About Us | BioAurea

Our purpose

We believe no territory can be protected if it cannot be measured. Estimating forest carbon the traditional way is costly, slow, and leaves out local communities, small carbon projects, and the institutions that need to track the state of their territory.

Our purpose is to change that. Every vereda in Colombia, from Chocó to the Amazon, should have data as rigorous as that of any capital in the world, and should have it in minutes rather than months. And that data should turn into decisions, from exports that comply with European regulation to carbon projects with a verifiable baseline and communities that know the value of their forest.

How BioAurea was built

Academic research

BioAurea was born from research into carbon markets that exposed a double gap. The communities with the potential to take part have no measurement, reporting and verification (MRV) technology within reach, and the tools that already exist lack the precision Colombia needs. Hence our purpose, to build Colombian technology, made for our territory and within reach of those who care for it.

AI models

We trained three carbon models, aboveground biomass, soil carbon, and necromass, with thousands of satellite data points and records from all of Colombia since 2005, calibrated with other tropical countries. We tested more than 50,000 configurations and evaluated them against the National Forest Inventory with points completely outside the training set.

Twenty satellite sources

The models integrate more than twenty satellite data sources calibrated for the Colombian territory, with automatic fortnightly updates for veredas and in real time for any coordinate.

Platform in production

Today it is a platform in production covering the country's 33,355 veredas, with published data in 32,050 in the latest snapshot. The satellite carbon base is 500 meters, and the higher-resolution layers on top of it, 10 to 30 meters, make it possible to resolve plots down to one hectare. It all reaches the user as five products, with clear reports, analysis, alerts and monthly dashboards.

Technological features

An ensemble of machine-learning models trained on multispectral and radar satellite imagery, calibrated against IDEAM's forest inventories. It estimates the four IPCC carbon pools at the vereda scale and across the entire time series.

Multisignal satellite modeling

Fuses multiple satellite and environmental data sources (spectral, climatic, topographic, and soil), combining optical and radar signals to estimate carbon even in areas with heavy cloud cover.

Complete carbon modeling

Three independent models for the IPCC pools, aboveground biomass, soil organic carbon and necromass. Belowground biomass is derived with the IPCC ratio, and everything is reported in Mg C/ha and in CO₂ equivalent.

Deep traceability

We reconstruct the history of any plot since 2005. For veredas, precomputed series every 8 days (around 959 observations). For specific coordinates, real-time queries with satellite data from the previous day.

Five integrated products

Five products on a single database. The Observatory Suite brings together the baseline, carbon flows, the change report and the calculator at vereda scale. The Carbon Explorer handles map, coordinate and polygon. The EUDR Passport issues the deforestation verdict, and the Fire Alert anticipates risk. The Impact Auditor quantifies the environmental liability of mining and hydrocarbons. Only mineral analysis became a private lab.

Anomaly detection and alerts

Automated detection of anomalous biomass changes via z scores over 20 years of history and national coverage, on a 500 m satellite base with higher-resolution layers for plot-level analysis. That is where the alerts for severe degradation, negative trend and accelerated carbon loss come from.

Calibrated artificial intelligence

Thousands of optimization experiments to tune each model. Validation with official field data from Colombia's IDEAM and IGAC, and evaluation against international benchmarks. The system updates with every new satellite observation.

How we validate, and against whom

The accuracy of a carbon model depends entirely on how it is measured. We publish the method before the result, and we publish the comparisons we do not win.

Rule 1

Against field data, not against another model

We measure against biomass and soil carbon measured by people in plots, with a published source and DOI. Comparing one map with another map is circular.

Rule 2

By spatial blocks, never at random

Neighbouring points resemble each other. Split them at random between training and test and the result comes out inflated. Here the whole block is held out.

Rule 3

A winner only if the margin excludes chance

Every comparison carries its 95 % confidence interval. If the interval touches zero, no victory is declared even when the number is favourable.

BioAurea above-ground biomass compared with ESA CCI, Spawn ORNL, CTrees and GEDI L4B on mean error and ranking ability

BioAurea soil carbon compared with SoilGrids and FAO GSOCmap on mean error and ranking ability

Deadwood: 2.83 Mg C per hectare of mean error and no comparable global product

And in Colombia, against the national forest inventory

Above-ground biomass

0.79

correlation against 953 plots of the National Forest Inventory, with 23.8 Mg C/ha of mean error.

