Annex XII
APDC/01/8

LONG-RANGE CLIMATE FORECASTS FORAGRICULTURE AND FOOD SECURITY PLANNINGAND MANAGEMENT IN ASIA AND THE PACIFIC

By
A.R. Subbiah
Kamal Kishore
Asian Disaster Preparedness Center (ADPC)^*
Thailand

Abbreviations and Acronyms

While climate variability associated impacts on crop production vary from year to year, there has been a significant increase in crop production loss during certain years due to occurrence of extreme climate events. For instance, in Indonesia, the rice crop area affected by drought during 1997-98 extended up to 867,997 hectares resulting in a loss of around 3 million tonnes (Govt. of Indonesia, 2000). The reduced rice production coinciding with economic crisis lead to 300 per cent increase in price of rice. The Government of Indonesia imported around 5 million tonnes of rice in order to maintain price levels and ensure food security to substantial number of households (ADPC, 1998). In the Philippines, during 1997-98, drought affected 600,000 hectares of rice and corn lands resulting in loss of around 1.2 million tonnes of food-grain (ADPC, 1997). In Fiji, 50 per cent of the annual sugarcane production was lost affecting 200,000 farmers during 1997-98 (Kaloumaria, 2001). In Bangladesh, a severe flood in 1998 caused crop loss of around 2.2 million tonnes affecting 25 million people (FAO, 1998).

It has long been recognized if society could have advance information on weather the adverse effects associated with it could be minimized. Prevalence of traditional forecast practices in various parts of the world reflects the demand for long-range forecasts to manage uncertainties associated with climate variability. Recent advancements in the climate prediction promise huge benefits for the society. However, 1997-98 El Niño management experiences reveal that a large gap exists between potential value of forecast information and the actual utilisation of this information for managing agricultural systems for societal benefits. There is a need to take concerted efforts to address these gaps to take full advantage of climate prediction advancements.

This paper focuses on issues relating to evolution of long lead forecast over a period of time in the Asia-Pacific monsoon region, recent advances in ENSO prediction and its potential value for managing agricultural systems. The paper while drawing experiences from the application of long lead forecast in recent years also highlights emerging issues for future action.

The paper draws information from Australia, India, Indonesia, the Philippines and Vietnam. The paper is structured into the following sections.

Section - 1

1.1 Evolution of long-range forecasts in the tropics

Traditional forecast

1. In many parts of the Asia-Pacific regions, agriculture plays a significant role in sustaining livelihood systems of the communities. The year to year variability of monsoon behaviour prompted agrarian communities to search for potential clues to put in place advance measures to manage risks. They have developed their own methods of climate forecasting based on generations of experience, local religious beliefs and close observations of their environment to anticipate weather patterns (Sukradi W., 1998). The communities rely on interpretation of cloud colour and form, animal behaviour, flowering of certain plants etc. as indicators of seasonal conditions. Many of these observations have been coded in folk songs, rhymes, and thus passed through generations. While use of these prediction methods varied from community to community, the prevalence of traditional forecasting methods reflects the demand for advance climate information to cope with climate variability in planning for agricultural operations (Eakin 2000).

1.2 Indian monsoon

2. Efforts were made to forecast monsoon behaviour one month and beyond on scientific basis in the late 19th Century in India and subsequently in Indonesia.

3. In 1877, India experienced a serious famine caused by highly deficient monsoon rainfall. In response to this, the Government of India called upon H.F. Blanford to prepare monsoon forecasts (Krishna Kumar, 1994). Thus, Blanford was the first to attempt a forecast of the monsoon based on the hypothesis that "varying extend and thickness of the Himalayan snows exercise a great and prolonged influence on the climate conditions and weather of the plains of northwest India" (Blandford 1884) The success of Blanford's tentative forecasts during 1882-85 encouraged him to start operational Long Range Forecast (LRF) of monsoon rainfall covering the whole of India and Burma in 1886. Since then the LRF of the monsoon has become an important operational task of the India Meteorological Department (IMD). Sir John Elliot (1895) utilised the weather conditions over the whole of India and surrounding regions to prepare an elaborate forecast of the monsoon rainfall. The forecasts after 1895 were based on: (i) Himalayan snow cover, October to May (ii) local peculiarities of pre-monsoon weather in India, and (iii) local peculiarities over the Indian Ocean and Australia (Thapliyal 1987).

4. Sir Gilbert Walker (1910) did pioneering work on the Southern Oscillation and published a prediction formula for India monsoon forecasting containing 22 predictors in six forecast formulae differentiated regionally (Banerjee, 1950).

5. Walker also succeeded in removing the subjectivity in the earlier forecast methods by involving, for the first time, the concept of correlation in the field of LRF of monsoon rainfall. Subsequent to Walker's work, very little progress was made in LRF of monsoon rainfall until the early 1980s. In 1988, a new forecast model known as "Parametric and Power Regression Model" was developed and put into operational use by the India Meteorological Department. This model uses 16 regional and global land-ocean-atmospheric parameters, which are physically related with the Indian monsoon. Combining these parameters, a Power Regression Model has been developed to provide quantitative monsoon rainfall forecast for the country as a whole. The values of individual parameters that go into the model calculations are based on observations over different periods specific to the individual parameters. Some of these parameters are derived from the Indian area itself, while some of them are obtained from as far as South America. For a few of the parameters, observations right up to end of May are required, while some others are known during the preceding winter itself. Thus IMD is in a position to issue the monsoon forecast only by the end of May i.e. before one month of the start of the monsoon season (June-September).

6. The forecast issued with the help of this technique since 1988 has been found to be fairly accurate. A table showing the forecast and actual performance of the monsoon since 1988 is shown in the table 1.

