Making the unpredictable
more predictable
From frost forecasting and flower-quality assessment to production planning, fci explores how AI is being developed and applied in Colombian floriculture.
By Marta Pizano
Weather-sensitive production and tightly timed export markets make predictability particularly valuable in floriculture. In Colombia, researchers and growers are exploring how AI—including image analysis and predictive modelling—can help anticipate frost, forecast harvest volumes and assess flower quality.
The work extends beyond the growing area. Linking information from these systems with inventories and sales could help businesses anticipate shortages and surpluses, while giving staff better information for day-to-day decisions.
Chrysanthemum production in Colombia, where computer-vision research is exploring ways to forecast flower harvests. Photo: Capiro/Universidad de Antioquia.
Anticipating production challenges
How is AI transforming modern floriculture? The integration of AI into Colombian floriculture is an exciting journey. Take, for instance, the progress being made with AI-generated models that help predict weather conditions with a high degree of accuracy, allowing growers to anticipate frost episodes, which can cause significant damage to flowers.[1]
Roses are particularly sensitive to low temperatures, and these conditions often occur shortly before Valentine’s Day, arguably the most important season for Colombian rose exporters. With AI-assisted predictions, preventive measures can be taken to minimise damage, including sprinkler irrigation, thermal covers and mulching.
Precision fertigation, combining AI with specialised sensors, can be used to help determine exact quantities of water and fertiliser required by each crop at different stages of the production process and according to prevailing environmental conditions. This can lead to improved production efficiency, better crop quality, greater energy efficiency and reduced environmental impacts.
Using image analysis and computer vision, either through drones or cameras, plant development, harvesting times and flower volumes can be estimated with a high degree of precision. This offers significant advantages to exporters targeting specific peak consumption dates, such as Valentine’s Day and Mother’s Day. Demand for specific flowers, and even particular varieties or colours, can therefore be met more efficiently, starting from the early stages of the production process. [2]
Specialised algorithms applied to image analysis can detect pests or diseases at very early stages, allowing preventive or corrective measures to be implemented quickly. This can help minimise pesticide use and reduce crop losses.[3]
We are able to detect differences in water, botrytis infection and carbohydrate content in roses, chrysanthemums and alstroemeria, which result in differences in vase life
Dr John Dole North Carolina State University
Dr Johan Ramirez Gil and Jiyoung Ham image roses for research linking spectral measurements with hydration, sugar content and disease status. Photo: Tatiana Osorio Montoya.
What is hyperspectral imaging?
An advanced imaging technology that combines digital photography and spectroscopy.
Assessing flower quality Hyperspectral imaging
FloraCulture International approached Dr John Dole of North Carolina State University (Raleigh, NC, USA), who has been working on the use of AI in Colombian flower farms for several years.
Although the volume of data generated is extremely large, making data analysis a lengthy process, Dr Dole kindly provided an overview of his research findings: “Using hyperspectral imaging, we are able to detect differences in water, botrytis infection and carbohydrate content in roses, chrysanthemums and alstroemeria, which result in differences in vase life.”
This allows for the early detection of water or carbohydrate deficiencies, as well as the presence of botrytis, enabling timely and appropriate pre- and post-harvest treatments. These applications can have a direct impact on flower quality and post-harvest performance.
Furthermore, Dr Dole’s team is using AI and multispectral imaging to create a system capable of detecting substandard flowers. This offers significant commercial benefits; such a system could allow flower quality to be assessed quickly and efficiently, giving growers immediate information about what is available for sale at a given point in time.
Dr John Dole, North Carolina State University, is researching imaging technologies to assess cut-flower quality.
PhD students tests flower samples for carbohydrate and water content as part of research into assessing post-harvest quality. Photo: John Dole.
Planning with better information La Gaitana Farms We asked La Gaitana Farms, the 2025 AIPH International Grower of the Year, about its current use of AI in its operations.
The farm’s priority is to develop tools that connect information across different areas of the operation. AI and advanced analytics help strengthen operational planning, maintain clear and up-to-date product inventories, support commercial analysis and provide staff with better information for decision-making.
The company is “moving progressively from a management model that traditionally depended heavily on manual analysis, toward an operation that is more preventive, agile and supported by data”.
