From Flat to Fab: Greening Landscapes with Landscape Information Modeling and Artificial Intelligence

DIANA CRISTINA BINCIU: Hi, and welcome to our session-- From Flat to Fab, Greening Landscapes with AI and LIM. About the session, shortly, we think it's time for landscape to step up the digital game and claim a bigger slice of the design spotlight with architects and engineers. The From Flat to Fab session focuses on exploring the process of transitioning from 2D landscape designs to Landscape Information Modeling, as in LIM, to gaining insights on how we can use AI at the moment and discover the possible role of AI in enhancing sustainability outcomes and processes.

My name is Diana. I am an urban architect and the BIM specialist for Henning Larsen Architects in the landscape team. And I've been working in this role for the past two years, trying to help the team enhance their 3D capabilities. And this is my colleague, David.

DAVID FINK: Hi, my name is David Fink. I'm the digital manager here at Henning Larsen. So I work with our digital platform for developing projects and also help with the tools, develop working processes, and look for the new technologies that are coming up in the future. So I've been involved with digitalization since around 2007 with the Danish BIM mandate, which was a requirement that all public buildings be done in BIM. So I have a pretty good background in digitalization now.

DIANA CRISTINA BINCIU: Now, we both work at Henning Larsen. For those of you who don't know, the company was founded in '59 by the Danish architect Henning Larsen, whose legacy of creativity and learning we still carry today. We navigate through complex connections that aim to bind together our built environment, ecological systems, and societies in large.

We work across multiple disciplines and we try to maintain acute attention to details whilst never losing sight of the bigger picture. We're also quite big. We have over 600 employees across studios in eight countries.

Like I'd mentioned before, we have a big portfolio. It spans through a range of disciplines-- through architecture, urbanism, landscape, interior, and graphic design, and then research and innovation.

The agenda for today-- we will start by selecting a study case, in this case, a project called Faelledby that we will take from 2D to 3D to LIM. And then we will show how we use artificial intelligence technology that exists today and then what we envision for the future.

The project is called the-- Faelledby. So the Faelledby project is designed by Henning Larsen, is a sustainable residential development located in Copenhagen, Denmark. It envisions an eco-friendly community where urban living harmonizes with nature. The design aims to preserve 30% of the natural landscape and integrates local biodiversity. It's aiming to blending greening spaces with residential areas, and biodiversity, and so on.

It's modeled after small villages. Faelledby fosters social interactions while emphasizing environmental sustainability throughout the use of eco-conscious, materials, renewable energy, and nature-based solution. This project aims to strike a balance between the human factor and the natural environment.

In short, it's Copenhagen's first timber neighborhood. Faelledby brings the urban environment into harmony with the natural landscapes, establishing a community of ecologically responsive housing units organized according to the model of a rural village.

Some facts about the project. It's around 44 to 45 acres in land area. 2 million square feet of building area that's supposed to house around 7,000 residents. 80% of the structural materials should be timber. And 45% of the site is used for natural habitats. And, of course, a project so complex involves multiple stakeholders.

This is how we envision the project. All buildings are supposed to use sustainable materials, primary timber construction-- primarily timber construction. It's supposed to mix nature into the building environment and increase biodiversity.

And, of course, we are actually building it. It's on its way. You could also see the Ramboll building there in the background. So, hopefully, in a few years maybe you can come visit and actually see the project.

DAVID FINK: So-- but before we get into looking at AI and how we use it for our landscape workflows, I thought we'd just take a little journey to see how we implemented using 3D landscape and LIM in-- at Henning Larsen.

So we started this journey about five years ago. Before, we were primarily using MicroStation and AutoCAD for landscape design. But now we've moved to Revit and Rhino as our primary platform. So this move was not easy. And we're still not finished, but we've definitely come a long way since the beginning.

