While it is much simpler to define the duration (time) of ONE complete day (for earth), by simply calling it to be the period between sunrise to next sunrise (which we have currently divided into 24 hours), how is the period of an earth year defined? I.e. how do we know that we have come to the same position after completing one revolution around the sun? If it is with respect to other stars / celestial bodies, can we consider that to be realistic / accurate? - because the other bodies are also changing their position continuously.
Even the first item (defining what a "day" is) is not as obvious as would appear. Using sunrise to sunrise (or equivalently, using sunset to sunset) is just about the worst option, for multiple reasons:
- One sunrise to the next is about a year at the South Pole.
- People at the same longitude but different latitudes will disagree on when the start of a day occurs. Even though Capetown, South Africa and Stockholm, Sweden are at close to the same longitude, there will be marked disagreement between them regarding when a day starts based on sunrise (or sunset).
- Sunrise (and also sunset) are hard to predict because changes in atmospheric conditions can have a marked effect on when sunrise (and also sunset) appears to occur.
- The equation of time also makes a mess of things.
- The Earth is not quite as perfect of a timekeeper as people once thought. Atomic clocks do a much better job of timekeeping than does the Earth's rotation. People to this day agree to disagree regarding whether the concept of leap seconds is a good or bad idea.
The first can be addressed by only using locations between the polar circles. Switching from one sunrise to the next from one noon to the next addresses the next two issues. That leaves the equation of time, which is still an issue with regard to timekeeping.
I'll ignore the last issue for a bit. A sidereal day is the time it takes for the stars to repeat their positions. The sidereal day is not subject to the equation of time. Ignoring that last tricky issue, the current approach is to use the sidereal day as the basis for timekeeping and then scale to result in a mean solar day that more or less stays in sync with one noon to the next, after accounting for the equation of time.
A mean solar day ("mean" means average in this case) should have 86400 seconds in it per this concept. Except it doesn't. The Earth's rotation rate with respect to the stars varies. It varies over the course of days to months due to weather. It varies over the course of seasons to multiple years due to transfers of angular momentum between the solid Earth, the atmosphere, and the oceans. It varies over the course of years to millennia due to transfers of angular momentum between ice coverage, the mantle, the outer core, and the inner core. Over even longer periods, it varies due to transfers of angular momentum from the Earth's rotation to the Moon's orbit.
These variations matter a lot to astronomers that perform milliarcsecond radio astronomy. (Some now need sub-milliarcsecond accuracy.)
Regarding a year, there are multiple ways to define the concept from an astronomical point of view. Two obvious approaches based on the Earth's orbit about the Sun are the time it takes for the Earth to revolve by 360° degrees with respect to the stars. This is the sidereal year. On 1 January 2000 the sidereal year was 365.256363004 days (days of 86400 seconds) long. The other obvious approach is the time it takes from one perihelion passage to the next. This is the anomalistic year. On 1 January 2000 the sidereal year was 365.259636 days long. This is a bit longer than the sidereal year due to perturbations by other planets.
Both of these definitions of a year are not all that good with regard telling farmers when to plant their crops. Telling farmers when it was a good time to plant and a good time to harvest was one of the key motivating concepts that drove the development of ancient astronomy.
This leads to a third astronomical definition of a year, which is the time from one vernal equinox to the next. This is the tropical year, 365.24219 days long. This is shorter than the sidereal year, primarily due to the precession of the Earth's rotation axis.
Some want the year to be highly predictable for planning purposes. This leads to the concept of a calendar year. The Roman empire used the Julian calendar, which had a leap year every four years. This results in a year that is 365.25 days long on average. This made the calendar get out of sync with the tropical year. It's better to not add a leap day in some years that would have been a leap year per the Julian calendar. The Gregorian calendar skips leap years on years that are divisible by 100 but not by 400. This results in a year that is 365.2425 days long, which is close to (but not the same as) a tropical year.
There are many other definitions of a year.
One takes a reference point and simply observes repeatedly and carefully - and counts day to the next solistice, thus when the sun is again at the exact same place in the sky.
That method is believed to have been used already in neolithic times, and there is evidence that places like Stonehenge, Ring of Brodgar served this purpose to make it easy to keep track of the time and to make sure that re-occurance of a solistice is easily observed. Similar argument can be made for various other sites all over the world, including the pyramids in Egypt and Southern and Middle America.
You might also find this article interesting dealing with the precession of the equinoxes in neolithic times - and includes a long section on determination of equinoxes and solistices in neolithic times.