Home     About Us     Services
Home About Us Goals Projects Forecasts Links
2003 CFAM • PartnersPersonnelContact Us CalendarGalleryReportsLinksContact Us

OVERVIEW

Agricultural experts have long held that forecasts on 15-30 day time scales are optimal. In monsoon regions around the globe these time scales contain the largest variations in monsoon rainfall (intraseasonal). Forecasts on these time scales would allow for potential adjustments in agricultural planning, e.g. timing of planting and harvesting, crop choice and irrigation optimization.

Such a 20-25 day forecasts could have reduced significantly the impact of the Indian mid-summer drought of 2002.

The forecast scheme is described here and its relevance to a large range of forecasting problems in the monsoon regions are identified. This forecasting scheme and its application to real problems over a wider geographical domain forms the basis of a new project: Climate Forecasting Applications for Monsoons (CFAM).

 
Monsoon Weather and Climate Variability

Significant climate variability exists in monsoon regions of South and East Asia that directly impact regional agricultural practices, health issues, water resource management, and the general welfare of a large percentage of the planet’s population. There are four major time scales for climate variability:

 

• Weather events (1-5 days) such as tropical cyclones, storm surges, flash floods;
• Sub-seasonal (15-30 day) variability of monsoon rainfall as the monsoon waxes and wanes between active (wet) and break (dry) periods;
• Interannual (> 1 year) variability of monsoon rainfall; and,
• Long-term (decadal and centennial) that includes climate change and global warming.

 

Monsoon rainfall is characterized by a series of active (wet) and break (dry) periods lasting from 10-30 days.
These figures show the variability of the 5-day average observed rainfall (mm/day -black curves) over the Ganges Valley for the four summers from 1999 through 2002 plotted against pentad numbers. These years are very different but the largest variation in each takes place sub-seasonally with the monsoon oscillating between rainy (“active” periods of the monsoon) and dry (“break” periods) events. The red curves show the 20-day forecasts. The forecasts determine accurately the amplitude and phase of the intraseasonal variability.
The Ganges Valley or catchment is depicted as the purple region on this map.
The timing of these events is critical for planting and harvesting of crops and the allocation of water for irrigation and industrial, urban, and hydropower usages. For example, in 2002, a delayed commencement of rains in early June, coupled with a severe monsoon break in late June and early July caused devastating crop losses over India. Even though the overall Indian seasonally averaged rainfall was 19% below average, greater problems occurred because of the timing, intensity and duration of the July drought rather than the overall seasonal deficiency.

 

Climate Forecasts on 20-25 day Time Scales
Before our current work, little progress had been made in the forecasting of the sub-seasonal peaks and valleys in rainfall in the monsoon regions, such as those seen above.
Most effort had been put into trying to forecast the seasonal rainfall averaged over vast regions such as India. The results have been mixed, however. Traditional methods using large-scale climate indicators such as the state of El Nino-Southern Oscillation (ENSO) have failed in recent times, even though there has been some success during a large portion of the last century.
Even if the traditional seasonal forecasts were more successful, one could argue about their utility when applied to real agricultural and water resource problems.
So why do we want these 20-25 day forecasts?

• The sub-seasonal monsoon rainfall variations of the monsoon are far larger than the differences from year-to-year and the timing and magnitude of the sub-seasonal variation is critical for agriculture.
• Even if a forecast for an above or below average seasonal rainfall were accurate, it would not indicate which parts of the large scale area would be above average, normal or below average. Nor would it indicate when during the summer the active and break periods of the monsoon might occur.

Who would benefit from these forecasts?

A forecast of the July monsoon break 20 days in advance would have minimized the damage to agriculture during the 2002 monsoon.

A. R. Subbiah (Asian Disaster Preparedness Centre: ADPC Bangkok) notes that:

The minimum length of a forecast which will allow a farming community to respond and take meaningful remedial actions against either flood or drought is about 10 days although a forecast period of 3 weeks would be optimal. Assuming that a three week prediction were available by the third week of June 2002 …. farmers could have been motivated to postpone agricultural operations; saving investments worth billions of dollars…water resource managers could have introduced water budgeting measures …. Similarly, the prediction of the revival of the monsoon in the second half of July would have motivated the planners and farmers to undertake contingency crop-planning…

(Preliminary Assessment of the 2002 Indian Drought: ADPC, Bangkok, Thailand).