So the idea is, at Henning Larsen, that most of our projects start with a story. In the case of Faelledby, it was the merging of the small town and the big city and nature areas. So we started with, of course, hand sketches designed-- from a design architect. And this was quickly translated into a Rhino model, which was the primary sketching tool for the early phases.

We use Rhino primarily because it was easier to sketch and generate designs and also produce multiple design iterations. Then after the design became more fixed, we moved the project to Revit. But Revit, as we know, is a tool designed for building, so there's a lot of functionality that's missing in the tool. So, therefore, we have to use Revit with a series of plugins in order to make Revit a landscape tool.

So some of the advantages of using these plugins is it gives us the ability to do some fine terrain adjustments, model roadways and sidewalks, and then also get the graphic outputs that we need as far as deliverables. The benefit of having the entire model in Revit is it gives us a platform for coordinating with our architects and engineers.

And you can ask yourselves-- a project like this-- why didn't we do it in Civil. And the simple answer is just because Civil does not offer the tools and the workflows that work for our landscape architects. So we decided to use Revit.

So when we started, we pretty much started with an idea of how we wanted to model in the software. But we need to find out how we can use the software. And when we look at software today, software is the tools of our trades. It's the-- the tools that traditional architects have used in the past is their pens and pencils. And we see the software, as our-- the tools of our craftsmen. And we treat our architects and designers as craftsmen.

So, for this, we decided to reduce our software portfolio to Revit and Rhino, eliminate the MicroStation, and then use AutoCAD only when necessary for importing and exporting information. Then we had to align the entire workforce in the office of how to use the tools more effectively. So we developed a set of whitepapers that describe how we work, and the tools that we use, and how do we use the tools.

And this-- these whitepapers became the basis of our education, of how do we upskill, and how do we teach the landscape designers how to use Revit and Rhino in the most effective way. So, with this, we also had to explain what we wanted in the end. So we have a set of sample models that we use. And these sample models are from different levels of development and different phases so that everyone is aware of what we need to deliver, when we need to deliver it, and how detailed the models should be so we don't re-- minimize the risk of overmodeling or undermodeling and we can meet the demands of our clients.

But then, of course, we still need to deliver drawings and drawings is probably the biggest challenge because we have a lot of project architects in the office and they have their own idea of what a good drawing looks like. So we've also tried to make an office standard of the way the drawings should look after we come out of Revit. And this is also how we communicate to the craftsmen on-site many times.

So when we look at our design processes, we start with competitions. A lot of our work is won by competition. And then these particular projects go very fast. So we use Rhino as our sketching tool. And this is the tools that our designers know.

In Denmark we have two different schools of architecture. There's the design school and the technical school. The technical schools use Revit, the design schools use Rhino, and the two schools don't talk to each other. So we have to rely on the tools that they know when they come into the field.

So, with this, we also have this set of sample drawings to show what we want because it's not easy to explain in words. It's much better to explain in graphics so we can harmonize the office standards by having these office standards of what we want the drawings to look like out of a competition and the conceptual design phases. So this is-- we found to be the most effective way, especially when you're communicating with architects who think more graphically.

Then as we move through the design phase to schematic design, this is typically where the geometry of the project becomes more fixed and we make this migration from Rhino to Revit. If you are-- there's many ways you can make this transition, but we found that we have some pretty good workflows for moving geometry from Revit to Rhino for landscape. So this is also where we make the move from the design teams to our production teams.

And then, also, when we have projects at this phase, we have BIM execution plans involved-- or in place for all of our projects so we know what to deliver, when we deliver it, and we've also calculated the extra cost of some of the services we provided into our contracts.

And then when we move into final design, we're mostly using Revit. I have to say we still use Rhino for some design sketching, but it's-- Revit is our primary tool and this is what is handed over to the client. And because we're in Denmark and we have these BIM requirements, many times we hand over IFC models in addition to our Revit models, and still some of the contractors are asking for DWG exports. These DWG exports are often used for grading the site.

