Results

The frequency with which forecast selected strategies outperformed long-term strategies, and vice versa, over the 1981/1982 to 1992/1993 seasons is presented in Fig. 21.3 for each catchment. The frequency with which the two selected strategies performed equally well, is also shown. The provinces in which the catchments are located are shown at the top of the figure to indicate geographic locality.

The frequency with which forecast selected strategies performed better than long-term strategies ranged from 0 to 75% across the different catchments. The outcome whereby forecast selected strategies performed better than long-term strategies was the most frequently occurring outcome in three catchments, these being located in KwaZulu-Natal province. In contrast, the outcome whereby long-term strategies performed better than forecast selected strategies was the most frequently occurring outcome in two catchments in Mpumalanga province. In the remaining 10 catchments, the outcome whereby forecast selected and long-term strategies performed equally well, was the most frequently occurring outcome or, alternatively, was equal in proportion to the outcome of long-term strategies performing better.

Figure 21.3 indicates the frequency with which a certain strategy performs better than another, but does not give any indication of the extent to which its yields are higher. This was assessed in Fig. 21.4 for cases where the forecast selected strategies yielded more than the long-term strategies. The maximum, mean and minimum differences in yield between the two strategies, over the 12 seasons of simulation, are plotted in the graph for the relevant catchments. The mean differences in yield are also shown in brackets as a percentage, with the relevant number of data points indicated

Fig. 21.3. Frequency with which different crop management strategies performed better over the 1981/1982 to 1992/1993 seasons

3500 3000 2 2 500

Limpopo Mpumalanga North West j J..J J..J i„r

KwaZulu-Natal

Free State

-Eastern-Cape

D41F W22J V40A

C52C C52L T34K

Catchments

▲ Maximum difference

- Mean difference

Minimum difference

Fig. 21.4. Differences in yields obtained from forecast selected and long-term strategies (cases where forecast selected strategies yield more than long-term strategies) over the 1981/1982 to 1992/1993 seasons

Fig. 21.4. Differences in yields obtained from forecast selected and long-term strategies (cases where forecast selected strategies yield more than long-term strategies) over the 1981/1982 to 1992/1993 seasons below. The provinces in which the catchments are located are again shown at the top of the figure. Figure 21.4 shows that the mean differences in yield for the catchments in KwaZulu-Natal (where forecast selected strategies performed better than long-term strategies most frequently) ranged from 28 to 505%. The mean yield difference for catchment W22J, when expressed as a percentage (505%), is inflated as a result of two seasons of crop failure when applying the long-term crop management strategy in this catchment. Nevertheless, the mean yield difference, when expressed in kg ha-1 (670), has the same order of magnitude as the other KwaZulu-Natal catchments, and the forecast selected crop management strategy can be concluded to have performed appreciably better than the long-term strategy in this province. The yield differences in the other provinces are also appreciable, although it should be borne in mind that there were considerably fewer occurrences of the forecast selected management strategy outperforming the long-term strategy (cf. Fig. 21.3).

21.4

Discussion and Recommendations

The usefulness of the crop yield forecasts, as defined by their potential to improve crop management decisions, varied across the catchments assessed, with the greatest forecast usefulness being detected in KwaZulu-Natal province. The gains in yield derived from applying yield forecasts in this province were also shown to be appreciable.

Ideally, climate forecasts should be available for the entire growing season when generating crop yield forecasts. In this study, rainfall forecasts were only available for the December to March period, while many of the crops simulated were growing outside of this period, thus requiring assumptions to be made about the rainfall in these periods. The yield forecasting methodology needs to incorporate current climate fore cast formats (terciles with associated probabilities), and the usefulness of the resulting yield forecasts needs to be assessed. If there are an insufficient number of forecast seasons for this analysis, as was the case in this study, an historical set of climate forecasts could be generated retrospectively. The incorporation of general circulation model (GCM) derived climate forecasts in the yield forecasting methodology also needs to be assessed, as these are becoming more readily available for South Africa. GCM derived climate forecasts may have advantages over forecasts derived from statistical climate models. For example, the finer spatial and temporal scale of modeling in GCMs produces information in a format more suited to application in crop yield models.

Crop forecasts were only produced for maize in the current study. Forecasts could be produced for other crops, which would then allow for crop selection to be included in crop management recommendations. Crop management strategies giving rise to the highest maize yield were selected in the study. In practice, a small-scale/subsistence farmer's objective may not be to maximize yield, but rather to minimize risk. To minimize risk, a farmer could avoid adopting strategies that give rise to a wide range in yields under different seasonal climate conditions, thus minimizing the impact of a forecast being wrong. Farmers apply a variety of management practices to spread the risk of a particular strategy failing. As confidence in the forecasts grows, forecast selected strategies could be applied more extensively.

