Previous Page Table of Contents Next Page

Parameters for weed-crop competition - M. Sattin and A. Berti


In most developing countries, agriculture employs more than three-quarters of the labour force and provides a major source of GDP with a share of 35-40 percent (Maskey, 1997). A crucial policy issue is how to raise the income of resource-poor farmers without despoiling the natural resource base. A holistic approach based on Integrated Pest Management (IPM), as well as on sound economic principles, would provide a useful framework to protect both resources and farmers’ income.

Since its adoption, IPM (Integrated Pest Management) and its component Integrated Weed Management (IWM), has become the basis for all FAO plant protection activities because it contributes directly towards the achievement of sustainable agriculture in developing countries (Labrada and Parker, 1994).

It is worth noting that IWM presents some important differences in relation to other IPM sectors:

1. The weed flora usually includes several species that contemporarily infest the same plot/field so, in practice, it is normally necessary to estimate the overall loss caused by the whole set of species instead of the effect of a single species. However, situations do exist in which the infestation is monospecific, or can be considered as such in relation to the competitive results (e.g. wild oat infestations in wheat, barnyardgrass in rice, gramineae in dicotyledon crops already treated in pre- or post-emergence with herbicides for controlling broad-leaved weeds).

2. Weeds have a strong periodicity, with plants co-existing at different stages of development; each species and each development stage within a species has a different impact and different sensitivity to the control measures, especially chemical weed control.

3. Herbicides do not generally control a single species but more than one, each with a different level of efficacy; although herbicides with a highly specific action spectrum exist. It is therefore not possible merely to extend to weed control the approach used for some time in the entomology sector, where each insect is controlled with specific products.

To implement an IWM strategy successfully, weed management should match the specific problems in a field and therefore some basic knowledge on weed and crop ecology and biology is needed to correctly predict the impact of a weed infestation on crop yield. Within this context, weed-crop growth characteristics and the dynamics of weed emergence are important aspects (Akobundu, 1998; Forcella, 1998). Many farmers in the developing world are unaware of several aspects of weed interference and the best time for weed removal (Akobundu, 1998; Labrada, 1996 and 1998), although there are exceptions. Ellis-Jones et al. (1993) found widespread recognition in Zimbabwe of the importance of early weeding for both weed suppression and improvement of rainfall infiltration.

Weed germination patterns generally result in cohorts of seedlings emerging over an extended period of time and are heavily influenced by weather conditions, soil type and cropping system (Vleeshouwers, 1997). The initial emergence time differs from year to year and varies according to the species ecological requirements (mainly temperature and soil moisture content - Forcella et al. 1997). It is also well established, and has been experimentally quantified for several crops and types of weed infestation (Zimdahl, 1988; Berti et al. 1996), that the relative time of crop-weed emergence and the time of weed removal strongly ifluence crop production.

On small-scale farms in developing countries more than 50 percent of labour time is devoted to weeding, and is mainly done by the women and children in the farmer’s family (Ellis-Jones et al. 1993; Akobundu, 1996). In traditional farming systems knowledge of the so-called ‘critical period’ of competition would enable farmers to make the most efficient use of limited labour resources. In conditions of medium-high weed pressure, the critical period is approximately centred on the first one-third of the crop growing cycle. For example, several major arable crops of temperate climates (e.g. maize, soybean, sunflower) take 100-140 days after emergence (DAE) to mature and the critical period is often between 25-40 DAE (Zimdahl, 1988; Doll, 1994). Of course, the critical period varies with the relative competitivity of the crop and the weeds: the lower is the crop competitivity (and/or the higher the weed flora competitivity) the longer the period the crop must be kept weed-free to prevent significant yield losses.


The presence of weeds in a crop leads to an increased number of plants within a certain area. Given that the crop density is already set at a level that optimises yield for that cultivar in that environment, the presence of weeds will lead to a reduction in the average yield of the crop.

In an infested field it is possible to identify different components of the overall competitive effect:


Assuming that crop density does not vary significantly, i.e. that the effect of intraspecific competition between plants of the cultivated species can be considered as constant, a simplified situation can be analysed: the effect of a monospecific infestation on crop yield. The yield loss caused by the weeds can be predicted with an acceptable level of precision through the use of mathematical models. These can be empirical or mechanistic. Empirical models are based on mathematical empirical relationships between some important independent variables (e.g. weed density or cover, growth rates, time of weed emergence in relation to that of the crop) and a dependent variable (usually crop yield) the variations of which should be interpreted and predicted. Mechanistic models are instead based on the growth processes (i.e. light interception, photosynthesis, dry matter partitioning between different plant parts of the competing plants (crop and weeds)); they are obviously much more complex than the former and require a good knowledge of the mechanisms and interactions involved in the crop-weed system and a large amount of data or information as inputs. These models have a limited direct practical use, but are good study and research tools from which much simpler empirical models can be derived.

