Draft: The #datadriven Grower Wiki
Draft: The #datadriven Grower Wiki

Draft: The #datadriven Grower Wiki

A primer on data-driven growing and technologies for optimal crop production in the greenhouse

Foreword. This book is designed to provide the technical training and decision support tools to greenhouse growers. The book is written for growers ... by a collective group of plant scientists, physicists, technologists, and growers. WIP: the book is a work-in-progress and still in a very-early stage. If you would like us to prioritize the writing of certain topics, don't hesitate to send us a request to [email protected]

3. Crop growth and response to environmental factors

3.1. Main biological processes in plants

3.2. Carbohydrate distribution and plant growth

3.3. Crop response to environmental factors

3.4. Crop steering

3. Crop growth and response to environmental factors

Most textbooks start with Physics and climate modeling as the foundation. In this book, we start with the plants first, explaining what the plants need to thrive and how they respond to different climate conditions.
Placeholder image
Placeholder image → need to draw a new diagram connecting photosynthesis, transpiration, and sugar distribution.


Plants breath CO2 through their stomata, which allow the gas exchange. Depending on the CO2 in the atmosphere, which can be controlled by either removing CO2 from the atmosphere or adding more from gas tanks. The stomatal conductance depends on the need for the plant to breath and control its water relations, since the stomata are both needed for the gas exchange and the evaporation of water to disperse heat. More information on this can be found in the water relations section. Carbon taken up through the stomata is metabolized via photosynthesis to store the intercepted light energy in the form of sugar, which is used as a fuel for the metabolism to create proteins. The rate of carbon per energy unit of light is described as the assimilation rate.

Assimilation rate

The assimilation rate describes the rate at which carbon from the air is taken up into the plant and used in the metabolism. Photosynthesis is responsible for this, breaking up CO2 to access the carbon inside, metabolizing it into sugar. Oxygen is a byproduct of this reaction, which incidentally allowed the development of higher life on earth. Looking back into history we can roughly pinpoint the time at which photosynthesis developed in cyanobacteria, which came along with a mass extinction of all species that were unable to adapt to this sudden rise in oxygen in the atmosphere. For more information please look at the “great oxidation event [Wikipedia].

The biochemical reactions that allow photosynthesis to happen in detail, its dependency on nutrient supply, Vmax (the maximum rate of carboxylation) and temperature are too complex to break down here in detail. In short, the photosynthesis apparatus works better at higher temperature, but like all enzymes and proteins it has a maximum temperature after which it starts to break down, or denature. For the photosynthesis enzymes the “optimal” temperature lays around 20°C, with higher temperatures slowly decreasing its effects. Above 40°C the enzymes start to break down quickly. For more information on the effects of temperature on crop photosynthesis please read this article by Moore et al (2021) “The effect of increasing temperature on crop photosynthesis: from enzymes to ecosystems.

In our growth models and harvest predictions we are often separating the yield into two parts, on one side the pure biochemical machinery that, roughly speaking, turns light energy into biomass in “sources”, and the plant development where this biomass is send into various “sinks”. Over the course of the plant development the focus of the “source/sink” ratio is changing. One of the best examples for this are the so-called “early” and “late” variety of various vegetables like cabbage.

Early varieties shift from leaf and root development to fruit development early, resulting in a quick harvest of small yet already developed fruits. Late varieties go the other way, focusing on large leaf and root development first before they begin to produce very large fruits. Depending on what the market demands the choice of the appropriate variety if important. Local markets prefer early fruits of manageable size while larger commercial buyers prefer large fruits for their higher yield.

The canopy density, meaning how tightly the plants are placed together, also plays into this calculation. Early plants are grown closer to another since they will not expand in size as much as late plants. These need to be separated either during the growth or from the beginning so that enlarging leaves and fruits don’t compete for space and light. Of course a higher canopy density increases the yield per area, but only as long as the negative effects of competition can be compensated. Additional lighting or intercanopy lighting are common ways to increase the supply of light in densely packed plants.

Light interception

Plants regulate their leaf positions and leaf angles [Wikipedia] depending on the time of day (Circadian cycle/nastic movement), the light intensity (sun avoidance), the position of the sun (heliotropism), water relations in the soil and plant (drought response), impact and touch, heat/cold and the shading by other plants, structures or even itself (shade avoidance).

The three main driving patterns of this leaf movement are the Circadian cycle, the Heliotropism mechanisms and the shade avoidance:

  • The circadian cycle is the internal clock of the plant, a massive interconnected biochemical network of feedback loops that pick up light and darkness signals, enabling the plant to measure the length of the day and night cycle and adapting to it. In most plants, this leads to leaves taking horizontal positions during the end of the night for maximized sun interception during the day and lowering/raising the leaves at the end of the day. While the mechanical basics [Ueda and Nakamura, 2007] are quite well understood by now, them being a chemically controlled shifting of water pressure in the pulvinus cells [Wikipedia] the evolutionary benefits are still under research. Experiments have shown that even under artificial patterns like six hours of light and darkness, the leaf movement adapted, with many lowering and raising reactions of the leaves happening before the lights turned on and off.
  • Heliotropism is the ability of plants to follow the path of the sun or a suitable artificial light source. Sunflowers are the most common plant showing a very defined reaction, with their large flowers following the suns path throughout the day. Just like with the circadian cycle a variety of possible reasons for this phenomenon have been proposed [Wikipedia]
  • Shade avoidance is the ability of plants to detect shade and shading, activating the leaf and stem elongation mechanisms to escape this shadow, reaching the light again. Regulating this response are the phytochrome complexes in the plant, photoreceptors that switch between and active and inactive form based on the ratio of red to far red light they detect [Wikipedia]

When sunlight is intercepted by a leaf not all wavelengths are absorbed by the chlorophyll complex. If that were the case, leaves would be black and not green. Instead the so called “green gap” [Second.wiki] in the visible light and those wavelengths “above” red (far red from 710 to 850 nm) are passing through the leaf undisturbed. As such the ratio of red to far red below a leaf is very different than above, with the far red ratio increased.

The phytochrome complexes are capable of detecting this ratio and can send the signal of shade down the response ladder, activating the elongation response. In dense canopies this effect can lead to stem and leaf elongation where plants are spending more energy on growing upwards faster, decreasing fruit development [Greenhouse product news].

The self shading of plants is partly prevented by the plants of leaf shape and position on the stem. This orientation is called phyllotaxis [Wikipedia].

