I've got lots of tiny streams on a hydrography shapefile. Historic texts indicate that certain ones have headwaters at "sources" or springs, and I want to try to find these. Browsing the historical imagery on Google Earth was my first attempt, but the ground cover and tree canopies are too dense to properly identify a pool. What kinds of satellite imagery types would help to identify such features?
As someone who did feature capture from imagery for a while, I would caution you against expecting a pool at a spring. The majority of the ones I've encountered (both in capture and on the ground in person) don't have one. We often referred to ancillary sources to suggest/confirm a spring. Depending on your purposes, USGS quad sheets or hydrography datasets may prove useful.
As for imagery, time of year would be key. Imagery taken during spring or fall would be best - minimal snow and canopy/vegetation cover makes for better visibility. For this reason the US Ag Program (NAIP) imagery wouldn't be ideal, since their purpose is showing peak crop growth. Google takes their imagery from a variety of sources and it's rather hit or miss on time of year. I know some local county governments time their acquisitions to minimize canopy cover. Depending on your area of interest location, some times of the year may be better than others for spring activity/flow.
If you have tiny streams then you will want to have satellite imagery with better resolution than Landsat (30 meter pixels). However, Landsat has the best historical coverage. I would do a combination of imagery and elevation (DEM) data. Using a DEM to make a hillshade, or some hydro analysis (flow direction) will provide you a great combination of options to identify tiny streams and potential head waters that are much smaller than a 30 meter pixel. Otherwise you will be looking to get IKONOS or QuickBird four band imagery, which can get very expensive.
Landsat imagery would be helpful. Different bands can be utilized separately or together depending on your needs, in your case delineating water and land boundaries would be near infared. If the streams are as tiny as stated, they may not appear due to the resolution or lack thereof.
I have spent many years surveying rivers in the UK and have visited the sources of many streams. My experience in the UK a spring is rarely a pool of standing water but are "flushes", basically water seeping out of the ground. Springs could be swampy areas or heathland dominated typically with Juncus. But we do have the classic water bubbling out of the ground pools for sources (Eg. R. Itchen).
Such a feature as identified by the others is too small to capture or obscured by vegetation. You don't say where in the world you are working? If it was in the UK our national mapping agency supplies those annotations as part of MasterMap. You could convert those to a point and snap them to the nearest end point of a stream network, that would be a good first pass filter?
Also when you say spring are you talking about water seeping out of the ground or an Aquifer creating a pool?