Frequently Asked Questions (FAQ) about Long Range Forecasting (LRF)

In science, the forecasting means the process of estimation of the value of some variable at some future time. One of the primary functions of the national Weather services is forecast of weather parameters such as rainfall, temperature, wind, humidity etc. over a region averaged over a particular time period. For example forecast of daily rainfall (rainfall averaged over a day).

As per the World Meteorological Organization (WMO) definition, long range forecast is defined as the forecast from 30 days’ up to one season’s description of averaged weather parameters. The monthly and seasonal forecast comes under long range forecast.

In general, three approaches are used. These are (i) statistical method (ii) numerical weather prediction or dynamical method and (iii) dynamical cum statistical method. From the beginning, the main approach towards the long range prediction has been based on statistical methods. IMD’s operational forecasts for the monsoon rainfall are based on this technique. The statistical method involves identification of predictive signals (predictors) that having significant and stable historical relationship with the predictand and predicting the value of predictand at future time. For this purpose, it is assumed that the observed preidctand-predictor relationship persists in the future also and that the predictor values corresponding to the future predictand value to be forecasted are known.

An alternate approach towards the prediction of ISMR is based on numerical models or so called General Circulation Models (GCM) for the simulation of atmospheric and oceanic conditions. Though the numerical prediction models have the potential for providing predictions over smaller spatial and temporal scales as per the user demands, they have not so far shown required skill for simulating the salient features of the mean monsoon rainfall and its interannual variability. For improved rainfall simulations, GCM models should be able to account for local sub grid features and sub-seasonal variability of the climatic fields.

The third approach of dynamical cum statistical method is based on the fact that GCMs have better skill in simulating large scale atmospheric circulation features and that there exits a semi empirical relation between rainfall over a region and prevailing large scale circulation features over both global and regional scales. Therefore, it is possible to derive recalibration equations between the rainfall and circulation features simulated by GCM models and assuming that these relationships will hold good in the future, regional rainfall can be predicted. The dynamical cum statistical method is a recent development in the long range prediction.

India Meteorological Department is solely responsible for issuing operational long range forecast for India. The forecasts are prepared at the National Climate Center of IMD located at Pune. At present, empirical (statistical) methods are used for the preparation of operational long range forecasts.

In addition to India, there are several other countries like United States, United Kingdom, Australia, South Africa, Brazil etc., which use empirical methods extensively for long range forecasting. For example, for the long range forecasting of ENSO, many international climate centers use empirical models.

IMD issues operational long range forecast for the rainfall during SW Monsoon Season (June- September). These forecasts are issued in two stages. The first stage forecast is issued in mid-April and consists of quantitative forecast for the season (June to September) rainfall over India as a whole. The second stage forecasts issued by the end of June consist of update for the forecast issued in April, a forecast for July rainfall over the country as whole and forecasts for seasonal rainfall over broad rainfall homogeneous regions of India.

IMD also prepares forecasts for winter (Jan- March) precipitation (issued in the end of December) over Northwest India and northeast monsoon (October-December) rainfall over Southern Peninsula (issued in October). However, these forecasts are issued only to the government.

Looking at the potential of numerical models, IMD has also established an experimental prediction system based on General Circulation Model (GCM) in addition to its existing operational forecasting system based on statistical models. For this purpose, IMD uses the seasonal forecasting model (SFM) developed at the Experimental Climate Prediction Center (ECPC), Scripps Institute of Oceanography, USA. The skill of the numerical model based forecasting system is to be validated for some more years before the same can be used for operational purpose.

The monsoon prediction in our country is being done with reasonable accuracy. The success rate of IMD forecasts since 1988 has been high. During the last21 years (1988-2008), IMD forecasts were qualitatively correct in 19 years (i.e. 90% of years). The exception was during years 2002 and 2004 both of which were drought years. However, in some years (1994, 1997, 1999, 2002, 2004 and 2007) the forecast error (difference between actual rainfall and forecast rainfall) was more than 10%. The 2002 drought was due to exceptionally low rainfall during the month of July (46% of long term period) caused by unexpected sudden warming of sea surface over equatorial central Pacific that started in the month of June. It may be mentioned that the exceptionally deficient rainfall of July, 2002 was not predicted by any prediction group in India or abroad. It is not possible to have 100% success for forecasts based on statistical models. The problems with statistical models are inherent in this approach and are being faced by forecaster world wide.