Fixed targets, variables and control parameters

In our binary system we can only really influence the parameters that are not controlled by other parameters. The actual price on the market for example is not ours to control, we can only set the price when we sell our product on the market. As such we are left with the following parameters that we can “control”: Ecological costs, Social costs, Fixed costs, Variable costs, Harvest index, Quality, Time of harvest/sale, Loss factor, CO2 atmosphere, Water, Temperature, Light, Nutrients and, despite not mentioned in the diagram so far, the Substrate.

We further divide these into the fixed “targets”, variables and control parameters. To explain:

  • Fixed targets are the settings that we can directly control, but which can’t be changed during the growth without severely affecting the entire system. As such we treat them as simply being unchangeable. These are the time of harvest/sale, the harvest index, the quality and the substrate. All of these have to be set before the growth begins and depend either on the production goal or the conditions of the greenhouse. We can’t simply just replace the substrate.

Substrate in this setup is treated as a storage for water and nutrients, the details of root growth are ignored for the moment but could be included later.

  • Variables are parameters that are influenced by our setup but are only regulated indirectly. Ecological costs, Social costs, Fixed costs and Variable costs are goals on which we can develop towards, but aren’t changed directly. We can demand that our production reduces the social costs, reduces the ecological costs while allowing it to rise in the variable costs. But it is impossible to simply declare a reduction in the ecological costs directly.
  • Control parameters are all those settings that we can directly control and which effect all other parameters. In our setup we count CO2, Water, Temperature, Light and Nutrients. These control parameters are not directly included in the binary tree because each affects multiple, if not all other parameters. Light effects the energy uptake, leaf development, but also the temperature in the greenhouse and the costs of energy. Same for the temperature inside of the greenhouse. Water and nutrients, their precise application time, concentration and makeup, determine quality, productivity, photosynthesis and more.

And only now are we slowly approaching the actual parameters that we can directly control in the greenhouse, the ones where the plant growth is concerned. Pests and diseases take away energy and need to be cured or prevented, assuming the cost of the cure is not higher than the loss of crop. Constant checkups and preventive measures are costly as well after all. If the risk of an infection is higher at higher temperatures, but higher temperatures mean more yield, is this an acceptable trade-off? What about the heating costs? Slowly the network needed to answer the questions “What is my profit” and “How can I maximize profit” is revealing itself. We could go even further into detail, talk about the large variety of pests and loss factors, the calculations between leaf size and leaf area index, the physical and chemical reactions that convert the light on the leaves into energy, how this energy is distributed through the plant and which stages in a plants life are the most influenceable to steer towards certain goals. The light interception for example is determined through the light in the greenhouse or field, in combination with the size and orientation of the leaves. The light use efficiency is determined through the photosynthesis, which in turn depends on water relations, which depend on transpiration and temperature, which can be controlled by heating, which influences the price for the greenhouse and so on and on.

For example, imagine a grower wants to change the nutrient setup due to high costs of the nutrient mix (Nutrients → Fixed costs). An AI could point this out, if it were to be tasked with reducing the fixed costs. The nutrient mix used is tagged with multiple stats, such as the costs for its use, its effect on the assimilation rate and the fruit quality, but also the ecological costs. Changing the nutrient mix to a different one changes these tags, with new costs for usage, effect on quality and ecological costs. With these new fixed effects the AI can recalculate the control parameters, determining new optimal light, temperature and water controls. All of which come again with new tags for fixed price, quality, assimilation rate and ecological costs.

This is the power of AI. The ability to build a network that calculates all these parameters in real time, using data from a central network and direct data from the growers. But in order to function one question has to be answered first:

What exactly are you looking to produce? And how are you planning to produce, grow, market, store, transport, sustain, defend, protect, harvest, sell, use and plant it?

Section: GIGO

What do a space probe crashing into mars and a computer pausing Tetris have in common? GIGO. In 1999 Nasa tried to send the “Mars Climate Orbiter” on a stable course around Mars. However, when the probe, far away from earth, was calculating the optimal height around mars a system failure approached. The company that had developed one piece of the ground software had set it to operate in feet and inches. Nasa uses metrics. As such the internal computer accepted the data, was unable to check the correct unit, and calculated a path that steered the probe straight into mars [Wikipedia].

In 2013 an AI system was developed to play a variety of old Nintendo games, learning based on its final score. Actions that caused a rise in points were rewarded, actions that lost points were “punished” and failing a game was seen as a large point reduction. And the best summary of what failed is perhaps the game “Tetris”, in which the ai learned to stack the blocks as fast as possible and then pause the game right before it lost. Since the game never stopped it had never lost. (The first level of Super Mario Bros. is easy with lexicographic orderings and time travel… after that it gets a little tricky. [Murphy, 2013])

What can we learn from these amusing stories? Well, amusing to use at least, Nasa was probably not laughing quite as much. We can learn to check our data and check our model. The binary model above shows how we can define the different parameters used to describe the plant growth operation and if one of these parameters is not brought into the plan then the system will ignore it. An optimization program is a tool to let you know which constraints you forgot. And an AI is a very fast optimization program [Hofstätter, 2021]. If a system is set to increase profit but the cost of water is ignored then an AI will use this, increasing water use to near infinity simple because it “knows” that plants grow better when they have more water. Same for nutrients. When we ignore the damage of excessive light the AI will increase light intensity simply because it “knows” that plants grow better when they get more energy. When we don’t tell the AI what we harvest from a plant then the ai will ignore it. When we don’t include pests and infections into the model the AI will be unable to calculate these, instead for example reading the damages as results of heat. If we don’t calculate the social and ecological costs then AI will ignore them, pushing the maximum amount of cost on it.

The cost of perfection is infinite

But we can at least come close to it by making sure we precisely know what we are looking for when we declare our goal for plant growth. What do we grow, at what time, for what purpose, for which market, at what cost. And in order to feed these data points into the system we split each parameter into two more, creating the binary tree.



The photosynthesis process evolved 450 million years ago. With this process, plants convert carbon dioxide and water into carbohydrates using light energy. Carbohydrates synthesized in the photosynthesis process will be distributed to all plant organs to fuel their activities. The photosynthesis process can be described as a chemical equation:


H2OH_{2}O is absorbed via root, and the gases CO2CO_{2} and O2O_{2} enter and leave the plant through tiny pores in the leaf called stomata.

Photosynthesis occurs within a special cell compartment called the chloroplast. When light is intercepted by leaves, individual photons (particles of light) are absorbed by a pigment called chlorophyll (also responsible for the green color in leaves). Chlorophyll is stashed in membranous sacs called thylakoids. Stacks of thylakoids fuse to form single units called grana. Thylakoids and grana are filled with lumen, and the chloroplast is filled with stroma (see figure below).

