Latvia’s forest data - insights from 3 million stands

Welcome back to Arbonics’ newsletter, where we share developments in forestry, voluntary carbon markets, and climate solutions more broadly. In this edition, our Data Lead Linda walks through the careful process of turning millions of data rows into growth curves that reflect the underlying rhythm of Latvia’s forests.

👉 Also, don’t miss our upcoming podcast on Tuesday, April 1, where our founder Lisett Luik joins

Cecil

co-founder

Alex Logan

and Head of Science Tom Walker for a 45-minute conversation on how forest monitoring is evolving, which tools are delivering value today, and what high-integrity monitoring could look like in the years ahead.

Forests cover about half of Latvia’s land area, making them one of the nation’s most valuable natural resources. To put it in perspective, while the Netherlands is famous for having more bicycles than inhabitants, Latvia boasts more forestry parcels than people!

With ~3 million forestry plots and a population of roughly 2 million, the sheer abundance of forest land underlines its importance, not only for timber production but also for biodiversity, carbon sequestration, and recreation.

Fig. 1. Sentinel-2 satellite mosaic showcasing the country’s greenery

At Arbonics, our mission is to scale nature-based solutions through technology. We strive to use local data to make our carbon credit estimates as accurate as possible in each country we operate.

That's why tapping into the rich, detailed records of Latvia’s forests was so vital. With a registry documenting all registered forest lands, we have a unique window into the growth patterns and health of these woodlands— key information for both sustainable management and reliable carbon accounting.

From a sea of data to a clear picture

Our journey began with 3 million rows of raw data— each representing a unique forest stand. However, extracting meaningful insights from such a vast dataset isn’t straightforward. We had to meticulously navigate through multiple steps of data processing and analysis. Starting with the first round of processing and exclusions, we:

• Translated and de-coded information into understandable labels

• Ensured no invalid records spoilt our analysis, such as those with negative values, missing information or implausibly high stock figures— likely a measuring, reporting or data handling error

These initial steps already refined our dataset to about 1.5 million high-quality records, providing a solid foundation for deeper analysis.

Fig. 2. Tree stock (in m³/ha) distribution with stand age, broken down by species (rows) and bonity class (columns). The denser, warmer-coloured hexes reveal where most data points cluster

Some species, like pine, appear across all bonities, while others thrive only under specific site qualities. At the same time, the variability within each panel is striking— even stands of the same primary species and bonity can differ substantially in their measured stock, reminding us that no two stands are exactly alike.

Building growth curves

To transform these data points into meaningful insights, we embarked on a multi-step analytical journey. We filtered the data even further to focus on stands where a single species is clearly dominant, ensuring that our analysis accurately reflects the growth patterns of that particular species.

Then, using Generalized Additive Models (GAMs), we let the data guide us in crafting average growth curves that capture the nonlinear nature of tree growth— from rapid starts to slower mid-life growth and eventual plateauing.

Each step, from data cleaning and filtering to smoothing and curve refinement, built on the previous one, ultimately revealing the underlying rhythm of these living forests.

Fig. 3. Final average growth curves for most common species in Latvia

While the growth curves capture the overall growth trends well, one striking pattern emerged from our analysis: a lot of forests are harvested or managed before they reach an old age. This can create a data gap for older stands, making it challenging to model the later years of growth.

What would happen if these forests were allowed to mature for decades— or even a century? Would biomass continue to climb, level off, or decline? While trees do eventually die and decompose, old trees transform into critical components of the ecosystem, serving as habitats and nutrient reservoirs and new trees naturally replace them.

While these curves offer valuable insights, they are, at their core, models— simplified representations that smooth over the inherent variability of nature. Factors such as soil differences, climate, management practices and random measurement errors all contribute to the scatter seen in each panel. Therefore, while these curves help guide forest management decisions, they must be interpreted with an understanding of their limitations.

At Arbonics we have developed software that predicts forest growth for any mature forest for the next 40+ years, assessing whether a given forest type is suitable for a carbon project and providing landowners with actionable insights.

Latvia’s forests are a living resource that provides timber, vital wildlife habitats, and even cultural identity. Covering nearly half the country, these woodlands form an essential backbone for both the economy and the environment, and our work aims to ensure that this natural legacy endures for generations to come.