The long-range forecast of monsoon rainfall is very crucial as the inter-annual variation of monsoon rainfall has many social and economic impacts. The total monsoon rainfall during the season has a statistically significant relationship with the crop yield, generation of power, irrigation schedule etc. over the country. In general, a weak monsoon year with significantly low rainfall can cause a low crop yield. On the other hand, a strong monsoon is favorable for abundant crop yield, although sometimes too much rainfall may cause devastating floods. There is in phase variation of the rice production in India with the all India summer monsoon rainfall. Over India, the monsoon rainfall accounts for about 75-80% of the total annual rainfall; in large areas of central and northwest India, the monsoon contribution to the annual rainfall is 90% or more. Thus there is a pressing need to understand the Indian monsoon and forecast its interannual variability on long range scale.

The predictability of day-to-day weather patterns in the tropics is restricted to 2-3 days. The seasonal mean monsoon circulation in the tropics, on the other hand, is potentially more predictable. This is because the low frequency component of monsoon variability is primarily forced by slowly varying conditions like sea surface temperature, snow cover, soil moisture etc. Therefore, it is possible to develop models for the long range forecasts of monsoon seasonal rainfall over the country as a whole. However, there is some limit in the seasonal predictability as the mean monsoon circulation is also influenced by the internal dynamics/variability.

Governments and industries, to whom the knowledge about the future weather patterns will help in making decision such as determination of how much food material have to be procured and stored each year, when and how much fertilizers or seeds have to transported to each part of the country, which of the areas has to be prepared for natural calamities such as floods, droughts etc. Long range forecast are also useful for farmers who are looking to get the most yield out of their arable land and crop insurance companies in deciding area based premium of their weather related insurance policies.

A time series of any variable has got a mean and a standard deviation. In general when the value of a variable is within the 1 standard deviation of either sides of its mean value, we can say that of the variable is within the normal range or simply “normal”. When the value of the variable is 1 standard deviation above (below) its value, we say the value is “above (below) normal”. In case of monsoon season (June to September) rainfall over India as a whole, the mean value (generally mentioned as long term average) is 89 cm and standard deviation is 9cm (about 10% of mean value). Therefore, when the rainfall is within ±10% of its long term average, the rainfall is said to be “normal” and when the rainfall is 10% more (less) than its long term average, the rainfall is said to be “above (below) normal”.

A computer model with mathematical equations that describe the physical process of the atmosphere.

Southern Oscillation or “SO” is a "see-saw" in the surface air pressure between eastern and western tropical Pacific. It is characterized by simultaneously opposite sea level pressure anomalies at Tahiti, in the eastern tropical Pacific and Darwin, on the northwest coast of Australia. The SO was discovered by Sir Gilbert Walker in the early 1920's. Later, the three-dimensional east-west circulation related to the SO was discovered and named the Walker Circulation. The SO has periodicity of about 2-5 years. A most common index of SO is computed as the difference between standardized sea level pressure anomalies at Tahiti and Darwin (Tahiti – Darwin).

El Nino and La Nina are oceanic manifestation of opposite phases of SO, which is an atmospheric phenomena. El Nino is characterized by the warming of the sea surface temperatures in the central and eastern equatorial Pacific Ocean, beginning at about Christmas time (hence the name "El Nino", which is a reference to the Christ child). This is called the warm phase of the SO. The cold phase of the SO, called "La Nina" is characterized by high pressure in the eastern equatorial Pacific, low in the west, and by anomalously cold SST in the central and eastern Pacific.