Photosynthesis can be divided into light-dependent (Light reaction) and light-independent (Calvin cycle) processes.
Photosynthesis can be divided into light-dependent (Light reaction) and light-independent (Calvin cycle) processes.

Photosynthesis can be divided into light-dependent and light-independent processes. The light-dependent process occurs within the thylakoid membrane and requires a steady photon stream. In this process, photons transfer energy to chlorophyll, and light energy is converted into chemical energy in the form of the molecules ATP and NADPH. The light-independent process (the Calvin Cycle) occurs in the stroma and does not require light. During this process, energy from the ATP and NADPH molecules is used to assemble carbohydrate molecules, like glucose, from carbon dioxide.


Transpiration is an important process within plants that occurs when water vapor leaves the plant through leaf stomata. The exit of water molecules through transpiration is responsible for the plant’s ability to pull water from its growth media up through its roots. Water’s cohesive property is responsible for this, since water molecules like to stick together; as molecules begin to evaporate through the stomata, the remaining molecules inside the plant’s vascular system are pulled upwards. This pull occurs all the way down the plant from the leaf to the root, where the plant can then pull more water up through its roots (see figure below).

Rendering of transpiration: water is absorbed by the plant through its roots. The water then travels up the plant’s stem, providing turgor to stabilize the plant and keep it upright. Water continues upwards to the leaves where it is used for photosynthesis and respiration. Excess water molecules are shed from the plant as water vapor, which evaporates through stomatal openings on the underside of the leaf.
Rendering of transpiration: water is absorbed by the plant through its roots. The water then travels up the plant’s stem, providing turgor to stabilize the plant and keep it upright. Water continues upwards to the leaves where it is used for photosynthesis and respiration. Excess water molecules are shed from the plant as water vapor, which evaporates through stomatal openings on the underside of the leaf.

Transpiration allows for the constant flow of water required by plants to perform photosynthesis and other growth processes, as well as providing turgor within the stems for the plant to stand upright and not wilt. Transpiration has a cooling effect on the plant in a similar way that sweating has on humans- heat held by water molecules leaves the plant as the molecules evaporate. Plant’s rate of transpiration can be modeled using mathematical equations, and used in data-driven growing as a growth factor.

2.2. Carbohydrate distribution and plant growth

2.3. Crop response to environmental factors

This section follows closely the order presented in Chapter 7 of [Stanghilini et al]

In this section, we will explain how the crop responds to each environmental factor such as light, temperature, carbon dioxide, humidity, etc. However, it’s critical to note that crops do not respond to these factors in a mono-factorial way or in any linear fashion.


Crops require light in order to carry out photosynthesis, produce carbohydrates, and increase biomass (grow!). Plants absorb light through their leaves, and therefore the more or bigger leaves in their canopy the more light they can absorb. As the amount of light absorbed by a plant increases, the photosynthetic rate also increases to a certain point as long as other nutrients are not limited. Heat as the result of light or photosynthesis is released from the plant through transpiration.

Too much light, either in length or power, can have negative effects on plant health. Plants can experience sun damage in the form of chlorosis- a yellowing of the leaf surface where the chlorophyll is lost. Chlorosis is not only caused by too much light, but can also be a sign of nutrient deficiencies and some diseases. Another way plants can protect themselves from too much light is the buildup of anthocyanins- deep purple or red molecules that act as a physical barrier to light.

Since plants require light to grow, if a light source is placed on one side of a plant, then that plant may grow in the direction of the light. This phenomenon is called phototropism.

Light can be supplied by natural sunlight or by artificial light sources.

4. Climate dynamics inside the Greenhouse

4.1. The Principles of Energy Balance

The indoor greenhouse climate is affected by the outdoor environment and greenhouse design which includes heating, insulation, shading, cooling, CO2 enrichment, humidification, and de-humidification. The scheme below describes energy exchange processes that control the climate inside the greenhouse. For example, processes that rule the canopy temperature are: PAR and NIR radiative heat exchange between canopy and Global radiation, between canopy and lamps; FIR radiative heat exchange between canopy and inter-lamp, between canopy and heating pipe; sensible heat exchange between canopy and greenhouse; and latent heat exchange between canopy and greenhouse air via transpiration.

Scheme of energy balance. All the items in grey exchange FIR between each other.
Scheme of energy balance. All the items in grey exchange FIR between each other. Source

4.2. The Dynamics of Humidity

All vapour fluxes inside greenhouse are depicted in the figure below. The vapour flux from canopy to the greenhouse air (MVCanAir)(MV_{CanAir}) is transpiration. Plant transpiration is the process of water movement through a plant and its evaporation. Up to 95% of the water is lost by transpiration, and only a small amount of water taken by the roots is used for growth and metabolism. Plant transpiration occurs through pores on leaves, which are called stomata. Transpiration rate is influenced by relative humidity, temperature of the greenhouse air, wind, and sunlight. At low relative humidity, high temperature, strong wind, and high sunlight intensity the transpiration rate will be high. In this circumstance, plants need to be provided more water than usual. Usually, greenhouse air receives vapour fluxes from canopy via transpiration, or humidification systems such as pad, fogging or blowing. And there are vapour fluxes from greenhouse air to the top air, screen, and outside air, and mechanical cooling system.

Vapour fluxes inside greenhouse.
Vapour fluxes inside greenhouse. Source

4.3 The Dynamics of CO2CO_{2}

CO2CO_{2} fluxes inside greenhouse. Source

5. Crop-specific Physiology

5.1. Lettuce and Leafy Greens

Leafy greens is a term for plant leaves eaten as a vegetable for example Lettuce, Kale, Microgreens, Collard Greens, Spinach, Cabbage, Beet Greens, Watercress, Swiss Chard, Arugula, Endive, Bok Choy, Turnip Greens, Purslane, Radicchio, etc. They are all rich in nutrients and can be eaten raw or cooked.

