I wrote a few weeks ago about resolution and stressed the equally important (and often under appreciated) concept of Minimum Mapping Unit (MMU). And, that if you have a significantly large MMU, it almost doesn’t matter what the input resolution of the imagery was. I’m going to build on that previous discussion a bit and make a slightly finer distinction between level of spatial detail (represented by those two previously covered concepts) and mapping accuracy.
The fight I often have a hard time making is that a product is not necessarily less accurate just because it is at a coarser resolution / maps less spatial detail (has a larger MMU). While the lack of detail may directly affect the usefulness of the product for your need, and that should obviously be a consideration, the fact that is misses some detail does not relate to accuracy at all (unless these features are larger than the stated MMU), just the size of the features the product is intended to map. The two factors are independent of each other.
The high-resolution aerial photograph to the right highlights the types of features that are included within a 30-meter pixel’s area (imposed grid). Pixel 1 is a homogeneous pixel covered by trees and would be considered as forest. Though it is a mix of evergreen and deciduous forest types, so if you were mapping to that level of detail it would be mapped as mixed forest. Pixel 2 is a mix of both vegetation and a large house. Since the house makes up more than 20% of the area, this pixel would be designated as low intensity development (in the C-CAP scheme). Pixel 3 is even more mixed, with the structure and paved surface making up less than 20 % of the total area. This pixel would be classified as either open space developed or forested. While none of these classifications capture all that is present within the imagery, they are the calls that are best at the 30 meter mapping scale. This could be thought of as spatial rounding. This might be a good example of an area that could use more detailed mapping, but you can also image the reverse… where we might be mapping large agricultural fields, where that level of detail is not gaining any additional information (and might end up making the work more difficult and costly).
Key point: It’s not that the one value/class mapped for that coarser pixel is wrong (it may be the best depiction at that scale)… It’s that one category over that area isn’t detailed enough.
So why doesn’t NOAA just map more spatial detail?
Short answer: Cost (and the time involved to analysis the same are in greater detail).
Longer answer: Many people do map to a finer level of spatial detail. Typically just not over a very large area. NOAA does too, but we just can’t afford to develop these products over the entire coastal U.S. area for which we are responsible for mapping. The average price for developing our 30 meter products, for instance, is typically less than $2/square mile. On the other hand, our high resolution land cover tends to cost a bit more, ranging from $100 to $250/square mile (depending on the size and complexity of the area).
So, it is up to the user to decide what level of detail is necessary to fill their need… and what price it is worth to get (i.e. Is the juice is worth the squeeze?).
Footnote: NOAA maps where there is a need for that additional detail, and that means that we are partner up with groups that have a need at a more local scale (FYI – we’re always looking for partners). If you’re interested in learning more about these products check them out at:http://coast.noaa.gov/digitalcoast/data/ccaphighres
Image source: University of Connecticut Center for Land Use Education and Research, n.d. Web. 12 Jul 2011.