In their view, crop production is one of the areas where AI offers significant potential. Traditionally, production planning relied on the experience of technical staff, historical production curves and continuous crop monitoring. Now, the company is developing a predictive model that combines historical production data, current estimates, climate variables and operational information to estimate exportable production by variety and colour.
The goal is to anticipate potential shortages or surpluses before they occur. These improved and more accurate forecasts can allow sales staff to plan resources more effectively and respond more quickly to customer requirements.
The post-harvest stage is another area where tangible applications of AI are emerging. Not only can logistics and traceability be positively impacted, but bouquet design, visual concepts and communication materials can also be enhanced.
La Gaitana Farms is actively developing AI capabilities across the organisation, with increasing numbers of employees receiving training through its “Strategic AI: From Trend to Execution” programme. Its goal goes beyond learning how to use AI tools; it is focused on turning these tools into practical business solutions. This is not an AI intelligence project, but rather a gradual transformation in the way the operation is managed: more accurately anticipating available inventories, measuring performance more closely, connecting different roles and operational segments within the organisation through robust information systems, anticipating scenarios and responding to demand.
The company emphasises that AI is by no means a replacement for the knowledge and experience on which it has built its business. Export floriculture is a complex activity in which agronomic expertise, operational experience, commercial understanding, personal relationships and good judgement remain essential.
With this in mind, AI provides a way to combine that experience with data, analytical capabilities and new technologies.
Post-harvest operations at La Gaitana Farms, where the company is developing AI tools to support planning and decision-making. Photo: La Gaitana Farms.
An industry-wide perspective Connecting technology and grower knowledge Finally, we asked ASOCOLFLORES, the Colombian Association of Flower Exporters, for a broader perspective on how AI is currently affecting the Colombian flower sector.
Anderson Paez, director of the cut flower innovation centre Ceniflores, explained that they recently carried out a study on “Vigilance and competitive intelligence on automatisation technologies in the floriculture sector”. It found that AI and related technologies are already beginning to have an impact on several parts of the cut flower value chain.
One of the clearest areas of development is precision agriculture. Sensors, digital platforms, imaging and data analysis are increasingly being used to monitor climatic conditions and plant nutrition. This allows growers to adjust irrigation, fertigation and other crop management practices according to the specific needs of a crop, or even of individual cultivars within a crop.
Water and fertiliser management is one example where these technologies can make a tangible difference. Commercial solutions such as GrowSphere from Netafim combine real-time monitoring with automated fertigation.
The Ceniflores study found that these systems can reduce water consumption by up to 30 per cent and fertiliser use by up to 25 per cent, without affecting yields.
Another important development is the early detection of plant health problems and diseases. Artificial vision systems combined with predictive analysis can help identify symptoms, detect insects and recognise infections at an early stage. Robots capable of applying treatments in a targeted way have also been developed, while other solutions are currently being explored.
AI and digital technologies are also opening up new possibilities for traceability and decision-making. A particularly interesting example is GHT (Growers Hub Trading), a Colombian company that combines intelligent business platforms, RFID technology and AI tools to centralise information from its farms. This makes it possible to trace flower stems from planting all the way through to sales. At the same time, the information can support better production planning, help anticipate plant health problems, estimate production volumes and improve distribution.
In logistics and inventory management, radio-frequency identification (RFID) technology is being used to track reusable flower boxes, follow orders and monitor product movements throughout the logistics chain. It also helps reduce errors associated with manual record-keeping. The study highlights the example of Dutch-based Hoek Group, which uses this technology to manage around 100,000 reusable boxes.
The impact of these technologies extends beyond production and logistics into post-harvest operations and marketing. Applications include automatic grading, bouquet production, internal transportation, quality control and demand forecasting. By automating and improving these processes, AI can help reduce losses, shorten the time between harvest and sale, and maintain flower quality throughout the cold chain.
Paez stresses that AI should not be viewed as a standalone technology in this context. Rather, it forms part of a broader technological ecosystem that includes sensors, the Internet of Things, robotics, artificial vision, cloud platforms, RFID and data analysis.
For floriculture, the real opportunity lies in bringing these technologies together with the agronomic knowledge and practical experience of production teams.
