Microorganisms can mediate a wide range of chemical transformations with minimal inputs and are widely used for water treatment. Harnessing microorganisms in this way however requires not only that the microorganisms have the appropriate metabolic capabilities, but also that the conditions allow expression of those capabilities and, for reliable and sustained operation, that the desired microorganisms are competitive in the ecology of the system. Reliability engineering seeks to model complex systems and inform decisions related to improving reliability; for example, by identifying the steps which are most likely to fail and bolstering the system design in those areas. We seek to apply reliability engineering techniques to investigate the microbial community dynamics in an engineered ecosystem and thereby improve support for design and operation of such systems.
Our model system is denitrifying biofilters, which remove nitrate from subsurface agricultural drainage and release it as nitrogen gas. These biofilters consist of a trench filled with woodchips and two flow control structures installed at the outlet of subsurface tile drain systems in agricultural fields. From a microbial perspective, these biofilters slow the flow and provide appropriate conditions for denitrification (available carbon substrate and low oxygen). In both field-scale and laboratory-scale biofilters, we monitor relevant environmental parameters such as temperature, nitrate, and dissolved oxygen and characterize the bacterial, fungal, and denitrifying microbial communities using DNA fingerprinting techniques.
Results/Conclusions
The denitrifying biofilters contain a complex microbial community which is structured by depth and shows a bi-annual pattern. Applying reliability engineering, we can define the mean time to failure for a given population as the average time interval before the population falls below detection. For example, at one location in a denitrifying biofilter, 375 individual microbial populations were detected in a temporal dataset covering 430 days. The persistence of individual populations varied from being detected on only one date to being detected in every sample. For approximately 20% of the populations, there was only one continuous period of detection, resulting in only one calculated time to failure, while other populations appeared and then fell below detection up to nine times during the study period. Our goal is to adapt reliability metrics such as mean time to failure for analysis of microbial population dynamics, to relate these metrics and environmental factors to performance, and through this approach to identify new metrics and approaches for designing and predicting reliability of denitrifying biofilters and other engineered ecosystems.