In this chapter, only empirical models will be considered.

Many empirical models have been developed to describe, and possibly predict, the effect of weeds in crops, many of which are based on the relationship between crop yield loss and weed density (Cousens, 1985b). In an attempt to solve the problems linked to predicting yield loss in relation to weed time of emergence, alternative models have been proposed based on the relative leaf area of the weeds and crop.

Prediction of yield loss on the basis of weed density

The model most widely used to describe the yield loss depending on the weed density is based on the rectangular hyperbola model (Cousens, 1985a):


where YL is the relative yield loss, D is the weed density, i is a parameter that represents the initial slope of the curve and a represents the maximum yield loss found with a very high weed density (Figure 1).

Figure 1. Rectangular hyperbola (from Cousens, 1985a) that links the relative yield loss to the density of a weed species.

Both parameters also vary with the crop-weed association as well as with crop density, time of emergence of the weed and crop and soil fertility. The values of i and a can therefore be used to compare various crop-weed associations in additive competition experiments (i.e. when the weed is added to a crop sown at a fixed density).

Equation (1) provides the relative yield loss, that cannot be measured directly, but only calculated starting from the yields observed with and without weeds. However, even the measurements taken in weed-free control plots can be affected by experimental error, so equation (1) can be adapted to include YWF among the equation parameters:


the third parameter YWF represents the yield of the weed-free control, which can thus be estimated using all the observed data and not just those from the control plots.

Equations (1) and (2) assume that the weeds emerge contemporarily with the crop, but in practice this does not happen. Cousens et al. (1987) therefore introduced a modification to the above-described model to take into account contemporarily the weed density and relative time of emergence of the crop and weeds:


where YL, D, i and a have the same meaning as in (1), te is the relative time of emergence of the crop and weeds and c is the regression parameter that expresses the variation of the competitiveness of the weeds depending on the delay with which they emerge.

Model (3) is an improvement on (1) as it allows the experimental data to be better described when the emergence is not contemporary with that of the crop, but with very gradual emergences the determination of te becomes difficult and time-consuming and the estimate of the parameters arguable. Moreover, the estimate of parameters i and a, being dependent also on that of parameter te, generally show a higher variability over the years and with locations. This drawback can, at least partly, be obviated using the temperature sum instead of days as a ‘biological scale’.

Weed density is the variable most commonly used to explain the variations in crop yield loss. This has clear advantages, such as: a) simplicity of its control in experiments; b) determination in the field is relatively easy and fast. It also shows some disadvantages: a) in the field it is difficult to control the time of emergence of the weeds in relation to that of the crop, particularly when emergence occurs in flushes; b) the models based on it have weak eco-physiological bases.

For these reasons some authors have tried to identify another variable that would at the same time be easy to measure, take into account weed competitiveness determined by different times of emergence of crop and weeds, and be eco-physiologically sound.

Prediction of yield loss based on leaf area or relative weed cover

Kropff and Spitters (1991) proposed a relationship based on the relative leaf area (Lw) defined as the ratio between the leaf area index (LAI) of the weed and the total LAI of the crop plus weeds:


where LAIw and LAIc are the leaf area index of the weed and crop, respectively. Lw can vary from 0 (absence of weeds) to 1 (leaf cover of the weed alone). The authors, starting with the processes that regulate crop growth and the results of many simulations done with an eco-physiological competition model, demonstrate that the relative leaf area of the weeds compared to that of the crop at the moment of “row closure” could be a crucial measurement of the competitive process and very well correlated with the final yield loss.

The relationship between relative yield loss YL and Lw is expressed by:


where q is an index of competitivity typical of a given weed in a crop, called relative damage coefficient.

Kropff et al. (1995) modified equation (5) by inserting a parameter m that represents the maximum damage caused by the weeds (asymptote of the hyperbole):


Model (6) is comparable to (1) with relative leaf area instead of density.

