# Making a Model, Doing A Design, Estimating Unknowns – How to Urban Plan

Draft article for JAPA

What is urban planning?  What do urban plans that don’t work all have in common?

This is not a philosophical article or an introduction to a 300 page book on planning theory, this is about what  urban planning is and should be at its rawest level.  This is the reason why this peice, quite deliberately, has no citations

Urban planning and urban design being the same thing, just different angles, different ways in to the same set of problems.

Plans that don’t work have made a mistake in their underlying model of place.  Planning by its nature being based on a necessarily a simplified model of place and a forecast of that model into the future.

These models are always wrong, but somethings fall apart catastrophically:  models that project forward employment levels but don’t square with the employment zoned, models that project forward household housing needs but somehow assume they will be magically met without housebuilding, the examples are endless.  All plans are an assumed series of equations and mathematical identities about activities in space:

An example Yield (for a zoning district) = Area_HA * ER*Density_DPH

In a spreadsheet (good) or a gis (far better).  Note: I would always recommend the NASA rule for not crashing into planets- do all dimensions and calculations in metric, specify the units in your formula and then convert to the locally idiosyncratic measuring system, like rods or furlongs etc. only at the end.  ER here being exaction ratio, or gross to net ration as it is called in some countries.  Density as dwelling per hectare.

A masterplan is the same.  The term ‘balancing the land budget’ really means doing one which is not mathematically impossible and can be built.  masterplans are driven by a series of mathematical identities, relating to population, units, blocks, streets, neighbourhoods etc.  If you have got these wrong and haven’t, for example, provided enough space to fit in the schools your density requires in terms of pupil product then your land budget doesn’t balance and you haven’t found a solution to the implicit solution set the zoning regulations and standards for the area imply.  It is also true for traffic models of a planning intervention, showing LOS RED for example., and viability models, where your discounted cash flow and peak debt models show your scheme wont make  a profit and no-one would fund it.

So producing a plan where the underlying model is explicit, transparent, tractable, understandable and manipulable is key to test what works and doesn’t and whether the reasonable assumptions made are resilient to stress testing is key to success, at least for a period of time.

We never have all the information we need for a perfect model and some spend so much time chasing it they never get down to plan or find their evidence base is now so out of date it has to be redone – rinse repeat.  So we have to estimate.  Formalisation of this is apace – so we talk of City/Construction/Civil Information modelling – for which we don’t yet have an accepted acronym and accepted open source standard as we do for BIM.  But we dont have to be held back by doing nothing till everything is standard, agile simple swift models, and wiring the models we have together,  are the way ahead and plugging these into the oncoming tsunami of models to model the world eventually is the way forward. I hope within five years we will be able to realign a road in cad and have all zoning and planning parameters, and the 3d CIm City model for the model, and traffic model and all utility models update automatically in the cloud presenting a dashboard of metrics to help determine design intent.

Sometimes this estimation is hard – we might be planning in a country where we have no census, or even count of dwelling, GDP information but not at the geographical level we need etc. etc.  So planners have developed an ingenious suite of techniques for ‘zero data planning’ where we make best mathematical guesses of the Rumsfeldian problem set of known unknowns.  These guesses all involve in some form use of a geostatistical model (in GIS)  to guess the unknown in space by what we do know in space. Then with the unknowns plugged into the ‘model’ we have now completed to run this model again in estimation mode – to find the most optimal solution to the model from the (more than atoms in the galaxy series of possible states) using a series of swiss army knife approaches that make the problem tractable.  This should not be scary, you don’t need a PHD in GIS or statistics, you just need to learn and share best practice in this emerging field.

Many other fields are facing similar problems and are trying similar solutions, but are often struggling.  In economics for example it even has names like the ‘quantification problem’ and the ‘replication problem’ (no-one can repeat your model results – if you cant its not a science, crisis – crisis!)

Don’t worry planning and design is not a science its an art.  We seek an imagined vision of future based on our model of the present  switched to forecast mode.  A parametric model where the computer does the hard work and we can experiment with good design and good planning at every scale.  So lets design some models to help solve our design problem.