Suzhou Space

suzhou space header image

Before starting at LMN, I lived in China for over three years, where I became fascinated by the classical gardens of Suzhou. These gardens, unique to southern China, are populated by small pavilions and halls, sculpted rock outcroppings, shallow pools, and planted areas, all contained within a perimeter wall and arranged in an aesthetically pleasing manner. While a pleasing arrangement of rocks may be enough to satisfy bored tourists, architects and designers find the gardens endlessly fascinating for the way they orchestrate experience, frame space, and create illusory depth in their limited boundaries.

Suzhou’s gardens are characterized by ambiguity of interior and exterior space, collisions of disparate forms and materials, and above all the intricate layering of space to create subtle perceptual shifts that serve to destabilize the viewer’s sense of space and scale, inspiring a kind of sublime wonder, a worthy goal for designers in any era or cultural context.

How are these effects achieved? Is Chinese garden design a lost art, or subject to codes buried in ancient design manuals? Can we leverage our computational tools to quantify what makes these gardens so beguiling? Could spatial analysis of these gardens hold any lessons for designers today?

intelligibility simple_0001_Layer 1

[single point isovist diagram – windows as obstacles]

intelligibility simple_0000_Layer 2

[single point isovist diagram – windows as void]

As a starting point, we can analyze a floor plan based on “spatial intelligibility” – a metric that indicates the relative visual accessibility or enclosure of a space. Intelligibility is best understood as an aggregation of vectors, a summation of the length of direct sightlines in all directions. From any point in the space, we draw projecting rays outward until they hit obstacles, measure the distance from the origin point to the intersection point, and add the values. (For this example, we’ve used Grasshopper‘s Isovist component.)

intelligibility simple grid lines

[isovist grid with radial projecting lines]

intelligibility simple grid color

[sample points with color map gradient – low density of samples]

Extending this logic to a grid, we can find this intelligibility value for every grid point and create a gradient map that visualizes the relative visual legibility of the space. In this simple example, we can see clearly that the ‘compressed’ space to the left has lower intelligibility values than the ‘open’ space to the right.

While this type of analysis, based only on physical barriers, could be useful (for retail planning, for instance, or urban design), we have so far assumed only solid obstacles. What would happen if we considered barriers that allowed visual, but not physical connections between adjacent spaces?

intelligibility simple grid color-dense

[sample points with color map gradient – high density of samples, windows as obstacles]

intelligibility simple grid color-dense - window voids

[sample points with color map gradient – high density of samples, windows as voids]

In Chinese gardens, there is a disjunction between physical accessibility and visual accessibility: windows often open to adjacent spaces that may be accessible only through convoluted, indirect routes. Screen walls divide space but still allow views through. While it may not be the most efficient layout, this complication is the root of many of the gardens’ pleasures, and is worth studying further.

For this analysis, we’ve picked the Garden of the Master of Nets, one of Suzhou’s most celebrated gardens. A moderately sized garden, it contains several areas of interest: a relatively enclosed section with halls and courtyards along a linear axis, a large open area around a shallow pool, and several intriguing areas where covered walkways collide with small pavilions and rockeries. By analyzing the plan for both physical and visual intelligibility, we can begin to quantify and visualize the unique qualities of each space.

Master of Nets - Sketch Plan

Master of Nets - Base Plan

Master of Nets - Physical Access

Here, we’ve mapped physical intelligibility; how accessible is each point on the grid? Note the values around the pond: though visitors can see across, we’re treating the pond edge as a boundary.

Master of Nets - Visual Intelligibility

Here, we’ve mapped visual intelligibility, and so we’ve removed the obstacle objects wherever a window or screen wall allows views through to spaces beyond. Note the pond edge again, some of the longest sightlines are to be had around the pool.

Master of Nets - Disparity

Finally, by calculating the difference between the physical and visual intelligibility values, we can create a map of the spaces where the disjunction between the two is greatest, and we can surmise that these are the areas where the surprising perceptual effects of the garden are most evident. The pond edge is worth noting again, as the views are open, but movement is restrained. More surprising, areas that are dense with rockeries and covered walkways are highlighted as well. These areas are physically divided but visually permeable thanks to the corridors’ open edges and prevalence of screen walls and windows.

Visual connectivity without physical access could be frustrating, but here is is an operative method of the design, and a technique that could be adapted for future building projects if such an effect is desirable. Designers today often deal with situations where circulation paths are kept separate for security, or sanitary, or social reasons. In such cases, visual connectivity may be encouraged as a way to enliven a users’ experience of the space. In retail design, intelligibility maps can be used to anticipate shopfront visibility, and thus sales. Extended to a sectional analysis, or a three-dimensional matrix, similar techniques could be used to analyze the overall legibility of a stadium or auditorium. In urban planning, these metrics could be used to anticipate traffic loads, or to speculate on usage patterns of public spaces. In all cases, intelligibility ties computational power to qualitative effects, and is worth further study as a metric that help designers bridge the gap between the digital and the physical.

