Dr. Sandra Castro and Dr. William Emery
Colorado Center for Astrodynamics Research, University of Colorado, Boulder, CO 80302
Program Manager: Dr. Stan Wilson, NOAA/NESDIS
Related NOAA Strategic Plan Goal:
Goal 2. Understand climate variability and change to enhance society’s ability to plan and respond.
Goal 3. Serve society’s needs for weather and water information.
The complementary nature of present infrared and microwave sea surface temperature (SST) products provides opportunities for combining the data, but significant differences among the products must first be understood and characterized. In this project we explore which environmental and sensor parameters contribute most to the uncertainty in existing infrared Advanced Very High Resolution Radiometer (AVHRR) and microwave Advanced Microwave Scanning Radiometer (AMSR-E) and Tropical Rainfall Mapping Mission Microwave Imager (TMI) SST products and present a method for specifying the errors in terms of available satellite-derived products.
Method: Collocations between the satellite retrievals and SST measurements from moored and drifting buoys are constructed and used to derive the error estimates. Multidimensional look-up tables are generated to provide the expected bias and standard deviation of individual retrievals as functions of various combinations of parameters influencing the differences between the products. Objective evaluations then determine what combinations are most effective in reducing these differences.
During the first three years of this project, we completed the derivation of error statistics for AVHRR, AMSR-E, and TMI as a function of satellite-derived sensor and environmental parameters. Bias adjustments derived from the satellite zenith angle, channel 4–5 brightness temperature difference, and SST for the infrared, and wind speed, water vapor content, and SST for the microwave products were the most effective single combinations of parameters in reducing the variability of differences both relative to the buoys and in gridded difference maps between the products. Further improvements are possible through inclusion of additional independent corrections based on the climatological SST anomaly and aerosol optical depth. These corrections enable reductions in the monthly rms difference between the products of as much as 42% and in differences with independent buoys of as much as 10%. The largest individual corrections were to AVHRR data in regions of low SST, but larger numbers of microwave retrievals were significantly changed by the adjustments.
The last two years of this project are devoted to understanding and characterizing uncertainty estimates in the presence of diurnal warming (DW). Diurnal warming of the ocean surface layer at low wind speeds and sufficient insolation can lead to significant and highly variable differences between the skin and subsurface temperatures. These differences represent a source of uncertainty in estimates of the subsurface temperature. This stage of the work explores whether improved accuracy can by achieved through retrieval of the skin temperature and explicit consideration of diurnal warming effects.
1. Based on our previous findings regarding the large biases in the AVHRR sensor for regions of cold SSTs (high latitudes), we are collaborating with Dr. Doug May of NAVO (the AVHRR operational agency and data provider) in order to understand the sources of the cold SST bias in the Navy retrieval algorithm. This is a global algorithm trained with buoy SST data, but since there is sparse buoy coverage at high latitudes, the algorithm fails to capture the unique characteristics of high latitudinal regions leading to the biases. To this end, we have extended the derivation of error statistics for the latest AVHRR sensor on board the NOAA-18 spacecraft. This required construction of a whole new set of satellite-buoy collocated matchups for the period from January 2006 (roughly the beginning of the N-18 AVHRR) through the present. Binned N-18 AVHRR SST – Buoy SST differences as a function of satellite SST are shown in Figure 1 for nighttime (top) and daytime (bottom). The figure reveals a persistent bias for cold SSTs present in the new sensor. This is an ongoing investigation and as such, we do not have a complete understanding as to what is causing the biases. We are required to keep a simpler approach than before based on what NAVO can likely implement operationally.
2. We have been participating in the Diurnal Variability Warming Group (DVWG), a working group within GHRSST-PP and its US counterpart, MISST. The objective of the DVWG is to provide a new consensus model for diurnal warming that can be used within the GHRSST project. Several physically-based models are currently being evaluated for their accuracy and suitability. Through collaborations with Dr. Gary Wick from NOAA/ERSL, we are currently testing a version of the Kantha and Clayson second moment turbulence closure model (MODKC), as well as an enhanced version of the TOGA COARE warm layer model. For homogeneity in the runs, a comprehensive data set of has been put together consisting of direct measurements of radiative fluxes from SEVERI (MSG satellite) and forcing conditions from numerical weather prediction (NWP) model outputs over the Western Mediterranean Sea at a 0.1 degree resolution. This region was selected in agreement with other NOPP investigators as part of a collaborative study of different diurnal variability models and their associated uncertainties. Model runs are validated using radiometric skin SST measurements and corresponding subsurface temperatures. As part of our involvement in the DVWG, we are concentrating our validation efforts using skin data from the CIRIMS radiometer, provided by Dr. Andrew Jessup (University of Washington).
Specifically, we have addressed the following issues:
• Supplementing the CIRIMS radiometric skin SST data with corresponding environmental data needed to study diurnal warming predictions.
• Explicit evaluation of DW models using CIRIMS data (Figure 2).
• Comparison of temperature profiles from the MODKC model as computed from idealized forcing data (Figure 3).
• Comparison of modeled DW with observations from SEVERI on the Western Mediterranean grid (Figure 4).
• A problem identified is the performance of the MODKC at lower wind speeds. We are currently looking at different turbulence formulations to see if the problem could be solved (Figure 5).
3. Submission of paper to Journal of Geophysical Research describing the results obtained in years 1-3. We just completed a round of revisions and have submitted the revised manuscript.
Castro, Sandra L., Wick, Gary A., Jackson, Darren J., and Emery, William J. “Error Characterization of Infrared and Microwave Satellite Sea Surface Temperature Products for Merging and Analysis”, submitted to J. Geophys. Res., 2007.
Summary of Interaction with NOAA
This work was done in close partnership with Dr. Gary Wick and Mr. Darren Jackson from NOAA/OAR/ESRL/PSD.
Additional project partners that we interacted with included scientists of the Office of Research and Applications within NOAA/NESDIS (Dr. Andy Harris and Ms. Eileen Maturi). Future planed impact studies incorporating our results will be conducted by NOAA/EMC within NESDIS.
Summary of Education and Outreach Activity
Results were presented at the DVWG workshops held in Key Largo, FL in February 2007 and Melbourne, Australia in May 2007. The next DVWG will take place in Edinburgh, Scotland in September, 2007. We will present a related poster in the AMS/EUMETSAT and SEAFLUX meetings in Amsterdam, The Netherlands, following the DVWG workshop. We will be directly working with the NAVO group in Stennis in October 2007 to address the issues with the N-18 AVHRR sensor.