• Abraham, J. P., J. Gorman, F. Reseghetti, K. Trenberth, and W. Minkowycz, 2011: A new method of calculating ocean temperatures using expendable bathythermographs. Energy Environ. Res., 1, 211, https://doi.org/10.5539/eer.v1n1p2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Abraham, J. P., J. M. Gorman, F. Reseghetti, E. M. Sparrow, and W. J. Minkowycz, 2012: Drag coefficients for rotating expendable bathythermographs and the impact of launch parameters on depth predictions. Numer. Heat Transfer, 62A, 2543, https://doi.org/10.1080/10407782.2012.672898.

    • Search Google Scholar
    • Export Citation
  • Abraham, J. P., and et al. , 2013: A review of global ocean temperature observations: Implications for ocean heat content estimates and climate change. Rev. Geophys., 51, 450483, https://doi.org/10.1002/rog.20022.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Abraham, J. P., R. Cowley, and L. Cheng, 2016: Quantification of the effect of water temperature on the fall rate of expendable bathythermographs. J. Atmos. Oceanic Technol., 33, 12711283, https://doi.org/10.1175/JTECH-D-15-0216.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Anderson, A. L., and et al. , 1979: Bearing stake acoustic assessment (U). NOSC Tech. Rep. TR 466, 279 pp.

    • Crossref
    • Export Citation
  • Anderson, E. R., 1980: Expendable bathythermograph (XBT) accuracy studies. Naval Ocean Systems Center Tech. Rep. 550, 201 pp.

    • Crossref
    • Export Citation
  • Argo, 2000: Argo float data and metadata from Global Data Assembly Centre (Argo GDAC). SEANOE, accessed 28 September 2018, http://doi.org/10.17882/42182.

    • Crossref
    • Export Citation
  • Bailey, R., H. Phillips, and G. Meyers, 1989: Relevance to TOGA of systematic XBT errors. Proceedings of the Western Pacific International Meeting and Workshop on TOGA COARE, J. Picaut, R. Lukas, and T. Delcroix, Eds., ORSTOM, 775–784.

  • Bailey, R., A. Gronell, H. Phillips, E. Tanner, and G. Meyers, 1994: Quality control cookbook for XBT data. CSIRO Marine Laboratories Rep. 221, 83 pp.

  • Beatty, W. H., III, J. G. Bruce, and R. C. Guthrie, 1981: Circulation and oceanographic properties in the Somali Basin as observed during the 1979 southwest monsoon. Naval Oceanographic Office Tech. Rep. TR-258, 78 pp.

    • Crossref
    • Export Citation
  • Boyd, J. D., and R. S. Linzell, 1993: The temperature and depth accuracy of Sippican T-5 XBTs. J. Atmos. Oceanic Technol., 10, 128136, https://doi.org/10.1175/1520-0426(1993)010<0128:TTADAO>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Boyer, T., and et al. , 2016: Sensitivity of global upper-ocean heat content estimates to mapping methods, XBT bias corrections, and baseline climatologies. J. Climate, 29, 48174842, https://doi.org/10.1175/JCLI-D-15-0801.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Bringas, F., and G. Goni, 2015: Early dynamics of Deep Blue XBT probes. J. Atmos. Oceanic Technol., 32, 22532263, https://doi.org/10.1175/JTECH-D-15-0048.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Budéus, G., and G. Krause, 1993: On-cruise calibration of XBT probes. Deep-Sea Res. I, 40, 13591363, https://doi.org/10.1016/0967-0637(93)90116-K.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Cheng, L., J. Zhu, F. Reseghetti, and Q. P. Liu, 2011: A new method to estimate the systematical biases of expendable bathythermograph. J. Atmos. Oceanic Technol., 28, 244265, https://doi.org/10.1175/2010JTECHO759.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Cheng, L., J. Zhu, R. Cowley, T. Boyer, and S. Wijffels, 2014: Time, probe type, and temperature variable bias corrections to historical expendable bathythermograph observations. J. Atmos. Oceanic Technol., 31, 17931825, https://doi.org/10.1175/JTECH-D-13-00197.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Cheng, L., and et al. , 2016: XBT Science: Assessment of instrumental biases and errors. Bull. Amer. Meteor. Soc., 97, 924933, https://doi.org/10.1175/BAMS-D-15-00031.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Cheng, L., H. Luo, T. Boyer, R. Cowley, J. Abraham, V. Gouretski, F. Reseghetti, and J. Zhu, 2018: How well can we correct systematic errors in historical XBT data? J. Atmos. Oceanic Technol., 35, 11031125, https://doi.org/10.1175/JTECH-D-17-0122.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Cowley, R., S. Wijffels, L. Cheng, T. Boyer, and S. Kizu, 2013: Biases in expendable bathythermograph data: A new view based on historical side-by-side comparisons. J. Atmos. Oceanic Technol., 30, 11951225, https://doi.org/10.1175/JTECH-D-12-00127.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Cunningham, S. A., and et al. , 2000: RRS Discovery cruise 242, 07 Sep–06 Oct 1999. Atlantic–Norwegian Exchanges. Southampton Oceanography Centre Cruise Rep. 28, 128 pp.

  • Dammann, P., 1982: Acoustic tracking of expendable bathythermographs. J. Acoust. Soc. Amer., 72, S38, https://doi.org/10.1121/1.2019863.

  • Daubin, S. C., 1973: Church Gabbro technical note: Systems description and performance. University of Miami RSMAS Tech. Rep. AD-763460, 160 pp.

  • Demeo, R. P., 1969: The validity of expendable bathythermograph measurements. The Decade Ahead, 1970–1980, Marine Technology Society, 155–179.

  • Denner, W., 1969: Separation of the residual instrument noise from the significant variability for expendable bathythermographs. Proceedings of the Fourth National ISA Marine Sciences Instrumentation Symposium, F. Alt, Ed., Marine Sciences Instrumentation, Vol. 4, Plenum Press, 635–641.

  • Di Nezio, P. N., and G. Goni, 2011: Direct evidence of a changing fall-rate bias in XBTs manufactured during 1986–2008. J. Atmos. Oceanic Technol., 28, 15691578, https://doi.org/10.1175/JTECH-D-11-00017.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Emery, W. J., W. Lee, W. Zenk, and J. Meincke, 1986: A low-cost digital XBT system and its application to the real-time computation of dynamic height. J. Atmos. Oceanic Technol., 3, 7583, https://doi.org/10.1175/1520-0426(1986)003<0075:ALCDXS>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Fahrbach, E., W. Krauss, J. Meincke, and A. Sy, 1984: Nordostatlantik ’83. Christian-Albrechts-Universität Institut Für Meereskunde Data Rep. 134, 65 pp.

  • Fenner, D. F., and W. J. Cronin Jr., 1978: Bearing stake exercise: Sound speed and other environmental variability (U). Naval Oceanographic Laboratory Ocean Acoustics Division Rep. NORDA-18, 73 pp.

  • Fenner, D. F., K. W. Lackie, B. A. Watrous, and L. A. Banchero, 1974: IOMEDEX sound velocity analysis and environmental data summary. NOO Tech. Rep. TR-244, 67 pp.

  • FICARAM Group, 2013: FICARAM-15 cruise report 20th March–22nd May 2013 on board BIO Hespérides. 54 pp.

  • Flierl, G., and A. Robinson, 1974: XBT-CTD intercomparison. Instrument description and intercomparison: Report of the MODE-I Intercomparison Group, 137–139.

  • Flierl, G., and A. Robinson, 1977: XBT measurements of thermal gradients in the MODE eddy. J. Phys. Oceanogr., 7, 300302, https://doi.org/10.1175/1520-0485(1977)007<0300:XMOTGI>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Fofonoff, N. P., and R. C. Millard Jr., 1983: Algorithms for computation of fundamental properties of seawater. UNESCO Technical Papers in Marine Science 44, 54 pp., http://unesdoc.unesco.org/images/0005/000598/059832EB.pdf.

  • Fonteles, C. S., and M. M. Mata, 2009: Estimativas de erros em medidas de XBT (expendable bathythermograph) no Atlântico Sudoeste. Universidade Federal do Rio Grande VIII Mostra de Produção Universitária, 3 pp.

  • Francis, S. A., and G. C. Campbell, 1965: A low cost expendable bathythermograph. Proceedings of the Third National Marine Sciences Symposium, W. C. Knopf and H. A. Cook, Eds., Marine Sciences Instrumentation, Vol. 3, Plenum Press, 85–89.

  • Georgi, D. T., J. P. Dean, and J. A. Chase, 1979: XBT probe-to-probe thermistor temperature variability. POLYMODE News, No. 71, WHOI, Woods Hole, MA, 1, 7–10.

  • Georgi, D. T., J. P. Dean, and J. A. Chase, 1980: Temperature calibration of expendable bathythermographs. Ocean Eng., 7, 491499, https://doi.org/10.1016/0029-8018(80)90048-7.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Goni, G. J., and et al. , 2010: The Ship of Opportunity Program. Proceedings of the OceanObs’09: Sustained Ocean Observations and Information for Society, J. Hall, D. E. Harrison, and D. Stammer, Eds., Vol. 2, ESA Publ. WPP-306, https://doi.org/10.5270/OceanObs09.cwp.35.

    • Crossref
    • Export Citation
  • Good, S. A., 2011: Depth biases in XBT data diagnosed using bathymetry data. J. Atmos. Oceanic Technol., 28, 287300, https://doi.org/10.1175/2010JTECHO773.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Gould, W. J., 1991: RRS Charles Darwin cruise 50, 29 June–22 July 1990; Oceanography of the Iceland Basin Oceanography of the Iceland Basin: The fate of Iceland and Scotland overflow water. IOS Deacon Laboratory Cruise Rep. 221, 41 pp.

  • Gould, W. J., R. J. Bailey, and M. Szabados, 1990: Errors in XBT probes. International WOCE Newsletter, No. 10, WOCE International Project Office, Southampton, United Kingdom, 10–11.

  • Gouretski, V., and K. P. Koltermann, 2007: How much is the ocean really warming? Geophys. Res. Lett., 34, L01610, https://doi.org/10.1029/2006GL027834.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Gouretski, V., F. Reseghetti, S. Kizu, S. Wijffels, G. Goni, P. DiNezio, and J. Trinanes, 2010: XBT Bias and Fall Rate Workshop: Summary report. University of Hamburg, KlimaCampus, 14 pp., https://icdc.cen.uni-hamburg.de/fileadmin/user_upload/icdc_Dokumente/xbt_ws_presentations/xbt_workshop_summary_report_final.pdf.

  • Gouretski, V., and F. Reseghetti, 2010: On depth and temperature biases in bathythermograph data: Development of a new correction scheme based on analysis of a global ocean database. Deep-Sea Res. I, 57, 812833, https://doi.org/10.1016/j.dsr.2010.03.011.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Green, A. W., 1984: Bulk dynamics of the expendable bathythermograph (XBT). Deep-Sea Res., 31A, 415426, https://doi.org/10.1016/0198-0149(84)90093-1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Grilli, F., and N. Pinardi, 1998: The computation of Rossby radii of deformation for the Mediterranean Sea. MTP News, Vol. 6, No. 4, University of Barcelona, Barcelona, Spain, 4–5.

  • Grilli, F., and N. Pinardi, 1999: Le cause dinamiche della stratificazione verticale nel Mediterraneo. ISAO Tech. Rep. ISAO-TR-3/99, 132 pp.

  • Hallock, Z. R., and W. J. Teague, 1992: The fall-rate of the T7 XBT. J. Atmos. Oceanic Technol., 9, 470483, https://doi.org/10.1175/1520-0426(1992)009<0470:TFROTT>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Hamon, M., G. Reverdin, and P. Y. Le Traon, 2012: Empirical correction of XBT data. J. Atmos. Oceanic Technol., 29, 960973, https://doi.org/10.1175/JTECH-D-11-00129.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Hanawa, K., and H. Yoritaka, 1987: Detection of systematic errors in XBT data and their correction. J. Oceanogr. Soc. Japan, 43, 6876, https://doi.org/10.1007/BF02110635.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Hanawa, K., and Y. Yoshikawa, 1991: Re-examination of the depth error in XBT data. J. Atmos. Oceanic Technol., 8, 422429, https://doi.org/10.1175/1520-0426(1991)008<0422:ROTDEI>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Hanawa, K., P. Rual, R. Bailey, A. Sy, and M. Szabados, 1995: A new depth time equation for Sippican or TSK T-7, T-6 and T-4 expendable bathythermographs (XBT). Deep-Sea Res. I, 42, 14231451, https://doi.org/10.1016/0967-0637(95)97154-Z.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Heinmiller, R., C. Ebbesmeyer, B. Taft, T. Olson, and O. Nikitin, 1983: Systematic errors in expendable bathythermograph (XBT) profiles. J. Oceanogr., 65, 287299.

    • Search Google Scholar
    • Export Citation
  • Hill, P. J., 1995: Swath-mapping ADEDAV survey by RV L’Atalante from Adelaide to Davao along the continental margin of Western Australia, 1994: Post-cruise and ADEDAV/TRANSNOR data processing report. Australian Geological Survey Organization Record 1995/55, 40 pp.

  • IGOSS, 1972: Manual on data acquisition for IGOSS. UNESCO/IOC, 2nd Draft, 250 pp.

  • IOC, 1992: Ad hoc meeting of the IGOSS Task Team on quality control for automated systems, Marion, Massachusetts, USA, 3–6 June 1991. Intergovernmental Oceanographic Commission IOC/INF-888, 144 pp.

  • IOC, 1995: Third session of the Task Team on Quality Control Procedures for Automated Systems (TT/CAS), Ottawa, Canada, 23–25 October 1995. Intergovernmental Oceanographic Commission IOC/INF-1017, 34 pp.

  • IPCC, 2013: Climate Change 2013: The Physical Science Basis. Cambridge University Press, 1535 pp., https://doi.org/10.1017/CBO9781107415324.

    • Crossref
    • Export Citation
  • Ishii, M., and M. Kimoto, 2009: Reevaluation of historical ocean heat content variations with time-varying XBT and MBT depth bias corrections. J. Oceanogr., 65, 287299, https://doi.org/10.1007/s10872-009-0027-7.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • JCGM, 2008: Evaluation of measurement data—Guide to the expression of uncertainty in measurement. Joint Committee for Guides in Meteorology Working Group 1 Rep. 100:2008, 120 pp.

  • Johnson, W. P., and R. E. Lange, 1979: Rapid sampling of temperature and temperature gradient using XBT’s. SIO Reference Series 79-4, 39 pp.

  • Kennelly, M.A., M.D. Prater and T.B. Sanford, 1989: XBT and XSV data from the Gulf of Cadiz expedition: R/V Oceanus cruise 202. University of Washington APL Tech. Rep. APL-UW TR 8920, 217 pp.

    • Crossref
    • Export Citation
  • King, B. A., 1991: Circulation and structure of the Bay of Biscay and north east Atlantic out to 20°W and 41°N. RRS Discovery Cruise 189, 09 Mar.–08 Apr. 1990, Institute of Oceanographic Sciences Deacon Laboratory Cruise Rep. 225-1991, 45 pp.

  • Kizu, S., and K. Hanawa, 2002a: Start-up transient of XBT measurement. Deep-Sea Res. I, 49, 935940, https://doi.org/10.1016/S0967-0637(02)00003-1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Kizu, S., and K. Hanawa, 2002b: Recorder-dependent temperature error of expendable bathythermograph. J. Oceanogr., 58, 469476, https://doi.org/10.1023/A:1021261214950.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Kizu, S., H. Yoritaka, and K. Hanawa, 2005a: A new fall-rate equation for T-5 expendable bathythermograph (XBT) by TSK. J. Oceanogr., 61, 115121, https://doi.org/10.1007/s10872-005-0024-4.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Kizu, S., S. Ito, and T. Watanabe, 2005b: Inter-manufacturer difference and temperature dependency of the fall-rate of T-5 expendable bathythermograph. J. Oceanogr., 61, 905912, https://doi.org/10.1007/s10872-006-0008-z.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Kizu, S., C. Sukigara, and K. Hanawa, 2011: Comparison of the fall rate and structure of recent T-7 XBT manufactured by Sippican and TSK. Ocean Sci., 7, 231244, https://doi.org/10.5194/os-7-231-2011.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Koso, Y., H. Ishii, M. Fujita, and H. Kato, 2005: Application of the new depth conversion formula of XBT (T-5). Tech. Bull. Hydrogr. Oceanogr., 23, 8992.

