can see:
Planting density and pattern by crew or contractor.
Which areas received thinning and when.
Where survival appears low (planting points with no later maintenance or no seedlings visible on subsequent imagery).
Mixed species or provenance batches can complicate automated logging if crews switch between types without manual tagging, potentially requiring post-processing to assign provenance metadata accurately. Compared with drones, STA gives higher precision on who did what, when, and where but only where human crews work. Compared with robots, it is deployable immediately using existing labor, and the capital cost is relatively modest.
Beyond operational logging, STA-like systems also enable cost-based benchmarking, such as plants per hour per operator and contract-compliance verification. For large forest companies, this creates a transparent basis for comparing contractors, improving quality assurance, and documenting that regeneration obligations are fulfilled.
Other players (GPS handhelds, mobile apps, equipment-mounted loggers) also map planting points, but they usually depend on user interaction or dedicated mapping passes. STA’s differentiator is its “zero-interaction, tool-mounted” design tuned for continuous operations logging.
Approach 3: Robotic Planting and Mapping
The logical end-point of seedling mapping is a robot that both plants and logs every tree with high precision. Södra’s BraSatt project is a leading example of this vision.
Concept and Technique
BraSatt is an autonomous machine designed to:
Navigate across clearcuts inside a geofenced area defined by a planning app.
Scan the environment with cameras and sensors to find obstacles and suitable planting spots.
Scarify small patches of soil and plant seedlings with millimeter-level control of depth and placement.
Every planting action is recorded: GPS location, time, soil prep parameters, local obstacle context and machine state. The system aims not only to automate labor but to improve planting quality, Södra’s stated target is around 90% survival three years after planting versus roughly 70% for many current practices.
The underlying research (e.g., AutoPlant’s vision-based planting-point selection) shows that machine-selected spots plus controlled scarification can significantly increase survival compared to manual planting on rough terrain.
Status and Ecosystem
BraSatt completed an initial development phase in 2024 and has secured funding for further refinement and early series production preparation. Södra positions it as long-term infrastructure for members rather than a quick commercial product; the main economic upside (better regeneration, less labor bottleneck) accrues to forest owners.
Other actors include excavator-mounted planting heads (Risutec, M-Planter) and global experiments with drone seeding and robotic planters. BraSatt’s distinctive contribution is the combination of autonomous navigation, intelligent spot selection, precise scarification, and built-in data logging in a coherent system owned by a major forest cooperative.
Putting It Together: Three Layers, One Data Future
The three approaches differ but reinforce each other:
Skyforest drones: fast, objective overview of planting success and competition at the stand and plot level, ideal for broad coverage and early failure detection.
STA tool-mounted logging: detailed, per-seedling operational history for all work done by human crews, capturing the practical reality of mixed manual operations.
BraSatt robots: high-quality, fully logged regeneration with consistent biomechanics and the richest machine-generated metadata per seedling.
An important distinction across the three technologies is the nature of their outputs.
Drone inventories provide statistical coverage with quantifiable confidence intervals.
STA logging delivers deterministic coordinates wherever crews worked.
Robotic planters generate deterministic full-coverage datasets since every seedling is planted and logged by the machine.
Avoiding a mechanisation bias in digital twins
An important implication of combining these approaches is that future digital twins do not become biased toward mechanised operations alone. Drone inventories, tool-mounted logging and robotic planting each capture different parts of operational reality. Together, they ensure that digital representations of forests reflect how regeneration actually happens across varied terrain, methods and labour models.
This makes seedling-level data not just richer, but more representative, supporting better long-term modelling, fairer benchmarking and more robust decision-making across the full spectrum of forestry operations.
This combination provides both breadth and precision, allowing managers to balance statistical inference with exact operational records.
Used together, they can describe a seedling from seedlot to satellite:
Nursery and logistics systems provide provenance and genetics.
Planting is logged either by STA or BraSatt with precise coordinates and method.
Early inventories by drones validate survival and guide corrective actions.
Later LiDAR, AI thinning support and harvester data connect early-life conditions to final yield and value.
The Missing Piece: Big Data Platforms
To fully exploit seedling-level information, forestry needs shared, scalable data platforms rather than siloed vendor systems.
Key requirements:
Common schemas. A standard seedling record (location, time, lot ID, genetics, method, soil prep, follow-up ops) so data from Skyforest, STA, BraSatt, nurseries and harvesters can interoperate.
Cloud-native spatial infrastructure. Systems similar to Forest Cloud or Earth Engine that can store and analyze billions of points and time series for large regions.
Governance and access. Forest owners must retain control while still being able to share anonymized or aggregated data with researchers, cooperatives or certification bodies where it adds value.
Seedling-level mapping is also the natural entry point for stand-scale digital twins.
Once seedlings are linked to provenance, soil preparation, micro-site conditions and early survival, stands can be simulated dynamically over time with far greater resolution. This places seedling data at the foundation of the long-term vision for climate-smart forestry: continuously updated, spatially explicit digital representations of every regeneration unit.
Once these foundations are in place, seedling mapping stops being a niche technical exercise and becomes the backbone of precision forestry: every tree traceable, every treatment measurable, every decision testable against outcomes over decades.