The Earth-like planet I'm working on is the culmination of years of labor by the galaxy's best scientists. It consists of fully artificial ecosystems, and is, for all intents and purposes, like Earth. The life-forms are carbon-based and have similar structures, metabolic processes, and habits as animals on Earth. Since the scientists and engineers set things in motion, they have been monitoring the planet from the equivalent of Low Earth Orbit, in a number of spacecraft equipped to monitor the atmosphere, oceans, weather patterns, etc.
A new instrument has been proposed: a satellite capable of monitoring the movements of large groups of animals. In an effort to interfere as little as possible with the current ecosystems, the scientists removed almost all equipment and temporary artificial structures from the planet. This means, for instance, no tracking collars. Instead, the group wants to use what is essentially an Earth observation satellite, capable of taking pictures of herds of animals.
The test target would be the near-equatorial regions, including savannas, where there is little tree cover. The various species of quadrupeds there are roughly zebra-sized, and move in groups of about 50, give or take. They aren't migratory, but they do move around in search of new food sources, as well as to avoid predators.
What I'm currently thinking about is using optical imaging, as a reconnaissance satellite might do. The thing is, I have no idea if that's enough to observe herds this small. The satellite will have to be cobbled together using spare parts, so there's not a lot of flexibility when it comes to instrumentation. I'm therefore open to other methods besides optical imaging - although I can't see to come up with any. The technology is essentially that of present-day Earth (2018).
Using optical imaging or some other related technology, can my scientists correctly identify and track herds of these quadrupeds from Low Earth Orbit? Ideally, this would be done by directly taking pictures of the herds, but if another method is possible, I'm open to suggestions.