Considering Lw instead of density takes into account, at least indirectly, that the damage caused by the weeds depends on the relative development of crop and weeds and therefore also their relative time of emergence. A certain value of Lw can be given by a few early-emerging plants or by many late-emerging ones.

The major problem of this approach is the difficulty in measuring Lw quickly and reliably. In fact, the precise measurement of leaf area is possible with destructive sampling of the vegetation (weeds and crop), but this method is neither rapid nor economical. An alternative is a visual survey, with the operator estimating the leaf area ratios between crop and weeds. With adequate training it is probably possible to obtain good results, but a level of subjectivity in the evaluations of Lw cannot be eliminated.

In place of Lw, the competitivity of an infestation can be evaluated on the basis of the partitioning of the upper layer of plant cover between crop and weeds (Relative Cover, RC). Observing the canopy vertically from above, the upper part of the plant cover will be formed by leaves of the crop and weeds: the ratio between the ground area covered by leaves of weeds and the total cover (i.e. weeds plus crop) represents the RC.

With W and C being the areas covered by weeds and crop, respectively, RC is given by:


The higher this value is, the greater the share of solar radiation intercepted by the weeds will be, and therefore the competition caused by them will be more intense. This method assumes that: 1) interference for light is a measure of interference by all mechanisms: the leaf canopy may serve as an ‘integrator’ of the combined effects of competition for light, water and nutrients, and possibly also allelopathic effects, since these all reduce height, shoot weight and therefore leaf area and radiation interception; 2) the competitive effect of weeds that are shorter than the crop at canopy closure is negligible; in other words, only the plants that are able to overgrow or, at least, reach a height similar to the crop can successfully compete. The use of this variable was proposed with crops of medium-low height (e.g. peas, soybean) (Berti and Sattin, 1996): under these conditions, the use of the RC was shown to be valid. The applicability of this method still has to be evaluated with more widely spaced crops and on taller ones, such as maize and sunflower.

The ratio between RC and crop yield loss is similar to (6):


Both Lw and RC proved to be good yield loss predictors at canopy closure stage (Kropff, 1988; Pike et al. 1990; Lutman, 1992; Berti and Sattin, 1996), which is far too late for any control treatment. The early estimation of these yield loss predictors involves estimating their evolution from the time of assessment to canopy closure. This represents a further source of variability that has to be determined by means of appropriate experiments.

The main advantage of RC is the easier measurement. While Lw is based on LAI ratios and therefore requires all the leaf areas to be determined, RC requires the measurement of the ratios of cover between crop and weeds by observing the canopy vertically from above. This can either be done by means of a subjective visual estimate or, with greater precision, by measurements starting from photographs taken from 2-3 m above the crop. The possibility of using optical equipment that automatically take this measurement can also be foreseen.

It should be stressed that the basic data used for the determination of Lw and RC do not give any indication on weed flora composition, information that is fundamental for a correct choice of type of control (e.g. herbicide).


Information on single crop-weed associations, although interesting, is of limited practical interest because infestations normally include various species and the control measures, in particular herbicides, have a well-defined spectrum of efficacy. The choices between control options (if and how to treat) must therefore be made, taking into account the damage that would be caused in the absence of control as well as the damage caused by the weeds that survive a given treatment. Both these evaluations require the competitive effect of a mixed infestation to be estimated.

In the case of the ratios between density and yield loss, the commonly used methods are based on the transformation of the observed densities into values that can be considered additive. It is obvious that two different weeds, even at the same density, will usually cause different yield losses: it is not therefore possible to directly sum up the observed density values to estimate the competitive effect of the infestation as a whole.

There are three main approaches: the first, proposed by Wilkerson et al. (1991) involves calculating the total competitive load (TCL). This method is used in the HERB program for the choice of post-emergence control options in maize and soybean. The second approach is based on the Density equivalent concept (Deq) (Berti and Zanin, 1994) and is the basis of a decision support system (GESTINF) adapted to Italian conditions and currently at the experimental stage at farm level. The third refers to relative cover and is still at the experimental stage.

Total competitive load method

The competitive ability of the various weed species is assigned through competition experiments by using an indexing method, i.e. assigning an arbitrary value K to the most competitive species and ranking the others according their relative competitivity with the reference species. Indexing is based on linear relationships between crop yield or biomass and weed density, which, for low density values, represent a good approximation of the hyperbolic relationship that exists between these two variables.

