<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by KPF Urban Interface on Medium]]></title>
        <description><![CDATA[Stories by KPF Urban Interface on Medium]]></description>
        <link>https://medium.com/@kpfui?source=rss-e33d27980495------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*kYVuOYqmWfotzPRStxZv6g.png</url>
            <title>Stories by KPF Urban Interface on Medium</title>
            <link>https://medium.com/@kpfui?source=rss-e33d27980495------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Tue, 17 Nov 2026 02:32:23 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@kpfui/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[Bridging Data Science and Architectural Practice]]></title>
            <link>https://medium.com/@kpfui/bridging-data-science-and-architectural-practice-aceb3f23fd95?source=rss-e33d27980495------2</link>
            <guid isPermaLink="false">https://medium.com/p/aceb3f23fd95</guid>
            <category><![CDATA[product-development]]></category>
            <category><![CDATA[computational-design]]></category>
            <category><![CDATA[architecture]]></category>
            <category><![CDATA[cloud-computing]]></category>
            <category><![CDATA[data-science]]></category>
            <dc:creator><![CDATA[KPF Urban Interface]]></dc:creator>
            <pubDate>Tue, 18 Jan 2022 15:01:17 GMT</pubDate>
            <atom:updated>2022-01-19T16:01:43.273Z</atom:updated>
            <content:encoded><![CDATA[<p><em>Emergent applications of cloud computing in architectural practice and academia</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*rXFDwpX6iUzy551V37T1sg.png" /><figcaption>Rhino Compute Workflows</figcaption></figure><p><a href="https://medium.com/u/9615e23a2246">David Andres Leon</a> — Head of Computational Design <a href="https://iaac.net/">IAAC</a></p><p><a href="https://medium.com/u/9ff5c7b1b5f0">Brandon Pachuca</a> — Urban Data Scientist + Web Developer</p><p>Architecture and computer science are becoming increasingly intertwined. The data science industry found a promising niche to leverage big data, machine learning, and web development to create viable products to automate workflows and bring new insights to market. Architecture can benefit from these trends, adapting product and web development to support advanced simulation and visualization practices.</p><p>At first, this phenomenon invited a few daring architects to step into the realm of computer programming, shifting the architecture industry’s long traditional curriculum into a hybrid profile of a designer with programming skills: the so-called computational designer. At the heart of this esoteric trade is an attempt to shift the focus away from merely designing buildings toward designing tools utilized for their conception.</p><p>Computer scientists stepped into specialized departments of architectural firms to develop automations that enhance the design practice, creating tools to analyze building performance and craft user-friendly interfaces. These allow designers to make better informed decisions in the creative process and unveil aspects of design that are often hidden beyond two-dimensional plans.</p><p>Among other software, <a href="https://www.rhino3d.com/">McNeel’s <em>Rhinoceros</em></a><em> </em>has become a key player in this interdisciplinary venture. First, through the <a href="https://www.grasshopper3d.com/"><em>Grasshopper</em></a> plugin, the software allows easy access to programming through a visual interface, lowering the entry barrier for non-experts and providing relatively simple access to a vast ecosystem of plugins created by a thriving community of developers. The plugins, most of them free to use, encompass a wide range of know-how from different trades, formerly siloed in various other software.</p><p>Secondly, and perhaps more importantly, by providing well documented access to its development tools, McNeel has allowed for the software to become highly customizable, not only by exposing most of its functionality through their cross-platform SDK <a href="https://developer.rhino3d.com/api/RhinoCommon/html/R_Project_RhinoCommon.htm">Rhinocommon</a>, but also allowing for innovative workflows like <a href="https://github.com/mcneel/rhino.inside"><em>Rhino.Inside</em></a>, which allows for Rhino to be injected into other <a href="https://github.com/mcneel/rhino.inside#samples-by-host-application">software</a>.</p><p>One of the most interesting recent developments in the Rhino ecosystem is the <a href="https://github.com/mcneel/compute.rhino3d"><em>Rhino Compute</em></a><em> </em>project. Built on top of the seventh version of the software, it provides access to most of the features of Rhino through SDK functions via a <a href="https://www.ibm.com/cloud/learn/rest-apis">REST API</a>, allowing the software to run as a service for geometry calculations, either locally on a client computer or on a cloud-based Windows server. Simply put, Rhino Compute allows for alternative applications to harness the power of Rhino and Grasshopper through a programing interface, without the need to have Rhino installed.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/712/0*YQXGPFMOd7w1qQH_" /><figcaption><em>Rhino Compute as a Cloud web service</em></figcaption></figure><p>Rhino Compute helps remove a barrier to entry by abstracting complex workflows from both typical architectural designers while increasing extensibility for computational designers. Used as a geometry solver back-end, it enables the creation of user-friendly front-end applications accessible via a web browser, enabling architects to execute complex parametric workflows with minimal technical entry requirements.