Useful color studies / thumbnails keep the local hues in mind.
Stuff like this
>>5975308 is nonsense where they just ignore the local hues and apply colors randomly because they read about a color palette or used a tool to recommend random 3 color harmonies.
Decide on a local hue arrangement first.
Most of it will be logical (grass is probably green, sky is probably blue, tree is probably brown).
But there can be a lot of design based decisions too (cars can be any color, dogs can be a subset of colors, an apple can be green/red/purple etc, so your decision should either be story based or composition based rather than just random).
Local hues aka colors don't need to harmonise but you can still choose low vs high saturation potential or low vs high value potential colors for objects that are important or not important.
For example if you pick a yellow car it will likely be a light value since yellow tends towards saturation at a high value and if it is surrounded by mid or dark values it will draw attention to itself.
If you pick a blue car it will likely be a dark value unless you choose a low saturation blue or go towards a cyan blue.
Once you have the local hues decided, next you decide the lighting. This is what you are doing color thumbnails for. You are seeing how different lighting conditions affect the local hues of the scene.
Once you have the local hue, you should be able to predict what a blue light or a yellow light or a red light or a green light etc does to that local hue. This is what you study from color & light courses. The interaction of color light sources on local object hues.
Most lighting setups are a temperature contrast between the lights and the shadows. The classic is a yellow-orange light with a blue-purple shadow color that simulates direct sunlight and the ambient shadow color of the blue sky. But you can introduce weather effects that act as diffuse light sources and filters or artificial lights like green and purple.