Garden of the Master of the Nets (1)

Garden of the Master of the Nets (5)

Garden of the Master of the Nets (15)

Garden of the Master of the Nets (20)

Garden of the Master of the Nets (21)

Suzhou - Master of the Nets Garden (6)

Suzhou - Master of the Nets Garden (1)





  • Marc Syp

    An intriguing layering of computational analyses… I would be interested to see some observational research mapping actual user experience in the garden. That is, in the same way the Space Syntax conducted detailed and methodical observation of visitors to the Tate Modern gallery in their research on the correlation between intelligibility and movement in non-directed space, you could observe peoples’ responses as they move through the gardens in an effort to map these spatial affects that you describe.

    This, of course, would be significantly trickier than mapping movement, which is objective and directly observable, but that is in accordance with your ambition to map qualitative aspects which are not easily quantified. There is a strong body of work in UX research that begins to point to measurement methods that begin to scratch the surface of this problem through the scientific method in the social sciences.

    You formulate a hypothesis by answering questions like “What characteristics can we observe that might indicate an experience of the sublime?” … and you begin to measure things like the number and specific location and frequency of pauses in a wanderers journey. You measure heart rate. You track skeletons using Kinect sensors and look for irregularities in gait, subtracting incidents around known topological features that may bias the sample. Or perhaps you introduce topological features and count the number of stumbles that happen that indicate a state of distractedness. And then you straight up ask people to point to the places that they think are the most sublime.

    As the correlations become clear, the features that describe the phenomenon become more specific, the methodology for evaluation becomes more effective, and parameterization of the spatial conditions with the characteristics in question perhaps become more attainable…

  • Evan Chakroff

    Marc, thanks for the comment! I’ll have to check out Space Syntax (again), the Tate analysis sounds fascinating. The UX approach is intriguing as well, as is the concept of proprioception (in psychology), loosely defined as a subconscious spatial awareness of one’s environment, which, if quantified somehow, could lead us towards a much richer understanding of the psychological map of a given space. For designers of architecture and environments, lessons from UX design could come in handy, especially as digital displays and interfaces become more ubiquitous.

    I would also love to get cracking on an art installation in which we ‘introduce topological features and count the number of stumbles’ – know any gallery owners who’d be game?

  • Addison Godel

    Very intriguing.

    It seems like the next step has to be a more elaborate theory of what constitutes “intelligbility” – there are lots of criteria/ideas in play here, would be good to unpack which exact set you mean. e.g. intelligbility as reading of a discursive sign (Venturi and Scott Brown), intelligbility as “I can tell where the edge of this space is” (hmm, maybe gestalt theory? not sure how to categorize this), intelligibility as “i understand this as a place/space and can orient myself in it with reference to other things” (mix of the above, I’m thinking here of the analysis in “The Image of the City”).

    And of course, there’s the variable of view in *section*, where I’d want to turn back to analyses of Raumplan, etc – sometimes the place you are at can be seen, but you can’t see the seer. This would be modified further by material conditions and light: looking out from a jali or indeed the screen windows here is different than looking in, and also depends on inside/outside lighting conditions, the so-called two-way mirror effect.

    But I do love the imagery and the implied potential use as a design tool for taking in spatial/perceptive complexity as, ultimately, a parametric variable! I do agree with Marc’s point that you stake out a claim that your tools actually aren’t prepared to test (re: experience, the sublime) but I think it’d be better to change the claim than to try and figure out those tools. For one thing, the user-experience survey would be a nightmare to organize, as well as to structure (SO many testing problems here… subjective variables, varying ways people might verbalize things, etc.) and you would also hit another big definitional wall: what do you mean by the sublime? Burke’s? Kant’s? Someone else’s? *Is* it a physiological thing? Or does it happen in the mind? Or does it exist in the objects themselves?

    Forgive me, as I’ve been knee-deep in 18th century aesthetic theory this week – I just think it’s a swamp you don’t want to, or need to, enter. What’s really at stake is a modeling and analysis technique for vision, with the allowed caveat that vision is here assumed to be “modern” vision (the all-seeing, camera-like perspectival I/eye). It’s about things being blocked or revealed, and testing how much of the overall plan is made visible/illegible when you add or subtract opaque matter at a particular point in plan. That’s got tons of potential in itself.

    I’d also encourage testing it on some other things/spaces so we can see how much it’s modeling something specific to Suzhou gardens, and make sure that the tool doesn’t make ALL spaces seem intriguingly ambiguous, etc. Alternately, though, you could use it to illustrate different KINDS of ambiguous and disappearing-reappearing space. Would be great to get it into section and run Vaux-le-Vicomte’s ha-has through it… etc.

  • Thomas


    great work – and deffinitely something to look into further!

    Would it be possible to share a screenshot or so of the Grasshopper file so to know what exactly you did here?

    Thank you very much

    cheers Tom