 |
|
 |
 |
|
 |
 |
|
 |
 |
 |
IntroductionThe through-cloud capabilities of microwave radiometers provide a valuable picture of global sea surface temperature (SST). To utilize this, scientists at Remote Sensing Systems have calculated a daily, Optimally Interpolated (OI) SST product at quarter degree (~25 kilometer) resolution. This product is ideal for research activities in which a complete, daily SST map is more desirable than one with missing data due to orbital gaps or environmental conditions precluding SST retrieval. In addition to the MW OI products, RSS also produces a MW+IR OI SST product that combines the through-cloud capabilities of the MW data with the high spatical resolution of the IR SST data. This daily SST product has a 9-km resolution. Improved global daily NRT SSTs should be useful for a wide range of scientific and operational activities.
Development of the MW OI SST ProductsMicrowave (MW) OI SSTs include data from these satellite radiometers:
The TRMM Microwave Imager (TMI), carried on NASA's Tropical Rainfall Measuring Mission (TRMM) satellite, was the first well-calibrated microwave radiometer capable of accurate through-cloud SST retrieval. NASDA's Advanced Microwave Scanning Radiometer for EOS (AMSR-E), carried on NASA's AQUA satellite, was the first microwave radiometer capable of accurately measuring global through-cloud SSTs. Soon after Aqua, the Naval Research Laboratory (NRL) and Air Force Research Laboratory launched the Coriolis satellite which carried the WindSAT instrument, also capable of measuring SST through clouds. TMI, AMSR-E, and WindSAT observations are used to retrieve SST, wind speed, columnar water vapor, cloud liquid water, and rain rate (Wentz, 1999). These environmental variables are calculated simultaneously using a multi-stage linear regression algorithm derived through comprehensive radiative transfer model simulations. SST retrieval is prevented only in regions with sun-glitter, rain, and close to land where there is side-lobe contamination. Since only a small number of retrievals are unsuccessful, the combination provides nearly complete global coverage each day. These MW SSTs are then used to calculate daily, quarter degree (~25km) OI SST products as described below. Although these SST products are at a lower spatial resolution than standard infrared SSTs, they represent a significant improvement over the highly used weekly, 1 degree (~100 km) NCEP OI (Reynolds) SST product.
Sensor-specific Error CalculationThe following corrections and analyses of errors are necessary first steps towards producing the MW OI SST products. Each step is further described below.
Correcting for TMI’s Emissive AntennaThe antenna coating of the TMI sensor was oxidized in orbit soon after launch, causing errors in the TMI observations. A correction was developed (Wentz, 2001), but proved to be incomplete in removing the error. A bias still exists in TMI data, which is a function of local observation time (Gentemann, accepted JGR). To account for this, an additional correction is applied before TMI data are included in the OI analysis.
Estimation and Removal of Diurnal WarmingBefore blending the satellite data, we consider the data sampling of each instrument. For example, the sun synchronous orbit of MODIS and AMSR-E on Aqua yields retrievals at a local time of approximately 1:30 AM and 1:30 PM. During the daytime over-pass, solar heating of the ocean surface can cause warming of up to 3° C (Price et al, 1986, Yokohama, 1996). Currently, many OI SST algorithms either ignore daytime retrievals or assign them a higher error than nighttime retrievals. While simply removing the daytime retrievals from the objective analysis does prevent warm retrievals from 'contaminating' the final product, the number of samples can be reduced by half. In well-sampled regions this may not impact the final product, but the infrared SSTs used in most analyses have large regions where few retrievals exist each month due to persistent cloud cover, making the daytime retrievals extremely valuable. Assigning the daytime retrievals a higher error (and therefore a smaller weight in the objective analysis) reduces diurnal 'contamination' of the dataset, but at the risk of still including some component of diurnal warming. The AMSR-E and TMI OI SSTs include day and night observations. To optimally utilize MW daytime retrievals, a simple empirical model of diurnal warming was developed that depends on solar insolation, wind speed, and local time of observation (Gentemann, 2003). Solar insolation is calculated as a function of latitude and day of year; wind speed is simultaneously retrieved with SST from radiometer observations. Using this diurnal model, all MW SSTs are 'normalized' to a daily minimum SST, defined to occur at approximately 8 AM, local time.
Sensor Errors for OI AnalysisMicrowave SST retrieval errors are mainly a function of wind speed and SST. These errors are added in a root-sum-squared sense to the daily standard deviation (STD) derived from buoy collocations to obtain a total retrieval error.
Additional Quality ControlSome rain contaminated SSTs exist in the microwave data. At the edges of rain cells, there is often undetected rain that causes a biased SST retrieval. Two tests attempt to remove rain contaminated SSTs. First, at each SST retrieval the STD is calculated using all data within one day and 100 km of the cell. SSTs falling outside of 3 STDs are flagged and removed from the data set. This process is then further repeated to remove outliers. Next, the SST is compared to the previous day's OI SST value. Any SSTs within 100 km of a rain pixel that are more than 0.6 C warmer than the previous day's OI SST value are removed. Some cloud contaminated SSTs exist in the infrared data. At the edges of cloud cells, there is often undetected cloud that causes a biased SST retrieval. We use a similar test as described above, to remove spurious cloud contaminated retreivals from the infrared SSTs.
