You need reference sites and degraded sites to make these comparisons. You also need to know what types of impacts are affecting your degraded sites.

It depends on various parameters like soil characteristics. Therefore statistical analysis may be somehow inefficient unless you count on all aspects. other approaches like ANN may be more helpful. However, If you could find the spectrum of gradual decrease of pollutants from the riparian zone to the stream line (Latitude not longitude), it may show some results.

You might consider RDA and variance partitioning (VARPART) as describes in Borcard (2011).  

Borcard, Daniel, François Gillet, and Pierre Legendre. Numerical ecology with R. Springer, 2011.

I would also consider non-metric multidimensional scaling with the envfit function that is part of the R package VEGAN.  Other possibilities are ADONIS and MANTEL tests.

See the VEGAN  tutorial 

http://cc.oulu.fi/~jarioksa/opetus/metodi/vegantutor.pdf

Thank you all. Your answers are very wonderful. In the real field situation, these days it is hard to get reference situation. However we have designed the study in such a way that at least two sampling sites are located at similar habitat condition but different vegetation type and similarly at least two sampling sites are located at similar land use type. As has been mention, other parameters like soil type and other factors could affect. The suggested statistical methods could tried, but the limitation with ANN is size of dataset. However we will see ANN and VARPART.

Hi! I don't know much about statistics, but I can suggest some publications that have been carried out in pampean streams (rich in macrophytes diversity), Argentina. I hope they help! Regards

Thank you so much Ana, We will go through the articles.

Hi

There are other factors that could influence the diversity of macroinvertebrates such as heterogeneity of fluvial habitat (for instance presence of several microhabitats). This could be another variable that can help to explain differences in diversity between sampling points. Good luck

Hello again. What Alvaro said is so true. In fact, I've found another publication that establishes the influence on habitat characteristics on water and biota quality. I hope it to be usefull too. Kind regards

Thank you once again, its useful.

Hello Prof. Argaw. Your research very interesting! :)

I suggest you to read Limnology written by Goldman and Horne, and Limnology by Wetzel and about River Continuum Concept by Vannote. I hope it can strengthen your research. As far as i know, if we talk about macroinvertebartes community in the river, it is very influenced by organic matter that can be coarse (CPOM) or fine particulate (FPOM) depends on site characteristic and ecological phenomena like forest fire. Im sorry i cannot give some suggestion about statistical analysis. 

In fact, to establish relationships between biological quality elements is a complex subject , and is dependent from the scale of the study as well as from the type of the aquatic ecosystem (biogeographical location).  The partition of the variance could be a possible way to evaluate the contribution of each drivers (water  quality and riparian vegetation) on macroinvertebrate communities. However to do it, you had to evaluate for each driver a set of relevant descriptors, because otherwise the effect of each driver can be under evaluated. Another possibility is the effect of riparian vegetation as ecosystem engineering, increasing mesohabita diversity, conducting to the higher diversity values, even under a lower water quality. to have ana idea of possible relationships among biological quality elements I can provide one of my papers

Have you considered using structural equation modeling for estimating causal relationships? 

No, I didn't try structural equation modeling. May be that could help to identify causal relationship.

Could you try  to applied NMDS for statistic analysis ?