Can Underground Agriculture Feed the World?

Greenforges uses advanced engineering tools to iterate a novel way to grow crops sub-surface.

This video is sponsored by SIEMENS.

Feeding the population of the planet of 8 million and growing, is a fundamental challenge for the 21st century. The green revolution that began in the 1950s relied on massive chemical inputs, in fertilizers, pesticides and herbicides. Today, environmental concerns, plus a warming climate and limited agricultural land has been the impetus for new ideas in agriculture. Could we go below the surface and use advanced technology to allow food production anywhere, including cities?   

Montreal, Canada-based Greenforges has developed a novel vertical system that allows agricultural production almost anywhere, without the traditional constraints of weather, irrigation or land-use. It’s harder than it looks to grow food underground, and development uses advanced tools to iterate cost-effectively. Joining engineering.com’s Jim Anderton to describe the technology and how simulation was essential in its development is Jamil Madanat from Greenforges and Carl Poplawsky, Engineering Services Manager at Maya HTT.

Learn how Simcenter unleashes your creativity, granting you the freedom to innovate and allowing you to deliver the products and processes of tomorrow today.

The transcript below has been edited for clarity:

Jim Anderton: Hello everyone and welcome to Designing the Future. Feeding the population of the planet, 8 billion and growing, oh, it’s a fundamental challenge for the 21st century. The green revolution that began in the 1950s, well, it relied on massive chemical inputs in terms of fertilizers, pesticides and herbicides. Today, environmental concerns plus a warming climate and limited agricultural land has been the impetus for new ideas in agriculture. Could we go below the surface and use advanced technology to allow food production anywhere, including cities?

Well, Montreal, Canada based GreenForges has developed a novel vertical system that allows agricultural production almost anywhere without the traditional constraints of weather, irrigation or land use. Now it’s harder than it looks to grow food underground and development uses advanced tools to iterate cost-effectively. Joining me to describe how the technology works and how simulation was essential as development is Jamil Madanat from GreenForges and Carl Poplawsky, engineering services manager at Maya HTT.

Jamil is a bachelor’s degree in mechanical engineering from McGill University where he specialized in machine design and project management. He has five years of professional experience in the impact startups world with a focus on social entrepreneurship and sustainability. Jamil is currently the CTO of GreenForges, the first underground controlled agriculture environment farm early next year.
Carl holds a master of science degree from Purdue University in Mechanical Engineering, and before his appointment as engineering services manager at Maya HTT, he was a senior applications engineer. Previously, Carl was VP of engineering at the Engineering Sciences and Analysis Corporation, and was technical consultant with the Structural Dynamics Research Corporation. Carl and Jamil, welcome.


Jamil Madanat:
 Hi, Jim. Thank you.


Carl Poplawsky:
 Thank you. Glad to be here.


Jim Anderton:
 Jamil, can we start with you? This is an intriguing solution to a pressing problem, feeding 8 billion plus people and growing. The stresses on the environment, in the Western world, we’re paving over agricultural land. We’re looking at a change in climate. There are a lot of factors here that are putting constraints on agricultural production. How big is this problem? I mean, do we need to find these ad advanced technology solutions to feed ourselves?


Jamil Madanat:
 Absolutely. Actually, this is how the idea came to me. So our founder, Phil, was going through a report describing projected food shortages all around the world, and the report was looking into different methods and means that will alleviate this food crisis issue by looking into urban agriculture. The study looked into, okay, how can we leverage urban agriculture to provide more food in the cities, and was looking into rooftop farming, indoor farming, shipping containers. A couple of days after that, Phil was happened to be thinking about this problem looking outside a window and saw a water well.
That’s when it clicked in his mind, why can we use the underground for food production? So the conclusion of that report saw that even if we leverage urban agriculture, we would still cover 4 to 5% of that food shortage. But now with underground being an option, I think we have a lot of more space to utilize and a lot more means to alleviate the projected food crisis that’s approaching faster than what we’re ready for.


Jim Anderton:
 Jamil, can you give me a brief overview of the GreenForges’ system? You mentioned underground. Underground, sometimes we think of sort of an abandoned coal mine or a cavern or something, but these are rather more like missile silos, aren’t they sort of a cylindrical vertical shaft?