Lettuce is one of the most popular vegetables, and it has grown more and more all around the world in past decades. According to the shape of the lettuce head and predominant use, lettuce cultivars can be classified into five types:

  • The crisphead (or iceberg) lettuce has firm, densely packed heads with pale green edible leaves and is resistant to mechanical damage. Crisphead lettuce is crisp and hearty but less flavorful compared to other types of lettuce.
  • The butterhead lettuce has loose, open heads with soft leaves and is sensitive to mechanical damage. There are two types of butterhead lettuce Boston and Bibb, both are suitable for cooked ground chicken or shrimp.
  • Leaf lettuce also shares fragile nature with butterhead lettuce. There are three different types of leaf lettuce: red, green, and oak. Red and green leaves have a burgundy tint and mild flavor, while oak leaf is spicier and nuttier. Leaf lettuce has more nutrient content compared to butterhead lettuce.
  • Cos (or romaine) lettuce has loose, loaf-shaped heads with erect, long, and sturdy leaves. Outer leaves of cos lettuce are slightly bitter, but it is sweeter in the center of the head.
  • Stem lettuce is popular in China and is also called Chiness lettuce. Stem lettuce has narrow and very bitter leaves.Therefore, stem lettuce leave is discarded while the stem is eatable.

Optimal Environment for Lettuce Production

Much research has been conducted looking into the optimal environment to produce lettuce cultivars, and key environmental values have been well established.

  • Light: When CO2 level is low and vertical air fans are used, the optimal DLI can be up to 17 mol/m2/daymol/m^{2}/day . If vertical fans are not present we should lower DLI. If DLI is too high, lettuce is at high risk of tip-burn and bolting. If DLI is too low plants will be small with elongated and smaller leaves and typically had a paler green color. Both cases produce low-quality lettuce and would not be impressed by consumers. For lettuce grown in the winter/short day season, supplemental lighting will be beneficial to reach this DLI value. Of course, DLI can be lower in production but crop cycle may be lengthened as a result.
  • Temperature: Temperature is the main factor affecting the growth rate of lettuce in the early stage. Optimal temperature for lettuce production is from 7 o^oC to 24o^oC with the average of 18°C18°C. Lettuce is a cool-weather plant, and therefore its temperature requirements are cooler than most other indoor-grown crops. This can be difficult to maintain in greenhouses or glasshouses in warmer climates.
    • Rising temperature can become a problem as it promoted bolting, when the lettuce begins to grow tall instead of wide and leads to flowering. Bolting also affects the flavor of the crop, often resulting in bitter taste.
    • Low temperature can result in slowing the crop cycle. When temperature and light are optimal, lettuce can grow in 35-day crop cycles. The more temperature and light deviate, the longer the crop cycle can take, up to 80 days or more.

To learn more about lighting and temperature’s effects on different cultivars of lettuce, check out this publication.

  • Nutrient solution for hydroponic growing: pH of 5.5 to 6.0 and fertilizer electrical conductivity (EC) of about 1.5 mhos/cm2 are optimal for growing hydroponic lettuce.
  • VPD: According to this study, lettuce grows well when the vapor pressure deficit is in the range of 0.4 to 1.0 kPa. Above 1.0 kPa, the air is too dry leading to an increase in plant water stress. When the VPD is lower than 0.4 kPa the air is too humid and the transpiration process will be inhibited. Poor transpiration over a long time increases the risk of tipburn as plants do not receive enough Ca for building new tissues.

Most Common Problems in Lettuce Production


Source by Neil Mattson

Tipburn is primarily caused by the calcium deficiency at the tip of young leaves. This frequently happens to the young inner leaves. The underlying cause is two fold:

  1. Lack of airflow around the inner leaves. This causes high boundary layer resistance and hence poor transpiration. Poor transpiration leads to poor nutrient (including Ca) transfer to the new inner leaves.
  2. Strong growth relatively to the transpiration capacity. Strong growth promote the formation of new tissues. However, since they don't receive enough nutrients to survive, they die young and become tipburn.
Tipburn is not a burn. It's a common myth, fueled by the term, that the plant got burned and hence growers solve it by reducing/turning off supplemental light or deploy shading so that the plants don't get burned. While there is a confounding factor between lighting and photosynthesis/transpiration and hence tipburn, lighting is not the primary cause of tipburn and suppressing it isn't always the optimal way of addressing the tipburn problem.

How to avoid tipburn

Better ventilation and airflow with vertical fans

Vertical fans provide better airflow for the vulnerable inner leaves and hence improve the transpiration to the tip of these young leaves.

Control the growth/transpiration ratio.

This is counter-intuitive and unfortunate but you do need to control the growth rate if you want to avoid tipburn. Additionally, keeping humidity low, between 50-70%, will help control the transpiration rate and occurrence of tipburn.

Discoloration other than tipburn


Iron Deficiency

Source by Neil Mattson

The most common nutrient deficiency in lettuce grown using hydroponics is iron deficiency. This deficiency appears as chlorosis (yellowing) of the leaves around the veins, usually affecting younger leaves first. The underlying cause can be two-fold:

  1. Lack of iron in the nutrient solution.
  2. High pH of nutrient solution, which negatively affects the absorption of iron. When pH rises too high (>6.5), iron may oxidize or precipitate, thus reducing its availability for absorption.

How to avoid iron deficiency:

Check the amount of iron in your nutrient solution

Most nutrient solutions provide 1-3ppm of iron

Check the nutrient solution is mixed thoroughly, and that your nutrient injectors are functioning properly

Proper quality assurance and maintenance will go a long way in preventing avoidable problems.

Monitor and maintain proper pH and nutrient levels in your hydroponic solution

Constant monitoring for pH and electrical current (EC) at several points in your hydroponic solution can ensure nutrient deficiencies are avoided. Several sensors are available on the market for such purposes. When sensors are also connected to a controller, maintenance of pH and EC can be automated and adjustments can be made based on real-time readings.

To learn more about iron deficiency in lettuce, read this article.

5.2. Tomato

Optimal Environment for Tomato Production

The optimal environment for tomato is shown in the figure below.

Optimal environment for tomato in different stages. T: temperature, RH: relative humidity, DLI: daily light integral.
Optimal environment for tomato in different stages. T: temperature, RH: relative humidity, DLI: daily light integral.

Light: Tomato requires a daily light integral (DLI) of more than 20 mol/m2/d (this value can increase toward the end of the season) with a maximum photoperiod of 18 hours per day. For tomato growth in the Winter when natural light is limited, supplemental light is necessary for plants to grow normally. The most common supplemental light sources are HPS and LED. Compared to HPS, LED is getting more and more attention due to its high efficiency and long lifetime. Additionally, LED generates less heat compared to HPS making it possible to bring the lamps into the crop canopy for inter-lighting. However, HPS is cheaper to set up and can provide an intense amount of light for plants. The combination of HPS as grow-lighting and LED as inter-lighting was shown to increase the yield of tomatoes by 20% compared with using grow-lighting HPS only (Moerkens et al., 2016). CO2: CO2 is an essential factor for plants to photosynthesize. CO2 enrichment in semi-(close) greenhouses can result in a 10-20% higher yield compared to an open greenhouse. Though the leaf- and truss-appearance rates are not affected by CO2 concentration, the fruit set increases when the CO2 concentration increases. Temperature: The fruit growth period is primarily dependent on temperature. Other factors like light, CO2, humidity, plant density, and nutrient contents have a small effect on the fruit growth period. According to the study of De Koning, 1994, 2000; Adams et al., 2001, the fruit growth period decreases with increasing temperature from 14 to 26°C, and high temperature also promotes fruit ripening. Humidity: The recommended value of vapor pressure deficit (VPD) for tomato is from 0.2 kPa to 1.0 kPa. Too low VPD increases the risk of diseases such as Botrytis. On the other hand, too high VPD reduces the photosynthesis rate due to stomatal closure. Both cases lead to a reduced leaf area and hence reduce fruit growth rate and yield.