So when we have this Revit in model-- Revit model in and-- in our common data environment, which we use Autodesk Construction Cloud, we have the ability to do a high level of coordination between the architectural and landscape. So we know exactly how the landscape will behave when it meets the buildings. We can ensure that we have the right slopes, we have the right access, and we also have the ability to visualize the project quickly, merging both the landscape and architectural models together.

But in order to get this to work, we also need to have a good content library. And this is easier said than done. We try to develop our content from projects because that way we are ensured that the content we develop is used on projects. But this is very complex and tedious task. And like our drawings, the 3D models and the objects, they also are under scrutiny about what they look like. So we've also decided for the office of what the graphic appearance should be for the 3D objects in the models.

And one of the hot topics is what a tree should look like in 3D and what a tree should look like in 2D. And this is just one of the examples of the discussions we have.

So, recently, we've started using something from Autodesk called Content Catalog. And this is an online, cloud-based content management system, previously known as UNIFI. And we see this is a real game changer for us. Because-- since we're an international office, we can have these collections tailored to the different geographies and we're moving from where we previously stored our content on our file server to a more graphic way that's searchable and more easily used by the entire office.

So we're in the process of rolling it out. It's just come off of beta about a month ago. So we're pretty excited about this. And I think everybody in the office is excited as well to start using this.

So we need to take a look at also how we start in Revit. And when we start with our projects, we always start with a site survey. And this site survey becomes the basic foundation for the project. Because this has-- its positioned on Earth and we have the coordinate systems, so all of our projects start with some site survey.

And we use this site survey from the start and it's important that we can round trip back to the site survey in the end, especially on a site like Faelledby, where we have the 45 acres of land with the GPS-controlled machinery to grade the site. It's really important that we end up in the same place that we started.

At the same time, we also use this 3D information that's captured in the site surveys to make our initial topographies that we use for developing the site. So this site survey is the word in the office of where we start. And without the site survey, we don't know where we are in space. Of course, we can start a project without it, but we will quickly position the site in space.

And one thing we've learned is this site survey-- the survey point in Revit is the controlling point of everything for landscape. We found out that modeling the terrain from the site survey elevation and also the position in relationship to the site is critical.

So you can move the site survey close to the site by unpinning it and moving it. But we found out that it's much more easy for all of our landscape architects to understand if we leave it at the 0, 0. That means they know exactly where the origin point in the AutoCAD site surveys are and they know exactly where the site is in space.

So as I mentioned before, we need to use additional tools to model landscape. And one of the packages of-- plugins that we use and we really appreciate is Environment. And Environment, we found out, fills in many of the holes in Revit, especially for landscape. I mean, if you look at the landscape tools in Revit, there aren't that many. There's just maybe four or five. But with the-- a tool like Environment, we get an entire portfolio of tools.

And one thing about Environment that we really appreciate is it's a tool designed by landscape architects for landscape architects. So the workflows in this-- the plugin really work for the landscape architects that we have in-house. It's easy for them to understand and easy for them to use. So we really like to use a tool like this. And it's also easy to explain the workflows because it's a lot of the same terms that we use in our daily work when working in the software.

We also found out that just because we have landscape doesn't mean, we can't-- we don't need to do coordination. We do a lot of coordination for our architectural projects. And we need to do the same for our landscape projects.

The sites are getting more and more complex with underground utilities. And we want to also ensure that we can place our trees without placing them on top of the utilities. So the coordination of landscape models is just as important as the building models. And, for this, many of the time-- many of the projects use Navisworks. But because we have these IFC models that we need to deliver, we also use Solibri. So we have our rule sets all set up for landscape. So when we need to do these coordinations, we can quickly set up a Navisworks project or a Solibri model and do this coordination.

So, also, the tools that we use need to play together. So interoperability between Revit and Rhino is really important. So we have some tools in Environment that can help us out with this interoperability, but we also use other tools as well. We're looking at some cloud-based tools and also like Speckle and as well as the Data Exchange.