Jamil Madanat: 
Precisely. So we’re taking a more simpler approach here. So what we’re starting with is think water wells or just pile foundations, same ones you would may use for building foundations. Initially, we’re starting with a diameter size of 60 inches, so closer to 1.5 meters. So it’s nothing too big. But the advantage that you would really get is going underground and now we’re experimenting with a model going 15 meters underground. With that you’d be surprised how many plants you can fit in there and grow. The scalability would just really grow much faster as you arrange these forges in a grid system.
As you mentioned in the intro, it is a controlled environment agriculture system. With that, it means we get more control over the plants that we want to grow, expedite the harvest cycle, get more precise and refined flavors and control the crops that we want to use. So it’s not just only we’re leveraging the underground system, filling it with plants, but also expediting harvest cycles one, and two, just becoming weather independent. So that will give a lot of additional advantages with keeping the food production running all year long too.


Jim Anderton: 
Jamil, is this a hydroponic process? Are you growing plants in soil? How does it work?

Jamil Madanat: Yes, it is a hydroponic system. Basically for those unfamiliar, a hydroponic system would be just water mixed with nutrients and oxygen that the plants need and runs continuously just touching the roots, the back of the roots, giving the plants the nutrients that they need. It is a continuous loop system. So really we just keep continuing using this water that’s getting fed to the plants while monitoring at the surface kind of the nutrients being consumed, how much extra oxygen it needs, fill it, and then recycle the water.
So on one hand, this saves a lot of water, almost 90% of the water is, if not more, is getting recycled back within the system. The second advantage is pest management. So we see a lot of contamination that happens when using soil based systems, but with hydroponics, you’re really creating that barrier for preventing a pest and contamination.


Jim Anderton: 
What sort of crops do you anticipate will be used with this system? I see some green leafy vegetables. I believe it is in your background.


Jamil Madanat: 
Yeah. What you see here is a couple of what we call grow modules being harvested from underground, extracted and just organized in a radial manner here for harvest. Initially, we’re starting with leafy greens and herbs. Now, the good thing about leafy greens is that they require less input from nutrients to energy to light, and they have faster harvest cycles. So starting with those gives us the advantage of running harvest cycles faster. So we get to iterate faster. Plus, these plants just tolerate variations in the environment, in nutrients much better. So this way, any changes and tweaks we run to the system will still end up with a higher success rate of production.


Jim Anderton:

Now, it’s interesting. So you’ve found a way to take that farming underground, and of course we’re interested in the engineering aspects of this. By the sound of it, there are several. It sounds a bit like a closed loop system. So we’re talking about gas, we’re talking about water, we’re talking about heat. So there are energy flows and there are physical flows of things happening inside these chambers, and there’s also a structural component, in that you have a vertical shaft. Are these things made of ferrous, non-ferrous metals, reinforced concrete? What’s the basic structure made of?

Jamil Madanat: 
So the casing would be made of steel. We’re using special coating. The special coating has to take into account multiple factors. One, obviously being non-corrosive, nothing to contaminate the plants. Second has to be antimicrobial. So it just doesn’t promote any growth of algae. Third, we’re adding a white coat layer that incentivizes light reflection. So this way, in this tight space that you have underground, you’ll be able to capture most of the lights.

Jamil Madanat:
In this tight space that you have underground, you’ll be able to capture most of the light that’s feeding the plants. So the casing is still with coating on the outside and on the inside. As for the internal structure, it’s a pretty simple structure. I can’t get into too much detail about the current design that we’re working on, that’s still being developed and just has a patent protecting behind it. But I wouldn’t say anything too complicated and we’re always keeping food safety in mind and operational ease in mind when designing these facilities.


Jim Anderton: 
Well, it sounds like you have multiple systems that are sort of interrelated and overlapping at the same time. You have a mechanical engineering task, you also have sort of dynamic systems operating at the same time here. How complex is this from an engineering perspective? What sort of tools do you use to design these things? Are we talking about a conventional CAD system, FEA simulation? How do you go about this?


Jamil Madanat:
 Absolutely. So generally when we look at the design of the forge, we split it between structural, mechanical, electrical, and digital systems. Now, the biggest challenge that we found with designing the forges is that once you go underground, there’s very little literature on the climatization of these farms, which has to be done very finely tuned, very controlled environment. As a side note, you’d be surprised how plants sometimes can be very sensitive, certain variations, as we plan on expanding the crops. So we got to be very certain what the environment looks like underground. And to do that, we had to work with Maya HTT to simulate how the heat transfer happens at different soil levels, at different soil types, different humidities. So maybe that’s something Carl can tell us a little bit more is they provided cell work to us, helping us understand better how to climatize the environment underground.


Jim Anderton: 
Carl, tell us about it.