    • Search Google Scholar
    • Export Citation
  • Levitus, S., J. I. Antonov, T. P. Boyer, R. A. Locarnini, H. E. Garcia, and A. V. Mishonov, 2009: Global ocean heat content 1955–2008 in light of recently revealed instrumentation problems. Geophys. Res. Lett., 36, L07608, https://doi.org/10.1029/2008GL037155.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Little, A. D., 1965: Experimental evaluation of expendable bathythermographs. Dept. of the Navy Bureau of Ships Rep. 4071165, 51 pp.

  • Little, A. D., 1966: Expendable bathythermograph (XBT) system evaluation for tactical sonar application. Dept. of the Navy Naval Ship Systems Command Rep. 4150866, 85 pp.

  • Magruder, P. M., Jr., 1970: Some characteristics of temperature microstructure in the ocean. M.S. thesis, Dept. of Oceanography, Naval Postgraduate School, 155 pp.

    • Crossref
    • Export Citation
  • McDowell, S., 1977: A note on XBT accuracy. POLYMODE News, No. 29, WHOI, Woods Hole, MA, 1, 4.

  • McDowell, S., 1978: A cautionary note on T-5 XBTS. POLYMODE News, No. 58, WHOI, Woods Hole, MA, 4.

  • Meincke, J., 1991: WHP cruise summary information of section A01E. WOCE, 76 pp.

  • Mied, R. P., G. J. Lindemann, and A. F. Schuetz, 1981: The hydrography and dynamics of the FREDDEX eddy. Naval Research Laboratory Memo. Rep. NRL-MR-4603, 35 pp.

  • Miura, T., M. Konda, T. Takikawa, and H. Ichikawa, 2004: Depth error in time-depth equations of the T5 XBT probes. JAMSTECR, 49, 7380.

    • Search Google Scholar
    • Export Citation
  • Newman, F. C., and P. Dammann, 1983: Acoustic measurement of XBT fall rates. SAI Tech. Rep. SAI-83/1220, 35 pp.

  • Paden, C. A., and M. C. Hendershott, 1986: Observations of temperature finestructure in the Gulf of California—XBT data report November 1984/March 1985. SIO Reference Series 86-14, 275 pp.

  • Paillet, J., 2001: Report de Campagne. Campagne POMME 1. Service Hydrographique et Océanographique de la Marine 065 MOA/NP, 20 pp.

  • Plessey, 1967: Plessey–Sippican: The only system that lets you make a bathythermograph plot at 30 knots. Int. Hydrogr. Rev., 44 (1), 1718.

    • Search Google Scholar
    • Export Citation
  • Raiteri G., A. Bordone, T. Ciuffardi, and F. Pennecchi, 2018: Uncertainty evaluation of CTD measurements: A metrological approach to water-column coastal parameters in the Gulf of La Spezia area. Measurement, 126, 156163, https://doi.org/10.1016/j.measurement.2018.05.058.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Reid, W. L., Jr., 1964: Expendable bathythermograph evaluation. U.S. Naval Oceanographic Office Informal Manuscript Rep. NOO-IM-I-1-64, 70 pp.

  • Reseghetti, F., M. Borghini, and G. M. R. Manzella, 2007: Factors affecting the quality of XBT data—Results of analyses on profiles from the Western Mediterranean Sea. Ocean Sci., 3, 5975, https://doi.org/10.5194/os-3-59-2007.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Reverdin, G., F. Marin, B. Bourles, and P. L’Herminier, 2009: XBT temperature errors during French research cruises (1999–2007). J. Atmos. Oceanic Technol., 26, 24622473, https://doi.org/10.1175/2009JTECHO655.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Ribeiro, N., M. M. Mata, J. L. L. de Azevedo, and M. Cirano, 2018: An assessment of the XBT fall-rate equation in the Southern Ocean. J. Atmos. Oceanic Technol., 35, 911926, https://doi.org/10.1175/JTECH-D-17-0086.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Ridgway, K., R. Bailey, and R. Coleman, 1999: Tasman-Coral Sea mass and heat transport/ satellite verification. National Facility Oceanographic Research Vessel, CSIRO Marine Research Cruise Summary RV Franklin FR 02/99, 18 pp.

  • Roemmich, D., and B. Cornuelle, 1987: Digitization and calibration of the expendable bathy thermograph. Deep-Sea Res., 34A, 299307, https://doi.org/10.1016/0198-0149(87)90088-4.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Russell, J. S., and K. M. Leavitt, 1984: Development of a 2000 meter aircraft expendable bathythermograph. Sippican Ocean Systems, Inc., Final Rep. R-1435B, 17 pp.

    • Crossref
    • Export Citation
  • Saur, J. F. T., and D. D. Stewart, 1967: Expendable bathythermograph data on subsurface thermal structure in the eastern North Pacific Ocean. U.S. Fish and Wildlife Service Special Scientific Rep.—Fisheries 548, 70 pp.

  • SBE, 2002: Conversion of pressure to depth. Sea-Bird Electronics Application Note 69, 1 pp.

  • Seaver, G. A., and S. Kuleshov, 1979: XBT accuracy. POLYMODE News, No. 72, WHOI, Woods Hole, MA, 1, 5–9.

  • Seaver, G. A., and S. Kuleshov, 1982: Experimental and analytical error of the expendable bathythermograph. J. Phys. Oceanogr., 12, 592600, https://doi.org/10.1175/1520-0485(1982)012<0592:EAAEOT>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Singer, J. J., 1990: On the error observed in electronically digitized T-7 XBT data. J. Atmos. Oceanic Technol., 7, 603611, https://doi.org/10.1175/1520-0426(1990)007<0603:OTEOIE>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Sippican Corporation, 1971: R-603G Instruction manual for the expendable bathythermograph. Sippican Corporation Doc.

  • Stark, J., J. Gorman, M. Hennessey, F. Reseghetti, J. Willis, J. Lyman, J. Abraham, and M. Borghini, 2011: A computational method for determining XBT depths. Ocean Sci., 7, 733743, https://doi.org/10.5194/os-7-733-2011.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Stegen, G. R., D. P. Delisi, and R. C. Van Colln, 1975: A portable, digital recording, expendable bathythermograph (XBT) system. Deep-Sea Res. Oceanogr. Abstr., 22, 447453, https://doi.org/10.1016/0011-7471(75)90067-4.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Sy, A., 1989: Summary about preliminary results from studies of depth fall rate errors of “Deep-Blue” probes. Integrated Global Ocean Services System (IGOSS) Summary of Ship-of-Opportunity Programmes and Technical Reports, Intergovernmental Oceanographic Commission IOC/INF-804, 185–192.

  • Sy, A., 1991: XBT measurements. WOCE Operations Manual, WHP Operation and Methods, WHP91-1, WOCE Rep. 68/91, 1–19.

  • Sy, A., 1992: Report on field tests in 1990 on the evaluation of the XBT depth fall rate equation. Ad hoc meeting of the IGOSS Task Team on Quality Control for Automated Systems, Marion, Massachusetts, USA, 3–6 June, 1991, Intergovernmental Oceanographic Commission IOC/INF-888, 121–130.

  • Sy, A., J. Ulrich, and M. Stolley, 2000: Status of ship-of-opportunity activities in Germany 1999. JCOMM Ship-of-Opportunity Programme Implementation Panel—Third Session, La Jolla, CA, USA, 28–31 March 2000: SOOP status reports, SOOP scientific and technical developments, April 2000, WMO/TD-1005, JCOMM Tech. Rep. 3, 50–55.

  • Szabados, M. W., 1992: Evaluation of the expendable bathythermographic (XBT) fall rate equation. Ad hoc meeting of the IGOSS Task Team on Quality Control for Automated Systems, Marion, Massachusetts, USA, 3–6 June, 1991, Intergovernmental Oceanographic Commission IOC/INF-888, 19–97.

  • Szabados, M. W., and D. Wright, 1989: Field evaluation of real-time XBT systems. Proceedings of the Western Pacific International Meeting and Workshop on TOGA COARE, J. Picaut, R. Lukas, and T. Delcroix, Eds., ORSTOM, 811–821.

  • Thadathil, P., A. K. Saran, V. V. Gopalakrishna, P. Vethamony, N. Araligidad, and R. Bailey, 2002: XBT fall rate in waters of extreme temperature: A case study in the Antarctic Ocean. J. Atmos. Oceanic Technol., 19, 391396, https://doi.org/10.1175/1520-0426-19.3.391.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • USNAVO, 1975: Instruction manual for obtaining oceanographic data. U.S. Naval Oceanographic Office Publ. 607, 3rd ed. 232 pp.

  • Volkmann, G., 1977: Four short XBT sections across Mid-Atlantic ridge. POLYMODE News, No. 33, WHOI, Woods Hole, MA, 3.

  • Wannamaker, B., 1980: XBT measurements near the sea surface: Considerations for satellite IR comparisons and data bases. Saclant ASW Research Centre Memo. SM-132, 13 pp.

  • Wijffels, S. E., J. Willis, C. Domingues, P. Barker, N. White, A. Gronell, K. Ridgway, and J. Church, 2008: Changing expendable bathythermograph fall rates and their impact on estimates of thermosteric sea level rise. J. Climate, 21, 56575672, https://doi.org/10.1175/2008JCLI2290.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Zanasca, P., 1994: On board XBTs calibration. NATO Undersea Research Centre Internal Notes, 17 pp.

  • View in gallery

    T5/20, T5, and DB XBT probes are shown from top to bottom. T5 and T5/20 differ only by the amount of wire on the canister spool and the junction between the canister spool and the probe spool. DB is shorter (no plastic cylinder between the zinc nose and the plastic afterbody), lighter (also for the amount of wire on the probe spool), and with a different amount of wire on the canister spool.

  • View in gallery

    (a). Number of profiles for nine major XBT probe type groups per year from 1966 to 2016 (T5 in black). Two groups for unknown probe type are also included (DX: Deep Unknown with maximum depth > 550 m and SX: Shallow Unknown with maximum depth < 550 m). (b) Percentage of T5 data in WOD13 at each depth from 1966 to 2016.

  • View in gallery

    The ∆T between the output of different Sippican RSs and the expected reading based on a set of high-accuracy reference resistors. The room temperature conditions were reasonably similar (23°–27°C).

  • View in gallery

    (a). Geographical distributions of T5–CTD pairs used in the present analysis. Different datasets are shown by different colors: GO_T5 (blue), JapT5 (cyan), CanT5 (light green), MedT5 (red), Brazil_T5 (purple), and WODT5 (olive). (b). Zoomed-in view of the Mediterranean Sea—MedT5 dataset: different colors identify different years when comparisons have been made. Stars indicated pairs from ArgoT5.

  • View in gallery

    FREC by using the FRE model of D = AtBt2 − Offset. (top) Coefficient A vs B and (bottom) coefficient A vs Offset shown with dots for each individual XBT–CTD pair: red (MedT5), blue (GO_T5), light green (canT5), cyan (JapT5), green (Brazil_T5), and gray (WODT5). The weighted mean of the FREC is shown (yellow star), compared with the T5FRE coefficient for T5 (yellow square). The linear regressions between A and B, and A and Offset are also shown with black lines.

  • View in gallery

    (a). The ∆D as a function of the probe weight in air for MedT5 probes. (b). The ∆D as a function of diameter of the hole in the zinc nose for MedT5. The linear regressions are shown with dashed lines.

  • View in gallery

    Offset vs h shown with dots: red (MedT5), blue (GO_T5), green (CanT5), cyan (JapT5), purple (Brazil_T5), and gray (WODT5). The mean (median) of Offset at each launching height is shown in green (dashed green), with two standard errors in black. A linear fit for the green curve is shown with a dashed orange line. Test results in the shallow water are shown as yellow triangles and rectangles.

  • View in gallery

    The speed term A as a function of 0–100-m averaged temperatures shown with dots: red (MedT5), blue (GO_T5), light green (CanT5), green (Brazil_T5), cyan (JapT5), and gray (WODT5). The black solid curve (and two standard error bar) is the median (two SD) of A at each 2.5°C temperature interval. The black dashed line is the linear regression for all data, while the orange dashed line is the regression for the dataset after the exclusion profiles from WODT5.

  • View in gallery

    The value of ∆D as a function of T: (a) 0–2500 m: mean ∆D vs averaged temperature for the whole profile: (b) 0–500 m: mean ∆D vs 0–500-m mean temperature; (c) as in (b), but for 500–1000 m; (d) as in (b), but for 1000–1500 m; (e) as in (b) but for 1500–2000 m. The mean ∆D at each 2.5°C temperature window is denoted by black curves, with two times standard error bars. In all panels, a linear regression between ∆D and T is shown with a dashed pink line. The different colors identify the different datasets as in Fig. 8.

  • View in gallery

    The ∆TXBT–CTD profile (a),(c) before and (b),(d) after individual ∆D correction for MedT5. Median of the ∆TXBT–CTD at each depth before correction (after correction) is also presented in black (green), and the SD is shown with a dashed line. The ∆TXBT–CTD profile at (top) 0–300 and (bottom) 300–2300 m.

  • View in gallery

    As in Fig. 10, but for all the other data except MedT5.

  • View in gallery

    PTU after individual depth correction as a function of depth. The linear regression for 10–2200 (blue) and 100–2000 m (red). For comparison, the pressure effect quoted in RC87 is shown (black).

  • View in gallery

    (a) Median and SD of PTU for T5 with (blue) or without (black) calibration in bath (MedT5 only). The median PTU after removing the pressure effect is shown in light blue and gray for T5 with or without calibration respectively. (b) Median and SD of PTU for T5 with (purple) or without (black) correction as a result of the use of Tester (MedT5 only). The median PTU after removing the pressure effect is shown in light purple (Tester) and gray (no Tester).

  • View in gallery

    Column-averaged PTU as a function of column-averaged temperature for all XBT data shown with dots. For data except MED, PTU was grouped into 1°C bins and the mean was calculated in each bin, which is shown with dark blue line, with an error bar indicating two standard errors of the mean. Linear trends are shown with dashed lines for MED and other data separately.

  • View in gallery

    Time variation of different FRE coefficients and PTU. Red lines show updated results based on global-scale dataset. The results of side-by-side datasets (all other data except MedT5) are shown in blue, with two standard errors shown with blue error bars. In side-by-side data, data in three successive years are collected together to give an estimate in each year, in order to increase data coverage and reduce uncertainty. The green line and error bars are for MedT5.

  • View in gallery

    T5 profiles vs CTD from MedT5, with (left) manufacturer T5 FRE, (center) new FREC and (right) new FREC and PTU correction. The time interval between the CTD and the last T5 was 76 min, and launching height h is also indicated. The large steplike structure of the CTD temperature profile highlights the double diffusion phenomenon.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 98 98 6
PDF Downloads 78 78 6

Assessment of Quality and Reliability of Measurements with XBT Sippican T5 and T5/20

View More View Less
  • 1 Italian National Agency for New Technologies, Energy and Sustainable Economic Development, Pozzuolo di Lerici, Italy
  • | 2 International Centre for Climate and Environment Sciences, Institute of Atmospheric Physics, Chinese Academy of Sciences, Beijing, China
  • | 3 Institute of Marine Sciences, National Research Council, Pozzuolo di Lerici, Italy
  • | 4 Bedford Institute of Oceanography, Department of Fishery and Oceans Canada, Dartmouth, Nova Scotia, Canada
© Get Permissions
Full access

Abstract

The T5 expendable bathythermographs reach the greatest depth within the current XBT family. Since the early 1970s, in several areas they have been providing a significant part of available temperature profiles below 1000 m and therefore represent an important resource for ocean climate study. In this paper we present new results from laboratory tests of Sippican T5 and T5/20 probes and analyses of more than 350 XBT–CTD matched pairs from our own field trials and the World Ocean Database (WOD), and we propose an improved fall rate equation (coefficients: A = 6.720 ± 0.025 m s−1, B = 0.001 60 ± 0.000 15 m s−2, Offset = 1.00 ± 0.65 m). Possible influences of probe physical characteristics and initial launch conditions on the probe motion have also been investigated with launching height and probe weight being identified as important factors. Analyses also confirm that fall speed and pure temperature error increase with water temperature, as previously reported for other XBT types. The uncertainties in depth and temperature measurements are then calculated. Finally, a new correction for a global T5 dataset is proposed, with an update of the currently available schemes.