For low values of weed density (D), the crop yield can be given by:

Y = a + b D


The ratio b/a represents an index of the competitive ability of the considered species. A competitive index (CI) can therefore be defined for the ith species, obtained from the following:


where br and ar are the regression parameters of the reference species.

K is a scale factor that can assume any value; the authors have chosen a value of 10. In this case, the weeds are ranked following a decimal scale with CI=10 for the reference species. CIs are the basic data used for the evaluation of the yield loss in a real field situation. Then, multiplying the density observed of the ith species by its CI, the competitive load (CL) for this species is obtained:

CLi = CIi Di


The sum of these values for the various species present represents the TCL of that particular infestation:


Crop yield loss can be estimated on the basis of the TCL of the infestation. According to Wilkerson et al. (1991) the yield loss for low infestation levels can be calculated with a linear relationship. With the increase of the CL, the weeds start to interfere with one another as well as with the crop and the competitive effect caused by each single plant decreases. Under these conditions the crop yield loss follows a hyperbolic trend. In the case of soybean, Wilkerson et al. (1991) fixed the passing from the linear relationship to the hyperbolic one at a TCL of 50. It is worth mentioning that this model was calibrated for the central-southern United States, where soybean was grown at wide interrow spacing.

The complete expression of the yield loss as a function of TCL will therefore be:


Density equivalent method

The Deq of a given weed species is defined as the density of a reference species that determines a yield loss equal to that caused by the studied species at the measured density.

The crop yield loss in competition with the reference species is:


While for the ith species present you have:


From the above definition, the Deq of the ith species (Deqi) is the value of the density of the reference species that makes the two preceding equations equal:


A series of algebraic steps leads to:


The equation can be simplified choosing a hypothetical species as reference with the i and a parameters both equal to 1. This assumption gives:


Adding up the Deq of the various species present gives a total Density equivalent (Deqt):


The crop yield loss will then be obtained from:


Knowing the indexes i and a of each weed species it is possible to calculate the damage that can be caused by any combination of these weeds.

Relative leaf area and relative weed cover methods.

If the competitivity of the infestation is evaluated using the Lw instead of density, the equation that expresses the effect of a mixed population becomes simpler. The formula that is applied for a single species can in fact be developed in an additive way, giving the following:


where qi, mi and Lwi indicate the relative damage coefficient, the asymptotic yield loss and observed value of Lw for the ith species, respectively. Where a maximum asymptotic value of yield loss is not considered, the value of mi is equal to 1 for all the species.

When using RC instead of Lw as a descriptive variable for weed-crop competition, the extension to a multi-species situation is straightforward. RC per se measures the ratio between the horizontal projections of crop and weed leaves, so integrating the effects of both leaf area and leaf posture of different weed species. This implies that RC is intrinsically a multi-species descriptor. Measurements taken in the field can therefore be directly used (see eq. 8) for estimating yield loss caused by a mixed weed infestation.


Given the importance of early growth (Sattin and Sartorato, 1997), the trend of the relationships between yield loss and time of emergence and removal of weeds can easily be understood. The competitive effect of a given density of weeds emerging with the crop depends strongly on the length of the period they remain in the field (i.e. the time of weed removal). The relationship between the duration of competition and crop yield reduction is approximately sigmoidal: weeds competing for a short period have little effect on crop yield; allowing the weeds to compete for a longer time, the yield reduction increases, until a plateau is reached corresponding to the yield loss caused by weeds competing over the entire growing cycle. Crops like maize and soybean show a relatively long initial period when the damage caused by weeds is relatively low, while most horticultural crops are more sensitive, as shown in Figure 2.

Figure 2. A: Crop relative yield, as a percentage of the weed-free test, as a function of duration of competition (i.e. time of weed removal) for a weed density of 10 plants m-2. 1 = maize in competition with Abutilon theophrasti (Sattin et al. 1992); 2 = soybean in competition with Amaranthus cruentus (Berti et al. 1990); 3 = onions in competition with Helianthus annuus (Dunan et al. 1995).

Figure 2. B: Soybean relative yield, as a percentage of the weed-free test, in relation to time of emergence, time of removal and density of Amaranthus cruentus (Berti et al. 1990).

The relationship between the time of emergence of weeds and crop yield loss mirrors the curves in Figure 2.A. Of course these relationships are influenced by weed density: for a specific weed time of emergence or removal, the higher the density the lower is the relative yield (Figure 2.B).

