A Wikiblog E-Book by Norman Uphoff with many others
Chapter 7: EVALUATIONS OF SRI STARTING WITH FACTORIAL TRIALS
The first evaluation of SRI using standard scientific methodology was done in Madagascar in the 1997-98 season by Joeli Barison for his baccalaureate thesis from the University of Antananarivo, reported in Chapters 3, 4 and 6. Joeli’s thesis analyzed data from 171 trial plots on the fields of two farmers in the peripheral zone around Ranomafana National Park. These replicated trials, each 2 × 2 meters and laid out according to random block design, were managed according to established protocols. Joeli’s results confirmed the large increases in yield that were being observed and measured on rice farms around Ranomafana. However, Joeli’s study was more exploratory than conclusive, so more definitive evidence was needed.
THE NEED FOR FACTORIAL AND OTHER KINDS OF ANALYSIS
After a presentation on SRI in 1998 at ICRAF, the International Center for Research on Agroforestry in Nairobi, the center’s director for research, Peter Cooper, told me that for SRI to gain acceptance among agricultural scientists, we would need to do, or have somebody else do, a full set of factorial trials, evaluating the SRI practices separately and collectively.
Such a research design would evaluate concurrently the effects of all possible combinations of the practices that were being modified. Factorial results would enable us to speak with more confidence about the respective and combined impacts of the several different practices that constituted SRI.
We hoped that this work could be done by or at least with an established rice scientist so that the results would have more credibility. When I spoke in June 2000 with a rice researcher who had headed IRRI’s agronomy department, Kenneth Cassman, about cooperating with us on this task, he declined to get involved in such an evaluation. Such an undertaking appeared to him to be too large and complicated given the number of factors involved.
To evaluate the effects of six factors (practices), even if considering just two alternative values for each, and with three replications of each combination of factors, 192 separate test plots would need to be set up and managed. Such a large number was necessary to evaluate all of the possible combinations with statistical rigor. This was a daunting challenge that even an experienced crop scientist preferred to pass up.
Fortunately, Joeli’s mentor and our colleague at the University of Antananarivo, Prof. Robert Randriamiharisoa, understood how important factorial trials would be to establish the validity of SRI methods. Also fortunately, Robert was able to recruit the top students in two successive classes of agronomy graduates from his university to undertake this mammoth task for their respective baccalaureate-degree research projects under his personal supervision. Bright, energetic students who had never before taken on such a large assignment were not deterred by the magnitude of the effort required. So, in 2000 and the next year, Jean de Dieu Rajaonarison and Andry Andriankaja proceeded to carry out the desired factorial trials both thoroughly and capably.
After their results were completed and reported in 2002, systematic analyses followed in other countries, although none were as comprehensive as this first two sets of trials done in Madagascar. In this chapter, we review first the initial factorial analysis of SRI methods’ impact on rice yield in Madagascar. Then we consider subsequent evaluations of SRI done independently in various countries, Cambodia, Laos, Sri Lanka, India, Bangladesh, Nepal, China and Indonesia, by a variety of universities, research institutions, donor agencies, and private-sector organizations.
Although the studies that followed were not replications of the first study (we would like to have had such replications done), each added to our understanding of SRI in different ways. Some produced results from a large number of on-farm comparison trials, over 12,000 in Indonesia, which gave confidence in the results reported. Other evaluations shed light on issues such as the variability of SRI results under different soil conditions, the adaptation of SRI methods to rice cultivation under rainfed conditions, the impact of SRI on women, the extent to which SRI methods could reduce farmers’ costs of production, and how much more (or less) labor was required for SRI practice. Within a few years of publication of the results from the first factorial analyses in Madagascar, a number of other evaluations had confirmed the broad applicability of SRI’s initial insights.
FACTOR ANALYSES DONE IN MADAGASCAR, REPORTED IN 2002
The purpose of factor analysis is to determine what is the effect from any one individual factor when a number of factors (variables, in this case, agronomic practices) are altered concurrently. All possible combinations of the factors are assessed at the same time, to ascertain their respective impacts, all other things being equal. This also shows the collective effects of the whole set of factors taken together.
This sounds complicated, and it is. The task becomes very large if more than two or three factors are evaluated at the same time because each factor adds exponentially to the number of possible combinations, and all combinations of treatments need to be repeated identically several times as replications in order to minimize the effects of chance occurrences. Doing three or more repetitions of each different set of practices means that the results being compared are less likely to reflect random influences.
The research design that Prof. Robert worked out with Jean de Dieu and Andry was devised to shed as much light as possible on SRI effects with just two successive large-scale experiments. To make the results realistic, Robert arranged for the sets of trials to be done under field conditions rather than on experiment stations, conducting the trials under different, indeed contrasting agroecological conditions.
The first trials, done in 2000, were on the west coast of Madagascar, at Morondava 600 km from the university. They were conducted at the Centre de Baobab, an NGO farm demonstration site that was evaluating and popularizing agroecological practices. The farm was situated near sea level, and the climate there was tropical. The soils at the Centre were relatively poor and sandy, typical for the region.
The second set of trials was conducted the next year on the high plateau in central Madagascar, at Anjomakely, 17 km south of the capital Antananarivo. The elevation there was about 1200 meters above sea level. Its climate was temperate, and the soils were better than at Morondava. 
The trials were designed so that all of the trials at Morondava (N=288) were on the same soil, poor and sandy. Half of these trials were conducted with a modern high-yielding variety (number 2798), while the other half of the trials were planted with a local, ‘unimproved’ variety known as riz rouge (red rice).
At Anjomakely, all of the trials (N=240) were planted with the same variety, riz rouge, but half of these were carried out on fairly good (sandy loam) soil, while the other half were done on better (clay loam) soil. The plot size of all the trials was 2.5 × 2.5 meters.
Another four variables were evaluated in each set of trials: the age of seedlings transplanted (8 days for the SRI trials vs. 16 or 20 days for the conventional trials); the number of seedlings per hill (one vs. three); water management (controlled and limited irrigation vs. continuous flooding); and soil nutrient amendments (none vs. organic compost vs. NPK chemical fertilizer).
A sixth factor evaluated was spacing, comparing planting distances of 25 × 25 cm between hills vs. 30 × 30 cm. Each of these spacings was used on half of all the plots in both experiments. However, this could be seen as a flaw in the research design because both of these spacings were within an SRI range. Farmers’ practice was to plant their hills of seedlings only 10-15 cm apart: 50 to 60 hills per m² rather than 12 to 16, with 3 to 6 plants in each hill, with sometimes even more seedlings clumped together.
As might be expected, the yields from the trials with 25 vs. 30 cm spacing were not significantly different from each other in either set of trials. Indeed, there was no difference at all in the yield between the two spacings in the Morondava trials (each spacing had 144 replications). At Anjomakely, the difference between the two spacings (N = 120 trials for each) was only 80 kg per hectare (0.8%), which was nowhere near statistical significance. So with no real difference between the two sets of spacing trials, we could analyze the results in terms of five factors instead of six, and each of the treatments was evaluated with 6 replications rather than 3. This made the results much more reliable and statistically significant.
Unfortunately, the research design did not include evaluation of one additional factor, the SRI-recommended practice of soil-aerating mechanical weeding. To have evaluated one additional factor would have doubled the number of plots required, increasing the respective totals to 576 and 480, huge numbers of trials to conduct concurrently at one site. As it was, Jean de Dieu and Andry had to do an incredible amount of work to establish, manage, and gather data from their 288 and 240 plots, respectively.
In addition to measuring the yield from each of these 528 plots, Jean de Dieu and Andry also took representative samples of the rice plants in each plot, measured them, and calculated the average number of tillers per plant; the number of panicles (fertile tillers) on each plant; average lengths of panicles and roots (in cm); and root-pulling resistance, discussed in Chapter 4. This required an immense amount of effort from both Jean de Dieu and Andry. Surely no other theses done for university degrees in 2000 and 2001 required more work.
The results of these trials were reported in the two theses in terms of yield (kg per hectare) because presenting and explaining this output variable was already quite complicated with six factors involved. Reporting similarly on five additional parameters would have made communicating and understanding the results inordinately complicated. However, Prof. Robert and I went over all of the raw data from both sets of trials, i.e., 6 measurements per plot from 240 or 288 plots, the measurements themselves being averages for 10 plants chosen randomly from each plot, to inspect the numbers for any repetitive patterns or dubious outliers.
I looked particularly for any suspicious regularities that might indicate fabrication of the data, always to be guarded against in any large study. There was impressive internal consistency within the data sets, but not ‘too much’ consistency. Prof. Robert and I were both satisfied that the measurements had been made properly and carefully. Robert himself traveled fortnightly to the study locations during the two growing seasons to supervise the work personally, to maintain both morale and thoroughness. We both knew how important this kind of analysis would be for gaining a better understanding of SRI.
What appeared to be most significant was the similarity in the patterns of yield response to SRI practices in these two very different environments: one at sea level with tropical climate and poor soils; the other at 1200 m elevation, with temperate climate and better soils. The design of the trials enabled us and others to evaluate also any differences in response between a high-yielding variety and a local landrace at Morandava, and between having poorer vs. better soils at Anjomakely.