1.3 Indonesia and Australia Monsoon

7. The inter-annual variability of rainfall over Indonesia attracted the attention of Dutch meteorologists. As early as 1919, Braak recognized the essential relationship between long-term pressure, wind and rainfall variations in the Indonesian regions. He established that forecasting for the east monsoon rainfall (June-September) was possible with atmospheric pressure as sole indicator (Jose A.M., 1989).

8. Of great interest for Indonesian monsoon forecasting was Nicholl's (1981, 1983) direct follow-up of Braak's (1919) ideas more than 60 years earlier, namely to use Darwin pressure in the first half of the calendar year to predict Java rainfall during the second semester. Nicholl's (1981, 1983) constructed a linear single-parameter regression model of Djakarta September-November rainfall versus Darwin August pressure during 1951-1969 and then used this relationship to predict the rainfall during each of the years 1970-1980. The model proved to be capable of explaining 44 per cent of the inter-annual rainfall variance.

9. Nicholls and Woodcock (1981) and Nicholls (1981) demonstrated that spring (September-November) rainfall in North Australia region could be predicted some months in advance by verifying successfully, earlier work of several pioneers of long-range forecasting (e.g. Walker, Braak, Berlage and Qualyle). The only predictor used in the studies was monthly mean Darwin surface pressure, observed several months earlier. Nicholls et. al., (1982) extended the earlier work by defining a date of wet-season onset in the area around Darwin and demonstrated that this onset date could also be predicted some months in advance.

10. Improvements have been made in the seasonal prediction of the Indonesian rainfall and the current scheme has been in operation since 1993. This is essentially a statistical-analogue scheme based on a very detailed analysis of rainfall data for 102 regions. The seasonal forecasts are based on the following techniques;

11. Most of the studies in Long-range forecasts are primarily based on statistical and empirical techniques. Diagnostic studies of historical datasets over the years have produced several predictors for monsoon rainfall forecasting. These parameters represent different components of the coupled atmosphere-ocean-land system.

12. Although a large number of predictors have been identified so far, uncertainty still prevails in identifying the best set of predictors in view of multi-co linearity, temporal variations of the relationships, lack of knowledge about the exact physical mechanisms relating to cause-and-effect.

13. These long lead forecasts have been used as general alerts to the national policy makers to mobilize resources to manage potential natural hazards.

1.4 El Niño and Southern Oscillation (ENSO) prediction

14. The monsoon forecasts of Australia, India and Indonesia considered southern oscillation as one of the predictors since early part of the 20th century. However, the connection between El Niño and Southern Oscillation was appreciated in the late 1960s principally through the work of Jacob Bjerkins. He discovered the tropical coupling between El Niño and Southern Oscillation (ENSO).

15. The last decade witnessed a major advance in understanding the predictability of the atmosphere at seasonal to inter-annual time-scale (Palmer and Anderson, 1993; NRC, 1996; Carson, 1998). The major impetus in current seasonal to inter-annual time scale prediction efforts was provided by the Tropical Ocean and Global Atmosphere (TOGA) Programme, which was carried between 1985-1994. Results from TOGA demonstrated that it is possible to predict Pacific Ocean El-Niño and Southern Oscillation (ENSO) related Sea Surface Temperatures (SST) over time scales extending from a few months to over 1 year.

16. ENSO, is one of the known key drivers to inter-annual variability, and have been associated with worldwide extreme climate anomalies, including changes in the space-time patterns of floods, droughts, cyclone/severe storms activity, cold/hot spells etc. (Rasmusson and Wallace, 1983; Cane et at., 1986; Ogallo, 1988; Ropelewski and Halpert, 1989; NRC, 1996).

17. The following ENSO patterns have intrinsic value for predicting future climatic conditions.

18. Annual Phase locking: No two El Niño events are ever the same. They differ in intensity and duration. Yet, as research continues on El Niño - Southern Oscillation phenomena, certain patterns are discernible. One such significant pattern is that El Niño episodes are generally phase-locked to an annual cycle. This means that if an El Niño (or its reserve, La Nina) becomes established by the middle of a calendar year, it will not alter until sometime early in the following year. As a consequence, most El Niño begins and end during the period between March and June.

19. This `phase-locked' pattern is particularly critical for countries like Australia Indonesia and the Philippines since such patterning has its effects on different seasonal monsoon. The most important period of rainfall occurs during crucial monsoonal months. The intensity of an El Niño episode during these months has a considerable effect on rainfall and cropping systems in virtually all areas of high production and high population (Fox J., 2000).

20. Amplified climate variability: One feature of rainfall fluctuations in areas affected by ENSO is the large inter-annual variability. Conrad (1941) examined the dependence of inter annual rainfall variability on the long-term mean annual rainfall. He found a strong relationship between relative variability (defined as the mean of the absolute deviations of annual rainfalls from the long-term mean, expressed as a percentage of the long-term mean) and the long-term mean precipitation. The relative variability decreased, in general, as the mean precipitation increased. Some of these deviations were due to the influence of ENSO phenomenon on rainfall. Nicholls (1988) and Nicholls and Wong (1990) confirmed on recent data that the ENSO amplifies rainfall variability in the areas it affects, relative to other areas. The amplification factor is substantial in certain areas, which also depend upon latitude and mean rainfall.

21. Biennial cycle: The phase locking, related to a biennial cycle, is a fundamental element of ENSO variability (Rasmusson et. al. 1990). The biennial mode means the El Niño events will often be preceded and/or followed by La Nina episodes, and vice-versa. In terms of rainfall, this means that year-to-year changes in rainfall can be extreme. The change from El Niño related drought to La Nina and wet conditions can be rapid, and usually occurs early in the calendar year (Nicholls 2000).

