Horukuru,
As per my post on the Zoom F3, the bit depth is all to do with the difference between the loudest and quietest sound in the recording. If you have a good signal to noise ratio (i.e. and loud bird vocalising close to you) and you adjust gain towards the top of the meters (without going over), then 16 bit should be fine. I actually have a Wildlife Acoustics SM Mini, which with two mics fitted, at high sampling rates drops at about 14 bit and still sounds OK. CD quality is actually only 44.1KHz at 16bit.
There has been much debate about HD streaming music, and whether humans can detect any difference in higher bit and sample rates. I saw a post on the internet where someone took a 24 bit music file, converted it to 16 bit, then played the two track 180 degrees out of phase, so that they cancelled each other out. The result sounded like silence, with just a small amount of noise visible if you really magnified the wave diagram. I suppose this makes sense, as the silent bits in a 16bit CD sound to all intent and purpose like silence!
As per the other post, I think 24 bit gives a more headroom to increase levels in post production, in the case where the gain setting was a bit low, but you still had a good signal to noise ratio at the mic.
Sample rates are more of a puzzle to me. Basically say you have a 10KHz sound, then the sound is oscillating 10,000 a second. If you sampled the wave at the same rate, you would always be sampling the same part of the wave (i.e. the crest or the trough etc.). From this sample information it would be impossible to recreate the wave signal. As Vollmeisse states, as a minimum you need to sample the wave at twice the frequency of the sound. Some people will say that you are still only sampling an analogue wave digitally at two points, so the sound is not as good at the original analogue signal. Others will say that the sound wave that is created from the digital sampling at twice the sound frequency is indistinguishable from the original. I don't have the expertise to comment, but what is for sure is that even with 48KHz sampling rates, the argument of limited sampling would only apply to high frequency sounds - a wave of a bird calling at 4Khz would still be sampled at 12 points along the wave!
After making an overlong response, to be honest, there isn't much of a downside to recording at the highest bit and sample rates available on your recorder. I generally record at 24 bit 96Khz, for no better reason than my recorder can do it and 192Khz seems OTT - I have however recently ventured into 32bit float, which is a different issue altogether (see Zoom F3 post). Basically the only downside of high sample rates and bit rates is perhaps file size and battery life. It is for this reason that passive recorders for long term deployment for days or weeks, normally record in 16bit 48kHz. But for normal recording work, you probably won't see a massive difference in battery life, and memory should not be a problem - a 32GB card should give you over 15 hours of stereo recording at 24 bit 96Khz - you may want to reduce the sample rate if you are away on a trip, and want to make your cards last longer - halving the sample rate will halve the file size. In you make the original recording at a high sample and bit rate, in software you can generally save a version at a lower bit rate or sample size, so can easily save to CD quality once you have completed any edits.
Regards
Jon Bryant