Extending the above concepts to a mixed-weed infestation that can be found in any cropped field, the yield can be expressed as a function of maximum yield of the crop kept weed-free, weed-competitive load and time of emergence and removal of the weeds. The first two factors are clearly site-specific, while contrasting data have been reported for the variability of the weed-free period (WFP) and duration of tolerated competition (DTC) curves for different years and/or locations. The experiments designed to obtain this sort of information are tedious and very expensive. An important question is therefore how variable are these data in space and time. Using three data sets showing the effects of mixed weed infestations in maize, soybean and durum wheat and by means of the Deq approach (Berti and Zanin, 1994), Sattin et al. (1996) analysed the variability of DTC and WFP curves over reasonably homogeneous areas in temperate environments. Despite the differences in weed flora among the experiments within each data set, the pattern of these relationships appeared to depend more on crop characteristics than on the composition of the weed infestation. If these results are confirmed for other areas and crops, for a given area the prediction of yield loss caused by a mixed-weed infestation with any time of emergence and/or removal will require the knowledge of the weed-free yield, the relationship weed competitive load-yield loss and only one set of parameters (i.e. only one large experiment) linked to the effects of the time of weed emergence and removal. Information on weed competitive load-yield loss relationships already exists for several sites and crops and, if unavailable, it is much less tedious to obtain compared to the determination of the relationships between DTC, WFP, weed density and crop yield loss.


The critical period has been defined as the period during which weeds must be controlled to prevent yield losses. Since the concept of critical period was introduced, it has been used to determine the period when control operations should be carried out to minimise yield losses for many crops (Zimdahl, 1988). Historically, critical periods have been calculated by mean separations (hereafter referred to as the classical approach) in experiments that evaluated the impact of time of weed emergence and time of removal on crop yields. Using the classical approach, it is possible to identify a period within which no statistically detectable yield losses occur. It has also been concluded that for most field crops it is unnecessary to control weeds in the first few weeks after crop and weed emergence (Zimdahl, 1988).

Several problems inherent to the classical approach have been pointed out and the use of regression analysis (hereafter referred to as the functional approach) was suggested as a better alternative. In changing from the classical to functional approach, the existence of a period when weeds do not cause any yield reduction became doubtful because of the continuous relationship between yield loss and time of weed emergence and removal. To avoid this problem, fixed yield loss thresholds were used to define critical periods (e.g. Van Acker et al. 1993). Within this framework it was concluded that early weed control is unnecessary (Hall et al. 1992). This conclusion was not the result of an accurate evaluation of when to start to control weeds, but of the way critical periods were calculated. The functional approach fails to recognise that since there is a continuous relationship between crop yield and time of weed removal, controlling weeds pre-plant or pre-emergence or post-emergence cannot be compared without considering yield losses that occur between planting and post-emergence control.

The establishment of a fixed yield loss threshold indirectly considers the economic aspect in the calculation of critical periods. Within this framework, Dunan et al. (1995) developed an economic approach to calculate critical period. They defined the economic critical period as the time interval when the marginal income of weed control is higher than the cost of control, and its limits are called early and late economic period thresholds.

Figure 3. Theoretical representation of the concept of optimum time of application for post-emergence treatments.

(A) Hypothetical pattern of weed emergence as a function of time after crop planting.

(B) Economic loss as a result of weeds emerging before the post-emergence treatment (ELB).

(C) Economic loss as a result of weeds emerging after the post-emergence treatment (ELA); and

(D) Total economic loss (TEL) as a function of time of control. Early and late represent two different times of post-emergence weed control and it is assumed that the efficacy of post-emergence treatment is 100 percent.

The period when weed control can be carried out starts before sowing with a pre-plant weed control tactic and continues until the phenological stage of the crop, which precludes any further weed control. Within this time lapse the three approaches (classical, functional and economic) all define a period when weeds should be controlled.

However, identification of a period alone does not provide information on how many or when control practices should be performed. Considering the time dependence of the weed competitive effect, treatments carried out within this period cannot be regarded as having the same net margin (i.e. difference between the value of the crop with control minus the cost of treatment and the value of the crop without control) and therefore an optimum time for one or more weed control tactics exists which provides the highest net margin (Figure 3).

Weed emergence can begin from the moment of final seedbed preparation and then continue while environmental conditions are favourable for the germination process. As previously shown, the time of weed emergence greatly affects weed competitiveness, with the first emerging weeds being far more competitive than the late emerging ones. Most post-emergence treatments have low or nil residual activity. Because of this they can control the weed population present at the moment of application, but have little or no effect on subsequent germinations. The economic result of post-emergence treatments will vary according to the time of application and is related to the efficacy of the treatment, crop characteristics (WFP and DTC curves) and weed germination pattern.

