Weeds in a crop field are distributed irregularly, with patches of high density as well as patches with few weeds (Cardina, Johnson & Sparrow, 1997). Spatially these patches may be relatively stable from year to year, a product of localized seed rain, a relatively immobile seed bank, the clonal spread of vege-tatively propagated weeds, and the patchiness of the soil environment (Colbach, Forcella & Johnson, 2000).
Although many weed species may be present in a field, only a limited number are important for crop management (Johnson et al., 1995). Each of these species has a defined life cycle with a relatively defined phenology around which weed and crop management practices are usually organized.
This patchiness of weeds presents difficulties for monitoring and recording weed abundance and composition at the field and farm scale. Weeds reduce crop losses at the scale of individual crop plants. Control practices, while usually applied over the entire field, act against individual weeds. In addition, fields across the agricultural landscape are different in their weed patchiness and composition, due to founder effects, differing patterns of cropping and weed control practices, soil and drainage, and location in the landscape. For example, Johnson et al. (1995) found that a particular weed species will not necessarily have the same degree of clumping in different fields. How to improve farmer planning and decision-making through simpler and more accurate scouting of weed patches, estimation of damage, and extrapolation of weed dynamics is a major challenge to farmers, extensionists, and weed scientists.
In addition to being patchy, the presence of weeds in crop fields and across the agricultural landscape is uncertain. For a specific field in a specific season, when weeds will germinate, how fast they will grow in relation to the crop, how much seed they will produce, and how effective crop growth and weed control practices will be are difficult to predict (Ghersa & Holt, 1995).
First, the individuals of a weed species have a wider range of response to weather conditions than the individuals of a crop planted in the same field. Although each weed species has a relatively defined life cycle, individual weeds within a species show a range of responses to moisture and temperature cues (Chapter 9) (Dekker, 1997).
Second, unpredictable variations in weather during and between seasons affect weed germination and growth, the relative development of the weeds and the crop, and the effectiveness of weed control measures. In a single location in Minnesota, for example, variable weather conditions from 1991 to 1994 included lingering snowpacks, a late cold snap, an exceptionally wet spring, an exceptionally dry spring, and midseason droughts (Forcella et al., 1996). These affected date of soil preparation and planting, date of weed germination, and herbicide effectiveness. Thus, although weed patches may show some stability across years, actual weed density, weed phenology in relation to the crop, and weed seed production may be much more variable than weed patch location and thus harder to predict.
Third, over several cropping seasons, nondirectional random shifts in weed composition due to weather fluctuations and semipredictable directional shifts in weed composition due to cropping patterns occur simultaneously. This interaction contributes an additional dimension of uncertainty to weed management. This may be further complicated by the occasional invasion of new weed species.
Lastly, farmers manage weeds according to different criteria and constraints depending on the year. Small farmers are routinely affected by family illnesses, economic crises, and low crop prices. Large-scale farmers often suffer from machinery breakdown, labor problems, and unexpected changes in the cost of inputs. All these factors can affect the nature and timeliness of weed control measures.
Thus, on the one hand, weeds are sufficiently predictable that farmers can use routine control practices to produce crops without being overrun by weeds. On the other hand, weeds are rarely eliminated altogether, due to the localized mismatch between routine control practices and the uncertainty of weed patchiness. Farmer decision-making in weed management aims to minimize this mismatch for more efficient and less risky crop production. How have advances by researchers and extensionists taken into account weed patchiness and uncertainty for the wide diversity of the world's farmers.?
Since the early 1900s, the routine use of uniformly applied agrichemical inputs on the better croplands has produced impressive increases in yields and labor productivity, first in temperate and later in tropical agriculture. Through multiyear replicated experiments, scientists conducted input-output research to identify the best broadly applied levels and combinations of different inputs, each of which has a specific, short-term purpose. Extensionists and later private crop consultants promoted the use of improved varieties, chemical fertilizers, and pesticides. This simple production model based on the efficient assemblage of purchased inputs into an end product resembles an industrial process (Levins, 1986). In the USA maize production quadrupled from 1940 to 1990 with fewer farmers and less land in production (Hossner & Dibb, 1995). In more recent years, in China rice and wheat yields have doubled and quadrupled, respectively (Hossner & Dibb, 1995).
Recently, science and society have begun to realize that a crop field is not a factory, but rather part of a living and responsive system. Herbicide resistance in weeds, floristic shifts to harder-to-control weeds, ground and surface water pollution, and human health effects are now routinely recognized as part of the risks and external costs of using herbicides and other agrichemicals (Chapter 1).
In many tropical countries, the standardized, high-input approach to increased crop yields has not fit productively with the complex landscapes, incipient infrastructure, and the diverse human cultures and cropping systems of smallholders (Pretty, 1995, pp. 31-3). As a result, input use has generally been irregular and crop yield responses modest and inconsistent. In the countries of Central America, Phaseolus bean yields have not increased consistently during the past 30 years (FAOSTAT, 1999). In Nicaragua, for example, since 1965 bean yields have fluctuated from 0.5 to 0.9 Mg ha \ but the long-term yield has increased only slightly (unpublished Nicaraguan Central Bank files, 1998). Similarly, coffee yields have fluctuated from 0.3 to 0.8 Mg ha ^
The Nicaraguan Coffee Growers' Union (UNICAFE, 1998) found that coffee yields could double with improved management and on better sites could quadruple with higher inputs and better management. During recent years, pesticide poisonings and the expenditure of foreign currency for herbicide imports have also fluctuated, although the fluctuations have not been correlated with crop yields (Beck, 1997).
These efforts in temperate and tropical regions to improve methods ofcrop production, including weed control, point to an important and still unfolding lesson. Progress has been possible, but not without costs and failures. Further progress will depend on the ability of scientists, extensionists, and farmers to work within increasingly complex expectations. At one time the objective was simply higher crop yields. Current goals include higher crop yields, protection of human health and the environment, improvement of soil and water quality, and greater market competitiveness. These factors are represented on the horizontal axis in Figure 3.2 as the increasing ecological, social, and economic complexity affecting crop production. In the case of weed management, this complexity is a product of several factors: changes in the larger social and economic context of agriculture, increasing understanding of weed ecology and crop production, and a need to ameliorate past negative results, such as herbicide resistance, and minimize them in the future.
In confronting this complexity, weed management has progressed from initially simple input-output research (e.g., trials on herbicide rates and cover crop species) through site-specific recommendations (e.g., herbicide rates by soil types) and problem-solving research (e.g., limiting competition between crops and cover crops) to predictive models and decision aids (e.g., Kropff & van Laar, 1993; Forcella et al., 1996). On one hand, these changes can be interpreted as the successive fine tuning of recommendations for broadscale, uniform weed management. This possibility for the vertical axis in Figure 3.2 originates from the perspective that the natural world can be increasingly predicted and controlled. However, these advances can also be seen as first steps in the development of adaptive management (Holling, 1978, pp. 1-21; Roling & Wagemakers, 1998). In conventional modern agriculture, most research aims to provide farmers with general technologies and recommendations suited for average situations. In contrast, the aim of adaptive management is a progressive increase in farmers' ability to develop and adapt a range of technologies for a local fit under variable and uncertain situations. Adaptive management assumes that decisions on weed control are based on less than perfect information, that control measures are not completely effective, and that each crop season provides additional data for the farmer on the development of more effective weed management. This is an appropriate response to the
increasing ecological, social, and economic complexities that confront programs for the improvement of on-farm weed management.
Was this article helpful?