15

I'm working on a project that will involve both "raster data" (aerial imagery and DEMs) and "gridded data" which I understand to be other kinds of numerical measurements taken at grid points, and which could be completely non-image related (eg, population density).

I'm having trouble finding much on "gridded data", so I'm wondering if it's just a different term for raster data? Is there any fundamental distinction? Are there different tools (and particularly servers) for these two types?

(If relevant, this is in the context of Australian state governments...)

Improve this question
0

2 Answers 2

Reset to default
15

No, the two terms aren't equivalent: 1) Not all raster data is gridded -- not even all Earth observation imagery raster data. 2) And while gridded data is usually stored in a raster file format, not all gridded data represents a straightforward rectangular raster data structure.

To illustrate this:

  1. "Gridded" presumes that data in question represents values taken at some set of grid points. These points are presumed to be regular in some fashion, for example, the next row of values in the data structure corresponds the next row of the grid, geographically speaking, in some direction, such as the next row to the south. However, if you look at basic satellite remote sensing imagery, the "raw" (that is, sensor-calibrated, but not gridded) raster data usually called Level 1 corresponds to rows of imagery data as seen by the sensor. For example, a sensor might image 10 rows at an east-west swipe, with the next swipe taking in the next 10 rows, and each data pixel has a lat/long pair (on the ground) associated with it. But as the field of view of a satellite-borne sensor widens at the swath edges, the first row of the next swipe might well, at least partially, overlap and intertwine with the last row of the previous swipe. This makes gridding raster data from a satellite-borne sensor non-trivial in practice.
  2. Conversely, "raster" primarily refers to a characteristic of a data structure, that is, a 2D array. If the gridded data is on a non-rectangular grid (such as triangular, hexagonal), it may be stored in a non-raster data structure such as a tree of some sort.

In your case, the main question is not whether you are dealing with "gridded" or "raster" data. Chances are, the data is either gridded or available on a raster with lat/long attributes in each cell and without overlaps and weird artefacts: processed data is usually corrected for this. The main question for you is whether your tools are compatible with the data format you have.

Improve this answer
6
  • Ah, I see - thanks very much for that explanation.
    – Steve Bennett
    Sep 11, 2013 at 5:40
  • 2
    I am having trouble comparing your use of "gridded" and "raster" to other definitions, such as thefreedictionary.com/grid and thefreedictionary.com/raster. These imply that point (1) is wrong and illustration (1) is actually about raster data.
    – whuber
    Sep 11, 2013 at 14:41
  • You're pointing to a general-purpose non-specialist dictionary (and not one that has a high reputation for lexicography at that), so there's bound to be some uses of the word (many!) that are not compatible with how it is used referring to scientific / geospatial data. My point 1 is an elaboration on the difference between swath data and gridded data when referring to remote sensing data (eg. MODIS from NASA). Both are obviously rasters. Point 2 is merely to say that grid data isn't necessarily in a raster format, though it usually is.
    – chryss
    Sep 11, 2013 at 15:01
  • 1
    Cognizant of the limitations of dictionaries and the lack of real conventions in GIS, I chose those references out of very many that appear to agree with one another. I would appreciate seeing, then, an authoritative "scientific/geospatial" reference corroborating your uses of these words, because your assertion that raster data are not necessarily gridded flies in the face of everything I have ever read about raster data.
    – whuber
    Sep 11, 2013 at 17:24
  • For a sensor such as MODIS for example, the data products in swath format are not distributed reprojected on a grid. At the image swath edges latitudes from successive rows will not be monotonous (see bowtie-effect: dkue3ufa3e1f8.cloudfront.net/files/images/ModisBowTie.jpg and . One way to remove those artifacts is to reproject the image on a pre-defined grid. Some URLs: cires.colorado.edu/~tharan/ms2gt/00README.txt (swath-to-grid toolbox), geospatialmethods.org/documents/ppgc/ppgc.html
    – chryss
    Sep 11, 2013 at 20:30
1

I'll add that DEMs (Digital Elevation Models) are almost certainly originally derived from a set of control points (a "Digital Terrain Model") that are not on a grid. Generally they are measured along ridges and furrows and collected at a density that will allow interpolation in between to fill out a raster at a certain resolution, e.g. 3 m, within a certain margin of error. Sometimes they are first interpolated in between to create contours (such as on US topo maps), which are subsequently further interpolated to a raster.

Improve this answer

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.