4.3 Land-yield combinations, major crops

Contents - Previous - Next

The aggregate crop production is projected to grow at 2.4 percent p.a., down from the growth rate of 2.9 percent p.a. in 1970-90. The reasons why future growth may be lower than in the past were explained in Chapter 3. The overall combinations of harvested area expansion and yield increases underlying the projections for major crops are shown in Table 4.7. The contributions to total growth of increases in area (and, within it, those of expansion of the physical arable-land area and those obtained from higher cropping intensities, i.e. expansion of multiple cropping and shorter fallows) and in yields are shown in Table 4.8. It is noted that these average yields (all crops, aggregated on the basis of the 1979/81 price weights, all agroecological zones, all countries) are not very useful for understanding the agronomic factors underlying the projections.

The data and projections in Tables 4.7 and 4.8 are, however, useful for forming an overall idea of the extent to which the production projections depend on the further expansion of land and irrigation, their more intensive use (increasing cropping intensities) and the continuation of yield growth. In particular, they shed some light on the question whether the future may or may not be like the past, although, as noted in the preceding section, the historical data do not always provide a sufficient basis for making this comparison. It is emphasized that these land and yield projections are definitely not extrapolations of the historical trends. The reader is invited to contemplate what the projections would have been if this study had just extrapolated the explosive area growth rates of the historical period for soybeans and sugarcane in a major country like Brazil, which were 10.8 percent p.a. and 7.4 percent p.a., respectively, in the period 1970-90.

Table 4.7 Area and yields for major crops: developing countries (excluding China)

Note: Sometimes the changes in the annual growth rates between the historical and projection period appear to be large. Often this is a continuation of a change already begun in the historical period or an expected change in one country which has a large weight in the total. For example, annual growth in sugarcane production in the developing countries excluding Brazil is projected to remain the same as in the historical period, namely 2.2 %. Likewise, the area allocated to soybeans in Brazil (currently more than half of total soybean area in developing countries) grew at 21.2% annually in the 1970s, but this growth rate fell to 3.5% in the 1980s.

*Revised data for 1991/92 as known in May 1994, but not used in this study.

Table 4.8 Sources of growth in crop production and in harvested area, developing countries, excluding China (%)

The overall conclusions are that: (a) for the major crops (e.g. cereal, soybeans) the growth rates of average yields can be expected to be much below those of the last 20 years, e.g. 1.6 percent p.a. for wheat compared with 2.8 percent p.a. in the past, 1.5 percent p.a. versus 2.3 percent p.a. for rice, etc. (Table 4.7); (b) harvested area expansion will continue to play a significant role in total crop production growth, though, like in the past, much less significant than that of yield increases. In parallel, increases in the cropping intensities of, mainly, the irrigated areas will play a predominant role in the land-scarce regions (South Asia, Near East/North Africa); and (c) as noted in the preceding section, the expansion of irrigated land will probably proceed at a much slower pace than in the past.

Table 4.9 Production of major cereals by land class in developing countries (excluding China)*

*Data on land and yields by land class at the country level do not exist in any systematic form. They have been assembled for this study based on whatever information was available (country/project reports, expert judgement, etc.). They should therefore be interpreted with care.
^†For an explanation of the land classes see Section 4.2 of this chapter.
^‡A, Y, P: Area in million ha; Yield in tonnes/ha; Production in million tonnes.

Land-yield combinations in the cereals sector

In the developing countries (excl. China), rice (in terms of paddy) is the most important cereal crop, accounting for 48 percent of total cereals production in 1988/90, followed by wheat (21 percent), maize (18 percent), sorghum (6 percent) and barley and millet (each less than 4 percent). The growth in production of wheat and rice is expected to slow down considerably over the projection period as compared with growth in the last two decades. Coarse grains would maintain their past annual growth in the future reflecting in part the strong growth of demand for cereals used for feed.

By far the greater part (82 percent) of the production of wheat in the developing countries (excl. China) is located in South Asia and Near East/ North Africa. Rice production is concentrated in South and East Asia (89 percent) and barley is mainly produced in Near East/North Africa. These are land-scarce regions with higher than average dependence on irrigation. Both necessity and potential dictate that much of the production increases of these three cereals will come from higher yields. Maize and sorghum are mainly produced in Latin America and sub-Saharan Africa, while millet production is evenly distributed between sub-Saharan Africa and South Asia. By and large, the predominance in the production of coarse grains (except barley) of the two land-abundant regions with rainfed agriculture indicates that area expansion will play a comparatively larger role, than for wheat and rice, in the growth of production.

The data and projections in Table 4.7 demonstrate this prospect. For example, a 2.0 percent p.a. increase in rice production may be achieved by a 0.5 percent p.a. increase in harvested rice area (and much less in arable area under rice). At the other extreme a 2.5 percent p.a. growth rate in sorghum production will be based on the harvested sorghum area growing at 1.4 percent p.a. The fact that the production of these coarse grains is overwhelmingly rainfed with, for millet and sorghum, a high proportion being in the two semiarid land classes, explains why growth of yields may be expected to make a less important contribution to production growth than for wheat and rice. The overall picture of possible area-yield combinations by agroecological land class underlying the production projections of cereals is shown in Table 4.9.