When the four SRI practices evaluated were used together -- young seedlings, one per hill, with no flooding of the fields, and with compost to enhance soil nutrition – the yield increases were substantial, ranging from 140% to 245% more than the results from plots on which more mature seedlings were transplanted, three per hill, with flooding, and with NPK fertilizer applied.
In both sets of trials, as seen below, the increments to yield increased most when all of the SRI practices were used together at the same time. Going from no SRI practices to one SRI practice, or from one to two SRI practices, or from two to three SRI practices added, on average, about 1 ton of yield per hectare. But going from 75% SRI to 100% SRI added almost 2 tonnes per hectare, an indication of synergy among the practices.
The impact of SRI practices on yield is summarized in the simple bar graphs below. The bars compare yield results from the two locations, differentiating between HYV and traditional-variety results and the use of SRI methods on better vs. poorer soils (clay and loam soils were at Anjomakely; sandy soils were at Morandava). A breakdown on the yields achieved from all the different combinations of practices is given then in the large summary table that follows.
The practices combined in the respective trials are: SS = saturated soil vs. AS = aerated soil; OS = older seedlings (16 or 20 days old) vs. YS = young seedlings (8 days old); 3 = three plants per hill vs. 1 = one plant per hill; NPK = chemical fertilizer vs. Comp = compost.
If the factor of soil-aerating weeding could also have been assessed in these trials, there is reason to think that there would have been even more enhancement of yield, based on the data from Madagascar, Nepal and Afghanistan reviewed in Chapter 11. However, we could not draw any quantified conclusion about the effect of this particular practice in conjunction with the other SRI practices because active soil aeration/ weeding was not evaluated.
The analysis of these two sets of factorial trials was completed in time for Prof. Robert and myself to present the results to international audiences twice in April 2002: first at the SRI conference held in Sanya, China, discussed in the next chapter, and the next week at a workshop on ‘water-wise rice production’ at Los Baños in the Philippines, organized by IRRI and Wageningen University. For the second forum, we combined the main results from the two experiments into a single table.
Multi-factorial trial results are unavoidably complicated. The numbers in the table above were calculated from the original data sheets compiled from the two sets of trials. The table shows the effects on yield of varietal and soil-quality differences, comparing SRI vs. non-SRI practices. The number of trials for calculating the average yield for each combination of practices is shown in parentheses.
Further analysis of these data showed that using young seedlings contributed the most to these differences in yield when all the other practices were equal. Having non-flooded, aerobic soil was the next most important factor influencing yield, other things being equal. Organic fertilization rather than chemical fertilizer was the third most influential factor, other things being equal; followed by transplanting single seedlings instead of clumps of plants. As expected, soil fertility had a bigger impact on yield than any particular practice.
These factorial trial results shed light on the respective advantages of organic versus inorganic soil fertilization. The data showed that when the yield response of a high-yielding variety to SRI methods is compared with that of an ‘unimproved’ variety on poor sandy soils, the HYV grown with just some SRI practices produced more with NPK fertilizer than with compost, as expected. HYVs are deliberately bred to be more responsive to inorganic fertilizer. On poorer soils, the traditional variety with different combinations and numbers of SRI practices gave somewhat lower yield, but the yields were roughly similar whether fertilization was organical or inorganical.
One important learning from these trials was that when all four SRI practices including compost were used with the high-yielding variety, its yield was 28% higher with compost than when the other SRI practices were used together with NPK fertilizer. Thus, compost could outyield or at least match inorganic fertilizer when combined with young seedlings, one per hill, and unflooded soil, the full set of SRI practices.
When the local landrace (riz rouge) was grown with all SRI practices, its yield with compost was 33% greater than with chemical fertilizer, a big difference. Even on soils with low inherent fertility, there was synergy among the SRI practices -- young seedlings, one per hill, and aerobic soil. Under SRI management, rice plants responded more vigorously to increased organic matter in the soil than when inorganically-sourced nutrients were provided to the plants directly. Feeding the soil was thus more important than feeding the plants, as suggested at the end of the preceding chapter.
When looking closely at the results of trials on the better soils of Anjomakely, we saw that the application of NPK fertilizer produced higher yield than with compost when only one or two of the SRI practices were used, especially when young seedlings were used in transplanting. This was true for both the better clay soils and the not-as-good loam soils (which were still better than the poor sandy soils of Morondava). However, when all of the SRI practices were used together with compost, the superiority of NPK fertilizer on these soils was no longer evident.
When three SRI practices – young seedlings, one per hill, and aerobic soil – were used with compost, there was an 18% yield advantage over NPK fertilizer on the better clay soil (10.35 t/ha compared to 8.77 t/ha), and on the loam soil, a 28% advantage with compost (6.39 t/ha vs. 5.00 t/ha). These results helped to shape our understanding of how the respective SRI practices boosted SRI productivity by supporting root systems’ growth and soil biological activity, discussed in Chapters 4 and 5. The results of trials reported in Chapter 6 underscored the importance of transplanting seedlings while still very young.
That there were six replications of all the treatments evaluated made these results more reliable. However, as discussed in Chapter 22, even when these findings were published in proceedings by the International Rice Research Institute and Wageningen University, they were ignored as if they had never been presented for the scientific community to consider.
The results were, admittedly, rather complicated and required some study to absorb fully their message and significance. By themselves, these data might have been considered only as matters of some scientific interest. However, by 2002 we were seeing the desirable impacts of SRI practices on ever more farmers’ fields beyond Madagascar, making this ambitious factorial analysis more than a matter for curiosity.
INDEPENDENT EVALUATIONS IN OTHER COUNTRIES
Perhaps one reason why the factorial analyses were not paid attention to or taken seriously was that the research was done by baccalaureate students in a poor developing country, even though Jean de Dieu and Andry were both outstandingly intelligent, hard-working, and well-trained. Maybe it was because their theses were written in French, although these were made available on the SRI website so anyone could evaluate their scientific rigor for themselves. More likely, the results were discounted and neglected because they clashed with certain presuppositions. This was hard to determine. Perhaps one problem was that the results were being reported by persons who, being impressed by what they were seeing and learning, could be discounted as ‘not objective.’
When we reported SRI results, this was often seen as advocating SRI, instead of what was intended: to encourage others to evaluate SRI ideas and methods for themselves. We insisted that we were not advocating the adoption of SRI, but rather its evaluation. If there was sufficient evidence for others to want to use SRI practices for themselves, this was fine and free since no patents or royalties or intellectual property rights were involved.
Some of the resistance to SRI could have been prompted by its unconventional, non-proprietorial approach to improving agricultural production. By the close of the 20th century, agricultural technology was becoming increasingly owned and sold commercially rather than being open-access, as it had been before. Since our main concern at the time was to understand better what was happening as a result of SRI crop management methods -- and why -- we did not devote much thought to understanding others’ lack of interest and enthusiasm.
Our own confidence in SRI methods was boosted by a succession of independent studies that assessed SRI methods in diverse countries, on farmers’ fields rather than on experiment stations. It was to be expected that results would vary from country to country. But the studies cumulated into a body of knowledge that impressed us and farmers more than it did most scientists.
There were, fortunately, some scientists who were curious about anomalous SRI results being reported. Their work and contributions are reviewed in Chapter 9. Here the results of studies done by independent evaluators in eight countries are summarized. All but one reported remarkably positive results that were made more comprehensible by what was learned from the 2000 and 2001 factorial trials.
The first really large independent evaluation was done here in 2004 by the German aid agency GTZ (now GIZ) which commissioned an evaluation led by an agronomist reported his findings to the agency and to a forum on international agricultural research for development. The data base was a survey of 400 SRI farmers and 100 non-SRI farmers randomly selected within five provinces.
The ‘SRI farmers’ were not necessarily using all of the recommended methods, and most were doing rainfed cultivation with no assured irrigation. Even so, the average yield with SRI methods was 41% higher than from farmers’ usual methods, 2.29 tonnes per hectare (using less fertilizer) compared with 1.63 tonnes. Most SRI farmers had started by using the new practices on just 20% of their land; but those farmers with more than one year of experience had doubled their share of land under SRI management, and 17% had gone to full SRI use.
Because SRI had already gotten a reputation in scientific and policy circles as being ‘too labor-intensive,’ as discussed at the end of this chapter, the most significant finding of the GTZ study was that SRI use in Cambodia was on average labor-neutral, not increasing labor requirements per hectare. This finding was from a large sample of farmers who gave detailed information on all of their labor inputs for production.
While learning the tasks, new SRI users needed more time than average for their operations the first year. But once gaining confidence and skill, their labor requirements were less than before. For the whole sample, there was on average no increase in labor inputs. Also, farmers reported that SRI reduced their need for labor at the time of peak labor demand, during the planting season. This was a big advantage from SRI use not reflected in survey data but revealed in focus groups. With SRI practices, family labor was usually sufficient, so there was less need to hire labor.
