High quality fluxes from ICOADS ship reports

Marine meteorological observations routinely made at sea by Voluntary Observing Ships (VOS) form a key component of the climate record. These observations have been widely used in studies such as: validation and verification of satellite observations; validation and assimilation in atmospheric reanalysis models; national and international assessments of climate change (e.g. the Intergovernmental Panel on Climate Change, IPCC) and many others. The VOS also provide the only widely distributed source of in situ observations of all the variables needed to estimate the turbulent heat fluxes between the ocean and atmosphere. These observations have been used to derive climatologies and atlases of the surface fluxes but none of these datasets have contained associated uncertainty estimates. A new approach is presented to estimate daily and monthly fluxes from VOS reports which allows uncertainty estimates to be made for those fluxes.

The observations used are from the International Comprehensive Ocean-Atmosphere Dataset (ICOADS) which contains surface meteorological reports from VOS, moored and drifting buoys, and other ocean and coastal platforms. Daily fields of the basic variables (air temperature, humidity, sea surface temperature (SST), wind speed and pressure) are first generated from the VOS reports using an Optimal Interpolation (OI) scheme. The OI scheme takes into account the uncertainty in the individual observations and produces uncertainty estimates for the analysed fields. These daily fields of the basic variables and their uncertainties are then used to calculate the turbulent heat fluxes along with the flux uncertainties. The method used also allows an estimate of the sampling uncertainty to be made.

Initial flux fields have been developed using only VOS reports but the methodology used can incorporate any observations with known random and bias error characteristics. Including buoy observations will not significantly alter the analyses or analysis errors except in limited geographic areas or for the SST. This is because drifting buoys report only a subset of the observations needed (e.g. most often SST and much less frequently air temperature, wind speed or pressure). Moored buoy reports usually contain a wider range of the necessary variables for flux calculation, but only for a fixed point in space. Hence, drifting buoy observations only make a significant impact on the quality of SST fields, and the impact of moored buoy observations is in limited geographic areas. Excluding the buoy observations from the analysis maintains the independence of the two sources and allows for an inter-comparison and validation of the VOS fluxes and data fields with other in situ sources not included in the analysis.

In data rich regions, such as the North Atlantic, the output from the OI scheme and the associated fluxes compare favourably with similar products and independent in situ data on a daily time scale. In more data sparse regions there is still a good agreement between the OI fluxes and the other flux products on a monthly time scale. The uncertainty estimates show the decline in this important part of the in situ observing system, with the uncertainty in the basic variables and fluxes increasing by between 20 and 40 % over the last 20 years due to a decreasing number of VOS observations. Whilst including the buoy observations reduces the decline in the SST field their inclusion does not significantly reduce the decline in the quality of other fields.