Whole-World Images
  
   _This answer is slanted towards Global AVHRR Land datasets. Anyone
   care to talk about other images?_
   
   Why create whole-world images?
          _Because they're fun, of course! :-)_
          Continental to global scale images are useful if they show
          information that is studied at a large scale, such as the state
          of the global biosphere. One major measure is NDVI, which
          characterises 'greenness' (see RS/Vegetation FAQ for details).
          Global NDVI images taken regularly over time - at intervals
          between one and two weeks - enable scientists to study change
          in the biosphere in detail.
          
   How do they create whole-world images
          The AVHRR Pathfinder and Global 1KM projects have created
          global land datasets showing NDVI (together with lower-level
          data) from AVHRR imagery, at resolutions up to 1.1KM. The
          global images are created by mosaicing a large number of
          individual scenes, taken over ten-day periods. Individual
          scenes are first stitched to generate half-orbits (in principle
          south to north pole, but generally broken because only daytime
          data is used)! The half orbits are then stitched together, with
          reference to a digital chart of the world.
          
          The key to compositing for NDVI is that each point on the
          Earth's surface is replicated in several images over the
          sampling period. Only the _best_ NDVI value is selected, so bad
          data (such as cloud cover) is discarded.
          
   Why AVHRR? Why not, say, Landsat?
          Yes, Landsat data is just as well-suited to computing NDVI as
          is the AVHRR.
          
          The NOAA satellites, in a polar orbit at an altitude of 833 KM,
          orbit the Earth fourteen times per day. The AVHRR instrument
          images a 2400-KM wide swath as it passes. Thus every point on
          the Earth's surface is viewed at least about once per day (the
          exact frequency of course varies with latitude).
          
          The Landsat series (4-5), in near-polar orbits at 705 KM, also
          orbit the Earth fourteen times per day. However, the swath
          imaged is just 185KM, so a point on the equator may be viewed
          only once in sixteen days. The data with which to generate
          weekly, ten-day or fortnightly global composites is simply not
          available. A sixteen-day composite would of course be subject
          to considerable cloud-cover (see below).
          
          Having said that, it is certainly possible to make large-area
          Landsat mosaics. NASA's Landsat Pathfinder Project (see
          http://pathfinder-www.sr.unh.edu/pathfinder/) has created such
          datasets for the study of tropical deforestation.
          
   How do they get rid of the cloud?
          As noted above, only the best NDVI values from each input
          dataset is used. Clouds will necessarily generate very low NDVI
          values - _clouds are not green!_. Hence clouds are
          automatically filtered out in the compositing process, provided
          there is at least one cloudless view of a point within the
          sample. Thus cloudlessness is not in fact guaranteed, but is
          statistically far more likely than for a single pass.
          Alternatively, it can be assured by collecting data over an
          unlimited time period; c.f. the GeoSphere project).
          
          Clearly this will work if and only if the characteristics being
          studied are dissimilar to any cloud in at least one of the
          available bands!
          
   Further reading:
   
   http://sun1.cr.usgs.gov/landdaac/1KM/1kmhomepage.html
          Global Land 1-KM Project Front Page from USGS/EDC. Includes
          extensive description of the project, and access to the data.
          
   http://atlas.esrin.esa.it:8000/
          Global AVHRR 1KM Server from ESA/ESRIN. The contents is
          essentially the same as the EDC server; readers should normally
          use whichever is closer to you in terms of Net connections.
          
   http://shark1.esrin.esa.it/
          _Ionia_ browser - AVHRR scenes and a browse version of a global
          composite from ESA/ESRIN
          
   http://xtreme.gsfc.nasa.gov/
          AVHRR Land Pathfinder from NASA/GSFC - various global
          composites.
          
   http://infolane.com/infolane/geosphere/geospher.html
          The GeoSphere project (commercial)
          
   All the above references deal with global land datasets. NASA's
   pathfinder program created also Ocean and Atmospheric datasets:
   
   http://sst-www.jpl.nasa.gov/
          SST Pathfinder from NASA/JPL
          
   http://pegasus.nesdis.noaa.gov/pathfinder.html
          Atmosphere pathfinder from NOAA
          
General Questions