Elsevier

Remote Sensing of Environment

Volume 201, November 2017, Pages 196-218
Remote Sensing of Environment

Review and assessment of latent and sensible heat flux accuracy over the global oceans

https://doi.org/10.1016/j.rse.2017.08.016Get rights and content

Highlights

  • Establishing reference input dataset maximizing the use of remotely sensed data

  • Performing a cross-comparison of different heat flux algorithms and approaches

  • Generating an ensemble of turbulent fluxes, including multiple approaches

  • Evaluating the quality and consistency of ensemble realizations

  • Exploiting integral heat constraints at local, regional and global scales

Abstract

For over a decade, several research groups have been developing air-sea heat flux information over the global ocean, including latent (LHF) and sensible (SHF) heat fluxes over the global ocean. This paper aims to provide new insight into the quality and error characteristics of turbulent heat flux estimates at various spatial and temporal scales (from daily upwards). The study is performed within the European Space Agency (ESA) Ocean Heat Flux (OHF) project. One of the main objectives of the OHF project is to meet the recommendations and requirements expressed by various international programs such as the World Research Climate Program (WCRP) and Climate and Ocean Variability, Predictability, and Change (CLIVAR), recognizing the need for better characterization of existing flux errors with respect to the input bulk variables (e.g. surface wind, air and sea surface temperatures, air and surface specific humidities), and to the atmospheric and oceanic conditions (e.g. wind conditions and sea state). The analysis is based on the use of daily averaged LHF and SHF and the associated bulk variables derived from major satellite-based and atmospheric reanalysis products. Inter-comparisons of heat flux products indicate that all of them exhibit similar space and time patterns. However, they also reveal significant differences in magnitude in some specific regions such as the western ocean boundaries during the Northern Hemisphere winter season, and the high southern latitudes. The differences tend to be closely related to large differences in surface wind speed and/or specific air humidity (for LHF) and to air and sea temperature differences (for SHF). Further quality investigations are performed through comprehensive comparisons with daily-averaged LHF and SHF estimated from moorings. The resulting statistics are used to assess the error of each OHF product. Consideration of error correlation between products and observations (e.g., by their assimilation) is also given. This reveals generally high noise variance in all products and a weak signal in common with in situ observations, with some products only slightly better than others. The OHF LHF and SHF products, and their associated error characteristics, are used to compute daily OHF multiproduct-ensemble (OHF/MPE) estimates of LHF and SHF over the ice-free global ocean on a 0.25° × 0.25° grid. The accuracy of this heat multiproduct, determined from comparisons with mooring data, is greater than for any individual product. It is used as a reference for the anomaly characterization of each individual OHF product.

Introduction

Accurate estimation of the ocean surface turbulent and radiative fluxes is of great interest for a variety of air-sea interaction and climate variability issues. Surface fluxes of heat, moisture, momentum, and gases play a key role in the coupling of the Earth's climate system and control many important feedbacks between the ocean and the atmosphere (Gulev et al., 2013). Furthermore, consistency studies of turbulent flux estimates and ocean heat storage estimates are also essential for constraining the Earth's energy budget in order to “track” the energy flows through the climate system, which in turn is critical for improving understanding of the relationships between climate forcings, the Earth system responses, climate variability and future climate change (Trenberth et al., 2009, von Schuckmann et al., 2016). The longest time series of surface fluxes going back to the mid-19th century can be derived from the Voluntary Observing Ship (VOS) data (Woodruff et al., 2011, Gulev et al., 2013). However, these data are characterized by insufficient and time-dependent sampling (Gulev et al., 2007a, Gulev et al., 2007b), and by inaccuracies in state variables used for flux computation (e.g. Josey et al., 1999, Josey et al., 2014). In contrast, atmospheric re-analyses, as well as remotely sensed data, potentially provide much more homogeneous time series of atmospheric state variables for surface flux computation. However, remotely sensed data are limited in time to a few decades while reanalyses can be strongly influenced by variations in the type and amount of data assimilated, particularly across the transition to the satellite era in the early 1980s.

