Reliable high resolution weather forecasts are in increasing demand to the
governments, industry, traffic, media, farming community and risk management
departments of most of the countries worldwide (Majewski, 1997). The Indian Summer
Monsoon Rainfall (ISMR) forecast for two weeks to one month in advance is one of the
most challenging task to the scientific community due to complex interactions between
land-air-sea and also small scale convective activities with large scale flow. The
summer monsoon season (June to September) contributes more than 70% of the
annual rainfall over India (Parthasarathy et al. 1994). The active equatorial intraseasonal oscillations (ISO) enhanced the convective activity over the north Indian
Ocean and moves northward to the Indian landmass. The onset of Indian Summer
Monsoon (ISM) over the southern tip of Indian peninsula marks the beginning of rainfall
season and ending of hot summer over the India. Though onset of ISM over Kerala on
1st June is considered the normal date for onset of ISMR according to India
Meteorological Department (IMD), it generally occurs during end of May or early June.
After the onset of ISM over Kerala, the monsoonal system marches towards the north
associated with rainfall. One of the important features of monsoon is the monsoon
trough, which, in general passes through the northern part of India such as Punjab,
Rajasthan, Uttar Pradesh, Bihar, West Bengal & north of Bay of Bengal. The fluctuation
of the ISMR mainly depends on the oscillation of the monsoon trough. In the active phase of monsoon the trough shifts to south of its mean position causing good amount
of rainfall over the country; on the other hand when it shifts to foothills of Himalaya, the
rainfall reduces over the central parts of the country and monsoon break occurs. In the
second half of September strength of monsoonal westerlies gradually decreases
leading to the withdrawal of southwest monsoon.
The Indian summer monsoon plays a crucial role for the agro economic country
like India. Major parts of the Indian population (more than 70%) explicitly depend on the
agriculture and their economies are highly dependant on the crop productions during
the summer monsoon season. Though the monsoonal system is a regular
phenomenon, but the Indian summer monsoon has a large abnormality in the global
climate systems. This abnormality varies from region to region and time to time. The
advance intimation of likely behavior of monthly and seasonal rainfall helps the farmer
to avail the opportunities and to make decisions that could enhance the farm
productivity and maximize returns or minimize the loss. Among various types of
forecasts made for different temporal scales viz. short range, medium range and long or
extended range, the extended range forecasts are highly valuable to the farming
community, government, industry for long term planning, decision making, management
and mitigation. The extended range prediction of monsoon rainfall over smaller regions
such as met-subdivision scale (Parthasarathy et al. 1994) is one of the challenging
tasks to the scientific communities.
The forecast products from General Circulation Models (GCMs) are being
effectively used all over the world for generating seasonal forecasts. The GCMs are the
important tools to simulate the atmospheric circulation. Present day, most of the GCMs
are coupled with oceanic models to take into account the interactions between the
oceans and the atmosphere. Although these numerical tools are required to understand
complex interactions between land-ocean-atmosphere systems globally, they are
computationally intensive and then, can only produce relatively low spatial resolution
simulations which in turn provide data in coarse resolutions on model spatial grid.
Therefore, direct application of GCMs output is often inadequate because of their limited representation of mesoscale atmospheric processes, topography and land sea
distribution in GCMs (Cohen 1990; von Storch et al., 1993). Consequently, the
performance of these model are poor in capturing small scale physical processes which
drive some important local/regional surface variables and their high resolution
properties such as precipitation (frequency of occurrence and intensity) and its strong
variability (Wood et al, 2004). Also, it is difficult to compare GCMs output to local
present observations (Vrac et al., 2007) and even more for extreme climate/weather
events (Vrac and Naveau, 2007) due to coarse resolution of GCMs. However,
comparison between local observations with the model simulation output is essential to
understand physical and dynamical processes of the atmospheric circulation in local
scale. In order to overcome these scale issues, it is important to reproduce information
from GCMs output in higher resolutions for better understanding the regional/local
weather/climatic phenomena though, this reproduced information for specific
geographic location may not coincide with the model grid.
A number of methods are used to convert GCMs output to required region. The
simplest method is to consider the nearby model grid points as the representative points
of the target region. This method often is not able to reproduce realistic features since
the representative points are in general, far away from the targeted region and the
surface characteristics of the representative points are also different. To improve the
nearest point forecast, a number of procedures are present that fall in general
calibration and downscaling techniques (Barnston and Smith, 1996; Goddard et al.,
2001; Landman and Goddard, 2002; Stephenson et al., 2005). These downscaling
techniques work as the bridge between climate forecasts and weather (Wilby and
Wigley, 1997; Huth and Kysely, 2000). In other words, downscaling is a technique which
links the state of some variables representing large space to the state of some variables
representing a much smaller space (Benestad et al., 2008). The field of downscaling is
divided into two approaches namely a) “Dynamical downscaling” based on nesting of
high-resolution regional climate models (RCMs) to simulate finer scale physical
processes consistent with large scale weather evaluation prescribed from a GCM
(Giorgi et al., 2001; Mearns et al., 2004; Lim el al., 2007) and b) “Statistical downscaling” adopts statistical relationships between the regional climate and statistical
characteristics of desired fields from the coarse resolution of GCM data (von Storch et
al., 1993; Wilby et al., 2004; Goodess et al., 2007). The downscaled high resolution
data can be used for forecast and as input into other types of numerical simulation tools
such as hydrological, agricultural and ecological models. Therefore, use of proper
downscaling techniques is the key issue for extended range prediction systems.
A brief overview has been given in 10.2 on the Extended Range Forecast
System (ERFS) and its present status with skill evaluated by various scientists
worldwide. Descriptions and methodologies of different downscaling techniques for
ERFS have been discussed in 10.3. Preliminary efforts with some experimental results
have been given in 10.4. Finally, the conclusions of this study have been presented in
10.5.