As a first approximation, crop yield loss is given by the sum of the yield loss caused by weeds emerging before the treatment (competing with the crop until the treatment) and that caused by weeds emerging after the treatment (competing from emergence until harvest). These two yield losses follow opposite trends (Fig. 3). With an early treatment, the loss caused by the weeds emerged before the spraying is low because the duration of competition is short. On the other hand, a consistent number of weeds can germinate after the treatment and, remaining until harvest, can produce an important yield loss. With a late treatment, damage caused by weeds emerging after the treatment is reduced, but there is a marked increase in the damage as a result of the weeds emerging before the treatment. The sum of these two losses gives a total economic loss curve characterised by a minimum which identifies the optimum time of treatment (Berti et al. 1996).

Figure 4. Relationships between time of application of a post-emergence treatment and net margin for maize and soybean. Maize: weed-free yield = 10 t ha-1; cost of post treatment = 45$ ha-1; grain price = 155$ t-1. Soybean: weed-free yield = 4 t ha-1; cost of post treatment = 45$ ha-1; grain price = 250$ t-1. For both crops a weed infestation capable of causing a yield loss of 50 percent if untreated was considered.

Figure 4 shows an example of these calculations for maize in the Po valley - Italy and for soybean in southern Ontario - Canada (Sattin et al. 1996). The curves were calculated using the Time Density Equivalent approach (TDE - Berti et al. 1996) and the DTC and WFP relationships calculated by Sattin et al. (1996) were used. With the TDE approach, the weed population is divided in daily cohorts depending on their time of emergence. For each cohort the TDE is equal to the density of a cohort of weeds emerging with the crop and competing until harvest, giving the same yield loss as the considered cohort. TDEs can be considered additive and their sum is a Total Time Density Equivalent (TTDE), which gives a good estimate of the competitivity of the weed population depending on the emergence pattern and, considering herbicide efficacy, on the time of treatment. The relationship between time of weed control and net margin depends on the pattern of weed emergence and on the competitiveness of the crop.


The approaches presented above would allow to decide if, how and when to treat. This obviously requires some knowledge on the quantitative and qualitative composition of the weed flora. Depending on the approach used, different measurements should be taken (i.e. weed density, weed and crop LAI, relative cover). This phase of scouting is crucial: it is clear that the higher the level of precision the better it is, being the basis for subsequent calculations and for the selection of the weed control option to be adopted; on the other hand, scouting requires time and represents a cost, decreasing the net margin of weed control. Furthermore, weeds usually have a patchy distribution in fields and this requires the sampling techniques to account for this. In the case of weed counts, Berti et al. (1992) proposed a relatively easy sampling method. Scouting is done by throwing a metal frame (25 × 30 cm) at random: the plants present within the rectangle are counted and classified by species. On the basis of previous studies (Berti et al. 1992), it has been verified that 20-30 throws are enough to guarantee a good estimate of weed density, assuming that the infestation is relatively homogeneous; if not sub-samples should be taken. The time required for this type of scouting is around 20 minutes per hectare when it is done by well-trained personnel, but can be more than doubled with less skilled people (Berti et al. 2002). The time requirement and the need for specialised personnel can become a major problem in the application of Decision Support Systems (DSS) in the field, especially in developing countries.

Other approaches have been proposed based on visual estimates of weed pressure. In the United States, Harvey and Wagner (1994) used a visual estimate of weed pressure for estimating crop yield loss. Weed pressure is defined as the visual estimate of the percentage which weeds contribute to the total volume of both crop and weeds in a plot, with a value of 0 indicating the complete absence of weeds and 100 indicating the complete absence of crop plants. A common problem with this type of approach is the high variability of ratings assigned by different personnel for the same level of infestation, leading to an uncertainty of yield loss estimates, particularly with low levels of infestation, which represents, on the other hand, the situation where the use of a DSS is more useful. A further constraint is that the visual estimation of weed pressure is easily feasible when both crop and weeds have already reached an advanced stage of development, while it is very difficult to give a correct rating when plants are at the seedling stage. This limits the usefulness of this approach for the selection of present-year weed-control measures, while it might be used to predict future crop yield losses in the same field in order to assist in making sound weed-management decisions for the following year (Harvey and Wagner, 1994).

