evidoria

← Back to browse

Good practice Imported

Controles por Monitorización — Castile and León Cuts CAP Field Inspections to 0.2% of Parcels With Satellite AI

Spain · Valladolid · See the Spain profile

Evidence: Observational / pre–post Top 22% 73/100 · Ask Evidence Copilot about this practice

Spain's Castile and León region classifies farm parcels from Sentinel-2 time series with machine learning across 26 crop and 9 non-crop types, replacing most physical CAP inspections. An EU Court of Auditors audit found only 0.2% of parcels still needed a field visit.

0.2 %
Parcels requiring an on-the-spot field inspection (2019)
9
CAP aid schemes covered by the monitoring method
26 crop / 9 non-crop
Crop and non-crop classes distinguished by the classifier

Details

Maturity
Established
Promoter
ITACyL (Instituto Tecnológico Agrario de Castilla y León), Junta de Castilla y León
Period
2016–present
Region (NUTS)
ES41
Keywords
agriculture, machine learning, compliance, earth observation

Context

Since 2016 Castile and León's agricultural technology institute, ITACyL, has combined Sentinel-2 satellite imagery (upgraded to a 10-metre, near-daily product from 2017) with LIDAR elevation, slope, rainfall and the region's SIGPAC land-parcel registry to classify farmland automatically. A machine-learning algorithm assigns each declared parcel to one of 26 crop classes or 9 non-crop classes, feeding Spain's 'Controles por Monitorización' scheme for verifying Common Agricultural Policy claims.

Results

The European Court of Auditors' 2020 special report names Castile and León as one of the first paying agencies to run the method at scale, applying it across nine CAP aid schemes; its audit table records that only 0.2% of the region's parcels required an on-the-spot inspection, versus a traditional EU sampling rate of around 5%.

Conclusions

As of the 2019/2020 audit, only 15 of 66 EU paying agencies had operationalised checks-by-monitoring, and most early deployments — including Spanish ones — applied it only to certain schemes and farmer groups, so full national coverage across all Spanish regions had not yet been achieved.

Implementation
Indicative cost
Medium (€50k–€500k) — Run by a regional public technical institute since 2016; no specific budget figure disclosed in sources.
Time to results
Long (> 3 years) — Piloted from 2016, upgraded to near-daily imagery in 2017, operating across nine CAP schemes by the time of the 2019/2020 EU audit.
Staffing & skills
ITACyL (Instituto Tecnológico Agrario de Castilla y León) technical staff, Junta de Castilla y León regional agriculture administration

Conditions for success

  • Near-daily 10m Sentinel-2 imagery (available from 2017)
  • Integration with the SIGPAC land-parcel registry and LIDAR/slope/rainfall data
  • Machine-learning classification pipeline for 26 crop and 9 non-crop classes

Where it fits

Governance type
regional public agency
Scale
regional
Income level
high-income

Commonly funded by

National / regional programmes

Indicative funding routes for practices of this type — always check each programme's current calls and eligibility rules.

Do you run this practice? Claim it — verified implementers get a public contact pathway and can propose corrections.

Data sources

Where this practice's information was retrieved from, and when.

Similar practices you may find useful