By Victor King,2014-05-19 14:06
8 views 0
However, it is also clear from figure 2 that the quality control on the buoy observations is not adequate, as there are some matchups that are situated in



    Jacob L. Høyer

    Center for Ocean and Ice, Danish Meteorological Institute,

    Lyngbyvej 100, København Ø, Email :

    ABSTRACT 2. THE DMI ANALYSIS SYSTEM DMI is currently producing several regional level 4 Sea The DMI analysis system consists of several different Surface Temperature (SST) analysis, one for the North steps. The major tasks are: Sea/Baltic Sea area, which has been produced for the

    last 5 years, and a new one for the Arctic Ocean. The 1. Retrieval of the satellite L2P observations interpolated SST products are based upon the same

    multiplatform optimal interpolation analysis scheme 2. Pre-processing: quality control, collating the that: pre-process the available GHRSST-pp L2P SST satellite observations products, perform quality control, bias adjust,

    interpolate and generate monitoring and validation 3. Optimal Interpolation, including ice masking statistics. Local covariance statistics empirically derived

    from the satellite data, are used for the analysis and the 4. Generation of monitoring and validation performance of the two products is validated against statistics drifting buoys. Future areas of developments for the

    Arctic Ocean analysis are presented in the end. A fundamental part of the scheme is the ingestion of the

    satellite data. All the satellite observations that are used 1. INTRODUCTION to produce the SST fields are obtained via ftp from the

    GHRSST-pp project as level 2P observations (Donlon et. The Sea surface temperature (SST) is a vital parameter al, 2007). The different sensors and their characteristics that is used in both oceanic models and Numerical are listed in table 1 Weather Prediction systems. The temperature of the sea

    surface can be observed from space using both infrared Sensor Satellite Sensor Resolution and microwave sensors. The infrared sensors typically

    have the best spatial resolution and highest accuracy but AATSR ENVISAT IR 1 km they are limited by cloud cover. In the microwave part

    of the spectrum, the SST can be observed in the MODIS Aqua IR 1km

    presence of cloud cover, but the spatial resolution is not Modis Terra IR 1km as good as for the infrared satellite sensors. The

    different sampling characteristics of the satellites and AVHRR NOAA IR 2km + 9 km the model demand for high resolution SST fields 17+18

    without gaps, makes it very relevant to perform an Seviri MSG-1 IR 5 km analysis of the SST observations, whereby the different

    satellite observations are referenced to each other and AMSR-E Aqua MW 25 km interpolated to produce a daily high resolution field

    without gaps. For this purpose, DMI has developed an

    optimal interpolation algorithm that uses statistics Table 1: Satellite observations that are currently derived from the data itself to produce level 4 fields for included in the DMI level 4 analysis system for the the North Sea Baltic Sea, the Arctic Ocean and several Arctic Ocean. IR stands for infrared and MW stands for other regions around the world (see Microwave. The analysis system for

    the Arctic Ocean is used in this paper as an example but In addition to the satellite SST observations, ice the other analysis products are very similar except for information is included from the Ocean and Sea Ice the region specific statistics. SAF project. The ice product is the Northern

    Hemisphere ice edge product in a 0.1 degrees resolution.

    The individual L2P data products are being quality

    controlled, corrected for biases and averaged to the

    analysis grid before they are being used in the analysis.

    All satellite observations within 24 hours from the 3. VALIDATION AGAINST IN SITU analysis time is included in the Arctic analysis. Both OBSERVATIONS day and night time observations are included and all Validation of the SST products is included in the observations are referenced to subskin observations. operational system. Figure 2 shows the positions and

    anomalies of the drifting buoys for the first half of the In the analysis step, the collated files are ingested and year in 2008, and figure 3 shows the time series of the anomalies are generated by subtracting the previous comparisons. days analysis as a guess field. All grid points with ice

    present are treated as satellite observations with an SST oC. If more that one satellite observation is of -1.8available in a grid point, a weighted average is

    calculated for the given point using the individual error

    values. Optimal interpolation is performed on the

    resulting collated grid to produce the SST analysis and

    error estimates. After the interpolation, monitoring

    statistics is generated and figures produced about e.g.

    the number of satellite observations included, anomalies

    of the individual satellite products, the size of the

    increments etc.. Finally, a validation routine compares

    the drifting buoy observations to the SST analysis to

    produce error statistics.

    The analysis system has been running for the Arctic

    Ocean in a 0.05 degrees resolution. Analysis fields are

    available from January 2006 up to present. Figure 1 thshows an example of the SST field for September 9 in

    2007, which was the day with the minimal ice extent.

    Figure 2: Positions of the drifting buoy observations

    used for validation. The colors indicate the differences

    between the in situ observations and the analysis (DMI

    analysis in situ).

    Figure 3: 10 days averages of differences between Figure 1 Example of the DMI Arctic Ocean level 4 SST thdrifting buoy observations and the Arctic level 4 SST product for September 7 2007, the date with the products. minimal ice extent.

It is clear from the figure that the performance of the

    Arctic Level 4 SST product is good in terms of standard

    deviation, whereas the bias seems to be too high in the

    beginning of the year. However, it is also clear from

    figure 2 that the quality control on the buoy

    observations is not adequate, as there are some

    matchups that are situated in regions with permanent ice

    cover. These positions probably correspond to buoys

    located on drifting ice, measuring the air temperature.

    The validation should therefore be interpreted as a high

    estimate on the error on the SST product and more

    rigorous quality control on the in situ observations will

    be part of the future work.


    There are several factors in the Arctic Ocean that makes

    the production of an accurate L4 SST product difficult.

    Among these are:

    ? Very few SST retrievals in the vicinity of the

    ice edge.

     ? Altered error covariance in the marginal ice

    zone Figure 4: Example of days since the pixels were filled

    with a satellite observation. Only grid points older than ? Questionable accuracy of the SST observations 2 days are shown and coincide with the location of the in the marginal ice zone. ice edge.

     ? Large gradients in SST in the vicinity of the ice

    edge (e.g. in the East Greenland Current). 5. CONCLUSION

    The DMI analysis system ingests all available The future work with the Arctic Ocean will focus upon GHRSST-pp L2P products to produce daily high resolving and determining these issues in the marginal resolution SST products for the Arctic Ocean. The ice zone. The DMI processing system carries validation of the SST product against drifting buoys information about the age of the satellite observations in oC and a shows a standard deviation of around 0.6each grid point. As an example of the poor retrieval, positive bias, which is highest in the wintertime and figure 4 shows the age of grid points with satellite age > decreasing in the spring time. The error estimates are 2 days. All data within the marginal ice zone are at least high limits and a more careful selection of the buoy 2 days old, outlining the very poor data return in the observations should discard obviously erroneous vicinity of the ice edge. observations. In the Arctic Ocean, there are several

    factors that complicate the production of an accurate

    SST analysis product, such as the quality of the

    marginal ice zone SST retrievals, and altered statistics

    close to the ice edge. These issues will be included in

    the future work in order increase the quality of the DMI

    Arctic Ocean SST analysis.


    1. Donlon et al, 2007, The GODAE High Resolution

    Sea Suurface Temperature Pilot Project (GHRSST-

    PP), Bulletin of the American Meteorological

    Society, 88, (8) (august 2007).

Report this document

For any questions or suggestions please email