</p><p>Moreover, through the recent inclusion of the <a href="https://developer.rhino3d.com/guides/grasshopper/hops-component/">Hops</a> component into Grasshopper in Rhino 7, Rhino Compute can be used to serve as a link between all the plugins that Grasshopper has to offer and any other platform that is able to serve geometry by speaking the language of Rhino, whether that is RhinoCommon or the multi-language Rhino independent open source library <a href="https://github.com/mcneel/rhino3dm"><em>rhino3dm </em></a>(formerly openNurbs), or to a <a href="https://github.com/python/cpython">CPython</a> <a href="https://flask.palletsprojects.com/en/2.0.x/">Flask web server</a>.</p><p>Such is the case of the <a href="https://discourse.mcneel.com/t/create-cpython-components-using-hops-in-grasshopper/120517">workflow </a>described by the McNeel folks in their forum a few months ago, where they showcased the connection between Hops and CPython through the popular Flask framework as middleware, by creating an HTTP server that allows for the transfer of data back and forth between both platforms, both locally and remotely.</p><p>This innovative workflow has suddenly opened the door for ubiquitous and powerful tools widely used in CPython for data science, machine learning and complex networks to be used along with Grasshopper: tools such as <a href="https://numpy.org/">Numpy</a>, <a href="https://scipy.org/">Scipy</a>, <a href="https://www.tensorflow.org/">TensorFlow</a>, <a href="https://keras.io/">Keras</a>, <a href="https://scikit-learn.org/stable/">SciKit</a>, <a href="https://opencv.org/">OpenCV</a>, <a href="https://networkx.org/">NetworkX</a>, python libraries that formerly could only be used in Grasshopper through other complex workarounds.</p><p>In this article we would like to showcase a few of the possible workflows that this new technology can offer to the design practice with two case studies: one from <a href="https://www.kpf.com/">Kohn Pedersen Fox</a> (KPF) architectural practice and the other from an academic perspective, as taught in the <a href="https://iaac.net/">Institute for Advanced Architecture of Catalonia</a> (IAAC). Both showcase the early adoption of this technology and how it is contributing to bridge the gap between emergent technologies and the architectural practice.</p><h3>Compute Workflows</h3><p>The following workflows enable geometry calculation and solving on either a local or cloud-based Windows Rhino Compute instance. Each workflow leverages a different stack of technologies, which can often be interchanged or even combined depending on the desired workflow. For instance, a specialist/designer might calculate a computationally expensive Rhino command, such as solving a complex facade calculation. Traditionally, if the designer did not know how to use Grasshopper, they would contact a computational specialist in the office to write the workflow, copy the script to their computer, and ensure their local computer had all the required plugins.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/655/0*v-5YUSyYLBbK6Noy" /><figcaption><em>Comparison between a Grasshopper script and a Rhino Plugin interface using Rhino Compute</em></figcaption></figure><p>With Rhino Compute, this could now be solved by the specialist writing the workflow and adding it to an office web service available to anyone in the office. Hardware solutions could include bolstering the office on premise or leveraging a cloud server with additional memory or processing power without the need to purchase high-end computers for each individual designer — effectively offloading both the administrative management and hardware requirements to run geometry workloads.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/787/0*hOVIclj_0tD99PQb" /><figcaption><em>Types of users profiles for the workflow diagram</em></figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*asgdfG3yj0xPOrxb" /><figcaption><em>Rhino Compute workflows</em></figcaption></figure><p>The introduction of cloud services enables an entire office to solve calculations and use web services to both act as an authentication layer to ensure privacy and perform any additional data transformations, perhaps convert to a WebGL format or push to a machine learning data pipeline. The numerous combinations of workflows enable immense flexibility based on available skill sets and cloud platforms.</p><p>We will briefly discuss each workflow and provide links to get started.</p><p><strong>A + B. Web Server→ Rhino Compute</strong></p><p>The specialist team can write geometry-based functionality in <a href="https://github.com/mcneel/rhino3dm#rhino3dmpy-python">Rhino3dm.py</a> and solve calculations using <a href="https://developer.rhino3d.com/guides/#rhinopython">Compute-rhino.py</a>. <a href="https://flask.palletsprojects.com/en/2.0.x/">Flask.py</a> web server enables functionality to be available via a web API endpoint, creating centralization of logic and scripts on the web server, while Rhino Compute can be hosted either locally on a user’s computer through a localhost, or on a cloud-based Windows Virtual Machine. An example of the local workflow can be found on the <a href="https://developer.rhino3d.com/guides/grasshopper/hops-component/#calling-a-cpython-server">McNeel Hops developer page</a>. Instead of using a <a href="https://flask.palletsprojects.com/en/2.0.x/">Flask.py</a> web server, <a href="https://nodejs.org/en/about/">Node.js</a> can be used as a web server with <a href="https://github.com/mcneel/rhino3dm/blob/main/docs/javascript/RHINO3DM.JS.md">Rhino3dm.js</a> and <a href="https://github.com/mcneel/compute.rhino3d.appserver">Compute-rhino.js</a>. Integrating WebGL libraries like <a href="https://threejs.org/">Three.js</a> enables visualization of geometry and solutions from Rhino Compute, and the Node.js server acts as a communication layer between the client web browser and Rhino Compute.