The study also found that SRI enabled farmers to reduce their costs of production by 40%, with seed costs cut by two-thirds and with a sharp reduction in the purchase of fertilizer and agrochemicals. This was particularly appreciated by Khmer farmers because most of their expenditures for rice farming were at the start of the planting season, when their cash on hand was at its lowest.
Net income from rice production per hectare was increased by 74%, from US$120 per hectare to US$209 per hectare. Net returns per day of labor were raised by 65%, from US$1.55 per day to US$2.54. Even when SRI required more labor from farmers, it clearly raised their labor productivity.
In terms of social impact, the percent of households in the sample who could not produce enough rice to meet their own subsistence needs was reduced from 34% to 28%, while the percent of SRI users moving into a position of surplus, so that they could sell some of their rice for income, increased from 20% to 33%.
A risk assessment showed that with SRI methods, the probability of these households not reaching a net income of $100 per hectare was 17%, while with conventional methods, this probability was 41%. Such data from a well-designed and large sample survey, provided very strong evidence of SRI merits. Below is the picture from the cover of the GTZ report.
In this neighboring country, comparative SRI evaluation trials were undertaken in 2002, but they were not conducted properly enough to be considered conclusive. IRRI’s country representative in Laos, Klaus Goeppert, when he learned about SRI was receptive to its ideas and wanted to try them out under local conditions.
After organizing a national workshop on SRI in April 2002, the IRRI program in Laos got several agricultural programs and NGOs there to do their own respective trials. Unfortunately, as the IRRI representative reported in an email to IRRI and Cornell, when the organizations met in 2003 to compare their results, there was agreement that most of the SRI trials had not applied the recommended practices correctly, and several had not established proper control plots that made meaningful comparison possible.
Karl wrote that these organizations and IRRI/Laos regarded the trials as “inconclusive” and not a proper test of SRI methods (email, May 23, 2003). The results of the 2002 trials were attached to his email. In three of the sets of comparisons, SRI methods had performed more poorly than standard methods, while in another set, there was no difference. In two of the sets, SRI methods performed quite impressively, with yields of 6-7 tonnes per hectare, compared to 3 tonnes for the controls. But those who conducted the trials did not regard them as proving anything very definite about SRI one way or the other.
The consensus of the meeting was that trials should be conducted again in the following year. But when Karl left his position in Laos, there was no leadership for following up with better-managed trials in 2003. The results of these trials were taken seriously enough by some, however, to be included in a meta-analysis of SRI results. And they may have been taken seriously at IRRI headquarters in Los Baños, even though they were discounted by its staff in the field as not being a valid assessment.
A more systematic assessment of SRI was done in Sri Lanka in 2003, more like the evaluation done in Cambodia, conducted by the International Water Management Institute (IWMI), one of the centers supported by the Consultative Group for International Agricultural Research (CGIAR). Its research team selected two districts and then used random-selection methods to identify 30 farmers in each district who were using SRI techniques and a matched set of 30 who were not. The whole sample was made up of 120 rice-growing farmers, half using SRI methods, and half using their usual cultivation practices.
When invited to review a first draft of the team’s report, I noted and commented that the tone and conclusions of the text did not seem to match very well the data reported. There was a consistent tone of ambivalence in the text, while the numbers presented in the tables were quite positive.
The average yield increase recorded with SRI management was 44%, while profits per hectare were almost doubled in comparison with conventional production. Farmers’ assessments of SRI were highly positive. But the report itself characterized SRI as no more than a “niche production method.” The picture below is from the cover of the IWMI research report on SRI in Sri Lanka.
The IWMI study assessed farmers’ economic risks when they adopted SRI, as did the GTZ study in Cambodia although with a different methodology. IWMI researchers calculated from their data the probability that a farmer would have a net financial loss from his rice crop (costs greater than revenue) when cultivated with either with SRI or with conventional methods. These calculations were made first figuring the costs of family labor as being zero, which is how farmers usually reckon since they do not pay themselves, and the grain that they produce for home consumption is considered as their compensation for their work.
Then the numbers were recalculated, valuing family labor at the prevailing wage for hired agricultural labor, attributing a market value to the family’s labor inputs. Then the numbers were reworked once again, assuming that all labor was paid at the current off-farm wage rate. This method assessed labor at its opportunity cost. The risks of economic loss with SRI vs. conventional methods were found to be quite different, with the probability that farmers would experience a net financial loss from growing their rice with SRI methods being much less in both the wet and dry seasons than with their usual methods.
It was also interesting that the study found SRI was more likely to be adopted by poorer and by richer farmers than by middle-scale farmers, contradicting an earlier finding in Madagascar that SRI would not be attractive for poor farmers to adopt. The survey found that the number of irrigations for SRI crops was reduced by 24%, and the number of hours of irrigation needed was 23% less. While such savings would be a boon for farmers, the IWMI report dismissed SRI water-saving as being of little benefit unless there was widespread adoption.
One would have expected that a water management institute, dedicated to raising irrigation water productivity and reducing water constraints (and conflicts), would be interested in investigating ways to make it more feasible for Sri Lankan farmers to adopt an innovation that could reduce their water requirements by one-quarter, while raising rice yields by almost half. However, the IWMI report on its evaluation made SRI seem uninteresting.
This evaluation was resurrected four years after publication of the IWMI report. Two of the study team members joined by two colleagues rewrote the findings from the study’s findings and published them in a German agricultural journal. With no editorial interference in the language of the text, this rewritten assessment of SRI in Sri Lanka reads like an entirely different study. A comparison of the data reported, however, confirmed that it was summarizing the same research findings. Unfortunately for SRI, the 2008 journal article had practically no impact compared to the initial IWMI research report that was published by a CGIAR institution.
In neighboring India, there was more interest in SRI than in Sri Lanka, although work with SRI got started in Andhra Pradesh state and was accelerated in Tamil Nadu state as a result of visits that Indian researchers made to Sri Lanka, as seen below and in Chapter 41.
SRI evaluation began in 2000 at Tamil Nadu Agricultural University (TNAU) through the Wageningen University project discussed in the next chapter. Dr. T.M. Thiyagarajan, known as TMT and director of the university’s Center for Soil and Crop Management Studies, was an ideal person to undertake evaluation of SRI, in part because he had good professional and personal connections to IRRI. He was able to get several PhD students at his university to do SRI evaluation research, and he did his own studies on the effects of mechanical weeding as green manure for SRI rice crops. TMT’s work on SRI got a boost from his making a visit to Sri Lanka in 2002, supported by CIIFAD and hosted by Gamini Batuwitage, at the time a Senior Assistant Secretary of Agriculture in Sri Lanka.
Small-scale trials conducted on the TNAU research station led to the university’s undertaking a comprehensive on-farm evaluation of SRI in the 2003-04 season. One hundred farmers participated in comparison trials in the Thamirabarani river basin, with TNAU technicians supervising their trials on 1000-m² side-by-side plots where the farmer used SRI methods on one and his usual methods on the other. The same varieties were planted on the same soil, with climate and management being identical for both plots.
Only 36 of the farmers used all of the SRI methods as recommended, however. The other 64 used most of the SRI methods but not necessarily up to the standard expected. Even so, the farmers harvested 28% more paddy rice from their SRI plots than from those with their own practices. 31 of the 100 farmers got more than 8 tonnes per hectare from their SRI plots, while with their usual methods, only 3 of the hundred reached this level of productivity.
Aggregate results were impressive. Farmers’ seed rate was reduced by 90%, and their water savings were 40-50%. Water productivity (kg of rice produced per m3 of water) went up by 40%. Farmers had a 10% reduction in their per-hectare costs of production. Part of the cost-saving was because farmers were no longer spraying herbicides and also used less fertilizer.
Farmers’ net economic return per hectare was calculated to be higher by 114% with SRI. Our expectation at the time was that these multiple benefits would be achieved at the cost of farmers having to make greater labor inputs for SRI cultivation. However, the data gathered showed there was an 8% reduction in the number of hours of labor required per hectare for SRI.
Some of the most interesting data from this study was SRI’s impact on the gender division of labor. In Tamil Nadu as in much of India, hand weeding of rice crops is considered to be ‘women’s work,’ so weeding is invariably performed by women. However, SRI weeding with a mechanical weeder got classified as ‘men’s work’ because it was mechanical, so men took over this operation from women.
The reassignment of this task meant that men’s per-hectare labor inputs for SRI cultivation went up by 66%, while women’s labor for rice production was reduced by 25%. Overall, there was an average reduction of 8%. The costs that this transfer of tasks imposed upon men was well rewarded since the profitability of rice production per hectare was more than doubled with SRI methods. This evidence supported the Tamil Nadu state government’s inclusion of SRI in a large World Bank-funded irrigation improvement project (IAMWARM) that was started in 2006, as discussed in Chapter 30.
The Director of Extension for the state’s agricultural university, known as ANGRAU, visited Sri Lanka in 2003 to observe SRI methods and results there. This eventful visit is described in Chapter 41. Of relevance here is that upon returning to Andhra Pradesh, Alapati Satyanarayana quickly got 200 on-farm evaluations set up for the summer (monsoon) season in each of this large state’s 22 districts. The state’s extension service supervised 150 of the trials, while 50 were overseen by ANGRAU technicians.
