In the past two decades, the Argo Program has collected, processed and distributed over two million vertical profiles of temperature and salinity from the upper two kilometers of the global ocean. A similar number of subsurface velocity observations near 1000 dbar have also been collected. This paper recounts the history of the global Argo Program, from its aspiration arising out of the World Ocean Circulation Experiment, to the development and implementation of its instrumentation and telecommunication systems, and the various technical problems encountered. We describe the Argo data system and its quality control procedures, and the gradual changes in the vertical resolution and spatial coverage of Argo data from 1999 to 2019. The accuracies of the float data have been assessed by comparison with high-quality shipboard measurements, and are concluded to be 0.002°C for temperature, 2.4 dbar for pressure, and 0.01 PSS-78 for salinity, after delayed-mode adjustments. Finally, the challenges faced by the vision of an expanding Argo Program beyond 2020 are discussed.

We report on profiling float technology used in the Southern Ocean Carbon and Climate Observations and Models (SOCCOM) program, a 6 year study of the interaction of ocean physics and the carbon cycle in the Southern Ocean. A central part of this program is to produce and deploy 200 profiling floats equipped with CTD units and chemical sensors capable of measuring dissolved oxygen, nitrate, pH, chlorophyll fluorescence, and particulate backscatter. The performance of the first 63 floats deployed in SOCCOM is examined, and examples of the design criteria used in producing these floats are shown. Some of the sensors require surface measurements to be made in the dark at regular intervals, and the probability of ascending to the sea surface in the dark is estimated as a function of year-day and latitude. An energy budget derived from laboratory measurements shows that only about 25% of the total energy stored in the batteries is used by the biogeochemical sensors, which bodes well for the long-term survivability of the floats. The ice-avoidance algorithm is discussed in detail, and it is shown that it is working as designed and allowing unprecedented numbers of profiles to be collected beneath the wintertime ice cover. The overall reliability of the first group of SOCCOM floats is compared with a much larger ensemble of Argo floats; the results show that the SOCCOM floats are surviving at a rate similar to the Argo floats, which have been shown to have lifetimes in excess of 5 years.

This .zip archive contains Quality Controlled float data for biogeochemical profiling floats deployed by the Southern Ocean Carbon and Climate Observation and Modeling (SOCCOM) program. Data were processed by the SOCCOM data management team at the Monterey Bay Aquarium Research Institute (MBARI). Certain early archives may contain additional University of Washington/MBARI floats deployed outside of the SOCCOM array. FORMAT: The ascii files contained herein were formatted to be compatible with Ocean Data View (ODV). ODV is freely available at https://odv.awi.de/. In addition, a Matlab function has been provided in each .zip archive for parsing the .txt files into data structures within Matlab (see get_FloatViz_data.m). Please note that the data files within this .zip archive represent a snapshot of all SOCCOM float data processed at MBARI as of the date listed in the file name. Therefore, be aware that processing updates (** AND THUS CHANGES TO THE DATA **) may have occurred since the time the snapshot was created. For the most up-to-date files (processed every 4 hours), visit ftp://ftp.mbari.org/pub/SOCCOM/FloatVizData/ or http://www.mbari.org/science/upper-ocean-systems/chemical-sensor-group/floatviz/ ALTERNATE FORMATS: NetCDF and Matlab Matlab and NetCDF formatted files are provided for each ODV text file. The Matlab format is loaded as structure, FloatViz, with the ODV parameter names as the structure's fieldnames. The NetCDF format is similar to ARGO Float NetCDF format in its structure. The parameter names, however, match the ODV text parameter names. In addition to the quality control flag strings that ARGO profiles use, an array of quality control flags is provided in the NetCDF files for programming convenience. QUALITY CONTROL DOCUMENTATION: Quality control (QC) of SOCCOM float data is performed routinely by SOCCOM data managers at MBARI. QC notification emails are currently being sent out on a monthly (or bi-monthly) basis to inform users of any recent updates to QC and/or sensor calibrations for specific floats. Comments on recent processing updates are also included. All QC emails as of the date of this snapshot are included in each downloadable zip file. PARAMETERS: For information pertaining to float identification, sensor arrays, data parameters, and quality control please refer to descriptions within the file headers. Snapshots created after Jan 01, 2017 include estimated total alkalinity and derived carbon parameters for DIC and pCO2 using one of three algorithms (LIAR, MLR, or CANYON). Floats without a pH sensor will not have these additional parameters within their respective data files. See file headers for details. Additionally, files located at the urls listed above will include carbon parameters derived using observed pH and total alkalinity estimated by the LIAR method. RESOLUTION: This archive contains either low resolution or high resolution data. The format is defined by the folder name: SOCCOM_LoResQC_METHOD_ddmmmyyyy or SOCCOM_HiResQC_METHOD_ddmmmyyyy (where METHOD = LIAR, MLR, or CANYON). Note that for APEX floats, the low resolution files only report data at depths where biogeochemical sensors sample, while the high resolution files merge this low resolution data with higher resolution pressure, temperature and salinity data (sampled every two meters in the upper 1000 meters). Be aware that, due to the merging of the two separate sampling schemes by interleaving the LowRes samples into the HiRes sample structure, HiRes files could potentially contain separate sets of samples with duplicate pressure values. For NAVIS floats all biogeochemical sensors except nitrate are sampled every 2 meters in the upper 1000 m. For NAVIS floats all data is contained in the LoResQC files (no HiResQC files exist). DISCLAIMER: These data are provided as-is. We do our best to provide high-quality, complete data but make no guarantees as to the presence of errors within the data themselves or the algorithms used in the generation of derived parameters. It is the user's responsibility to ensure that the data meets the user's needs. However, please report any observed discrepancies in the data to the contact listed below and we will do our best to fix them. CONTACT: Please report any discrepancies, problems or concerns to the following and include FLOATVIZ SNAPSHOT PROCESSING in the subject line of the email. Tanya Maurer tmaurer@mbari.org Josh Plant jplant@mbari.org SOCCOM Data Management MBARI 7700 Sandholdt Road Moss Landing, CA 95039

