I just discussed this with a colleague, and we both agree that the use of the term "scene" is sometime misleading. However, the scene doesn't refer to an exact location. For sure, a scene has a size which is usually expressed in km at ground level, while an image size is expressed in pixel. Then why isn't a "scene" referring to a given "map sheet" ? Probably because there is no exact coordinate matching from a scene to another, even with Landsat. Landsat is very convenient since you usually get very close match within some classes of images. But there are still some differences from one image to another, especially if you perform an analysis over a long period of time : corner coordinate mismatch, row and column count mismatch, resolution mismatch (30m pixel was actually 28.5m), etc...

In the "NASA’s Global Orthorectified Landsat Data Set" (http://glcf.umd.edu/library/guide/PERSMarch_04_313-322.pdf), one can read "Data at these spatial resolutions can provide a high potential mapping accuracy of natural vegetation and alterations to it, if and only if highly accurate scene-to-scene within- and among-date registration is achieved."

Another thing : here, we more and more perform studies with mixed products, combining different sensors. Do you think it is relevant to re-define this term for Landsat-only studies ? Is it really still so informative to refer to a "scene" as being the sensor-specific covered area in this context ?

Anyway, very interesting question. This is not really an answer, just my two cents...

I always used "scene" as a synonym for "tile". So, I agree with your point. I think that we should verify if there is already a standardized definition for "scene" (e.g., within the elements of RS imagery metadata).

1. I did a rapid documentation on the subject, and found these definitions (somehow standardized, if I understand well), which indirectly bring some clarification:

(1) Collection type

B005## [Collection type] must be used if (B001##) is used. It is the way in which the instrument gathers data from the SCENE observed.

(2) Distortion

Compound tag (B250##) [Distortion] must be used if (B075##) is used, and if applicable. It is the departure of positions in the image from those in the SCENE imaged.

("Guidebook for NCDC Template for FGDC RSE Metadata" by Charles S. Phillips, 2006).

I'm sure that there are more focused and convincing documents.

2 @ Mr. Firoz Khan: previously I referred to a question about Landsat images, which had been addressed in the "RS" Topic . Anyway, in reply to your remark, I add that I appreciate art, landscape, photography and drama very much. So, nothing on your side to worry as regard me destroying cultural or/and natural scenes.

I just discussed this with a colleague, and we both agree that the use of the term "scene" is sometime misleading. However, the scene doesn't refer to an exact location. For sure, a scene has a size which is usually expressed in km at ground level, while an image size is expressed in pixel. Then why isn't a "scene" referring to a given "map sheet" ? Probably because there is no exact coordinate matching from a scene to another, even with Landsat. Landsat is very convenient since you usually get very close match within some classes of images. But there are still some differences from one image to another, especially if you perform an analysis over a long period of time : corner coordinate mismatch, row and column count mismatch, resolution mismatch (30m pixel was actually 28.5m), etc...

In the "NASA’s Global Orthorectified Landsat Data Set" (http://glcf.umd.edu/library/guide/PERSMarch_04_313-322.pdf), one can read "Data at these spatial resolutions can provide a high potential mapping accuracy of natural vegetation and alterations to it, if and only if highly accurate scene-to-scene within- and among-date registration is achieved."

Another thing : here, we more and more perform studies with mixed products, combining different sensors. Do you think it is relevant to re-define this term for Landsat-only studies ? Is it really still so informative to refer to a "scene" as being the sensor-specific covered area in this context ?

Anyway, very interesting question. This is not really an answer, just my two cents...

You raise an interesting point. My question was specific to Landsat, but it is true that in this time and age rare is the occasion where we use a single data source. If we open the discussion to other sensors, then I think we could use the common definition that a 'scene' is an "Area on the ground that is covered by an image or photograph".

Many vendors sell satellite imagery by the scene, and in this case the scene doesn't refer to a specific location as you say. For example, SPOT scenes are 60 by 60 km, and even if there is a reference grid based on the satellite ground track, the scenes can be shifted. But in the case of Landsat, the scenes are fixed, so they do refer to a specific portion of the earth. Different acquisitions of the same scene will have slightly different no-data areas, and the corners of the actual images will rarely coincide as you pointed out, but the official boundary of the scene is always in the same place.

I reckon is a bit awkward to think of Landsat scenes as map sheets, because the latter do not overlap while the former do. But if you define 'Landsat scene' as "a fixed portion of the earth, of center coordinates indicated by the satellite path and row, which when represented in a planar projection such as UTM has the shape of a slightly tilted parallelogram of dimensions 185 by 180 km," then it makes sense. So I think we should reserve the term 'Landsat scene' to refer to that parallelogram, otherwise it is ambiguous.

I agree with Richard, with the remark that in my understanding the footprint may be also a part of scene. The latter case is quite frequent, e.g., when you are allowed to order a quarter or a half of the scenes (in order to buy more images on your area of interest with the same budget, let's say for monitoring soil moisture based on radar data). Moreover, I think that "footprint" stands always for the geographic area on which RS data are acquired, while "scene" has an additional, commercial sense, related to the geographic area for which the data are ordered.

Hello Guillermo! YES!!! If you are taking votes, put me on the side of supporting scenes as the relatively invariant geographic area represented by a given path / row, with the images representing that area. A single date study could have the number of scenes and images being equal, a multitemporal study will have images in greater numbers than the scenes. With the compositing work (aka best available pixel) that is becoming increasingly common it is good for you popularize this distinction.

I am sure most of you Landsat-fans have seen this, but just in case, here is a terrific video that shows how "the instruments aboard the satellite are able to view a swath 185 kilometers wide and collect images along that swath as the satellite proceeds through its orbit." http://landsat.gsfc.nasa.gov/?p=7909

Thanks a lot to all of you for your insightful comments and feedback!

It seems the consensus is that the term 'Landsat scene' should be reserved for the geographic area enclosed by each of the polygons in the WRS2 shapefile (http://landsat.usgs.gov/tools_wrs-2_shapefile.php), especially in multitemporal studies as Mike pointed out (btw, awesome video, Mike, I hadn't seen it).