Soil carbon

556

points from IDEAM and IGAC, evaluated holding out 56 km blocks. It is the hardest pool: veredas less than 5 km apart differ by 25 Mg/ha.

Deadwood

2.49

Mg C/ha of mean error against Colombian field data, with an R² of 0.70.

Resolution

66.8 %

accuracy in telling which of two plots 500 m or more apart holds more carbon. Chance gets 50 % and the placebo yields P below 0.0001.

Beyond 500 metres the model ranks, it does not quantify. Estimates are for pre-assessment and do not replace accredited field verification.

Total carbon by bioregion

t C/ha, sum of the four pools

Pacific205.3

Andes164.9

Amazon162.0

Caribbean113.7

Orinoquia108.5

What the map shows

The Pacific holds the highest total carbon density per hectare, driven by above-ground biomass. The Andes, with two thirds of the country's veredas, owe their total to soil rather than to trees. Amazon and Pacific are the only regions where deadwood exceeds 4 Mg C/ha.

Bioregion	Veredas	Above-ground	Below-ground	Soil	Deadwood	Total t C/ha	CO₂eq/ha
National	32,050	46.3	11.1	95.0	3.2	155.6	570.5
Pacific	938	99.8	23.9	76.2	5.4	205.3	752.9
Andes	22,532	44.6	10.7	106.4	3.2	164.9	604.5
Amazon	2,397	65.4	15.7	76.4	4.4	162.0	594.1
Caribbean	3,926	36.7	8.8	65.6	2.6	113.7	416.9
Orinoquia	2,257	37.4	9.0	59.5	2.7	108.5	397.9

Snapshot of 31 July 2026, average per vereda. Below-ground biomass is derived from above-ground with the IPCC factor. All 33,355 veredas in the country are covered; 32,050 have published data in this snapshot.

Crop detection, where open sources fall short

For the European deforestation regulation the blind spot is that shade-grown coffee has a canopy and reads as natural forest. Our detector separates coffee, oil palm, cocoa, rubber, soy, timber, natural forest and pasture at 10 metres, and in Colombia it is evaluated against IDEAM's official land-cover map holding out 55-kilometre blocks so no neighbouring point leaks information.

Coffee detection in Colombia compared with open reference sources, and multiclass accuracy in Colombia, Vietnam and Ghana

What we do claim

In Colombia the detector is national and validated: it separates coffee from natural forest without error in the spatial test. The method replicates in any country that contributes its own field truth, as it did in Vietnam and Ghana.

What we do not claim

It is not a world map that works in a country without data from that country. Where no local field truth exists we only state forest versus non-forest, which is the one thing that transfers.

Where we do not win

For oil palm in Southeast Asia we are a sensitive screen, not a fine detector: we catch 93 % but with lower specificity than the specialised detector. That is why outside Colombia palm is never asserted without a second source corroborating it.

Indicative land-use attribution. It does not constitute EU-accredited verification; the legal deforestation verdict rests on Hansen and JRC, in a separate layer.

Validated with real institutions

We presented BioAurea to 104 representatives from local governments across more than 15 municipalities of eastern Antioquia and the rest of the country. This is what they answered.

71.2% raised their interest in carbon markets after seeing BioAurea

77.9% would trust a digital platform to manage carbon projects

69.2% had an immediate positive reaction to the platform

79.8% value involving strategic partners in the process

Validation study of BioAurea with municipal representatives, 2026.

Where we're headed

Environmental agency

To become the primary source of analysis and information on Colombia's ecosystems, and an official source for the media and for government, with two editorial analyses each month.

Better technology

To keep refining the models, incorporate new sensors and field data, and narrow the margin of error until we approach direct measurement.

Colombia's communities

To put data in the hands of the communities, cooperatives and smallholders who care for the forest, so they can reach fair markets with their own information.

The environment

To contribute to conservation with evidence. Every hectare measured is a hectare that can be protected, restored and defended against deforestation.

We measure the pulse of Colombia's forests