© 2018 American Meteorological Society. For information regarding reuse of this content and general copyright information, consult the AMS Copyright Policy (www.ametsoc.org/PUBSReuseLicenses).

Corresponding author: Lijing Cheng, chenglij@mail.iap.ac.cn

Abstract

The T5 expendable bathythermographs reach the greatest depth within the current XBT family. Since the early 1970s, in several areas they have been providing a significant part of available temperature profiles below 1000 m and therefore represent an important resource for ocean climate study. In this paper we present new results from laboratory tests of Sippican T5 and T5/20 probes and analyses of more than 350 XBT–CTD matched pairs from our own field trials and the World Ocean Database (WOD), and we propose an improved fall rate equation (coefficients: A = 6.720 ± 0.025 m s−1, B = 0.001 60 ± 0.000 15 m s−2, Offset = 1.00 ± 0.65 m). Possible influences of probe physical characteristics and initial launch conditions on the probe motion have also been investigated with launching height and probe weight being identified as important factors. Analyses also confirm that fall speed and pure temperature error increase with water temperature, as previously reported for other XBT types. The uncertainties in depth and temperature measurements are then calculated. Finally, a new correction for a global T5 dataset is proposed, with an update of the currently available schemes.

© 2018 American Meteorological Society. For information regarding reuse of this content and general copyright information, consult the AMS Copyright Policy (www.ametsoc.org/PUBSReuseLicenses).

Corresponding author: Lijing Cheng, chenglij@mail.iap.ac.cn

1. Introduction

Following a request from the U.S. Navy, prototypes of expendable bathythermograph (XBT) probes for marine applications were developed in the early 1960s. The version by Francis Associated was the winning model, and its production began in 1964 by Sippican Corporation [currently Lockheed Martin Sippican (Sippican)]. Since then, several versions of XBTs have been developed (see Table 1), customized for different depths and launching from fast ships, submarines, and airplanes. An XBT system [including an XBT probe falling in water, a recording system (RS), a computer, and a launcher with a connecting cable] is cheap, convenient, and easy to use. Therefore, between 1970 and 1990, XBTs collected most of the temperature data in the upper 2000 m of the ocean, especially along the main commercial ship lines. Consequently, the ocean cover was inhomogeneous in terms of both space and time. Presently the number of XBT probes launched annually has greatly diminished in favor of profiling floats. Nevertheless, since they are versatile, inexpensive, easy to use, and complementary to other manned observations, XBTs remain a useful source of oceanographic information (http://www.aoml.noaa.gov/phod/goos/xbtscience/index.php; Goni et al. 2010; Abraham et al. 2013; Cheng et al. 2016).

Table 1.

Main characteristics of Sippican XBT types. The difference between rated and actual maximum depth is due to the tolerance in the amount of wire on the probe spool. The last two columns on right show the coefficients of the FRE describing the probe fall (provided by Sippican). We note that since 1996 but only for T4, T6, T7, and DB probes, the Integrated Global Ocean Station System (IGOSS) proposed to use different values based on H95: A = 6.691 m s−1 and B = 0.002 25 m s−2. The characteristics of T12 and LMP5-T1 are from Sippican software versions MK21 v2.3.1 and v2.7.1, respectively; 1 kt ≈ 0.51 m s−1.

Table 1.

The values of XBT measurement uncertainties in depth D and temperature T stated by the manufacturers have been reassessed in various occasions (e.g., Flierl and Robinson 1974, 1977; McDowell 1978; Anderson 1980; Seaver and Kuleshov 1982; Heinmiller et al. 1983; Hanawa and Yoritaka 1987; Bailey et al. 1989, 1994). Recently, climatologists have also highlighted both the importance of historical XBT datasets and the need for accurately estimated uncertainties, which are crucial for climatological analyses (e.g., Gouretski and Koltermann 2007; Abraham et al. 2013; IPCC 2013; Boyer et al. 2016).

This article examines the properties and consistency of measurements made exclusively by Sippican T5 and T5/20 probes (Fig. 1), which can reach about 2000 m. The total amount of T5 profiles available worldwide, up to 2000 profiles per year in the late 1990s (Fig. 2a) or up to 10% of the ocean subsurface profiles below 700 m (Fig. 2b), is much smaller than that of the shallower XBT types. We do not consider T5 probes manufactured by Sparton of Canada and TSK. The properties of TSK T5 probes have been analyzed thoroughly (Kizu et al. 2005a,b; Miura et al. 2004; Koso et al. 2005). In this paper, unless otherwise stated, “T5s” will denote both Sippican T5 and T5/20 probes.

Fig. 1.
Fig. 1.

T5/20, T5, and DB XBT probes are shown from top to bottom. T5 and T5/20 differ only by the amount of wire on the canister spool and the junction between the canister spool and the probe spool. DB is shorter (no plastic cylinder between the zinc nose and the plastic afterbody), lighter (also for the amount of wire on the probe spool), and with a different amount of wire on the canister spool.

Citation: Journal of Atmospheric and Oceanic Technology 35, 10; 10.1175/JTECH-D-18-0043.1

Fig. 2.
Fig. 2.

(a). Number of profiles for nine major XBT probe type groups per year from 1966 to 2016 (T5 in black). Two groups for unknown probe type are also included (DX: Deep Unknown with maximum depth > 550 m and SX: Shallow Unknown with maximum depth < 550 m). (b) Percentage of T5 data in WOD13 at each depth from 1966 to 2016.

Citation: Journal of Atmospheric and Oceanic Technology 35, 10; 10.1175/JTECH-D-18-0043.1

In section 2 there is a summary of the articles where T5 probes were used with a list of problems, followed in section 3 by results of laboratory measurements, calibration in bath, and recording systems’ tests. Selection criteria for profile pair matching and techniques applied in our analysis are reviewed in section 4. The new fall rate equations, the effects of water temperature, launching height, and other physical parameters on the fall rate are presented in section 5, while uncertainties of temperature values are analyzed in section 6. A comparison of T5s versus Argo in the Mediterranean and the global-scale analysis are presented in section 7. The final section contains a brief discussion of the main results. Characteristics and properties of an XBT system are reviewed in appendix A, and the approach used for the uncertainty estimate is in appendix B.

2. Short history of the T5 probes

XBT probes and measurements have recently been discussed (e.g., Reseghetti et al. 2007; Kizu et al. 2011; Abraham et al. 2013; Cheng et al. 2016 and references therein). Here we recap only key aspects of T5s, leaving the most relevant characteristics of XBT systems for appendix A, where we also define properties and nominal values of uD and uT, the standard uncertainties associated with the depth and temperature measures, respectively.

T5s are scarcer than some other XBT types: in the World Ocean Database 2013 (WOD13), January 2017, they are only 1.4% of the total amount, with yearly contributions of 1%–15% of available profiles in the range 800–1800 m during 1990–2002 (Fig. 2). T5s are also currently known as the 6000-ft probe, although there are incomplete references to a 5000-ft (1520 m) version (IGOSS 1972; USNAVO 1975). T5s can be easily identified within the XBT family by visual inspection and weight measurement (Fig. 1): they are ~13 cm longer (because of a plastic cylinder between the zinc nose and the plastic afterbody) and at least 250 g heavier than the other XBTs with which they share all the other components (zinc nose, thermistor, plastic afterbody with terminal fins, plastic internal spool, and copper wire). The production of T5s started in 1968, and it was immediately evident how “delicate” that probe was, probably linked to the probe's very long deploying time. Sippican indicated a drop rate success of ~80% for T5s (to be compared with more than 90% for shallower XBT types) and provided a standard fall rate equation (FRE) D(t) = AtBt2, where AS = 6.828 m s−1 and BS = 0.001 82 m s−2 (T5FRE), to describe its fall motion. Table 2 summarizes the currently available fall rate equation coefficients (FRECs) for T5s: Boyd and Linzell (1993, hereafter BL93) provided an alternative version, but the values quoted in Mied et al. (1981) were not confirmed. Sippican T5 probes suffer a hindrance caused by the low maximum speed (6 kt = 3.06 m s−1) allowed to the ships so that the probe can reach its maximum depth. Since the 1990s, Sippican was asked to study the feasibility of a 2000-m XBT version for fast ships. After several trials (see Table 1), in 2007 Sippican released the T5/20 version with a canister spool hosting a very long wire coupled with a T5 probe (see Fig. 1) that can reach the same depth as a standard T5 but from fast ships (20 kt).

Table 2.

FRE coefficients available in literature for T5 and similar probes. Results concerning T5 TSK probes are shown in italics for comparison.

Table 2.

There are several articles and reports describing the use of T5 probes that showed problems caused by incomplete acquisition (frequent wire breakage) and the reliability of measurements, especially in the early 1970s. Daubin (1973) reported that only 38% of 176 probes dropped in November–December 1972 had a good full-length acquisition (and a further 28% were reliable down to a D ~ 760-m depth). He quoted that Sippican had experienced difficulties with quality control on probe production. The failure rate was also discussed by Fenner et al. (1974) (41% of full good acquisition and an additional 30% with good measurements at D < 760 m), McDowell (1978) (~30% of failed drops), Anderson et al. (1979) (45% of failed drops), Anderson (1980) (~30% of failed drops caused by apparent insulation problems in the upper 50 m), King (1991) (3 of 19 as the failure rate, mainly attributable to operators and computer), Cunningham et al. (2000) (18 out of 82 as the failure rate but with small uT), and Paillet (2001) (failure rate of ~25%).

Volkmann (1977) noted that at D > 1000 m, the isotherms from T5 data were up to 80 m deeper than from CTD measurements. McDowell (1978) did an accurate analysis on T5s used during POLYMODE research cruises. He pointed out the effect of pressure on the thermistor and a lighter-than-stated weight as possible sources of bad measurements. Then, his comparison of profiles jointly recorded by an analogic recording system (ARS) and a digital recording system (DRS) showed uncertainties linked to variability of T5 weight and shape and to the digitization process of ARS profiles. In addition, isotherms from T5 readings were always deeper than those from concurrent profiles from a salinity–temperature–depth device. Anderson (1980) found slightly lower temperature values than those from collocated T4 probes. Seaver and Kuleshov (1979, 1982) observed that the depth difference ∆D for T5s has the opposite sign with respect to T7 at D > 120 m, and they also suggested that T5 should have a more stable configuration than T7 during the fall because of different shapes and larger weight. They concluded that 7 m ≤ uD ≤ 23 m in the 800 m ≤ D ≤ 1500 m range if a DRS is used and systematic components on uD are removed. Over a sample of 150 drops, Kennelly et al. (1989) found that the difference in temperature values between XBT and CTD (∆TXBT–CTD) was ∆TXBT–CTD = +0.075°C (but assuming that D values calculated by T5FRE are correct). Szabados and Wright (1989) verified that T5FRE describes the fall motion adequately (but their conclusion is based on only 10 drops). They also showed that the mean of |∆TXBT–CTD| for T5s is different from one of the other XBT types and depends strictly on the RS used (0.11°C ≤ ∆TXBT–CTD ≤ 0.24°C). After an accurate analysis, BL93 calculated FRECs with lower fall speed and smaller deceleration than in T5FRE, and found ∆TXBT–CTD = +0.07°C, probably linked to the DRS. In IOC (1995) there is an indication of comparisons conducted by researchers from France, Japan, and Germany in northwest Pacific and in the tropical and northwest Atlantic but without final data analysis. Hill (1995) found temperature values consistent with concurrent CTD casts. Kizu et al. (2005a) did not find a systematic bias for T5s, while Kizu et al. (2005b) reported that the differences were usually within the nominal tolerances and identified a correlation between T and the fall rate. In tests conducted in the subequatorial Atlantic Ocean in 2000 and 2005, Brazilian researchers found that T5FRE usually overestimates D (Fonteles and Mata 2009). The analysis of 12 T5 probes dropped during the fifteenth Flux Interchanges of Carbon Dioxide in a Atlantic Meridional Section (FICARAM-15) cruise in the Atlantic Ocean showed 0.10°C ≤ ∆T ≤ 0.36°C with respect to a temperature reference, only a single probe had ∆T < 0°C (FICARAM Group 2013). Meincke (1991) and Sy et al. (2000) found variability in the relationship between manufacturing year and D, based on tests conducted in 1991–97 in the North Atlantic; Ridgway et al. (1999) confirmed similar behavior, but their datasets are no longer available. Finally, we note that the drag coefficient for T5s was first evaluated by computational fluid dynamics (CFD) calculation in Stark et al. (2011).

3. Laboratory tests

This section discusses the results of laboratory tests conducted on the components of an XBT system and T5s probes to try to quantify their influence on the measures. As for the probes’ physical dimensions, we wanted to determine their range of values and their possible variability over time, given that Sippican stated that the values of all the parameters remained constant from the beginning within the industrial tolerances.

We tested the DRS to determine its contribution to the uncertainty, the possible influence of the specific device used, and the time stability of the measurements. Finally, we used calibration bath tests to check the possible correlations between uncertainty in T values, XBT type, temperature, and components of the recording circuit. The methodology followed in the uncertainty estimate is shown in appendix B.

a. Tests on probe dimensions

The copper wire linear density (WLD), the probe weight in air W, the diameter of the central hole δ, and the external maximum diameter of the zinc nose ∅ of several tens of T5s manufactured in 2002, 2003, 2007, 2008, 2010, 2014, and 2017 were measured in laboratory tests.

We cut from more than 100 T5 probes a sample of wire 10.00 ± 0.02 m long (including 55 probes dropped in the Mediterranean Sea in comparison tests). The results of the weight measurements (with an analytic scale resolution < 0.0005 g) are shown in Table 3 as well as 60 values of ∅ (of which 12 dropped in the sea) and 91 (18 dropped) measured values of δ with Vernier calipers.

Table 3.

Wire linear density, nose diameter, and hole diameter for T5 and T5/20 probes measured in laboratory tests.

Table 3.

The weight values are within the range provided by the manufacturer without any evident correlation with the manufacturing year and are nearly coincident with the values from other contemporaneous XBT types.

Table 4 summarizes the W values for 131 T5s manufactured in the period 2007–17 (including 33 probes dropped during tests in the Mediterranean). The nominal values provided by Sippican (2014, private communication) are W = 998.0 ± 5.0 g in air and 680.5 ± 2.0 g in water. We found a certain batch-to-batch variability, whereas Kizu et al. (2005b) quoted 714.3 ± 1.7 g as the average weight in water for T5s manufactured in 2003. The weight of two zinc noses was found to be of 613.5 and 614.0 g (with an analytical scale resolution of 0.05 g), respectively, while the stated range is 611.9–613.7 g.

Table 4.

Weight in air for T5 and T5/20 probes separately measured in laboratory tests as a function of the manufacturing year.

Table 4.

