Spencer Blackwood, Yale University

Class Year:

Mentor:

Abstract:

The Arctic is warming two to three times more rapidly than the global average, causing benthic communities in the region to undergo rapid and pronounced ecological changes. Researchers have compiled time series of benthic biological, biogeochemical, and physical data in the Bering and Chukchi Seas which document northward shifts among benthic species. However, Arctic data collection is expensive. Our work has two objectives: (a) to minimize costs of future research by identifying redundant information reported among research stations; (b) to augment existing knowledge of how Arctic benthic communities are responding to rapidly changing conditions. Conventional Euclidean distance metrics for comparing time series can only provide simultaneous comparisons; however, Arctic research stations may report similar information at a time lag due to changing conditions. In this study, we applied dynamic time warping (DTW) to these time series of Arctic water column and benthic community data, realigning their time indices to highlight similarities in their shapes. A cluster analysis was then performed using the distance values output by DTW for each time series comparison. For each variable under study, we produced two clusterings using one globular and one density-based clustering algorithm. The time series of water column variables were largely clustered by region, highlighting the existence of similar spatial dynamics within, but not across, regions. For each of the sedimentary and benthic community variables, the stations were grouped into one large cluster with the exception of one or two extreme observations. Removing these outliers could improve the clustering results, rendering them more informative. Further research which integrates a holistic approach (e.g. multidimensional DTW) with a region-specific analysis (e.g. formal tests for trend synchrony) will be necessary to conclusively identify sources of redundancy among the stations of study.