Pixel size, Schmixel size

“We need 3 meter imagery”, one of our prospects said to us. The higher the better. More detail, more insights. Or not? Should this be the starting point of a discussion or assessment on the usefulness of satellite imagery? Find out what the weather forecast and a digital camera have in common with satellite data analytics. And how the success of using remote sensing should be defined in this story by our Director Arjen Vrielink.

In Remote Sensing there currently is a rat race for the highest pixel size, spatial resolution or temporal frequency. Providers are competing for ever bigger satellite constellations providing imagery at increasingly higher spatial and temporal resolutions.

But not only satellite companies are doing that, also the ‘traditional’ private and scientific Remote Sensing sector companies are guilty of the same thing. Remote Sensing products are usually accompanied by an overly technical fact sheet rambling on about spatial, temporal and radiometric resolution. Usually, the higher the better. More detail, more insights. Or so it is presumed.

The consequence is that not only satellite image providers and value adders are competing on attributes of self, also the market is requesting products with an associated pixel size. “We need 3 meter imagery”, one of our prospects said to us. I wonder if they themselves even realised what that means.

I find it very puzzling and strange that some prospects are concerned with attributes of (the sensors that are) parts of the supposed solution as opposed to the actual problems that they are facing. It made me think of two examples that put the debate about pixel size somewhat in perspective. Imagery and map examples where pixel size doesn’t seem to be an issue (anymore): digital cameras and the weather forecast.

When digital cameras were introduced, somewhere in the late 90’s, early 00’s, the number of Megapixels was a big deal. The higher the Megapixel number, the ‘sharper’ the photos, the more detail could be captured by the sensor. Cameras, shops, sales persons, your uncle at a family gathering, everybody was talking and inquiring about how many Megapixels your shiny new gadget had. More was better. Less was worse, Simple.

Each new generation of cameras introduced ever higher numbers of Megapixels. Reviews would take Megapixel as a major contributor to the final score of a review. Until suddenly, or maybe I stopped paying attention, the emphasis on Megapixels was gone. Nowadays I don’t even know how many Megapixels my phone camera is.

Why is that? Apparently, Megapixels is not relevant anymore, because the pictures I take are fine. They are good enough for the purpose I take them for: showing random snippets of my life to people I meet. People who never inquire about the Megapixels of my camera.

The picture below shows the effect of a not so many (left) and a lot (right) Megapixels.

People see sawing bowling balls


The Weather Forecast
Next, the weather forecast. I have been watching the weather forecast pretty much all my life. And, rationally thinking, the quality of the forecast must have been improved by a lot in those 40 years. But emotionally, I do not perceive it as such. The weather forecast has always been pretty decent. At least in The Netherlands and in my perception.

Yes, we do have an animated radar now showing the location and progress of rain showers, which is very helpful. And we do have a chill factor and UV alarm in addition to the temperature, wind and precipitation forecasts. But in general, the weather forecast has been the same for a long time. And it’s great! I actually never wondered about the spatial resolution of the weather forecast. Why is that?

A typical weather forecast map for The Netherlands looks like this:

Weather forecast Netherlands; source: www.knmi.nl


Considering the size of The Netherlands, I estimate the spatial resolution of above at about 50–100 kilometer. The temperature and wind speeds are indicated in 0 decimal precision and the precipitation, sunnyness or cloudiness are indicated from a limited set of 5 to 10 icons. Wind direction is one of 8 options. And I’m fine with that. Along with 17 million other Dutch people.

Why don’t we care if it is going to be 5.24 degrees Celsius rather than 5.13 degrees? And why don’t we care if there is going to probably be 1.3456 millimeter (mm) of rain in your backyard as opposed to 1.3433 mm in your neighbours backyard?

Because those are not the questions people have when watching the weather forecast. What people are concerned about is:

  • Do I need to wear a jacket tomorrow or can I go out in short sleeves?
  • Do I need to bring rain gear?
  • Do I have to be careful in the morning on the highway ramp because it’s freezing with east wind?

That is why nobody is asking about the resolution of the weather forecast: people want to know if they need to wear a jacket tomorrow. Not if their neighbours garden is 0.002 warmer than theirs.

Again, like for the digital cameras, there are probably professional groups that are concerned with more detailed forecasts. But they have their own specific reasons and problems to solve that require that.

Pixel size, schmixel size!
Should we forget about pixel size, spatial resolution, temporal frequency and radiometric attributes in Remote Sensing then? No, we should definitely not forget about those. But it shouldn’t be the starting point of a discussion or assessment of the usefulness of Remote Sensing based products and services. We should start with the problem of the end user. What is actually the problem that we are trying to solve? How can Remote Sensing help? What would a solution look like?

Once the answers to those questions are clear, you can look into attributes of Remote Sensing related sensors and techniques to see if some of those attributes are blocking a solution and if there is any room for improvement to overcome those blockers.

Take, for example this animation of a breathing Mekong Delta. The first reaction of most people is remarks about the flooding and seasonality, obvious from the animation, not pixel size.

The success of Remote Sensing is not about who has the highest resolution satellites, it’s about listening to the end user and providing solutions to their problems.

Written by Arjen Vrielink