We also measured the length of the plastic cylinder L connecting the nose with the plastic afterbody of 57 T5s and the results are 12.69 ≤ L ≤ 12.75 cm (resolution = 0.005 cm). Finally, we observed an unexpected gap between the zinc nose and the plastic cylinder in some probes with a maximum value of 0.10 cm, as well as some variability in the length of their plastic fins (Gouretski et al. 2010).

b. Tests on recording systems

Both ARS and DRS have been acknowledged as factors influencing the measurement process (Roemmich and Cornuelle 1987, hereafter RC87). The intrinsic uncertainty of an RS on a temperature reading and its time response to quick and sharp temperature changes indicates how an RS works and how an RS may influence the whole uncertainty uT on T in an XBT measurement. We used two different tester probes (hereinafter Tester) manufactured with high-accuracy (~0.01%) resistors and working at two fixed reference values (Tester_1 at T = 12.20°C and T+ = 26.75°C; Tester_2 at T = 12.65°C and T+ = 27.95°C). We also used a set of high-accuracy (0.01%) resistors and the “CSIRO XBT system comparison timer” (which provides the same value of resistance to two DRSs and simulates quick-step T changes) to evaluate the performances of some Sippican DRSs (two MK12, three MK21-USB, and one MK21-Ethernet). Measurements were repeated in rooms at slightly different conditions (23°C < Troom < 27°C), and the results can be summarized as follows:

  • The calibration of DRSs with respect to a set of high-accuracy resistors giving values from −2° to 36°C resulted in different calibration curve shapes (Fig. 3) of the ∆T profiles depending on the DRS type. Two MK12 with −0.01°C ≤ ∆T ≤ 0.05°C had an almost linear increase with temperature, whereas the MK21 family shows a parabolic behavior with a minimum ∆T value at 12.0°C ≤ T ≤ 14.0°C and a smooth sideways rise.
  • Two out of three MK21-USB had an almost random behavior in time with a variability up to ~0.10°C on a time scale of some hours, while the MK21-Ethernet had a linear ∆T increase on the same time scale with 0.01°C drift. On the other hand, the MK12 DRSs did not show such a variability.
  • Different DRS types detected the start of positive or negative step temperature changes with different response times (∆t ~ 0.5 s as maximum), while the final stable temperature was reached after 0.2 ≤ ∆t ≤ 0.4 s. MK12 DRS had a slightly faster response time (by ∆t ~ 0.1 s) than MK21 versions while detecting temperature changes.
Fig. 3.
Fig. 3.

The ∆T between the output of different Sippican RSs and the expected reading based on a set of high-accuracy reference resistors. The room temperature conditions were reasonably similar (23°–27°C).

Citation: Journal of Atmospheric and Oceanic Technology 35, 10; 10.1175/JTECH-D-18-0043.1

c. Tests in calibration bath

The calibration of XBT probes in a laboratory bath provides a reliable reference for temperature measurement (e.g., Georgi et al. 1980; Budéus and Krause 1993; Zanasca 1994). More recently, Reseghetti et al. (2007) found that XBT readings were always higher than the Tbath values for a few T4 and Deep Blue (DB) probes with ∆T and a standard deviation (SD) increasing with T, in substantial agreement with the DB-based values by Sy (1989).

Since 2007, we conducted calibration tests in a thermalized and salty bath (35.00 psu) at the Centre for Maritime Research and Experimentation (CMRE) Laboratories in La Spezia (Italy) on T5s coupled with different DRSs (MK12 and MK21-USB) and hand launchers having variable lengths. First, all probes in the test (from a minimum of 2 to a maximum of 10) were inserted about 10 cm deep in the calibration bath to thermalize both the nose and thermistor (similar to RC87), and they remained a few hours in the bath during the measurements at different T values. Water circulation was maintained with a propeller installed in the bottom. The Tbath values were selected to match the conditions in the Mediterranean (12°C ≤ Tbath ≤ 28°C). In 2008 and 2010, the same XBT probes were read using Sippican LM3A hand launchers having different cable lengths (two 20- and two 50-m versions). The readings (Table 5) were always nearly coincident (|∆T| ≤ 0.01°C): nominal accuracy mirrored the reading resolution. If the last calibration test is not considered, then there is a trend showing higher ∆T at higher T. Moreover, two different DRSs provided different Toffset when reading the same probes (February 2008) and two calibrations (August 2007 and February 2008) showed large negative Toffset. A possible explanation is in terms of the bias variability caused by the DRS (up to ~0.10°C or more) as shown in some subsequent tests on the DRS itself. Unfortunately, only in the last test did we use a Tester to measure the possible Toffset introduced by DRS.

Table 5.

Coefficients of the linear fit of data from different calibrations in bath at CMRE: ∆T = Toffset + AT. Number of calibrated probes, T range, and number of different temperature values are also indicated. The RS used was always the same, MK21-USB, with the exception of probes labeled with *, when the test was done by an MK12 checking the same probes labeled with a °.

Table 5.

4. Data and methods

Given that the main aims of this paper are to analyze T5s fall motion, to revise FRECs, to estimate the influence of some parameters on T5 fall, and to evaluate the uncertainties uT and uD on temperature and depth measures, we prepared a collection of T5s–CTD pairs, by merging profiles from our own archives, from those of some colleagues, and from the WOD13. Table 6 contains some information about the individual datasets used in this collection, while Fig. 4a covers their geography. The data distribution is far from homogeneous: most pairs come from the North Atlantic and the Mediterranean, a few tens from the Pacific Ocean, and a small number from southern oceans (no data from the Indian Ocean).

Table 6.

Some characteristics of the different datasets and calculated values of FREC (BL93 and Sippican and Hanawa95 FRE are also included for comparison). The last two columns show the depth after AT = 290.6s (the value rated by Sippican) and after AT = 360.0s, nearly coincident with the maximum allowed.

Table 6.
Fig. 4.
Fig. 4.

(a). Geographical distributions of T5–CTD pairs used in the present analysis. Different datasets are shown by different colors: GO_T5 (blue), JapT5 (cyan), CanT5 (light green), MedT5 (red), Brazil_T5 (purple), and WODT5 (olive). (b). Zoomed-in view of the Mediterranean Sea—MedT5 dataset: different colors identify different years when comparisons have been made. Stars indicated pairs from ArgoT5.

Citation: Journal of Atmospheric and Oceanic Technology 35, 10; 10.1175/JTECH-D-18-0043.1

The CTD profiles used as reference are typically reduced to 1-m/dbar bins for data storage. For consistency, we vertically interpolated both XBT and CTD profiles every 1 m. XBT raw profiles were interpolated to equidistant 1-m profiles by linear inverse distance weighting using the closest neighbors within a range of X m. No further filtering was applied. Profiles previously sampled at 2-m intervals were linearly interpolated to a 1-m sequence. On the other hand, CTD profiles were accepted only with 2-m/dbar resolution or less [conversion from dbar to depth followed Fofonoff and Millard (1983)].

All instrumental measurements do have errors and uncertainties: for example, the Sea-Bird Electronics (SBE) sensors installed on the Argo floats in the Mediterranean have nominal accuracy ranging between 2.4 and 3.0 db for the pressure and between 0.002° and 0.03°C for the temperature. Because of the higher accuracy, we consider the chosen CTD profiles as the “true” representation of T and D, attributing the ∆TXBT–CTD deviations from zero to inaccuracies in the XBT readings. For the current purpose, we ignore all the issues related to the quality of CTD sensor operation, performance, and calibration (see Raiteri et al. 2018).

Consistently, TXBT and TCTD are measured in degrees Celsius (°C) and D is measured in meters. The main characteristics of the merged datasets are described below.

a. MedT5 dataset

T5s and CTD profiles were recorded by the authors from July 2007 to November 2016 during cruises on Italian research vessels in the Mediterranean (Fig. 4b), a marginal but warm and salty sea, with highly homogeneous temperature values at D > 400 m. The dataset consists of 55 pairs: 48 T5s were dropped at height h ~ 2.5 m and 7 T5s from a higher platform, with a delay from the beginning of the CTD cast ranging from a few minutes to about 3 h. Two MK21-USB DRSs and two laptops were used with an additional MK12 DRS in August 2012. Each T5 acquisition was estimated as “good” until the wire broke or some problem occurred in the data taking. The values of acquisition time (AT) vary from 22.1 to 369.1 s: 21 probes (out of 55) have AT < 290.6 s (the nominal value), but for 32 T5s AT > 320 s. Since late October 2007, the DRS efficiency was checked by a Tester, and the T5s readings were corrected before using them in our analyses for 42 out of 55 profiles.

CTD profiles were recorded using always the same Sea-Bird SBE 911 plus automatic profiler, calibrated before each cruise. The CTD underwater unit was always lowered at a nominal speed of 1.0 m s−1 (with an uncertainty of ~0.1 m s−1) sampling at 24 Hz. The static nominal accuracy of the CTD is 0.001°C in T, and 0.0003 S m−1 in conductivity, but the accuracy in operational conditions is poorer. The (static) time constants are 0.065 s for the conductivity and temperature sensors (nominal spatial resolution is 0.065 m), and 0.015 s for the P sensor (spatial resolution is 0.015 m). CTD measurements were processed using Sea-Bird’s software (data conversion, alignment, cell thermal mass correction, filtering, derivation of physical values, bin averaging, and splitting of down and up trace profiles); afterward, they were qualified with SeaDataNet protocols.

b. Other datasets

Other datasets include pairs selected under slightly different criteria in space and time, and recorded during both side-by-side and other different research activities.

The only established side-by-side dataset is JapT5, which consists of 23 T5–CTD pairs recorded in 2003 by Japanese researchers in the Pacific Ocean close to Japan and previously analyzed in Kizu et al. (2005a) (by courtesy of S. Kizu). The XBT profiles had already been reduced by 1 m.

The geographical and temporal constraints applied to other datasets are summarized in Table 7. We underline that the GO_T5 dataset includes raw profiles from a large geophysical survey conducted in the Gulf of Cadiz, within the EU project GO. T5s were dropped from a moving ship, while the CTD cast was performed from another ship, therefore making the simultaneous deployment at the same position impossible. Because of the strong water exchange in that area between the Mediterranean and the Atlantic Ocean at D ≥ 500 m, resulting in a significant variability on very short distances and time scales, only pairs in which |∆TXBT–CTD| < 1.0°C (i.e., 5 uT_XBT) at D > 400 m, an arbitrary value but slightly above the exchange region, were accepted. On the other hand, the CanT5 dataset includes pairs recorded within the Atlantic Zone Off-Shelf Monitoring Program (AZOMP), which yearly monitors the Labrador Sea and the western North Atlantic off Nova Scotia, Canada.

Table 7.

Selection criteria used for the datasets but excluding specific side-by-side activities. All XBT probes are T5 with the exception of CanT5 (T5/20 probes).

Table 7.

The WODT5 dataset consists of profiles extracted from WOD13 (January 2017), selecting code = 210 (and D > 1000 m as an additional request) and code = 999 (and D > 1250 m) and excluding profiles from the Pacific Ocean or from Japanese vessels to prevent possible contamination by TSK probes.

c. ArgoT5 dataset

This study also includes an XBT–Argo comparison in the Mediterranean as an independent comparison from the previous XBT–CTD comparison, because we found a different behavior of XBT error in MedT5 as compared to other regions (this will be shown in the following sections), so we aimed to prevent CTDs errors from influencing our results. Therefore, the XBT–Argo comparison could further reinforce our key findings based on XBT–CTD comparisons.

Profiles from Argo floats (usually equipped with SBE41 Sea-Bird sensors), which have been available in the Mediterranean since 2004, have been downloaded (http://www.coriolis.eu.org) if |ΔLat| ≤ 0.10°, |ΔLon| ≤ 0.15°, and |Δt| ≤ 7 days compared to T5s (Argo 2000). These constraints are weaker than those of T5s–CTD pairs (in particular Δt), but in the Mediterranean the variability caused by the time factor is less significant than that induced by a change of location. The maximum distance is about 15 km (a bit more than the Rossby radius for the Mediterranean; Grilli and Pinardi 1998, 1999) but less than the usual sampling distance for XBT within the Ship of Opportunity Program (SOOP) monitoring (~20 km). The selected Argo profiles were processed by converting the pressure into depth values [SBE (2002), which in turn uses the formula proposed in Fofonoff and Millard (1983)] and then interpolated linearly every 1 m as for raw XBTs.

The dataset consists of 15 pairs of profiles (Fig. 4b), originating from 7 T5/20 probes and 15 Argo profiles (from five different floats).

5. Fall rate equation

Hanawa et al. (1995, hereafter H95), below, summarized the state of the art of the most popular XBT models and proposed a method, which later became standard, to derive the coefficients of a fall rate equation “à la Sippican” based on the comparison between the T-gradient profiles of a XBT versus a reference device. In 2011, two articles described alternative methods. Cheng et al. (2011, hereinafter CH11) proposed using T profiles (instead of T-gradient profiles) and a second degree equation, including an offset term, to tune the description of the XBT probe motion. A completely different approach was suggested in Stark et al. (2011) and Abraham et al. (2011, 2012), where CFD techniques were applied to the description of XBT motion.

We summarize below the characteristics of an improved version of CH11 developed to better describe the fall rate of XBTs. In our approach it is currently impossible to isolate the influence of a single parameter, because the probe’s specific weight, copper wire density, diameter of the zinc nose, diameter of the central hole, launching height, and so on, all contribute together to the probe motion actually registered. This interdependence must be taken in account if a possible correlation between a single specific physical parameter and the characteristics of the motion of T5s is looked for.

a. Technique for FRE calculation

H95 and CH11 calculated the ∆TXBT–CTD profiles in “windows” centered at different depths and selected the best FRECs by minimizing ∆TXBT–CTD, but the window size is set arbitrarily. Furthermore, the first window considered is located at D = 100 m. Subsequently, the FRECs are also applied to the initial part of the profile, which is not used to calculate the FRECs, since in that area the motion of the XBTs is considered too dependent on the initial conditions. We have tried to match the size of the windows to the typical length scale of the temperature variation with depth in order to distinguish the scales of noises and signals (especially in the deeper regions). The characteristics of the profiles in the Mediterranean (very small temperature variability along a large part of the profile) provided strong support for this improvement. We tried to identify the best window size based on a sensitivity test by varying their size and then running CH11 for each value so that we had a set of FRECs. The uncertainties were calculated for each run and used as a metric to identify the best value for the window. For MedT5, the best solution is a 20-m window at D ≤ 200 m and a 120-m window at D > 200 m. For the remaining datasets, we used a 20-m window at D ≤ 200 m and a 60-m window at D > 200 m.

b. FRE coefficients

Given the following FRE:
e1
we searched for the best FREC values within the following ranges: 6.1 ≤ A ≤ 7.2 m s−1, −0.003 ≤ B ≤ 0.003 m s−2, and −10.0 m ≤ Offset ≤ 10.0 m running CH11. To obtain the best coefficients for a dataset, because of the variable length of the profiles, we first summed up the selected value of each coefficient multiplied by the length of the corresponding profile and then we divided by the sum of the lengths of all the profiles. The results for the global dataset are (see Table 6)
e1a
with 6.684 ≤ Ai ≤ 6.822 m s−1, 0.0011 ≤ Bi ≤ 0.0020 m s−2, and −1.0 ≤ Offseti ≤ 3.0 m as the range of variability. We note that ABL93ATW < AS, while BTW ~ BBL93 < BS. Individual results are shown in Fig. 5, where B and Offset are plotted versus A, and the error bars indicate two standard errors of the mean, which is also used in the analyses below.
Fig. 5.
Fig. 5.

FREC by using the FRE model of D = AtBt2 − Offset. (top) Coefficient A vs B and (bottom) coefficient A vs Offset shown with dots for each individual XBT–CTD pair: red (MedT5), blue (GO_T5), light green (canT5), cyan (JapT5), green (Brazil_T5), and gray (WODT5). The weighted mean of the FREC is shown (yellow star), compared with the T5FRE coefficient for T5 (yellow square). The linear regressions between A and B, and A and Offset are also shown with black lines.

Citation: Journal of Atmospheric and Oceanic Technology 35, 10; 10.1175/JTECH-D-18-0043.1

According to Cheng et al. (2014, hereafter CH14), we then calculated a linear regression BTW versus ATW, and OffsetTW versus ATW after eliminating values near the boundary of the provisional algorithms. We suppose that reaching the boundary implies a higher possibility that those FREC might not render the best coefficients. The results are
e2
e3
We used the weighted least squares rather than the traditional least squares method to reduce the impact of outliers.

c. FRE coefficients versus probe instrumental factors

The analysis of T5s motion as a function of probe parameters gave interesting results. Unfortunately, we have the full required information only for the MedT5 dataset. Furthermore, for this dataset, T should make a negligible contribution, since its range of variability is less than 1°C. Kizu et al. (2005b) found that the weight W of an XBT might influence its fall rate. We have information about weight for 33 probes (out of 55), and we found no significant relationship [at the 95% confidence level (CL) based on a Student’s t test] between their FREC and W. Because of the correlations between the A, B, and Offset, the three coefficients might be adjusted to minimize the total uD, as reported in H95: different combinations of A/B values along a line in the A/B plane reveal a similar uD. This is also why CH14 used A/B and A/Offset correlations to calculate B and Offset from A. Therefore, we additionally calculated ∆D by using the new FRECs compared with T5FRE,
e4
where Aj, Bj, and Offsetj are the individual FREC calculated in this interval. We found that ∆D “linearly” decreases with W (Fig. 6a), namely,
e5
Fig. 6.
Fig. 6.

