Transcript Slide 1
Ocean synthesis inter-comparison using OceanDIVA Alastair Gemmell Keith Haines Greg Smith Jon Blower Environmental Systems Science Centre University of Reading, UK D D D s http://www.resc.rdg.ac.uk Color-coded model-obs T misfits Outline Methods – OceanDIVA – a Java web application to visualize and compare gridded model data, and insitu point observations Geospatial representation of the data. Statistical representation of the data Conclusions OceanDIVA – Ocean Data Inter-comparison and Visualization Application Input data can be local to the web service or read in remotely via OPeNDAP protocol For this study: • Used EN3 (ENACT/ENSEMBLES) observations dataset • Compared a range of CLIVAR GSOP ocean syntheses accessible via OPeNDAP • Analysed Sept ’04 (Sept ’01 for syntheses finishing before ‘04) Geospatial representation of data in Google Earth Probability Density Functions (PDFs) Covering the north Pacific. Model is Reading ¼ degree • • • • Binned data into bins of 10m by 0.2oC Blues = bins with lower data density Reds = bins with higher data density Data density normalised to depth level Minimum bin content = 3 1 2 Probability Density Functions (PDFs) Covering the north Pacific. Model is Reading ¼ degree Depth Temperature Misfit Salinity Misfit Temperature Observed Depth Depth Misfit Salinity Misfit Regional Variability This example: Reading ¼ degree model showing S(T) Atlantic Pacific Obs. Obs. Misfit North Trop. South Misfit North Pacific z(T) across syntheses Observations ECCO-JPL GFDL ECMWF World Ocean Atlas ‘05 CERFACS 2001 ECCO-SIO 2001 SODA MERCATOR INGV 2001 GECCO 2001 Reading 1o control Reading 1o assim. WOA ‘05 ECCO-GODAE Reading 1/4o control Reading 1/4o assim. Bias v Standard Deviation 25 Misfit Std. Dev. (m) 65 North Pacific – z(T) – over T range 12-22 oC 0 CERFACS ‘01 ECCO-GODAE ECCO-JPL ECCO-SIO ‘01 ECMWF Misfit Mean (m) GECCO ‘01 GFDL INGV ‘01 MERCATOR Reading 1o control 50 Reading 1o assim. Reading ¼o control Reading ¼o assim. SODA WOA 2005 North Pacific S(T) across syntheses Observations ECCO-JPL GFDL ECMWF ECCO-GODAE CERFACS 2001 ECCO-SIO 2001 SODA MERCATOR INGV 2001 GECCO 2001 Reading 1o control Reading 1o assim. WOA ‘05 ECCO-GODAE Reading 1/4o control Reading 1/4o assim. Bias v Standard Deviation 0.05 Misfit Std. Dev. (PSU) 0.13 North Pacific – S(T) – over T range 5-17 oC 0.0 CERFACS ‘01 ECCO-GODAE ECCO-JPL ECCO-SIO ‘01 ECMWF Misfit Mean (PSU) GECCO ‘01 GFDL INGV ‘01 MERCATOR Reading 1o control 0.14 Reading 1o assim. Reading ¼o control Reading ¼o assim. SODA WOA 2005 Bias v Standard Deviation 0.04 Misfit Std. Dev. (PSU) 0.11 North Pacific – S(T) – over T range 17-30 oC 0.0 CERFACS ‘01 ECCO-GODAE ECCO-JPL ECCO-SIO ‘01 ECMWF Misfit Mean (PSU) GECCO ‘01 GFDL INGV ‘01 MERCATOR Reading 1o control 0.08 Reading 1o assim. Reading ¼o control Reading ¼o assim. SODA WOA 2005 Conclusions OceanDIVA is a useful tool for visualizing data, and comparing model data with observations. Useful for validation in fields of • Ocean reanalyses • Operational oceanography Outputs shown which appear to reflect differences between synthesis techniques – e.g. methods of data assimilation. • E.g. mode waters, S(T) relationships Interesting Future work planned including • different and longer time periods • using isopycnals • more syntheses. Provided correct metadata and standards are used, there is the exciting prospect of increasing amounts of data available on OPeNDAP servers etc, leading to more collaborative work and comparisons being carried out.