22. Seasonal variations in predictability: The SOI has a strong degree of persistence, such that it tends to retain the same value over the following months. Thus once an ENSO event is established, its persistence from one season to another can be used as a tool to predict seasonal rainfall. For example in Indonesia, following the methodology applied to Australia by McBride and Nicholls (1983), maps of lag correlation have been produced between the SOI and Indonesian rainfall in the following season. The greatest simultaneous relationship with SOI is for both June, July, August and September, October, November rainfall has been recognized. For this reason, three-month lag correlation between SOI and these two seasonal rainfalls has been established.

23. With regard to the June, July, August rainfall, statistically significant correlations exist over the central part of the country. Thus it opens the possibility of predicting June, July, August rainfall using the average value of the SOI from March to May. On the other hand, insignificant predictive information is found for most stations located over the `non-monsoonal' areas, i.e. the northern part of Sumatra and Kalimantan as well as central Irian Jaya.

24. Stronger evidence for predictability using the precursor value of SOI is found for the September, October, November rainfall. This suggests that the low 1997 September, October, November rainfall experienced by most of these stations could actually have been predicted in advance (Kirano, 1998).

25. Thus, the ability to forecast some aspects of ENSO signals for time scales of months to over one year are currently being used to extrapolate the potential occurrences ENSO related extreme weather/climate events for specific seasons and regions of the world which have strong ENSO signals. Such information now forms crucial components of early warning systems, including the planning, management and operations of agricultural activities in some parts of the tropical regions.

Section - 2
Potential value of ENSO forecasts

2.1 Farm level

26. The long range forecast could provide the indications of monsoon rainfall variability. There were at least four significant aberrations in the rainfall behaviour could upset established crop calendars and crop yields. These are:

1. The commencement of rains may be quite early or considerably delayed
2. There may be prolonged `breaks' during the cropping season
3. There may be spatial and/or temporal aberrations
4. The rains may terminate considerably early or continue for longer periods.

27.To deal with the above-mentioned aberrations, the farmers could respond to forecasts to undertake the following measures:

• Change variety for one with shorter or longer duration.
• Change crop species or mix of species, especially combinations of cash and food crops.
• Implement soil and water conservation techniques.
• Increase or decrease area planted, either total, by crop, or by upland or lowland location.
• Adjust timing of land preparation.
• Increase or decrease borrowing for inputs.
• Sell or purchase livestock depending on anticipated cost and availability of feed.
• Remain in village or migrate to seek off-farm employment or better grazing for livestock.

28. In Java and Eastern Indonesian region, for example, in El Niño years, farmers are frequently misled by initial rains, which offer promise but then cease. The false rains tempted farmers to plant. However, as the rain cease later, the crops usually die due to dry spell. Most farmers keep some seed reserves in case they are forced to plant a second time during the wet season. Rarely do farmers have sufficient seed reserves for a third attempt at planting and by the time such a third planting seems necessary, there is little likelihood of success. In most of the El Niño years, the incidences of false rains were noticed. A long lead forecast could help farmers to wait till setting in of regular rains (Fox J., 2000).

2.2 Provincial level

29. Water Resource Management: The water resource management managers at the catchment/watershed level/river basin level could undertake pro-active measures to manage water resources. There is a potential possibility to introduce water budgeting arrangements to prioritise water use and allocate water resources among various competitive users. In areas where water availability for irrigation purposes is scarce, a campaign can be launched to advise farmers to provide minimum irrigation only at the critical crop stages. The lead-time available could be used for augmenting water resources by constructing small scale water harvesting structures and rehabilitate old irrigation structures.

30. Compensatory Cropping Programme: This has two dimensions. One is to try to compensate the crop loss in most severely affected areas (MSA) by intensifying the production programme and increasing yield in the most favorable areas (MFA) where there was expectations of good rainfall and availability of assured irrigation sources. The second is to make up the crop loss in the same area by taking up short duration cultivars.

31. Alternate Cropping Strategy: This strategy involves shifting of crops, which could be grown on the availability of soil moisture less than normal conditions. Farmers in Indonesia usually adopt this strategy by replacing paddy crop with maize and other secondary crops. The success of this strategy could depend on the government intervention in providing input and market support to the farmers.

32. The above-mentioned approaches needed to be matched with irrigation potential and agro-climate zonation maps to evolve suitable cropping pattern, keeping in view El Niño influences on rainfall pattern in various regions.

33. The provincial level institutions would have lead-time to provide agricultural input support, credit arrangements and technical advisories to enable the farmers to undertake contingency crop plants. The provincial level administration could also provide support for marketing the agricultural products.

2.3 National level

34. The national level institutions could provide the necessary support to provincial administration and the farming communities in terms of resources. The national government can undertake policy decisions to map out potential impact areas and target resources for mitigation measures. The national government could also undertake policy measures for export/import of agricultural commodities. The National Government could undertake measures to plan for food logistics such as procurement of food grains, transport and distribution to the potentially affected areas.

2.4 Insights from Crop Models

35. Climate patterns translate via rainfall variability into associated crop production variability. However, rainfall anomalies are not the only determinant of yield and factors such as starting soil moisture, temperature, planting dates and timeliness of rainfall strongly influence final yields. The crop simulation models capture these effects. As there is a potential value to integrate climate forecast information into crop models, certain initiatives have been taken recently to put in place-integrated climate crop models in Australia. Based on Australian experiences, these experimental projects are being tested in certain sites of the developing countries.

