Abstract
Fluvial particulate organic carbon (POC) is a complex mixture that undergoes rapid and complicated shifts in source during storm events. High-temporal resolution sampling and source-sensitive chemical analyses, such as those for organic geochemical biomarkers, are necessary to investigate the dynamic POC source behaviour during storm events. However, experimental designs that accommodate those requirements inevitably yield large datasets that require a new data analysis approach. Here, we adapt one of the widely used data visualization techniques, heatmaps with clustering analysis, to seek patterns in source mobilization and transition and pinpoint their timing during storm events more effectively and intuitively. Biomarker concentration data are scaled and used to construct a biomarker heatmap using the ComplexHeatmap package in R. Hierarchical clustering is performed to reorder the biomarkers based on (dis)similarities in their concentration fluctuations during storm events. We implemented our approach to visualize our high-frequency biomarker data obtained from storm POC samples collected in the well-characterized field site of Clear Creek, Iowa. The results demonstrated clear sequential source changes from algal and microbial OC to vascular plants- and soil-rich OC during the event, with an additional source transition identified within the vascular plant biomarkers. The sensitivity analyses results showed that the additional source transition was lost as the temporal resolution of sampling was reduced to 25% of the original data. The sensitivity of the identified clustering to varying scaling methods and number of biomarkers was also examined. Comparison with principal component analysis (PCA) showed that the biomarker heatmap performed better in visualizing temporal changes of individual biomarkers. This biomarker heatmap approach will help scientists to understand the complex storm-induced POC source changes by offering a new perspective to explore the data and generate hypotheses to be tested in follow-up analyses.