Validation of the Glob Colour Full product set

Validation of the Glob. Colour Full product set (FPS) over open ocean Case 1 waters David Antoine Laboratoire d’Océanographie de Villefranche Inputs from the Glob. Colour team Special thanks to Gilbert Barrot, Julien Demaria and Christophe Lerebourg Glob. Colour / Medspiration user consultations, Nov 20 -22, 2007, Oslo

Objectives - questions Ø Are the overall geographical distributions valid in the merged data set? (e. g. , any artificial boundaries? ) Ø Are the statistics derived from the match up analysis of the FPS with field data at least not worst (and hopefully better) than the individual -sensors’ statistics? Ø Is the data set usable for delivery of operational services (GMESMCS) and for “carbon cycle research”? Ø Recommendations for the next steps Glob. Colour / Medspiration user consultations, Nov 20 -22, 2007, Oslo

Plan Ø Examination of global composites Ø Validation against field data (OBPG, NOMAD, BOUSSOLE) Ø Looking at time series and histograms over selected areas Ø A bit more sophisticated analyses as well (distributions, anomalies) Ø Conclusions Ø Recommendations Glob. Colour / Medspiration user consultations, Nov 20 -22, 2007, Oslo

The Glob. Colour product list Ø Chlorophyll-a concentration (Chl-a) derived from reflectance ratios or, for the GSM method, from aph Ø Error bar on the Chlorophyll concentration (GSM) Ø Colored dissolved + particulate (“detrital”) organic matter (CDM) either for MERIS or from the GSM 01 method Ø Particle backscattering at 443 nm (bbp 443) from the GSM algorithm Ø Total suspended matter (TSM) Ø Diffuse attenuation coefficient for downward irradiance (Kd 490) Chl-based algorithm (original Kd’s from Sea. Wi. FS and MODIS are not used) Ø Fully normalized water leaving radiances (n. Lw’s) Ø Excess of radiance at 560 nm (turbid Case 2 waters) Ø Photosynthetically available radiation (PAR) Ø Aerosol optical thickness Ø Cloud fraction Ø Data quality flags Glob. Colour / Medspiration user consultations, Nov 20 -22, 2007, Oslo

The Glob. Colour products we have considered in the open ocean validation (in red) Ø Chlorophyll-a concentration (Chl-a) derived from reflectance ratios or, for the GSM method, from aph Ø Error bar on the Chlorophyll concentration (GSM) Ø Colored dissolved + particulate (“detrital”) organic matter (CDM) either for MERIS or from the GSM 01 method Ø Particle backscattering at 443 nm (bbp 443) from the GSM algorithm Ø Total suspended matter (TSM) Ø Diffuse attenuation coefficient for downward irradiance (Kd 490) Chl-based algorithm (original Kd’s from Sea. Wi. FS and MODIS are not used) Ø Fully normalized water leaving radiances (n. Lw’s) Ø Excess of radiance at 560 nm (turbid Case 2 waters) Ø Photosynthetically available radiation (PAR) Ø Aerosol optical thickness Ø Cloud fraction Ø Data quality flags Glob. Colour / Medspiration user consultations, Nov 20 -22, 2007, Oslo

1 st step: looking over some global composites Examples for may 2006 Glob. Colour / Medspiration user consultations, Nov 20 -22, 2007, Oslo

n. Lw(412), may 2006 Glob. Colour / Medspiration user consultations, Nov 20 -22, 2007, Oslo

n. Lw(443), may 2006 Glob. Colour / Medspiration user consultations, Nov 20 -22, 2007, Oslo

n. Lw(490), may 2006 Glob. Colour / Medspiration user consultations, Nov 20 -22, 2007, Oslo

n. Lw(555), may 2006 Glob. Colour / Medspiration user consultations, Nov 20 -22, 2007, Oslo

Chlorophyll (weighted average), may 2006 Glob. Colour / Medspiration user consultations, Nov 20 -22, 2007, Oslo

Chlorophyll (GSM), may 2006 Glob. Colour / Medspiration user consultations, Nov 20 -22, 2007, Oslo

