MONITORING GLOBAL MONTHLY MEAN SURFACE TEMPERATURES

An assessment is made of how well the monthly mean surface temperature s for the decade of the 1980s are known. The sources of noise in the d ata, the numbers of observations, and the spatial coverage are apprais ed for comparison with the climate signal, and different analyzed resu lts are compared to see how reproducible they are. The data are furthe r evaluated by comparing anomalies of near-global monthly mean surface temperatures with those of global satellite channel 2 microwave sound ing unit (MSU) temperatures for 144 months from 1979 to 1990. Very dis tinctive patterns are seen in the correlation coefficients, which rang e from high (>0.8) over the extratropical continents of the Northern H emisphere, to moderate (approximately 0.5) over tropical and subtropic al land areas, to very low over the southern oceans and tropical weste rn Pacific. The physical difference between the two temperature measur ements is one factor in these patterns. The correlation coefficient is a measure of the signal-to-noise ratio, and largest values are found where the climate signal is largest, but the spatial variation in the inherent noise in the surface observations over the oceans is the othe r major factor in accounting for the pattern. Over the oceans, sea sur face temperatures (SSTs) are used in the surface dataset in place of s urface air temperature and the Comprehensive Ocean-Atmosphere Data Set (COADS) has been used to show that 80% of the monthly mean air temper ature variance is accounted for in regions of good data coverage. A de tailed analysis of the sources of errors in in situ SSTs and an overal l estimate of the noise are obtained from the COADS by assessing the v ariability within 2-degrees longitude by 2-degrees latitude boxes with in each month for 1979. In regions of small spatial gradient of mean S ST, individual SST measurements are representative of the monthly mean in a 2-degrees box to within a standard error of 1.0-degrees-C in the tropics and 1.2-degrees to 1.4-degrees-C in the extratropics. The sta ndard error is larger in the North Pacific than in the North Atlantic and much larger in regions of strong SST gradient, such as within the vicinity of the Gulf Stream, because both within-month temporal variab ility and within-2-degrees box spatial variability are enhanced. The t otal standard error of the monthly mean in each box is reduced approxi mately by the square root of the number of observations available. The overall noise in SSTs ranges from less than 0.1-degrees-C over the No rth Atlantic to over 0.5-degrees-C over the oceans south of about 35-d egrees-S. Greater daily variability in surface marine air temperatures than in SSTs means that two to three times as many observations are n eeded per month to reduce the noise in the monthly mean air temperatur e to the same level as for SST. Tests of the reproducibility of SSTs i n analyses from the U.K. Meteorological Office (UKMO) and the U.S. Cli mate Analysis Center (CAC) and from COADS reveal monthly anomaly corre lations on a 5-degrees grid exceeding 0.9 over the northern oceans but less than 0.6 in the central tropical Pacific and south of about 35-d egrees-S. Root-mean-square differences between CAC and UKMO monthly SS T anomalies exceed 0.6-degrees-C in the regions where the correlation is lower than about 0.6. With the marked exception of the eastern trop ical Pacific, where the large El Nino signal is easily detected, there are insufficient numbers of SST observations to reliably define SST o r surface air temperature monthly mean anomalies over most of the ocea ns south of about 10-degrees-N. The use of seasons rather than months can improve the signal-to-noise ratio if careful treatment of the annu al cycle is included. For seasonal means, SST anomalies cannot be reli ably defined south of 20-degrees-S in the eastern Pacific and south of approximately 35-degrees-S elsewhere except near New Zealand. In ligh t of the noise estimates and the much fewer numbers of observations in the past, difficulties in establishing temperatures from the historic al record are discussed.