Predicting Crops – From Brazil to Thailand

Agriculture is one of the oldest means of food production. It has been a vital part of human civilization since time began.

Historically it was impossible to tell if it would rain later that day, let alone predict the weather in advance. All humanity could do was make vague assumptions based on the clouds, animal behaviours and praying to their gods.

Today we not only do we predict the weather to within 1 degree but we can also see crops growing around the world, predicting yields and tracking harvests.


Satellites have revolutionized the farming industry  – from weather satellites enabling the highly accurate predictions to near-perfect mapping services via GPS.

Satellite imagery is another way in which technology is transforming farming. This data is now being used to understand, monitor, and predict crops.

Fusing together images from different satellites and using supervised machine learning models enables highly precise information to be obtained about crops. By building an accurate picture of crops , farm by farm, is possible to know exactly what is being grown, across an entire region and plan accordingly.

Easy v Hard

Areas such as Brazil, have high-quality data about their crops. This is because it is much easier to collect data – with large farms, heavy use of drones is possible, and satellite analysis is easier. 

Other areas, such as Thailand have far more challenges for a combination of reasons: Farms in Thailand are small, complex, and routinely covered in the cloud.

Small and Complex Farms

The average size of a Thai farm is very small compared to Brazil. In Thailand, a typical farm is just 4 hectares (around 10 acres) – in Brazil, most farms are over 200 hectors (500 acres) and nearly 40% are over 400 hectares (1000 acres).

The farms in Thailand are also complex – they have mixed crops – growing sugar, rice, and cassava alongside each other. They become even more complex as the crops can change every year – with no fixed boundaries. This means that mapping fields via satellite is not an effective method in an area like Thailand. 


Finally – to make Thailand even more challenging is the fact is also has a lot of cloud cover.  All of these factors combined mean that satellite analysis of Thailand’s crops has not been effective.

Pixel by Pixel. Satellite by Satellite

Technology, computing power, and analytics are improving, which means that new methods can be deployed.

The issue of cloud cover can be negated by using more satellites and radar data (which can see through the cloud) and combining them.  The fused images have an increased “temporal resolution” and increased “spatial resolution”.  In other words more frequent images of better quality.

With more frequent images the the “gaps in the cloud”  can be combined together to produce a single image. This means that the cloud is effectively removed.

With the increased data quality the fields can now be analyzed at a pixel by pixel level, rather than trying to map out ever-changing polygons onto the fields.

This combination of data and analytics can result in predictions with over 96% to 98% accuracy – for complex places like Thailand and China.