GSM Bbp(443), may 2006 Glob. Colour / Medspiration user consultations, Nov 20 -22, 2007, Oslo

GSM CDM, may 2006 Glob. Colour / Medspiration user consultations, Nov 20 -22, 2007, Oslo

Kd(490), may 2006 Kd(490) = 0. 0166 + 0. 0835[Chl]0. 633 (Morel et al. , 2007) Glob. Colour / Medspiration user consultations, Nov 20 -22, 2007, Oslo

2 nd step: match ups with field data (NOMAD + OBPG + BOUSSOLE) Glob. Colour / Medspiration user consultations, Nov 20 -22, 2007, Oslo

Matchups with the merged products 1 – NASA’s OBPG data set Location of the OBPG validation dataset used for the Glob. Colour Level-3 validation Parameters Chl = 69 L 443 = 54 L 531 = 01 KD = 25 L 490 = 41 L 555 = 54 L 412 = 47 L 510 = 26 L 670 = 49 Glob. Colour / Medspiration user consultations, Nov 20 -22, 2007, Oslo

Matchups with the merged products 2 – NASA’s NOMAD data set Werdell and Bailey, 2005: An improved bio-optical data set for ocean color algorithm development and satellite data product validation. Remote Sensing of Environment , 98(1), 122 -140. First contributed by the NASA SIMBIOS Program (NRA-96 -MTPE-04 and NRA-99 -OES-09) Glob. Colour / Medspiration user consultations, Nov 20 -22, 2007, Oslo

Matchups with the merged products 3 – BOUSSOLE data set 3 years of data From Sept. 2003 to Sept 2006 n. Lw’s 412, 443, 490, 510, 560, 665, 681 nm HPLC TChl-a during monthly servicing cruises Glob. Colour / Medspiration user consultations, Nov 20 -22, 2007, Oslo

Matchups using all three data sets Glob. Colour / Medspiration user consultations, Nov 20 -22, 2007, Oslo

Matchups using all three data sets: overall statistics Glob. Colour / Medspiration user consultations, Nov 20 -22, 2007, Oslo

Do we perform better after the merging process ? Evolution of the statistical indicators from the individual-sensors’ products to the merged products The answer is definitely YES Glob. Colour / Medspiration user consultations, Nov 20 -22, 2007, Oslo

3 rd step: 9 -year time series Overall consistency, trends? Jumps? Glob. Colour / Medspiration user consultations, Nov 20 -22, 2007, Oslo

Analysis of time series: the selected areas Glob. Colour / Medspiration user consultations, Nov 20 -22, 2007, Oslo

Analysis of time series: global ocean (50 S-50 N, depth>1000 m) Glob. Colour / Medspiration user consultations, Nov 20 -22, 2007, Oslo

Analysis of time series: South east Pacific Glob. Colour / Medspiration user consultations, Nov 20 -22, 2007, Oslo

Analysis of time series: Mediterranean Sea Glob. Colour / Medspiration user consultations, Nov 20 -22, 2007, Oslo

Analysis of time series: Chl at the BOUSSOLE site only Glob. Colour / Medspiration user consultations, Nov 20 -22, 2007, Oslo

Analysis of time series: bbp(443) at the BOUSSOLE site only Glob. Colour / Medspiration user consultations, Nov 20 -22, 2007, Oslo

Analysis of time series: Main results from what we have seen and from the other areas as well - The best agreement between the 3 sensors is for L(490) - GSM Chl is always larger for MERIS than for MODIS-A & Sea. Wi. FS - GSM Chl is often smaller than the weighted average Chl - L(555) is often smaller for MODIS-A than for MERIS and Sea. Wi. FS, and this is due to lower values for clear waters. It is also “flatter” (less seasonality) in many occasions. The average value (0. 3 m. W cm-2 mm-1 sr-1 is, however, closer to the “clear water radiance” (Gordon and Clark, 1981). - Good results for the preliminary validation of the bbp(443) at BOUSSOLE - The merged data set is close to the Sea. Wi. FS data set Glob. Colour / Medspiration user consultations, Nov 20 -22, 2007, Oslo

