The surface ocean productivity and bottom water oxygen signals in deep water benthic foraminiferal assemblages

A previous study of the relative abundances of benthic foraminifera in the deep Pacific revealed that surface ocean productivity and bottom water oxygen signals were strongly imbedded in the foraminiferal data. The present report examines the nature of the species associations that carry the productivity and bottom water oxygen concentration signals. Principal components analysis is used to define species associations independently of the environmental variables. Principal components analysis of the correlation and the covariance matrices of species percents, and of taxon accumulation rate indices, was undertaken. The correlation analysis gives equal weight to all taxa while the covariance analysis weights results to the more abundant species. The accumulation rate index analysis reduces the distorting effects of calculating percents, and matrix closure, on the taxon abundance patterns. Deep water benthic foraminiferal assemblages from the Eastern Pacific Ocean show species distribution patterns dominated by a response to surface ocean productivity. Principal components analysis of either relative abundances or accumulation rate indices reveals a consistent set of species bearing the productivity signal. These include species normally found at depth (e.g. hispid Uvigerina) and taxa more commonly found on continental margins which are associated with infaunal microhabitats. The principal components analysis of both relative abundance and accumulation rate indices also shows that there are no taxa which are exclusively responsive to changing bottom water oxygen concentration. Rather, this factor acts in concert with other environmental variables (productivity in this study) to control benthic assemblage abundance patterns. Using relative abundance data convolves the oxygen concentration signal with the productivity response of the benthic assemblage. However, experimentation with absolute abundance indices suggests that the oxygen signal can be extracted from the assemblage data. This result is consistent with the previous analysis (Loubere, 1994) which yielded an r(2) of 0.96 for regression of assemblages against oxygen.