PILOT : PRECISION AGRICULTURE IN VEGETABLES_2 (POTATOES)

The goal of this pilot is to provide the potato farmers information during the growing season about the potential and actual yield predictions and the actions they can take to mitigate the foreseen yield losses. The pilot will supply the farmers with benchmark data about their crops compared to the region, previous growing seasons etc. The data provided could also be the basis for timely and more location specific treatment.

Method

The basis for the yield predictions will be the combination of data from different sources in a self-learning system. Using historical yield data and historical earth observation data, machine learning will be applied to model the potato growth and calculate yield prediction for the current year, based on recent earth observation data. The more yield (historical) yield data is supplied, the better the predictions will be. Specifically, as part of pilot solution, an online platform will be used to provide satellite imagery, weather data and yield predictions. The farmer can use the satellite imagery (biomass index, 10m resolution) to monitor and benchmark their field productivity potential relative to production levels achieved in the region. After one year, yield prediction can be implemented using the test data from year one, and other historical data (when available).

Relevance to and availability of Big Data and Big Data infrastructure

VITO has archives of earth observation (EO) data based on several satellite platforms. The recently released Sentinel 2 will be the most detailed sources, which we hope to be able to use. The study groups have collected detailed field data for some years which will useful for calibrating the system. 

Benefit of pilot

This pilot will focus on the family farms in the Veenkoloniën region in the Netherlands, which are members of the AVEBE cooperative. The main research partner will be VITO. Especially these farms will benefit from this pilot, but as a spinoff the farms will be able to grow their crops in a more sustainable way, which will be beneficial to all, farmers and the people in the whole region.

The main expected results of this pilot are:

• To identify the potential of using of satellite data and machine learning to benchmark and optimize the yield and quality of the potato crops through the development of a monitoring and yield prediction model based on weather and EO data

• To identify the potential of yield prediction to improve capacity planning and sales forecasts.

The main KPIs of the pilot are:

• Prediction quality: Evaluate the correctness of the model by testing on historic data that the system was not trained on.

• Revenue potential with potential yield vs. what happened: Quantify the value of the potential yield and the actual realization on historical data, and motivate the price of the service which would be profitable for the farmer.

• Improved ratio of realized yield to potential: Visualize the trend of yield optimization over time.