Playing around a bit with the Google Timelapse project (sponsored by Time magazine, I guess?).
1: It isn't clear when these "snapshots" are taken in time. For example, the 1984 shot of Mendenhall clacier is probably taken in Northern Hemisphere winter (a guess based on snowfall over Glacier National Park). So is the 1989 and most of the late 90's images. Whether the final snapshot's indications of climate change are actually so dramatic, and not the cause of aliasing by comparing two different places in the seasonal cycle isn't clear. This needs to be corrected.
2: Smoothing is playing a big role in producing the image. Of course it takes a lot of passes to create an image of the globe, so this is necessary. But just look at what happens in the Atlas mountains between 1988 and 2003. LANDSAT catches the mountains in a time where they are ice-capped in 1998 , but it doesn't in the following years. The result is an unphysical, linear interpolation between that snowy year and the snow-free years that looks a little weird.
Additionally, that question of timing I raised about Alaska glaciers is different around the Rocky mountains, by watching Lake Tahoe cover with ice we can infer that the first few years in Colorado were summertime, the remaining were taken in the winter. We need timestamps!
3: Chinese growth happens fast! Search for Fengdu, for example, and watch an empty grassland turn into a huge city between 2000-2005. You can pretty much pick any place along the Yangtze River and watch massive growth occur in the early 2000's until today.
4: Speaking of China, check out the Three Gorges Dam! You can watch as huge regions upstream are flooded as a result of its opening in 2006. Kind of sweet.
Oceanographer, Mathemagician, and Interested Party