The overlap would be a tangled morass. On the frontiers of the field, there would be a hodgepodge mix of understanding and knowledge. But there are vast tracts of the mechanisms of Evolution that we understand rather than merely have knowledge of. Including natural selection.And in doing so maybe there's an appearance of Understanding of evolutionary laws and not "just" knowledge of them.
Testability would be used to justify knowledge, but has less to do with understanding. Understanding is like suddenly seeing the picture after connecting a certain number of dots. Knowledge is that point A connects to point B. Testability would be confirming that point A connects to point B. Prediction would be a proposition that point C connects to point D. Understanding is that connecting A through Z makes a smiley face.Doesn't testability allow science to link knowledge and understanding?
You can arrive at understanding without having full knowledge. However, with full(or near-full) knowledge, you can be more confident that your understanding is correct. If you have only sparse knowledge, you may have an understanding of it, but that understanding could be false. We've had enough knowledge to understand evolution for a while now.
One problem with this is, it's a lot of knowledge, and takes a while to learn. Another problem is that having full knowledge does not entail truthful understanding. You could connect dots A through Z, but stubbornly demand that the picture is a horse's ass rather than a smiley face. If you're determined to see a horse's ass rather than a smiley face, then no amount of knowledge would persuade you otherwise.
Explanation comes before prediction. You start with a premise and a universal law, and the output is an explanation(as long as the explanation logically follows). Evolution does well at explaining.If a theory is not in a position to predict, can it be in a position to explain?
You can use an explanation(hypothesis or theory) along with measurements to formulate predictions. The problem with evolution are those measurements, also known as initial/boundary conditions.
If you were to predict where the Earth would be in a few thousand years, you have very few initial or boundary conditions to consider. But if you were to predict what a rabbit will do in ten seconds, there are trillions(or likely more) initial and boundary conditions to figure in to the equation. Just to "model" a bunny rabbit to produce a prediction, you'd need the fastest computer on Earth with an exact virtual replica of the rabbit's brain. Even then, due to the sensitivity to initial conditions in all areas of life, your prediction could be dead wrong merely because you didn't correctly measure a single external variable. You didn't note the physical similarity of a nearby stump to a wolf's head. So the rabbit bolts, and you didn't see it coming.
The rabbit example is from behavioral science, but it serves as a good example of the complexity of initial and boundary conditions. Evolution shares that complexity, and is arguably even more complex than behavioral science.
In other words, evolution can explain, but due to the complexity of conditions, it is hard to form predictions. Not to say there aren't predictions. It wouldn't be science without them, right?
http://answersinscience.org/evo_science.html