Most Common Problems in Tomato Production

Topping Tomato Plants

What and why is topping? Topping is literally cutting the top of the main stem off of your tomato plants.

There are two great reasons to top your tomato plants. The first reason is when tomato plants are outgrowing your infrastructure, they can snap easily, especially when the wind is strong. Once the plant snaps, it will get damaged and rotting. To control the growth, we need to top them and cut the head off so plants stop growing taller. The other reason is that we want to maximize the chances of setting fruit and fruit production. By removing the heads of plants, they stop encouraging leaf growth and put more energy into flowers and fruits.

When to top? We should top tomato plants around 45 to 60 days before the season ends. That will allow the roots to concentrate as much energy as possible on fruit set and production and make fruits grow as quickly as possible.

How to top? Use a sanitized, washed, and dried pair of shears to cut off the head right behind the top flower cluster of the main stem. After topping, the main stem will not grow any further, and the flower cluster is at the top of this stem. While topping, we also want to remove all suckers because we don't want them to develop individual main stems on their own.

5.3. Strawberry

More technical materials

J(x)=ex×GJ(x) = \int e^x \times\nabla G

5.4. Other Vine Crops

Most of the sections for tomato are also applicable for other vine crops such as cucumber, bell pepper, etc. Here, we only highlight the key differences.

Differences between Cucumber and Tomato

Fruit characteristics

  • Cucumbers don't grow in truss. There is 1 fruit and 1 leaf for each node of the cucumber plant.
  • The fruit development rate of cucumber is relatively faster than the the fruit development rate of tomato
  • Fruit aborting is a common phenomenon in cucumber. Fruit pruning is a common practice to avoid that.
  • Temperature sum for the fruit to reach full potential: 275C275^{{\circ}}C (compared to 1035C1035^{{\circ}}C of tomato)

Leaf characteristics

Specific leaf area index (m2 of leaf per mg of dry matter): 3.78e-5

(higher than that of tomato, which is 2.66e-5)

Climate preferences

  • Temperature: cucumbers enjoy a higher temperature than tomatoes do. According to [Marcelis], the daily averaged temperature for cucumbers is: 10 (lower bound), 20-25 (ideal), 35 (upper bound)
  • Humidity: cucumber plants are also more tolerant to higher humidity [Hao]

Topping and pruning

We do not have to top cucumber unless plants are spreading out too much and the end of the season is approaching, you can top off your plants to stop the new growth and let the plant focus on finishing its current fruit. However, for cucumber, we should do pruning once a week during the growing season. Pruning is to avoid vines grow out of control, diseases (such as powdery mildew), poor airflow..

6. IPM

The main idea of Integrated Pest Management (IPM) is using biological control when you can, chemical control when you must. IPM is different in different stages: before planting, early stage of pest, and lesion appears on plant.

6.1. IPM before planting


Soil health: Soil with sufficient nutrients and moderate porosity ensures the growth of healthy roots. It helps plants grow healthy and resistant to many pests and diseases. In addition, soil sanitation helps eliminate some pathogens available in the soil.

Choice of variety: Choose disease-free and insect-free certified seeds and plants if possible. It is important to start with sturdy plants with healthy roots because diseases and insects in young plants can later cause heavy losses in your greenhouse. Precision sowing: placing seed at a precise spacing and depth, by doing so, each plant has optimum space to grow and develop. Precision sowing also makes week management easier and reduces the risk of pest and disease.

Crop Hygiene: Remove and destroy the infested or diseased plants/seeds so they do not become sources of re-infestation.

6.2. Early stage of pest


Pests and diseases can be early detected by human eyes, traps (sticky trap, pheromone traps), magnifying glasses or special cameras on drones. In the early stage, when the symptoms are still subtle growers can apply mechanical and/or biological managements such as using natural enemies, using trapping techniques and pheromone traps, using plant nutrition, matching IPM and crop, climate management. Application of these methods should be considered in combination with the economic thresholds.

6.3. Lesion appears on plant


If the IPM in before planting and in the early stage are not well done, the symptoms of pests and diseases will become more serious with time, and when the lesion appears on plants growers will have to apply biopesticides or even chemicals which is usually the last recommended solution because pesticides and chemicals can kill both beneficial insects and pest species. On the other hand, pesticides and chemicals can be harmful to environment and pesticide residues in fruits or vegetables are hazards to users if they are overused.

7. Hardware tech to improve crop production

In this chapter, we focus on improvements that can be added to your existing greenhouse to achieve better productivity. For that reason, we will not discuss technologies that you can't easily adopt, such as greenhouse structure and glazing materials.

7.1. Movable Screens

Movable screens: There are three types of screens for different purposes to help control the climate inside the greenhouse.

  1. Blackout screen: it completely blocks light and is usually used during nighttime when supplemental light is on to avoid light pollution or in case crop requires a carefully regulated day length. It can be deployed on a sunny day to reduce sunlight entering the greenhouse as well. However, when using a blackout screen to reduce sunlight, the temperature inside the greenhouse will increase, and we may need to activate the fogging system to cool the greenhouse.
Blackout screen,
Blackout screen, source
  1. Thermal screen: it is also called energy screen. The thermal screen keeps warm air inside the greenhouse and cool air outside the greenhouse. As the thermal screen is transparent, about 75% of light can go through. It is usually used on cold nights to save heating costs, and sometimes it can also be used to reduce light intensity on sunny days.

Thermal screen,
Thermal screen, source
  1. Shade screen: We sometimes have to cool the greenhouse air down to avoid overheating in the summer. Outside of opening the window, the door, and vents or using fans, the most affordable way is to use shade screens. There are different percentage amounts of shade screens like 20%, 30%, … 90%. That means they block sunlight but not completely, which allows us to reduce heat inside the greenhouse while letting some sunlight enter the greenhouse for plants to grow.

