OS31A-01 INVITED
The Next Generation of Sea Surface Tempoerature Climate Data Records
A new generation of integrated sea surface temperature (SST) data products are being provided by the Global
Ocean Data Assimilation Experiment (GODAE) High-resolution SST Pilot Project (GHRSST-PP). These combine
in near-real time various SST data products from several different satellite sensors and in situ observations and
maintain fine spatial and temporal resolution needed by SST inputs to operational models. The practical
realisation of such an approach is complicated by the characteristic differences that exist between
measurements of SST obtained from subsurface in-water sensors, satellite microwave and satellite infrared
radiometer systems. Furthermore, diurnal variability of SST within a 24 hour period, manifested as both warm
layer and cool skin deviations, introduces additional uncertainty for direct inter-comparison of data sources and
the implementation of data merging strategies. The GHRSST-PP has developed and now operates an
internationally distributed system that provides operational feeds of regional and global coverage, high-resolution
SST data products (better than 10 km and ~ 6 hourly). A suite of on-line satellite SST diagnostic systems are
also available within the project. All GHRSST-PP products have a standard format, include uncertainty estimates
for each measurement and, are served to the international user community free of charge through a variety of data
transport mechanisms and access points. They are being used for a number of operational applications. The
approach will also be extended back to 1981 by a dedicated re-analysis project. This paper provides a summary
overview of the GHRSST-PP structure, activities and data products. For a complete discussion, access to data
products and services see http:www.ghrsst-pp.org.
http:www.ghrsst-pp.org
OS31A-02 INVITED
Assuring the Quality and Stability of Sea Surface Height Series from Satellite Altimetry
Since the launch of TOPEX/Poseidon in 1992 the community has assembled nearly 15 years of continuous sea surface height data from a variety of satellite altimeters. As this record was developed, methods were devised to evaluate the quality of the data via comparisons with in situ sea level observations from the global tide gauge network. In particular, the tide gauge observations have been used to evaluate temporal drift in the altimetric time series that would create difficulties for analyses aimed at studying low frequency variations in the ocean. A brief history of the development of these methods is given, with emphasis on an error analysis for global sea level rise estimates from altimetry. We will also describe the present method for doing these comparisons and show results for a number of satellite altimeter datasets.
OS31A-03
Assessment of the Ice Extent Climatic Parameters for the Arctic Region and Regional Seas Based on Historic Ice Charts and the Last Decades Passive Microwave Climatic Records
Current report presents results of the assessment of the Arctic ice cover climatic parameters in terms of statistical analysis of sea ice total concentration probability distribution functions' (PDF) centre, scale and type and ice extent temporal variability. Close to centennial historical 5-10 days ice charts collection for 1933-present and ice extent data for 1900-present from the WMO "Global Digital Sea Ice Data Bank" (GDSIDB) are used as the basis material to assess the special features of the PDF's as well as robust parameters of regional ice extent variability. Commonly-used last decades passive microwave Nasateam and Bootstrap algorithm daily ice concentration patterns and extent distributed by the NSIDC are used to complement and compare results of the previous analysis either for areas with insufficient information from ice charts or underline the effect of shorter time series. Using results of analysis of the various basic and quantile statistics representing the sea ice total concentration PDFs' center, scale and type it is shown what within the area of Arctic Basin and regional seas and in seasonal cycle non-Gaussian i, j and U PDF's types are well pronounced. Subsequent PDFs' classification is proposed on the basis of Pirson curves. The same analysis based on passive microwave data shows that these data source though containing absolute errors complement classification of PDFs type for such areas as the high Arctic due to higher relative resolution of concentration up to 1% in comparison to 5% from ice charts. Verification (confirmatory) discriminate analysis (DA) of the discriminate variables (DV) chosen from both basic and quantile statistics is used to specify the shares and spatial position of the revealed i, j, U and Gaussian-form PDF types. Analysis of the DA results carried out in the space of canonical discriminate functions shows the presence of mixtures of distributions and allows to draw a conclusion, that the class of PDF revealing specific system of thermodynamic processes in the ice cover is the initial and the system of DV defined through the quantile and basic statistics is not optimum. Analysis of the form of the variability of the ice extent in the Arctic Seas based on historical ice collection shows the high probability of existence of quasi-periodical oscillations of different period. Wavelet analysis reveals the cycles of about 10, 20 and 50-60 years in the long-term changes of the ice extent during the XX century occurring against the background linear trends different in space and time. Maximum decrease of ice extent evaluated in linear term was noted during the first part of XX century for Eurasian sub-region including Greenland, Barents and Kara Seas, whereas for the second part of the century magnitude of linear trend estimated for the same region is several times less. Level of significance for the linear trend for sub- region including Laptev, Eastern-Siberian and Chukcha Seas is close to zero, the same is observed for the Beaufort Sea. The same analysis based on the passive microwave data shows that though ice extent values may slightly differ in absolute values they follow the previous ones based on ice charts, however, shorter longevity of the data source can not reveal periods prior to 1978 visible on that from the historical ice collection.
