maybe this paper will be usefu:

http://www.sciencedirect.com/science/article/pii/S161650471300205X

I think that one of the problems  is that you are using large data sets that  were not intentionally collected  for different scales. Having said that, you need to define  a general problema at the large scale ( ecological, economic and  social) , and interprete the ways of expression of the problem(s)  for a territorial scale and finally for the  farm scale. The indicators to be used for similar variables at the three levels will be with no dougbt very  different, and of different nature. At the higher level you can base yourself in national and international oficial statistics(with all its limitations). But at the territorial and  farm level, you will hardly find  statistics and you will have to go  to primary information taking.

Regrads

I would suggest that you determine the measurable socioeconomic/biophysical explanatory variables that have a strong association with the response surface( fragmentation). Use of exploratory surveys and empirical considerations would help identify these explanatory variables. For example from an exploratory survey, it is possible to determine the influence of human impacts on fragmentation through assessments of farming practices (land-uses), poverty levels/income levels, literacy levels among others. Socioeconomic variables interact with biophysical factors. It may be necessary to use Geo-spatial technologies to determine patterns of cover change over time as a means of evaluating the effect of both socioeconomic and ecological factors on fragmentation.