(a). The ∆D as a function of the probe weight in air for MedT5 probes. (b). The ∆D as a function of diameter of the hole in the zinc nose for MedT5. The linear regressions are shown with dashed lines.

Citation: Journal of Atmospheric and Oceanic Technology 35, 10; 10.1175/JTECH-D-18-0043.1

This seems to suggest that heavier T5s have a slightly slower fall rate than lighter probes, similar to Kizu et al. (2005b).

We also have information about δ for 18 T5s: despite the poorness of the sample, nevertheless ∆D averaged over the whole depth interval (0–2500 m) varies from positive to negative values as δ increases (Fig. 6b). Data can be linearly interpolated giving
e6

Therefore, it might seem that probes with larger δ have a slower fall rate.

Copper wire on the probe spool contributes to T5 weight and to its decrease during the fall, but it also influences the motion in the first meters because of the trapped air, which is then quickly expelled as a result of increasing pressure. It is hard to perform measurements of this effect, so we checked the only possible influence of the variability of WLD on the T5 motion because of the slightly different mass loss. WLD values are available for the whole MedT5 dataset (55 probes). The FRECs and ΔD slightly diminish with WLD, but the variation is always negligible: this seems to indicate that WLD plays a very marginal influence on FRECs.

We have information on maximum diameter ∅ for only 12 probes, so the current results are without statistical significance.

A certain variability of the shape of terminal fins, the length of the plastic cylinder, and the roughness of the zinc nose was also observed, but the reliability of the measurements was too low to justify their use in our analysis.

d. FRE coefficients versus launching height

Launching platform height h over sea level has long been suspected to influence the first seconds of XBT fall and the correctness of D values provided by the standard FRE. In the past, several authors added an offset in their FRE to better match their reconstructed XBT profiles versus CTD profiles mainly in upper region (e.g., Singer 1990; Hallock and Teague 1992; Reseghetti et al. 2007). Sippican quotes an uncertainty uD ≥ 5 m (uD = 0.02D at D > 250 m), and this hinders analyses in the upper region. Field tests show that the actual depth is almost always different from the value calculated by the FRE standard if h = 2.5 m. In literature, papers by Green (1984) and Hallock and Teague (1992) and also in papers based on CFD approach (e.g., Abraham et al. 2012) accurately discussed the XBT motion in the upper layer. Reseghetti et al. (2007) published the results of a small field test with a dozen launches from two platforms at different heights but without conclusive results. Bringas and Goni (2015) reported a thorough analysis on measurements in a laboratory tank for DB-like XBT type and proposed a correction formula.

The range of h values in the datasets is shown in Table 6: sometimes h has been directly measured (MedT5) or empirically estimated with high accuracy (GO_T5 and CanT5), while h has been deduced after a check on pictures showing the ships used for the remaining datasets.

In our approach, Offset is the main indicator of ∆D at the surface. In Fig. 7, Offset values are plotted versus h: there is a decrease and the slope (−0.191 ± 0.158 m m−1) is significantly different from zero based on a Student’s t test at the 95% CL. MedT5 (red) and JapT5 (cyan) have higher-quality data: their statistics have less scattering than WODT5, GO_T5, and Brazil_T5, and a more significant dependence on h. Within MedT5, 48 probes were dropped at h ~ 2.5 m, 2 probes at h = 4.5 m, 4 probes at h = 7.0 m, and a probe at h = 12.0 m. The comparison with CTD confirms that the depth calculated by T5FRE is deeper than the actual value (∆D ~ 2.5 m) and the difference decreases at higher platforms: ∆D ~ 1.7 m when 2.5 ≤ h ≤ 4.5 m and ∆D < 1.0 m when 5.0 ≤ h ≤ 12.0 m.

Fig. 7.
Fig. 7.

Offset vs h shown with dots: red (MedT5), blue (GO_T5), green (CanT5), cyan (JapT5), purple (Brazil_T5), and gray (WODT5). The mean (median) of Offset at each launching height is shown in green (dashed green), with two standard errors in black. A linear fit for the green curve is shown with a dashed orange line. Test results in the shallow water are shown as yellow triangles and rectangles.

Citation: Journal of Atmospheric and Oceanic Technology 35, 10; 10.1175/JTECH-D-18-0043.1

To further investigate the influence of h on ∆D, tests in shallow waters were conducted (see Table 8): T5s were launched from a selected height, and the time when the probes hit the bottom (20 m ≤ D ≤ 100 m) was recorded. Since we were able to measure Dbottom with good accuracy in shallow water, the difference between the value calculated by a given FRE DFRE and the actual depth Dreal represents the error ∆D. The critical points of this analysis are the correct identification of the hitting point and the poor depth sampling interval (~0.68 m) as a result of the sampling time interval (0.1 s). The error on depth (∆Di) has been calculated as follows:
e7
Table 8.

Results of tests for T5 dropped from different h with different bottom depth. Depth is estimated by T5FRE. The uncertainty on ∆D is not indicated. Reference measurements have been done by wire (w) or acoustic device (ac).

Table 8.

We have not considered the Offset term in the calculation of DFRE: ∆Di between the calculated and measured bottom values largely reveals the significance of Offset. If the bottom has D > 40 m, then the uncertainty on A and B might impact uD, so we checked two different sets of A and B coefficients (As, Bs) and (AMed, BMed). Each ∆Di at the bottom is shown in yellow in Fig. 7, and the two choices of FRECs are represented by yellow triangles (AS, BS) and rectangles (AMed, BMed). The deployments at h = 2.5 m always show ∆D > 0 but ∆D < 0 for h = 12.0 m. This supports the previous findings of reducing uD with increasing h.

e. FRE coefficients versus water temperature

Since the 1980s, there have been indications that the temperature does influence the XBT fall rate (e.g., Seaver and Kuleshov 1982; Thadathil et al. 2002; Kizu et al. 2005a, 2011; Hamon et al. 2012; Cowley et al. 2013, hereafter CW13; CH14; Abraham et al. 2016; Ribeiro et al. 2018). The explanation is in terms of decreasing viscosity at higher T, so the drag force is larger in colder waters. Similar findings were obtained within a CFD approach by Abraham et al. (2012). Just to quote some numerical values, Kizu et al. (2005a) indicated ~2 m °C−1 for TSK T5 probes (or 0.0131 m s−1 °C−1 when linearly extrapolated), while CH14 calculated values ranging from 0.0050 m s−1 °C−1 for T4/T6 to 0.0025 m s−1 °C−1 for T7/DB. Abraham et al. (2016) also proposed the expected variability of the speed term linked to the near-surface temperature. Ribeiro et al. (2018) have recently shown that in the Antarctic region, T7/DB probes have a fall rate slower than initially proposed by Sippican, confirming a strong correlation between T and probe motion.

Following CH14, we divided A values into a 2.5°C large bin with the average temperature over a 0–100-m interval T100 (see Fig. 8, where we note the large spread for data from WODT5 and GO_T5). The black solid curve is the median, while the error bar represents two standard deviations. After a linear interpolation on two datasets, the results are (TW = black dashed line; TW-WODT5 = orange dotted line)
e8
e9
Fig. 8.
Fig. 8.

The speed term A as a function of 0–100-m averaged temperatures shown with dots: red (MedT5), blue (GO_T5), light green (CanT5), green (Brazil_T5), cyan (JapT5), and gray (WODT5). The black solid curve (and two standard error bar) is the median (two SD) of A at each 2.5°C temperature interval. The black dashed line is the linear regression for all data, while the orange dashed line is the regression for the dataset after the exclusion profiles from WODT5.

Citation: Journal of Atmospheric and Oceanic Technology 35, 10; 10.1175/JTECH-D-18-0043.1

In both the cases, A increases with T100 and the high-quality datasets (MedT5, GO_T5, JapT5, and CanT5) show an even larger slope (larger speed change) than TW, unfortunately with large uncertainties in assessing the way A depends on T.

Usually T decreases almost monotonously with depth in most profiles, so a slow decrease in the fall rate should be observed following the combined action of decreasing T and deceleration acting for a longer time. This combination should not occur in areas with homogeneous waters, such as the Mediterranean (high T) or the polar regions (low T). We tried to check this different behavior in the global dataset using ΔD, which has been identified as a good indicator of phenomenon strength. We looked for a possible relationship between ΔD and T at different depths on different depth ranges (i.e., 0–2500, 0–500, 500–1000, 1000–1500, and 1500–2000 m; see Fig. 9). The correlation is evident at each depth interval, confirming that T influences the fall rate. In the full 0–2500-m window (Fig. 9a), a mean ∆T = 10°C could lead to uD ~20 m. We note that MedT5 has the warmest (and saltiest) water among the analyzed datasets, while in JapT5 there are profiles from cold regions (the northwestern Pacific Ocean is almost “cold;” see Fig. 9d and 9e). In JapT5, at D > 1000 m, we note that ∆D increases with D more than in the other datasets: ∆D ~ 30 m for 1000 < D < 1500 m, ∆D ~ 50 m for 1500 < D < 2000 m.

Fig. 9.
Fig. 9.

The value of ∆D as a function of T: (a) 0–2500 m: mean ∆D vs averaged temperature for the whole profile: (b) 0–500 m: mean ∆D vs 0–500-m mean temperature; (c) as in (b), but for 500–1000 m; (d) as in (b), but for 1000–1500 m; (e) as in (b) but for 1500–2000 m. The mean ∆D at each 2.5°C temperature window is denoted by black curves, with two times standard error bars. In all panels, a linear regression between ∆D and T is shown with a dashed pink line. The different colors identify the different datasets as in Fig. 8.

Citation: Journal of Atmospheric and Oceanic Technology 35, 10; 10.1175/JTECH-D-18-0043.1

6. Uncertainty on temperature reading

By applying its own FREC to each profile, we get the best description of the T5 fall: the T5 profiles are more or less “stretched,” so that the thermal structures are moved up or down and the depth discrepancy between the XBT and CTD profiles also decreased. The ∆TXBT–CTD cannot be further reduced by changes in FRECs and, once the component of ∆TXBT–CTD linked to depth is minimized, the remaining part can be attributed to the different reading of the same thermal phenomenon by the two instruments. We thus define the pure temperature uncertainty (PTU) based on the ∆TXBT–CTD profile obtained after the application of individual FRECs to T5s. Figure 10 (for MedT5) and Fig. 11 (for the remaining datasets) show that the residuals are everywhere more consistent at any depth (especially at deeper depths), so the SD values of the residuals decrease when compared with profiles without depth corrections. Besides, we underline that uT = uT (D, T).

Fig. 10.
Fig. 10.

The ∆TXBT–CTD profile (a),(c) before and (b),(d) after individual ∆D correction for MedT5. Median of the ∆TXBT–CTD at each depth before correction (after correction) is also presented in black (green), and the SD is shown with a dashed line. The ∆TXBT–CTD profile at (top) 0–300 and (bottom) 300–2300 m.

Citation: Journal of Atmospheric and Oceanic Technology 35, 10; 10.1175/JTECH-D-18-0043.1

Fig. 11.
Fig. 11.

As in Fig. 10, but for all the other data except MedT5.

Citation: Journal of Atmospheric and Oceanic Technology 35, 10; 10.1175/JTECH-D-18-0043.1

The following paragraphs describe the results concerning MedT5. New FRECs do not significantly diminish ∆TXBT–CTD (Fig. 10): in other words, T5FRE provides a reasonable description of the probe motion in the Mediterranean but the SD with new FRECs is smaller than for the remaining datasets (one SD error bar attached),
e10
e11

In addition, the median PTU ≅ 0°C and becomes negative when D < 1000 m (PTU > 0°C for the remaining datasets). The most significant improvement occurs when D < 100 m: SD diminishes from 0.503° to 0.389°C, whereas median ∆TXBT–CTD varies from 0.045° to −0.023°C (with a good improvement where the upper thermocline occurs), and is more homogeneous along the profile.

The median of the ∆TXBT–CTD profiles over two intervals (Fig. 12) can be interpolated as follows:
e12
e13
Fig. 12.
Fig. 12.

PTU after individual depth correction as a function of depth. The linear regression for 10–2200 (blue) and 100–2000 m (red). For comparison, the pressure effect quoted in RC87 is shown (black).

Citation: Journal of Atmospheric and Oceanic Technology 35, 10; 10.1175/JTECH-D-18-0043.1

RC87 suggested the action of pressure on the thermistor (~10−5 °C m−1) as a possible source of the drift of ∆TXBT–CTD values with D, and the slopes in Eqs. (12) and (13) are in good numerical agreement with their proposal (black line in Fig. 12). A steeper slope occurs if the correlation between PTU and T (shown in the next section) is included: (1.206 ± 0.031) × 10−5 °C m−1 and (1.265 ± 0.026) × 10−5 °C m−1, respectively. These values are marginally compatible with the results of similar analyses in Reseghetti et al. (2007) on T4 and DB types, while the offset term is nearly coincident with the value for DB. We note that Ribeiro et al. (2018) quoted a negative slope (−2.52 × 10−5 °C m−1) for T7/DB probes in Antarctic waters but calculated over the range 0–700 m. The pressure effect on the thermistor (offset of ~0.15°C at 2000-m depth) was also noted by Sippican in the development of the 2000-m version of the aircraft XBT (Russell and Leavitt 1984). Also, Budéus and Krause (1993) indicated the action of pressure on the thermistor as a possible source of significant errors on T values. We have tried to remove this depth effect (evident in Figs. 13a and 13b and also Figs. 10c and 10d): the new PTU is almost constant with depth, and the depth-averaged value is PTUOffset = −0.009°C for MED data.

Fig. 13.
Fig. 13.

(a) Median and SD of PTU for T5 with (blue) or without (black) calibration in bath (MedT5 only). The median PTU after removing the pressure effect is shown in light blue and gray for T5 with or without calibration respectively. (b) Median and SD of PTU for T5 with (purple) or without (black) correction as a result of the use of Tester (MedT5 only). The median PTU after removing the pressure effect is shown in light purple (Tester) and gray (no Tester).

Citation: Journal of Atmospheric and Oceanic Technology 35, 10; 10.1175/JTECH-D-18-0043.1

We have also calculated PTU for different MedT5 subsets, namely, calibrated (18 probes) versus uncalibrated (37 probes) and with Tester correction (42 probes) versus without Tester correction (13 probes). Later, the pressure effect [Eq. (12)] is removed: the depth-averaged results are reported in Table 9, while the PTU profiles are shown in Figs. 13a and 13b. These PTU values (sometimes very close to zero and with a small SD) can be reasonably estimated as the probes’ intrinsic thermal offset.

Table 9.

Depth-averaged PTUOffset values calculated in different subsets of the MedT5 dataset. Equation (12) is applied as the pressure correction.

Table 9.

Despite the small sample size, the results seem to be stable and provide interesting indications: it would appear that the calibration in bath introduces an additional component (−0.024°C) to PTU. Moreover, calibrated T5s show lower speed and smaller deceleration (Table 6) and lower temperature reading. This strongly suggests that there may be procedures followed in the calibration in bath that change probes irreversibly. As a possible physical explanation, D. Roemmich (2017, private communication) suggested that a thin film of CaCO3 acting as an additional insulator remains on the thermistor and other probe components after bath calibration, while Sy (1992) proposed that saltwater can induce corrosion on the zinc nose, increasing its roughness and decreasing its fall rate. After these tests, the onboard calibration following the procedure described in Budéus and Krause (1993) seems to be a viable option.

On the other side, it is evident that the use of a Tester improves the PTU value, strongly reducing the component caused by RSs.

PTU versus temperature

Studies on other XBT types (Reverdin et al. 2009; CW13; CH14; Ribeiro et al. 2018) highlighted that PTU increases with T. To look at the behavior of T5s, first we column-averaged PTU against the column-averaged temperature for each pair (Fig. 14). This should serve to better describe each “averaged” profile: there is good coverage of different water conditions but with a reasonable smoothing of possible anomalies. This approach works well with pairs from waters having very different physical characteristics but shows ambiguous results when the sample is small or with highly homogenized water (such as for MedT5 or polar regions). Data plotted in Fig. 14 can be linearly interpolated as follows:
e14
e15
Fig. 14.
Fig. 14.