In addition, surface flux products from reanalyzes and remote sensing are also subject to biases and uncertainties and require further improvement for turbulent flux determination. These include; improvements in spatial and temporal resolution, the accuracy, and the characterization of the spatial and temporal distribution of errors of each flux component. It is one of the priorities of the World Climate Research Program (WCRP) to improve the accuracy of surface fluxes for climate studies to within “a few W/m2” and 10 W/m2 for individual flux components and the large scale net heat fluxes, respectively (e.g. WGASF, 2000, Bradley and Fairall, 2007). The Southern Ocean Observing System (SOOS) group recommends a better flux observation density for improving heat flux accuracies at regional scales (Gille et al., 2016). These requirements impose challenges including the development of new parameterizations, achievement of global and regional heat budget closure, reducing sampling uncertainties, and better scaling parameters for surface flux estimates.

To meet these community requirements, the European Space Agency (ESA) launched a project called Ocean Heat Flux (OHF (http://www.oceanheatflux.org/) aiming at development, validation, and evaluation of satellite-based estimates of surface turbulent fluxes and their documentation, particularly those derived from ESA satellite/mission earth observation (EO) data, as well as all bulk parameters needed for turbulent flux calculations over the global ocean. OHF involves a number of objectives and studies. The main OHF objectives include (but are not limited to); establishing a reference surface flux dataset (to maximize the use of remotely sensed data including ESA products), development and accuracy assessment of an ensemble of ocean heat turbulent flux products available over decadal or longer timescales (in order to foster the use and validation of ESA mission data).

For these purposes, OHF uses in-situ, satellite-based, blended or synthetic, and reanalysis-derived surface fluxes over the global ocean, with synoptic and sub-synoptic spatial resolution for the period 1999–2009. The project makes use of the most modern global satellite surface flux data sets such as those from IFREMER (Institut Français pour la Recherche et l'Exploitation de la MER; France), HOAPS (the Hamburg Ocean Atmosphere Parameters and Fluxes from Satellite; Germany), SEAFLUX (Woods Hole Oceanographic Institution, Woods Hole (WHOI); USA), and J-OFURO (Japanese Ocean Flux Data sets with Use of Remote Sensing Observations; Japan). These are used along with surface turbulent fluxes from three modern reanalyzes: ERA-Interim (Dee et al., 2011), NCEP-CFSR (Saha et al., 2010) and NASA MERRA (Rienecker et al., 2011), as well as the synthetic OAFLUX product (Yu et al., 2008) and the VOS based NOCS2 surface flux climatology (Berry and Kent, 2009). Because these flux products were derived using different approaches and data sources, they all have their strengths and weaknesses. Wide use of these products for different climate applications such as (among others) forcing ocean models (e.g., Ayina et al., 2006), analyzing ENSO dynamics (Mestas-Nuñez et al., 2006, Mestas-Nuñez et al., 2013), and/or evaluating the intra-seasonal variability (Grodsky et al., 2009) requires a detailed quantitative assessment of each product's limitation and of and inter-product differences.

This study presents pilot results from the OHF project that describe uncertainties of the different flux products. Such intercomparison is supplemented by the validation of individual surface flux components against estimates based on in-situ buoy and ship data, especially buoy data included in the Flux reference OceanSites network (http://www.oceansites.org/). Consideration is also given to a new approach to using observations that are themselves incorporated into the flux products that are being validated.

The datasets used in this study are described in Section 2, while OHF products, all available at the same space and time resolution, are described in Section 3. Section 4 demonstrates the impact of recalibration on each OHF product. Regional product inter-comparisons are introduced in Section 5. The accuracy and quality of each OHF flux product, and the ensemble mean flux product, is discussed in 6 Buoy comparisons, 7 Ensemble versus standardized products.

Section snippets

IFREMER

In this study, we use the new IFREMER turbulent fluxes (version 4) available daily over the global ocean on a 0.25° regular grid. It is an updated version of (Bentamy et al., 2013). The bulk variables such as surface wind speed (U10) and specific air humidity (qa) at 10 m height are estimated from remotely sensed observations. U10 is mainly obtained from scatterometers onboard ERS-1 (1992–1996), ERS-2 (1996–2001), and QuikSCAT (1999–2009) satellites. More specifically, the main change with

Standardized flux products

Table 1 provides the spatial and temporal resolution characteristics of flux products used in this study. The spatial resolution of products varies from 0.25 to 1° and the highest temporal resolution varies from 3 hourly to daily. For further intercomparisons, we interpolated all products onto a standard 0.25° grid and at daily time resolution.