</p><p><strong>C. Hops → Rhino Compute</strong></p><p>Hops enables a specialist to call remote Grasshopper scripts and functionality into the Grasshopper environment, allowing access to <a href="https://github.com/python/cpython">CPython </a>and other Python 3.x+ libraries, such as <a href="https://scikit-learn.org/stable/">Scikit-learn</a> for machine learning and <a href="https://networkx.org/">Network-x</a> for graph-based solving. Rhino Compute can be accessed either locally or remotely through Hops based on the URL parameter input for the Hops component.</p><p><strong>D. Jupyter Notebook → Rhino Compute</strong></p><p>The data science field is growing at an exponential rate as datasets on the urban context continue to emerge. <a href="https://jupyter.org/">Jupyter notebooks</a> are a popular medium to share and discuss data science workflows, from extracting walking distances to extracting information on lot level in New York City. Spatial data can now be moved between a Jupyter Notebook (Python 3.x+) and Rhino interface using Rhino3dm.py. The workflow invites data scientists to interact with Rhino as any other Python library.</p><p><strong>E + F. Rhino Compute as a Service</strong></p><p>Since Rhino Compute can be used as a back-end web service to solve geometry calculations, the complexity can be abstracted from a designer through an intuitive interface either in the web or through a Rhino plugin. This enables a user to input a series of settings then run a GH script in the cloud — removing local hardware limitations and allowing specialist teams to effectively manage scripts and updates. There is no more need to pass along a GH script, when you can deliver the workflow directly to a user through an interface, where the designer can focus on the design decisions rather than on whether they have the correct GH plugins installed or hardware limitations. GH scripts can be combined with machine learning and front-end web visualization to continuously create additional functionality.</p><h3>Case Study #1: Rhino Compute in Practice</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/379/0*sE1UCmC7AMuEE-Kh" /><figcaption><em>Collaboration using Rhino Compute</em></figcaption></figure><p>KPF leveraged Rhino Compute as far back as Fall 2019 on a competition project in Vancouver. The design team needed a method to quickly iterate through design options and present to both internal and external teams. Each team, however, placed value on a different metric. This meant the design team needed to test design options against a variety of analyses — views, thermal comfort, shadows and more. The <a href="https://ui.kpf.com/">KPFui</a> team created a Rhino plugin connecting the Windows cloud server running Rhino Compute and front-end web interface <a href="https://scout.build/">Scout</a>. The workflow facilitated faster insight for design teams and augmented the capability of a KPF computational designer to work on more projects across the global firm.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/842/0*9WJ98A73jc5SMzNd" /><figcaption>Computational efficiency</figcaption></figure><h3>Centralization</h3><p>KPFui has many Grasshopper scripts, which in a typical workflow would be sent to a designer to execute on their local computer. This would require the designer to understand Grasshopper and navigate a Grasshopper canvas for settings and outputs. This workflow proved to be too difficult for adoption of tools throughout the office. In the end few teams would actually use the tools provided to them. Instead, the Rhino Compute plugin enabled the Grasshopper script to run on the cloud server with settings exposed to the designer through the plugin interface. This lowered the designer’s barrier to entry to execute complex Grasshopper scripts, creating an iterative workflow when changing design options and testing them against various metrics.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*2v2Riif_iWu9V3r2" /><figcaption>Bentall Centre’s Computational Design model in KPF’s Scout web platform</figcaption></figure><p>In 2019, KPF leveraged Rhino Compute to support Hudson Pacific Properties’ repositioning of Bentall Centre in downtown Vancouver. The Bentall Centre project introduced the opportunity to rethink one of Vancouver’s most central urban precincts as the city’s most exciting destination, and as a place that is more than the sum of its individual parts. The site had unused area that was to be placed in relation to four existing towers, built in the late 1960s and early 1970s. It was important that this area was located thoughtfully — that it would not negatively impact sun and view lines from the existing Bentall towers and urbanistically would not take up existing public plazas. The site contained a complex set of constraints, such as transit lines, sloped grading, and protected site line requirements from the city. The design team needed to weigh various tower options for unobstructed views from each tower configuration and direct sunlight to the ground plane in order to best locate a sun-filled plaza. KPF leveraged the Rhino Plugin using Rhino Compute, which enabled the design team to quickly iterate through various design options without the need for a Grasshopper expert. The Grasshopper scripts were stored on the cloud server and solved on the Rhino Compute Windows server, enabling the architects to glean insights rapidly during the design process, thus increasing their ability to make informed design decisions and subsequently present them to the client for approval.