Automatic water-quality control for good recirculation water in greenhouse horticulture

Greenhouse horticulture irrigation water, when recirculated, acquires a complex composition. The impact this has on crops is for the most part unknown. The possibility of determining whether the water quality is good or poor is essential for the good management of the irrigation water system, with a view to plant health and the continuous operation of the water system. This project therefore aims to develop a water-quality model on the basis of plant, water and climate sensors, with control measures for the purpose of producing a healthy crop. Control threshold values will be studied in operational and pilot greenhouses, by connecting water parameters to plant growth, climate and microbiological parameters using machine learning. A user-friendly dashboard will be created for the presentation of information and control parameters.


Water-quality sensors can be used for the rapid and autonomous measurement of water quality. There is a wide variety of such sensors available, and several of them are have already been installed and tested for application in greenhouse horticulture [e.g., TKI project ‘Waterkwaliteit in Beeld’ (Picturing Water Quality)]. This experience has shown that although sensors do produce information about the water composition, no single individual sensor is able to assess whether irrigation water is of good quality, because it is the total of all parameters that establishes whether the water quality is sufficient for reliable reuse. A combination of sensors does however make it possible to measure a ‘fingerprint’ of the water. Experience in other sectors has shown that such a fingerprint can be used to reach conclusions about water quality, which would not be possible on the basis of the individual sensors. This project will investigate the possibility, through the use of a combination of water-quality, climate and plant sensors, of producing a decision-support measurement system, in which, on the basis of the monitoring of the water-quality fingerprint, guidance can be provided for the control and adjustment of the recirculation process and water treatment based on a set of threshold values.


Although there are guidelines for good irrigation water that define concentration levels for nutrients and trace elements for specific crops, there is not enough information about when water is good or healthy for the plant for irrigation through recirculation. Water quality is today a ‘black box’ for horticulturalists, and there is not enough information to permit the timely detection of problems.


The project will research the criteria for good water quality. This will be done in horticulture practice, by connecting sensors for water parameters and climate to sensors for plant growth and production. In addition, experiments will be carried out at pilot sites to establish causal connections between sensor data for water quality and plant growth. With the application of ‘deep learning’, correlations will be derived between the measured parameters, and a model will be developed which will be able, on the basis of water-quality measurements, to predict the effects on plant health and, drawing on this, also provide guidance for the water treatment applied during recirculation. The water-quality data from the sensors will be complemented with discontinuous data sources, such as chemical and microbiological laboratory analyses, including Next Generation Sequencing results.


This project will contribute to increasing the efficiency of water recirculation in greenhouse horticulture, by providing the horticulturalist with the tools to safeguard and manage the quality of irrigation water under conditions of full reuse. To this end, the project will develop a user-friendly, decision-support system (dashboard), so that the end-user (horticulturalist), on the basis of measurement values and threshold values, will have information about the water quality and support in the control of recirculation and water treatment processes. The dashboard will be based on the integrated model developed in this project.