7.2. Supplemental Lighting

Supplemental lighting is used in greenhouses to increase crop production during time periods with low levels of solar radiation. These time periods usually occur during the winter months, but cloudy summer days can be as dark as some of the darker winter days. Thus, if crop production is on a tight schedule, supplemental lighting may be required year-round. Sometimes, photoperiod lighting is also defined as supplemental lighting. But since the light intensities required are very low, and photoperiod lighting consumes limited amounts of energy, it is not considered in the context of this discussion. Despite the installation and operating costs associated with supplemental lighting systems, growers are discovering the benefits. These systems can help improve crop quality, keep production on schedule and reduce the length of the growing cycle. Thus, growers produce a higher-quality product while keeping their production schedules on target, and they are able to produce more crops per year.

High-Pressure Sodium (HPS)

High-intensity discharge lamps, especially high-pressure sodium (HPS) lamps, have been traditionally used for supplemental lighting in commercial greenhouses to increase photosynthesis. High-pressure sodium lamps are the most widely used lamp type because of their relatively high efficacy (conversion of electricity into photosynthetic light) and lifespan of 10,000–12,000 h. However, approximately 70% of the energy consumed by the fixtures is not converted into PAR and, instead, is emitted as radiant heat energy. The surface temperature of HPS lamps can reach as high as 450 C, which requires the separation of lamps from plants (Fisher and Both 2004; Nelson 2012; Spaargaren 2001). Additionally, HPS lamps primarily emit light in the range of 565–700 nm, which is predominately yellow (565–590 nm) and orange (590–625 nm) light. They only emit 5% blue light, which is low compared to solar radiation which contains 18% blue light (Islam et al. 2012).

Light-emitting diodes (LED)

Light-emitting diodes are a promising supplemental lighting technology for the greenhouse industry as they surpass in many aspects capabilities of commercially available lamps commonly used in horticulture (Morrow, 2008). As described by Bourget (2008), LEDs are robust, solid-state semiconductor devices that can emit narrow-spectrum light to maximize photosynthetic quantum efficiency for specific crop species. In 2008, LEDs were as electrically efficient as fluorescent lamps and slightly less efficient than HPS lamps at converting electrical energy to light (Bourget, 2008). As of 2012, blue and red LEDs are up to 50% and 38% efficient, respectively (Philips Lumileds Lighting Co., 2012). Unlike traditional high-intensity discharge (HID) light sources used in commercial greenhouses today, the relative coolness to the touch of LED photon-emitting surfaces allows them to operate in close proximity to plant tissue without overheating or scorching plants, thereby increasing available PAR at leaf level using less energy. With ongoing improvements in terms of energy efficiency and availability of photosynthesis-driving wavebands, LEDs provide potential solutions to the profitability and sustainability issues that greenhouse growers face.

Research in plant growth chambers and tissue culture laboratories has proven that LEDs are an efficient light source for plant lighting in controlled environments (Hoenecke et al., 1992Jao and Fang, 2004Jao et al., 2005Nhut et al., 2000Poudel et al., 2008Schuerger et al., 1997Shin et al., 2008Tanaka et al., 1998). However, the potential of LED supplemental lighting for large-scale greenhouse operations continues to be explored. Dueck et al. (2012) compared the effect of different irradiation directions of LEDs and HPS on growth and production of greenhouse-grown tomatoes in The Netherlands. They suggested that a combination of overhead HPS and LEDs is the most promising alternative for their climate, when taking into consideration different production parameters and energy costs (lighting + heating) of using the different systems. Another experiment measured responses of photosynthesis and yield for cucumber (Cucumis sativus) grown under either a combination of LEDs (80% red + 20% blue) within the canopy + overhead HPS or HPS only during a winter production cycle (Trouwborst et al., 2010). They reported no improvement in net crop photosynthesis and fruit production when using LEDs + HPS compared with overhead HPS only but attributed their results to low irradiance (light-limited crops) throughout the experiment, regardless of treatment.


The most efficient HPS and LED fixtures have equal efficiencies, but the initial capital cost per photon delivered from LED fixtures is five to ten times higher than HPS fixtures. The high capital cost means that the five-year cost of LED fixtures is more than double that of HPS fixtures. If widely spaced benches are a necessary part of a production system, LED fixtures can provide precision delivery of photons and our data indicate that they can be a more cost-effective option for supplemental greenhouse lighting.

Manufacturers are working to improve all types of lighting technologies and the cost per photon will likely continue to decrease as new technologies, reduced prices, and improved reliability become available.

Light Units

The preferred unit for measuring light for plant production is µmol m-2 s-1 (pronounced: “micromol per meter squared per second”). This unit expresses the number of particles (photons or quanta) of light incident on a unit area (m2) per unit of time (second). The portion of the light spectrum the plants use for photosynthesis is called Photosynthetically Active Radiation (PAR, 400-700 nm, nm = nanometer), and it is expressed in the unit of µmol m-2 s-1. Sensors used to measure PAR are called quantum sensors and have carefully designed filters such that no light outside the PAR waveband is measured. Our human eye is able to detect light in a slightly larger waveband of approximately 380-770 nm. To measure light in this waveband, a foot-candle meter (or a lux meter) can be used. But measurements with a foot-candle meter include some light with wavelengths outside the waveband used by plants for photosynthesis. Therefore, using a foot-candle meter introduces a small error when we are only interested in measuring the amount of light available to plants for the process of photosynthesis. For this reason, the use of a foot-candle meter is not recommended when evaluating the light environment for plant production. It is possible to convert a measurement taken with a foot-candle meter into a µmol m-2 s-1 value, but the correct conversion factor depends on the light source and is, in the case of mixed light sources, not always easily determined.

Red Light

Red photons of light have a wavelength between 600 and 700 nm. One of the most common roles for red light is to participate in the physiological process of photosynthesis. Specifically, many LED arrays emit red wavelengths at 660 nm, which is very close to the absorption peak of chlorophyll (Massa et al. 2008). Thus, red LEDs can be used to efficiently drive photosynthetic activity, resulting in increased biomass and overall plant productivity. However, red light alone is not sufficient for the optimum production and quality of most crops. When exposed to solely red light, many dicotyledonous crops develop extensive hypocotyl elongation (Hoenecke et al. 1992). Additionally, Arabidopsis plants grown under only red light develop abnormal morphological characteristics (Goins et al. 1998). However, both red and blue light (400–500 nm) control stem elongation (Kigel and Cosgrove 1991). Specifically, blue light, when combined at a low irradiance with red light, can prevent excessive elongation of hypocotyls, stems, and petioles and deter other morphological abnormalities observed under solely red wavelengths (Goins et al. 1998; Hoenecke et al. 1992).

