I would like to run a Kernel density analysis on point features in ArcGIS 10.2 for hotspot detection, and to do it is recommended to apply a buffer zone to the study area in order to reduce the edge effects during the kernel analysis. I would like to know how to do to figure out the most suitable buffer distance, knowing that my study area is about 331 000 km², my output cell size will be 130 m, and my search radius will be about 150 km (distance of the first statistically significant peak z-scores given by an incremental spatial autocorrelation analysis).
If your search radius is 150km then try that as a buffer distance. The inclusion of a buffer distance is to ensure your densities are not cut off at the boundaries so your suitable boundary will depend on where your hotspots are / how many points there are in those clusters and how dense they are. There's more explanation of kernel densities at geospatial modelling environments here:
And an example of applying bandwidth as the buffer distance can be found in the Journal of Modern Transport:
Identification of crash hotspots using kernel density estimation and kriging methods: a comparison. Authors: Lalita Thakali, Tae J. Kwon, Liping Fu, 2015