OS31A-04
Ocean Color Climate Records
Individual ocean color missions have finite lifetimes, so it is critical to produce a consistent time series across ocean color missions if we are to address fundamental questions of Earth science importance, especially how the ocean biogeochemical system is changing. Developing Ocean Color Climate Records (OCCR's), which meet the definitions of the National Research Council has been a challenge. Consistent algorithms and processing methodologies are considered essential for such data records, but our experience with SeaWiFS and MODIS- Aqua indicates that this is not enough. They differed by 10% globally in overlapping time segments, late 2002- 2004 (see McClain et al., 2006). For perspective, the maximum change in annual means over the entire SeaWiFS mission era was 5%, and this included an El Niño-La Niña transition. We define OCCR's to meet the broad definitions of NRC, but additionally impose more stringent requirements specific to the science of ocean color: 1) all mission-dependent biases removed or quantified, 2) no obvious interannual discontinuities unattributable to natural variability, 3) similar data quality and structure. Our overall plan is to look at the missions as a time series, and develop methodologies that maximize the usefulness of the time series. The primary mission-long problems are aerosols, calibration, and other residual biases. We provide solutions to these problems using consistent processing methodologies, in situ data fusion techniques, and data assimilation. Our criteria for success are: 1) Low bias as determined by in situ chlorophyll comparisons, with regression slope near unity, 2) No change in trends annually and globally and no discontinuities using contemporary, overlapping data, when one mission is substituted for another 3) Observed trends in agreement with related climate data trends 4) Comparison and broad agreement with coupled global ocean biology models 5) Methodological maturity 6) Community approval
OS31A-05
A Validated 20-year SSM/I Satellite Wind Dataset for Climate Monitoring
A 20-year time series of satellite winds has been constructed for climate applications. To accomplish this, the Special Sensor Microwave Imager (SSM/I) brightness temperature dataset was recalibrated to remove small intersatellite offsets and instrumental drifts. Improvements to the SSM/I ocean retrieval algorithm were also implemented to bring this algorithm up to date with those currently being used for AMSR-E and the TRMM microwave imager (TMI). The entire 20-year SSM/I dataset was then reprocessed. The new wind retrievals represent the Remote Sensing Systems 6th generation of SSM/I ocean products. Most climate applications are very demanding when it comes to radiometer calibration and intersatellite debaising. Long-term stability accuracies (after applying on-orbit corrections) need to be at the 0.1 m/s/decade level or better, and it is essential that this required performance be validated by independent means. Accordingly, the SSM/I winds are compared to in situ measurements coming from three moored buoy arrays and to scatterometer wind retrievals. The scatterometer retrievals are particularly useful in that one expects scatterometers to be less prone to calibration drifts than radiometers. The scatterometer wind retrieval is based on the ratio of received power to transmitted power, and this ratio should be insensitive to instrument drift. Care must be taken when doing the buoy validation. Due to their particular location with respect to ocean currents, upwelling area, atmospheric stability, and wind fetch, each buoy has a unique wind speed bias relative to the satellite retrieval of wind stress. This wind bias must be removed before doing the long-term stability validation or else buoys at new locations coming on-line and buoys at old locations going off-line will corrupt the analysis. The SSM/I versus buoy analysis shows good agreement between the satellite and buoy winds. There is no significant bias (because the SSM/I were specifically calibrated to remove the overall bias) and the standard deviation is 1.21 m/s and 0.85 m/s for the NDBC and TAO/Pirata arrays, respectively. The NDBC results are less accurate because the NDBC buoys are not calibrated as well as the TAO/Pirata buoys and they are located in areas of higher wind gradients, both temporally and spatially. The long-term stability analysis shows a +0.08 (- 0.13) m/s/decade relative trend of SSM/I versus the TAO/Pirata (NDBC) buoys. Giving equal weight to each buoy array results in an overall trend difference of -0.02 m/s/decade. Although the overall trend difference is very small, the time series of SSM/I minus buoy winds shows features similar to those that are found with SSM/I as compared to the NASA scatterometer QuikScat. Thus we apply a small post-hoc adjustment to the SSM/I winds to bring them into agreement with the buoy winds. This adjustment is a simple table of 20 numbers that correspond to the SSM/I minus buoy wind speed difference for each year. The magnitude of these annual adjustments is about 0.1 m/s or less. When the adjusted winds are compared to the European scatterometer ERS-1 during the 1992-1996 period and to QuikScat during the 1999- 2006 period, the SSM/I minus scatterometer trend difference is +0.03 m/s/decade for both.
