The atmosphere over the Indian Ocean perhaps holds the key to seasonal rainfall, and scientists are just beginning to understand its role.
Yet another monsoon is over and, with that, yet another test for the prediction skill of the statistical long-range forecast (LRF) model of the India Meteorological Department (IMD); this time it was a new model with a new methodology (Frontline, June 1). The years monsoon was bountiful a bumper kharif crop is expected as well but the new LRF model of the IMD, circa 2007, failed to predict the total seasonal (June 1 to September 30) rainfall for the country as a whole the Indian Summer Monsoon Rainfall (ISMR) accurately. Like last year, the temporal and spatial distributions of rainfall have been unusual (Frontline, November 3, 2006). However, it is too early either to say that this is reflective of a changing pattern of monsoon rainfall or to throw the new model out of the window.
For the past four years, the IMD forecast has been issued in two stages. The first one is given in April based on the March values of the parameters of the model. The second, an updated one issued in June, is based on the May-June data, taking into account the development of key forcing meteorological conditions such as El Nino/La Nina (the temperature anomalies over the tropical Pacific Ocean), the parameter sets being different for the two forecasts.
This year, the methodology of ensemble forecasts was adopted (Frontline, June 1) and appropriate inputs from the experimental forecasts of various Indian institutions as well as operational/experimental forecasts of international institutions were incorporated.
The statistical forecast model used eight parameters in all; while the April forecast was based on a five-parameter subset, the June one used a six-parameter subset. The April model predicted the ISMR to be 95 per cent of the Long Period Average (LPA), which is taken to be 89 cm (based on 1941-1990 data), with a model error of plus or minus () 5 per cent.
In June the following forecasts were issued: The ISMR was revised downwards to 93 per cent of the LPA with an inherent model error of 4 per cent. The July rainfall was predicted to be 95 per cent of the LPA (for July) with an error of 9 per cent. The seasonal rainfall for the four broad geographical regions of North-West India (comprising Jammu and Kashmir, Himachal Pradesh, Punjab, Rajasthan, Haryana, Chandigarh, Delhi, Uttarakhand and Uttar Pradesh), North-East India (Arunachal Pradesh, Meghalaya, Assam, Nagaland, Manipur, Mizoram, Tripura, Sikkim, West Bengal, Bihar and Jharkhand), Central India (Gujarat, Madhya Pradesh, Chhattisgarh, Maharashtra, Goa and Orissa) and South Peninsula (Andhra Pradesh, Karnataka, Tamil Nadu, Kerala, Lakshadweep and Andaman & Nicobar Islands) was predicted respectively to be 90 per cent, 98 per cent, 96 per cent and 94 per cent of the individual regional LPAs, all with an inherent model error of 8 per cent.
As the weekly progress of the monsoon (see diagram) shows, rainfall activity was delayed in early June. This was mainly because of the cyclonic storm Akash in the Arabian Sea in May, which disrupted the monsoon flow, and the super cyclone Gonu soon after over the East-Central Arabian Sea, which prevented its advance. From then on, rainfall activity has been more or less consistently above normal on a weekly scale except for the usual monsoon break in the second half of July. Peculiarly enough, with a slower-than-normal withdrawal of the monsoon, rainfall activity picked up towards the end of the season. There was significantly above normal rainfall even in the last week of September. The temporal distribution was as follows: + 19 per cent above LPA in June, - 3 per cent in July, - 1 per cent in August and +18 per cent in September. As a result, the total seasonal rainfall turned out to be far greater than the predicted amount.
The actual seasonal rainfall was 93.7 cm, an excess of 5 per cent over the LPA against the predicted deficit rainfall of 82.8 cm (93 per cent of the LPA). This is, in fact, 8 per cent above even the upper error margin of +4 per cent. The July rainfall (of 97 per cent of the LPA) was, however, as per the prediction of 95 (9) per cent.
The spatial distribution too has been peculiar. Of the 36 meteorological subdivisions, 30 received excess/normal rainfall and six deficient rainfall. Unusually, 13 subdivisions received excess rainfall (see Table 1). Of the deficient subdivisions, five (West Uttar Pradesh, Haryana-Chandigarh-Delhi, Punjab, Himachal Pradesh and East Madhya Pradesh) experienced moderate drought conditions (26-50 per cent deficit rainfall). Some subdivisions, such as Saurashtra, Kutch & Diu (+85 per cent) and Rayalaseema (+101 per cent), received unusually excessive rainfall (see map).
This skewed distribution is reflected in the deviations of the cumulative rainfall for the four regions (Table 2) against the regional LPAs. In terms of forecast, the model is quite off the mark for North-West India and South Peninsula and just about within the upper error margin for Central India. Only for North-East India, the rainfall was within the forecast window.
