Rather than find just the right HDRI type image to put in the GI environment slot, it can be useful to simply create a gradient. The problem, and this goes for images too, is understanding how the mapping translates to your scene. I did a few tests to make sure I understood what was happening, and now want to share some of the results.
First, the V direction. This one is very easy to understand. The gradient is 2D-V and 'stands up' vertically with the left portion defining the bottom, the right the top. So whatever color is on the left will shine UP towards the model, whatever color is on the right will shine down. The exact center will shine horizontally, all of these being equal when seen in plan (a top view). The areas between the ends and center angle in between that end and the center/horizon. The very ends will be weaker than the center for the same 'amount' of color:
The gradient in U is more complicated. Initially I tried the offset angle only, found that as it rotated, it drastically changed the 'weight' of one end or the other, so one end of the gradient would overpower the entire lighting scheme. This unbalance seemed apparent at an offset of zero, so I went with 90 degrees as a default for the direction test.
The scene is shown in plan/top view. The lighting is even horizontally, meaning that if you look at an elevation/side view, you see no up/down shadowing, only sideways.
So with the offset at 90 you have the center (50% position on the gradient, which is 2D-U) shining 'up', meaning from the front. The two ends wrap around to form the back direction. The two sides produce from-the-left light at the 25% position and from-the-right light at the 75% position. Obviously you get mixed angles by using the positions in between these.
Then you mix the U and V in a layer or fusion shader to get a full lighting model.
I combined a U and a V gradient to make a scene lighting. I combined them by using 'multiply', though you could use 'soft light' or simply use 'normal' with the fade at 50% or so. It depends how you've set up the horizontal and vertical gradients. What I've done here is fairly even, so the lighting is fairly even as a result. I think with more darkish parts of the gradients I could create stronger lighting differences. But typically we use the GI environment for soft, overall lighting. What I've done is control where colors are.
Here is a mash-up of the gradients and how they look on the scene objects:
Even if you are using an image, if you use it with spherical mapping and the U offset at 90, you should be able to predict where the brightness and color will come from by observing the map relative to the directions I showed.
I hope this will helps all