Using this 2-25 day prediction scheme, the extent and duration of the 2002 drought would have been evident at the peak of the June rainfall (near day 170) with the eventual resumption of substantial rains by early July.
In summary, a perfect forecast of the seasonally averaged Indian summer rainfall of –19% would have been useful. But, since the devastation of agriculture resulted from a cessation of rains from late June through most of July, and not from a general reduction of rainfall throughout the summer (see above - Figure for 2002), a two to three-week forecast of the intraseasonal variability of the monsoon would have allowed substantial mitigation of the drought’s adverse effects of the disastrous drought.


A Forecasting Scheme for 20-25 day Variability

During the last few years a new understanding has developed regarding the nature of the variability of the monsoon on sub-seasonal and interannual time scales. A detailed morphology of the sub-seasonal monsoon variability is described in the literature. [Webster, P. J., T. Palmer, M. Yanai, R. Tomas, V. Magana, J. Shukla and A. Yasunari, 1998: Monsoons: Processes, predictability and the prospects for prediction. J. Geophys. Res., 103, 14451-14,510]

Through observational, diagnostic and modeling efforts the sub-seasonal variability has been shown to be a very large-scale phenomena occupying, at least, the entire Indian Ocean basin. The phenomena possess much the same evolution 2-4 times per summer and also year after year.

The predictors are:
• Precipitation over central equatorial Indian Ocean and central India;
• Sea-level pressure over central India;
• Soil moisture over central India;
• Intensity of the low-level Somalia Jet stream, upper tropospheric equatorial zonal wind, surface winds over the equatorial Indian Ocean and the Arabian Sea and location and intensity of the upper tropospheric easterly jet stream;
• SST over equatorial Indian Ocean.
The quantity to be predicted (the predictand) is a quantity associated with the modulation of the monsoon on sub-seasonal time scale. One important predictand is the 5-day average rainfall over an area approximating the Ganges catchment region determined from GPCP satellite-based precipitation.

Wavelet analysis is applied to the predictand to determine the major bands in which variability resides. Four major bands are identified (<10 days, 10-40 days, semiannual and annual). The predictors are broken into the same bands and a linear regression technique is used to make the forecast. Each of these predictors is readily available in near real time so that timely forecasts are possible.

 
20-25 day Forecasts over India and Bangladesh
   
To test the technique for a smaller region, 20-day forecasts were made for the summer of 2001 and 2002 for the Indian States of Orissa and Rajasthan. Considerable skill still exists even for these smaller regions
These figures show the variability of the 5-day average observed rainfall (mm/day -black curves) over the Indian state of Orissa for the two summers of 2001 and 2002 plotted against pentad numbers.The red curves show the 20-day forecasts. The area is much smaller than the Ganges Valley but still considerable skill is depicted in the results. Likewise for the state of Rajasthan (not depicted here).
The Indian state of Orissa is depicted as the yellow region on this map.
The same technique has been used for the prediction of river discharge into Bangladesh with similar accurate results. Examples of the forecasts for the Brahmaputra and Ganges are given here for the year 1998.
The observed 5-day average discharge of the Brahmaputra and the Ganges at the boundaries of Bangladesh for the year 1998 (black curves) compared to forecasts of the same fields 20 days ahead (red curves). During the 1998 monsoon season there was extensive flooding in Bangladesh where 60% of the country was under water for over 3 months.
   
Utility and Scope of the Monsoon Forecasts
We anticipate that the scheme will be applicable wherever the slow physics of the sub-seasonal monsoon climate dominates such as in South and East Asia. North Australia and Indonesia are also candidate regions. We are confident that the discharge forecasting methodologies will be useable in other catchment areas such as the Mekong. We are currently investigating the utility of the techniques to Equatorial Africa and the African monsoon regions.
We anticipate that the use of such forecasts by decision makers may have a significant impact on agricultural and water resource practices in the monsoon regions. Currently, water resource management and cropping strategies are based on the climatological evolution of precipitation. Similar to the “Green Revolution” of the 1960s, substantial increases in yield may be expected if the forecasts are properly assessed, disseminated and acted upon. Significant lead times in the forecasting of floods as presently being tested in Bangladesh, can be expected.
To achieve this goal, there needs to be a strong interaction between scientists, government officials, policy makers and the user community. Optimization of impact requires the establishment of a set of common goals between these diverse groups. If these enactments were to occur, the impact of the intraseasonal forecasts may go beyond the substantial increase in yields. It may herald a truly “green” agricultural revolution as the use of pesticides and fertilizer would be used more efficiently and not increased as was necessary in the 1960s.
The monsoon regions are arguably the most vulnerable regions of this planet. This will be especially true over the next century in a world that is warming due to a changing climate. We are of the strong opinion that a society that develops an infrastructure to adapt to, and take advantage of, short-term climate forecasts is a society that will be best equipped to accommodate climate changes associated with global warming.