These agroecological distinctions convey significant information about the sources of growth in average yields, because they make it possible to distinguish between "genuine" yield increases (due to more fertilizer, better varieties, management, etc.) in a production environment with given physical characteristics, from increases in average yields due to the relative shift of production from the lower to higher potential land classes. For example, part of the I tonne/ha increase in the average rice yield is due to the possibility that the share of irrigated land in total rice area will rise from 43.5 percent p.a. in 1988/90 to 47.9 percent in 2010. When this happens, it is possible for the average yield to increase even if irrigated and rainfed yields remained the same.

However, by far the most important factor in raising average yields will be the yield growth in each and every land class. The extent to which such yield increases in a given agroecological zone may depend on the generation by the research system of new varieties (e.g. those that would make possible quantum jumps in the rates of yield growth) or of varieties contributing to slower (evolutionary) growth of yields and for the periodic replacement of existing ones subject to erosion of their yield potential, is central to the issue of research requirements and priorities to sustain the growth of production (for a useful discussion see Byerlee, 1994 and Plucknett, 1994).

The agroecological characteristics used to classify agricultural land into categories in this study provide some useful background for addressing this issue. How useful they are depends on whether the resulting land classes may be considered to be fairly homogeneous as to their potential for yield growth. For example, this condition would be fulfilled if land in the rainfed category "sub-humid" (AT3) had the same physical and biotic characteristics (LOP, soil, terrain) in the different countries, say Argentina and Mexico. In this land class Argentina achieves maize yields of 3.5 tonnes/ha but Mexico only 1.4 tonnes/ha. If the physical production environments were the same it would be reasonable to consider that existing varieties provided considerable scope for yield growth in Mexico. In this case, increasing maize production in Mexico would depend less on new research breakthroughs to increase maize yield ceilings for this type of land and more on a combination of adaptive research and the socioeconomic factors that would enable the Mexican farmers to raise yields by adopting the varieties which produce the higher yields in Argentina.

However, the agroecological criteria used in this study are not sufficiently fine to permit the derivation of firm conclusions based on this type of argument. For example, the rainfed land class "sub-humid" comprises land with: (a) moisture regimes ranging from LGP 180 days to 270 days; (b) soil/ terrain characteristics in classes "very suitable" and "suitable"; and (c) the yield potential of "very suitable" land goes from 80 percent to 100 percent of the maximum constraint-free yield (MCFY) and that of "suitable" from 40 percent to 80 percent of the MCFY. It follows, therefore, that the land class "sub-humid" spans a wide range of production environments. In the event, land classified as sub-humid can have potential maximum yields from as low as 40 percent of the MCFY to as high as 100 percent. In conclusion, if most of the "sub-humid" land of country X is nearer the lower limit of the range and that of country Y is nearer the upper limit, the potential for raising yields in country X mainly by means of policies favouring the transfer, adaptation and diffusion of existing technology and varieties used in country Y would be severely circumscribed. It is noted that even land in the irrigated class is far less homogeneous than suggested by the apparent flexibility afforded by irrigation to relax the moisture (LGP) constraints. This is because, the quality of the irrigated land can range from "fully equipped and not subject to water shortages" to that of "partly equipped and subject to water shortages".

Table 4.10 Cereal yields in major agroecological land classes and inter-country differences, developing countries (excluding China)

* Yields of countries with at least 50 000 ha in the land class and crop shown.
** Simple averages of the yields of the bottom 10% and top 10% of the countries ranked by yield level (not always the same countries in the top or bottom deciles in each year).

Table 4.11 Inter-country gaps in average yields for wheat and rice, developing countries, excluding China* (tonnes/ha)

* Data and projections for countries with over 50 000 ha under wheat or rice in the year shown. Average yields are simple averages (not weighted by area).

These limitations of the agroecological classifications notwithstanding, their use in the analyses of this study in association with the distinction of individual cereals, rather than cereals or coarse grains as groups, is a useful step in the direction towards enlightening the debate on the extent to which future yield growth depends on new research breakthroughs. It certainly provides a sounder basis for judging this issue than the mere comparison of inter-country differentials in average yields, let alone differentials among large country groups, e.g. between average yields of the developed and the developing countries. For example, average country yields for wheat in the Near East/ North Africa region range from nearly 5 tonnes/ha in Egypt and Saudi Arabia (all irrigated) to 0.8-0.9 tonnes/ha in Algeria and Iraq (mostly rainfed and in semi-arid conditions). Obviously, these differentials in average country yields convey no useful information for the issue at hand. But it is more relevant to know that rainfed wheat yields in the "sub-humid" land class are twice as high in Turkey as in most other countries of the region. This land-specific yield gap comes closer than the average one to identifying the scope for yield gains in the lagging countries through a catching-up process (for a discussion of this issue based on an analysis of the inter-country differences in average yields see Plucknett, 1993).