In seasonal ice zones (SIZs), such as the one of the Greenland Sea, the sea ice growth in winter and subsequent melting in summer influence the phytoplankton activity. However, studies assessing phytoplankton activities over complete annual cycles and at a fine temporal resolution are lacking in this environment. Biogeochemical-Argo floats, which are able to sample under the ice, were used to collect physical and biogeochemical data along vertical profiles and at 5-day resolution during two complete annual cycles in the Greenland Sea SIZ. Three phytoplankton activity phases were distinct within an annual cycle: one under ice, a second at the ice edge, and a third one around an open-water subsurface chlorophyll maximum. As expected, the light and nitrate availabilities controlled the phytoplankton activity and the establishment of these phases. On average, most of the annual net community production occurred equally under ice and at the ice edge. The open-water subsurface chlorophyll maximum phase contribution, on the other hand, was much smaller. Phytoplankton biomass accumulation and production thus occur over a longer period than might be assumed if under ice blooms were neglected. This also means that satellite-based estimates of phytoplankton biomass and production in this SIZ are likely underestimated. Simulations with the Arctic-based physical-biologically coupled SINMOD model suggest that most of the annual net community production in this SIZ results from local processes rather than due to advection of nitrate from the East Greenland and Jan Mayen Currents. e annual Net Community Production occurs under ice and the other half at the ice edge

