Yes, I have thought for decades how to implement it. Here are some of the ideas
In draft: I recorded the wind sound noise in the city, flat countryside and sea at 1 meter (Different wind intensity). I use that recording stereo audio as a random wind speed and direction generator. The direction is made by the stereo and the intensity by the volume and frequency. This gives me a nice starting real-time parameter that I loop it over time. The atmosphere has a curve value multiplier in function of the altitude. It follows the indication of the Nasa records and is calculated in real-time in function of the glider altitude. It also gets the Re numbers that is used for selecting the correct airfoil aerodynamic chart. This is just for starting because then there are layers from 0 to 100m, 100 to 700, etc… that [layer*] can go up and down and has different pre-calculated aerodynamic properties like Nc (turbulence) rotors, waves, the thermal characteristics. So that at low altitude when landing the pilot suffers the big gust bursts. The variation and altitude are in function of the day conditions. Probably the pilot will be able to move the traditional soaring dew point atmospheric aerodynamic conditions chart. And the chart will be given by history data of that particular place. That on flat land. The LOD houses and trees game objects include a trigger collider in direction of the wind and trigger that learps to the turbulent city wind audio model. The intensity is in function of the distance. It also creates a rotor and wave that is calculated in real-time. In 2000 I made an Excel algorithm that simulates all that in 2D. My intention is to use a lot of this 2D calculation at start. So there are real-time rays from the glider pointing in the direction of the wind looking for obstacles, rays looking down for calculating the terrain locally as well as looking up searching for other layers*. These layers are meshes. There will be, also meshes where air temperature changes direction for the base of the clouds, rotor delimitation, etc… These meshes that mimic the terrain with the addition of the waves [or rotors] are pre-calculated at the start. And let me collect data for the real-time calculation. It will become more complex with mountains, change of atmospheric condition, enter of breeze and storm. But at the beginning will be just flat land, and obstacles. Thermals are calculated by temperature, and colour of the surface texture [point generator]. Also can at some point be move buy wind. Then the bubbles move in real-time inside that atmosphere and shift following the wind direction just like a ballon. It can contain an attraction to other thermals or it can be divided in 2 like a tree. I was also considering upside down tree meshes for this. When the glider is more close to the centre of the column (sphere collider media) the additional wind shift will follow the thermal chart’s curve[of the zone]. Data collected is use to define the size and intensity in function of the altitude. So that the mix with the atmosphere creates more thin and broken at low altitude.
The intention is to put the pilot in normal conditions and dangerous conditions. As for example landing with gust that makes the glider float, thermals that flip the glider, cumulus that sucks at the base, c. congestus that throw water decreasing the air Nc efficient value. Mosquitos or ice in the leading edge, spiders and serpents in the VR cabin! 