The benefits of open data in the agriculture and forestry sector: the case of Fruktia and Arbaria

As in other industries, digital transformation is helping to change the way the agriculture and forestry sector operates. Combining technologies such as geolocation or artificial intelligence and using open datasets to develop new precision tools is transforming agriculture into an increasingly technological and analytical activity.

Along these lines, the administrations are also making progress to improve management and decision-making in the face of the challenges we are facing. Thus, the Ministry of Agriculture, Fisheries and Food and the Ministry for Ecological Transition and the Demographic Challenge have designed two digital tools that use open data: Fruktia (crop forecasting related to fruit trees) and Arbaria (fire management), respectively.

Predicting harvests to better manage crises

Fruktia is a predictive tool developed by the Ministry of Agriculture to foresee oversupply situations in the stone fruit and citrus fruit sector before the traditional systems of knowledge of forecasts or gauges. After the price crises suffered in 2017 in stone fruit and in 2019 in citrus fruit due to a supervening oversupply, it became clear that decision-making to manage these crises based on traditional forecasting systems came too late and that it was necessary to anticipate in order to adopt more effective measures by the administration and even by the sector itself that would prevent prices from falling.

In response to this critical situation, the Ministry of Agriculture decided to develop a tool capable of predicting harvests based on weather and production data from previous years. This tool would be used internally by the Ministry and its analysis would be seen at the working tables with the sector, but would not be public under any circumstances, thus avoiding its possible influence on the markets in a way that could not be controlled.

Fruktia exists thanks to the fact that the Ministry has managed to combine information from two main sources: open data and the knowledge of sector experts. These data sources are collected by Artificial Intelligence which, using Machine Learning and Deep Learning technology, analyses the information to make specific forecasts.

The open datasets used come from:

• Information from weather stations of the Spanish Meteorological Agency (AEMET).
• Information from agro-climatic stations.

With the above data and statistical data from crop estimates of past campaigns (Production Advances and Yearbooks of the Ministry of Agriculture, Fisheries and Food) together with sector-specific information, Fruktia makes two types of crop predictions: at regional level (province model) and at farm level (enclosure model).

The provincial model is used to make predictions at provincial level (as its name suggests) and to analyse the results of previous harvests in order to:

• Anticipate excess production.
• Anticipate crises in the sector, improving decision-making to manage them.
• Study the evolution of each product by province.

This model, although already developed, continues to be improved to achieve the best adaptation to reality regardless of the weather conditions.

On the other hand, the model of enclosures (still under development) aims to:

• Production forecasts with a greater level of detail and for more products (for example, it will be possible to know production forecasts for stone fruit crops such as paraguayo or platerina for which we currently do not have information from statistical sources yet).
• Knowing how crops are affected by specific weather phenomena in different regions.

The model of enclosures is still being designed, and when it is fully operational it will also contribute to:

• Improve marketing planning.
• Anticipate excess production at a more local level or for a specific type of product.
• Predict crises before they occur in order to anticipate their effects and avoid a situation of falling prices.
• Locate areas or precincts with problems in specific campaigns.

In other words, the ultimate aim of Fruktia is to achieve the simulation of different types of scenarios that serve to anticipate the problems of each harvest long before they occur in order to adopt the appropriate decisions from the administrations.

Arbaria: data science to prevent forest fires

A year before the birth of Fruktia, in 2019, the Ministry of Agriculture, Fisheries and Food designed a digital tool for the prediction of forest fires which, in turn, is coordinated from the forestry point of view by the Ministry for Ecological Transition and the Demographic Challenge.

Under the name of Arbaria, this initiative of the Executive seeks to analyse and predict the risk of fires occurring in specific temporal and territorial areas of the Spanish territory. In particular, thanks to the analysis of the data used, it is able to analyse the socio-economic influence on the occurrence of forest fires at the municipal level and anticipate the risk of fire in the summer season at the provincial level, thus improving access to the resources needed to tackle it.

The tool uses historical data from open information sources such as the AEMET or the INE, and the records of the General Forest Fire Statistics (EGIF). To do so, Artificial Intelligence techniques related to Deep and Machine Learning are used, as well as Amazon Web Services cloud technology.

However, the level of precision offered by a tool such as Arbaria is not only due to the technology with which it has been designed, but also to the quality of the open data selected.

Considering the demographic reality of each municipality as another variable to be taken into account is important when determining fire risk. In other words, knowing the number of companies based in a locality, the economic activity carried out there, the number of inhabitants registered or the number of agricultural or livestock farms present is relevant to be able to anticipate the risk and create preventive campaigns aimed at specific sectors.

In addition, the historical data on forest fires gathered in the General Forest Fire Statistics is one of the most complete in the world. There is a general register of fires since 1968 and another particularly exhaustive one from the 1990s to the present day, which includes data such as the location and characteristics of the surface of the fire, means used to extinguish it, extinguishing time, causes of the fire or damage to the area, among others.

Initiatives such as Fruktia or Arbaria serve to demonstrate the economic and social potential that can be extracted from open datasets. Being able to predict, for example, the amount of peaches that fruit trees in a municipality in Almeria will yield helps not only to plan job creation in an area, but also to ensure that sales and consumption in an area remain stable.

Likewise, being able to predict the risk of fires provides the tools for better fire prevention and extinction planning.

Content written by the datos.gob.es team