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/716/0*ValBXpC0AhvO3jnx" /><figcaption>Unobstructed View Bentall Centre analysis to best locate additional FAR on site</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/711/0*C2VhzCJA_lUrU8rj" /><figcaption>direct sunlight analysis on the site</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*JxcXqv69VKm3-v0R" /><figcaption>Early draft of KPFui’s Rhino plugin leveraging Rhino Compute in 2019</figcaption></figure><h3><strong>Case Study #2: Rhino Compute in Academia</strong></h3><p>IAAC introduced the use of Rhino.Compute to the students of the <a href="http://iaac.net/macad">Master in Advanced Computation for Architecture &amp; Design</a> within the <em>Digital Tools for Cloud-based Data Managemen</em>t seminar of the <em>BIM and Smart Construction </em>module in early 2021. During the course, students learned how to leverage their Grasshopper skills by turning their definitions to functions accessible on the browser with the use of Rhino.Compute. Later on, within the <em>Geometrical Optimization </em>course of the <em>Artificial Intelligence in Architecture </em>module, students were instructed in the use of Hops as a means to interoperate between their machine learning models and Grasshopper. This allowed them to exhilarate their workflows by being able to transform geometry to data and vice-versa on their local machines, eventually allowing for user input in the browser through the use of Compute.</p><p><a href="https://www.iaacblog.com/programs/design-space-exploration-variational-autoencoders/"><em>Design Space Exploration with Variational Autoencoders </em></a>is a thesis project developed within the context of the MaCAD program. The project focuses on the application of machine learning to create a diverse design catalogue accounting for structural performance and the creation of a web-based user interface that enables users to explore these options in early stages of the design process.</p><p>The main software environment used is Rhinoceros and Grasshopper, with the Karamba plug-in for structural analysis and a Grasshopper Python script to automate the dataset generation. A Machine Learning model was then trained in Google Collabolatory Notebook using Tensorflow and Keras libraries. The Hops plug-in for Grasshopper was used to link Machine learning libraries in Python and Grasshopper for geometry recreation. Finally, the whole workflow was brought online with Node.js and Rhino Compute through an <em>appserver</em> which enabled hosting the Grasshopper script in a web application.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*_zVQWmaC-zKKKJkW" /><figcaption>Computational workflow</figcaption></figure><p>The workflow starts by creating a parametric model for a desired geometry which is then analyzed for a specific performance, for instance, structural or environmental. A dataset of samples is created by randomly generating parameters and running analysis for each geometry created by those parameters. Using this dataset, a Variational Autoencoder (VAE) model was then trained. The decoder recreates geometries from the latent space and a performance map was created which enabled exploring the samples generated. By bringing the inputs and results to a web interface, the outcome is a user-friendly platform for interactive design space exploration.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*WVe4FQzuUNeYhe4VpEkPyQ.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/1*J86iuU78K1gA6x_4v1OqNw.gif" /><figcaption>Dataset generation</figcaption></figure><p>As a case study, a canopy truss structure composed of a column and a roof was investigated. The column was created with variable radii at different heights and the roof with variable control points and lateral shift. A dataset was then created by randomly generating parameters that describe the geometry and running structural simulation for each sample. The calculations were performed with the <em>Karamba </em>plug-in Grasshopper. Each sample had the same number of elements, and for each, the cross section optimizer found the optimal cross section for each member. The results, such as overall mass of the structure, maximum displacement, and elastic energy, were saved in the database together with parameters describing the geometry of each of the samples.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*gtAXR-YJXqtecZd2" /><figcaption><em>Exploration of the Latent Space</em></figcaption></figure><p>The collected data was then fed into a conditional Variational Autoencoder, which produces a trained machine learning model of probability distribution over a latent space, steering the generation of samples into a specific area of interest — in this particular case study, structural performance. After the training process, geometries are reconstructed by linking the Python model to Grasshopper using Hops. The Python file containing the model can be run inside Grasshopper through Hops and operated through slider inputs which enable interactive design space exploration within this environment. The output of the Hops component are the parameters generated by the decoder which are used to recreate geometries using the same grasshopper definition that generated the dataset originally.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/963/0*Ev7RC6Y5XAaql05U" /><figcaption><em>Web interface for Design Exploration</em></figcaption></figure><p>In order to extend this tool outside of the scope of Rhino and into the browser, the whole workflow is leveraged as a web app that can be used to interact with the samples in the design space in the browser in a user-friendly interface.