Home/Offers/Agriculture & Environment/Satellite Crop Monitoring API

Agriculture & Environment

Satellite Crop Monitoring API

Free satellite imagery covers every field on earth every few days. Turning it into something a grower or a loss adjuster can act on is the part that does not come free.

Agritech platforms
Crop insurers and loss adjusters
Agricultural lenders
Carbon and sustainability programmes
Land managers

Discuss This Build See What It Returns

What your customers would see

example interface

MockupIllustrative mockup of the interface we would build. Figures shown are examples, not results from a live deployment.

The problem this solves

There is an odd gap in agricultural software. Sentinel-2 and Landsat imagery is open, global and refreshed every few days, with enough spectral resolution to track canopy development and detect stress. Anyone can download it. Very few products turn it into field-level answers, because the pipeline between a raw scene and a usable number is longer than it looks.

The work is in the unglamorous parts. Atmospheric correction, cloud and shadow masking, field boundary intersection, resampling across sensors with different resolutions and revisit patterns, handling mixed pixels at field edges, and building a time series that is comparable week to week despite varying sun angles. Get any of those wrong and the index series shows changes that are artefacts of processing rather than events in the field.

Then there is cloud. Optical imagery sees nothing through cloud, and in maritime and tropical climates that can mean weeks without a usable observation during the period that matters most. Products that quietly interpolate across those gaps produce smooth, confident series that are partly fiction. The honest approaches are to report the gap, or to bring in radar, which sees through cloud but measures something different and needs its own interpretation.

Who buys this

Agritech platforms, Crop insurers and loss adjusters, Agricultural lenders, Carbon and sustainability programmes, Land managers

Category: Agriculture & Environment

What the API returns

What comes back

Vegetation index time series per field — NDVI, NDRE, EVI and others
Within-field variability maps and management zone delineation
Crop stress detection with change-point timing
Cloud and quality masking with explicit observation coverage reporting
Radar-based observations for periods where optical imagery is unusable
Field change detection — cultivation, harvest, cover crop establishment
Crop type classification where sufficient training data exists
Imagery tiles and index rasters for your own mapping layer

Endpoints

Method	Path	Returns
POST	/v1/fields/indices	Vegetation index time series for a set of fields
GET	/v1/fields/{id}/variability	Within-field variability and management zones
GET	/v1/fields/{id}/stress	Detected stress events with timing and extent
GET	/v1/fields/{id}/changes	Detected field events — cultivation, harvest, cover
POST	/v1/fields/classify	Crop type classification for a set of field geometries
GET	/v1/fields/{id}/raster	Index raster or imagery tile for map display

Request and response

illustrative

Request

{
  "fields": [
    { "field_id": "f_22181", "geometry": { "type": "Polygon", "coordinates": "..." } }
  ],
  "indices": ["ndvi", "ndre"],
  "from": "2026-03-01",
  "to": "2026-06-20",
  "include": ["quality", "radar_fallback", "variability"]
}

Response

{
  "field_id": "f_22181",
  "area_ha": 41.2,
  "observations": [
    { "date": "2026-05-04", "sensor": "sentinel2", "ndvi_mean": 0.812, "ndre_mean": 0.431,
      "usable_pixel_pct": 97.4, "quality": "good" },
    { "date": "2026-05-09", "sensor": "sentinel2", "usable_pixel_pct": 11.2,
      "quality": "cloud_obscured", "ndvi_mean": null },
    { "date": "2026-05-11", "sensor": "sentinel1", "vh_backscatter_db": -14.8,
      "quality": "radar_substitute",
      "note": "Radar observation during 17-day optical gap; not directly comparable to NDVI" },
    { "date": "2026-05-21", "sensor": "sentinel2", "ndvi_mean": 0.744, "ndre_mean": 0.388,
      "usable_pixel_pct": 94.1, "quality": "good" }
  ],
  "coverage": { "requested_days": 112, "usable_optical_observations": 9,
                "longest_optical_gap_days": 17, "gaps_filled_with_radar": 2 },
  "stress_events": [
    { "detected_between": "2026-05-04/2026-05-21", "index_drop": 0.068,
      "spatial_extent_pct": 34.0, "pattern": "south_west_corner",
      "confidence": "medium",
      "caveat": "17-day observation gap means onset timing is uncertain to within two weeks" }
  ],
  "variability": { "cv_within_field": 0.118, "zones_suggested": 3 }
}

Where the data comes from

Sentinel-2	Open optical imagery at 10 to 20 metre resolution with roughly five-day revisit. Free to use including commercially, with a published attribution requirement that must appear in your product
Landsat	Open optical imagery from USGS with a long historical archive, valuable for multi-decade baselines. Also free with attribution
Sentinel-1	Open radar imagery, unaffected by cloud. Essential in cloudy climates and measures surface structure and moisture rather than greenness, so it is a substitute with caveats rather than a replacement
Commercial high-resolution imagery	Sub-metre providers where field detail genuinely requires it. Licensed per area or per scene, usually with firm restrictions on redistribution and on derived-product resale
Digital elevation models	Open terrain data for slope, aspect and drainage context when interpreting within-field variability
Field boundary data	From your platform, from open cadastral or parcel registers where licensing allows, or derived by segmentation from imagery

On data licensing. Being allowed to read a dataset does not automatically mean you may redistribute it, resell it through an API, or train a model on it. Before any build starts we check the licence, terms and derived-data rules of every source, and we design around the ones that do not permit your intended use. Where a commercial licence is required, that cost is identified during discovery rather than discovered later.

The intelligence layer

The difference between a demo and a service is mostly in this section. These are the parts that keep the answer trustworthy once real traffic is hitting it.