So we use a tool called Beam. And we see Beam is really good for single objects. So if we want to move those from Rhino to Revit-- and we have a interiors department and design departments and their primary tool is designing in Rhino. So we use Beam to push the geometry we get in Rhino to Revit as families. So it's a good way of trying to keep as many native objects in our Revit models as possible.

But when we talk about larger projects and larger amounts of data, Beam is not so good for moving entire projects, so we use tools like Speckle. And this is how we migrate our geometry back and forth between Revit and Rhino for landscape. We're looking at using the Data Exchanges for topography, but we haven't developed our plugins for that as well.

So if you're interested in seeing how you can use Data Exchange for Revit and Rhino, you can see the session that I also presented at this year's Autodesk University. So now I'll hand it back to Diana and she'll talk about some of the challenges we have in landscape.

DIANA CRISTINA BINCIU: Yes, looking back at what we did from 2D to LIM, we found out we have some challenges. So one of them is that there is a limited scale. So big data and big scale projects more than 32 kilometers we cannot-- we simply cannot work with.

While Revit has a Toposurface tool, it is relatively basic compared to the needs that the landscape architects need. It lacks the ability to handle more and more complex grading, more detailed terrain, precise contour modeling. So advanced tools for retaining walls, slope analysis, and cut and fill calculations are limited.

There are no native tools for managing or analyzing soil types, subgrades, or creating detailed planting bed designs that could reflect soil depth, for example, or drainage conditions. This type of information is quite critical for a landscape architect in order to ensure proper plant health and site sustainability. Simply, Revit does not allow for multilayered terrain models that will show different types of soils or layers beneath the surface, which is also important for planting, for the root depth, for stormwater infiltration, and hardscape foundations.

And regarding irrigation, Revit doesn't include specialized tools for designing irrigation systems, such as automatic piping layout for sprinkles-- sprinklers, sorry, drip systems, or controllers. So integrating water-efficient landscape design elements is quite difficult without third party add-ons.

And landscape architects often rely on GIS data to inform their designs. So Revit has some ability to import external data, but it lacks the integration with GIS and this-- it's making it hard to incorporate topographic surveys.

Another thing is that it's quite difficult to produce the plans and the sections that are graphically appealing to landscape designers-- the lack of comprehensive plant databases. Revit includes some basic plant families. But these are limited in variety and don't reflect the diversity of real world species. They don't really have advanced information like growth patterns, seasonal changes, or maintenance needs.

It doesn't allow for material assignments-- but it does allow for material assignments, but doesn't have a rich library or customization tools specific to landscape paving materials. And, overall, the smoothness of elements is quite limited.

Some pros and cons. We think it's quite powerful for integrating landscape with architectural designs and engineering models. It's also great for 3D modeling and BIM workflows, but I do think we need some plugins for advanced landscape features.

You might ask if it's enough to have a good workflow, enough to have a good database, or a good program, or good tools? I would say no. We need to deal with the new hot topic of today, which is AI. So I'm going to show you how we use the current AI technology in order to promote and to choose our design ideas.

We've been looking at multiple platforms for it and we came-- we selected these three that match our studio identity-- Midjourney, RunDiffusion, and Runway, which are cutting edge platforms that leverage AI to create stunning renders.

So Midjourney is known for its ability to generate detailed and artistic visuals from text prompts, making it a powerful tool for exploring landscape designs. In this case, for example, we can decide to change the sky and the atmosphere. It's also a very flexible platform to bring new ideas to life with basically a blink of an eye and have good results, could add cyclists or children. Overall, Midjourney simplifies the rendering process and it's enabling us to produce impressive visuals.

The second tool is called the RunDiffusion, which offers similar capabilities to Midjourney, but it's more focused on realistic landscape. And we use this tool to create scenarios for client meetings, for example, or for media content, or just analyzing different camera views for our submissions.