</p><h3>Outlook</h3><p>As showcased in the case studies, the introduction of these technologies help bridge design and computation, which allow for compelling possibilities.</p><p>To begin with, Rhino Compute enables design projects to have accessible interfaces that empower users to easily navigate the design space, explore parametric variations and reveal layers of relevant data that otherwise would be challenging, if not impossible, to display in traditional formats. Not only does this render a more comprehensive visualization of the many dimensions that comprise a design project, but it encourages a more participatory design process. Furthermore, it releases this experience from any particular software knowledge by bringing it to what arguably is nowadays the most ubiquitous platform — the web browser.</p><p>In parallel, they could also enable the interaction of commonly used software for design (CAD) with other platforms, which can be of great interest for informing the design process. Either by lowering the entry barrier for designers to make use of advanced computational workflows such as machine learning, or by providing them with readymade components created by expert developers, these technologies help expand the vision of the designer and aid in making better informed decisions throughout the design process.</p><p>If we consider that the inclusion of these technologies in the design practice can only enhance the view of the designer, and that therefore such an expansion of the architectural trade has a positive impact in the resulting projects, it would seem that architectural practice can greatly benefit from investing in potential employees with these skill sets. The same can be said for academics, by imparting such technologies to endow future professionals with the skills that are currently required to tackle these computational challenges.</p><p>However, to implement such technologies comes with a cost that often design practices think twice about investing in: long established design habits must be rethought in order to fit these new workflows and it can take considerable development time for these tools to become widespread and useful. As any other emerging technology, depending on the scale of the project, or the number of projects that can benefit from these tools, the benefit of introducing emergent technologies may not be appealing enough to stop the orchestrated process of design logistics that architecture requires.</p><p>Whether that is the case or not, what is true is that computational workflows are evolving rapidly and increasingly being introduced into architectural practice, which invites us to turn our attention back to the fundamental question: What is the role of the designer?</p><p>How can the scope of the architect continue to evolve, while allowing other specialties to develop tools that aid the design process? Or should design be pushed to break free from the tools that currently frame it, by endowing designers with the skills required for the creation of its own toolset, in order to allow for more innovative processes?</p><p>What is certain is that we are currently in the midst of an enticing conversation, one in which creativity and technology entwine in order to expand our vision of how we conceive the cities and communities of the future.</p><h3>Acknowledgements</h3><ul><li><em>Design Space Exploration with Variational Autoencoders</em> is a thesis project by Aleksandra Jastrzebska developed within the context of the Master for Advanced Computation for Architecture and Design and under the supervision of David Andres Leon.</li><li>Steve Baer, Luis Fraguada, Will Pearson and the McNeel team for their support in the development of these case studies and revision of this article</li><li><a href="https://www.kpf.com/about/leadership/directors/darina-zlateva">Darina Zlateva</a> — Director at KPF for all the support in beta testing and giving valuable feedback.</li><li><a href="https://www.kpf.com/">Kohn Pedersen Fox Associates</a> for the continued support to research new technologies and workflows to improve the design practice.</li></ul><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=aceb3f23fd95" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Sidewalk Crowding in the Age of COVID-19]]></title>
            <link>https://medium.com/@kpfui/sidewalk-crowding-in-the-age-of-covid-19-3ca9d20039a7?source=rss-e33d27980495------2</link>
            <guid isPermaLink="false">https://medium.com/p/3ca9d20039a7</guid>
            <category><![CDATA[urban-planning]]></category>
            <category><![CDATA[spatial-analysis]]></category>
            <category><![CDATA[covid19]]></category>
            <category><![CDATA[data-visualization]]></category>
            <category><![CDATA[new-york-city]]></category>
            <dc:creator><![CDATA[KPF Urban Interface]]></dc:creator>
            <pubDate>Mon, 31 Aug 2020 20:19:11 GMT</pubDate>
            <atom:updated>2020-09-02T12:25:20.571Z</atom:updated>