Each flux product listed in Table 1 is interpolated onto the same regular 0.25° latitude/longitude grid, using two methods, namely spline interpolation

Inter-comparisons

The nine LHF products exhibit quite similar latitudinal variations in the zonal means, averaged over the Atlantic, Indian, and Pacific Oceans (Fig. 5). Notably, all standardized LHF products, including MERRA, are within one STD from OHF/MPE. For the three basins, a local minimum in LHF is present near the equator due to the combination of low wind speed and relatively small range of surface humidity departures from saturation. This equatorial minimum is apparent in the Atlantic and Pacific

Product calibration considering correlated errors

Spatial and temporal coverage of the intersection of two (or more) datasets can be orders of magnitude smaller than the coverage of just one gridded dataset. In this sense, collocations only allow one to infer the bias and performance of a full dataset and sometimes such inferences may be lacking (cf., Josey et al., 2014). However, it is common for gridded products to benefit from observations in assimilation windows that are typically as large or larger, than the grid interval on which a true

Statistical results

The statistics aiming at the characterization of comparisons between buoy and flux products (and the associated bulk variables) are determined from collocated buoy (see Section 2.9) and product data. Daily fluxes for each product are collocated in space with buoy estimates. The collocation criterion separating buoy and product is that the distance should be less than the product spatial resolution (Table 1). For the standardized products, the spatial criterion is 25 km. The statistics are

Ensemble versus standardized products

The results of the buoy data comparisons indicate that OHF/MPE is more accurate than any of the contributing products. Hence, it is now employed for the characterization of the spatial and temporal errors of each standardized product.

The evaluation is first performed over global oceans for the period 2000 through 2007. Mean and the associated STD characterizing the difference between OHF/MPE and each product (in this order) are shown in Fig. 11 for LHF and SHF. About 95% of LHF (resp. SHF) mean

Probability distribution results

For the validation of different surface turbulent flux products against buoy measurements we also applied an approach developed by Gulev and Belyaev (2012) focused on the analysis of probability distributions of surface turbulent fluxes. In this approach, probability distributions of surface turbulent heat fluxes are approximated by the 2-parameteric MFT (Modified Fisher-Tippett) distribution which allows for the analysis of the probability density functions (PDFs) and high percentiles of

Summary and conclusion

Over the last twenty years, there have been various attempts to estimate accurate LHF and SHF over the global ocean with high space and time resolution. LHF and SHF are estimated based on the use the aerodynamic bulk approach requiring the knowledge of variables such as surface wind speed, specific air and surface humidities, and air and surface temperatures. The LHF and SHF characteristics taking into account space and time resolutions may lead to significant differences between available heat

Acknowledgments

The authors are grateful to ESA, EUMETSAT, CERSAT, JPL, ECMWF, NCEP, NASA, Météo-France, NDBC, PMEL, and UK MetOffice for providing numerical, satellite, and in-situ data used in this study. We would like to thank D. Croizé-Fillon and IFREMER/CERSAT team for data processing support. The authors would also like to thank the strong support of the reviewers to improve this study. This study is supported by the ESA Support to Science Element (STSE) program under contract 4000111424/14/I-AM. SKG

References (63)

  • A. Gruber et al.

    Recent advances in (soil moisture) triple collocation analysis

    Int. J. Appl. Earth Obs. Geoinf.

    (2016)
  • P. Schlussel et al.

    Retrieval of latent-heat flux and longwave irradiance at the sea-surface from SSM/I and AVHRR measurements

    Adv. Space Res.

    (1995)
  • H. Akima

    1970: a new method of interpolation and smooth curve fitting based on local procedures

    J. ACM

    (October 1970)
  • E. Andersson et al.

    Variational quality control

    Q. J. R. Meteorol. Soc.

    (1999)
  • A. Andersson et al.

    The Hamburg ocean atmosphere parameters and fluxes from satellite data - HOAPS-3

    Earth Syst. Sci. Data Discuss.

    (2010)
  • A. Andersson et al.

    Evaluation of HOAPS-3 ocean surface freshwater flux components

    J. Appl. Meteorol. Climatol.

    (2011)
  • R. Atlas et al.

    A cross-calibrated, multiplatform ocean surface wind velocity product for meteorological and oceanographic applications

    Bull. Am. Meteorol. Soc.

    (2011)
  • L.H. Ayina et al.

