ABSTRACT. – Arid regions lack models predicting shifts in biological and environmental conditions, thus precluding the implementation of efficient water policies. The aim of this study was to evaluate how macroinvertebrate communities respond to habitat and chemical conditions in arid-land streams of North Eastern Algeria, and how these communities are affected by human activities. We used Self-Organizing Maps (SOMs) to classify the sites of the Belezma National Park according to environmental conditions (including habitat and chemical variables), thus generating a template against which biotic conditions could be matched. The macroinvertebrate inventory included 93 taxa covering over 4 phyla, 8 classes, 22 orders and 58 families. Overall, coarse substrates were the most taxon-rich habitats (48 taxa). Wadi Chaaba, located in the integral zone of the park, hosted the most diversified fauna (40 taxa). The SOM delineated 3 clusters of stations (A, B and C), which combined environmental characteristics (altitude, flow velocity, physicochemical descriptors of water, type and substrate surfaces, human pressure) and characteristic benthic macroinvertebrate communities. The distribution patterns of the benthic macroinvertebrates were examined with reference to their pollution-sensitive or pollution-resistant status, and to the associated environmental variables. Wadi Chaaba, which is located in the protected area of the Park, was the healthiest stream, in that it hosted the most pollution-sensitive, indicator species.