Search

In situ observations are vital to improving our understanding of the variability and dynamics of the ocean. A critical component of the ocean circulation is the strong, narrow, and highly variable western boundary currents. Ocean moorings that extend from the seafloor to the surface remain the most effective and efficient method to fully observe these currents. For various reasons, mooring instruments may not provide continuous records. Here we assess the application of the Iterative Completion Self-Organizing Maps (ITCOMPSOM) machine learning technique to fill observational data gaps in a 7.5 yr time series of the East Australian Current. The method was validated by withholding parts of fully known profiles, and reconstructing them. For 20% random withholding of known velocity data, validation statistics of the u- and υ-velocity components are R2 coefficients of 0.70 and 0.88 and root-mean-square errors of 0.038 and 0.064 m s−1, respectively. Withholding 100 days of known velocity profiles over a depth range between 60 and 700 m has mean profile residual differences between true and predicted u and υ velocity of 0.009 and 0.02 m s−1, respectively. The ITCOMPSOM also reproduces the known velocity variability. For 20% withholding of salinity and temperature data, root-mean-square errors of 0.04 and 0.38°C, respectively, are obtained. The ITCOMPSOM validation statistics are significantly better than those obtained when standard data filling methods are used. We suggest that machine learning techniques can be an appropriate method to fill missing data and enable production of observational-derived data products. Significance Statement: Moored observational time series of ocean boundary currents monitor the full-depth variability and change of these dynamic currents and are used to understand their influence on large-scale ocean climate, regional shelf–coastal processes, extreme weather, and seasonal climate. In this study we apply a machine learning technique, Iterative Completion Self-Organizing Maps (ITCOMPSOM), to fill data gaps in a boundary current moored observational data record. The ITCOMPSOM provides an improved method to fill data gaps in the mooring record and if applied to other observational data records may improve the reconstruction of missing data. The derived gridded data product should improve the accessibility and potentially increase the use of these data. [ABSTRACT FROM AUTHOR]

The Earth system is accumulating energy due to human-induced activities. More than 90% of this energy has been stored in the ocean as heat since 1970, with ∼60% of that in the upper 700 m. Differences in upper-ocean heat content anomaly (OHCA) estimates, however, exist. Here, we use a dataset protocol for 1970–2008—with six instrumental bias adjustments applied to expendable bathythermograph (XBT) data, and mapped by six research groups—to evaluate the spatiotemporal spread in upper OHCA estimates arising from two choices: 1) those arising from instrumental bias adjustments and 2) those arising from mathematical (i.e., mapping) techniques to interpolate and extrapolate data in space and time. We also examined the effect of a common ocean mask, which reveals that exclusion of shallow seas can reduce global OHCA estimates up to 13%. Spread due to mapping method is largest in the Indian Ocean and in the eddy-rich and frontal regions of all basins. Spread due to XBT bias adjustment is largest in the Pacific Ocean within 30°N–30°S. In both mapping and XBT cases, spread is higher for 1990–2004. Statistically different trends among mapping methods are found not only in the poorly observed Southern Ocean but also in the well-observed northwest Atlantic. Our results cannot determine the best mapping or bias adjustment schemes, but they identify where important sensitivities exist, and thus where further understanding will help to refine OHCA estimates. These results highlight the need for further coordinated OHCA studies to evaluate the performance of existing mapping methods along with comprehensive assessment of uncertainty estimates. [ABSTRACT FROM AUTHOR]

Expendable bathythermographs (XBTs) have been widely deployed for ocean monitoring since the late 1960s. Improving the quality of XBT data is a vital task in climatology. Many factors (e.g., temperature, probe type, and manufacturing time) have been identified as major influences of XBT systematic bias. In addition, the recording system (RS) has long been suspected as another factor. However, this factor has not been taken into account in any global XBT correction schemes, partly because its impact is poorly understood. Here, based on analysis of an XBT–CTD side-by-side dataset and a global collocated reference dataset, the influence of RSs on the pure temperature error (PTE) is examined. Results show a clear time dependency of PTE on the RS, with maximum values occurring in the 1970s. In addition, the method used to convert thermistor resistance into temperature in the RS (using a resistance–temperature equation) has changed over time. These changes, together with the decadal changes in RSs, might contribute a small error (10% on average) to the RS dependency. Here, an improvement of global XBT bias correction that accounts for the RS dependency is proposed. However, more than 70% of historical global XBT data are missing RS-type information. We investigate several assumptions about the temporal distribution of RS types, and all scenarios lead to at least a ~50% reduction in the time variation of PTE compared with the uncorrected data. Therefore, the RS dependency should be taken into account in updated XBT correction schemes, which would have further implications for climatology studies. [ABSTRACT FROM AUTHOR]