The above-average performance of the monsoon has thus been essentially because of the excess rainfall observed over South Peninsula and the significantly higher-than-average rainfall in Central India. This kind of rainfall distribution, with extreme activity in South Peninsular region and Central Indian region and suppressed activity in the North and the North-East is quite unusual. We have to go back to past records to see when we had such a peculiar monsoon, says M. Rajeevan, Additional Director-General of Meteorology (Research), and Director, Climate Research Centre, IMD, Pune.
Even as the IMD was predicting 2007 to be a deficit monsoon year in June, major climate research centres of the world, such as the European Centre for Medium Range Weather Forecasting (ECMWF) in the United Kingdom, and the National Centres for Environmental Prediction (NCEP), the International Research Institute (IRI) for Climate and Society, and the Experimental Climate Prediction Centre (ECPC) in the United States were predicting above-normal precipitation over the Indian subcontinent.
These forecasts using dynamical models (where inputs are based on physics) as against the IMDs statistical model (where inputs are based on empirical correlations) were based primarily on the prediction made in May of the occurrence of a La Nina event in the Pacific after July. In fact, the IRIs multimodel probability forecast for precipitation, issued in June, predicted that there was 40 per cent probability of an above-normal rainfall in the peninsular region.
El Nino and La Nina are components of the El Nino-Southern Oscillation (ENSO) system, which has been found to influence the inter-annual variability of the monsoon in a major way. El Nino refers to the warm ocean waters in the central and eastern tropical Pacific (off the Peruvian coast). The inverse of El Nino, namely the occurrence of abnormal cooling of the Pacific waters, is called La Nina. These conditions are said to arise when sea surface temperature (SST) anomalies are 0.5C or more. The associated changes in the atmosphere in the form of a see-saw pressure condition between Tahiti (in the Pacific) and Darwin (in the south-eastern Indian Ocean) are referred to as the Southern Oscillation. These result in the warm El Nino phase being associated with enhanced convection in the atmosphere above and the cold La Nina phase with a suppressed convection. A La Nina may follow an El Nino event, but not always. This time, however, there was an El Nino event beginning in September 2006 and a La Nina phase beginning in summer 2007 following it (see graph).
It has been found that the monsoon rainfall is more strongly correlated with the convection over the central Pacific (called Nino 3.4: 120-170 W, 5 N-5 S) than over the eastern Pacific. Generally, an El Nino condition has been found to result in deficit ISMR and La Nina in an above-normal ISMR. The El Ninos of 1982 and 1987 were associated with droughts and the La Nina of 1988 was associated with excess rainfall.
Data over the past 50 years show that whenever ENSO was favourable (La Nina) there were no droughts and whenever it was unfavourable (El Nino) there were no excess monsoons. However, 1997 was an exceptional year when there was a severe El Nino the strongest of the century, in fact but the seasonal rainfall, instead of being deficit, turned out to be above average (101.9 per cent of the LPA). It was not an excess monsoon though.
The predictions of a possible La Nina this year notwithstanding, the IMD forecast a deficit rainfall on the following grounds. During May and June, ENSO-neutral conditions had prevailed in the tropical Pacific. While most statistical models predicted continued ENSO-neutral conditions, most dynamical models indicated the development of La Nina in the following three months. Moreover, the IMD argued that, for the past months, dynamical models had predicted a stronger and more rapid cooling of the Pacific than had actually occurred. For the week ending June 13, the SST anomaly in Nino 3.4 was, in fact, marginally positive. Indeed, this had led to some centres revising their predictions to near-neutral conditions in the equatorial Pacific for July and August.
However, a mild La Nina developed after late July in the Nino 3.4 region (see graph) but, according to Rajeevan, this was too weak to have a significant influence over the Indian monsoon rainfall. In fact, even today meteorologists give a 50-50 chance of this developing into a full-fledged La Nina in 2007. Also, argues Rajeevan, the influence could not have been so immediate to cause increased rainfall in August itself. He says the reason for the excess rainfall in August and September may lie somewhere else and only detailed analysis will tell.
Monsoons in recent years have led to a rethink on ENSOs influence on the Indian rainfall. The strong El Nino event of 1997, which actually resulted in above-average rainfall, led K. Krishna Kumar of the Indian Institute of Tropical Meteorology (IITM), Pune, and associates to suggest in 1999 that the relationship between ENSO and the ISMR had weakened. This conjecture only seemed to be confirmed with the drought of 2002 which no model could predict when there was only a very weak El Nino.
It had been known for some time that SST anomalies in the Indian Ocean seemed to have some kind of countervailing effect on ENSOs influence on the monsoon. In 2001, T. Yamagata and his associates in the Institute for Global Change Research, Japan, identified a possible mechanism with that. They proposed that the newly discovered ocean-atmospheric coupled phenomenon in the equatorial Indian Ocean, called the Indian Ocean Dipole (IOD), was responsible for the offsetting effect. The IOD refers to the SST difference between the western equatorial Indian Ocean (off the eastern coast of Africa, from the northern half of Madagascar to the northern edge of Somalia) and the southeastern Indian Ocean (off the northern coast of Australia and throughout Indonesia).