    The impact of satellite winds and latent heat fluxes in a numerical simulation of the tropical Pacific Ocean

    J. Clim.

    (2006)
  • A.C.M. Beljaars

    The parametrization of surface fluxes in large-scale models under free convection. Q.J.R

    Meteorol. Soc.

    (1995)
  • A. Bentamy et al.

    Gridded surface wind fields from Metop/ASCAT measurements

    Int. J. Remote Sens.

    (2011)
  • A. Bentamy et al.

    Satellite estimates of wind speed and latent heat flux over the global oceans

    J. Clim.

    (2003)
  • A. Bentamy et al.

    Improvement in air–sea flux estimates derived from satellite observations

    Int. J. Remote Sens.

    (2013)
  • A. Bentamy et al.

    Homogenization of scatterometer wind retrievals

    Int. J. Climatol.

    (2016)
  • D. Berry et al.

    A new air-sea interaction gridded data set from ICOADS with uncertainty estimates

    Bull. Am. Meteorol. Soc.

    (2009)
  • M.G. Bosilovich

    NASA's modern era retrospective analysis for research and applications: Integrating Earth observations. Earthzine

  • E.F. Bradley et al.

    A Guide to Making Climate Quality Meteorological and Flux Measurements at Sea

    (2007)
  • L. Brodeau et al.

    Climatologically significant effects of some approximations in the bulk parameterizations of turbulent air-sea fluxes

    J. Phys. Oceanogr.

    (2017)
  • K.S. Casey et al.

    The past, present and future of the AVHRR pathfinder SST program

  • H. Charnock

    Wind-stress on a water surface

    Q. J. R. Meteorol. Soc.

    (1955)
  • S.-H. Chou et al.

    Surface turbulent heat and momentum fluxes over global oceans based on the Goddard satellite retrieval, version 2 (GSSTF2)

    J. Clim.

    (2003)
  • S.-H. Chou et al.

    A comparison of latent heat fluxes over global oceans for four flux products

    J. Clim.

    (2004)
  • C.A. Clayson et al.

    SEAFLUX Version 1: a new satellite-based ocean-atmosphere turbulent flux dataset

    Int. J. Climatol.

    (2013)
  • R.E. Danielson

    Collocations of ICOADS and Ocean Heat Flux (centered five-day) latent and sensible heat flux estimates. SEANOE online data archive

  • R.E. Danielson et al.

    Exploitation of error correlation in a large analysis validation: GlobCurrent case study

    Remote Sens. Environ.

    (2017)
  • D.P. Dee et al.

    The ERA-Interim reanalysis: configuration and performance of the data assimilation system

    Q. J. R. Meteorol. Soc.

    (2011)
  • C.W. Fairall et al.

    Bulk parameterization of air–sea fluxes: updates and verification for the COARE algorithm

    J. Clim.

    (2003)
  • E. Freeman et al.

    ICOADS release 3.0: a major update to the historical marine climate record

    Int. J. Climatol.

    (2017)
  • S. Gille et al.

    New approaches for air-sea fluxes in the Southern Ocean

    Eos

    (2016)
  • S.A. Grodsky et al.

    Intraseasonal latent heat flux based on satellite observations

    J. Clim.

    (2009)
  • S.K. Gulev

    Influence of space-time averaging on the ocean-atmosphere exchange estimates in the North Atlantic mid-latitudes

    J. Phys. Oceanogr.

    (1994)
  • S.K. Gulev et al.

    Probability distribution characteristics for surface air-sea turbulent heat fluxes over the global ocean

    J. Clim.

    (2012)
  • Cited by (71)

    • Systematic improvement in simulated latent and sensible heat fluxes over tropical oceans in AMIP6 models compared to AMIP5 models with the same resolutions

      2022, Atmospheric Research
      Citation Excerpt :

      This is rather remarkable, and is partly due to better boundary conditions (e.g., sea surface temperature as shown in Fig. 8) (Eyring et al., 2016) and model physics in AMIP6 than AMIP5. Earlier studies related to flux evaluation (e.g., Cao et al., 2015; Bentamy et al., 2017; Găinuşă-Bogdan et al., 2018) could not conclude whether the improvement was due to increased resolution, model physics or model boundary conditions. Although we compared the models with same horizontal resolutions, four AMIP6 models have higher vertical resolutions compared with AMIP5 (Table 1).

    View all citing articles on Scopus
    View full text