            <content:encoded><![CDATA[<h4><em>Using Data to Inform the Reopening of New York City</em></h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Xixg3yWUgr3zWFZJU5_lUQ.png" /></figure><p>The novel coronavirus has fundamentally changed how we inhabit urban space. Sidewalks have become hosts to a variety of new activities: seating for restaurants and cafes, spaces for queuing to enter buildings, and retail pick-ups/drop-offs, among others. In addition to continuing to provide space for pedestrian circulation and other pre-COVID19 uses, municipal governments must now make room for these curbside newcomers and do so in accordance with social-distancing and other public safety protocols.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/600/1*NVVHGOlQZBiR4ldbVkLRSA.jpeg" /><figcaption>NYC outdoor seating regulations for restaurants. [Source: New York Times]</figcaption></figure><p>In a city like New York, which is already known for dense sidewalks, safely adapting to new sidewalk conditions raises pressing questions. Where is there not enough sidewalk space to accommodate these new activities? In a city as large as NYC, how can the city and community groups logically and comprehensively plan and design for a safe reopening through the pedestrianization of streets, sidewalk expansions, and other novel adaptations of public space? Enter data!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*IQCT9ajd4H0boKGZDtcdSg.gif" /><figcaption>Sidewalk Density Web Tool [Source: KPFui]</figcaption></figure><p>We created a set of web tools that helps everyone understand sidewalk population density in order to aid NYC decision makers in the immense challenge of reopening the city in an effective, safe, and equitable way. These tools leverage detailed datasets and simulate new, COVID-era sidewalk uses. They aim to empower the city and community groups to make decisions to reduce the risk of sidewalk crowding in New York City more rapidly and with a higher degree of confidence.</p><p>The suite is comprised of two primary tools — a City-Wide tool that highlights areas where high levels of sidewalk crowding are likely and a Neighborhood tool that uses routing analysis to locate specific street segments that are within high instances of pedestrian, movement, queuing, and business activity.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*qUpHm4GOLbr8meNQg1ut6g.png" /><figcaption>In the city-wide analysis, constituent datasets are weighted on the fly and can be done so to model different real-world scenarios. For instance, we can simulate Phase 2 (displayed here) reopening by reducing the weights for pre-COVID job population, office buildings, etc. and increasing the weights for residential population and restaurants. [Source: KPFui]</figcaption></figure><h4><strong>City-Wide Tool</strong></h4><p>The City-Wide Tool (<a href="https://sidewalk.kpfui.dev/">https://sidewalk.kpfui.dev/</a>) determines which areas of New York City are most likely to have high levels of sidewalk crowding. The tool aggregates a variety of constituent datasets — ranging from resident and job population figures, business locations, and building information — into a composite score of square feet of sidewalk per pedestrian.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*j1aRy2wR6RyR5Pehzjs-7g.png" /><figcaption>The different constituent datasets that were incorporated into the analysis include resident and job population figures, business locations, and building use and size, among others. [Source: KPFui]</figcaption></figure><p>Given the fluid and complex nature of cities, the varying priorities of different NYC agencies and community groups, as well as the unpredictable nature of the novel coronavirus, creating a model that could be adjusted and calibrated in real time was essential. The utility of the dynamic web map we created is clear when simulating different scenarios — weekday and weekend peak hours, for instance.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*IsYdge1OWxvVNoq0I1u7sQ.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*1s7Pyc8jH--5LPUvxUl9Qw.png" /><figcaption>Maps of weekday (left) and weekend (right) peak-hour sidewalk stress make clear the difference in where people are out on NYC sidewalks (and for what reasons). Notice the concentration of sidewalk stress in Midtown Manhattan and other business districts during weekday peak-hours and the more dispersed distribution of sidewalk stress throughout NYC with significantly more people out in the outer boroughs (closer to their homes). Our web-based dynamic model allows users to adjust weights of the different component data to quickly and intuitively simulate various conditions. [Source: KPFui]</figcaption></figure><p>Sidewalks become crowded for many different reasons — hordes of shoppers, people heading home, or a lack of sidewalk space, for example. Clustering the most stressed grid cells allowed us to better understand the factors that were most important in sidewalk crowding in our analysis. The clustering algorithm groups areas of the city together based on similar reasons for crowding: Midtown and FiDi score high in offices and working population, the Upper East Side and Flatbush experience crowding because of large residential populations, etc.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*3EbAIhx0eF1ezVNPWD6M5Q.png" /><figcaption>When clustering based on restaurant and retail density, four typologies appear. Each of these typologies is characterized by distinct restaurant and retail conditions, which exist across the city. [Source: KPFui]</figcaption></figure><h4>Neighborhood Tool</h4><p>From the city-wide tool, we identify neighborhoods with high sidewalk crowding and zoom in for detailed study of pedestrian density along sidewalks at different times of day relative to the effective width of each sidewalk (<a href="https://neighborhoods.kpfui.dev/">https://neighborhoods.kpfui.dev/</a>). It does so by simulating pedestrian traffic to local destinations — transit, businesses, etc. — and include both moving and standing pedestrians. Some wide sidewalks might still not have enough space given business and pedestrian flow, whereas some narrow sidewalks might be so lightly trafficked that they do not experience much stress. This tool is a resource for identifying hotspots within the neighborhood where large number of pedestrians congregate and the demand for more sidewalk space is high.