Weather and urban growth modelling

Written September 2nd, 2014 by

Gavin Schmidt presenting at TEDI came across this TED talk by Gavin Schmidt. He is a NASA climate expert and his presentation was about weather models and climate models. In his talk, Gavin explains why we can be confident about predicting how the climate will react into the long-term future.

He discusses the enormous complexities of modelling climate patterns that span a range of 14 orders of magnitude across both time and space. I though modelling urban growth was complex! We are only dealing with 2 orders of magnitude. Even at that level, there is a large body of work required to build and test an urban growth model. I certainly appreciate the mammoth effort needed to build climate models at those scales of magnitude and complexity.

Always wrong

What caught my attention in Gavin’s talk were statements that he made about weather and climate modelling and how they are just as relevant to the modelling of other systems and patterns, such as urban growth.

“Models are not right or wrong. They are always wrong. They are always approximations.”

I think sometimes we can lose sight of this as we strive to perfect a model. Yes, we have to spend time making sure the inputs to the models are accurate as possible. We don’t want models suffering from GIGO… Garbage In, Garbage Out.

However, we also need to be mindful we are not spending an inordinate amount of time casing perfection. As I pointed out in a earlier post on how to establish a baseline, only 11%-13% of properties have development capacity in an urban growth model. Spending hours trying to make sure you have the correct land use for that 50 square metres of retail space that cannot be developed, in the big scheme of things, probably is going to have very little impact on the model. Spend your time where it matters most and remember models are always “approximations”.

Having skill

Gavin also spoke about models having “skill”.

A model result is skilful if it gives better predictions than a simpler alternative.

I have never thought about models having this characteristic. It is an attribute that I associate with humans. Reflecting on what Gavin is saying.

Does a model tell you more information than what you would have had otherwise? If it does, it is skilful.

I find myself agreeing with him. It is “skilful” to be able to take the existing state of a system, apply highly complex algorithms to produce very reasonable predictions. I now get a certain satisfaction knowing that I am building something that is skilful… a feature that is almost human.

If you build models of systems, then watch Gavin’s talk and take pride that you are building something that is skilful. If you don’t build models, watch it anyway. It certainly gives a compelling insight into what the models are predicting for the climate and our future on this fragile planet.


About the Author:
Bradley is the founder and Managing Director of Sizztech. He is an enterprise ICT consultant with over twenty years of experience in the Information Technology industry. In the last eight years Bradley has specialised in delivering web based geo-spatial solutions that focus on supporting capital works planning and delivery. This has lead to Sizztech developing Forecaz Modeller, an automated tool that performs urban growth modelling and forecasting.

Leave a reply