36. The following paragraphs capture the potential value of integrated climate crop models in Australia and India.

2.4.1 Australia

37. Significant, physically based lag-relationships exist between an index of ENSO and future rainfall amount and temporal distribution in eastern Australia. Australian scientists have shown how phases of the Southern Oscillation Index (SOI) are related to rainfall variability and are useful for rainfall forecasting for many locations in eastern Australia. For large parts of Eastern Australia, they have shown that a rapid rise in SOI over a two months period is related to a high probability of above long-term average rainfall at certain times of the year. Conversely, a consistently negative or rapidly falling SOI pattern is related to a high probability of below average rainfall for many regions in Australia at certain times of the year. As the SOI pattern tends to be `phase-locked' into the annual cycle (from autumn to autumn), the SOI phase analysis provides skill in assessing future rainfall probabilities for the season ahead.

38.The wheat crop simulations for various locations in eastern Australia pointed to a consistent median yield reduction in years of negative SOI index during May. However, provided there were adequate soil moisture reserves available at the time of planting, the simulations pointed to high chances of economically viable yields even during El-Niño years. Through a detailed on-farm monitoring programme it was established that wheat grown on good soil moisture reserves in 1997 across northeastern Australia performed well although the seasonal rainfall was substantially below average. The crops were assisted by timely rain at flowering.

39. The information from rainfall analysis, crop simulation and categorization for seasonal outlooks has been used to examine tactical production decisions relating to wheat, sorghum and chickpeas, and on decisions about potential relative benefits from planting wheat or chickpeas in particular years depending on the prevailing climate outlook. Chickpeas have a shorter growing season and a later planting date so that they are perceived as a potentially less risky option that wheat when rainfall is scarce.

2.4.2 System for Analysis, Research & Training (START) in Global Change Climate Prediction for Agriculture Programme

40. In September 1997, START established a task group to develop a draft strategic plan for the development and implementation of the programme on Climate Prediction and Agriculture (CLIMAG). The CLIMAG strategic plan provides a conceptual framework for the utilisation of climate prediction in agriculture. The broad strategy for conducting the project in a specific region would be

1. To determine the baseline relationship between climate variability and crop production in the region;
2. To establish awareness in the region of the potential for climate predictions to be used to increase crop yield;
3. To mobilise a multi-disciplinary team to design and execute the project in the region;
4. To identify agriculture practices in the region that may be modified through knowledge of future climate variations;
5. To design a project in which the impact of changes in agriculture practice can be quantified;
6. To conduct a trial with farmers over a number of years where climate information is used to modify agriculture practice as required; and
7. To analyse and disseminate the results of the trial.

41. The preliminary results of the CLIMAG obtained from two sites in India are given below:

42. CLIMAG, Tamil Nadu, India: Agricultural production in the Indian state of Tamil Nadu experiences problems due to erratic monsoon seasons, crop failures and improper resource management. Average annual rainfall is 640 mm with most rainfall received during two monsoon seasons, namely the southwest monsoon (172 mm; June to September) and northeast monsoon (321 mm; October to December) in the western part of Tamil Nadu State. For this part of India, the use of SOI phases has shown considerable skill. There is a significant relationship between seasonal rainfall and negative phases of SOI, with a stronger signal during the northeast monsoon. In both seasons variability is less in negative SOI phases with chances of getting at least median rainfall considerably higher than in other years. The average dry land crop production in this region is about 0.6 t/ha. The key management decisions are selection of crop, dates of sowing, land management, fertiliser rates, intercultural operations and harvesting. Seasonal climate forecast could help to increase production by making better-informed management decisions. For instance, preliminary simulation results indicated that if the median August-November rainfall probability is above 400 mm and the soil profile contains at least 50 per cent of stored soil moisture on August 15th, it is advisable to plant cotton. If the condition is not satisfied it would be better to consider planting sorghum between September 15th and October 15th, provided the median September-December rainfall probability exceeds 300 mm and stored soil moisture is above 30 per cent. Otherwise millets, sunflower and chickpea could be considered.

43. CLIMAG, Andhra Pradesh, India: Groundnut is the most important oilseed crop in India with a total production of about 8.2 million hectares of which over 80 per cent is rained. Anantapur is at the heart of the groundnut region over the Indian peninsula, in the state of Andhra Pradesh, which accounts for about one third of the groundnut production of the country. The district-average yield (the district is about 10,000 sq kms and the area under groundnut covers about 8,000 sq kms) varies considerably from year to year. The yield varies from less than 500 kg/ha to 1.5 tonnes (which is considered as a failure of the crop). It has been observed that the variation in yield arises to a large extent from the variation in the total rainfall during the growing season (Gadgil S., 2000).

44. The seasonal rainfall up to 50 cm is required to sustain a successful groundnut crop. When the seasonal rainfall is below this value, there are several years with yield below 700 kg ha-1 i.e. very low yields on the farmers' field. On the other hand when seasonal rainfall is above 50 cm, the probability of yield above 1-5 t ha-1 is over 50 per cent whereas that below 700 kg ha-1 is zero. This suggests that a prediction of whether seasonal rainfall will be below 50 cm will be useful.

45. It has been observed that the seasonal rainfall is less than 50 cm in 21 El Niño years out of 24. It is rather high only with the exceptional El Niño of 1953 during which the all India monsoon rainfall was also in excess. Thus in 87 per cent of the El Niño years, the seasonal rainfall is below the threshold of 50 cm, which implies a high probability of low levels of yield. Hence if an El Niño is predicted, it would be worthwhile to minimize on investment such as chemical fertilizers, etc. or cultivate some other crop such as horse gram. Thus a prediction for El Niño has potential for application for farm-level decisions.

46. However, it should be noted that of 58 years that are characterized by seasonal rainfall less than 50 cm only 21 coincide with El Niño. Hence while prediction of an El Niño will be of considerable use, other predictors for prediction of such years have to be explored.

47. These climate crop models are being tested in a homogeneous environment. These models provide information about the potential value of utilising climate forecast information for managing agriculture systems.

