Selecting an appropriate spatial scale in landscape ecology!?


I am currently studying the effect of organic farming on honeybee colonies. I have calculated the percentage of organic land in several buffer areas around the hives (from 100m to 3000m in 100m steps) and I would like to study how an increase in this percentage changes honeybee performance. Unfortunately we do not know much about the foraging behaviour of bees, but we know that they forage at variable scales, depending e.g. on the flower availability.

I know some people do this based on maximal correlation between the response variable and the landscape variable. It appears to me like data mining without correcting for it...
However I should definitely choose a scale where I expect an effect.

Some considerations: The smaller the scale (regarded buffer area)
1) the more area that potentially affects bees is disregarded
2) the more sure can we be that distance to the hive within the regarded area is negligible

Looking at the histograms of organic farming at different scales it gets apparent that, the smaller the scale
1) the higher the maximum percentage of organic farming (some hives are placed next to organic fields)
2) the more hives without any organic land in their buffer area (e.g. at 100m: 90 %)

I would be happy about any suggestions to identify an appropriate scale. Thanks in advance :)