Therefore, at the present time the use of DSS for selecting weed-control measures faces a paradox: They are especially useful to less skilled farmers, because they can greatly improve their management techniques, while more experienced people can normally select the ‘right’ control measure based on their own experience. On the other hand, to be done rapidly and reliably, thus not determining an excessive increase of control costs, a proper assessment of weed competitive pressure requires a relatively high level of knowledge.


The use of models relating weed density or other independent variables expressing weed competitivity can greatly improve the weed control selection procedure, with a great impact on both yield and economics. However, applications of DSS derived from these competition models are, so far, relatively limited, mainly because the use of DSS requires a relatively high level of knowledge to be really effective.

The wide literature on this subject shows that the beginning of the growing cycle is crucial in determining the intensity and outcome of the subsequent weed-crop competition. It also appears that in hotter climates where the crop growing cycle is often shorter, as is found in many developing countries, the importance of early weed-crop competition is accentuated (Mohamed et al. 1997). The relative time of weed and crop emergence appears to be crucial and can overwhelm many other factors: a few days’ difference in emergence can create an unrecoverable gap between plants, and therefore any agronomic practice which delays weed emergence and favours a satisfactory crop establishment plays an important role. There are indications that the variability (in space and time) of yield loss caused by mixed weed infestations can be relatively low. Therefore, if these results are confirmed for different areas, for various crops and for a reasonably homogeneous area, only a few low-tech experiments, although labour-intensive, are required to predict yield loss in relation to WFP, DTC and weed density.

This type of information could give useful indications and provide a framework particularly in situations where the technological level is low (i.e. where hand-weeding is prevalent), and then where simple rules directing weed control are more likely to be adopted instead of more complicated computer-based technologies. It would therefore be very useful to have, at least for the most important agricultural areas and crops, some basic data sets related to weed-crop competitivity, timing of weed competition and weather conditions.


Akobundu, O. 1996. Principles and prospects for integrated weed management in developing countries. Proc. of the Second Int. Weed Control Congress, Copenhagen. pp. 591-600.

Akobundu, O. 1998. Basic elements for improved weed management in the developing world. In Report of the Expert Consultation on Weed Ecology and Management. pp. 93-101. FAO, Rome.

Berti, A. & Zanin, G. 1994. Density equivalent: a method for forecasting yield loss caused by mixed weed populations. Weed Res. 34: 327-332.

Berti, A. & Sattin, M. 1996. Effect of weed position on yield loss in soybean and a comparison between relative weed cover and other regression models. Weed Res. 36: 249-258.

Berti A., Sattin, M. & Zanin, G. 1990. Soybean-Amaranthus cruentus L.: effects of the period of duration of competition. Proc. First Congress of the European Society of Agronomy, Paris, Session 5: P01.

Berti A., Bravin, F. & Zanin G. 2002. Application of a farm decision-support system for post- emergence weed control. Weed Sci. (in press).

Berti A., Dunan, C.M., Sattin, M., Zanin, G. & Westra, P. 1996. A new approach to determine when to control weeds. Weed Sci. 44: 496-503.

Berti, A., Zanin, G., Baldoni, G., Grignani, C., Mazzoncini, M., Montemurro, P., Tei, F., Vazzana, C. & Viggiani, P. 1992. Frequency distribution of weed counts and applicability of a sequential sampling method to integrated weed management. Weed Res. 32: 39-44.

Cousens, R., Brain, P., O’Donovan, J. T. & O’ Sullivan, P. A. 1987. The use of biologically realistic equations to describe the effect of weed density and relative time of emergence on crop yield. Weed Sci. 35: 720-725.

Cousens, R.D. 1985a. A simple model relating yield loss to weed density. Annals Applied Biology 107: 239-252.

Cousens, R.D. 1985b. An empirical model relating crop yield to weed and crop density and a statistical comparison with other models. J. of Agricultural Sci. 105: 513-521.

Doll, J.D. 1994. Dynamics and complexity of weed competition. In Labrada R., Caseley, J.C. & Parker, C., eds. Weed Management for Developing Countries, pp. 29-34. FAO, Rome.

Dunan, C.M., Westra, P., Schweizer, E.E., Lybecker, D. & Moore, F.D. 1995. The concept and application of early economic period threshold: the case of DCPA in onions (Allium cepa). Weed Sci. 43: 634-639.