Red light is involved in much more than simply photosynthetic activity. Phytochrome is one of the primary families of photoreceptors that absorb red light as well as far-red (700–800 nm) radiation. When exposed to light, phytochromes exist in two interconvertible forms, the red-absorbing (Pr) and far-red-absorbing (Pfr) forms (Smith and Whitelam 1990). The relative proportion of Pfr to the total amount of phytochrome (phytochrome photoequilibrium) regulates a variety of photomorphogenic responses including stem extension (Runkle and Heins 2001; Stutte 2009) and flowering (see Chap. 14). These photomorphogenic responses are known to vary with plant species and cultivar, age, light quantity and quality, and temperature. Plants are generally more compact when exposed to light with a high red-to-far-red ratio (R:FR). They are also more sensitive to red and far-red light at the end of the day (EOD), and 10–60 min of EOD red light may be as effective as a high R:FR during the entire photoperiod to inhibit extension growth (Ilias and Rajapakse 2005).

Installation Considerations

When installing supplemental lighting systems in greenhouses, several factors should be considered. First, the average amount of solar radiation for the location should be investigated.

This will give an idea of the range of solar radiation conditions at the site. One way to determine the amount of light available for crop production at a particular location in the United States is to consult the database of solar radiation data maintained by the National Renewable Energy Laboratory in Golden, Colo. (www.nrel.gov). This database contains solar radiation data for 239 locations across the United States and its territories. For plant production purposes, the solar radiation data can be converted into the units of mol m-2 d-1, indicating the daily sum (integral) of light available for photosynthesis (1 kWh m-2 d-1 = 7.49 mol m-2 d-1). Second, the type of greenhouse structure, glazing and equipment installed will have an impact on the transmission of sunlight. Third, the type of crop (or crops) grown in the greenhouse will determine the plant’s requirements (such as light intensity, duration or light integral). Fourth, the available space in the greenhouse to hang lamps will have an impact on the uniformity of supplemental lighting (the less space available for taller crops in lower greenhouses, the less uniform the light distribution). Finally, the plant’s requirements should be compared to the available amounts of sunlight to calculate the necessary amounts of supplemental lighting.

It is usually not economical to install lighting systems that provide high light intensities in greenhouses because of the large number of lamps required. Therefore, supplemental lighting systems can be designed to provide a certain light integral during a 24-hour period such that the sum of the supplemental light integral and the solar radiation integral meet the plant’s requirements for even the darkest day of the year. The light integral supplied by the supplemental lighting system depends on the average light intensity provided by the lamps and the duration of operation. The light intensity supplied by commercial supplemental lighting systems usually is not higher than 200 µmol m-2 s-1 (0.72 mol m-2 hr-1 or 17.3 mol m-2 per 24-hour period).

Light Uniformity

In addition to light intensity, light uniformity is an important factor to consider when designing lighting systems for greenhouses. In general, except when clouds are passing overhead or when structural elements create shading patterns, sunlight is uniform from one location to the next inside a greenhouse. However, due to the distance between lamps and the distance between the lamps and the crop, supplemental lighting systems will always provide non-uniform lighting patterns over a plant canopy. It is the task of the designer to optimize light uniformity by carefully calculating the light distribution from each lamp and the different paths the light can travel from each lamp to the crop underneath. Fortunately, computer software programs exist to assist the designer with this complicated task and in general, a careful design results in very acceptable light distribution and uniformity over a crop canopy.

7.3. Vertical Fans

7.4. Sensors


Greenhouses and indoor farms require constant monitoring and environmental regulation to ensure optimal plant growth. Many factors such as humidity, temperature, light levels, irrigation and more must be constantly kept in check or your plants may suffer. Manually checking these conditions in your growing environment can be time-consuming and tedious. Not to mention the paranoia over whether a door was left open or leaks in irrigation equipment causing a climate catastrophe. Sensors are an excellent tool to help you monitor your growing environment quickly from anywhere and begin automating your indoor growing environment.

There are many different types of sensors available for greenhouses and indoor farms to constantly monitor and measure such variables. Sensors equipped with Internet of Things (IoT) technology are especially capable of transmitting the data collected to a data management system for remote access. For example, instead of manually checking a thermometer, a temperature sensor could be used to automatically collect this data and report it to an online server or data management system. The data can now be accessed remotely instead of having to walk into your greenhouse and check a wall-mounted thermostat.

Sensors can also increase the amount of observations you can make about your greenhouse. For example, temperature sensors can check the micro-climate of every growing zone in your greenhouse. This can be important because although it is optimal for the entire greenhouse to have the same exact temperature, you likely know that is not always the case realistically. Clouds blocking the light over certain sections, uneven drafts from fans and ventilation equipment, and uneven plant crowding as your crop matures can all play a part in offsetting the balance and consistency in your greenhouse climate variables. Realizing such differences, via the data collected by sensors, means that each area can be handled exactly as it needs to be to restore optimal climate for plant growth. This is how sensors are paving the way for precision agriculture.

Not only can sensors monitor your grow environment, but many also include the capability to respond to changes in the environment either automatically or manually. For example, you may set your temperature system to automatically begin heating if your temperature sensors register a lower-than-optimal temperature. On the other hand, you may be able to review your temperature sensor data remotely and make the decision to begin heating based on your observations, likely from a desktop or mobile app. Using sensors to create a “smart greenhouse” can save a lot of time and effort from constant manual climate monitoring, mitigate response time to and losses from unpredictable climate interruptions (i.e. leaving a door open), and collect valuable data about your grow system and how your plants respond to environmental changes.

8. Software tech and data-driven growing

In this chapter, we focus on the software technology that can unlock your sensor data’s true value, and the basics of data-driven growing. The data collected by your sensors is valuable on its own for the ease of remotely monitoring several growing conditions. However, analysis of this data can do so much more for your greenhouse production, such as improve and increase your yield, reduce your resource usage, and overall increase your profits. The name of this game is data-driven growing: using data in a proactive way to guide your growing. We outline the steps below to achieve data-driven growing, which are data collection and visualization, data analysis, and digital decisions, all of which are taken care of by software technology such as Koidra’s.