Participating farmers each set aside 0.4 hectare of rice land (1 acre) and cultivated half of it with SRI methods and the other half with their respective, mostly ‘modern’ methods. The farm sites ranged from lowland coastal delta areas through middle elevations (Telangana) to interior upland regions (Rayalseema). Nineteen different varieties of rice were used.
While the farmers’ own practices on the trial plots yielded 4.89 tonnes per hectare on average, their average yield from SRI plots was 8.34 tonnes per hectare, 2.5 tonnes more and a 70% yield advantage for SRI. One of the most interesting findings was that the yield advantage with SRI was 1.15 tonnes per hectare in the coastal areas, which were commonly believed to be the ‘best’ areas for growing rice. In Telangana, the SRI advantage was 2.5 tonnes on average, while in upland Rayalseema it was 4.7 tonnes.
This latter region was regarded as being not particularly good for rice production because of soil-quality and water limitations. However, the well-drained upland soils of Rayalseema, less saturated than the low-lying heavy delta soils, gave very good rice yields if provided with organic matter that supported more aerobic life in the soil (Chapter 5).
Twenty per cent of participating farmers in the delta areas got an SRI yield over 10 tonnes per hectare, and 25% of the Telangana farmers reached this enviable level of rice production. In Rayalseema with its notoriously ‘poor’ soils, it was a very pleasant surprise that 60% of farmers’ SRI yields surpassed 10 tonnes per hectare. These results challenged and revised our and maybe others’ thinking about what are the ‘best’ soils for growing rice.
No precise measurement of changes in water consumption was possible, but farmers and extension staff reported that water requirements with SRI were lowered by 40-50% while giving higher yield. The next year, in 2005, an evaluation by the Andhra Pradesh extension service of 1,525 on-farm comparison trials showed average SRI yields to be 8.73 tonnes per hectare vs. 6.31 tonnes with farmer methods, again a difference of almost 2.5 tonnes per hectare.
Another interesting result of SRI evaluations conducted in Andhra Pradesh was the experience of a large commercial farmer who used the new methods on 40 hectares (100 acres) of his paddy land. He got an average harvested (not sampled) yield of 11.13 tonnes per hectare, as measured by extension staff and adjusted to a standard level of grain moisture. So while small farmers, in relative terms, could get the most benefit from SRI ideas and methods, these were seen as able to benefit also larger farmers.
As discussed more in Chapters 13 and 41, one of the first SRI initiatives in India was taken by the NGO PRADAN in Purulia district on the western side of West Bengal state. Just four farmers there were willing to try out the new methods in 2003, but in the next season, 150 farmers took up the methods. This was an impressive expansion, so in the summer of 2004, the India program of IWMI, the International Water Management Institute, sent an evaluation team to Purulia to evaluate the SRI experience there.
SRI methods had to be adapted in Purulia because its rice production was rainfall-dependent, with few irrigation facilities. Rainfall came in abundance during the summer monsoon, but there were increasing water shortages and stress during the latter months of the season. We thought that if widely-spaced young seedlings could be transplanted with the onset of the monsoon rains -- with compost provided to the soil and standing water on the fields minimized – the growth of larger, deeper root systems would equip rice plants for greater survival and success when water became scarce.
The IWMI team studied the experience of 110 farmers in two villages, bringing in qualified land surveyors to avoid any errors in measuring field areas. The team used the same methods for calculating yield from both the SRI and conventional fields, so any errors in measuring weight would not be biased in favor of one cultivation system or the other.
There was an average SRI yield advantage of 32%, with 50% higher yield in one village, and 12% in the other, which was harder-hit by drought in that season. The latter village had three dry spells rather than one, and the farmers there did less soil-aerating weeding, also having transplanted older seedlings.
The calculated net increase in income per-hectare from SRI rice production, averaged for the two villages was 67% in this adverse-climate season. Not all of the SRI methods were used as recommended; only one of the 110 farmers did 4 mechanical weedings, for example, and six did 3 weedings. The rest did only 1 or 2 weedings. The research team observed that the farmers who used SRI methods more fully had better results.
As in the Tamil Nadu evaluation, the researchers found that the number of hours of labor invested per hectare was reduced with SRI cultivation by 8% on average. Labor inputs were on average 1,005.5 hours per hectare with conventional cultivation and 874.5 hours with SRI. Farmers’ output of paddy per hour of labor invested was increased by 40%.
There was a 90% reduction in seed requirements, very important for farmers who are living in poverty since any rice not needed as seed can be consumed. The most dramatic and important impact of SRI was that farmers saw it increasing their seed multiplication rate by 20 times With conventional methods, planting one kg of seed yielded back 4 kg at harvest time. Under SRI management, with seed reduction and higher yield, one kg of seed was calculated to return 845.6 kg of paddy rice for farmers’ households.
Since this was rainfed agriculture, reductions made in water use were not calculated or meaningful, but farmers saw that they got more yield from whatever water was available from rainfall. Also, farmers could see the larger root systems on their SRI-grown plants, and a lighter, brighter color of the plant roots. And there were some truly amazing crop growth responses. Farmers also reported fewer pests, particularly fewer stem borers.
By 2007, after five seasons of SRI in Purulia, PRADAN reported that the number of farmers using SRI methods had risen to 3,793, on 528 hectares (rice areas there are very small). Average paddy yields in the district were 2.5 tonnes per hectare. The farmers who used SRI methods were probably somewhat better farmers because average yield on their conventionally-managed fields was 5.4 tonnes per hectare. (Possibly they were also integrating some SRI practices in their ‘usual’ cultivation.)
In any case, the average yield from SRI plots was 7.1 tonnes per hectare. The farmers who had four years of experience with SRI methods averaged 8.5 tonnes from their plots, so apprehension that SRI yields would decline over time without the use of chemical fertilizer was allayed.
The ensuing spread of SRI in West Bengal was not as rapid as hoped by the NGOs doing extension work, at least in part because the state government held back from supporting SRI’s spread. The data from these evaluations certainly warranted extensive diffusion. Of particular relevance for our learning about SRI was that its ideas and methods could be adapted productively to rainfed agroecosystems. Having an assured and controlled water supply gives the best results, but SRI principles could be adapted to raise yields from non-irrigated rice production, as discussed in Chapter 13.
In 1999, a staff member of the international NGO CARE brought a paper on SRI back to this country from a conference in Italy and circulated it. CARE-Bangladesh tried out the methods in Kishoreganj district, and it got the Department of Agricultural Extension there also to try out the new practices in the 2000 winter season. The next year, the Bangladesh branch of Syngenta and the NGO BRAC, formerly known as the Bangladesh Rural Advancement Committee, began their own evaluations.
In January 2002, these organizations plus some others, including the Bangladesh Rice Research Institute and the Bangladesh office of IRRI, formed a national steering committee to evaluate SRI and promote it as appropriate. Four members of the committee on behalf of the whole group -- BRAC and two other NGOs (SAFE and POSD) plus Syngenta/Bangladesh -- put together a two-year evaluation plan that was funded by a British aid project (PETRRA) which was being administered through the local IRRI office.
Between October 2002 and June 2004, under the supervision of these four organizations, 1,278 farmers (193 of them women) compared SRI methods with their own practices on their own fields in 8 subdivisions of 4 districts across Bangladesh.
The summary findings of this large study were that SRI methods raised farmers’ paddy yield by 30% on average, with 7% lower costs of production per hectare. The methods increased farmers’ net income per hectare by 58%. Although this study was conducted under IRRI auspices, with a methodology approved by the local IRRI office, it appeared to have little impact on thinking and practice beyond Bangladesh.
As discussed in Chapter 42, SRI trials were carried out here as early as 1998 and 1999 at Cornell faculty initiative. But these trials were unsuccessful as the water levels in the plots were not controlled during the monsoon season, and the young seedlings transplanted in the trial plots struggled and often drowned.
In 2002, a British-funded project in the Nepal plains (terai), implemented by the Dutch consulting firm (NEDECO), started evaluation of SRI methods in the Sunsari-Morang irrigation scheme. This was a large 65,000 hectare system greatly in need of better water management and productivity improvement. The project worked with farmers in a participatory way, giving training and doing experiments and demonstrations through the farmer field schools that it had helped to set up at the grassroots level.
In the monsoon season 2002, 12 groups of farmers did parallel trials of farmer practice vs. improved practice vs. SRI. Improved practice was the methods being recommended by the government’s extension service. SRI methods were with 10-day-old seedlings spaced 30 × 30 cm and with four mechanical weedings if possible.
The average yield results from the 12 sets of trials was 8.1 tonnes per hectare with SRI methods, 5.8 tonnes per hectare with ‘improved’ practices (mostly fertilizer), and 4.4 tonnes per hectare with farmer practice. As part of the experimentation, farmers simulated a stem borer attack by cutting off 10-20% of the stem 14-28 days after transplanting and defoliating the plants by removing 25-100% of the leaves at an early stage of plant growth. That they found no significant effect on yield showed to farmers the importance of having good early establishment of rice plant roots.
