Weather affects agriculture more than any other variable. For centuries, growers had to depend upon small bits and pieces of local climatological data collected and passed down in almanacs. Over the last 100 years, however, scientists have developed complex Numerical Weather Prediction (NWP) models that are able to forecast weather with increasing accuracy. The objective of this work was to use a probabilistic NWP model (the European Centre for Medium-Range Weather Forecasts (ECMWF) Ensemble Prediction System (EPS)) as a component to couple with agricultural decision-making tools and models. First, customized ECMWF EPS forecasts were used as an irrigation scheduling aid for a field trial. Next, the CROPGRO Cotton Model was used to simulate the field experiment as well as an additional irrigation scheduling strategy. Finally, a cotton canopy temperature model was developed and coupled with customized ECMWF EPS forecasts to generate hourly canopy temperature forecasts. These forecasts were used to create a heat stress warning system. Results from the field trial indicate that using precipitation forecasts to schedule irrigation could provide a convenient alternative relative to a standard method. Results from the simulated field trial suggest using precipitation forecasts issued on the day of irrigation could be more efficient than using forecasts issued one to two days prior. Last, results from the heat stress project indicate forecasts were skillful to 10 days, allowing enough time for growers to protect crops if needed. In light of the above, implications for the agricultural community could be significant. Coupled atmospheric-agricultural models have the ability to put weather forecasts in terms producers can understand and can quickly use to make strategic on-farm decisions, therefore, possessing the potential to make a large positive global impact.

Thesis (M.Sc.(Meteorology))--University of Pretoria, 2008. / Summary in English. Includes bibliographical references.

In order to make informed decisions, we need to resort to various types of information and we need to know how uncertain this information is. A commonly used source for information and subsequent action is weather forecasts. The communication of uncertainty in weather forecasts has been widely studied for English speaking audiences, resulting in a number of guidelines that practitioners can follow. For forecasts aimed at Russian speaking audiences there are very few, if no, such studies. The aim of this study is to extend previous research on the communication of uncertainties in weather forecasts to the Russian-speaking domain. The underlying hypothesis for this study is that it should be possible to distinguish texts from different types of forecasts, with different inherent uncertainty, by analysing the linguistic uncertainty markers in the text-based section of these forecasts. If this is not the case, this could in a first step be solved by applying the recommendations in the available guidelines, in a second step the guidelines themselves might need to be extended to meet the needs of the practitioners. To test the hypothesis, I analyse the expressed linguistic uncertainty in two different sources of meteorological information: weather forecasts and seasonal outlooks. The analysis shows that the original hypothesis can be confirmed: the differences between these two sources of information can be detected by analysing linguistic uncertainty markers. Further, the recommendations from the guidelines were met to a large extent, but both type of forecasts, in particular the seasonal outlooks, would benefit from a more consolidated approach. The analysis also shows that these guidelines could be improved by placing an increased emphasis on text-based forecasts, highlighting which linguistic means should be used for what purpose. The guidelines could be extended with language-specific best-practise examples. This way the guidelines would cater for a much larger user-base than they do at present.