Keeping up with urban change

@dereklieu • Development Seed

Rapid urban change means your maps are probably wrong.

1. Where?

2. By how much?

We built a tool for these questions. It's called Urchn, which is short for URban CHaNge.

We partnered with Radiant, formerly Digital Globe, and Azavea. (Radiant came up with the name)

Urchn starts with persistent change.

We do pixel analysis on time series Sentinel and Landsat imagery.

It works by examining three consecutive, cloud-free images.

If a change persists in all three, we look back at three prior images for confirmation.

This tells us where a change occurred, and when.

2017 02 09

2017 02 15

2017 02 20

2017 03 11

2017 03 16

2017 07 19

We also aggregate, by tile and admin boundary.

We're interested in a few use cases.

1. Identify where are official maps are out of date.

2. Identify where OSM is out of date.

github.com/azavea/osmesa

<3 Azavea

Here are a few more ideas we're experimenting with.

1. Predicted building footprint area using machine learning.

2. Urban land-use classification

3. Whatever else you've got?

Release goals:

• Open-source mapping platform code
• Open-source aggregation code
• Browseable Osmesa data globally
• Change data via some money-making scheme, possibly subscription
• Public beta release by SOTM Detroit, Oct 5 2018

Challenges:

• Improve recognition of re-development
• Improve aggregation speed to keep up with Osmesa

If you could design a tool to monitor urban change, what would it tell you?

Let's end on a game.

Identify everything that's changed between one image...

...and the next.

This is Sursum Corda.

Sursum Corda is a housing cooperative for low-income residents.

This is the future of Sursum Corda.

This is what it looks like now.

To me, understanding urban change means making local knowledge widely known.

We've love to explore what that means for you.

Thanks!

derek@developmentseed.org