This .zip archive contains Quality Controlled float data for biogeochemical profiling floats deployed by the Southern Ocean Carbon and Climate Observation and Modeling (SOCCOM) program. Data were processed by the SOCCOM data management team at the Monterey Bay Aquarium Research Institute (MBARI). Certain early archives may contain additional University of Washington/MBARI floats deployed outside of the SOCCOM array. FORMAT: The ascii files contained herein were formatted to be compatible with Ocean Data View (ODV). ODV is freely available at https://odv.awi.de/. In addition, a Matlab function has been provided in each .zip archive for parsing the .txt files into data structures within Matlab (see get_FloatViz_data.m). Please note that the data files within this .zip archive represent a snapshot of all SOCCOM float data processed at MBARI as of the date listed in the file name. Therefore, be aware that processing updates (** AND THUS CHANGES TO THE DATA **) may have occurred since the time the snapshot was created. For the most up-to-date files (processed every 4 hours), visit ftp://ftp.mbari.org/pub/SOCCOM/FloatVizData/ or http://www.mbari.org/science/upper-ocean-systems/chemical-sensor-group/floatviz/ ALTERNATE FORMATS: NetCDF and Matlab Lynne Talley has provided alternate formats, NetCDF and Matlab, for the ODV text data. Matlab and NetCDF formatted files are provided for each ODV text file. The Matlab format is loaded as structure, FloatViz, with the ODV parameter names as the structure's fieldnames. The NetCDF format is similar to ARGO Float NetCDF format in its structure. The parameter names, however, match the ODV text parameter names. In addition to the quality control flag strings that ARGO profiles use, an array of quality control flags is provided in the NetCDF files for programming convenience. PARAMETERS: For information pertaining to float identification, sensor arrays, data parameters, and quality control please refer to descriptions within the file headers. Snapshots created after Jan 01, 2017 include estimated total alkalinity and derived carbon parameters for DIC and pCO2 using one of three algorithms (LIAR, MLR, or CANYON). Floats without a pH sensor will not have these additional parameters within their respective data files. See file headers for details. Additionally, files located at the urls listed above will include carbon parameters derived using observed pH and total alkalinity estimated by the LIAR method. RESOLUTION: This archive contains either low resolution or high resolution data. The format is defined by the folder name: SOCCOM_LoResQC_ddmmmyyy or SOCCOM_HiResQC_ddmmmyyyy. Note that for APEX floats, the low resolution files only report data at depths where biogeochemical sensors sample, while the high resolution files merge this low resolution data with higher resolution pressure, temperature and salinity data which is sampled every two meters in the upper 1000 meters. For NAVIS floats all biogeochemical sensors except nitrate are sampled every 2 meters in the upper 1000 m. For NAVIS floats all data is contained in the "LoResQC" files (no "HiResQC" files exist). DISCLAIMER: These data are provided as-is. We do our best to provide high-quality, complete data but make no guarantees as to the presence of errors within the data themselves or the algorithms used in the generation of derived parameters. It is the user's responsibility to ensure that the data meets the user's needs. However, please report any observed discrepancies in the data to the contact listed below and we will do our best to fix them. CONTACT: Please report any discrepancies, problems or concerns to the following and include "FLOATVIZ SNAPSHOT PROCESSING" in the subject line of the email. Tanya Maurer tmaurer@mbari.org Josh Plant jplant@mbari.org SOCCOM Data Management MBARI 7700 Sandholdt Road Moss Landing, CA 95039

This .zip archive contains Quality Controlled float data for biogeochemical profiling floats deployed by the Southern Ocean Carbon and Climate Observation and Modeling (SOCCOM) program. Data were processed by the SOCCOM data management team at the Monterey Bay Aquarium Research Institute (MBARI). Certain early archives may contain additional University of Washington/MBARI floats deployed outside of the SOCCOM array. FORMAT: The ascii files contained herein were formatted to be compatible with Ocean Data View (ODV). ODV is freely available at https://odv.awi.de/. In addition, a Matlab function has been provided in each .zip archive for parsing the .txt files into data structures within Matlab (see get_FloatViz_data.m). Please note that the data files within this .zip archive represent a snapshot of all SOCCOM float data processed at MBARI as of the date listed in the file name. Therefore, be aware that processing updates (** AND THUS CHANGES TO THE DATA **) may have occurred since the time the snapshot was created. For the most up-to-date files (processed every 4 hours), visit ftp://ftp.mbari.org/pub/SOCCOM/FloatVizData/ or http://www.mbari.org/science/upper-ocean-systems/chemical-sensor-group/floatviz/ ALTERNATE FORMATS: NetCDF and Matlab Lynne Talley has provided alternate formats, NetCDF and Matlab, for the ODV text data. Matlab and NetCDF formatted files are provided for each ODV text file. The Matlab format is loaded as structure, FloatViz, with the ODV parameter names as the structure's fieldnames. The NetCDF format is similar to ARGO Float NetCDF format in its structure. The parameter names, however, match the ODV text parameter names. In addition to the quality control flag strings that ARGO profiles use, an array of quality control flags is provided in the NetCDF files for programming convenience. PARAMETERS: For information pertaining to float identification, sensor arrays, data parameters, and quality control please refer to descriptions within the file headers. Snapshots created after Jan 01, 2017 include estimated total alkalinity and derived carbon parameters for DIC and pCO2 using one of three algorithms (LIAR, MLR, or CANYON). Floats without a pH sensor will not have these additional parameters within their respective data files. See file headers for details. Additionally, files located at the urls listed above will include carbon parameters derived using observed pH and total alkalinity estimated by the LIAR method. RESOLUTION: This archive contains either low resolution or high resolution data. The format is defined by the folder name: SOCCOM_LoResQC_ddmmmyyy or SOCCOM_HiResQC_ddmmmyyyy. Note that for APEX floats, the low resolution files only report data at depths where biogeochemical sensors sample, while the high resolution files merge this low resolution data with higher resolution pressure, temperature and salinity data which is sampled every two meters in the upper 1000 meters. For NAVIS floats all biogeochemical sensors except nitrate are sampled every 2 meters in the upper 1000 m. For NAVIS floats all data is contained in the LoResQC files (no HiResQC files exist). DISCLAIMER: These data are provided as-is. We do our best to provide high-quality, complete data but make no guarantees as to the presence of errors within the data themselves or the algorithms used in the generation of derived parameters. It is the user's responsibility to ensure that the data meets the user's needs. However, please report any observed discrepancies in the data to the contact listed below and we will do our best to fix them. CONTACT: Please report any discrepancies, problems or concerns to the following and include FLOATVIZ SNAPSHOT PROCESSING in the subject line of the email. Tanya Maurer tmaurer@mbari.org Josh Plant jplant@mbari.org SOCCOM Data Management MBARI 7700 Sandholdt Road Moss Landing, CA 95039