ENSO (El Nino Southern Oscillation) is an acronym designed to stress the fact that the El Nino and SO are components of the same global Ocean-Atmosphere coupled phenomena.

Both monsoon and ENSO are ocean-atmosphere couple phenomena. There is a general inverse relationship between monsoon and ENSO. The warm phase of ENSO is generally associated with weaker than normal Monsoon and vice versa. During the period 1885-2007 there were 36 years of warm ENSO (El Nino) and 25 years of cold ENSO (La Nina). During the 15 of the 35 El Nina years (42%), Indian summer monsoon rainfall (ISMR) was below normal and 9 of the 25 La Nina years (36%), ISMR was above normal. This shows that there is no one to one correspondence between ENSO and ISMR.

The monsoon system is a planetary scale system, and it has large variability at different spatial and temporal scales. The long range forecasting is mainly used to predict the interannual variability of the monsoon rainfall over large region. In addition to the influence of various global and regional factors over the monsoon circulation, the monsoon rainfall over a region depends on the local factors such as geography of the area. As a result, smaller the area we consider, the larger will be the variability of the rainfall over the region. Therefore, it may not be easy to model such a large variability of rainfall with the help of predictors. That is why even during the recent series of successive normal monsoons (1988-2001) over the country as whole drought conditions were prevailing over many parts of the country.

Probabilistic LRF (Long Range Forecast) provide probabilities of occurrences or non-occurrences of an event or a set of fully inclusive events. Probabilistic LRF can be generated from an empirical model, or produced from an Ensemble Prediction System (EPS). The events can be classified into categories (above/below normal or above/near/below normal for example). A lthough equi-probable categories are preferred for consistency, other classifications can be used in a similar fashion.

The Indian Ocean Dipole (IOD) also known as the Indian Niño is an irregular oscillation of sea-surface temperatures in which the western Indian Ocean becomes alternately warmer and then colder than the eastern part of the ocean. The IOD also affects the strength of monsoons over the Indian subcontinent. A significant positive IOD occurred in 1997–8, with another in 2006. The IOD is one aspect of the general cycle of global climate, interacting with similar phenomena like the El Niño-Southern Oscillation (ENSO) in the Pacific Ocean.

The Madden Julian Oscillation (MJO) is one of the most important atmosphere-ocean coupled phenomena in the tropics, which has profound influence on Indian Summer Monsoon. The MJO is the leading mode of tropical intraseasonal climate variability and is characterized by organization on a global spatial scale with a period typically ranging from 30-60 days, which was discovered by Madden and Julian in 1971 in a published paper. It has the following characteristics:- • MJO is a massive weather event consisting of deep convection coupled with atmospheric circulation, moving slowly eastward over the Indian and Pacific Oceans. • MJO is an equatorial traveling pattern of anomalous rainfall that is planetary in scale. • Each cycle lasts approximately 30–60 days. Also known as the 30-60 day oscillation, 30-60 day wave, or intraseasonal oscillation (ISO). • The MJO involves variations in wind, sea surface temperature (SST), cloudiness, and rainfall. • Based on the place of convective activity the period of MJO is divided into 1-8 phases with each phase roughly last for 7 to 8 days. Since the MJO is the most important mode of tropical intraseasonal variability with potentially important influences on monsoon activity in the Asian regions on extended range time scale (beyond 7 days to on1 month), the capability of statistical or numerical models in capturing MJO signal is very crucial in capturing the active/break cycle of monsoon.

Air pressure, temperature, mountain ranges, ocean currents and many other factors combine to produce an enormous quantity of interacting variables all of which can alter the weather to a greater or lesser extent. However, greater understanding of the science, plus the use of powerful computer models, continue to improve our ability to make more accurate predictions with longer lead times.

The Standardized Precipitation Index (SPI) is a tool which was developed primarily for defining and monitoring drought. It allows an analyst to determine the rarity of a drought at a given time scale (temporal resolution) of interest for any rainfall station with historic data. It can also be used to determine periods of anomalously wet events. The SPI is not a drought prediction tool.