A superprocess is a mathematical object that models a cloud of particles which move randomly through space while also branching (splitting or dying) over time. When the process is "supercritical," branching is strong enough that the population tends to grow rather than die out. This paper studies how much time such a process spends in different regions of space, captured by a quantity called the occupation measure, which keeps a running total of where particles have been. The first main result is a law of large numbers: after rescaling by the exponential growth rate, this cumulative measure converges to a predictable shape in space, multiplied by a random variable that reflects the overall randomness of the population's long-run growth.
The more interesting and technically demanding results concern fluctuations, meaning how the occupation measure deviates from that large-scale average. The authors prove functional central limit theorems, which describe these deviations not just at a single time but as evolving processes over all time. The character of the fluctuations turns out to depend critically on a spectral property of the function being tested against the occupation measure, specifically on how quickly the function's influence decays relative to a threshold set by half the leading growth rate. Depending on which side of this threshold the function falls, or whether it sits exactly at it, the behavior changes qualitatively.
This gives rise to three distinct regimes. In the first, fluctuations look like a Gaussian process whose overall scale is itself random, inheriting randomness from the long-run population size. In the second, the fluctuations are still Gaussian but frozen in shape, again with a random variance tied to population growth. In the third regime, fluctuations are non-Gaussian and are driven by a martingale built from the dominant spectral component of the test function. These results extend and complement earlier work on instantaneous snapshots of superprocesses, replacing those with the richer and more practically relevant picture of time-averaged behavior.