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*gQ4rQU5tOPmKyL8P8cqWMQ.png" /><figcaption>The simulation shows the approximate amount of pedestrian queuing and walking to these businesses. This analysis was generated using urbano.io, a mobility modeling tool developed by the Environmental Systems Lab at Cornell University. It is seriously cool — check it out. [Source: KPFui]</figcaption></figure><h4>Notes from under the hood..</h4><p><em>(An assortment of interesting tidbits from the analysis)</em></p><p><strong>Incorporating Sidewalk Area </strong>Given that many regulations aimed at mitigating the spread of COVID-19 are explicitly tied to spatial thresholds/indicators, knowing the amount of sidewalk space per pedestrian throughout the city is a powerful resource for decision makers and is thus what this suite of tools seeks to quantify. Given the variation in sidewalk space throughout New York City, plotting the constituent and composite data relative to sidewalk area produced unexpected and informative findings.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Az9wOLJ6ldb4gSrFo117bA.png" /><figcaption>The map of residential population (left) looks as we would expect, with large concentrations of population in Manhattan and select locations in the outer boroughs. However, when we account for sidewalk area, many areas with strictly high population become less relevant, and acute hotspots throughout the city become visible. [Source: KPFui]</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*3F7lLWu9piUKgOQqkU0juA.png" /><figcaption>Establishing ratios and rules of thumb for these calculations proved to be a challenging aspect of the project. However, having a resultant figure that could be easily understood in a spatial sense (square feet of sidewalk per pedestrian) allowed us to adjust these calculations to compensate for values that were much too small (pre-COVID results of 2 square feet of sidewalk per person, for instance). [Source: KPFui]</figcaption></figure><p><strong>Sidewalk Stress in the Outer Boroughs </strong>Outside of obvious hot spots in mid and lower Manhattan, the analysis in a Phase 2 simulation reveals distinct hotspots in the outer boroughs. These include Co-op City and parts of Tremont in the Bronx, Jackson Heights and Flushing in Queens, and Flatbush and Sunset Park in Brooklyn.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*6gZlEN0foMkAghSBdym3jA.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*DU0_fQ_0TJD4VbIJs0Nc6A.png" /><figcaption>Co-op city (web map on the left and satellite imagery on the right) in the Bronx emerges as a distinct hot spot in the simulated Phase 2 analysis. [Source: KPFui and Google Earth]</figcaption></figure><p><strong>Office Workers on the Sidewalk</strong> The clustering mode can be used to quickly identify neighborhoods that share similar levels of sidewalk stress. If the tool for 100% occupancy for workers in retail, restaurants, and office businesses and 50% occupancy for residents and the results are clustered — Midtown Manhattan shares similar characteristics with Flushing, Queens. Both areas are simulated to have a high levels of worker-generated sidewalk stress. Perhaps solutions implemented in Midtown could also improve conditions in Flushing.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*R0pvkQ5K0z8ZLMk66uT48Q.png" /><figcaption>Areas in yellow also have the lowest number of residents and highest number of office workers affecting sidewalk stress. [Source: KPFui]</figcaption></figure><p><strong>Neighborhood Pedestrian Pattern </strong>The neighborhood analysis simulates the travel demands of three busiest times of day — morning rush hour, lunch time, and evening rush hour. Taking the aggregated pedestrian travel demand from Google Places data, the analysis model simulates pedestrians commuting or visiting businesses per sidewalk. The tool further breaks down the category of pedestrians on each street to understand what activities contributes to the most sidewalk stress. In Jackson heights during lunch time, we see the high pedestrian traffic to restaurants along commercial streets, attributing to its sidewalk stress. This information helps inform decisions about phased reopening for small businesses and strategies for locating outdoor dining space.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*QGsXAXOjhetp41-iPasb3g.jpeg" /><figcaption>Neighborhood Analysis of Jackson Heights, Queens showing street pedestrian traffic at 7:00PM on a typical Saturday night. [Source: KPFui}</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*K9UZFnd9CuYmZlV4BHrzcQ.jpeg" /><figcaption>Neighborhood Analysis showing sidewalk area per person in Jackson Heights, Queens, identifying streets that experience higher pedestrian traffic in relation to the sidewalk area available. Highlighted sidewalks indicate instances where less than 30 sqft of sidewalk is available per person — the minimum distance recommended for social distancing by the CDC. [Source: KPFui]</figcaption></figure><p><strong>Sidewalk Width determines Stress </strong>Street activity alone does not indicate sidewalk stress. A neighborhood with large sidewalks could support commercial activities while providing enough distance for pedestrians to social distance safely. Similarly, while residential sidewalks tend to have lower pedestrian traffic, neighborhoods with narrow sidewalks would still exhibit sidewalk stress as the area per pedestrian is low.