Carl Poplawsky:
 The simulation services group at Maya HTT focuses on what we call virtual prototyping. Virtual prototyping is a technique that we use to test the mechanical design long before it’s manufactured or before physical prototypes are produced. We use computer rate and engineering software, CAE software to do that. We standardize a Siemens software products in center 3D, and we look at, in this case, the thermal and flow situations going on within the vertical farm. Our contributions to this project focused on energy efficiency and water use. So we use the software to predict the cooling load within the silo or the vertical farm. That cooling load is a function of not only conduction to and from the surrounding earth, but also quite a lot of heat load caused by the lighting that is necessary to provide the solar heating for the plants to survive.


Jim Anderton:
 Carl, that sounds like you got several inputs here going on at the same time. It’s also got something which goes is vertical, that’s goes to quite a depth at this point. Is there a temperature gradient from the top to the bottom of this system?


Carl Poplawsky: 
Yes, absolutely. It starts with the earth itself in that the surface of the earth is basically an ambient temperature. As you go down the surface of the earth reaches the constant, relatively constant temperature within the depth that we’re talking about. And so you have a temperature gradient going down through the earth, and then when you’re pumping air down into the farm, you’re going to see heat transfer happening, picking up heat, and then you have to bring that hot air back up and through the heating, ventilation, air conditioning system. We call it the HVAC system. So we help to size the preliminary sizing for the HVAC system and also the piping and pump sizing in order to remove the condensation that collects at the bottom.


Jim Anderton:
 Carl, tell me about humidity. that’s something where anyone who’s a greenhouse operator, of course controlling humidity is a major factor here. This sounds much more challenging. This is rather a closed system with artificial light and the heat input coming from that, plus the aspect ratio of this operation seems to be quite high. It’s a tall, skinny cylinder. Is that a factor?


Carl Poplawsky: 
Well, first of all, the software handles humidity and condensation and it can predict the condensation that collects on the walls and also provides the relative humidity distribution throughout the system. And of course, that has to be controlled pretty tightly for the plant’s health. And that will certainly influence how the final HVAC design evolves.


Jim Anderton:
 When you’re designing an HVAC system, and Jamil, maybe I’ll throw us back to you the same way is that it’s much engineering development is iterative and in a lot of cases if you’re breaking out, breaking new ground and doing something which has not been done before in the way that you’re doing this, building prototypes, testing, breaking them, going back and redesigning is a very common way to design components in areas like automotive that I’m familiar with. You’ve got a very large and expensive and complex process here. You can’t dig a hundred holes and then iterate a hundred different designs and then go back and then figure out what works down there. How do you cut the corner on that? We know simulation is a great tool to do this, but even with simulation is that you’ve got multiple variables interacting at the same time here down there. Do you have simplifying assumptions you work with or do you just crunch the numbers in a brute force way? How do you attack it? How do you attack this problem?


Jamil Madanat:
 Absolutely. So, obviously following the engineering method, we go with subsystem testing. So you can’t test the whole system altogether. And as you said, drilling multiple holes on your ground, it’s an expensive process. Same thing is you can just keep flying rockets every time you want to test something there. So what we really try to do is take isolate systems and try to experiment, iterate, and prototype with them. So be it that the lighting system that we’re working with, digital systems and how they work with the lighting system, the climatization system, and how it works with the plants and then for things that we can’t really replicate.
So we do have multiple labs that are running multiple experiments in parallel, whether it’s the extraction system, whether it’s the integration of, call it the hydroponic system with the controllers and specifically for the HVAC and climatization we said, okay, we want to validate a couple of major assumptions, which is how much heat does this soil absorb during the day cycle? How much heat does it retain during the night cycle? And let’s run the simulations based on these assumptions. That would give us at least the groundwork for what we know is at least true. And then you build foundations from there. But yeah, always working from first principles and running subsystem tests, then you can validate and build on top of these building blocks is generally the easiest. Easy is an overstatement, but the best approach to get a more comprehensive design.


Jim Anderton: 
Plants are an interesting phenomenon. We know that early designers of space station systems had considerable difficulty with things like humidity control, temperature control, also in a closed system. In this case you’re looking at plants and the amount of biomass inside your system is considerable at this point. And of course transpiration is a factor here. So the plants are an active component of changing the environment that they work in. Is there a difference depending on the type of crop that you grow, are those lettuce leaves different from a different type of plant?


Jamil Madanat: 
Back to kind of controlling the climate of the plants, we have external factors and internal factors. On the external factor side, obviously, and I think I’d like to point out an important piece of the design we’re taking into consideration that I haven’t alluded to before, but when you go underground almost worldwide, below the seven meter mark, the temperature converges to the annual average regardless of what the surface variation is like. When you do the simulation, you want to take that upper piece of variation into the climatization model and then account for that kind of all the way consistent climate that you want to account for. Now, the interior internal variations that are taking place is one, depending on the crop, and second, depending on the growth stage of the crop.