The next monsoon season, the evaluations were continued with 13 farmer field school groups reporting their results from different locations within the Sunsari-Morang system. The average SRI yield in 2003 was 8.48 tonnes per hectare; improved practice 6.2 tonnes; and farmer practice 4.2 tonnes, the results very similar to those of the previous year.
Some farmer experiments assessed how much yield loss farmers would incur if they transplanted older seedlings or if they did not do soil-aerating weeding. SRI methods used with 21-day seedlings instead of 10-day seedlings yielded 6.74 tonnes per hectare, and using weedicides to control weeds instead of mechanical weeding gave 6.27 tonnes. These yields were very respectable, indeed most Nepali rice farmers would be very happy with such yields. But these yields were 20% and 26% less than what farmers produced when they used SRI practices fully.
These positive results elicited much farmer interest and enthusiasm. However, when DFID funding for the project ended, this initiative faded away despite having established strong and consistent evidence of SRI’s productivity. Nepali staff implementing the project formed an NGO that they hoped could get government or donor funding to continue and expand the SRI work, but they had no success.
Fortunately, an agricultural extension officer in the government’s district office in Biratnagar, Rajendra Uprety, took up the SRI challenge and started momentum for getting SRI used in other parts of Nepal, as discussed in Chapter 42.
In 2004, the dean of China Agricultural University’s College of Resources and Environmental Science, Dr. Zhang Fusuo, had learned from his field research in Sichuan province about some SRI adoption and impacts there. He encouraged the university’s Center for Integrated Agricultural Development to send a research team to study SRI uptake in Xinshen village in Jianyang county and to document its effects. This village’s experience had attracted Zhang’s attention because SRI use there had gone from 7 of the village’s 612 households in the first year, to 298 households the next year.
The research team learned that 2003 had been a drought year in Sichuan, and while average paddy yield in Xinshen village had declined by 26%, the handful of farmers who had used SRI methods (mostly members of the village committee or their relatives) had harvested 9% more than their usual yield. This gave impetus to more widespread adoption.
The study team interviewed 75 households, randomly selected to represent the whole community, and conducted focus-group discussions with both SRI and non-SRI farmers. What was most surprising to the research team was that the greatest advantage that farmers attributed to SRI was its labor-saving. This was calculated to be 7% per hectare, similar to what was found in two Indian studies reported above.
When asked what they liked most about SRI, farmers ranked labor-saving higher than the 48% increase in yield and SRI’s water saving of 46% and its 7% reduction in costs of production. This was a study done just in a single village, but it was done by experienced field researchers, giving insight into both farmer assessments and quantified agronomic and economic impacts. As such, it added to the confidence that could be placed in SRI ideas and methods.
All-China Evaluation: A decade later, an agronomist at the Northwest University of Agriculture and Forestry in Shaanxi province, Wei Wu, took it upon himself to do a meta-analysis of all the evaluations of SRI that had been done and published by Chinese researchers.
Wei searched through three internet databases to identify all the published research that compared the results of SRI trials with results from what the researchers had considered ‘best management practices’ (BMP). Wei found more than two dozen published studies that reported enough data and detail to analyze the results thoroughly. After he had done a meta-analysis of his large data set, he asked me by email if I could help with further analysis of the data and the write-up of results.
When the full data set was first analyzed, Wei found that SRI methods had an average yield advantage of 11% over researchers’ BMP. This was the opposite of a previous meta-analysis which had concluded that BMP had an 11% yield advantage over SRI. When the studies in the database were examined carefully, it was seen that only a few had used SRI practices fully or as recommended. For example, most of the SRI trials relied more on chemical fertilizer than on compost or did no soil-aerating weeding, relying on herbicides. This called for a more disaggregated analysis, considering the degree to an SRI protocol had been used.
When the extent of SRI utilization was quantified to assess the results from different degrees of SRI use, in the comparison trials where there had been ‘good’ use of SRI practices (scoring ≥20 points on a scale where 27 points represented full SRI), the SRI yield advantage over BMP was 20%. On the other hand, with ‘minimal’ use of SRI practices (<10 points), BMP had a 4% yield advantage over SRI. It is unfortunate that such disaggregation had not been done before; but ten years earlier there would not have been enough published research on SRI to make such a quantitative evaluation possible.
This was, after China, the next country outside of Madagascar where SRI’s effectiveness was confirmed by rice scientists. There was, however, considerable resistance to SRI within official circles given the success that Indonesia’s rice sector had had with a Green Revolution strategy, as discussed in Chapter 39.
Fortunately, several NGOs and the national integrated pest management (IPM) program supported by FAO moved ahead with SRI, although with little government backing. In Indonesia, the biggest thrust for SRI came, unexpectedly, from the private sector, from the Japanese consulting firm Nippon Koei.
The leader for this firm’s technical assistance team helping to implement an irrigation system improvement project in the eastern part of the country, Shuichi Sato, initially dismissed SRI claims when he first heard about them in 2002. But at the encouragement of two expatriate advisors, Sato decided to evaluate the methodology, initially in just two provinces.
By 2006, his team had assembled a huge data base for assessing SRI impacts. Over nine seasons across eight provinces, the team had monitored a total of 12,133 on-farm comparison trials covering 9,429 hectares. This was an empirical data base seldom matched in terms of numbers, time, and diversity of agroecosystems, and the results were truly impressive.
Farmers compared SRI methods side-by-side with their usual methods. With the new methods, there was an average yield increase of 3.3 tonnes per hectare. This was a 78% increase, achieved with a 40% average reduction in irrigation water issues and with a 50% reduction in the use of chemical fertilizers. Farmers’ per-hectare costs of production were calculated to be reduced by 20%. Below is a picture widely circulated by the project, showing a farmer displaying the phenotypical differences that were being observed with respective management practices.
An analysis of the amount of labor input required by farmers in Eastern Indonesia for their SRI production showed SRI to be labor-neutral as seen in Cambodia. Assessments of SRI’s labor requirements differed from country to country. Some concluded that labor needed with SRI increased, while others showed the opposite, that SRI would be labor-saving. This pointed to the need for both more refined analysis and evaluation of labor as a factor of production.
ASSESSING LABOR REQUIREMENTS WITH SRI CULTIVATION METHODS
It is unfortunate that the labor inputs for SRI were first reported from Madagascar. Traditional methods of rice cultivation in this country are quite different from rice production methods in Asia, where the density of population is about 3x higher. The agricultural labor force per unit of land is thus quite high in most of Asia, especially in comparison to Madagascar.
The first study characterized SRI as a labor-intensive methodology not only because it relies more on labor than on capital, which is a correct characterization, but because when Malagasy farmers started practicing SRI they had to expend more labor per hectare than before. In Madagascar, practically any effort to raise rice yields will require more work, at least at first. In Asia, on the other hand, agricultural land is relatively scarce and the labor for rice cultivation is relatively abundant. This has led to more intensive use of land resources in Asia with greater inputs of labor to get the highest possible output from the available land.
When SRI moved to Asia, different conclusions were drawn about SRI’s labor-intensity than in Madagascar. In most Asian countries, farmers and evaluators found that once SRI methods had been learned, they were actually labor-saving, or at least labor-neutral. Whether the new methods would be more labor-intensive (more labor-demanding) or not will depend largely upon what are farmers’ current methods of production.
How can SRI be labor-saving when it is based on managing all of the resources used for rice production – land, labor, water, seed, and purchased inputs – more carefully? A major explanation is that with SRI, plant populations are so greatly reduced, by as much as 90%. This lowers labor requirements right away. Also, the knowledge and skills required to practice SRI are reasonably simple and can be learned fairly quickly.
When rice fields are no longer kept flooded to control weeds, it is true that more weeding will be necessary. But mechanical weeding with a simple implement can save a lot of labor time compared to manual weeding, while it also reduces drudgery. Farmers have even told me that it is not more difficult to harvest an SRI crop even though it is twice as large. Why? Because SRI panicles ripen more uniformly and cutting them goes more quickly. Also, the panicles stand more evenly, at the same height, because they are growing on (large) individual plants, not on half a dozen different plants crammed together.
Where rice production is more labor-extensive than labor-intensive as in most of Africa, evaluations of SRI’s labor requirements have been mixed, as seen in the box in Chapter 1 reporting on SRI experience in the Mwea irrigation scheme in Kenya. In some areas of that scheme, farmers found that SRI required more labor, while in other areas it required less. Even in Madagascar, where rice farmers have traditionally expended a minimum amount of labor in their rice production, it was found that within several years those who took up SRI were getting more yield with less labor per hectare than when using their usual practices.
Almost everywhere, farmers have seen that with SRI management, the productivity of their labor increases, that is, they are able to produce more kilograms of rice per day or per hour of work expended. For farmers, this is their most important consideration: what are their returns to labor, as well as to their other factors of production, their land, their seed, their water, and their capital? Increasing the returns to any or all of these resources is a boon to farmers.
