||TMI Data Validation
RSS performs extensive algorithm validation to determine the quality of our data. We maintain a dataset of over 8 million buoy observations to use in validating the TMI products we produce.
Validation of Version-2 TMI Sea Surface Temperatures
Validation of satellite-derived SSTs is necessary to check the retrieval algorithm developed for the TMI 10.7 GHz channel. The validation process includes comparison of TMI SSTs to in-situ measurements made by moored buoys located in both the tropical Pacific and tropical Atlantic oceans.
The tropical Pacific moored ocean buoys consist of mostly TAO array buoys. Hourly data from the approximately 60 buoys that make up the array are obtained from the Pacific Marine Environmental Lab (PMEL). The remaining Pacific Ocean buoys are operated by the National Data Buoy Center (NDBC). The NDBC provides data from moored buoys along the US coasts and have approximately 20 buoys greater than 35 km from land on the Pacific (8) and Atlantic (11) coasts. We also compare our TMI SSTs to daily SST observations from the PIRATA buoys moored in the tropical Atlantic Ocean.
Data comparisons are made for the December 1997 to June 1999 time period. During this time we have over 9000 TMI to TAO/NDBC buoy collocations and approximately 1500 PIRATA daily collocations. TMI and buoy SST collocations were used only for rain-free atmospheric conditions. We found small differences between the TMI and buoy SSTs. The TMI minus TAO difference when averaged over all buoys is 0.1 degree Celsius and the mean TMI minus NDBC SST difference is 0.2 degree Celsius. The standard deviation for both was approximately 0.5 degree Celsius. Slightly better results were found when we averaged the hourly TAO/NDBC data to daily values and compared them with TMI 3-day mean SSTs. The PIRATA buoy data, though fewer in number, on average showed no difference from the TMI SSTs, and had a standard deviation of 0.4 degree Celsius.
A time series of two Central Pacific buoys located within 0.1 degree longitude of each other and TMI SST data is shown below in Fig. 1. The TAO buoy (51023, red) has fewer data than the NDBC buoy (51028, blue) and the TMI data (black). Increased SST variability exists at this location prior to the week of May 13th, 1998. It was during this week when the SSTs suddenly decreased as the warm El Niño waters dissipated. This change is also visible in the SST animations. A mean 0.4 degree TMI minus NDBC SST bias is present in this time series. We find the TAO and TMI SSTs agree.
Most of the PIRATA data are for four buoys longest running in this new array. Fig. 2 shows four time series that show the excellent agreement between the TMI SSTs and the buoys SSTs at these four locations.
Validation of Version-2 TMI Wind Speeds
TMI winds have been calibrated to buoys and the QuikScat scatterometer. The rms difference between the TMI wind (11 GHz) and buoys is 0.84 m/s (Fig. 3). The rms difference between the TMI wind (11 GHz) and QuikScat is 0.64 m/s.