Initially, the first about two weeks, we almost assume no humidity generation. The plants are just growing, evapotranspiration is very low, and then it just exponentially increases in majority of the crops. Now, different crop size also breathe differently and need different humidity levels, different temperatures and different light requirements. Taking all these into account, initially we’re working with all leafy greens that just have a very kind of narrow window of variation. Then as we validate one, we just build on top of the others. Hope that answers your question.


Jim Anderton: 
It does. Carl, if a mechanical engineer were to design, for example, a ground source heat pump, you can approximate that roughly into sort of a coiled two heat exchanger and think about what we think as the classic sort of convection conduction radiation equations of heat transfer, integrate them with your form factors, your shapes. But in this case, they’re so many other factors going on here that are complicating this issue. Is this something you could run on sort of when you think of a stock simulation software. Does this require a coded solution, a low code, no code solution?


Carl Poplawsky: 
No, it doesn’t require any additional coding. The software out of the box handles this problem. It is quite complex. Not only do you have the temperature gradients going down through the earth, but you have the humidity distributions and everything else going on. Well, actually in the initial simulations, we looked at just pumping air down in without the HVAC in order to calculate what sort of HVAC requirements we need. The software is quite sophisticated at this point can handle those kind of things. Again, what we’re really trying to do is shortcut the design process to some extent in that. As Jamil mentioned, and you mentioned you can’t go out and drill 50 holes. So we’re going to drill 50 holes in a virtual environment with software and look at the mechanical thermal float performance of design. Then when Jamil is ready to drill those holes, he’s going to drill only a couple because he’s going to have a much higher probability of his design being successful, thanks to the virtual prototyping,


Jim Anderton:
 Carl, many users simulation tend to think of it in turn as a validation tool as much as the development tool essentially. We know where we want to go, here are targets, we check our design, does it work or does it not work. However, we know the simulation can also feed useful information back in multiple ways into the design process. The aerospace industry, for example, sometimes they discover things they didn’t understand or didn’t realize about a design by actually sort of flying it virtually with simulation. Does that happen in this case too?


Carl Poplawsky: 
Absolutely. One of the major advantages of this technique is we can look at what I call coffin corner conditions. These are conditions that maybe you can’t physically prototype. Of course one example is in a spacecraft craft industry, when you’re flying something in outer space and we get involved in a lot of spacecraft applications, you can’t actually test all those things in a terrestrial environment.


Jim Anderton:
 It’s funny you mentioned Jamil, more than one engineers proposed that what you are doing may be the only way to actually feed colonies some places like Mars and the moon. Is there a possible connection here?


Jamil Madanat: 
Potentially. I mean, virtually you can drill these systems anywhere. The fact that they’re insulated from the environment that’s happening on the surface, regardless of weather conditions, so severe hot or cold climates. The fact also that you can just build this standalone system, lock it, seal it, let it run its harvest cycle and then move after. With little maintenance, most of the work can be done once you put in the plants and then once you want to harvest them, makes them potentially viable for multi planetary agricultural systems, so you never know.


Jim Anderton:
 Jamil, how did you go about approaching Maya HTT and Carl in this process? Did you have a problem and then say, “we have a complex problem here we need to solve,” or did you run up against a roadblock that seemed insurmountable? How do you get to that point where you say, “Wow, I need to consult with someone outside my industry just to fix this problem?”


Jamil Madanat: 
I’ll share with you kind of the thought process that we went through here, which is try to understand how the environment for the plants will look underground and ultimately you just want to make sure the crops are climatized based on the what we call the crop [inaudible 00:21:59]. Now, generally you design based on surface conditions. You say, okay, well the temperature outside is going to be as such, we want to climatize the environment up. Inside we want provide this much heating or cooling or dehumidification because outside the kind of temperature and climate variables are very well defined. Underground, while looking through different literature and even engineering formulas in front of me, I realize that the problem is just multidimensional or multifaceted. It’s not only I have to understand, well, okay, we have these LED columns running between 12 to 16 hours a day providing heat to the soil.

How much of this heat is the soil going to absorb? What kind of soil absorbs the most versus releases heat the most? Once you go underground is you have just a wider gradient of different soil conditions and different humidity. Based on the humidity, how much heat is going to be absorbed, how would it be retained? And then let’s scale this a little bit more. You have a grid of, let’s say three by three or 10 by 10 forges, how close should they be to each other so they won’t affect the heat between each other. That kind of dynamic or behavior of temperature distribution underground at different soil conditions is not something that you just can pull off at the back of the napkin. That’s definitely where the Maya HTT team came in very, very handy and useful at helping us understand it.