Section - 3
The experiences of application of ENSO index based long range forecast

48. In the previous section the potential value of climate forecast information has been highlighted. The study of the actual experiences of application of forecast information during 1997-98 (El Niño) in Australia, Indonesia, the Philippines and Vietnam are discussed in the succeeding paragraphs.

3.1 Australia

49. The bureau of meteorology has been preparing season climate outlooks each month for the past decade. The outlook consists of a comprehensive booklet (including tabulations of probabilistic prediction of rainfall in terciles: dry, near normal, wet). It is prepared at the start of the three month period, after a meeting involving meteorologists, oceanographers and some representatives of the agriculture sector (the main user sector for this forecast). From May 1997 the Bureau had been including indication of a likely El Niño event and hence an increase probability of low rainfall over eastern Australia. The outlook issued in early August indicates "El Niño persists: Dry weather likely to continue over south-eastern Australia. The summary went on to say that "there is a strong likelihood of significantly drier than normal conditions persisting and expanding across much of eastern and southern Australia. The tables included in the August Outlook indicated that rainfall in the dry tercile was, typically, two to three times more likely than the wet tercile. In the event, although there were areas where the August-October period was dry, there were also considerable areas with rainfall much above the average (and well into the "wet" tercile). Moreover, rainfall was good through much of the region in September, a critical time for crops.

50. A huge communication gap was noted. Part of the problem is attributable to different emphases placed by forecasters and users on certain critical words. It appears that users and forecasters interpret "likely" in different ways. Those involved in preparing the forecasts and media releases intended this to indicate that dry conditions were more probable than wet conditions, but that there was still some chance that wet conditions would occur. Many users, it appears, interpreted "likely" as "almost certainly dry, and even if it wasn't dry then it would certainly not be wet".

51. Lack of knowledge about interpreting and using ENSO forecasts caused many farmers to have inappropriate expectations of El Niño and took inappropriate actions to cope with it. As a result, the forecasts in some cases probably did more harm than good. It was reported that some farmers over reacted and sold large portion of their herds and not planting a crop etc. (Nichols N., 1998).

3.2 Indonesia

52. Prior to 1997, the Bureau of Meteorology and Geophysics (BMG) generally used to issue weather forecasts keeping in view meteorological parameters. From 1997 onwards, the BMG has taken the initiative to establish a broad based National Seasonal Forecasting Working Group drawing upon expertise from various sectors. This Working Group comprises BMG, Bureau of Assessment and Application of Technology (BPPT), the National Space Center (LAPAN), Agriculture Research Institute and Water Resources Management Research Institute.

53. The Working Group draws upon forecast information from ASEAN Specialized Meteorological Centre (ASMC), IRI, BOM Australia and UK Metro Office to prepare seasonal forecast guidance that includes the following:

1. Seasonal monsoon onset forecast indicating the dates of onset of monsoon with ten days intervals for 102 meteorological regions across the entire country.
2. Monthly forecast of rainfall for 102 meteorological regions for the country.
3. Seasonal cumulative rainfall status for the entire season for 102 meteorological regions.

54. Respective climate sensitive organizations at the national level, on receipt of climate forecast information from BMG, process the outlook with reference to past impacts and disseminate processed information to provincial sectoral organizations. At present, these forecasts are used as general alert. The information is received from the field agencies to the national level user agencies only when disaster events occur. The processed forecast information received at the national level is useful for taking general precautionary measures but cannot be used for comprehensive development planning.

55. During 1997-98, the Department of Agriculture, on receipt of the information from BMG, on the likely impact of El Niño, processed the information and disseminated to provincial agencies in a routine way. The details are given below:

56. It may be seen that it took about 6 weeks for the forecast information to reach the farmers with some general recommendations. The farmers did not benefit from the forecast information as they planted on schedule on receipt of the first rains in September/October. The rains ceased thereafter. The farmers had to replant again and again. As a result around 900,000 hectare of crop area was reported affected by drought in varying degrees.

57. As no concerted efforts were taken to take lead-time provided ENSO forecast, a loss of around 3 million tonnes of rice was reported. The Government had to import 5 million tonnes of food grains to ensure food security in the country.

3.3 Philippines

58. PAGASA, the Philippine meteorological agency incorporates ENSO indices as one of the major parameters in its long range forecast scheme. The PAGASA provided early warning about 1997-98 El Niño in 1996 itself and the first drought advisory was issued in May 1997. The advisories indicate the broad weather outlook. The following are the extract of the seasonal climate forecast issued by the PAGASA in May 1997.

59. "Based on these recent evolution and forecast of the atmospheric and oceanic conditions, it can be expected that warm episode will intensify during the next several months. This climate forecast on an impending warm episode will have global scala implications and for Philippines some climate tendencies during the seasons are indicated below.

60. Southwest Monsoon Season (May 1997 to September 1997): In view of this new development, the onset of the rainy season (which normally occurs during the second half of May) is expected to be delayed by about two weeks. With this, the duration of rainy season, which normally ends during the early half of October may be short, ended, although some bursts of heavy rainfall during the rainy season could also be expected mostly in the western section of Luzon and some parts of western Visayas.

61. Northeast Monsoon Season (October 1997 to March 1998): The impending warm episode in the central and eastern equatorial Pacific will have influence on the activity of tropical cyclones in the PAR. Below normal tropical cyclone activity will most likely occur during the coming northeast monsoon months. This will cause below normal rainfall condition in a bigger portion of the country".