Time-varying biases in expendable bathythermograph (XBT) instruments have emerged as a key uncertainty in estimates of historical ocean heat content variability and change. One of the challenges in the development of XBT bias corrections is the lack of metadata in ocean profile databases. Approximately 50% of XBT profiles in the World Ocean database(WOD) have no information about manufacturer or probe type. Building on previous research efforts, this paper presents a deterministic algorithm for assigning missing XBT manufacturer and probe type for individual temperature profiles based on 1) the reporting country, 2) the maximum reported depth, and 3) the record date. The criteria used are based on bulk analysis of known XBT profiles in the WOD for the period 1966-2015. A basic skill assessment demonstrates a 77% success rate at correctly assigning manufacturer and probe type for profiles where this information is available. The skill rate is lowest during the early 1990s, which is also a period when metadata information is particularly poor. The results suggest that substantive improvements could be made through further data analysis and that future algorithms may benefit from including a larger number of predictor variables. [ABSTRACT FROM AUTHOR]

A very large portion of the historical information on ocean temperatures has been measured using expendable bathythermograph (XBT) devices. For decades, these devices provided the majority of global information. It is, therefore, important to quantify their accuracy and identify biases in this important historical dataset. Here, calculations are made of the influence of water temperature on the rate of descent of the XBT devices into the ocean waters. In colder regions, the larger viscosity of the water is expected to cause a greater drag force on the device, which would slow the descent. It was found through computational fluid dynamic models that the impact of temperature and viscosity on the probe descent is approximately 2.2% for water temperatures that range from 0° to 27°C. Probe-specific temperature-dependent fall rate equations were applied to 269 collocated XBT/conductivity-temperature-depth (CTD) measurements from two different research cruises. It was found that the probe-specific descent equations were an improvement over the uncorrected method. Next, in an effort to automate the process, the fall rate coefficients were related to the topmost measured temperature in the water column. With this relationship, comparisons were made between the probe-specific descent calculations and 2937 high-resolution XBT-CTD pairs. It was found that again, the new methodology outperformed the standard fall rate equation. The new method was also compared with an independent correction method that was previously published. It was found that both new methods were improvements upon the industry-standard fall rate calculation. Subsequent calculations using the top-100-m water temperature were performed and were found to be statistically insignificant compared to the proposed simplified method. [ABSTRACT FROM AUTHOR]

Systematic biases in historical expendable bathythermograph (XBT) data are examined using two datasets: 4151 XBT-CTD side-by-side pairs from 1967 to 2011 and 218 653 global-scale XBT-CTD pairs (within one month and 1°) extracted from the World Ocean Database 2009 ( WOD09) from 1966 to 2010. Using the side-by-side dataset, it was found that both the pure thermal bias and the XBT fall rate (from which the depth of observation is calculated) increase with water temperature. Correlations between the terminal velocity A and deceleration B terms of the fall-rate equation (FRE) and between A and the offset from the surface terms are obtained, with A as the dominant term in XBT fall-rate behavior. To quantify the time variation of the XBT fall-rate and pure temperature biases, global-scale XBT-CTD pairs are used. Based on the results from the two datasets, a new correction scheme for historical XBT data is proposed for nine independent probe-type groups. The scheme includes corrections for both temperature and depth records, which are all variable with calendar year, water temperature, and probe type. The results confirm those found in previous studies: a slowing in fall rate during the 1970s and 2000s and the large pure thermal biases during 1970-85. The performance of nine different correction schemes is compared. After the proposed corrections are applied to the XBT data in the WOD09 dataset, global ocean heat content from 1967 to 2010 is reestimated. [ABSTRACT FROM AUTHOR]