8.1. Data Collection and Visualization

The first step to improve your data-driven growing game is having a good data platform. You need to be able to collect and unify your data under one (digital) roof. A large spreadsheet of data points is not useful, we want easy-to-read charts that convey a message. The data needs to be visualized with tools such as operational dashboards, which growers can directly use for crop monitoring and decision making. Figures 2 and 3 provide examples of customizable dashboards with critical growing metrics such as light levels, CO2, temperature and more.

Using sample climate dashboards on Koidra’s
Using sample climate dashboards on Koidra’s Krop Manager app as an example, users can zoom in to see the data in the same time range across multiple data groups (or charts). Moreover, by hovering over one data time point brings up all variables’ data at that time point simultaneously. Data can be easily grouped into logical chart groups such as Lighting, CO₂, Temperature & Humidity, etc.
Helpful apps can allow growers to compare the performance across zones or across seasons to help growers understand what caused, for instance, higher or lower yields. Using screenshots from Koidra’s Control Center as examples: the left shows the differences, in terms of yield and daily light integral (DLI), between the two zones growing the same crop (cucumber) during the same season; and the right shows the differences between the same zones but in terms of yield and CO₂. Which metrics growers want to see and compare are fully customizable.
Helpful apps can allow growers to compare the performance across zones or across seasons to help growers understand what caused, for instance, higher or lower yields. Using screenshots from Koidra’s Control Center as examples: the left shows the differences, in terms of yield and daily light integral (DLI), between the two zones growing the same crop (cucumber) during the same season; and the right shows the differences between the same zones but in terms of yield and CO₂. Which metrics growers want to see and compare are fully customizable.

Before investing in a data platform, there are some important questions to ask: Is it another silo in your existing mix of data silos (i.e. isolated, unintegrated data)? Can it be integrated with your existing or new control systems? Can the data be turned into actionable insights? Once you find the data useful, can you operationalize upon that data (i.e. can the control system use it)?

Crop Registration Data

In addition to sensor data, we also recommend manually registering crop measurements on a low-frequency basis (such as day, week).

Crop Registration Parameters

8.2. Advanced Decision Support System

8.2.1. Digital Decisions

Enterprises rise or fall based on the collective efficacy of all decisions made ... by leaders, ... by employees, ... by decision logic embedded in applications.

The decision logic embedded in applications are called digtal decisions. In a nutshell, digital decisions are operational decisions in real-time or near-real-time. They are optimized by

  1. expert knowledge (this book!);
  2. sophisticated models to distill actionable insights from data (aka. AI);
  3. and data collection at scale (fueled by IoT technologies).

For greenhouse operations, digital decisions are primarily the automated climate control decisions although they may include some other actions in special settings (such as as the decision to move the gutters in a lettuce greenhouse to optimize the plant density in real time). Below is a schematic of a digital decision making system for a greenhouse.

Sensors collect climate and crop data, and send these points to the
Sensors collect climate and crop data, and send these points to the μcontroller (microcontroller), which is basically a very tiny computer with a lot less features than normal. The μcontroller sends the data into the cloud for storage and data analysis. The data in the cloud is available for growers to access through the Control Center online platform, and growers can send more data about the crop to the cloud. While in the cloud, data is also analyzed using AI in order to send optimized crop handling instructions to the grower and automatically determine climate decisions. The climate decisions are sent back to the μcontroller, which then sends the message along to actuators, the part of the system that acts on the decisions within the greenhouse.

8.x. Be aware of the high-interest technical debt

[to be written]

Data Silos

A data silo is a collection of data that is kept in a system that is not easily or openly accessible, sometimes even to those who generated the data. Data silos are unfortunately common in smart greenhouses that use sensors. Oftentimes sensor companies will collect the data and “silo” it into their own private management system, while only giving consumers access to select data points. For example, a temperature sensor may show you the temperature for the last 24 hours, but will not allow you to access the temperatures it recorded last season or even last week.

Another way data can be “siloed” is if data is not kept altogether in a central location. This case occurs commonly in greenhouses or indoor farms that use different types of sensors from multiple companies. One company may store the temperature and humidity data in their silo, but the lighting company stores their data in their own silo. Even when the data is accessible, it is often “spaghetti-ed”, for example in long excel sheets that require massive amounts of scrolling. No simple cut-and-paste could seamlessly combine the data from different silos.

We know the climate and crop variables inside greenhouses interact with and influence each other. Therefore, in order to make any type of data-driven growing decision, growers need access to all of their data in one place.

Extra Materials

Digital Horticulture Roadmap

Further reading on
Further reading on greenhousecanada.com

9. References

10. About Authors

Andreas Fricke, Senior lecturer, Leibniz Universität Hannover · Institute of Horticultural Production Systems


Andreas is a senior lecturer at the University of Hannover, Institute of Vegetable Systems Modelling. He studied Horticultural Sciences and received his Ph.D. in 1992. He has specialized himself on stress physiology and harvest prediction models, both in the field of vegetable crops. As lecturer he teaches courses in the study programs ‘Molecular and Applied Plant Science’ (B.Sc.), ‘Plant Biotechnology’ (M.Sc.) and ‘International Horticulture’ (M.Sc.).

Simon Schmitz, Ph.D. candidate, Leibniz Universität Hannover · Institute of Horticultural Production Systems


Simon is a Ph.D student at the University of Hannover, Institute of Vegetable Systems Modelling. He studied Biotechnology (B.SC) and Plant-Biotechnology (M.Sc). He has specialized himself on short term stress response in the physiology of plants and remote sensing technology to detect such stresses.

Kenneth Tran, CEO, Koidra Inc.


Kenneth (Ken) is the founder and CEO of Koidra Inc. Before Koidra, Ken was a Principal Applied Scientist in the Machine Learning Group, Microsoft Research (MSR). While at MSR, he led the Sonoma team, winning the first autonomous greenhouse challenge (2018), becoming the only artificial intelligence (AI) team that outperformed expert Dutch growers. Ken recently led the Koala team comprised of Koidra and Cornell University researchers to win the first phase of the 3rd autonomous greenhouse challenge in 2021.

Ken’s research expertise and experience include Reinforcement Learning, Physics-informed Machine Learning, and Deep Learning. Tran received his Ph.D. in Computational & Applied Mathematics from The University of Texas at Austin.

Koidra Other Scientists

Many applied scientists and engineers at Koidra have contributed to the book. Only current and active contributors are listed here.

Hannah Gardner

@Ketut Putra and @Hanh Bui to add your names and some info about you here.

11. FAQ

Copyright and ownership

The contents here are owned and copyrighted by the lead authors. Reproduction is prohibited. Any derivative material or work benefiting from this book must cite and link to the book.

How can I ...
What is ...

Link to the "control of the grower" section
Content of the book