62. The relevant extract of drought advisory issued by PAGASA before the commencement of the Northeast monsoon 1997-98 is given below:

63. "Based on trends, climatological studies and the present atmospheric and oceanographic situation in the central and eastern equatorial Pacific, manifestations of the effects of the existing El Niño phenomenon on the Philippines climate will have its peak during the northeast monsoon season (October to March). Atmospheric sea level pressure in the eastern equatorial Pacific including the Philippines will be above the normal while sea surface temperature will be below the normal. Consequently, below normal tropical cyclone activity is expected in the PAR. With these factors, drier than normal weather conditions can be experienced in the Philippines starting October 1997 and continuing through March 1998".

64. The expressions such as "drier than normal weather conditions" and "bigger portion" of the Philippines would experience moderate to severe drought has been interpreted as that the whole of Philippines would be affected by devastating drought by the user departments. An El Niño task force was constituted at the national/provincial and other lower levels through out the Philippines. Resources were distributed to all the regions of the country.

65. The drought impact was confined to certain areas only. The farmers did not get the location specific advisories to change the crops etc. As such, around 600,000 hectares of corn and rice lands were affected by drought. Around 1.2 million tonnes of food grain production was reported.

66. Although establishment of a comprehensive climate forecasting applications system is well under way in the Philippines, there is still a great need to develop capabilities to process forecast information into more actionable formats at the local level. The information provided by the national agencies falls short of meeting the specific needs of users at the local level.

3.4 Vietnam

67. The Hydro Meteorological Services (HMS) uses antecedent parameters such as Eurasian snow cover, ITCZ etc. for making seasonal forecasts. In recent months after the initiation of Extreme Climate Events Program, HMS has begun to start incorporating ENSO into long-range forecast information into seasonal forecasts.

68. The seasonal forecast information provided by HMS is used by climate sensitive sector agencies like agriculture, water resources, Disaster Management Center only as a general alert. A lot of progress needs to be made to make full use of climate forecast information for development planning.

3.5 Fiji

69. The primary agency for meteorological monitoring, forecasting, and research in Fiji is the Fiji Meteorological Service (FMS). This organization has strong professional linkages with equivalent groups in New Zealand (NZ), Australia, and the Pacific ENSO Application Center in Hawaii. Forecasting is made using (1) general monthly rainfall pattern analysis, (2) analysis of past ENSO warm and cold events, (3) rain forecast models (both locally - and foreign developed), and (4) regional prediction models from foreign agencies (Kaloumaria A., 2000).

70. As no local specific action to minimise sugarcane losses which is the primary agriculture industry in Fiji has been undertaken during 1997-98, 50 per cent sugarcane production loss was reported affecting 200,000 farmers.

Section - 4
Emerging issues

71. The discussion in Section-2 indicates that skill in climate prediction offers considerable opportunities to managers to reap benefits (i.e. increased food production and profit and/or reduced risks). Our discussion in Section-3, however, reveals that realizing these opportunities are not straightforward as the forecasting skill is imperfect and approaches to applying the existing skill to management issues have not been developed and tested extensively. While much has been written about impacts of climate variability, there has been relatively little done in relation to applying knowledge of inherently imprecise climate predictions to modify actions ahead of likely impacts, i.e. applications of climate prediction. An effective application of a seasonal climate forecast is defined as use of forecast information leading to a change in a decision that generates improved outcomes in the system of interest.

72. A considerable body of effort in various parts of the world is now being focused on this issue of applying climate predictions to improve agricultural systems. This section will be devoted to discuss emerging issues for applying long range forecast information in the Asia-pacific region drawing lessons from the experiences of applying forecast information during 1997-98 El Niño and 1998-99 La Nina. These are:

1. Delimitation ENSO sensitive sectors, seasons and regions.
2. Utilisation of climate forecast with coarse resolution.
3. Articulation of users' needs.
4. Integration of intra-seasonal oscillations.
5. A system's approach for climate forecast and application system.

4.1 Delimitation ENSO sensitive sectors, seasons and regions

73. The climate exhibits only limited predictability and skilful forecast are available for some seasons and regions. While in some areas there are clear relationships between ENSO indices and local climate variables, other areas do not exhibit a linear relationship. It would take some time to obtain climate forecast with greater geographic resolution covering all factors governing the climate variability.

74. It is, therefore, necessary to delimit specific climate sensitive zones, which are relatively highly sensitive to ENSO indices, and specific relationship exists between ENSO indices and local climate variability when compared to other areas. After spatial delimitation of the geographic zones, a temporal delimitation of comparatively more ENSO sensitive season than other time periods need to be undertaken. For instance, summer season is more sensitive to ENSO than the dry season, which is by and large protected by assured irrigation systems. De-limitation of climate sensitive zones, sectors and seasons would facilitate application of forecast information at the local level.

75. The documentation of the past ENSO events could provide insight for mapping the specific areas likely to be affected by future El Niños. The programme on extreme climate events implemented by the Asian Disaster Preparedness Center (ADPC) implemented in Indonesia, the Philippines and Vietnam, is primarily to document the impacts of past extreme climate events and map out ENSO sensitive regions in these countries. This approach assisted the countries to prepare general vulnerability maps to manage future ENSO events.

4.2 Utilisation of climate forecast with coarse resolution

76. A commonly recognized problem in the application of climate forecasts in the agricultural community is that currently available seasonal forecasts are at too large a scale to be useful for site level planning. Both spatial and temporal scales need to be refined for agricultural management applications. There is a need to downscale the global ENSO index based forecast into local level climate outlook products. These climate outlook products need to be further downscaled, keeping in view the specific vulnerabilities at the local level with reference to different seasons and different cropping systems. The intermediary research organizations in the countries could be assisted to translate the very coarse ENSO forecast information into actionable format for specific uses at the end user level.

4.3 Articulation of users' needs

77. Most research has been driven from the climate and agro-ecological communities, and has tended to involve a top-down approach, where uses are sought for existing forecast information, and less commonly by a bottom-up approach, where a decision situation is examined to identify niches and needs for climate forecasts. Also, most current forecast products lack the spatial, temporal and element specificity that users seek for their specific decision-making needs.