The Southern Ocean Carbon and Climate Observations and Modeling (SOCCOM) program has begun deploying a large array of biogeochemical sensors on profiling floats in the Southern Ocean. As of February 2016, 86 floats have been deployed. Here the focus is on 56 floats with quality-controlled and adjusted data that have been in the water at least 6 months. The floats carry oxygen, nitrate, pH, chlorophyll fluorescence, and optical backscatter sensors. The raw data generated by these sensors can suffer from inaccurate initial calibrations and from sensor drift over time. Procedures to correct the data are defined. The initial accuracy of the adjusted concentrations is assessed by comparing the corrected data to laboratory measurements made on samples collected by a hydrographic cast with a rosette sampler at the float deployment station. The long-term accuracy of the corrected data is compared to the GLODAPv2 data set whenever a float made a profile within 20 km of a GLODAPv2 station. Based on these assessments, the fleet average oxygen data are accurate to 1 +/- 1%, nitrate to within 0.5 +/- 0.5 mu mol kg(-1), and pH to 0.005 +/- 0.007, where the error limit is 1 standard deviation of the fleet data. The bio-optical measurements of chlorophyll fluorescence and optical backscatter are used to estimate chlorophyll a and particulate organic carbon concentration. The particulate organic carbon concentrations inferred from optical backscatter appear accurate to with 35 mg C m(-3) or 20%, whichever is larger. Factors affecting the accuracy of the estimated chlorophyll a concentrations are evaluated.

This .zip archive contains Quality Controlled float data for biogeochemical profiling floats deployed by the Southern Ocean Carbon and Climate Observation and Modeling (SOCCOM) program. Data were processed by the SOCCOM data management team at the Monterey Bay Aquarium Research Institute (MBARI). Certain early archives may contain additional University of Washington/MBARI floats deployed outside of the SOCCOM array. FORMAT: The ascii files contained herein were formatted to be compatible with Ocean Data View (ODV). ODV is freely available at https://odv.awi.de/. In addition, a Matlab function has been provided in each .zip archive for parsing the .txt files into data structures within Matlab (see get_FloatViz_data.m). Please note that the data files within this .zip archive represent a snapshot of all SOCCOM float data processed at MBARI as of the date listed in the file name. Therefore, be aware that processing updates (** AND THUS CHANGES TO THE DATA **) may have occurred since the time the snapshot was created. For the most up-to-date files (processed every 4 hours), visit ftp://ftp.mbari.org/pub/SOCCOM/FloatVizData/ or http://www.mbari.org/science/upper-ocean-systems/chemical-sensor-group/floatviz/ ALTERNATE FORMATS: NetCDF and Matlab Lynne Talley has provided alternate formats, NetCDF and Matlab, for the ODV text data. Matlab and NetCDF formatted files are provided for each ODV text file. The Matlab format is loaded as structure, FloatViz, with the ODV parameter names as the structure's fieldnames. The NetCDF format is similar to ARGO Float NetCDF format in its structure. The parameter names, however, match the ODV text parameter names. In addition to the quality control flag strings that ARGO profiles use, an array of quality control flags is provided in the NetCDF files for programming convenience. PARAMETERS: For information pertaining to float identification, sensor arrays, data parameters, and quality control please refer to descriptions within the file headers. Snapshots created after Jan 01, 2017 include estimated total alkalinity and derived carbon parameters for DIC and pCO2 using one of three algorithms (LIAR, MLR, or CANYON). Floats without a pH sensor will not have these additional parameters within their respective data files. See file headers for details. Additionally, files located at the urls listed above will include carbon parameters derived using observed pH and total alkalinity estimated by the LIAR method. RESOLUTION: This archive contains either low resolution or high resolution data. The format is defined by the folder name: SOCCOM_LoResQC_ddmmmyyy or SOCCOM_HiResQC_ddmmmyyyy. Note that for APEX floats, the low resolution files only report data at depths where biogeochemical sensors sample, while the high resolution files merge this low resolution data with higher resolution pressure, temperature and salinity data which is sampled every two meters in the upper 1000 meters. For NAVIS floats all biogeochemical sensors except nitrate are sampled every 2 meters in the upper 1000 m. For NAVIS floats all data is contained in the LoResQC files (no HiResQC files exist). DISCLAIMER: These data are provided as-is. We do our best to provide high-quality, complete data but make no guarantees as to the presence of errors within the data themselves or the algorithms used in the generation of derived parameters. It is the user's responsibility to ensure that the data meets the user's needs. However, please report any observed discrepancies in the data to the contact listed below and we will do our best to fix them. CONTACT: Please report any discrepancies, problems or concerns to the following and include FLOATVIZ SNAPSHOT PROCESSING in the subject line of the email. Tanya Maurer tmaurer@mbari.org Josh Plant jplant@mbari.org SOCCOM Data Management MBARI 7700 Sandholdt Road Moss Landing, CA 95039

