With this, you mean that you would keep the leds closer one to another than I was planning?
It is not totally necessary, especially the higher from the water surface the light is..or the wider the light optics are. Most "on chip" lenses are 110 degrees or greater
"Clustering" colors and whites give a better blend, as their (if using 90 degree lenses) cones blend together..
This is an "on chip" example. It is called 'Ocean Coral White" and is 660nm red, cyan, and royal blue in a cluster (in this case all 3 on the same heat sink.)
Not so important w/ whites....
Any combination of red(ish) plus green(ish) plus blue(ish) blends to white..
another example of a DIY w/ blending:
Broad spectrum, fully controllable LED shoji lantern
13 X cool white Cree XT-E
13 X warm white Cree XT-E
2 X 660nm "deep red"
2 X 475nm "deep blue"
2 X 495nm "turquoise"
On the above "I" would have doubled the color clusters..and ran each on it's own channel w/ possibly combining the royal blue and cyan so those would be either 2 or 3 channels.
Whites would need at least 2 channels or cutting down the amount and, depending on preference changing the ratio or ww/cw.. BUT that is a matter of taste. 50/50 ww/cw has a tendency to be too "yellow" looking for many.
A lot depends on the actual K value..
alternatives include red/ww on the same channel.. Blue/cw on another.. Leaving green/cyan on one more.. This creates a natural R(ish)/G/B(ish) blending capability..
but back to your orig question..clustering does help blending of colors.. and your spacing would only need to be adjusted accordingly..
What your trying to avoid:
Early build w/ 2-10W 660nm red and 15 warm whites in a row..
this is 4 660nm red, 7 3500k and 7 6500k LED's (3W)(the ww/cw channel is ramping up..so it is not on full) Native optics..110-120 degree)
Disco effect is not nearly as pronounced ...
Ground cover is red granite/flourite/white quartz