We look at the impact of resolution on raster data.

More often than not, the first question asked when dealing with raster data is, “What’s the resolution?”.  Here, resolution refers to the size of each grid cell, or, if we’re talking about imagery, each pixel.  Since each grid cell is a square, we can further talk about its resolution, or size in the “real world” as, say, 25 m by 25 m, and often further shortened to just 25 m.  The level of resolution gives a sense of how detailed the data are, and in this post we’ll look into that in a bit more detail using elevation as our proxy.

To delve into this, we’ll do a bit of a cosmic zoom and look across a range of scales, starting zoomed out fairly far – here’s a DEM (well, more precisely, a hillshade from a DEM) with a resolution of 100 m at a scale of 1:500,000:

Some familiar territory and enough detail to figure out where we are.  If we zoom in to a scale of around 1:50,000, things start to pixelate:

Let’s switch from 100 m resolution over to a 25 m DEM hillshade, same scale:

Sharper images, yes, due to the higher resolution.  Let’s go in more now to a scale of 1:5,000 with this layer:

Oh my – looking quite grainy as we get closer to this layer’s resolution.  Here’s a 15 m DEM hillshade:

Not a vast improvement but there is a bit more detail visible.  Next up let’s throw on the 1 m DEM hillshade at this scale:

Wow!  Now that’s a bit more like it!  Here we’ve got a lot more visible detail given the much higher resolution.  These, of course, are LiDAR data so we should expect to see some rich detail.  Can you tell where we are?

Halswell Quarry is a nice demo for this with a mix of flat and elevated ground.  Switching back to our 100 m hillshade for a comparison, all we see are blocks of gray at this scale:

Here’s the 1 m hillshade again in a 3D scene  looking southeast to just give us a sense of how nicely detailed these data are:

I’d be remiss if I didn’t bring imagery into this discussion, as it’s there where we really start to see the impact of resolution.  We’ll look at a few examples from different satellites to show this.

First off, the Landsat satellite at 30 m resolution.  There are some great things about Landsat, including longevity and coverage, but with fairly major limitations due to resolution.  Here’s what the quarry looks like from Landsat (scale = ~1:6,000):

The Spot satellite has a much higher resolution at 2.5 m though at this scale, this still appears a bit pixelly (Ed. That’s not a real word, by the way):

The built in NZ Imagery base map gives us sub-metre resolution. The imagery here is a mix of the best available and might well be from Quickbird, but it’s a bit hard to tell for sure.  This gives us some really nice, human-scale detail:

As we zoom in, the resolution holds (scale = ~1:150):

Here’s Landsat for that same area and scale:

Minecraft anyone?

With both imagery and elevation (particularly with LiDAR) the past ten+ years have been nothing short of a revolution in resolution.  Over a short period of time resolutions for both have increased by orders of magnitude, which is great from an analysis point of view (but less so from a data storage perspective).

The main point here is that the resolution is critical to the level of detail we can show with raster data.  While this sort of resolution can be eye-popping, it’s really a horses-for-courses sort of thing.  You’ll always want to use the resolution that best suits your analysis, but this was just a bit of fun to show the effect of different pixel sizes. Viva la Resolution!

C