Column-averaged PTU as a function of column-averaged temperature for all XBT data shown with dots. For data except MED, PTU was grouped into 1°C bins and the mean was calculated in each bin, which is shown with dark blue line, with an error bar indicating two standard errors of the mean. Linear trends are shown with dashed lines for MED and other data separately.

Citation: Journal of Atmospheric and Oceanic Technology 35, 10; 10.1175/JTECH-D-18-0043.1

PTU increases at high T in all the analyzed datasets, but with different slopes and Toffset terms. Results for MedT5 are more controversial: all T values are in the range 13.5°–14.5°C but PTU increases with T as (and more than) in the World Ocean Database. Unfortunately, a large spread of PTU values at nearly coincident temperature renders any linear interpolation nearly meaningless.

7. Further results

a. T5 versus Argo in the Mediterranean Sea

T5s and Argo profiles were compared, and the mean value, SD, and median of ∆TXBT–Argo were calculated on the whole sample and on D ≤ 100 m and D > 100 m subsamples after the application of three FRECs (the original T5FRE, MedT5, and MedT5 Tester). We note that ∆TXBT–Argo also includes a small surviving component as a result of wrong D values. Results (Table 10) show an improvement when new FRECs are applied, mainly in the near-surface region even if the original FRECs do not provide a bad description of the probe motion in the Mediterranean, probably because of warm and homogeneous water temperature. We underline that the values obtained here are close to those obtained from the tests comparing T5s and CTD, confirming that where the procedures followed for the launch and operation of the instrumentation are sufficiently verified, the measurements made with the T5s probes in operational conditions are quite accurate and do not differ significantly from the results obtained during dedicated tests. However, when selecting the pairs of profiles to be compared, it is essential to apply constraints strictly related to the water characteristics in that particular area. Finally, we note that the launching platforms for these T5s have usually h ~ 10 m.

Table 10.

Results of the analysis on difference in T readings between T5s and Argo floats from the ArgoT5 dataset. Mean, SD, and median of ∆T for T5 profiles with depth calculated by different FRECs: original Sippican (T5FRE), MedT5, and MedT5 Tester.

Table 10.

b. Updated XBT correction scheme for T5 (after new FRE and new PTU)

Careful study of uncertainties in a side-by-side dataset provides useful information for the correction of the global XBT dataset. Currently, the CH14 scheme has been recommended as the best method to correct historical XBT data (Cheng et al. 2016, 2018): it uses A/B and A/Offset correlation, the dependence of PTU from T, and the coefficient A to construct the global correction. However, the original version of CH14 used relationships for T5s based on T7/DB data, but this study has shown that T5s have their own uncertainty. New relationships for T5s calculated here could lead to an update of the previously used values in global-scale XBT–Reference data and the T5 correction in CH14. In a global-scale XBT–Reference pair dataset, we compare T5 profile with the nearest CTD/Argo/optimal spectral decomposition (OSD) data within 1° and ∆t ≤ 30 days. There are ~3000 T5 pairs in the global-scale dataset.

Figure 15 shows the updated temporal variation of FRECs and PTU for global-scale (GS) data compared with the side-by-side (SbS) dataset. Terminal speed shows an increasing trend in global-scale dataset from AGS ~ 6.6 m s−1 (in the 1990s) to AGS ~ 7.0 m s−1 (after 2015). In side-by-side data since the 1990s, ASbS ~ 6.7 m s−1, while AGS ~ AS ≅ 6.8 m s−1 toward 2010, probably because MedT5 is the main dataset after 2007. After 2000, BGS > BSbS and BS (similar to A), because BSbS is deduced from A according to A/B correlation. There are also fluctuations in some years for side-by-side data because of both insufficient amounts of profiles and large uncertainties in some datasets (e.g., GO_T5). The Offset term shows time variability in a global-scale dataset very similar to that one in a side-by-side dataset. Finally, PTU has a major difference for the two datasets: PTUGS ~ 0.1°C, while it varies from PTUSbS ~ 0.04°C (before 2005) to PTUSbS ~ 0°C (after 2006), because of the importance of MedT5. A significant difference between global and side-by-side datasets is a long-standing issue (i.e., shown in CH14; CW13; Cheng et al. 2018), and the reason is not yet clear. The problem of side-by-side data is the influence of the small sample size, because it is very difficult to estimate the minimum number of data that can guarantee both satisfactory coverage in space and time and a statistically robust result. The global-scale analysis could in turn be influenced by mesoscale signals within 1° and a space–time window of 30 days that mediate interacting properties of the water mass (which change in space and time) and variability inside a 2000-m-high water column (also combined with an internal process on a short time scale). Therefore, CH14 (and this study) used high-quality side-by-side data to examine the influencing factors of XBT error (i.e., water temperature dependency) but use a large amount of global-scale data to get a more reliable estimate on the time dependency of XBT error, since the origin of the time variable error is not fully identified yet. This study follows CH14 to recommend using the time dependency from global-scale data as corrections for global T5 data.

Fig. 15.
Fig. 15.

Time variation of different FRE coefficients and PTU. Red lines show updated results based on global-scale dataset. The results of side-by-side datasets (all other data except MedT5) are shown in blue, with two standard errors shown with blue error bars. In side-by-side data, data in three successive years are collected together to give an estimate in each year, in order to increase data coverage and reduce uncertainty. The green line and error bars are for MedT5.

Citation: Journal of Atmospheric and Oceanic Technology 35, 10; 10.1175/JTECH-D-18-0043.1

8. Comments and conclusions

The updated knowledge on Sippican T5 probes obtained through laboratory tests and field measurements and some unresolved issues can be summarized as follows:

  • T5s have shown to be slightly more problematic than other Sippican XBT models. Nevertheless, full wire acquisition (up to about 360 s, despite the rated value of 290.6 s) has been obtained without decaying the quality of readings.
  • Laboratory tests are the important first step to identify the key influencing factors of the errors in XBT measures and can feed into the CFD method and the construction of a better global XBT correction method in the future. Our laboratory measurements (on a sample of T5s manufactured from 2002 to 2017) indicate that wire linear density, hole diameter, nose diameter, and length of the plastic cylinder have remained “reasonably” constant over the years. Only the probe weight has shown a larger variability (Table 4). Electronic components of an XBT system (thermistor, RS, and so on) do influence the quality of the reading in both D and T. The circuit time constant (which is variable and linked to RS type) causes a delay in detecting the start of a thermal event (up to 2–3 m). Our calibrations in bath (Table 5) confirm that the observed ∆T can be linearly described by a term that increases with T and a term likely depending on RS, circuitry, and working conditions. Field tests indicate that this procedure seems to induce nonreversible changes in the probe by modifying both the fall rate and the T reading (onboard calibration seems preferable).
  • Notwithstanding the small sample size of the tests in the Mediterranean, we have tried to identify correlations between probe dimensions, initial external conditions, and the probe fall rate. Although it is impossible to fully separate the influence of each individual parameter on the XBT error, and to use these parameters explicitly in the construction of correction scheme for the past observations, field tests suggest that the probe weight in air, launching height, and the size of the central hole seem to affect the values of the FRE coefficients. In short, heavier probes or those with a large hole diameter have a slightly slower fall rate and larger ∆D values.
  • The recording system also has a considerable impact on the error in temperature values. The use of a Tester to evaluate the efficiency of RS at the beginning and at the end of the acquisition period is strongly recommended as well as adding probe weight, hole diameter, and launching height within the transmitted metadata. Documentation of these parameters will help toward a better understanding of the error in the future, while the CFD method could directly benefit by the availability of this information to understand the error, too (see Abraham et al. 2012).
  • New FREC for T5s calculated by an improved CH11 version are close to A and B found in BL93 (Table 6), and for most datasets Ai < AS and Bi < BS. JapT5 has the smallest A and the largest B, but these profiles come from a region with the coldest water in this analysis, in full agreement with Ribeiro et al. (2018) for T7/DB probes in Antarctic waters. The Offset term shows mostly positive but small values (~1 m) with the exclusion of the MedT5 and JapT5 datasets, in agreement with CH11 for DBs and CW13 for T4, T6, T7, and DB probes. The Offset for T5s is close to the values for different XBT types, and this is a hint that the Offset could mainly be linked to launching and initial conditions, leaving a marginal dependence on the XBT type. If we use only ∆D at the rated bottom depth (1830 m) as an estimator (Table 6), T5FRE does not describe too badly the motion of the T5s for a large part of available dataset but often fails in the calculation of the correct depth values of thermal structures. Despite a rather small sample, h = 2.5 m and T5FRE do not work well together in the near-surface region: the actual depth of T5s is always shallower than calculated, while the disagreement diminishes for probes dropped from higher platforms.
  • Although our dataset does not offer global coverage (Figs. 8 and 9 show an evident undersampling at some temperature interval and a large spread of the fall rate values), a correlation between FRECs and T is undeniable. The dashed lines in Fig. 8 indicate that higher temperatures allow a faster fall (up to 0.01 m s−1 °C−1). On the other hand, Figs. 9a–e show that ∆D decreases at higher T (i.e., the deeper part of profiles from JapT5), but there is a large spread of individual ∆D values (e.g., WODT5). The fastest fall occurs in the Mediterranean, where T > 13.0°C over the whole profile (see also ∆TXBT–CTD profiles in Fig. 10).

The last comment is dedicated to an evaluation of the overall quality of measurements with T5s. Figure 16 shows a zoomed-in view of some T5 profiles belonging to MedT5 before and after the application of the procedures developed here. The upper region shows problems for T5FRE, while the new FREC makes T5 values much closer to CTD values. After this step, we emphasize the smallness of ΔT at D > 1000 m, almost exclusively of thermal origin. These results also support the description of the difference between XBT and CTD measurements in terms of two components, linked to T and D. Equations (12) and (13) indicate that when D increases, the T5 readings become slightly warmer than the actual value (∆TXBT–CTD increases with the depth). This difference can be linearly interpolated with a slope very close to the value proposed by RC87 as a result of the pressure acting on the thermistor (we suggest correcting all XBT readings for this effect in the future). Ultimately, if the XBT system works well under controlled modes and optimized procedures (see Sy 1991), T5s may provide a realistic representation of the existing thermal structures down to D ~ 2000 m with small differences when compared to the CTD readings (Fig. 16).

Fig. 16.
Fig. 16.

T5 profiles vs CTD from MedT5, with (left) manufacturer T5 FRE, (center) new FREC and (right) new FREC and PTU correction. The time interval between the CTD and the last T5 was 76 min, and launching height h is also indicated. The large steplike structure of the CTD temperature profile highlights the double diffusion phenomenon.

Citation: Journal of Atmospheric and Oceanic Technology 35, 10; 10.1175/JTECH-D-18-0043.1

During the preparation of this article, we found indications of old T5 profiles that were not archived and missing information on the instrumentation used and data processing. Therefore, we strongly recommend that all users of XBT probes do archive their data and begin to provide a more complete set of metadata (such as adding launch height, probe weight, and calibration of the registration system) to allow future improvements in the quality of XBT measurements. And for the global T5 dataset, for example, archived in WOD, we will use the updated T5 correction obtained in this study, which is recommended in Cheng et al. (2016).

Acknowledgments

This study is supported by National Key R&D Program (2016YFC1401800) and National Science Foundation (41476016). Some XBT–CTD pairs from 1999, 2003, and 2006 have been downloaded from NOAA NODC (http://www.nodc.noaa.gov/OC5/XBT_BIAS/xbt_bibliography.html). All of the T5 pairs are available at http://159.226.119.60/cheng/. The Argo profiles in the Mediterranean (http://www.coriolis.eu.org) were collected and made freely available by the Coriolis project and programs that contribute to it. We are grateful to the captains, crew, and technicians of the Italian R/V Urania and Minerva 1 for their help during field tests, and to E. Molinari and D. Galletti (CMRE, La Spezia, Italy) and S. Latorre (INFN, Milan, Italy) for laboratory tests. The Argo data were collected and made freely available by the International Argo Program and the national programs that contribute to it (http://www.argo.ucsd.edu, http://argo.jcommops.org). The Argo Program is part of the Global Ocean Observing System.

Many thanks to the following experts for their kind help during the preparation of this study: S. Kizu, for providing the profiles of the 2003 comparison tests and some very useful comments on technical aspects; V. Gouretski, for comments and help in recovering some old German data; A. Thresher and Australian colleagues from the CSIRO, for providing the time-box; M. Mata, C. Fonteles, and R. Torquato, for sharing many profiles of cruises conducted near Brazil after 2000; R. W. Hobbs, for providing the original raw data of the GO project (FP6; Contract 15603; NEST); D. Roemmich, for some very precious indications concerning his activity with XBT; T. Rossby, for some information on early tests and the encouragement to complete the work. We also thank R. S. Loewenstein for the help during manuscript preparation. We also want to thank the reviewers who helped us improve the quality of the article with their criticisms and comments.

APPENDIX A

Characteristics of an XBT System, Tolerances, and Uncertainties

An XBT is a torpedo-shaped probe falling through the water column dereeling a copper wire linked to a recording unit installed on a ship. XBT probes do not have a pressure sensor, so the depth D is not measured but estimated by an algorithm converting a time sequence of resistance values (sampled at a fixed rate by a thermistor inserted in the probe nose and read by the recording system) into a paired D–T sequence. The time-to-D conversion is obtained by the fall rate equation proposed by Sippican: D(t) = AtBt2, where D represents the depth at time t measured from the moment a probe hits the sea surface. The values of A (terminal speed) and B (deceleration) were determined using field tests: they depend on the XBT type but are constants and independent of water characteristics, ship motion, and launching conditions. Sippican developed different versions of XBTs (some properties are summarized in Table 1) that can be divided in four groups depending on dimensions and FRE coefficients: 1) T4, T6, T7, and DB (the most popular versions); 2) T5, T5/20 and FD; 3) T10; and 4) T11. FD, T5, and T5/20 are longer and heavier than the members of the previous group (see Fig. 1), whereas the T10 type has the same dimensions and shape as the probes in the first group but is lighter and used for measurements in shallower areas. The T11 type is a completely different probe with a very slow fall rate.

Sippican provided the values of accuracy and uncertainty to be expected in measurements with an XBT system. The accuracy on D depends on the RS sampling rate and on the XBT type: for the standard 10-Hz sampling rate, its initial value ranges from 0.18 m (for T11) to 0.68 m (for T5). The uncertainty on D is uD = 0.02D but uD ≥ 5 m. (see also Saur and Stewart 1967; Plessey 1967). While the sensitivity of T of the used negative temperature coefficient (NTC) thermistor is 0.01°C, the expected uncertainty on T because of the probe is uT = 0.10°C, but the global uncertainty on T of an XBT system stated by Sippican (independently on XBT type) is uT_XBT = 0.20°C (see also Saur and Stewart 1967; Demeo 1969; Denner 1969; Stegen et al. 1975; Anderson 1980).

