Open California data & sensor comparison

planet earth observation open california

I wrote recently about clipping data and shared an example in Beverly Hills. I wanted to use this data to explore and compare all the data sets available on Planet Explorer, but focusing on the terrific Open California data. I’ve been meaning for sometime to write a comparison between different sensors. My feeling is that this can get a bit overwhelming for some non-specialists; there is such a large choice of data today.

Open California

I’ve known for a while that Planet have made freely available all the data they have been acquiring in California. There are exceptions though; the main two are:

 

Are there any data limitations?

We currently limit the amount of data to 2GB per day for each user, mostly for bandwidth considerations. If you have a compelling use case that needs bulk access, please get in touch.

How recent is the Open California data?

You will have access to California imagery collected since October 2013, along with newer data we collect at a two-week delay. Same-day imagery will not be available.

Otherwise you are welcome to explore and use. More details on the programme are made available here. Get an account with Planet and you are good to go.

Access

Let’s take a look at the data in Beverly Hills as an example (I accessed this data on 24th/25th June 2017),

using my simplified polygon from OpenStreetMap.

planet earth observation open california

It is a pretty intuitive website – you can filter on cloud, area of your AOI and the source of the data (default is PlanetScope data and RapidEye data, but you can also search for Landsat8 and Sentinel2a data – nice one).

A comparison of four sensors

Access to this data gives a good opportunity to compare all four Satellite data sources (from a spatial, spectral and temporal resolution perspective). Let’s take a look at these four sensors.

Resolution examples

Landsat 8 – 30m RGB, (could be Pan sharpened to 15m with Band 8)

 

landsat8 EO

Sentinel 2a – 10m RGB

Sentinel2 EO

RapidEye – 5m RGB

RapidEye EO

PlanetScope – 3m RGB

Planet EO

A closer look

If I zoom into the Franklin Canyon Reservoir the difference in resolution is telling.

 

EO, Landsat 8, Sentinel2, RapidEye, Planet

Moving clockwise from top right, Landsat 8, Sentinel2a, RapidEye and PlanetScope data. The above imagery shows the spatial resolution as comparatively.

Spectral coverage

  • Landsat 8 has 11 bands. Covering visible light, near infrared, shortwave infrared and thermal.
  • Sentinel2a has 13 bands. Again covering visible, near infrared and shortwave infrared.

Spectral

source: http://landsat.gsfc.nasa.gov/wp-content/uploads/2015/06/Landsat.v.Sentinel-2.png

  • RapidEye has 5 spectral bands, Blue, Green, Red, Red Edge and Near Infrared.
  • Planet has 4 spectral bands, Blue, Green, Red and Near Infrared.

A comparison of the RapidEye and PlanetScope is here.

Temporal Resolution

  • Landsat 8 has a 16 day revisit time
  • Sentinel 2a has a 10 day revisit time (which will drop to 5 days once Sentinel 2b is in full service)
  • RapidEye has a daily revisit (off-nadir) or 5.5 days (at nadir)
  • Planetscope has a daily revisit time

It would be easy at this point to just say better spatial resolution and temporal resolution = best data, but a great deal depends on what you are ultimately trying to achieve / map with the data. With this in mind let’s take a look at two examples.

1 Edge Detection

Edge detection is usual for mapping features on the ground. By manipulating the canny edge algorithm it’s possible to optimise edge detection for whatever the target is. In this case edge detection is good for road-, building- and object-identification.

Edge detection planet rapideye

From top right running clockwise. Landsat 8, Sentinel2a, RapidEye and PlanetScope. All look like make your own camouflage!

2 Unsupervised Classification

Classification of remote sensing is a massive topic. There are numerous methods and approaches. Ultimately with a classification we are trying to break down the image into categories of land use or land cover, or perhaps build a terrain model.

 

unsupervised classification

As before from top right running clockwise: Landsat 8, Sentinel2a, RapidEye and PlanetScope. None of these models are perfect; they are for illustrative purposes only.

Simplified summary table

Hopefully this has been a relatively gentle introduction to four different sensors and a quick look at the open California data set.

 

I have grouped all my previous blogs (technical stuff / tutorials / opinions / ideas) are here http://gis.acgeospatial.co.uk

I am @map_andrew on twitter