Carl Poplawsky:
 Our simulations found that the performance is heavily dependent on the soil conditions, the amount of moisture in the soil, whether it’s sand or a clay or something like that. When you run these simulations, you have to make some judgment calls about how large the earth domain is going to be because it has to extend out way past the vertical farm in order to get correct results. So for instance, if you look at image number one, what we’re showing here is the temperature distribution in a cylinder of the earth, in which the shaft or the farm is contained. And we’ve got some color bars here. Red is hot, blue is cold. So you can see it’s cooler at the bottom. You can see how the temperature contours flatten out. That’s giving us a good indication that this cylinder, this basically arbitrary cylinder of earth is large enough so that the simulation will be sufficient. And of course, this was also going to tell us how closely you can space these things. If you look at image number two, this is the closeup of the air domain within the farm itself. This is just the top. That cylinder over on the left is the air handler.
You can see the little squares in the middle are the LEDs. You can see how they’re producing heat. Again, red is hot and blue is cold. So these are examples of the kind of results we can get. These are of course temperatures. If you look at simulation, I’m sorry, image number three, you’ll see that this is showing the temperature contours going down through the depth and it’s very easy to see now how we do have some significant gradients there. Image number four shows the velocity profile. So we’re pumping air down in and then it has to come up. And of course we don’t want to tear the leaves off the plants, so we have to be concerned about what that velocity profile actually looks like.


Jim Anderton: 
Remarkable. Intuitively growing things is something that’s as old as humanity. So intuitively we have a simple process. You’re going to take a greenhouse, we’re going to stick it underground. Carl, you just showed us, is that this is more akin to the engineering environment in a space station than it is to farming in a sense. There’s a lot of complexity, a lot of things going on here. Generically for companies that have things that have a lot of things going on, like you have here, Jamil, Carl in this case, how much does an engineer have to know to approach you with a problem? Do they simply have to say, “These are the parameters I have to hit with this design. Am I going to get there?” Or do they have to turn around and say, “I need results on this, this, this and this to understand how they interrelate?” Just how deep do they have to go?


Carl Poplawsky: 
Yeah, great question. Really, it’s the virtual prototyping techniques that provide the information for how all these parameters interact with each other. So as a mechanical engineer, we think about what we call control volumes. And so we have boundary conditions on those control volumes. Here, the control volume would be that cylinder of the earth, and Jamil is providing us certain boundary conditions that are going to influence the simulation. For instance, the heat dissipation of the LED lights, the transpiration rate of the vapor coming off the plants. So those are basic boundary conditions. And then we take it from there and providing the information on how all these parameters interact with each other. In particular, we can change the performance of the HVAC system and change the locations of air ducting, inlets, outlets, things like that, and look at the total performance of the system.


Jim Anderton: 
Jamil, did you have any idea when you joined this project that it would be as complex as it is?


Jamil Madanat: 
Not initially. Not initially, because you look into an idea and you’re like, “Okay, I think we can make it work.” But the more you dig in, the more you realize there’s just way more to it. It really is so multidimensional, as I mentioned, on the mechanical, the structural, the lighting, the just horticultural side. So ultimately we have the plans at the core of our design and we want to make them happy. Also, on the other side, we have our operators on the farm and you want to make it accessible for them to extract and clean and work with it. And the unit economics have to work out too, the carrier capital expenditure and operating expenditure. So try to balance all these parameters together is a challenge. But that’s engineering, right? It just keeps you going and it’s that excitement of discovering something new every time we approach a new problem. So yeah, so far really enjoying it.


Jim Anderton:
 It is exciting. And one last question, Carl. For design engineers that are working with complex systems and have problems that require the kind of professional help that you and Maya HTT can offer, what’s the number one piece of advice you could give them to prepare themselves before they approach you with a problem? What homework should they do, before they present you with an issue?


Carl Poplawsky:
 Yeah, I don’t think it’s really not that unusual. You have to think about what your goals are going to be. If somebody comes and says, “I need a thermal simulation.” I don’t have a lot of information there. If he tells me that I don’t want the material to exceed certain temperatures and I need to minimize my energy consumption, now we have something to work with. So it’s really just like anything else in life. You have to decide what your goals are going to be and then we help you meet those goals.


Jim Anderton:
 An exciting project. Jamil Madanat, GreenForges. Carl Poplawsky, Maya HTT. Thanks for joining me on the show.

Learn how Simcenter unleashes your creativity, granting you the freedom to innovate and allowing you to deliver the products and processes of tomorrow today.