In finance, the “Portfolio Effect” is a term used to describe the fact that the risk in a portfolio of non-correlated investments is lower than the risk of a single investment. The portfolio returns will tend to be less volatile than that of a single investment.
In oil and gas reserves, the “Portfolio Effect” (sometimes called the “Aggregation Effect”) refers to the fact that P90 volumes (or any other probability level other than P50) can not be summed.  Arithmetically summing 10 reserves at the P90 level will not yield a P90 result, because that assumes that the every one of the 10 individual items encounters its 1 in 10 “Low Case”. That would be equivalent to rolling a “1” on a 10 sided die 10 times in a row. The probability of the group total exceeding the sum of the P90 individuals will, in fact, be well over 95%.  If the portfolio is increased to 1000 items, achieving the sum of the P90s is virtually certain.
In practice, the evaluator may choose to relax the level of certainty assumed for individual items, knowing that the aggregate will me more certain than the parts. For example, the Canadian Oil and Gas Evaluation Handbook COGEH cites as a rule of thumb that an evaluator can arithmetically add entity- or field-level deterministic estimates of proved reserves of lesser probability (for instance, a judgmental > P65) that will result in a greater overall probability (P90) at the portfolio or aggregate level, provided enough entities are added together.

PRMS 2018 also speaks to this in section 4.2.5.2:

Two general methods of aggregation may be applied: arithmetic summation of estimates by category and statistical aggregation of probability distributions. There are typically significant differences in results from these alternative methods. In statistical aggregation, except in the rare situation when all the reservoirs being aggregated are totally dependent, the P90 (high degree of certainty) quantities from the aggregate are always greater than the arithmetic sum of the reservoir level P90 quantities, and the P10

(low degree of certainty) of the aggregate is always less than the arithmetic sum of P10 quantities assessed at the reservoir level. This “portfolio effect” is the result of the central limit theorem in statistical analysis. Note that the mean (arithmetic average) of the sums is equal to the sum of the means; that is, there is no portfolio effect in aggregating mean values.