78. Users are diverse and cannot be lumped into a homogeneous set. Even within the agricultural sector, the needs of agribusinesses such as seed suppliers and grain traders vary distinctly from primary producers. Among producers, limitations in access to resources or risk exposure condition responses to new information. One characteristic of users is self-sufficiency or survival (subsistence producers) or profit maximization (commercial). This plays a critical role in what information they would be willing to use.

79. Even if locally actionable climate forecast is available, the communication of probabilistic forecast information could be problematic to the end users. There is a need to undertake a communication package keeping in view socio-cultural peculiarities of communities for whom the forecast are intended to benefit.

80. Even if locally actionable climate forecast is available and communication is also perfect, there could be following constraints for the farmers to respond to the forecast:

• Lack of credit access or previous debt burden;
• Competing demands for labour;
• Limited access to ploughs, seed of suitable varieties, and other inputs;
• Limited land access;
• Untimely dissemination of forecasts: farmers need forecasts by the end of April at the latest;
• Priorities and strategies for risk aversion and risk management;
• Market access or stability of prices and demand for cash crops;
• Decision irreversibility;
• Inappropriate forecast information;
• Lack of confidence in forecast or in the source/provider of the information;
• Diversity and level of income, both on-and off-farm; and
• Local consumption preferences for crop varieties.

81. There is a need to evolve policies and programmes at the National level to address these constraints to enable the farmers to use climate forecast information.

4.4 Integration of intra-seasonal oscillations

82. The long range forecast would provide an indication of the behaviour of rainfall during the course of the season. However, there could be a meso-scale intra-seasonal oscillations which may result in long dry spells/wet spells/cyclones and storms. The farmers will encounter these disturbances in the course of the cropping season. During the course of crop growing season, certain midterm corrections will be required to minimise crop yield losses. Hence, short term forecast between 5 to 10 days will provide critical information for undertaking corrective measures.

83. A long range forecast system coupled with monitoring of seasons, weather behaviour would be necessary to taking into account intra-seasonal variabilities.

4.5 A system's approach for climate forecast and application system

84. The application of climate prediction information requires a close consideration of the roles of and interactions among the full span of climatic, ecological and social factors involved. This includes climate observation systems, the choice of climate prediction tools, the design of climate forecast products to suit users need, communication of the forecast products, sector system models (crop climate models) the decision behaviour, institutional constraints and social settings in which the decisions are made. A continuous dialogue mechanism between climate information producers, intermediary research organizations, and policy makers and end users need to be institutionalised.

85. The ADPC promoted establishment of climate information and user networks in Indonesia, the Philippines and Vietnam. These institutional arrangements would view climate variability as a continuous phenomenon and meet before the onset of wet and dry seasons to utilise the forecast for mapping out potential impacts and undertake measures to manage the likely impacts. This co-learning process would facilitate better appreciation of climate variability, limitations of forecast skills and constraints and opportunities for utilisation of climate forecasts.

REFERENCES

ADPC, 1997, "El-Nino 1997-98: Strategies for Risk Reduction in Water Resources and Agricultural Sectors in the Philippines", http://www.adpc.ait.ac.th

ADPC, 1998, "La Nina 1998-99: Challenges and Opportunities for Indonesia", http://www.adpc.ait.ac.th

Basher R., 2000, "Seaonal Climate Prediction and Managing for variability", IHDP Update, Newsletter of the IHDP, No. 2, 2000 pp. 8-9.

Banerjee, A.K., Sen, P.N. and Raman, C.R.V. (1978) On foreshadowing southwest monsoon rainfall over India with midtropospheric circulation anomaly of April. Indian F. Meterol. Hydrol, Geo-phys., 29, pp. 425-431.

Blanford, H.H. (1884) On the connection of Himalayan snowfall and seasons of drought in India. Proc. R. Soc. London, 37, pp. 3-22.

Braak, C. 1919, `Atmospheric variations of short and long duration in the Malaly Archipelago and neighboring regions, and the possibility to forecast them', Verhandelingen No. 5, Koninklijk Magnetisch en Meteorologisch Observatorium te Batavia.

CLIMAG, 1999 (December), Climate Prediction and Agriculture, START Network News, Issue No. 5, pp. 4-8.

Climate Prediction and Agriculture, http://www.start.org/Climag/ImplemDem.html

Cane, M.A. Zebiak, S.E., Dolan, S.C., 1986, Experimental forecasts of El-Nino, Nature 321, 827-832.

Carson, D.J., 1998 seasonal Forecasting. Q.J.R. Meteorol. Soc. 124, 1-26.

Conard, V. 1941. "The variability of precipitation'. Mon. Weath. Rev., 69:5-11.

Eakin H., 2000., "Seasonal Climate Forecasting and the Relevance of local Knowledge. Physical Geography forthcoming.

FAO, 1998, "Special Report FAO/WFP Crop and Food Supply Assessment Mission to Bangladesh, Global Information and Early Warning System on Food and Agriculture, World Food Programme.

Fox James J., 2000 "The impact of the 1997-98 El-Nino on Indonesia, in El-Nino - History and Crisis, edited by Richard H. Grove and Hohn Chappell, pp. 171-190.

Gadgil S., "Farming Strategies for a Variable Climate - An Indian Case Study" in Proceedings of the Interntional Forum on Climate Prediction, Agriculture and Development, April 26-28, 2000, International Research Institute for Climate Prediction, Palisades New York, USA pp. 27-37.

Govt. of the Philippines, 1998 "A united initiative: Task Force El-Nino", The 1997-98 Annual Report on Task Force El-Nino Phenomenon.

