AN EVALUATION OF CLIMATOLOGICAL DATA IN THE INDIAN-OCEAN AREA

This paper presents evidence that both land-records and marine products currently used to assess the interannual and decadal variability of the monsoon system are subjected to important systematic errors due to non-climatic factors. The main inhomogeneities in the NCAR's World Monthly Surface Climatology during the 1900-1984 period affect all Sea Level Pressure (SLP) time series over the Indian subcontinent. These discontinuities are mainly linked to the non-consideration of changes in the time of observation between sample periods when the data have been compiled at NCAR. The main example of such discontinuity is observed in 1961 with the change from the World Weather Records collection to the Monthly Climatic Data for the World series. Systematic biases are also revealed-for Sea Surface Temperature (SST), Air Temperature (ART) and SLP ship-measurements compiled over the Indian Ocean during 1900-1986. The main spurious jumps occur around 1932 for SLP, 1940 for both SST and ART, and 1954 for SST sampled in the east Arabian Sea and the Bay of Bengal. In addition, an artificial trend contaminates SST reports during 1954-1976. Though no attempts have been made to determine the exact causes responsible for these inhomogeneities, there is little doubt about their origins because all these dates and trends are in phase with important changes in the composition of ''source-decks'' merged into the marine dataset, and do not agree with fluctuations of corrected SLP and temperature measurements along the coasts of the Indian subcontinent. Finally, substantial biases in resolution of the annual cycle can be expected for all the parameters because the sampling over the Indian Ocean has a strong seasonal dependence after 1950. Many inconsistencies in the observational picture of interannual SST variations over the Indian Ocean are thought to be linked to such data problems. A comparison between land and marine trends has however suggested the existence of some significant decadal-scale fluctuations in the Indian region during the 1900-1986 period. These true climatic happenings include prominently a warming temperature trend and a general SLP decrease over both land and ocean during 1900-1939, and a sudden warming in the Indian Ocean after 1976. This recent warming does not affect the interior of the Indian subcontinent and is in phase with persistent positive SLP anomalies for the whole Indian sector. This corroborates the evidence of a climatic change in the Indian Ocean after 1976 (Nitta and Yamada, 1989).