Of course, it is desirable to reduce the amount of labor required of any household to grow its staple food supply, which is often also a major source of household income. Producing rice, unless there is mechanization, requires the expenditure of a lot of labor, a majority of which usually comes from women. Work in rice fields is difficult, arduous, and often deleterious for health, so mechanization is desirable for multiple reasons, as discussed in Chapter 19.
However, it must also be considered that reducing the amount of labor required for producing rice is not an unambiguous benefit. Many poor households, and especially their female members, depend on wage employment in rice production for much-needed income. So, while decreasing the amount of work necessary to grow rice is good for households in general, reducing the amount of employment in rice cultivation distributes benefits unevenly and somewhat equivocally.
We have been interested in innovations for SRI cultivation that can mechanize certain operations, saving farmers time and money and alleviating the physical stresses and hazards involved in rice farming, considered in Chapter 19. For households that are growing rice with their own labor, labor-saving is a boon. However, labor-saving can reduce income opportunities for those who depend on wage employment in rice fields, and they are often among the most needy and insecure.
There is no single or easy answer. From numerous evaluations, we know that SRI is not necessarily more labor-demanding, and most often it reduces farmers’ labor requirements. SRI requires some more time from farmers initially while they are acquiring experience and moving up on the learning curve. But whether or not SRI will require more labor from a given household depends on how labor-intensive its current rice-growing practices are.
Since about 90% of the world’s rice is grown in Asia, where most production is currently labor-intensive, introducing SRI is likely to reduce the requirements for labor, and almost always it will raise labor productivity. This should make it possible for farmers hiring agricultural laborers to remunerate them more generously.
OTHER THINGS LEARNED FROM EVALUATIONS
The main focus of most evaluations done of SRI has been on grain yield. This is not surprising since the most evident purpose of rice production is to provide food to meet basic human needs, and in the process to generate income for households. But the evaluations summarized here as well as others have produced extensive evidence on other considerations.
As climate change has become more evident and salient in the 15 years after the Sanya conference, we have learned more about how SRI management methods could make rice (and other) crops more resilient to anomalous temperatures and precipitation. This was seen in the evaluation carried out in Sichuan province of China, where resistance to drought on a large scale was documented. Much more evidence has accumulated since then that SRI methods help to buffer the stresses and hazards of climate change, considered in Chapter 12.
As noted in the evaluation in Tamil Nadu state of India, evidence started to build up, not just anecdotal, that SRI methods can contribute to greater gender equity in food production, which is discussed in Chapter 15. The new methods can reduce the burdens and disabilities forced upon women who do a majority of the work for growing the world’s rice. Also, as seen from Andhra Pradesh in India and Eastern Indonesia, the earlier dismissal of SRI as a ‘niche innovation’ is mistaken. This has been demonstrated most clearly by the spread of SRI around the world (Part III).
The evaluation from West Bengal state in India gave evidence that SRI ideas and practices could be adapted successfully to rainfed rice production, elaborated on in Chapter 13, so rice-growing can be improved in less-favored areas that must rely only on rainfall. An evaluation in Nepal, discussed in Chapter 11, confirmed an initial analysis in Madagascar that soil-aerating weeding significantly raised yield, and further that SRI methods can reduce the length of the crop cycle, shortening time to maturity, also discussed in Chapter 11.
These things and many more were learned in the years after the Sanya conference in 2002 as a growing number of researchers, civil society actors, government and donor agencies, and private-sector bodies became involved in their own respective evaluations. They often, although not always, built upon the body of knowledge which was slowly accumulating in the published scientific literature. More and more they connected on-line through the internet, which provided a ‘global commons’ for the dissemination, evaluation and promotion of innovation.
Here in Part I we are considering how an understanding of SRI was developed and diversified. This proceeded at the same time that SRI knowledge was being presented to scientific, policy and other audiences, to get this knowledge accepted and used (Part II) and to be tested and applied in specific country arenas (Part III). It was an undertaking of CIIFAD and then of SRI-Rice at Cornell to facilitate these processes as best we could (Chapter 36).
NOTES AND REFERENCES
 Ken Cassman, at the time on the faculty of the University of Nebraska, was attending a symposium on rice research in June 2000 in Ithaca, NY. This was organized by Cornell faculty in honor of IRRI’s first director-general, Robert Chandler, and Ken and I had an opportunity to talk about SRI between symposium sessions. I remember vividly how Ken calculated, using his fingers for the number of factors, how many trial plots would be needed, and then shaking his head.
 Jean de Dieu Rajaonarison, Contribution a l’amelioration des rendements de 2eme saison de la double riziculture par SRI sous experimentations multifactorielles: Cas des sols sableux de Morandava (2000); and Andry H. Andriankaja, Mise en evidence des opportunités de développement de la riziculture par adoption du SRI, et évaluation de la fixation biologique de l’azote (2001). Andry went on to do a PhD in agronomy at the Swiss national technical university (ETH) in Zürich and is now employed as an agricultural biotechnologist by BASF in North Carolina, USA.
 As noted below, SRI had been characterized early on as a ‘niche innovation,’ only beneficial on certain kinds of (marginal) soils or for certain kinds of (poor) farmers. Achim Dobermann, ‘A critical assessment of the system of rice intensification (SRI),’ Agricultural Systems 79: 261-281 (2004). This characterization was refuted by the state-wide evaluation of SRI done in Andhra Pradesh with on-farm trials across all of its districts in 2005, reported on below, with a wide range of soil conditions and farmers.
 TAFA (TAny Sy FAmpandrosoana, Earth and Development) was a Malagasy organization that worked in cooperation with CIRAD, the Center for International Cooperation in Agronomic Research for Development, with financial support from the French development agency AFD.
 At Anjomakely, on the poorer loam soil, no trials were done with no soil amendments, only with compost or NPK; on the better clay soils, all three nutrient practices were evaluated with replicated trials. This explains why the total number of trials at Anjomakely was 240, while it was 288 at Beforona.
 Age of ‘older’ seedlings differed between the two experiments because rice plants’ biological clock is affected by temperature, as noted in Chapter 6. At Anjomakely with its colder climate, seedlings transplanted at 20 days would be equivalent to 16-day seedlings planted at Beforona with its warmer temperatures. Both ages would be, phenologically, within the 4ᵗʰ phyllochron of growth, and thus ‘older’ seedlings. Both ages were considerably younger than typical with farmer practice in Madagascar, where seedlings were generally transplanted when 6 or 7 weeks old.
 At Morondava, the average yield with 30×30 cm spacing and all other practices being equal was exactly the same as for 25×25 cm spacing. As noted, the difference in yield at Anjomakely was negligible, with 30×30 cm spacing averaging just 80 kg more.
 The total number of trials differed because at Anjomakely, no trials were done on the poorer (loam) soils with no soil fertilization amendments. On the better soils at Anjomakely and in all the trials at Morondava, three soil fertilization alternatives were evaluated: no soil amendments vs. organic compost vs. NPK fertilizer.
 Travelling was physically challenging for Prof. Robert because of leg impairment due to complications from diabetes. Yet he voluntarily did this field supervision without complaint, and without extra payment, just getting his travel expenses reimbursed.
 This table does not include the results of trials in which there was no application of soil nutrients, either organic or inorganic. Such trials were done at Morondava and in half the trials at Anjomakely as explained in endnote 8.
 R. Randriamiharisoa and N. Uphoff, ‘Factorial trials evaluating the separate and combined effects of SRI practices,’ in N. Uphoff et al., eds., Assessments of the System of Rice Intensification (SRI): Proceedings of an international conference held in Sanya, China, April 1-4, 2002, 40-46, CIIFAD, Ithaca, NY (2002), and N. Uphoff and R. Randriamiharisoa, ‘Reducing water use in irrigated rice production with the Madagascar System of Rice Intensification.,’ in B.A.M. Bouman et al., eds., Water-wise Rice Production, Proceedings of the International Workshop, 8–11 April 2002, 71-88, International Rice Research Institute, Los Baños, Philippines (2002).
 This is Table 2 from Uphoff and Randriamiharisoa, cited in the preceding endnote, the data from which are presented below.
 Specific comparisons from the Anjomakely trials showed that, with all other factors being equal across all the trials, the young seedling effect added 2.48 tonnes per hectare to yield (8-day-old seedlings = 6.28 t/ha vs. 20-day-old seedlings = 3 .80 t/ha). Water management added 1.41 t/ha (aerobic soil = 5.75 t/ha vs. flooding = 4.34 t/ha), while organic fertilization added 1.01 t/ha (compost = 5.49 t/ha vs. NPK fertilizer = 4.48 t/ha).
The average yield on clay soils at Anjomakely without any fertilization was 4.25 t/ha, keeping in mind that the rice variety planted there was an unimproved landrace (riz rouge), not bred to be responsive to inorganic fertilizer. The plant-per-hill effect was 0.78 t/ha (1 plant per hill = 5.43 t/ha vs. 3 plants per hill = 4.65 t/ha). The spacing effect was only 0.08 t/ha at Anjomakely, and nil at Morondava. Both of these spacings were ‘wide’ and within SRI range.
