I find it best when thinking about problems like this to take the approach of thinking of our universe as static 4-dimensional thing — what physicists would call a Minkowski space-time. The “present” is a three-dimensional slice of this space-time that is transversal to the time axis. “The past” is simply that portion of space-time that is on one side of this slice, while the future is on the other side.
Of course, we don’t get to choose arbitrarily which side of our present slice is “the past”. This is in contrast to the spacial dimensions, where, for example, we get to choose arbitrarily which side of the universe is “below” the spiral of our galaxy. Several things distinguish the past from the future. The one that is probably the least philosophically problematic is the arrow of entropy: the 3D slices of space-time with smaller time coordinates are the ones with lower entropy.
However, the more interesting thing that distinguishes between the past and the future is the arrow of memory: we remember the past, but we don’t remember the future. Somewhat more precisely, you can tell that time slice S[sub]1[/sub] precedes S[sub]2[/sub] on the time axis if the brain of Tyrrell McAllister in S[sub]2[/sub] contains a record of the state of the world in the vicinity of Tyrrell McAllister in S[sub]1[/sub], but not vice versa. (I said that the arrow of memory is more problematic than the arrow of entropy because it is harder to define precisely what a “record” is. But it is even harder to rigorously distinguish between cause and effect without presupposing a distinction between the past and future, so I’m staying away from the arrow of cause and effect.)
In our universe, with one temporal dimension, the past is that half of the timeline that we remember. That is, at time coordinate t[sub]0[/sub], we remember those events with time coordinates t < t[sub]0[/sub].
Now consider an entity living in a universe with two temporal dimensions. So, instead of having a timeline this entity will think of time in terms of a “timeplane”. To specify a point in time will require giving two coordinates. By analogy with the above, I suggest that at time coordinates (s[sub]0[/sub], t[sub]0[/sub]), our entity will remember those events with time coordinates (s, t) such that s < s[sub]0[/sub] and t < t[sub]0[/sub].
Whereas for us, the “past” is the left half of the timeline, to our entity, the past will be the lower left quadrant of the timeplane.