Govt. of Indonesia, 2000, "Impact of Extreme Climate Events and Food Crop Management in Indonesia".

Hammer, G.L., 2000, A gernal approach to applying seasonal climate forecasts, in G.L. Hammer, N. Nicholls, and C. Mitchell (eds.) Application of seasonal climate forecasting in Agriculture and natural systems - The Australian experience Kluer Academic, The Netherlands.

Hastenrath, S. (1987) Predictability of Java monsoon rainfall anomalies: a case study. Journal of Climate and Applied Meteorology, 26, 133-141.

Hastenrath S. 1987, `Predictability of Java monsoon anomalies: a case study', J. Clim. Appl. Meteor., 26, 133-141.

Jose A.M., 1989 "Prediction of seasonal rainfall in the Philippines", Ph.D. thesis, College of Science, University of the Philippine, Diliman, Quezon City.

Kaloumaira A., 2001 "Fiji Country Case Study Impacts and Responses to the 1997-98 El-Nino Event" in Reducting the Impact of Environmental Emergencies through Early Warning and Preparedness: The case of the 1997-98 El-Nino, A UNEP/NCAR/UNU/WMO/ISDR Assessment, January, 2001, pp. 91-102.

Kirono, D., Tapper, N. and McBride, J. 1998, `Documenting Indonesian rainfall in the 1997-98 El-Nino Event', in preparation for submission to The International Journal of Climatology.

Krishna Kumar, K. (1994) Forecasting of Indian summer monsoon rainfall on smaller regional scales using canonical correlation analysis technique. Project Report, International Research Institute for Climate Prediction Pilot Project, Lamont-Doherty Earth Observatory, New York, 28 pp.

Kirono, D. 1998 (personal communication) Ph.D. student at Monash University, Australia, working on Indonesian rainfall variability.

McBride, J.L. and Nicholls, N. (1983) seasonal relationships between Australian rainfall and the Southern oscillation. Monthly Weather Review, 111, 1998-2004.

Nicholls, N. 1981, `Air-sea interaction and the possibility of long-range weather prediction in the Indonesian Arcipelago', Mon. Wea. Rev., 109, 2435-2443.

Nicholls, N., 1983 `The Southern Oscillation and Indonesian Sea surface temperatures', Mon. Wea. Rev., 112, 424-432.

Nicholls, N. and Woodcock, F. 1981. `Verification of an empirical long-range weather forecasting techique'. Quart. J. Roy. Met. soc. 107:973-976.

Nicholls, N. 1981. `Air-sea interaction and the possibility of long-range weather prediction in the Indonesian Archipelago' Mon. Weath. Rev. 109:2435-2443.

Nicholls, N. 1984. `The Southern Oscillation and Indonesian sea-surface-temperature'. Mon. weath. Rev., 112:424-432.

Nicholls, N. 1988a. `El-Nino - Southern Oscillation and rainfall variability'. J. Climate, 1:418-421.

Nicholls, N., 1985, Impact of the Southern Oscillation Australian Crops., J. Climatol. 5, 553-560.

Nicholls, N. and Wong, K., 1990. `Dependence of rainfall variability on mean rainfall, latitude, and the Southern Oscillation'. J. Climate, 3:163-170.

Nicholls, N., and Kestin T., 1999, http://www.esig.ucar.edu/lanina/report/kestin.html

Nicholls, N., 2000(a) "What the instrumental and recent historical record tells us about the El-Nino - Southern Oscillation" in El-Nino - History and Crisis , edited by Richard H. Grove and Hohn Chappell, pp. 79-88.

NRC, 1996. Learning to predict with El-Nino and Southern Oscillation. National Research Council (NRC), National Academy Press, Washington DC.

Ogallo, L.A., 1988. Relationship between seasonal rainfall over East Africa and the Southern Oscillation. J. Climatol. 8, 31-43.

PAGASA, 1997, "Update on seasonal climate forecast in the Philippines for 1997", dt. 15 May, 1997.

PAGASA, 1997, "Drought Advisory No. 9805-A" dt. September 1997.

Palmer, T.N., Anderson, D.L.T., 1993. Scientific assessment of the prospects for seasonal forecasting a European perspective PRESAO-1, 1998. First Forum on seasonal forecasting and its applications to early warning systems for food security in West Africa, Abidjan, Cote d'Ivoire, 4-8 May 1998, ACMAD.

Ramusson, E.M. Wallace. J.M., 1983, Meteorological asects of the El-Nino/southern oscillation. Science 222, 1195-1202.

Ropelewski, C.F., Halpert, M.S, 1989. Precipitation patterns associated with high index of the Southern Oscillation, J. Clim. 2, 268-284.

Stone R.C., and de Hoedt., G. 2000 "The development and delivery of current seasonalclimate forecasting capabilities in Australia", in G.L. Hammer, N. Nicholls, and C. Mitchell (eds.) Application of seasonal climate forecasting in Agriculture and natural systems - The Australian experience Kluer Academic, The Netherlands.

Sukardi W., 1998, "The meterological Meaning of Traditional Time Reckoning "Wariga" and its applications", Jurnal Agromet, Vol. XIII No.1, Indonesian Association of Agricultural Meteorology pp. 15 to 24.

Thapliyal, V (1987) Prediction of Indian monsoon variability evaluation and prospects including development of a new model. In: Ye, D., Fu, C., Chano, J. and Yoshino, M. (Eds.) Climate of China and global Climate, China Ocean Press, pp. 397-416.

Walker, G.T. (1908) Correlation in seasonal variation of climate (Introduction). Mem. India Meterol, Dept. (IMD Mem.), 20 pp.117-124.

^* ADPC is a technical resource center dedicated to disaster reduction for safer communities and sustainable development in Asia and the Pacific.