The performance of clipping depends on the number of vertices in the clipping mask and the number of triangles in your region and mesh attachments. You can find the stats for the latter in the metrics view: Metrics view - Spine User Guide

Please post them so we have a better idea about your model's setup.

Like Mario said, clipping has to be calculated for all of the attachment images that are being clipped, and it looks like basically everything needs to be clipped on your character when they are in the ground (since the only visible part is the top of their head).

I believe clipping works by calculating new mesh vertices for each attachment, so it is done on the CPU and is expensive if a lot of attachment images (with lots of vertices) need to be clipped.

I would look into using a Sprite Mask (https://docs.unity3d.com/Manual/class-SpriteMask.html), to mask out the area that is considered 'below ground', and then use the "Mask Interaction" setting on the SkeletonAnimation component to make it 'Visible Outside Mask'.

When you use Masking/Stenciling, it's calculated on the GPU (it basically checks whether or not the individual pixel should be drawn to the screen or not, based on if that pixel location has had its mask/stencil value set) and is very efficient.

There are some performance tips here:
Clipping attachments - Spine User Guide: Performance

The best case is making your clipping polygon have only 3 vertices (a triangle). You can make the triangle very large, it's size doesn't affect performance.

Next to get good performance you need to try to clip as few attachments having as few vertices as possible. Use region attachments (4 vertices) where possible. Hide attachments once they are fully clipped.

Doing all that may be a pain and if you still have lots of attachments, it may still not perform well. Doing the clipping other ways, as suggested above, may be easier.