The soil effect, averaged for an equal number of trials with compost and NPK fertilizer, was 3.03 t/ha (better clay soil = 5.08 t/ha vs. loam poorer soil = 3.72). As noted above, the average yield on better clay soil without any amendments was 4.25 t/ha. Such numbers are worth considering, but they reflect the particular confluence of factors and conditions in these trials. These results will be somewhat different in different growing environments and probably different with other rice varieties. As noted above, what was most important was the similar patterns of result.
 The average yield from four sets of trials with all four SRI practices including compost was 7.38 tonnes per hectare, while with three SRI practices and NPK fertilization, it was 5.77 tonnes.
 Jürgen Anthofer, Potential of the System of Rice Intensification (SRI) for Cambodia, report to the Food Security and Nutrition Policy Support Project, and the Rural Development Program, Gesellschaft für Technische Zusammenarbeit, Phnom Penh and Spaichenge, April, 2004; and ‘The Potential of the System of Rice Intensification (SRI) for Poverty Reduction in Cambodia,’ paper for Deutscher Tropentag, Berlin, October, 2004. When this book was being written, Anthofer was Executive Secretary of the European Initiative for Agricultural Research for Development, established by the EU.
 Within each of the five provinces, four villages were randomly selected, and then within each village, interviews were conducted with farmers randomly selected, 20 using SRI methods and 5 using their usual practices. The SRI farmers varied in how many seasons of experience they had with the new methods, between one and four, a variable that was analyzed in the study.
 Recall the discussion in a box in Chapter 1 of an evaluation of SRI in Andhra Pradesh state of India where a 50% reduction in the wages paid to agricultural laborers by SRI farmers was considered as a negative aspect of SRI introduction (although economic calculations showed that the three-fold increase in farmer incomes generated more than enough value-added to compensate for the social costs of introducing SRI). A. Gathorne-Hardy, D. Narasimha Reddy, M. Venkatanarayana and B. Harriss-White, ‘System of Rice Intensification provides environmental and economic gains, but at the expense of social sustainability: A multi-disciplinary analysis in India,’ Agricultural Systems, 143:159-168 (2016).
 In 2005, CEDAC, the NGO which spearheaded introduction and spread of SRI in Cambodia (see Chapter 24), did a study of 113 farmers who had been using SRI methods for five years. (SRI use in Cambodia had expanded from 28 farmers in 2000, to over 40,000 by 2005.) This study showed that fertilizer use among these farmers had declined by 60%, from 116 kg per hectare to 46 kg. And the percent of households using chemical pesticides had dropped from 28% to less than 1%. Chey Tech, Ecological System of Rice Intensification (SRI) Impact Assessment in 2001-2005, CEDAC Field Document, Cambodian Centre for Study and Development of Agriculture, Phnom Penh (2005).
 When some eight year later I was able to meet Jürgen Anthofer, the team leader, at the 2ⁿᵈ Global Conference on Agricultural Research for Development held in Punta del Este, he confided that he had taken on this assignment for GTZ with considerable ambivalence, not giving much credence to SRI claims. On the basis of his findings, however, he had become persuaded of SRI benefits (personal communication, October, 2012).
 Here are the results reported by Karl Goeppert, May 23, 2003: “Summary of SRI Trials during the rainy season June-November 2002, at various locations throughout Laos by different organizations”
 A. McDonald, P.R. Hobbs and S.J. Riha, ‘Does the system of rice intensification outperform conventional best management? A synopsis of the empirical record,’ Field Crops Research, 96: 31-36. This analysis is discussed in Chapters 22 and 28. The Laos data reported above constituted 10% of the data base of the McDonald et al. article, even though it had been rejected by IRRI’s office in Laos as not valid.
 R.E. Namara, P. Weligamage and R. Barker, Prospects for Adopting System of Rice Intensification in Sri Lanka: A Socioeconomic Assessment. Research Report 75. International Water Management Institute, Colombo, Sri Lanka (2003).
 The report’s conclusion was: “whether or not to pursue the avenues [which the report had identified to make SRI more attractive and feasible for Sri Lankan farmers] remains an open question.” (page vi).
 Farmer assessments reported in Tables 3 and 4 identified positive features of SRI practice as: SRI saves seed (100%), more tillers (98%), reduced need for herbicide (92%), less lodging of rice (91%), improved seed quality (91%), saves water (90%), less disease and pest attack (88%), reduced need for inorganic fertilizer (86%), reduces input costs (85%), more yield (83%), less labor needed for harvesting [despite the higher yield] (80%), less labor needed for transplanting (78%), and more milling output (77%).
Negative features reported by farmers were: SRI requires more effort (75%), requires well-drained soils (69%), weed control is necessary (60%), not enough organic matter available (58%), does not work on flooded fields (56%), transporting organic matter is problematic (51%), transplanting is difficult (37%), requires skilled labor for management (32%), transplanting requires special skills (25%), and more mice attack (2%).
 This characterization was echoed by Achim Dobermann, a senior IRRI rice scientist, in his article, ‘A critical assessment of the System of Rice Intensification (SRI),’ Agricultural Systems, 79: 261-281 (2004). In October 2006, I learned from Dr. Intizar Hussain, executive director of the International Network on Participatory Irrigation Management (INPIM) who had been a visiting fellow at IWMI in Sri Lanka at the time of the SRI study, that IWMI’s director for social sciences, Dr. M. Samad, was quite negative, even antagonistic toward SRI. (This was confirmed by acquaintances at ICRISAT and IWMI who knew him.) Samad edited and modified the research team’s original draft to make its wording more qualified or more critical toward SRI (personal communication, October 18, 2006). The claim (p. 23) that SRI methods tripled labor requirements, for example, was not supported by any data provided in the report.
 By the first calculation, neither SRI nor conventional methods had any risk of net loss in the rainy season, while in the dry season, SRI had a 1% probability of loss while with conventional rice-growing methods there was a 9% chance of experiencing loss. If family labor was costed at farm-labor rates, SRI’s probability of net loss in the two seasons was 1% and 4%, while with conventional methods rice production had 15% and 27% respective probabilities of loss. When labor inputs were valued at off-farm wage rates, with SRI the probability of net loss was only 2% and 6%, while with conventional methods there were 22% and 29% risks of loss (from Table 15 in the report)
 C.M. Moser and C.B. Barrett, ‘The disappointing adoption dynamics of a yield-increasing, low external-input technology: The case of SRI in Madagascar,’ Agricultural Systems, 76: 1087-1100 (2003).
 Regassa Namara, Deborah Bossio, Parakrama Weligamage and Indika Herath, ‘The practice and effects of the System of Rice Intensification (SRI) in Sri Lanka,’ Quarterly Journal of International Agriculture, 47: 5-23 (2008). One difference between this article and the preceding research report is that the article provided evidence of SRI resistance to drought stress (Chapter 12). The main season crop covered in the survey had experienced 75 days of severe drought, which was not addressed in the report, but was discussed in the article.
 A. Satyanarayana, T.M. Thiyagarajan and N. Uphoff, ‘Opportunities for water saving with higher yield from the system of rice intensification,’ Irrigation Science, 25: 99-115 (2007).
 As this chapter was being drafted, the World Bank announced that SRI promotion would be included in a new $318 million project, to benefit 500,000 farmers in Tamil Nadu.
 These are also reported in the article by Satyanarayana et al. referenced in endnote 29.
 For purposes of comparison, we note that a six-nation comparative evaluation of what IRRI and other scientists called ‘site-specific nutrient management’ -- the fine-tuning of inorganic fertilizer applications to match what were calculated from soil testing to be the rice crops’ nutrient needs -- gave an incremental yield of only 360 kg more paddy rice per hectare. Dobermann, et al., ‘Site-specific nutrient management for intensive rice cropping systems in Asia. Field Crops Research, 74:37–66 (2002). That SRI methods were increasing rice yield by 6x more than IRRI’s new fertilizer-management methodology might have colored the attitude of Institute staff toward SRI.
 This information was reported in Satyanarayana et al. (2007). The large-farmer experience was written up in a 2005 trip report, pages 1 and 10. This large farmer was known for being progressive not just in terms of agricultural experimentation but also socially, he freely interacted with the agricultural laborers he employed, taking care to give good instruction, and observing their work personally. One more instance where social relations are important for SRI.
 S.K. Sinha and J. Talati, ‘Productivity impacts of the system of rice intensification (SRI): A case study in West Bengal, India,’ Agricultural Water Management, 87: 55-60 (2007). This is a shorter version of the authors’ research report for the IWMI-Tata water policy program, Impact of the System of Rice Intensification (SRI) on Rice Yields: Results of a New Sample Survey in Purulia District, India (2005).
 These and other data here come from Sinha and Talati (2007).
 When I happened to meet S.K. Sinha at an IWMI-TATA policy workshop in Ahmedabad, India in 2005, he told me, like Anthofer (endnote 17), that he had been very skeptical about SRI reports before he went to Purulia. But had become quite convinced of its merits by his observations and measurements. He said that one SRI field, the best one, had given a yield of 15 tonnes per hectare, quite unheard of for rainfed rice since this was beyond what was usually achieved with irrigation. Sinha said that he had decided to harvest this field personally and to calculate the yield himself because he know that this result would be difficult for most people to believe. He was confident enough of the number to include it in their 2005 report because he had done the weighing and measurement himself.