Six profiling floats equipped with nitrate and oxygen sensors were deployed at Ocean Station P in the Gulf of Alaska. The resulting six calendar years and 10 float years of nitrate and oxygen data were used to determine an average annual cycle for net community production (NCP) in the top 35 m of the water column. NCP became positive in February as soon as the mixing activity in the surface layer began to weaken, but nearly 3 months before the traditionally defined mixed layer began to shoal from its winter time maximum. NCP displayed two maxima, one toward the end of May and another in August with a summertime minimum in June corresponding to the historical peak in mesozooplankton biomass. The average annual NCP was determined to be 1.5 ± 0.6 mol C m−2 yr−1 using nitrate and 1.5 ± 0.7 mol C m−2 yr−1 using oxygen. The results from oxygen data proved to be quite sensitive to the gas exchange model used as well as the accuracy of the oxygen measurement. Gas exchange models optimized for carbon dioxide flux generally ignore transport due to gas exchange through the injection of bubbles, and these models yield NCP values that are two to three time higher than the nitrate‐based estimates. If nitrate and oxygen NCP rates are assumed to be related by the Redfield model, we show that the oxygen gas exchange model can be optimized by tuning the exchange terms to reproduce the nitrate NCP annual cycle.

Oxygen is an important tracer for biological processes in the ocean. Measuring changes in oxygen over annual cycles provides information about photosynthesis and respiration and their impact on the carbon cycle. Long-term, accurate oxygen measurements over wide areas are needed to determine changes in ocean oxygen content and oxygen deficient zones. Oxygen sensors have been increasingly mounted on Argo floats that profile between 2000 m and the surface. Most of these measurements are currently too inaccurate to calculate the air-sea gas flux, which is the dominant flux of oxygen in the surface ocean and typically driven by surface oxygen supersaturation states of only several percent. In this study, we present data from 17 Aanderaa oxygen optodes mounted on 11 Argo floats modified to make atmospheric measurements for calibration. Optodes measure oxygen equally well in air and water, allowing the use of atmospheric oxygen to perform on-going, in situ calibrations throughout the float lifetime. We find that it is necessary to make atmospheric measurements at night, that raising optodes higher into the air reduces variance in measurements, and that multiple measurements each time a float surfaces provide the best calibration data. Initial optode calibration on deployment has an average uncertainty of +/- 0.1% (1 sigma) and drift can be calculated to +/- 0.1% yr(-1). Measurable drift was determined in 10-12 optodes out of the 14 that were deployed for similar to 2 yr. The maximum drift rate measured was -0.5% yr(-1), which is large enough to strongly impact calculations of air-sea oxygen fluxes.