Full preprocessing chain built explicitly — atmospheric correction, cloud and cirrus and shadow masking, topographic correction — because downstream analysis that skips these measures processing artefacts
Per-pixel quality flags aggregated to a field-level usable-pixel percentage, with observations below a threshold returned as unusable rather than as a mean over the clear corner
Observation gaps reported as gaps. Interpolation is offered as an explicitly labelled derived series, never as the default, since a smoothed curve through a three-week cloud period invents the data that matters
Sentinel-1 backscatter used as a cloud-period substitute with its own interpretation, not rescaled into a pseudo-NDVI that would imply comparability it does not have
Field-edge pixel exclusion with a configurable buffer, because mixed pixels at boundaries contaminate field means more than most users expect
Cross-sensor harmonisation between Sentinel-2 and Landsat, with the residual bias stated rather than assumed away
Change-point detection on the index series for stress and event timing, with timing uncertainty widened in proportion to the surrounding observation gap
Monitoring of scene availability and processing baseline changes — reprocessing of the archive by the data provider shifts historical values and must be tracked

What it changes for the business

Loss adjustment speed

Index history and change timing give an adjuster evidence before a site visit, and sometimes instead of one.

Portfolio-scale monitoring

An insurer or lender can watch every field in a book rather than sampling by inspection.

Agronomy targeting

Within-field variability identifies where a problem is, which is what turns an observation into an action.

Verification

Independent observation of cultivation, cover crop and harvest timing supports scheme and contract compliance.

How it fits into your product

How it embeds

Field geometry as GeoJSON, or boundaries derived by segmentation where you do not hold them
Batch endpoints for portfolio-scale processing across thousands of fields
Raster and tile output for your existing mapping layer, with index colour ramps
Scheduled processing triggered by new scene availability rather than a fixed calendar
Webhooks on stress detection and field change events
White-label delivery, with the satellite attribution requirements handled correctly in your interface

Included in every build

Authentication, scoped API keys and per-customer rate limiting
Input validation with predictable, documented error responses
Versioned endpoints so your integration never breaks without notice
Request and response logging for audit and dispute resolution
Latency, error-rate and usage monitoring with alerting
OpenAPI specification, sandbox environment and client examples
Backups, failover design and a documented recovery procedure
An agreed support arrangement once the service is live

How we would build it

about 20 weeks

Discovery and feasibility

Weeks 1–3

This one depends heavily on data access, so we start by proving the sources exist, can be licensed for your use, and actually carry the signal the model needs. If they do not, you find out here.

Architecture and baseline model

Weeks 4–6

A written architecture and a baseline model measured against a holdout period, so you know what accuracy you are buying before the full build starts.

Production build

Weeks 7–16

Data pipelines, model training and serving, historical archive, validation harness, API, security, monitoring and documentation.

Validation and integration

Weeks 17–20

Backtesting against held-out history, accuracy reporting your team can audit, integration support and an agreed retraining cadence.

Commercial models

There is no fixed price list, because the cost is driven by data complexity, integration count, model requirements, request volume and the support level you need. Scope is agreed in writing before development starts.

Fixed-scope prototype

A defined slice of the service, built and priced as one piece of work. The usual first step when the business case needs proving internally.

Fixed-price build

An agreed scope against a written architecture, delivered for an agreed price. Best when the requirements are already clear.

Build then retainer

We build the service, then operate and improve it under a monthly agreement covering monitoring, updates and model maintenance.

White-label product

The service runs under your brand, your domain and your pricing, with MLPivot invisible to your customers.

Questions buyers ask

What happens during extended cloud cover?

You get fewer observations, and in maritime or tropical regions that can mean two to four weeks with nothing usable — often during the growth period you most wanted to watch. We report the gap explicitly: observation counts, longest gap, and usable pixel percentage per scene. We do not interpolate across it by default, because a smooth curve through a three-week hole looks like data and is not. If you want an interpolated series it is available as a separately labelled field.

When is radar the right answer?

When cloud persistence makes optical monitoring unreliable — persistent maritime climates, monsoon seasons, high-latitude winters. Sentinel-1 sees through cloud and gives useful signals on canopy structure, biomass and surface moisture, plus reliable detection of events like harvest and cultivation. What it does not give is a greenness measure comparable to NDVI. We return it as its own quantity with its own interpretation rather than converting it into a number that looks like the optical series.

Can we resell the imagery to our customers?

Sentinel and Landsat are free to use commercially, including in a paid product, provided the attribution requirements are met and visible — we build that into the interface rather than leaving it to you to remember. Commercial high-resolution imagery is different: those licences are typically per-area, frequently forbid redistribution, and some restrict derived products as well. Permission to read a scene is not permission to resell it, and if your product needs to display sub-metre imagery that becomes a licensing conversation before it becomes an engineering one.

How small a field can this handle?

Sentinel-2 at 10 metres gives a usable field mean from roughly one hectare upward, allowing for an edge buffer. Below that, mixed pixels dominate and the number stops meaning much. Smallholder monitoring at a fraction of a hectare genuinely needs commercial high-resolution imagery, with the cost and licensing that brings. We would rather tell you that than return a confident mean over four pixels.

Can you identify the crop in a field?

Where there is enough labelled training data for that region and cropping system, reasonably well for major crops. Where there is not, accuracy drops sharply and mixed or minor crops are frequently misclassified. We report per-class accuracy from validation rather than a single headline figure, and for regions without training labels we would say classification is not supportable rather than shipping one.

All Agriculture & Environment offers →

Want this built for your platform?

Tell us what your product does today and what you want another system to be able to ask it. We will come back with an approach, a scope and a price.

Discuss This Build Browse All Offers