OS31A-06
Establishing a Climate Data Record for MODIS Sea Surface Temperatures - error characteristics and traceability to temperature standards
The establishment of Climate Data Records from satellite sensors requires extensive characterization of the uncertainties in the retrieved geophysical variables, ideally by comparison with measurements using independent sensors of known accuracy and traceability to a National Standard. To achieve an adequate description of the error characteristics the validation data sets should encompass the full climatological range of not only the retrieved variables, but also those that introduce uncertainties. Thus, for example, the validation of sea surface temperatures (SSTs) should cover not only the full range of SSTs, but also span the range of atmospheric water vapor distribution. This presentation describes the approach of validating the skin SSTs from MODISs (MODerate-resolution Imaging Spectroradiometers) on Terra and Aqua using shipboard M-AERIs (Marine-Atmospheric Emitted Radiance Interferometers) and drifting buoys. An additional approach of using microwave SST retrievals from AMSR-E (Advanced Microwave Scanning radiometer for the Earth Observing System) as a "transfer standard" is being explored. The uncertainty characteristics in the satellite retrievals that are considered acceptable depends on the intended applications, and while for many purposes knowledge of the globally averaged errors is sufficient, but for others it is necessary to specify the uncertainties in regionally or temporally constrained conditions. Here we discuss an approach of stratifying the error statistics in a multi- dimensional space, an error hypercube, that allows rapid and easy predictions of the expected retrieval errors on a pixel-by-pixel basis. This has been implemented in the Global Data Assimilation Experiment (GODAE) High Resolution Sea Surface Temperature Pilot Project (GHRSST-PP), and is argued should be adopted for the forthcoming NPP and NPOESS missions to extend the SST CDRs into the next decade.
OS31A-07
Small-Scale Variability in MODIS and Pathfinder Sea Surface Temperatures With Applications to Data Error Models for in Situ Observations
Sea surface temperature (SST) is arguably the most visible climate variable in the public forum of climate change debate. In climate change detection and attribution studies they are usually used in the form of gridded data sets which are analyzed statistically or serve as boundary conditions for atmospheric general circulation models. Therefore, it is of primary importance to ensure the optimality and reliability of gridded data sets, especially for the pre-satellite data period, including reliability of their error estimates. Most current methods of gridding blend together satellite and in situ data and involve a mixture of optimal interpolation (successive corrections), eigenvector reconstruction, bias correction techniques, and some forms of data assimilation. Analyses of the pre- satellite period depend on quite sparse in situ data as their inputs, but they usually try to make use of statistical information extracted from the satellite period. Therefore, the quality of these analyses and hence our ability to detect and properly attribute long-term climate change hinges on the quality of a priori statistical information obtained from the satellite data. Extensive satellite data was used in order to quantify and model in situ data errors, with the goal to improve pre-satellite era climate analyses. To this end intercomparisons of MODIS SST products, and their comparisons with Pathfinder V5 SST, and the in situ data collection ICOADS are presented. Small-scale and (within 1 degree bins) and short-term (within 1 month) variability of SST are estimated, using satellite data sets. These estimates are then successfully used to model the magnitude of the error in the binned in situ SST values from ICOADS.
OS31A-08
Climate Trend Detection using Sea-Surface Temperature Data-sets from the (A)ATSR and AVHRR Space Sensors.
Sea-Surface Temperature (SST) is an important indicator of global change, designated by GCOS as an essential Climate Variable (ECV). The detection of trends in Global SST requires rigorous measurements that are not only global, but also highly accurate and consistent. Space instruments can provide the means to achieve these required attributes in SST data. This paper presents an analysis of 15 years of SST data from two independent data sets, generated from the (A)ATSR and AVHRR series of sensors respectively. The analyses reveal trends of increasing global temperature between 0.13°C to 0.18 °C, per decade, closely matching that expected from some current predictions. A high level of consistency in the results from the two independent observing systems is seen, which gives increased confidence in data from both systems and also enables comparative analyses of the accuracy and stability of both data sets to be carried out. The conclusion is that these satellite SST data-sets provide important means to quantify and explore the processes of climate change. An analysis based upon singular value decomposition, allowing the removal of gross transitory disturbances, notably the El Niño, in order to examine regional areas of change other than the tropical Pacific, is also presented. Interestingly, although El Niño events clearly affect SST globally, they are found to have a non- significant (within error) effect on the calculated trends, which changed by only 0.01 K/decade when the pattern of El Niño and the associated variations was removed from the SST record. Although similar global trends were calculated for these two independent data sets, larger regional differences are noted. Evidence of decreased temperatures after the eruption of Mount Pinatubo in 1991 was also observed. The methodology demonstrated here can be applied to other data-sets, which cover long time-series observations of geophysical observations in order to characterise long-term change.