Reagent-free optical nitrate sensors [in situ ultraviolet spectrophotometer (ISUS)] can be used to detect nitrate throughout most of the ocean. Although the sensor is a relatively high-power device when operated continuously (7.5 W typical), the instrument can be operated in a low-power mode, where individual nitrate measurements require only a few seconds of instrument time and the system consumes only 45 J of energy per nitrate measurement. Operation in this mode has enabled the integration of ISUS sensors with Teledyne Webb Research's Autonomous Profiling Explorer (APEX) profiling floats with a capability to operate to 2000 m. The energy consumed with each nitrate measurement is low enough to allow 60 nitrate observations on each vertical profile to 1000 m. Vertical resolution varies from 5 m near the surface to 50 m near 1000 m, and every 100 m below that. Primary lithium batteries allow more than 300 vertical profiles from a depth of 1000 m to be made, which corresponds to an endurance near four years at a 5-day cycle time. This study details the experience in integrating ISUS sensors into Teledyne Webb Research's APEX profiling floats and the results that have been obtained throughout the ocean for periods up to three years.

Profiling floats with optical sensors can provide important complementary data to satellite ocean color determinations by providing information about the vertical structure of ocean waters, as well as surface waters obscured by clouds. Here we demonstrate this ability by pairing satellite ocean color data with records from a profiling float that obtained continuous, high‐quality optical data for 3 yr in the North Atlantic Ocean. Good agreement was found between satellite and float data, and the relationship between satellite chlorophyll and floatderived particulate backscattering was consistent with previously published data. Upper ocean biogeochemical dynamics were evidenced in float measurements, which displayed strong seasonal patterns associated with phytoplankton blooms, and depth and seasonal patterns associated with an increase in pigmentation per particle at low light. Surface optical variables had shorter decorrelation timescales than did physical variables (unlike at low latitudes), suggesting that biogeochemical rather than physical processes controlled much of the observed variability. After 2.25 yr in the subpolar North Atlantic between Newfoundland and Greenland, the float crossed the North Atlantic Current to warmer waters, where it sampled an unusual eddy for 3 months. This anticyclonic feature contained elevated particulate material from surface to 1000‐m depth and was the only such event in the float's record. This eddy was associated with weakly elevated surface pigment and backscattering, but depthintegrated backscattering was similar to that previously observed during spring blooms. Such seldom‐observed eddies, if frequent, are likely to make an important contribution to the delivery of particles to depth.

Ocean color, first measured from space 30 years ago, has provided a revolutionary synoptic view of near‐surface fields of phytoplankton pigments. Since 1979, a number of ocean color satellite missions have provided coverage of phytoplankton biomass and other biogeochemical variables on scales of days to years and of kilometers to ocean basin.Because of the nature of visible light and its interaction with absorbing and scattering materials in the ocean and atmosphere, these measurements are biased toward near‐surface waters and are obscured by clouds. As a consequence, ocean color satellites miss significant fractions of phytoplankton biomass, marine primary productivity, and particle flux that occur at depths beyond their sensing range. They also miss phytoplankton blooms and other events that occur during periods of extended cloud cover.

For timescales much greater than the local buoyancy period, the buoyant response of a RAFOS float is virtually dictated by its compressibility. As the compressibility of a thermally inert RAFOS float increases from zero, its oceanic equilibrium surface undergoes a smooth continuous deformation starting from an in situ density sur-face, eventually merging with an isopycnal surface and finally with a neural surface. Thus there is a continuum of operational modes available to RAFOS floats; each mode is associated with a critical compressibility. Hypothetically, the compressee that transforms an isobaric RAFOS float into an isopycnal float can be modified to make a neutral-surface drifter by altering the critical value of its compressibility. The ballast procedure used to target a float to a prescribed equilibrium surface can be viewed as an accurate (+/- 3%) laboratory measurement of the float's compressibility. For current ''isobaric'' RAFOS floats. the mean measured compressibility was approximately 2.71 X 10(-6) db-1 (i.e.. about 60% that of seawater). which can induce pressure deviations from an isobar as large as 125-175 db in the main thermoclines of the North Pacific and North Atlantic. Errors in ballasting a float yield targeting errors that depend on the float's compressibility and the local density stratification of the ocean. For isobaric floats deployed to 1000 db in the North Pacific Ocean. the targeting errors are approximately 23 db per gram of ballast error. By optimizing the ballast procedure. the ballast errors (and hence the targeting errors) can be minimized. For 59 shallow (1000 db) floats to which the optimized procedure was applied, preliminary estimates of the mean and maximum targeting errors are 25 and 50 db.