Reconstructing Subsurface Ocean Temperature and Salinity by the Reduced Space Objective Analysis With Dynamical Constraints

Subsurface ocean data are very sparse, especially in the pre-ARGO period. Yet, subsurface ocean temperature and salinity variations are crucial for our understanding of the wide range of climate phenomena: ocean response to the atmospheric increase in greenhouse gases, sea level rise, subsurface dynamics of ENSO events, and changes in the ocean circulation. Existing estimates of major integral characteristics of subsurface ocean, like upper ocean heat content of the top 700 m from 1950s to present, produced by different authors disagree with each other even in their annually-averaged global means. In this work the reduced space approach is used as the main methodological framework for extracting the large-scale climate variability from sparse and erroneous observations and for producing verifiable uncertainty estimates. Monthly sets of temperature and salinity profiles assembled in the U.K. Met Office Hadley Centre EN4 and in the U.S. National Oceanographic Data Center World Ocean Database have been binned separately since 1900s by grid cells of a three-dimensional spatial grid. Resulting binned data sets and reduced space analyses have been inter-compared and compared to other interpolated data sets at the variety of spatial resolutions. The effect of performing vertical interpolation of individual profiles prior to binning and the dependence of the results on the interpolation algorithms (traditional Reiniger-Ross weighted parabolas method, linear, cubic, and rational splines) were investigated and given statistical interpretation. Surface analysis results and their uncertainty estimates were verified by comparison with high-resolution satellite data sets. The analysis methodology was also augmented by switching to potential density as a vertical coordinate and by including the approximate conservation of the Bernoulli function as a weak dynamical constraint.