In practice, many factors influence a farmer's crop management decisions. It is recommended that the application of crop yield forecast information in crop management decisions be assessed in more detailed case studies where these factors can be taken into account. Field data would need to be collected to ensure that the crop model inputs, including the representation of crop management strategies, is realistic. Observed data would also be needed to verify forecast accuracy and usefulness. Greater collaboration with stakeholders would be required to facilitate these case studies. A research project has been proposed involving a number of organizations and individuals, where case studies will be implemented at identified sites in various provinces.

Lumsden and Schulze (2004) reviewed forecast information needs and forecast application constraints in South Africa. These needs include improved forecast quality, more extensive forecast verification, more relevant forecasts to users, forecast dissemination improvements and capacity building. Apart from these deficiencies in the available forecast information, farmers may also be constrained in their ability to respond to forecasts owing to a lack of resources such as draft power, healthy labor in HIV/AIDS affected communities, credit, water, land, fertilizers and favorable markets. The impact of these constraints on forecast application could be better understood in the case studies planned above. Efforts to improve the resources available to farmers need continued attention.

Based on the findings of the Lumsden and Schulze (2004) study, which included a review to determine what forecast information is currently available for South Africa, three potential applications of crop yield forecasts to small-scale/subsistence agriculture were identified for further research and implementation in the country. These applications, which have varying scales and functions, are outlined in Table 21.1.

An example of regional planning where crop yield forecasts might be applied is the coordination of aid to farmers. For regional planning, magisterial districts were suggested in Table 21.1 as an alternative to QCs as the scale at which forecasts could be produced. Although QCs are a convenient scale at which to produce forecasts because

Table 21.1. Potential applications of crop yield forecasts to small-scale/subsistence agriculture identified for further research and implementation in South Africa (Lumsden and Schulze 2004)

Yield forecast application

Scale of yield forecast

Potential users

Representation of soils

Representation of management

(1) Regional planning

Regional -QC,or

- magisterial district

- Government departments

- Sector organisations

Typical soil type and depth

Generic strategy

(2) Regional crop management recommendations

Regional -QC,or

- magisterial district

Extension services

Matrix of common soil types and depths

Regional strategies

(3) Local crop management recommendations

Local

- extension center

- representative farms

Extension services

Actual soil type and depth

Local strategies

of the associated agroclimatic databases available, this scale of forecasting may not be convenient for users. A single typical soil profile and a generic crop management strategy would be used as the influence of different soils and management strategies are of less importance for this application. According to Vogel (2000), the application of forecast information in regional planning may be the most feasible application of forecasts if the constraints faced by farmers in altering their crop management strategies in response to a forecast, are found to be too great. The production and dissemination of crop yield forecasts for regional planning is believed to be the most achievable of the forecast applications identified in Table 21.1.

For the forecast application targeted at regional crop management recommendations, magisterial districts were again suggested as an alternative to QCs as the scale at which forecasts could be produced, for the same reasons discussed above. A matrix of soil types, soil depths and crop management strategies would be represented in these forecasts in an attempt to represent the range of conditions occurring in the region. The regional crop management strategies would be focussed on management practices less subject to local effects, for example, planting dates and crop type selection. The crop yield forecasts produced in the current study would fit into this category of forecast applications. The Directorate of Agricultural Risk Management in the National Department of Agriculture (NDA-ARM) periodically disseminates regional agricultural advisories to farmers via extension services. At present the advisories include simple crop management recommendations based on climate forecasts and reports on current conditions from field workers. Crop yield forecasts like those produced in this study could be used as an additional source of quantitative information in formulating these advisories (Archer, personal communication in 2003; Walker, personal communication in 2003; Lumsden and Schulze 2004).

For the forecast application targeted at local crop management recommendations, agricultural extension centers/offices are suggested as possible sites for forecasting because the relevant extension officers would be familiar with the conditions prevailing at these centers. Alternatively, representative farms in the extension districts could be identified for yield forecasting, as farmers might find it easier to relate the condi tions on these farms to their own farms. The actual soil type and depth would be represented in the forecasts, as would crop management practices that are specific to the area. At this scale of application it might be possible to begin tailoring the recommended management responses to suit the typical livelihoods of households found in the area. The detailed case studies proposed previously would fit into this category of forecast application. While recommendations at this scale would be more applicable to farmers, a greater degree of downscaling of the climate forecast information would be required, which may limit the usefulness of the resulting yield forecasts. If the case studies prove successful, the sites studied could become demonstration sites showing the value of applying forecast information in decision-making. The widespread production of crop yield forecasts at this scale would be a longer term goal because of the research and resources required.

Was this article helpful?

0 0

Post a comment