4 th (and last) step: global Chl distributions and Global Chl “anomalies” Glob. Colour / Medspiration user consultations, Nov 20 -22, 2007, Oslo

Global chlorophyll distributions Oligotrophic Mesotrophic Eutrophic 51% 43% 38% 52% 44% 38% Glob. Colour / Medspiration user consultations, Nov 20 -22, 2007, Oslo

Global chlorophyll “anomalies” Differences between the global Chl stock (mg. Chl m-2) of a given month and the same stock for the corresponding “climatological month”, i. e. , the average for this month over 9 years (1998 -2006) From the Behrenfeld et al. (2006) paper in Nature Chl from the weighted average GSM Chl Which one is right? Raises the question about the validity of the FPS for long-term analyses Glob. Colour / Medspiration user consultations, Nov 20 -22, 2007, Oslo

Conclusions (1/3) - Based on the match up with the global data set of field data (Chl and n. Lw’s), the Glob. Colour FPS is validated. - The statistics favourably evolve for the normalized water-leaving radiance in the blue bands (412 and 443 nm), as compared to what they are for the individual sensors - In terms of Chl, the statistics for the GSM Chl are a little better than those for the product from the weighted average. - The normalized water-leaving radiance at 490 nm is by far the most homogeneous product among the 3 sensors, so the confidence in the merged product is higher for this peculiar band - The MODIS-A L(555) is often smaller than the L(555) for the two other sensors - The merged product has not degraded the situation as compared to each of the 3 single -sensor data set Glob. Colour / Medspiration user consultations, Nov 20 -22, 2007, Oslo

Conclusions (2/3) - The good results in term of global match ups are somewhat fortuitous, however, and may often result from compensating effects, in particular between MODIS-A and MERIS. - The Glob. Colour FPS is often close to the Sea. Wi. FS data set alone. - This is not totally satisfactory: the merged data set would not be validated in case another sensor, with its specific bias, would be added, or if one the presently used sensor would be removed from the merging process. - This is not due to the merging process, but to remaining uncertainties in the vicarious calibration of the various ocean color sensors, and to differences in algorithms (atmospheric corrections in particular) Glob. Colour / Medspiration user consultations, Nov 20 -22, 2007, Oslo

Conclusions (3/3) - The Glob. Colour FPS is definitely qualified and usable for operational uses, such as assimilation into global models (there is a pixel-by-pixel error bar delivered with the GSM Chl), or delivery of services. -The Glob. Colour FPS is not yet qualified to perform temporal analysis over the period 1998 -2007. - In other words, the Glob. Colour FPS doesn’t yet meet the standards for being qualified as a “climate quality data record” Glob. Colour / Medspiration user consultations, Nov 20 -22, 2007, Oslo

Recommendations - We need several in situ long-term time series, in order to add the temporal dimension to our statistical analyses (match ups). - An international effort is still needed to standardize & improve the vicarious calibration methodologies. We are still not at the desired level of confidence (see, e. g. , Ohring et al. , EOS Trans AGU, 88(11), 13 march 2007) - Establish a collaborative frame between space Agencies (ESA, NASA and others), so that vicarious calibration and related issues (atmospheric corrections) can be standardized. This may need a specific body where these issues are discussed and the methods are implemented (see, e. g. , The GHRSST). - Incorporate new approaches, for instance where the TOA level-1 observations of all instruments are processed the same way (same algorithm), which also means that they are all vicariously calibrated against the same standard. It might become obvious at some point that this approach is mandatory if one thinks to CDRs (“climate quality data records”). Glob. Colour / Medspiration user consultations, Nov 20 -22, 2007, Oslo

Thank you for your attention Glob. Colour / Medspiration user consultations, Nov 20 -22, 2007, Oslo

Statistical measures (indicators) Coef. of determination Glob. Colour / Medspiration user consultations, Nov 20 -22, 2007, Oslo

The full product set “FPS” q 10 years of global data from the 3 sensors q Systematic application of the 2 merging methods (weighted average and GSM) q Generation of all products Glob. Colour / Medspiration user consultations, Nov 20 -22, 2007, Oslo

Glob. Colour / Medspiration user consultations, Nov 20 -22, 2007, Oslo