NRT Validation and Bias Corrections Using In Situ DataSince each sensor /satellite combination has unique biases, inter-calibration of different SST retrievals is necessary. Additionally, a daily estimate of variability is important for accurately estimating sensor errors. These are both accomplished through comparisons of the satellite data with ‘ground truth’ in situ measurements. NRT in situ observations are retrieved from the Global Ocean Data Assimilation Experiment (GODAE) Monterey Server. Observations from ship engine room intake, fixed buoy, drifting buoy, ship hull sensors, and CMAN stations are included in the GODAE dataset. In situ observations are collocated with the closest microave SST observations. A collocation is made only if there is a satellite observation within 25 km and 6 hours. Collocations within 200 km of land are excluded as these are typically regions with highly variable (both temporally and spatially) currents. Satellite and buoy SSTs measure diurnal warming with different amplitudes (Donlon, 2002). Therefore, collocations between 12 Noon and 4 PM (local time) with wind speeds less than 6 m/s are excluded. These matchups provide daily mean bias and standard deviations for all microwave SSTs.
Optimum Interpolation (OI)After characterizing the errors listed above, the SSTs are blended together using the OI scheme described in Reynolds and Smith (1994). OI is a widely utilized method in oceanography and meteorology that makes use of the statistical properties of irregularly spaced data (in time and space) to interpolate the data onto a regularly sampled grid. For each dataset included in the analysis, error characteristics must be understood or at least estimated. A first-guess field, the previous day's OI SST, is employed to calculate data increments, which are all nearby data minus the first-guess field. The new SST estimate is formed by a weighted sum of increments, with the weights calculated by the OI method, added to the first guess data. Correlation scales of 4 days and 100 km are used in determining the weights used in our methodology.
Known Problems
Undetected Sea IceUndetected sea ice is causing some unrealistically warm SST values to appear in these products. The problem is most apparent near ice edges, especially as the ice edge advances or retreats. The first set of images (below) illustrates the problem occuring in the Beaufort Sea, Artic Ocean, over a seven day period. In the images on the left, the ~4° C (light blue, circled) SSTs are probably artifacts of a thin layer of sea ice or slush. As the sea ice solidifies, it becomes more accurately identified as the images progress in time towards the right.
The second set of images (above) tracks retreating Antarctic sea ice over ten days. Here we see probable ice causing up to ~5° C warm artifacts.
Missing DataSatellite instruments are occasionally unavailable. Near real time OI SST products will be created for the current day, even if no new observations exist. The OI method utilizes a first guess field, which in this analysis is the previous day's OI SST. If there are no new observations, the new SST estimate is the previous day's SST. Instrument observations are missing for the following dates:
For example, AMSR-E was unavailable September 13-19, 2002. For these dates, the combined TMI + AMSR-E product accurately represents detailed daily SSTs in TMI range (±40°), but at latitudes greater than 40° the OI SST values change little because no AMSR-E observations exist. As more satellites are added to this analysis, the chance of this 'frozen' data will diminish. All OI analyses suffer from this problem. Daily browse imagery for the TMI and AMSR-E instrument products can show the observations available on any given day.
MW OI SST ProductsSeveral optimally interpolated (OI) SST products are created from the microwave and infrared SSTs.
PLEASE READ!!! The tmi_amsre and mw_ir SST data both use WindSAT data from 10/2011 onward, due to the failure of AMSR-E. We are currently working in reprocess all the datasets. In the future there will simply be a microwave OI SST at 25 km and a microwave+infrared SST at 9 km. One analysis contains only TMI SSTs, another contains only
Gridded Binary Data File FormatEach binary data file available at ftp.discover-earth.org/sst consists of three 0.25 x 0.25 degree grids (1440 x 720 array) of single byte values representing a given day's SSTs, interpolation ERROR estimates, and data MASKing information. The MASK has bit values: leftmost bit (bit 0) = land if set to 1, bit 1 = ice, bit 2 shows whether IR data were used, bit 3 shows whether MW data were used, and bit 4 is set to 1 for bad data. Interim products ("rt") are updated several times per day until the data become final ("v03").
File names follow these conventions:
Where "yyyy", "doy", and "ver" stand for:
The center of the first cell of the 1440 column and 720 row map is at 0.125 E longitude and -89.875 latitude. The center of the second cell is 0.375 E longitude, -89.875 latitude.
Byte values range from 0 to 255. The read routines supplied insert the following specific values into the SST and ERROR data grids as below:
Byte values 0 - 250 need to be scaled to obtain standard units:
Thus, to convert SST byte values (0 - 250, inclusive) to degrees Centigrade, multiply by .15, then subtract 3.
All binary data files have gzip compression to reduce size and decrease transfer time.
Read Routines:Read routines are available in IDL, Matlab and Fortran at:
ValidationWe have performed extensive comparisons of our MW OI SST products to NCEP OI (Reynolds) SSTs. Following are two sets of statistics: one for collocations within the range of TMI data (±40°); another for global collocations (±90°).
Note that we expect to find differences, as these MW OI SST products accurately resolve real SST features that are smoothed out of the low resolution NCEP OI (Reynolds) SSTs. For example, we see a higher standard deviation when latitudes greater than 40 are included, likely due to the presence of more dynamic SST features at higher latitudes, such as the western boundary currents (Brazil/Malvinas, Atlantic Gulf Stream, Kuroshio/Oyashio) and the Antarctic circumpolar current.
Conclusions and Future WorkThese SSTs have proven useful for tropical cyclone intensity forecasting and improve upon currently available SST products. Further significant improvements to our MW OI SSTs will be made through better error specification, the addition of data from other satellites/sensors, and continued research in modeling diurnal warming.
References
Acknowledgement
Last updated: August 06, 2012 |
|
 |
 |
|
|
 |
 |
|
 |