The preceding discussion is relevant for examining the factors underlying the yield projections of this study. The dominant factor is the realization that the scope for raising yield ceilings by quantum jumps through the introduction of new varieties is more limited now than it was in the past (Ruttan, 1994). Therefore, much of the growth in average yields must come less from raising yields of the countries with the highest yields today and more from raising those of the countries, particularly the large ones, at the middle and lower ranges of the yield distribution. This is why the yield projections imply a narrowing of the inter-country yield differences for each land class. The relevant data and projections by land class are shown in Table 4.10.

Does this pattern conform to the historical experience? It is not possible to investigate this issue for individual land classes because there are no relevant historical data. Such data exist only for average yields (over all land classes) in each country. They show that the gap between the countries with the highest and the lowest yields (simple averages of the top and bottom deciles of the countries ranked by yield level) had widened between 1969/71 and 1988/90 (Table 4.10). This occurred mainly through a process whereby the yields of the countries in the top decile in 1969/71 rose by more than those of the countries in the bottom decile. The projections for average yields (over all land classes) imply that the future may be unlike the past and the yield gap may become narrower because the scope of yield growth in the top decile countries of 1988/ 90 is more limited than it was 20 years ago.

This pattern is illustrated in Table 4.11 with data of individual countries for wheat and rice (data for similar comparison for other cereals can be found in Appendix 3). For wheat, the countries in the top decile of the distribution had in 1988/90 yields which were nearly twice as high as the countries in the top decile of 1969/71. In contrast, there was much less yield growth in the countries at the bottom decile. These developments are even more pronounced for rice.

The dependence of the future growth of aggregate production of the developing countries on the narrowing of the inter-country yield gap (as measured here, i.e. the difference in the simple average of yields in the top and bottom deciles of countries) should not, however, be exaggerated. This is because the countries at the two ends of the distribution (top and bottom deciles) account for a relatively small part of the total production of the crop examined. This is true even when, as has been done for Tables 4.10 and 4.11, countries with less than 50 000 ha under the crop (and for Table 4.10 also under the given land class) are excluded from the analysis. In practice, the realism, or otherwise, of the projections of total production of the developing countries, depends crucially on that for yield growth in the countries which account for the bulk of the area under each crop.

For this purpose Table 4.11 also shows the relevant historical data and projections of the 10 percent of countries with the largest areas (top decile of countries ranked by area under the crop). It is seen that: (a) these countries have yields which are less than one-half those in the countries with the highest yields; (b) for wheat, their (simple) average yield is projected to grow by 43 percent which is below the 62 percent increase of the last 20 years; (c) for rice the corresponding percentages are 47 percent and 50 percent; and (d) even with these increases these countries, whose performance carries a large weight in the total, would still have in 2010 (simple) average yields around one-half those projected for the countries in the top decile. Thus, although the gap may narrow, particularly for rice, it would be the result of the more limited scope for yield growth of the countries in the top decile, not because the large countries with middle yields are projected to have higher growth than in the past.

The preceding rather lengthy discussion was considered necessary in order to provide the reader with sufficient material for thinking about the issue of the potential for further growth in yields to underpin the growth of production. This issue is discussed further in Section 4.4. No attempt is made here to translate these projected yields into concrete proposals for agricultural research (magnitude, modalities, priorities). No doubt, further growth in yields, even at the more modest rates projected for the future compared with the past, will not come about unless the research effort continues unabated. It is just that the effects of research on production growth may manifest themselves in different ways: more impact through the results of evolutionary, adaptive and maintenance research and less through achievement of quantum jumps in yield ceilings.

Evaluating the area-yield projections

The approach used in this study to project harvested areas and yields for the different crops is essentially the same one used in 1985 86 to make the projections to 2000 of the 1987 edition of this study (Alexandratos, 1988). Being now halfway into the projection period 1982/84 2000 (the three-year average 1982/84 was the base year of the 1987 study), it is interesting to examine how well those projections fared when confronted with the actual outcomes up to 1992, the latest year with firm data. Relevant comparisons for the major cereals of the developing countries (excluding China, for which, as in this study, no area-yield projections were made then) are presented in Figures 4.1 and 4.2. By and large, the projections tend to overpredict harvested area for all three cereals shown and underpredict yields for wheat and rice, while maize yields followed closely the predicted path. The net result of these pluses and minuses is that the aggregate production for all three cereals (with rice in milled form) is right on track, with the actual production of 462 million tonnes of the three-year average 1990/92 being equal to that obtained by interpolating areas and yields for each of the three cereals for 1991 on their projection trajectories 1982/84-2000. Similar comparisons for the other coarse grains (16 percent of the total cereals production of the developing countries, excluding China) cannot be made because subsequent revisions changed radically the base year (1982/84) data which had formed the basis for the projections, i.e. the area data have been revised from 91.9 million ha to 84.6 million ha and those for yields from 860 kg/ha to 933 kg/ha.

Contents - Previous - Next