The cited paper describes a sea-surface temperature analysis which is derived from both in situ and satellite observations. In situ observations are made from ships and from buoys. Ship observations have been made as far back as the early part of the 19th century; buoy observations are more recent, beginning in the 1970s. The advantage of these data is the long period of temporal coverage made by independent instruments which measure temperatures directly at the ocean's surface. Operational satellite instruments first became useful for sea-surface temperatures in November 1981 when the Advanced Very High Resolution Radiometer (AVHRR) became available on NOAA's polar orbiting satellites. This instrument directly measures radiances using multiple infrared frequencies or channels. The multiple channels allow better detection and elimination of cloud contamination and better correction for the effects of atmospheric water vapor. The clear advantage of satellite sea-surface temperature data is that the spatial coverage is far superior to the coverage from ships and buoys. The disadvantage is that satellite sea-surface temperature measurements are made remotely from one instrument. Thus, any errors from the instrument itself or from the algorithms that produce the retrievals may lead to systematic error biases in the retrievals.

The cited paper is important because it described a sea-surface temperature analysis which was one of the first to use both in situ and satellite sea-surface temperature data. I believe that I was the first to try to optimize the advantages of both types of data. The analysis uses the in situ data to correct any large-scale (roughly 10^o in latitude or longitude) biases in the satellite data and then combines the in situ data and the corrected satellite data into the final analysis. Thus, the final analysis uses the advantage of the ground truth of in situ data and the advantage of the high spatial coverage of satellite data. The analysis was designed for climate studies, such as global and regional long-term temperature trends, as well as variations such as those caused by El Niño and La Niña events. These events originate in the tropical Pacific and influence global temperature and precipitation, especially in the winter. For climate purposes it is necessary that the analysis be as consistent as possible in time and space. Thus, as periodic improvements are made to the analysis, it is recomputed for the entire period. The analysis is widely used for climate monitoring, prediction, and research as well as specifying the surface boundary condition for numerical weather prediction and for atmospheric reanalyses. The paper has been highly cited because of these improvements and because the analysis has always been freely available to anyone who wanted it.

I began my work on sea-surface temperatures after the 1982-83 El Niño warming event. This event was not well observed and forecasted partly due to poor sea-surface temperature analyses. At that time the satellite sea-surface temperature retrievals were negatively biased due to the presence of stratospheric aerosols from the April 1982 eruptions from the Mexican El Chichón volcano. I thought I could solve this problem in a few months. I have now been working on sea-surface temperature analyses for more than 20 years.

It is a very interesting and exciting time for sea-surface temperature data and analyses because new satellite instruments are now available and more will be available soon. When the sea-surface temperature analysis described in the cited paper was originally developed, there was only one satellite instrument available to operationally produce sea-surface temperatures. In roughly the last decade, new infrared sensors have become available on other satellites. Beginning in 1997, tropical sea-surface temperatures began to be available from the microwave instrument on the Tropical Rainfall Measuring Mission (TRMM) mission. Other microwave instruments have become available in 2002 and others are planned. Microwave instrument sea-surface temperatures have lower resolution than infrared instruments but are able to retrieve sea-surface temperatures in cloud-covered regions where an infrared instrument cannot. In addition, the instruments and algorithms differ from each other, which results in independent error characteristics. New analyses using multiple satellite products are now being produced and are being planned at much higher spatial and temporal resolutions.

The work I have been doing was successful partly because the sea-surface temperature analyses were easily available to users and well documented. Easy access and documentation is particularly helpful for developers of analyses because users quickly find problems or request additional information. This allows these products to be reevaluated and improved. Once the product is improved, more users become interested and the reevaluation and improvement cycle can begin again.

Richard W. Reynolds
National Climatic Data Center
National Oceanic and Atmospheric Administration
Asheville, NC, USA