Then we go through Runway, that goes a bit further. And it's combining AI with video and 3D rendering. And its-- allows the designers to create a more dynamic and interactive landscape scene.

Together, these platforms are revolutionizing, in a way, let's say, the way landscapes are visualized and designs. It's offering both efficiency and creativity in the rendering process. So combining all of these three provides quite a powerful tool kit in the office for visual content.

These were the results. But how did we get there, you might ask? Pretty simple. We start with this level of information. We can take it from our Revit-- can take a Rivet snip, we can take an Enscape view or just any view from any model that it's quite simplified.

And then after we select the desired view, we start dissecting by using text prompts. This case, because the project is in Denmark, we might want a Danish summer sky with clouds because we have maybe 90% of the time cloudy summers. Then we do want that sustainable timber wood facade look and to just populate with some more in the background. And then when we get to the desired outcome, we can go in deep and work with landscape.

So, for example, let's say we want some landscape designs featuring lush Danish vegetation. Then we could get some different scenarios that we can talk and create upon. Or maybe we want some windy, green fields with the Danish environment and then we get different options. And, at the end, maybe we would like to explore with water. Therefore, you could see that there are multiple and it's quite a variation of what we can get. All of this was obtained in 10 to 15 minutes of work.

The pros and the cons regarding AI, I would say it saves time. Like I mentioned, it took me 15 minutes to generate some images from a Revit model. I would say it saves money because we don't have to outsource this to-- outsource it to other offices. It increases the quality. Sometimes it can elevate from what we did before. It's quite easy to rendering iterations and upscaling.

The cons. I would say the result might not be what we expect. I don't know if you've noticed, I posted this picture earlier, but underneath the car we have a rock. Hopefully it's a rock and not something more dark. But, yeah, so you could get some bugs here and there.

You also risk that everything will look the same. If all the offices in Denmark will use Danish vegetation, windy and cloudy Danish skies, I assume we will get the same vibe. So, yeah, it could risk into going into that.

And then, of course, it does save time, but it also took me a lot of time to explore and research. So you do need to think about investing in that.

Next, I will show you how we envision a bit AI in-- yeah, in the future and how it could help us with the project so complex as Faelledby. We came up with the four main points. We think that AI systems could analyze real time weather data to optimize, for example, the irrigation schedules to ensure that landscape receives just the right amount of water without wasting any.

Models that predict plant growth patterns, enabling landscapers to allocate resources efficiently based on seasons. Or we could have AI-driven soil sensors that monitor moisture level, for example, and automatically adjust the water distribution.

We could also have empowering sustainability practices. So we could have tools that are integrated with LIM that automatically generate sustainable landscape design that could-- for example, by analyzing site-specific environmental data, such as wind or sun. AI-driven LIM platforms that recommend plant species based on the local biodiversity, depending on where you are on the globe and could be promoting ecological harmony and reducing costs.

The third point is about enhancing ecological balance and reducing the carbon footprint in landscape. Maybe some AI algorithms that can calculate the carbon sequestration of different plant schemes because sometimes when you think about landscape, you might think that it's all green and natural and forget that designing landscape also produces carbon. So, yeah, this could be taken into consideration.

Then we could maybe have some AI-powered analysis of urban heat islands, guiding the placement of greenery to reduce temperatures and improve the air quality. Or some predictive models that assess the long-term ecological impact of the landscape.

So the fourth point is regarding new standards for sustainability metrics. There could be some platforms that benchmark landscape energy consumption. This would help designers minimize the energy use through strategic planning. We could use maybe AI to track and reduce the carbon footprint of landscape construction and the maintenance, setting the-- setting new industry standards regards to sustainability or systems that ensure the ecological equilibrium by optimizing the balance between built and natural environments.

I would like to close the session by raising up a question, and that would be, can we create better landscape designs with AI? Thank you. If you have more questions, maybe you reach us out on LinkedIn.