I have to agree with @stadiumwave after looking at the individual years and the blend of 3 years.
the smoothing looks like there’s something there, but in this case, no. That’s why we have to be careful about combining years into a composite map, especially if there are so few years in the blend. It’s fine if they all show spatial similarities, but if they’re so different that they don’t individually match the blend, we’d have to toss the blend due to small sample size.
PS. This advice goes for me, too. I’ve been known to blend liberally on small sample sizes…