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Nk99eEKPszBQG5wti9Mf9w.jpeg" /><figcaption>Pedestrian traffic along Fulton Street in BedStuy is simulated to be far greater than Macon Street where an Open Street has currently be established. [Source: KPFui]</figcaption></figure><p><strong>Leveraging Open Streets </strong>As businesses adapt to new COVID-19 operation strategies, businesses have transformed the sidewalk as extensions of their stores through measures such as designating queuing lines along store perimeters or setting up outdoor seating. This information provides insight to better inform urban interventions led by the community or by city agencies to expand pedestrian zones on commercial streets, or to create better synergy with the current Open Streets initiative.</p><p><em>Special thanks to Assistant Professor Timur Dogan and PhD student Yang Yang from the Environmental Systems Lab at Cornell University for their mobility modeling tool urbano.io and their help customizing it for generating the analysis data for the the neighborhoods web map.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=3ca9d20039a7" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Regulating Density (Or Not)]]></title>
            <link>https://medium.com/@kpfui/regulating-density-or-not-adc466195377?source=rss-e33d27980495------2</link>
            <guid isPermaLink="false">https://medium.com/p/adc466195377</guid>
            <dc:creator><![CDATA[KPF Urban Interface]]></dc:creator>
            <pubDate>Thu, 16 Jul 2020 15:13:57 GMT</pubDate>
            <atom:updated>2020-07-16T15:13:57.407Z</atom:updated>
            <content:encoded><![CDATA[<h4>A look into Floor Area Ratio</h4><p>Floor Area Ratio, or Plot Ratio, is the ratio of the total building floor area to the site or lot area. It is a common measurement used to quantify the density of developments. Some cities, such as New York City, use Floor Area Ratio (FAR) to regulate the density of what can be built, while others like London use it to understand existing building density as part of the planning process. This shift in the use of FAR from regulatory to representational is often accompanied by a shift in how it is measured, resulting in different representations in density.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*WsCRtJBI3GdtVZ6VE0mk0A.jpeg" /><figcaption>FAR = Gross Floor Area / Lot Area</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*bWWhA1d2S1gWR8D8agJbMQ.jpeg" /><figcaption>FAR = Gross Floor Area / Street Center Line Area</figcaption></figure><p>In New York City, FAR is calculated relative to the lot area to regulate how much can be built. In London, FAR (referred to as Plot Ratio) is often calculated relative to the centerline of the streets bounding the lot when density is considered during the planning process. The method used in London gives a better representation of the built density of a neighborhood, as it includes streets and public spaces. However, it is comparatively more difficult to use as a regulatory mechanism, since it encompasses both private and public land. Below are several New York City neighborhoods measured using both FAR methodologies to illustrate the different representations of density.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*k1A40esZsomZb5ho_G-Uew.jpeg" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*sM5xrHHCc_6-dBnyxnMrWQ.jpeg" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*hyaT8zP_xbL09mB4HMc_Jg.jpeg" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*yENov_VxHAWRw_S5NQan8g.jpeg" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*S4sTHVcw6G9HPu7JCGjcsA.jpeg" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*0H8wNSpEqFVX0APGpfkaSA.jpeg" /></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=adc466195377" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[History of Grand Central Terminal Development]]></title>
            <link>https://medium.com/@kpfui/history-of-grand-central-terminal-development-b609f6dbb0c5?source=rss-e33d27980495------2</link>
            <guid isPermaLink="false">https://medium.com/p/b609f6dbb0c5</guid>
            <dc:creator><![CDATA[KPF Urban Interface]]></dc:creator>
            <pubDate>Tue, 14 Jul 2020 21:17:20 GMT</pubDate>
            <atom:updated>2020-07-14T21:17:20.638Z</atom:updated>
            <content:encoded><![CDATA[<p>From a combination of historic maps, photographs, and data sets, KPFui modeled the development of Grand Central Terminal and the surrounding neighborhood starting from the Grand Central Depot in the 1870s to today.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*4yH_r0kwTkJKy2VxU1BwVA.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*54nzvLooyBQKXCuVFr-LMQ.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*oBI34AEo2Wyfpkxbo8FeVQ.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*WBZ--eycdcHwgzNHBb5zGA.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*AzoxkcGZOHXJoziGt-f1Kw.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*GK0v-4TsVGtlKe5WbIC9fg.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*VSJOKaRCS_4s-p8zToV85w.png" /></figure><p>Sources: <a href="http://www1.nyc.gov/site/planning/data-maps/open-data/dwn-pluto-mappluto.page">PLUTO data</a>. <a href="http://www.nypl.org/node/165254">Historic Maps New York Public Library</a>. <a href="https://memory.loc.gov/ammem/index.html">American Memory for the Library of Congress</a>. <a href="http://www.amazon.com/Grand-Central-Terminal-City-Within/dp/0960689222">“Grand Central Terminal : City Within the City,” Deborah Nevins, Municipal Arts Society</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=b609f6dbb0c5" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>