Precision agriculture, redistributed
The tools of industrial agriculture, made available to smallholders, indigenous territories, and food-sovereignty projects.
The asymmetry
Industrial agribusiness has spent the past two decades layering satellite imagery, drone-based crop monitoring, soil-sensor networks, and machine-learning analytics on top of every hectare it farms. A multinational palm-oil operator in the Llanos can tell you the NDVI trajectory of an individual plot, the soil-moisture deficit by sub-block, and the yield forecast for the next harvest. A 3-hectare smallholder one valley over — running the same crop on the same soils under the same climate — has none of this. The technology is not the bottleneck. The bottleneck is access, affordability, and translation.
Capybara's wager: most of the analytic capability that industrial-scale agriculture pays for is available, for free, on Google Earth Engine and similar platforms. What is missing is the bridge — the scripts, the workflows, the training materials, and the partnerships that put these tools in the hands of smallholders, indigenous farming communities, agroforestry projects, and food-sovereignty initiatives. We build that bridge.
How we redistribute these methods
1 — Free satellite imagery as the equalizerSentinel-2 (10 m, every 5 days), Landsat (30 m, every 16 days), MODIS, PML/V2 GPP, CHIRPS precipitation, SMAP soil moisture — all open, all free, all running on Google Earth Engine. A smallholder cooperative with one technically-inclined member can monitor crop vigour, water stress, and forecast risk for hundreds of member plots without a single subscription fee. We provide the scripts and the workflows. |
2 — Affordable drone analytics through pooled accessDrone surveys are expensive per-individual but cheap per-cooperative. We work with farmer cooperatives, university extension programmes, and NGOs to pool drone access — one campaign covering many smallholder plots, with the imagery processed centrally and the per-plot reports delivered back to each producer. The economics that work for a 5,000-hectare estate also work for 200 × 25-hectare members of a cooperative. |
3 — Community-led ground truthSoil sampling, agronomic observation, traditional ecological calendar, generational knowledge of which plot floods first and which dries first — these are forms of ground truth that no satellite carries. We design field-sampling protocols community members themselves can run, pair them with lab analysis where useful, and feed the results back into the same dashboards. Indigenous, traditional, and smallholder knowledge enters as data, not as anecdote. |
4 — Climate-risk integrationThe Capybara flood, wildfire, and heatwave models already map crop exposure as a first-class output. For partner cooperatives we plug the model surfaces directly into the precision-ag workflow — anticipating climate shocks at the plot scale weeks in advance, and pairing the warning with the insurance, credit, and harvest-timing decisions that can mitigate it. |
Who this is for
| Smallholder cooperatives & producer associations | Coffee cooperatives, cocoa associations, smallholder fruit and vegetable producers, sugar / panela mills with smallholder supply chains. The unit-economics work when costs are shared across a member base. |
| Indigenous territories & traditional agricultural societies | Indigenous authorities, traditional resguardos, palenque and quilombo communities, ancestral agroforestry systems. We bring satellite imagery and modelling, the community brings the ecological calendar and the ground reality. The pairing is the analysis. |
| Agroforestry & food-sovereignty projects | Polyculture systems, agroecological transitions, peri-urban gardens, coca-substitution programmes, post-conflict reactivation projects. Standard precision-ag tools assume monoculture; we adapt the methods for diverse, mixed, and intercropped landscapes. |
| University extension & NGO partners | Agronomy schools, agricultural extension agencies, climate-adaptation NGOs working with farming communities. Capybara provides the open code, the methodology, and the volunteer support; the partner brings the local relationships and the agronomic expertise. |
What traditional agricultural knowledge teaches our models
The Bayesian flood model's per-basin precipitation deviation, the fire model's burnable-fuel mask, the heatwave model's shelter identification — each carries threshold values and weighting choices that were originally informed by, and are continually refined against, the knowledge held by farmers and territorial communities in the regions Capybara works in. We attribute these contributions in the situation reports and the model documentation. Indigenous and smallholder knowledge is treated as data, with the same epistemic standing as a satellite observation.
Track record & lineage
|
SIGETI — indigenous territorial governance 2022 · Gaia Amazonas Foundation |
The participatory data platform for indigenous authorities in the Colombian Amazon. Combined remote-sensing analytics with traditional knowledge for climate adaptation and territorial governance. The methodological template for how Capybara now works with non-industrial agricultural systems. |
|
Capybara flood & wildfire models — crop exposure 2024 — ongoing · Open code |
The current model family already reports per-pixel cropland exposure (MODIS land-cover + ESA WorldCover), oil-palm exposure (BIOPAMA), and protected-area exposure for any country. The infrastructure to monitor agricultural climate risk at the smallholder scale already exists in the codebase — what's missing is the partnerships to deploy it. |
|
Coffee roasting analytics — Jaguar Coffee Co. 2018 — 2019 · Specialty coffee startup |
Earlier sensor-data analytics for specialty-coffee quality. Demonstrated that data-driven optimisation can move cleanly between domains — from roasting curves to remote-sensing analytics — when the underlying methods stay open. |
What we don't do
| Sell data back to the farmers who produced it | Any field data a community contributes belongs to the community. We do not aggregate-and-resell, do not license downstream, and do not subscribe-gate the analyses. |
| Substitute algorithms for traditional knowledge | A satellite NDVI time series does not replace generations of agronomic knowledge about a specific landscape. The model layers are additions to local expertise, never replacements. |
| Take industrial-monoculture client work that displaces smallholders | We decline engagements where the analytic output would be used against the smallholder communities Capybara is built to serve. |
| Pretend a dashboard alone changes outcomes | Analytic infrastructure is necessary but not sufficient. Without partner cooperatives, extension agronomists, fair credit access, and supportive policy, even the best dashboard sits unused. Capybara works only inside partnerships that include these surrounding elements. |
Partner with us
Capybara works through partnerships — never as a standalone precision-ag vendor. Typical engagement scopes:
- A cooperative-scale crop-monitoring system built on free satellite imagery, with member-plot reporting.
- Climate-risk integration for a producer association — pairing flood / wildfire / heatwave model output with crop calendars and member geography.
- A participatory data platform with an indigenous-territory authority for ancestral agroforestry monitoring.
- Capacity-building workshops for university extension agronomists or NGO field staff on the Capybara open methods.
Capybara Center · Precision agriculture, redistributed · MIT-licensed open methods · Built on Google Earth Engine
Note: Only a member of this blog may post a comment.