This study examined the adoption of newly released rice varieties and variety groups in the farmers' fields and measured the replacement rate of newly released varieties using data on varietal spread from the Rice Research and Development Institute (RRDI) database. The data showed that Bg 94-1, Bg 300 and Bg 352 are enduring varieties from the 1970s, 1980s and 1990s, respectively and Bg 358, Bg 359 and At 362 are widely planted new varieties. The new improved varieties released from the mid-1960s were popularly planted with an observed decrease of the use of old improved varieties and traditional varieties. The replacement period of new rice varieties has declined over time and the current replacement period is around 16 years with lower adoption rates of newly released rice varieties. While policymakers are encouraged to continually support and strengthen the current efforts of rice breeding research and extension, the policy of variety release and promotion needs to be strengthened, focusing the emphasis on accessibility of seed and information on new varieties to farmers.
Development of rice varieties is crucial for sustaining food security in Sri Lanka. Indeed, notable progress in rice production has been achieved over the last four decades through development and adoption of new improved varieties. The rice research system has been continuing the release of new rice varieties. By 2010, the Department of Agriculture (DOA) had released 67 rice varieties for different agroecological conditions. Evenson and Gollin (2003) viewed that varietal releases are not necessarily good measures of the success of varietal development programmes: a better measure is the use of these varieties in farmers' fields. According to the paddy statistics (2009/10 Maha) published by the Department of Census and Statistics, about 98 per cent of the country's rice area is planted to new improved varieties released by the DOA. However, these statistics have grouped all rice varieties released since 1968 into three categories and adoption trends of specific rice varieties cannot be distinguished.
A more detailed understanding of the development and adoption trends by specific rice varieties in farmers' fields is important because it relates to the impact of continued public plant breeding research programmes. A recent study by Jayawardena et al. (2010) assessed the adoption by specific rice varieties and revealed that a high percentage of total rice area is still occupied by varieties that were released more than 15 to 20 years ago. However, this study used the data only at particular point of time and change of adoption over time with release of new varieties has not been studied. Further, at the Provincial Working Group Meetings and Coordinated Rice Varietal Testing (CRVT) meeting of the DOA, it is often discussed that replacement of old varieties with new varieties is very slow and it can be attributed to a weak research-extension linkage, an effective public sector extension system and the absence of seeds of new varieties in adequate quantities. However, to what extent farmers have been changing from the 'old' varieties to the 'new' ones, and the reasons for the low level of replacement of old varieties with new varieties are issues that have not yet been rigorously studied.
In this context, it is useful to have an updated measure of the adoption trends of varieties and to what extent farmers use these varieties. Adoption rates of new varieties indicate the impact of the breeding programme, which continues to develop and release new varieties, on rice production in the country. Brennan and Byerlee (1991) indicated that "for a given rate of variety release, a rapid rate of variety replacement in farmers' fields leads to higher returns to public plant breeding research because the lag between variety release and adoption by farmers is reduced". This paper aims to analyse the adoption trends of released rice varieties and rate of variety replacement in farmers' fields using the aggregate time series data set of the RRDI database.
The study used data from the RRDI database, which is the only time series data set available on rice varieties cultivated in Sri Lanka. The Department of Census and Statistics (DCS) publish data seasonally on rice cultivation but it is classified only to areas harvested to old improved varieties, new improved varieties and traditional varieties. One limitation of RRDI data set is that the 'other variety' category, which stands for varieties that are not identified by name, represents a substantial percentage (10-14 per cent). This is possible when the farmers are not able to correctly identify the variety by name or recall the varieties they planted.
Conceptually, a promising line undergoes a series of field performance tests and evaluation trials before it is released as a recommended variety. A promising line coming from preliminary and major yield trials conducted in the breeding station is subjected to National Coordinated Rice Varietal Testing (NCRVT), Varietal Adoptability Testing (VAT) and Large Scale Varietal Adoptability Testing (LSVAT) and the standard check varieties usually used are the best available existing variety. Therefore, newly released varieties are expected to perform similar to, if not better than, the existing varieties in the given environment for which it is recommended. When a farmer chooses to adopt a new variety in place of an older variety, it reflects the farmer's judgment that the new variety offers some net benefit or advantage (Evenson and Gollin, 2003).
With this in mind, simple indices were calculated to better understand the dynamics of rice variety adoption, especially of the newly released varieties. Indices such as the proportion of recent varieties and weighted average age of varieties were calculated to determine the adoption rate of the newly released varieties following Brennan (1984) and Launio et al. (2008). The proportion of recent varieties is the proportion of the total area that is planted to recently released varieties.
This index, qit, is computed as follows:
qit = pit if year of release >t - m,
qit = 0 if year of release <t - m,
where pit is the proportion of the area sown to variety, i, in year, t;m is the number of years used to define 'recent'. Then
It = Σqit
where I is the proportion of the total area that is sown to varieties released in the previous m years. In this study, it was assumed that the lag period between the release of a variety and its availability to farmers is 2 years and recent variety is defined as a variety that is available to farmers for 8 or more years. The number of years used in this study to define 'recent' is 10 years (m=10).
The weighted average age of varieties (following Brennan and Byerlee, 1991), on the other hand, consists of the weighted average age of varieties grown by farmers in a given year, measured in years from varietal release and weighted by the proportion of area sown to each variety at that time. This index, WAt, is computed for a given year, t, as follows:
Wat = S pit> Rit
where Rit is the age of the variety in terms of the number of years (at time t) since the release of variety i. This measure avoids the use of an arbitrary definition of 'new' or 'recent' varieties (Brennan and Byerlee, 1991). These indices (I and WA) were calculated at four points of time taking years; 1980, 1990, 2000 and 2009 as t. For year 2009, m =9. The oldest and newest varieties planted were also identified in the four points of time and provided an additional indicator of how old the varieties are that are observed in farmers' fields.
