A critical question in microbial ecology concerns how environmental conditions affect community makeup. Arctic thermal springs enable study of this question due to steep environmental gradients that impose strong selective pressures. I use microscopic and molecular methods to quantify community makeup at Troll Springs on Svalbard in the high arctic. Troll has two ecosystems, aquatic and terrestrial, in proximity, shaped by different environmental factors. Microorganisms exist in warm water as periphyton, in moist granular materials, and in cold, dry rock as endoliths. Environmental conditions modulate community composition. The strongest relationships of environmental parameters to composition are pH and temperature in aquatic samples, and water content in terrestrial samples. Periphyton becomes trapped by calcite precipitation, and is a precursor for endolithic communities.

Microbial succession takes place at Troll in response to incremental environmental disturbances. Photosynthetic organisms are dominantly eukaryotic algae in the wet, high-illumination environments, and Cyanobacteria in the drier, lower-illumination endolithic environments. Periphyton communities vary strongly from pool to pool, with a few dominant taxa. Endolithic communities are more even, with bacterial taxa and cyanobacterial diversity similar to alpine and other Arctic endoliths. Richness and evenness increase with successional age, except in the most mature endolith where they diminish because of sharply reduced resource and niche availability. Evenness is limited in calcite-poor environments by competition with photosynthetic eukaryotes, and in the driest endolith by competition for water. Richness is influenced by availability of physical niches, increasing as calcite grain surfaces become available for colonization, and then decreasing as pore volume decreases.

In most endoliths, rock predates microbial colonization; the reverse is true at Troll. The harsh Arctic environment likely imposes a lifestyle in which microbes survive best in embedded formats, and to preserve live inocula for regrowth.

ARISA is commonly used to assess variations in microbial community structure. Applying a uniform threshold across a sample set, as is normally done, treats samples non-optimally and unequally. I present an algorithm for optimal threshold selection that maximizes similarity between replicate pairs, improving results.