The possible sources of uncertainties in XBT measurements were roughly classified into two families linked to T (uT) and to fall rate and D (uD) (Reid 1964; Francis and Campbell 1965; Little 1965, 1966): the thermal response and pressure sensitivity of the thermistor, insulation, flow characteristics, weight of XBT, launching height, wave height, surface water turbulence, Vship, time to reach terminal speed, wire drag, water characteristics, lateral lift force, RS chart drive accuracy, and precessional motion inducing an upward lift component as a result of the Magnus effect (Johnson and Lange 1979). Strong shearing currents have been supposed to influence the probe motion as well as its failure rate (Beatty et al. 1981). Helical trajectory in the first meter with bubble formation has been observed (e.g., Seaver and Kuleshov 1982; and by the authors in 2009 during field tests). Small variations in the dimensions and shape of the probes (even if within the “industrial tolerance” admitted by Sippican) have been suspected to introduce unpredictable variability in the fall motion, reducing the accuracy of the description based on the standard FRE (e.g., Seaver and Kuleshov 1979, 1982; Hanawa and Yoshikawa 1991; Kizu et al. 2005b, 2011). According to Sippican, the physical characteristics of XBTs (thermistor properties, dimensions, shape, weight, and copper wire) remained essentially unchanged since the start of production. The only official known change on XBT probes was a different coating process (after 1995, when the factory moved to Mexico and the suppliers were changed), resulting in a slight reduction of WLD in air but without changes in water. In 2016 Sippican moved its factory back to the United States, and there may have been a further change in suppliers. In the literature there are indications (e.g., Saur and Stewart 1967) that Sippican did apply some changes in 1965–66 (substitution of T3 type by T4 type, better wire insulation, and better response of the thermistor at the water entry). Sy (1992) indicated that after 1968 there were no changes in the XBT, but the third JCOMM SOOP Implementation Panel (IP) in La Jolla (2000) made a specific request to Sippican as a result of increasing incidence of XBT malfunctions caused by production changes. With high probability Sippican did apply some technical changes in T5 manufacturing procedures after problems experienced in the first period (Daubin 1973). The original characteristics of copper wire were 0.010 in. in diameter and a breaking strength of 1 lb. (Little 1966). We point out that the spool onboard unwinds clockwise, while the XBT spool unwinds counterclockwise, and this counterrotation results in zero net twisting of the two conductor wires. The terminal fins induce a clockwise rotation as the probe falls for both probe stability and to help the cable feed out with minimum resistance.

The thermistor (R25 = 5 kΩ), whose resistance R is a nonlinear function of T, had an original nominal response time of 0.110 s (0.130 s as the maximum; Magruder 1970) with accuracy better than 1% in R and had to meet the TR curve through the full range of pressure within ± 0.1°C (Sippican 2014, private communication). GE Sensing is the company currently responsible for the production of thermistors (also for TSK products). Early tests have shown that the range of variability for thermistors was less than ±0.06°C at 95% CL if the Sippican ARS is not used (Georgi et al. 1979, 1980). Moreover, SD < 0.03°C at T = 0°C and decreases to SD < 0.01°C in the region 25°–30°C, probably because of the properties of used thermistors.

The uncertainties of D (uD) and T (uT) for XBT measurements were estimated by comparisons of XBT profiles with contemporaneous and collocated profiles from more accurate devices, usually CTD probes (e.g., Wijffels et al. 2008; Ishii and Kimoto 2009; Levitus et al. 2009; Gouretski and Reseghetti 2010; Good 2011; Hamon et al. 2012; CW13; CH14). The main difficulty of this procedure is that the check on T occurs at a unique D value for a probe usually dropped from a stationary vessel instead of a moving ship (as in operational activity). The T5 probes move initially at a speed 6 times greater than a CTD, so the time needed to complete the fall (about 6 min) is much shorter than the time for a CTD to complete the descent to the same depth (~35 min). In addition, ship drift, sea currents, and internal waves could partially influence the goodness of these comparisons.

We have found in literature other techniques to study the XBT fall rate (but not applied to T5s): acoustic techniques (Dammann 1982; Newman and Dammann 1983), multibeam devices to accurately evaluate the bottom depth (e.g., Gould et al. 1990; Gould 1991), and altimetric comparisons (DiNezio and Goni 2011). Different RS versions have been used to record XBT measurements, and it is also well known that RSs are a source of problems (e.g., Fenner and Cronin 1978; Sy 1991; H95). Since 1965 there were digitizers coupled with ARS to obtain a numeric output (Saur and Stewart 1967), but ARSs were used until the early 1980s, and then electronic devices substituted them almost completely (Emery et al. 1986). To the first order, T is a linear function of voltage in the circuitry, but a small correction has to be applied because of the nonlinearity of the TR relationship (Sippican Corporation 1971).

The ARS requires an accurate daily calibration: after this, voltage becomes a linear function of distance along the strip chart that is scaled such that the nonlinear TR relationship is considered (Paden and Hendershott 1986). The transformation of ARS profiles into digital sequences was done with different criteria, software, and uncertainties: McDowell (1977) quoted uT = 0.1°C and uD = 1 m, while Anderson (1980) indicated uT = 0.055°C (as in Wannamaker 1980) and uD = 0.95 m.

It is a hard task to retrieve the results of tests on ARS. Fahrbach et al. (1984) used both ARS and DRS and found that the average profiles of ∆TXBT–CTD have similar shapes and SD: ARS shows always ∆T > 0°C (∆T decreases with D from ~0.5°C at D = 100 m to ~0.2°C at D = 750 m), whereas ∆T for DRS varies from +0.05°C to −0.3°C with a transition to negative values at D > 250 m. Szabados and Wright (1989) found −0.06°C ≤ ∆TXBT–CTD ≤ +0.10°C among four different digital RSs in the range 0°C ≤ T ≤ 30°C. Szabados (1992) found that the differences in temperature readings between ARS and DRS were not statistically significant. Kizu and Hanawa (2002a,b) found ∆T ~ 0.1°C for some DRS manufactured in Japan, whereas D. Snowden (2008, private communication) obtained ∆T ~ 0.05°C in a comparison between Sippican MK21-USB and Devil, manufactured by Turo (Australia). Zanasca (1994) reported ∆T = 0.02°C between two Sippican MK12 DRSs reading the same set of probes. The total uncertainty for DRS was early estimated uT = 0.1°C (Stegen et al. 1975), one-half of uT for a standard XBT system.

The standard sampling rate of DRS is 10 Hz, with a time window very close to the thermistor response time, while in the literature 20 Hz for TSK RSs and 8 Hz for early bathysystem RS are also available. Finally, we note that the Sippican recorder bridge circuit includes a resistor that linearly corrects for the thermistor response (with ~0.5°C as accuracy) and the Sippican DRS then corrects in the host computer. On the other hand, ARS used a nonlinear T grid on the chart paper output for the correction (IOC 1992).

APPENDIX B

Procedure for Evaluating and Expressing Measurement Uncertainty

The approach we followed in the paper to determine errors and uncertainties is described below by means of an example. Let the water temperature (at a given D) be the variable to be measured, indicated as T: in accordance with JCGM (2008) its estimated value and its associated standard uncertainty u(T) are obtained from a distribution of possible values. This probability distribution may be frequency based (i.e., based on a series of observations Ti of T) and/or it may be an a priori distribution. Type A evaluations of standard uncertainty components are based on frequency distributions (e.g., by calculating the SD of the mean of a series of independent observations), while type B evaluations are founded on a priori distributions identified by collecting information from various sources (e.g., previous measurement data, general knowledge of the behavior of the instrument, manufacturer’s specifications, or data provided in calibration and other reports). In both cases the distributions are models that are used to represent the state of our knowledge. Once both type A and type B uncertainties are determined, they have to be combined in order to obtain uc(T), the standard uncertainty of the measurement result T, and taken to represent the estimated whole SD of the result itself.

REFERENCES

  • Abraham, J. P., J. Gorman, F. Reseghetti, K. Trenberth, and W. Minkowycz, 2011: A new method of calculating ocean temperatures using expendable bathythermographs. Energy Environ. Res., 1, 211, https://doi.org/10.5539/eer.v1n1p2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Abraham, J. P., J. M. Gorman, F. Reseghetti, E. M. Sparrow, and W. J. Minkowycz, 2012: Drag coefficients for rotating expendable bathythermographs and the impact of launch parameters on depth predictions. Numer. Heat Transfer, 62A, 2543, https://doi.org/10.1080/10407782.2012.672898.

    • Search Google Scholar
    • Export Citation
  • Abraham, J. P., and et al. , 2013: A review of global ocean temperature observations: Implications for ocean heat content estimates and climate change. Rev. Geophys., 51, 450483, https://doi.org/10.1002/rog.20022.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Abraham, J. P., R. Cowley, and L. Cheng, 2016: Quantification of the effect of water temperature on the fall rate of expendable bathythermographs. J. Atmos. Oceanic Technol., 33, 12711283, https://doi.org/10.1175/JTECH-D-15-0216.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Anderson, A. L., and et al. , 1979: Bearing stake acoustic assessment (U). NOSC Tech. Rep. TR 466, 279 pp.

    • Crossref
    • Export Citation
  • Anderson, E. R., 1980: Expendable bathythermograph (XBT) accuracy studies. Naval Ocean Systems Center Tech. Rep. 550, 201 pp.

    • Crossref
    • Export Citation
  • Argo, 2000: Argo float data and metadata from Global Data Assembly Centre (Argo GDAC). SEANOE, accessed 28 September 2018, http://doi.org/10.17882/42182.

    • Crossref
    • Export Citation
  • Bailey, R., H. Phillips, and G. Meyers, 1989: Relevance to TOGA of systematic XBT errors. Proceedings of the Western Pacific International Meeting and Workshop on TOGA COARE, J. Picaut, R. Lukas, and T. Delcroix, Eds., ORSTOM, 775–784.

  • Bailey, R., A. Gronell, H. Phillips, E. Tanner, and G. Meyers, 1994: Quality control cookbook for XBT data. CSIRO Marine Laboratories Rep. 221, 83 pp.

  • Beatty, W. H., III, J. G. Bruce, and R. C. Guthrie, 1981: Circulation and oceanographic properties in the Somali Basin as observed during the 1979 southwest monsoon. Naval Oceanographic Office Tech. Rep. TR-258, 78 pp.

    • Crossref
    • Export Citation
  • Boyd, J. D., and R. S. Linzell, 1993: The temperature and depth accuracy of Sippican T-5 XBTs. J. Atmos. Oceanic Technol., 10, 128136, https://doi.org/10.1175/1520-0426(1993)010<0128:TTADAO>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Boyer, T., and et al. , 2016: Sensitivity of global upper-ocean heat content estimates to mapping methods, XBT bias corrections, and baseline climatologies. J. Climate, 29, 48174842, https://doi.org/10.1175/JCLI-D-15-0801.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Bringas, F., and G. Goni, 2015: Early dynamics of Deep Blue XBT probes. J. Atmos. Oceanic Technol., 32, 22532263, https://doi.org/10.1175/JTECH-D-15-0048.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Budéus, G., and G. Krause, 1993: On-cruise calibration of XBT probes. Deep-Sea Res. I, 40, 13591363, https://doi.org/10.1016/0967-0637(93)90116-K.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Cheng, L., J. Zhu, F. Reseghetti, and Q. P. Liu, 2011: A new method to estimate the systematical biases of expendable bathythermograph. J. Atmos. Oceanic Technol., 28, 244265, https://doi.org/10.1175/2010JTECHO759.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Cheng, L., J. Zhu, R. Cowley, T. Boyer, and S. Wijffels, 2014: Time, probe type, and temperature variable bias corrections to historical expendable bathythermograph observations. J. Atmos. Oceanic Technol., 31, 17931825, https://doi.org/10.1175/JTECH-D-13-00197.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Cheng, L., and et al. , 2016: XBT Science: Assessment of instrumental biases and errors. Bull. Amer. Meteor. Soc., 97, 924933, https://doi.org/10.1175/BAMS-D-15-00031.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Cheng, L., H. Luo, T. Boyer, R. Cowley, J. Abraham, V. Gouretski, F. Reseghetti, and J. Zhu, 2018: How well can we correct systematic errors in historical XBT data? J. Atmos. Oceanic Technol., 35, 11031125, https://doi.org/10.1175/JTECH-D-17-0122.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Cowley, R., S. Wijffels, L. Cheng, T. Boyer, and S. Kizu, 2013: Biases in expendable bathythermograph data: A new view based on historical side-by-side comparisons. J. Atmos. Oceanic Technol., 30, 11951225, https://doi.org/10.1175/JTECH-D-12-00127.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Cunningham, S. A., and et al. , 2000: RRS Discovery cruise 242, 07 Sep–06 Oct 1999. Atlantic–Norwegian Exchanges. Southampton Oceanography Centre Cruise Rep. 28, 128 pp.

  • Dammann, P., 1982: Acoustic tracking of expendable bathythermographs. J. Acoust. Soc. Amer., 72, S38, https://doi.org/10.1121/1.2019863.

  • Daubin, S. C., 1973: Church Gabbro technical note: Systems description and performance. University of Miami RSMAS Tech. Rep. AD-763460, 160 pp.

  • Demeo, R. P., 1969: The validity of expendable bathythermograph measurements. The Decade Ahead, 1970–1980, Marine Technology Society, 155–179.

  • Denner, W., 1969: Separation of the residual instrument noise from the significant variability for expendable bathythermographs. Proceedings of the Fourth National ISA Marine Sciences Instrumentation Symposium, F. Alt, Ed., Marine Sciences Instrumentation, Vol. 4, Plenum Press, 635–641.

  • Di Nezio, P. N., and G. Goni, 2011: Direct evidence of a changing fall-rate bias in XBTs manufactured during 1986–2008. J. Atmos. Oceanic Technol., 28, 15691578, https://doi.org/10.1175/JTECH-D-11-00017.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Emery, W. J., W. Lee, W. Zenk, and J. Meincke, 1986: A low-cost digital XBT system and its application to the real-time computation of dynamic height. J. Atmos. Oceanic Technol., 3, 7583, https://doi.org/10.1175/1520-0426(1986)003<0075:ALCDXS>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Fahrbach, E., W. Krauss, J. Meincke, and A. Sy, 1984: Nordostatlantik ’83. Christian-Albrechts-Universität Institut Für Meereskunde Data Rep. 134, 65 pp.

  • Fenner, D. F., and W. J. Cronin Jr., 1978: Bearing stake exercise: Sound speed and other environmental variability (U). Naval Oceanographic Laboratory Ocean Acoustics Division Rep. NORDA-18, 73 pp.

  • Fenner, D. F., K. W. Lackie, B. A. Watrous, and L. A. Banchero, 1974: IOMEDEX sound velocity analysis and environmental data summary. NOO Tech. Rep. TR-244, 67 pp.

  • FICARAM Group, 2013: FICARAM-15 cruise report 20th March–22nd May 2013 on board BIO Hespérides. 54 pp.

  • Flierl, G., and A. Robinson, 1974: XBT-CTD intercomparison. Instrument description and intercomparison: Report of the MODE-I Intercomparison Group, 137–139.

  • Flierl, G., and A. Robinson, 1977: XBT measurements of thermal gradients in the MODE eddy. J. Phys. Oceanogr., 7, 300302, https://doi.org/10.1175/1520-0485(1977)007<0300:XMOTGI>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Fofonoff, N. P., and R. C. Millard Jr., 1983: Algorithms for computation of fundamental properties of seawater. UNESCO Technical Papers in Marine Science 44, 54 pp., http://unesdoc.unesco.org/images/0005/000598/059832EB.pdf.

  • Fonteles, C. S., and M. M. Mata, 2009: Estimativas de erros em medidas de XBT (expendable bathythermograph) no Atlântico Sudoeste. Universidade Federal do Rio Grande VIII Mostra de Produção Universitária, 3 pp.

  • Francis, S. A., and G. C. Campbell, 1965: A low cost expendable bathythermograph. Proceedings of the Third National Marine Sciences Symposium, W. C. Knopf and H. A. Cook, Eds., Marine Sciences Instrumentation, Vol. 3, Plenum Press, 85–89.

  • Georgi, D. T., J. P. Dean, and J. A. Chase, 1979: XBT probe-to-probe thermistor temperature variability. POLYMODE News, No. 71, WHOI, Woods Hole, MA, 1, 7–10.

  • Georgi, D. T., J. P. Dean, and J. A. Chase, 1980: Temperature calibration of expendable bathythermographs. Ocean Eng., 7, 491499, https://doi.org/10.1016/0029-8018(80)90048-7.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Goni, G. J., and et al. , 2010: The Ship of Opportunity Program. Proceedings of the OceanObs’09: Sustained Ocean Observations and Information for Society, J. Hall, D. E. Harrison, and D. Stammer, Eds., Vol. 2, ESA Publ. WPP-306, https://doi.org/10.5270/OceanObs09.cwp.35.

    • Crossref
    • Export Citation
  • Good, S. A., 2011: Depth biases in XBT data diagnosed using bathymetry data. J. Atmos. Oceanic Technol., 28, 287300, https://doi.org/10.1175/2010JTECHO773.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Gould, W. J., 1991: RRS Charles Darwin cruise 50, 29 June–22 July 1990; Oceanography of the Iceland Basin Oceanography of the Iceland Basin: The fate of Iceland and Scotland overflow water. IOS Deacon Laboratory Cruise Rep. 221, 41 pp.