 Report on SRI in Purulia District, West Bengal, Kharif season 2007, unpublished field report, 2007.
 Also, neither the International Water Management Institute nor its India program did any further evaluation of SRI, in India or elsewhere (Chapter 22).
 Information here is from the project report submitted to the IRRI office in 2004 by A.M. Muazzam Husain (project leader), Gopal Chowhan, P. Barua, A.F.M. Razib Uddin and A.B.M. Ziaur Rahman, Final Evaluation Report on Verification and Refinement of the System of Rice Intensification (SRI) Project in Selected Areas of Bangladesh (SP 36 02).
 The findings of a small parallel evaluation done by researchers at the Bangladesh Rice Research Institute, with a sample of just 20 farmers in one location, received much more attention from the scientific community because it was published in the literature, even though its empirical base was a fraction of that for the evaluation by BRAC, POSD, SAFE and Syngenta. The study by M.A. Latif, M.R. Islam, M.Y Ali and M.A. Saleque, ‘Validation of the system of rice intensification (SRI) in Bangladesh,’ Field Crops Research, 93: 281-292 (2005), was more detailed in terms of soil, nutrient and other analyses. But its conclusion that SRI methods give lower yield with higher costs of production and less profitability was contradicted by the much larger study done under IRRI auspices with data from eight locations, not just one. (It was contradicted also by many other evaluations of SRI as seen in this chapter.) The study by Latif et al. was included in the meta-analysis of SRI data by McDonald et al. (2006) in preference to the study by BRAC et al. which had a data base 600 times larger.
 Chris Evans, Scott Justice and Shyam Shrestha, ‘Experience with the System of Rice Intensification in Nepal,’ in N. Uphoff et al., Assessments of the System of Rice Intensification: Proceedings of an International Conference, Sanya, China, April 1-4, 2004, 64-66, CIIFAD, Ithaca, NY.
 ‘Sunsari-Morang Irrigation Project: Experience with SRI – Monsoon Season 2002,’ NEDECO-SNAP Team, Biratnagar (2003). The ranges of yield were SRI 5.45-11.1 tonnes per hectare, improved practice 4.85-7.5 tonnes, and farmer practice 2.5-6.5 tonnes.
 In Madagascar, we had reports from farmers that their young SRI transplants after being eaten down to the ground by cicadas, a 17-year cyclical pest in that country, recovered to give a normal yield. However, there were no planned and controlled experiments to verify this.
 Report on SRI-Rice website from the NEDECO-SNAP Team for monsoon season, 2003.
 Personal communication, Robert Davey NEDECO technical assistance team leader.
 Li Xiaoyun, Xu Xiuli and Li He, ‘A Socio-Economic Assessment of the System of Rice Intensification (SRI): A Case Study from Xinsheng Village, Jianyang County, Sichuan Province,’ unpublished paper, Center for Integrated Agricultural Development, China Agricultural University, Beijing, 2005; subsequently published in Chinese in China Rural Economics, March (2006), 1-13.
 2006 and 2010 were also drought years in Sichuan province. The Provincial Department of Agriculture reported that in these years, when SRI use in the province was on 57,466 and 204,467 hectares, respectively, up from only 1,133 hectares in the province in 2004, SRI methods’ yield advantage over standard practice was 11% greater than in the seasons between 2004 and 2012 with more normal rainfall, as discussed in Chapter 12. Zheng Jiaguo, Chi Zhongzhi, Li Xuyi and Jiang Xinlu, ‘Agricultural water savings possible through SRI for water management in Sichuan, China,’ Taiwan Water Conservancy, 61:50-62 (2013).
 I was able to visit Xinsheng village in August 2013 with Dr. Zheng Jiaguo and Dr. Lu Shihua, senior scientists with the Sichuan Academy of Agricultural Sciences and active promoters of SRI. Farmer enthusiasm was very evident from the large ceremonial welcome that the villagers gave us, and interviews confirmed the report of their 2004 experience. But we learned that in this village, rice production, both SRI and non-SRI, was receding because of the low market price that farmers got for their rice, so their resources and effort were being redeployed to vegetable production.
 W. Wu, B.Ma and N. Uphoff, ‘A review of the system of rice intensification (SRI) in China,’ Plant and Soil, 393: 361-381. Wei’s search of the Web of Knowledge, Google Scholar and China National Knowledge Infrastructure data bases, with ‘system of rice intensification’ and ‘China’ as key words, turned up 17 articles with 26 sets of data from field trials in 8 provinces that had sufficient detail on their methodology and data bases so that he could make a detailed assessment.
 See the article referenced in endnote 21.
 Below is the scoring table developed to assess the degree to which SRI practices were used. If all practices were in the first column (3 pts), this would give a total score of 27, representing fullest use. A score of 20 or above was considered ‘good’ use. From 15-19 points was described as ‘median’ use, and 10-14 points as ‘minimal’ use. Less than 10 points represented ‘poor’ use, not qualifying for the designation of SRI.
Scoring matrix for kinds and degrees of SRI crop management, developed to differentiate full use of SRI methods from non-use of these methods, with intermediate degrees of utilization (from Wei Wu et al., 2015).
 In evaluations where there had been ‘median’ use of SRI practices (15-19 points out of 27), SRI’s yield advantage over BMP was 12%. With ‘minimal’ use of SRI methods (only 10-14 points out of a possible 27, or even less), we would expect BMP might gave better results than SRI. For example, if young seedlings are transplanted into a continuously flooded field; this would suffocate the plants’ roots and make the soil inhospitable for most beneficial soil organisms.
 A. Gani, T.S. Kadir, A. Jatiharti, I. P. Wardhana and I. Las, ‘The System of Rice Intensification in Indonesia,’ in N. Uphoff et al., Assessments of the System of Rice Intensification: Proceedings of an International Conference, Sanya, China, April 1-4, 2004, 58- 63, CIIFAD, Ithaca, NY.
 One of Sato’s technical advisors for the project, Ed Vander Velde, had been an advisor for Cornell in the early 1980s implementing the Gal Oya irrigation water management project in Sri Lanka. Both Ed and I had attended a water management conference at the Asian Institute of Technology in Bangkok in April 2002, and Ed took my paper on SRI back to the project in western Indonesia. Ed’s urging that SRI be evaluated was supported by another advisor Harry Clark from the UK.
 S. Sato and N. Uphoff, ‘A review of on-farm evaluations of system of rice intensification methods in Eastern Indonesia,’ CAB Reviews, 2:54 (2007).
 See the article referenced in endnote 24.
 A study team sent by the Tata Trust in India to evaluate SRI experience in Purulia district of West Bengal state collected data from farmers and technicians on the number of hours that were invested in each rice-production operation using SRI vs. farmers’ current practices. It calculated that under SRI management, the number of hours of labor expended per hectare was reduced by 33%.
Similar calculations in Gaya district of Bihar state showed a 23% reduction in labor requirements. V.P. Singh, U. Shankar and P. Bora, Feasibility Study to Support System of Rice Intensification, Report to the Sir Dorabji Tata Trust, Mumbai (December 2007).
 A. Mrunalini and M. Ganesh, ‘Work load on women using cono weeder in SRI method of paddy cultivation,’ Oryza, 5: 58-61 (2008). These researchers calculated that the mechanical hand weeder reduced the labor required for weeding by 58%.
 C.B. Barrett, C. M. Moser, J. Barison and O.V. McHugh, ‘Better technology, better plots or better farmers? Identifying changes in productivity and risk among Malagasy rice farmers,’ American Journal of Agricultural Economics 86: 869-888 (2004). This analysis showed that after three years of experience with SRI methods, they became labor saving, even in comparison with farmers’ traditional labor-extensive methods.
 See discussion and reference in endnote 17.
 One of our biggest mistakes in the first decade of SRI work beyond Madagascar was to not challenge more directly the stereotype of SRI as ‘labor-intensive.’ This was a plausible characterization because SRI, by definition, depends more on labor than on capital. But the image of SRI as demanding ‘more labor’ or ‘too much labor’ derived from the fact that the initial economic evaluation of SRI was done in Madagascar, where the baseline was labor-extensive rice production.
The reference cited in endnote 57 above provided data that could help to correct the stereotyped image of SRI, but the authors’ second article received much less attention than their preceding report, cited in endnote 27. The conclusion of the first article was publicized by those who preferred to dismiss SRI, while the evidence from the second article, based on a larger, better and deeper data set, was mostly ignored. Once the categorization of SRI as ‘labor-intensive’ got accepted, correcting this wrong impression became very difficult.As former dean of the faculty at Harvard University, Henry Rosovsky observed, “Never underestimate the difficulty of changing false beliefs by facts.”
 This was argued in the references from IWMI and IRRI in endnotes 20 and 22.
 In a 2016 blog, David Bollens characterized the SRI network as an ‘eco-digital commons.’ ‘New Forms of Network-Based Governance.’
PICTURE CREDITS: Reports from GTZ and IWMI; Shuichi Sato (Nippon Koei)