In your case I would suggest using the stereo version because:
- You already said Kinect is technically not a possible solution because of the sun's IR light
- A reasonable good lidar sensors costs a fortune even for a 2D version
- Stereo vision is not expensive. You get 2 webcams for 100$, which is even less then a Kinect.
The major problem with the stereo is the computational requirement which is almost impossible to achieve on the UAV. I suggest sending the videostreams to an external computer and analyze the data (and send it back if necessary at all). wifi is able to handle this datarates easily, especially the 5GHz version (802.11n). You might even think about compresing the data with h264 codec, which can be accelerated by specifc chips.
I guess your plan is using it for a SLAM application. This is possible, but you have to keep in mind that the wifi connection is not reliable at all. Maybe a nearby weather radar station is interfering your connection and UAV gets no data update anymore until the connection is reastablished. Anyway a good UAV-software will be able to handle the wifi disconnects.
So relying on this might not the best solution. alternativly you can stack all your computational power directly onto the UAV, like I do it for my research project. Therefore you need a decent UAV which is able to carry a complete motherboard with modern CPU and maybe a GPU; the GPU is a sake for stereovision.
At the end (if SLAM is your final goal) you can look at other options like MonoSLAM.
Note: This answer does not say you should do this. I got way too less information about your project to give a high quality answer. This is more a collection of thoughts/ideas to your problem, which should help to find a feasable solution.