  • Gould, W. J., R. J. Bailey, and M. Szabados, 1990: Errors in XBT probes. International WOCE Newsletter, No. 10, WOCE International Project Office, Southampton, United Kingdom, 10–11.

  • Gouretski, V., and K. P. Koltermann, 2007: How much is the ocean really warming? Geophys. Res. Lett., 34, L01610, https://doi.org/10.1029/2006GL027834.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Gouretski, V., F. Reseghetti, S. Kizu, S. Wijffels, G. Goni, P. DiNezio, and J. Trinanes, 2010: XBT Bias and Fall Rate Workshop: Summary report. University of Hamburg, KlimaCampus, 14 pp., https://icdc.cen.uni-hamburg.de/fileadmin/user_upload/icdc_Dokumente/xbt_ws_presentations/xbt_workshop_summary_report_final.pdf.

  • Gouretski, V., and F. Reseghetti, 2010: On depth and temperature biases in bathythermograph data: Development of a new correction scheme based on analysis of a global ocean database. Deep-Sea Res. I, 57, 812833, https://doi.org/10.1016/j.dsr.2010.03.011.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Green, A. W., 1984: Bulk dynamics of the expendable bathythermograph (XBT). Deep-Sea Res., 31A, 415426, https://doi.org/10.1016/0198-0149(84)90093-1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Grilli, F., and N. Pinardi, 1998: The computation of Rossby radii of deformation for the Mediterranean Sea. MTP News, Vol. 6, No. 4, University of Barcelona, Barcelona, Spain, 4–5.

  • Grilli, F., and N. Pinardi, 1999: Le cause dinamiche della stratificazione verticale nel Mediterraneo. ISAO Tech. Rep. ISAO-TR-3/99, 132 pp.

  • Hallock, Z. R., and W. J. Teague, 1992: The fall-rate of the T7 XBT. J. Atmos. Oceanic Technol., 9, 470483, https://doi.org/10.1175/1520-0426(1992)009<0470:TFROTT>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Hamon, M., G. Reverdin, and P. Y. Le Traon, 2012: Empirical correction of XBT data. J. Atmos. Oceanic Technol., 29, 960973, https://doi.org/10.1175/JTECH-D-11-00129.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Hanawa, K., and H. Yoritaka, 1987: Detection of systematic errors in XBT data and their correction. J. Oceanogr. Soc. Japan, 43, 6876, https://doi.org/10.1007/BF02110635.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Hanawa, K., and Y. Yoshikawa, 1991: Re-examination of the depth error in XBT data. J. Atmos. Oceanic Technol., 8, 422429, https://doi.org/10.1175/1520-0426(1991)008<0422:ROTDEI>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Hanawa, K., P. Rual, R. Bailey, A. Sy, and M. Szabados, 1995: A new depth time equation for Sippican or TSK T-7, T-6 and T-4 expendable bathythermographs (XBT). Deep-Sea Res. I, 42, 14231451, https://doi.org/10.1016/0967-0637(95)97154-Z.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Heinmiller, R., C. Ebbesmeyer, B. Taft, T. Olson, and O. Nikitin, 1983: Systematic errors in expendable bathythermograph (XBT) profiles. J. Oceanogr., 65, 287299.

    • Search Google Scholar
    • Export Citation
  • Hill, P. J., 1995: Swath-mapping ADEDAV survey by RV L’Atalante from Adelaide to Davao along the continental margin of Western Australia, 1994: Post-cruise and ADEDAV/TRANSNOR data processing report. Australian Geological Survey Organization Record 1995/55, 40 pp.

  • IGOSS, 1972: Manual on data acquisition for IGOSS. UNESCO/IOC, 2nd Draft, 250 pp.

  • IOC, 1992: Ad hoc meeting of the IGOSS Task Team on quality control for automated systems, Marion, Massachusetts, USA, 3–6 June 1991. Intergovernmental Oceanographic Commission IOC/INF-888, 144 pp.

  • IOC, 1995: Third session of the Task Team on Quality Control Procedures for Automated Systems (TT/CAS), Ottawa, Canada, 23–25 October 1995. Intergovernmental Oceanographic Commission IOC/INF-1017, 34 pp.

  • IPCC, 2013: Climate Change 2013: The Physical Science Basis. Cambridge University Press, 1535 pp., https://doi.org/10.1017/CBO9781107415324.

    • Crossref
    • Export Citation
  • Ishii, M., and M. Kimoto, 2009: Reevaluation of historical ocean heat content variations with time-varying XBT and MBT depth bias corrections. J. Oceanogr., 65, 287299, https://doi.org/10.1007/s10872-009-0027-7.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • JCGM, 2008: Evaluation of measurement data—Guide to the expression of uncertainty in measurement. Joint Committee for Guides in Meteorology Working Group 1 Rep. 100:2008, 120 pp.

  • Johnson, W. P., and R. E. Lange, 1979: Rapid sampling of temperature and temperature gradient using XBT’s. SIO Reference Series 79-4, 39 pp.

  • Kennelly, M.A., M.D. Prater and T.B. Sanford, 1989: XBT and XSV data from the Gulf of Cadiz expedition: R/V Oceanus cruise 202. University of Washington APL Tech. Rep. APL-UW TR 8920, 217 pp.

    • Crossref
    • Export Citation
  • King, B. A., 1991: Circulation and structure of the Bay of Biscay and north east Atlantic out to 20°W and 41°N. RRS Discovery Cruise 189, 09 Mar.–08 Apr. 1990, Institute of Oceanographic Sciences Deacon Laboratory Cruise Rep. 225-1991, 45 pp.

  • Kizu, S., and K. Hanawa, 2002a: Start-up transient of XBT measurement. Deep-Sea Res. I, 49, 935940, https://doi.org/10.1016/S0967-0637(02)00003-1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Kizu, S., and K. Hanawa, 2002b: Recorder-dependent temperature error of expendable bathythermograph. J. Oceanogr., 58, 469476, https://doi.org/10.1023/A:1021261214950.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Kizu, S., H. Yoritaka, and K. Hanawa, 2005a: A new fall-rate equation for T-5 expendable bathythermograph (XBT) by TSK. J. Oceanogr., 61, 115121, https://doi.org/10.1007/s10872-005-0024-4.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Kizu, S., S. Ito, and T. Watanabe, 2005b: Inter-manufacturer difference and temperature dependency of the fall-rate of T-5 expendable bathythermograph. J. Oceanogr., 61, 905912, https://doi.org/10.1007/s10872-006-0008-z.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Kizu, S., C. Sukigara, and K. Hanawa, 2011: Comparison of the fall rate and structure of recent T-7 XBT manufactured by Sippican and TSK. Ocean Sci., 7, 231244, https://doi.org/10.5194/os-7-231-2011.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Koso, Y., H. Ishii, M. Fujita, and H. Kato, 2005: Application of the new depth conversion formula of XBT (T-5). Tech. Bull. Hydrogr. Oceanogr., 23, 8992.

    • Search Google Scholar
    • Export Citation
  • Levitus, S., J. I. Antonov, T. P. Boyer, R. A. Locarnini, H. E. Garcia, and A. V. Mishonov, 2009: Global ocean heat content 1955–2008 in light of recently revealed instrumentation problems. Geophys. Res. Lett., 36, L07608, https://doi.org/10.1029/2008GL037155.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Little, A. D., 1965: Experimental evaluation of expendable bathythermographs. Dept. of the Navy Bureau of Ships Rep. 4071165, 51 pp.

  • Little, A. D., 1966: Expendable bathythermograph (XBT) system evaluation for tactical sonar application. Dept. of the Navy Naval Ship Systems Command Rep. 4150866, 85 pp.

  • Magruder, P. M., Jr., 1970: Some characteristics of temperature microstructure in the ocean. M.S. thesis, Dept. of Oceanography, Naval Postgraduate School, 155 pp.

    • Crossref
    • Export Citation
  • McDowell, S., 1977: A note on XBT accuracy. POLYMODE News, No. 29, WHOI, Woods Hole, MA, 1, 4.

  • McDowell, S., 1978: A cautionary note on T-5 XBTS. POLYMODE News, No. 58, WHOI, Woods Hole, MA, 4.

  • Meincke, J., 1991: WHP cruise summary information of section A01E. WOCE, 76 pp.

  • Mied, R. P., G. J. Lindemann, and A. F. Schuetz, 1981: The hydrography and dynamics of the FREDDEX eddy. Naval Research Laboratory Memo. Rep. NRL-MR-4603, 35 pp.

  • Miura, T., M. Konda, T. Takikawa, and H. Ichikawa, 2004: Depth error in time-depth equations of the T5 XBT probes. JAMSTECR, 49, 7380.

    • Search Google Scholar
    • Export Citation
  • Newman, F. C., and P. Dammann, 1983: Acoustic measurement of XBT fall rates. SAI Tech. Rep. SAI-83/1220, 35 pp.

  • Paden, C. A., and M. C. Hendershott, 1986: Observations of temperature finestructure in the Gulf of California—XBT data report November 1984/March 1985. SIO Reference Series 86-14, 275 pp.

  • Paillet, J., 2001: Report de Campagne. Campagne POMME 1. Service Hydrographique et Océanographique de la Marine 065 MOA/NP, 20 pp.

  • Plessey, 1967: Plessey–Sippican: The only system that lets you make a bathythermograph plot at 30 knots. Int. Hydrogr. Rev., 44 (1), 1718.

    • Search Google Scholar
    • Export Citation
  • Raiteri G., A. Bordone, T. Ciuffardi, and F. Pennecchi, 2018: Uncertainty evaluation of CTD measurements: A metrological approach to water-column coastal parameters in the Gulf of La Spezia area. Measurement, 126, 156163, https://doi.org/10.1016/j.measurement.2018.05.058.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Reid, W. L., Jr., 1964: Expendable bathythermograph evaluation. U.S. Naval Oceanographic Office Informal Manuscript Rep. NOO-IM-I-1-64, 70 pp.

  • Reseghetti, F., M. Borghini, and G. M. R. Manzella, 2007: Factors affecting the quality of XBT data—Results of analyses on profiles from the Western Mediterranean Sea. Ocean Sci., 3, 5975, https://doi.org/10.5194/os-3-59-2007.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Reverdin, G., F. Marin, B. Bourles, and P. L’Herminier, 2009: XBT temperature errors during French research cruises (1999–2007). J. Atmos. Oceanic Technol., 26, 24622473, https://doi.org/10.1175/2009JTECHO655.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Ribeiro, N., M. M. Mata, J. L. L. de Azevedo, and M. Cirano, 2018: An assessment of the XBT fall-rate equation in the Southern Ocean. J. Atmos. Oceanic Technol., 35, 911926, https://doi.org/10.1175/JTECH-D-17-0086.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Ridgway, K., R. Bailey, and R. Coleman, 1999: Tasman-Coral Sea mass and heat transport/ satellite verification. National Facility Oceanographic Research Vessel, CSIRO Marine Research Cruise Summary RV Franklin FR 02/99, 18 pp.

  • Roemmich, D., and B. Cornuelle, 1987: Digitization and calibration of the expendable bathy thermograph. Deep-Sea Res., 34A, 299307, https://doi.org/10.1016/0198-0149(87)90088-4.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Russell, J. S., and K. M. Leavitt, 1984: Development of a 2000 meter aircraft expendable bathythermograph. Sippican Ocean Systems, Inc., Final Rep. R-1435B, 17 pp.

    • Crossref
    • Export Citation
  • Saur, J. F. T., and D. D. Stewart, 1967: Expendable bathythermograph data on subsurface thermal structure in the eastern North Pacific Ocean. U.S. Fish and Wildlife Service Special Scientific Rep.—Fisheries 548, 70 pp.

  • SBE, 2002: Conversion of pressure to depth. Sea-Bird Electronics Application Note 69, 1 pp.

  • Seaver, G. A., and S. Kuleshov, 1979: XBT accuracy. POLYMODE News, No. 72, WHOI, Woods Hole, MA, 1, 5–9.

  • Seaver, G. A., and S. Kuleshov, 1982: Experimental and analytical error of the expendable bathythermograph. J. Phys. Oceanogr., 12, 592600, https://doi.org/10.1175/1520-0485(1982)012<0592:EAAEOT>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Singer, J. J., 1990: On the error observed in electronically digitized T-7 XBT data. J. Atmos. Oceanic Technol., 7, 603611, https://doi.org/10.1175/1520-0426(1990)007<0603:OTEOIE>2.0.CO;2.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Sippican Corporation, 1971: R-603G Instruction manual for the expendable bathythermograph. Sippican Corporation Doc.

  • Stark, J., J. Gorman, M. Hennessey, F. Reseghetti, J. Willis, J. Lyman, J. Abraham, and M. Borghini, 2011: A computational method for determining XBT depths. Ocean Sci., 7, 733743, https://doi.org/10.5194/os-7-733-2011.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Stegen, G. R., D. P. Delisi, and R. C. Van Colln, 1975: A portable, digital recording, expendable bathythermograph (XBT) system. Deep-Sea Res. Oceanogr. Abstr., 22, 447453, https://doi.org/10.1016/0011-7471(75)90067-4.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Sy, A., 1989: Summary about preliminary results from studies of depth fall rate errors of “Deep-Blue” probes. Integrated Global Ocean Services System (IGOSS) Summary of Ship-of-Opportunity Programmes and Technical Reports, Intergovernmental Oceanographic Commission IOC/INF-804, 185–192.

  • Sy, A., 1991: XBT measurements. WOCE Operations Manual, WHP Operation and Methods, WHP91-1, WOCE Rep. 68/91, 1–19.

  • Sy, A., 1992: Report on field tests in 1990 on the evaluation of the XBT depth fall rate equation. Ad hoc meeting of the IGOSS Task Team on Quality Control for Automated Systems, Marion, Massachusetts, USA, 3–6 June, 1991, Intergovernmental Oceanographic Commission IOC/INF-888, 121–130.

  • Sy, A., J. Ulrich, and M. Stolley, 2000: Status of ship-of-opportunity activities in Germany 1999. JCOMM Ship-of-Opportunity Programme Implementation Panel—Third Session, La Jolla, CA, USA, 28–31 March 2000: SOOP status reports, SOOP scientific and technical developments, April 2000, WMO/TD-1005, JCOMM Tech. Rep. 3, 50–55.

  • Szabados, M. W., 1992: Evaluation of the expendable bathythermographic (XBT) fall rate equation. Ad hoc meeting of the IGOSS Task Team on Quality Control for Automated Systems, Marion, Massachusetts, USA, 3–6 June, 1991, Intergovernmental Oceanographic Commission IOC/INF-888, 19–97.

  • Szabados, M. W., and D. Wright, 1989: Field evaluation of real-time XBT systems. Proceedings of the Western Pacific International Meeting and Workshop on TOGA COARE, J. Picaut, R. Lukas, and T. Delcroix, Eds., ORSTOM, 811–821.

  • Thadathil, P., A. K. Saran, V. V. Gopalakrishna, P. Vethamony, N. Araligidad, and R. Bailey, 2002: XBT fall rate in waters of extreme temperature: A case study in the Antarctic Ocean. J. Atmos. Oceanic Technol., 19, 391396, https://doi.org/10.1175/1520-0426-19.3.391.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • USNAVO, 1975: Instruction manual for obtaining oceanographic data. U.S. Naval Oceanographic Office Publ. 607, 3rd ed. 232 pp.

  • Volkmann, G., 1977: Four short XBT sections across Mid-Atlantic ridge. POLYMODE News, No. 33, WHOI, Woods Hole, MA, 3.

  • Wannamaker, B., 1980: XBT measurements near the sea surface: Considerations for satellite IR comparisons and data bases. Saclant ASW Research Centre Memo. SM-132, 13 pp.

  • Wijffels, S. E., J. Willis, C. Domingues, P. Barker, N. White, A. Gronell, K. Ridgway, and J. Church, 2008: Changing expendable bathythermograph fall rates and their impact on estimates of thermosteric sea level rise. J. Climate, 21, 56575672, https://doi.org/10.1175/2008JCLI2290.1.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Zanasca, P., 1994: On board XBTs calibration. NATO Undersea Research Centre Internal Notes, 17 pp.

Save