The SST rises in the former and drops in the latter during a positive IOD phase and vice versa in the negative phase. They argued that a good monsoon is associated with a positive IOD phase, though the correlation of the IOD index with the ISMR was found to be poorer compared to correlation with the rainfall over equatorial western Indian Ocean and eastern Africa. This year too, according to Yamagata, the IOD has been in a positive phase. Yamagatas group had discovered the IOD phenomenon in 1999.
However, in recent analyses, Sulochana Gadgil of the Indian Institute of Science (IISc), Bangalore, and her associates have shown that while the IOD does not correlate as well with the ISMR, what they call the Equatorial Indian Ocean Oscillation (EQUINOO) correlates much better.
EQUINOO may be roughly regarded as the atmospheric component of the IOD; however, there is no one-to-one relationship between positive (negative) phases of the EQUINOO with positive (negative) phases of the IOD as SST anomalies depend not only on surface winds but also on the nature of oceanic upwelling in the region.
Sulochana Gadgil and associates showed that there existed a strong correlation between the extremes of ISMR (droughts and excesses) with a composite index of ENSO and EQUINOO, where the EQUINOO index is defined in terms of the anomaly of the equatorial component of the surface wind in the zone 60 - 90 E, 2.5 S - 2.5 N. It is surprising, they say in their most recent article in the July 25 issue of Current Science, that for such a complex phenomenon such as the Indian monsoon, the extremes can be explained in terms of links with just two modes: ENSO and EQUINOO. In fact, in a very recent analysis using data over 1881-1998, C. Ihara and others have conclusively shown that the ISMR is much better described by the EQUINOO index than by the IOD index.
Following Sulochana Gadgil and associates, the EQUINOO can be described as follows. A positive IOD phase is associated with a suppression of convection over the eastern equatorial Indian Ocean (EEIO) and an enhancement over the western equatorial Indian Ocean (WEIO) and a negative phase is associated with its reverse. While enhanced convection over the WEIO is associated with an easterly (east to west) anomaly of the equatorial surface wind, enhanced convection over the EEIO is associated with a westerly (west to east) anomaly. The oscillation between these two states is the EQUINOO.
An analysis of correlation of the ISMR data for the period 1979-2004 with the ENSO index and the EQUINOO index by Sulochana Gadgils group suggested that a positive EQUINOO, associated with enhanced convection over the WEIO, favoured a good monsoon. They found that each drought (excess rainfall) year was associated with unfavourable (favourable) phases of either ENSO or EQUINOO, or both. In 1997, both the indices were comparable but opposite in sign and the ISMR was normal.
In 1994, ENSO was unfavourable but EQUINOO was favourable and it was an excess monsoon. In 2002, though ENSO was weaker than 1997, EQUINOO was unfavourable as well and a severe drought occurred. Thus, they argue, with EQUINOO, we can explain not only the droughts that occurred in the absence of El Nino or in the presence of a weak El Nino, but also excess rainfall seasons in which ENSO was unfavourable. The worst droughts are associated with unfavourable phases of both the modes.
But, more significantly, what the group has found in an analysis of the ISMR for all seasons in the period 1958-2004 is that not only could extremes be explained but there was a clear separation between excess and deficit years when seen through the composite prism of ENSO-EQUINOO indices. However, the predictive power of this association of the ISMR with EQUINOO appears limited at present because, as its proponents emphasise, this relationship is simultaneous. That is, the EQUINOO index values in May, for instance, seem to have very poor correlation with the ISMR, whereas June values seem to have good correlation with the mid-season rainfall of July and August.
Similarly, July-August values seem to be well correlated with the September rainfall. This was clearly evident in the performance of the 2006 monsoon when, as late as late July the monsoon was below normal. Subsequent development of positive EQUINOO conditions led to the deficit being made up in August-September.
This year, according to Sulochana Gadgil, positive EQUINOO condition that is enhanced convection over WEIO has prevailed right from June. And that may well be the reason for the excess rainfall for the year and not La Nina as predicted by the dynamical models. She points out that dynamical models are yet to capture this Indian Ocean phenomenon.
In the earlier article (Frontline, June 1) it was pointed that the IMDs new statistical model had done away completely with regional parameters, which seemed to defy simple logic and was hard to understand. Maybe EQUINOO is the unpredictable component of the Indian monsoon and the missing key regional parameter for a statistical LRF model. Since the phenomenon is concurrent with the monsoon, it is not clear how the influence of EQUINOO could be incorporated into an LRF model.
Detailed studies focussing on EQUINOO are, therefore, now required to assess how well it is able to describe monsoons in the future and how best it can be used as a predictive tool. It would involve the untangling of the complex three-way interplay among ENSO, EQUINOO and the cross-equatorial flow of cloud-bearing systems over the subcontinent.
That may not be all that simple, points out Sulochana Gadgil. Until then, notwithstanding this years failure, the performance of the current IMD model should be observed for a few more seasons before tinkering with it.