Ellis-Jones J., Twomlow, S., Willcocks, T., Riches, C., Dhliwayo, H. & Mudhara, M. 1993. Conservation tillage/weed control systems for communal farming areas in semi-arid Zimbabwe. Brighton Crop Protection Conference - Weeds, Vol. 3: 1161-1166.

Forcella F. 1998. Application of weed seed bank ecology to weed management. In Report of the expert consultation on weed ecology and management. pp. 23-35. FAO, Rome.

Forcella, F., Wilson, R.G., Dekker, J., Kremer, R.J., Cardina, J., Anderson, R.L., Alm D., Renner, K.A., Harvey, R.G., Clay, S. & Buhler, D.D. 1997. Weed seedbank emergence across the corn belt. Weed Sci. 45: 47-76.

Hall, M.R., Swanton, C.J. & Anderson, G. 1992. The critical period of weed control in grain corn (Zea mays). Weed Sci. 40: 441-447.

Harvey, R.G. & Wagner, C.R. 1994. Using estimates of weed pressure to establish crop yield loss equations. Weed Tech. 8: 114-118.

Kropff, M.J., Lotz, L.A.P., Weaver, S.E., Bos, H.J., Wallinga, J. & Migo, T. 1995. A two-parameter model for prediction of crop loss by weed competition from early observations of relative area of weeds. Annals of Applied Biology 126: 329-346.

Kropff, M.J. 1988. Modelling the effects of weeds on crop production. Weed Res.28: 465-71 Kropff, M.J. and Spitters C.J.T. 1991. A simple model of crop loss by weed competition from early observations on relative area of the weeds. Weed Res.31: 97-105.

Labrada R. 1996. Weed management status in developing countries. Proc. of the Second Int. Weed Control Congress, Copenhagen. pp. 579-589.

Labrada R. 1998. Problems related to the development of weed management in the developing world. In Report of the Expert Consultation on Weed Ecology and Management. pp.7-12. FAO, Rome.

Labrada R. & Parker C. 1994. Weed control in the context of integrated pest management. In Labrada, R., Caseley, J.C. and Parker, C., eds. Weed Management for Developing Countries. pp. 3-26. FAO, Rome.

Lutman, P.J.W. 1992. Prediction of the competitive effects of weeds on the yields of several spring-sown arable crops. IXème Colloque Int. sur la Biologie des Mauvaises Herbes, Dijon, France: ANPP. pp. 337-45.

Maskey, R.K. 1997. Sustainable agricultural development in less developed countries. Outlook on Agriculture 26: 39-45.

Mohamed, E.S., Nourai, A.H., Mohamed, G.E., Mohamed, M.I. & Saxena, M.C. 1997. Weeds and weed management in irrigated lentil in northern Sudan. Weed Res. 37: 211-218.

Pike D.R., Stoller, E.W. & Wax L.M. 1990. Modeling soybean growth and canopy apportionment in weed-soybean (Glycine max) competition. Weed Sci. 38: 522-7.

Sattin, M, Zanin, G. & Berti, A. 1992. Case history for weed competition/population ecology: velvetleaf (Abutilon theophrasti) in corn (Zea mays). Weed Tech: 6: 213-219.

Sattin, M. & Sartorato, I. 1997. Role of seedling growth on weed-crop competition. Proc. of 10th EWRS (European Weed Research Society) Symposium, Poznan, Poland. pp. 3-12.

Sattin, M., Berti, A. & Zanin, G. 1996. Crop yield loss in relation to weed time of emergence and removal: analysis of the variability with mixed infestations. Proc. of the Second Int. Weed Control Congress, Copenhagen. pp. 67-72.

Van Acker, R.C., Swanton, C.J. & Weise, S.F. 1993. The critical period of weed control in soybeans. Weed Sci. 41: 194-200. (Ph.D. thesis)

Vleeshouwers, L.M. 1997. Modelling weed emergence pattern. Wageningen Agricultural University, The Netherlands.

Wilkerson, G.G., Modena, S.A. & Coble, H.D., 1991. HERB: Decision model for post-emergence weed control in soybean. Agronomy J. 83: 413-417.

Zimdahl, R. L. 1988. The concept and application of the critical weed-free period. In Altieri, M.A. & Liebmann, M., eds. Weed Management in Agroecosystems: Ecological Approaches. pp. 145-155. CRC Press, Boca Raton. Florida, USA.

Previous Page Top of Page Next Page