Abstract
The variability of the monsoon is investigated using a set of 90-day forecasts [MONEG (Tropical Ocean Global Atmosphere Monsoon Numerical Experimentation Group) experiments] and a set of AMIP-type (Atmospheric Model Intercomparison Project) long-term simulations of the atmospheric circulation with the ECHAM3 model. The large-scale aspects of the summer monsoon circulation as represented by differences of dynamical quantities between the two extreme years 1987 and 1988 were reproduced well by the model in both kinds of experiments forced with observed sea surface temperature (SST). At the regional scale the difference of precipitation over India during summer 1987 and 1988 was well reproduced by the model in the 90-day forecasts using interannually varying SSTs; however, similarly good results were achieved in forecasts using climatological SSTs.
The long-term simulations forced with interannually varying SST at the lower boundary of the atmosphere over a period of 14 years, on the other hand, only partly reproduce the observed differences of precipitation over India between 1987 and 1988. For the ensemble mean of five simulations averaged from June to September and for the whole of India an increase from 1987 to 1988 is simulated by the model as observed but with smaller values. The difference in observed precipitation between 1987 and 1988 is of opposite sign for May to that for September. The simulations and observations agree in the manifestation of this sense of opposing variability within a monsoon season for these two years and also for other years. The simulations and observations differ most during July.
The paper concentrates on the question why the interannual variability in the long-term simulations on one hand and the 90-day forecasts and in the observations of precipitation on the other hand differ so strongly during the peak of the monsoon in July. Large-scale dynamics over India are mainly forced by the anomalies of Pacific SST. For the variability of precipitation over India other forcings than the Pacific SST are important as well. Due to enhanced evaporation, warmer SSTs over the northern Indian Ocean lead to increased precipitation over India. Changes in the SST there within the range of uncertainty (0.5 K) can lead to clear impacts.
As a further boundary forcing, the impact of soil moisture is investigated. The use of realistic soil moisture differences between 1987 and 1988 in the MONEG forecasts resulted in improved skill of precipitation forecasts over India. Also the two individual AMIP simulations with realistic precipitation differences over India had more realistic soil moisture differences over east Asia in the beginning of the monsoon season between the two years than those experiments that failed to produce the correct precipitation differences.
The years 1987 and 1988 were quite different with respect to the phase of the stratospheric quasi-biennial oscillation (QBO). As atmospheric circulation models cannot yet reproduce stratospheric QBOs realistically, their impact was tested by nudging observed QBOs into AMIP simulations for July 1987 and 1988. Seven out of eight experiments showed an impact toward a more realistic simulation of precipitation over India; however, during the west phase of the QBO (1987) impacts are very small.
None of these forcings gave a dominant effect. If this finding is confirmed by further experimentation, improvements of practical long-range forecasts may be very difficult as two of these quantities are hardly known with the required accuracy (northern Indian Ocean SSTs and the Eurasian soil moisture) and because models are not yet able to simulate the stratospheric QBO realistically.
This study confirms that El Niño has two direct effects: it reduces the precipitation over India and reduces the surface winds over the Arabian Sea. Due to the latter, the SST of the Arabian Sea can increase as there is less mixing and upwelling in the ocean. Here it is suggested that because of this increased SST there would be more precipitation over India, thus counteracting the expected decrease from the direct El Niño effect.
Sensitivity experiments were carried out with the ECHAM3 model to substantiate this hypothesis. The results may be model-dependent and model deficiencies might influence sensitivities from boundary forcings adversely. Therefore observational data have been investigated as far as possible to seek independent confirmation of the findings obtained through the model simulations.
Corresponding author address: Dr. Klaus Arpe, Max-Planck-Institut für Meteorologie, Bundesstr. 55, D-20146 Hamburg, Germany.