Results and discussion
Varieties released by the Department of Agriculture
The number of varieties released in the 2000 decade was 12, compared to 18 in the 1990s,18 in the 1980s and 13 in the 1970s. Of 67 rice varieties released, 5 varieties fall into the 5-6 month age group, 23 into the 4-4.5 month age group, 23 into the 3.5-month age group, 14 into the 3-month age group and 2 into the 2.5-month age group. Varieties have been released by the DOA under two categories. The first category is for general cultivation. The second category is for specific regions/problem soils/quality attributes. Out of 67 varieties, 32 varieties are for general cultivation and 35 varieties are for specific regions/problem soils/quality attributes.
The potential yield of Bg 90-2 released in 1975 has not yet been surpassed by any new variety of that age group, although 5 varieties of the 4-4.5 month age group have been released for general cultivation since then. In addition, no new varieties of this age group have been released after 1993. For general cultivation, 13 varieties of the 3.5-month age group have been released and of these, 7 varieties have been released after 1993. Of these 7 varieties, 6 varieties surpassed potential yield of varieties released before 1993. Under 3-month age group, 10 varieties have been released for general cultivation. The yield of Bg 276-5 and Bg300, released in 1979 and 1987, respectively was surpassed only by Bg 304 and Bg 305, released in 1993 and 1999, respectively. It appears that breeders have given the priority to develop short age varieties in their breeding programmes, recognizing the importance of water saving and the demand of farmers for short age varieties.
Rice varieties cultivated by farmers
The percentage of area planted to rice varieties based on the major categories; old improved, new improved, and traditional are presented in Figure 1. It did not take long for farmers to adopt new improved varieties (NIV). After their introduction in the late-1960s, more than 50 per cent of the total rice area was planted to NIVs by 1978. This adoption rate steadily grew, and by 1984, around 95 per cent of the rice area was planted to NIVs. The 2009 data showed that almost 100 per cent of the total rice area is cultivated with NIVs.
The close look at the percentage areas planted by variety in different seasons at four points of time (1980, 1990, 2000 and 2009) revealed that the most popular varieties released in the 1970s were Bg 11-11, Bg 34-8 and Bg 94-1. Bg 94-1 and Bg 34-8 ranked first or second respectively until 1990. By 2000, Bg 300, a 3-month variety released in 1987, was widely cultivated along with 3.5-month Bg 352, released in 1992 and Bg 94-1, released in 1975. Bg 300 still ranked first while Bg 352, released in 1992, and Bg 358, released in 1999, were also widely used in both seasons in 2009. Some varieties have become popular to a certain level within a short period of time after their release and sometimes even before they are officially released. For example, Bg 300, released in 1987, has occupied 12 per cent of the total rice area by 1990 Yala season. At 362, officially released in 2002, was reported in farmers' fields in 2000 Yala. This is possible because seed of new varieties is given to farmers for research purposes (VAT and LSVAT) before release and through this seed new varieties are spread in farmers' fields.
Adoption of new rice varieties and replacement of old varieties
When the proportion of the total area planted to new rice varieties (varieties released in the previous 10 years from the point of time concerned) are examined in the 1979/80 Maha (wet season) and 1980 Yala (dry season) varieties released during the period from 1970 to 1979 covered around 77 to 78 per cent of the total rice area planted. In the 1999/00 Maha, new rice varieties (varieties released during the period from 1990 to 1999) occupied only 33 per cent of total rice area planted. In the 2000 Yala season it was only 26 per cent. This implies that rice varieties released before 1990 dominated in farmers' fields even in 2000. On the other hand, it indicates the low rate of adoption of newly released varieties.
The close look at the weighted average age of varieties as an index of adoption rates of new varieties indicates comparatively lower adoption of new rice varieties. The weighted average age of varieties planted has increased over time. This shows that the replacement period of rice varieties has increased over time despite the high rate of release of new varieties. In 1980, The rate was 5 to 6 years. In 2009, it was 16 years. This implies a low rate of adoption of new varieties or low rate of replacement of old varieties with new varieties. This confirms the previous observation that a higher percentage of rice areas are planted to old varieties released more than 10 years ago. The information on the old and new varieties cultivated indicates that some varieties such as Bg 94-1 (1975), Bg 379-2 (1980), and Bg 300 (1987) released more than 20 years ago are still cultivated by farmers.
Conclusions and policy implications
A progression in the use of NIV is evident from the decreasing trend in the use of traditional and old improved varieties and the increasing trend in the use of NIV. This information can be used to encourage policymakers to continually support and strengthen the current efforts of public rice breeding research and extension. However, data showed that varieties (new improved) released more than 20 years ago are still planted in farmers' fields and there is also evidence of the slow rate of adoption of newly released varieties.
On the adoption of new varieties, the study showed accounts of late adoption of new varieties. The replacement period of rice varieties in the country is increasing despite the high rate of release of new rice varieties. Does this reflect that in general, new varieties are inferior to the old ones? These findings should be considered in the policy of variety release and promotion, as it is possible that newly released superior varieties which may not have diffused yet will be crowded-out by more recent variety releases.
It is important that the seed network responsible for promoting new varieties be strengthened in order that all new varieties would have at least been tried by farmers and taken its potential advantage before they will be crowded-out by varieties that have tested superior in other locations. The successful performance of new varieties in village demonstration sites, proper information flow of the successes varieties, and the timely availability of certified and registered seed immediately after release must be ensured in order not to delay the diffusion of the variety. In addition, some varieties selected by farmers, such as Pokuru Samba, are adopted in farmers' fields even though it is not officially released. Thus, participatory breeding and consideration of farmers' selection as strategies for variety testing should continually be explored and pursued.
(References available upon request)