High-tech machine learning in the greenhouse?

The challenge of modern agriculture

Modern agriculture is tasked with providing more quality food to a growing population, in a sustainable manner, while maintaining healthy margins.

Farming practices have been growing more and more intensive, with methods that are aimed at stabilizing crop quality and improving yield. Greenhouses are playing an important and growing element in this movement – offering a controlled and tempered environment that is often used for high value, high yield crops.

Even with all the advances that have been made, agriculture will not be able to meet the growing demand without a fundamental revolution, and that revolution is managing agricultural operations with accurate data and automation.

Irrigation as an example

In our experience, about 80% of farmers’ work is centered around irrigation and nutrients. The decision of how much and when to irrigate is based on numerous factors starting with the type of crop, temperature and growth stage.

It doesn’t stop there.

The soil or substrate and their ability to retain water impact irrigation frequency. Production targets such as size or quality dictate different irrigation regimes.

Surprisingly, despite the importance of irrigation and the complexity of the decision, research shows that irrigation is the farming decision least supported by data (Survey of precision-ag by Croplife and Purdue University).

Traditionally, growers have continuously toured to inspect plants and growing media for signs of stress, and rely on a wide variety of data sources such as thermometers, humidity sensors and weather forecasts to make manual decisions and adjust the irrigation schedule.

This is a highly labor-intensive and inaccurate process that is prone to errors. In open-air agriculture the time consuming manual process may lead to waste, suboptimal yields or plant disease.

But greenhouses are a different matter. They are home to extreme growing: high-value produce is intensely cultivated to drive fast growth cycles and high yields. Substrate often has low water retention properties. In such an environment there is little margin for error. Late response that leads to a small amount of under or over irrigation can have a high impact on crop yield or quality. Worse – it can kill the crops.

 

It’s time for a change

If modern agriculture is to succeed in providing quality food, in a sustainable manner, while maintaining healthy margins, it needs to change. The smart agriculture of the future will not be able to win without turning to data driven automation. In irrigation, that means highly optimized precision irrigation that detects small deviations in soil moisture and corrects them immediately. Highly accurate, real time data, powerful analytics, closed loop irrigation control, cost effectiveness and above all – ease of use are the makings of such a future.

Viridix mission is to help farmers achieve the agriculture of the future, by simplifying and automating one of the most important and time-consuming decisions in farming. The irrigation data gathering process is still labor-intensive and decisions are manual. What attempts have been made to introduce data into the process have failed due to inaccurate moisture sensing technologies and the need for constant calibration and maintenance. With such low-quality data, automation is not possible. Another issue was the high costs of implementation that canceled any positive ROI that the systems generated may have generated.

 

Viridix has been working with growers – both greenhouse and others – for the past few years on a different type of system. Low maintenance RooTense^TM water potential sensors constantly monitor soil moisture at varying ground depths. Data is transmitted over standard wireless networks to the Viridix cloud where AI and machine learning are used to generate actionable insights. Integrated irrigation control systems then execute the plan to provide plants with just the right amount of water.

The system is in use in a wide variety of crops grown in greenhouses including pineapple, bell peppers, and others. Unlike many other moisture sensors, RooTense^TM sensors work in soilless substrates. It can also be used in vertical installations and measures in multiple locations.

The system is cost-effective and easy to set up, requires no manual intervention or maintenance for daily operation and it’s solar-powered.

 

Using high-tech machine learning techniques to meet modern agriculture challenges is not science fiction. In fact, it is much quicker and simpler than you might think.