This setup should determine an adequate configuration of rayds / raydsv_max.

The basic setup is a GLORIA simulation of a plane flying a circle and scnaning the inner part of the circle. The atmosphere has a grid sidze of about 50km x 50 km x 250 m.

This figure shows the current dilemma.

The y-axis is always given in „percent“, the other parameters are given in kilometers.

Reducing either RAYDS or RAYDSV_MAX reduces the (average and standard) deviation to the simulation run with a fixeds step length of 0.01.

However, reducing rayds_vmax increases the maximal difference of rays pointing upwards. Further, there seems to be a combination of rayds/raydsv_max which is „optimally“ bad, as there is a local maximum/minimum in the maximal/minimal difference at about the same location.

Apparently, most of the effect was introduced by an unlucky choice of location of line-of-sight points in relation to the segment they represent. By moving the line-of-sight points from the start of the segment to the middle a significant error reduction was achieved: