SURFACE TEMPERATURES IN THE POLAR-REGIONS FROM NIMBUS-7 TEMPERATURE HUMIDITY INFRARED RADIOMETER

Monthly surface temperatures in the Arctic and Antarctic regions have been derived from the 11.5-mum thermal infrared channel of the Nimbus 7 temperature humidity infrared radiometer (THIR) for a whole year in 1979 and for a winter and a summer month from 1980 through 1985. The d ata set shows interannual variability and provides spatial details tha t allow identification of temperature patterns over sea ice and ice sh eet surfaces. For example, the coldest spot in the southern hemisphere is observed to be consistently in the Antarctic plateau in the southe rn hemisphere, while that in the northern hemisphere is usually locate d in Greenland, or one of three other general areas: Siberia, the cent ral Arctic, or the Canadian Archipelago. Also, in the southern hemisph ere, the amplitude of the seasonal fluctuation of ice sheet temperatur es is about 3 times that of sea ice, while in the northern hemisphere, the corresponding fluctuations for the two surfaces are about the sam e. The main sources of error in the retrieval are cloud and other atmo spheric effects. These were minimized by first choosing the highest ra diance value from the set of measurements during the day taken within a 30 km by 30 km grid of each daily map. Then the difference of daily maps was taken, and where the difference is greater than a certain thr eshold (which in this case is 12-degrees-C), the data element is delet ed. Overall, the monthly maps derived from the resulting daily maps ar e spatially and temporally consistent. are coherent with the topograph y of the Antarctic continent and the location of the sea ice edge, and are in qualitative agreement with climatological data. Quantitatively , THIR data are in good agreement with Antarctic ice sheet surface air temperature station data with a correlation coefficient of 0.997 and a standard deviation of 2.0-degrees-C. The absolute values are not as good over the sea ice edges, but a comparison with Russian 2-m drift s tation temperatures shows very high correlation (with correlation coef ficient at 0.998) and a standard deviation of 1.1-degrees-C. Overall, the rms error is estimated to be from 1-degrees to 2-degrees-C, depend ing on the surface, while the average bias when compared with in situ data is less than 2-degrees-C.