With just about any system, you have 0 guarantee that an objext fuether than your camera will be at lod0.
Thats likely the problem in a nutshell.
You either have to adjust LODs size on screen so that a worse case scenario (you sit right at the middle of a component and look across into another one or at the corner where another 3 meet - the corner one of which will oikely be the smallest size on screen) is still LOD0 and loose out on performance - which works in any case - or you have to ditch the LODs entierly.
With landacape, if you insist, you can maybe try to manually bump up the componet to lod 0 based on a sweep from the actor.
Make a cone that matches the camera fov, set the distance of it to the max of your gameplay element.
Then, detect the landacape components in a sweep collision using that cone.
Force every component in that cone to LOD0.
Should work, but likely you need to do it in c++ for deeper access to the landacape in order to override the LOD.
Idially you do this in a camera matching sweep so that the player doesn’t see the lods pop.
The further out you need somerhing in a desert area to be lod/collision accurate, the better the model has to be.
You can’t cover it up with bushes or hide it out in any way, so maybe even no LODs at all could be a thing…