Predicting hydrologic function with the streamwater mircobiome
Abstract
Recent advances in microbiology allow for rapid and cost-effective determination of the presence of a nearly limitless number of bacterial (and other) species within a water sample. Here, we posit that the quasi-unique taxonomic composition of the aquatic microbiome is an emergent property of a catchment that contains information about hydrologic function at multiple temporal and spatial scales, and term this approach `genohydrolgy.' As first a genohydrology case study, we show that the relative abundance of bacterial species within different operational taxonomic units (OTUs) from six large arctic rivers can be used to predict river discharge at monthly and longer timescales. Using only OTU abundance information and a machine-learning algorithm trained on OTU and discharge data from the other five rivers, our genohydrology approach is able to predict mean monthly discharge values throughout the year with an average Nash-Sutcliffe efficiency (NSE) of 0.50, while the recurrence interval of extreme flows at longer times scales in these rivers was predicted with an NSE of 0.04. This approach demonstrates considerable improvement over prediction of these quantities in each river based only on discharge data from the other five (our null hypothesis), which had average NSE values of -1.19 and -5.50 for the seasonal and recurrence interval discharge values, respectively. Overall the genohydrology approach demonstrates that bacterial diversity within the aquatic microbiome is a large and underutilized data resource with benefits for prediction of hydrologic function.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2017
- Bibcode:
- 2017AGUFM.H21H1581G
- Keywords:
-
- 1816 Estimation and forecasting;
- HYDROLOGY;
- 1869 Stochastic hydrology;
- HYDROLOGY;
- 1873 Uncertainty assessment;
- HYDROLOGY;
- 1942 Machine learning;
- INFORMATICS