Thesis (MScAgric (Agronomy))--Stellenbosch University, 2008. / The development of herbicide resistance in weeds is one of the major factors hampering profitable crop production worldwide. In South Africa resistance to herbicides in weeds is also a big problem, in particular in the Winter Rainfall Region of the country. The lack of sufficient different mode of action herbicide groups that can be rotated in these conditions necessitate the implementation of integrated weed management programmes to curb the development and spread of herbicide resistance. One of the alternative physical weed management strategies is to maximize crop competition to the weed population. One aspect of such a strategy is to plant crop cultivars that have greater interference potential than others.

Multivariate count data are found in a variety of fields. For modeling such data, one may consider the multivariate Poisson distribution. Overdispersion is a problem when modeling the data with the multivariate Poisson distribution. Therefore, in this thesis we propose a new multivariate Poisson hidden Markov model based on the extension of independent multivariate Poisson finite mixture models, as a solution to this problem. This model, which can take into account the spatial nature of weed counts, is applied to weed species counts in an agricultural field. The distribution of counts depends on the underlying sequence of states, which are unobserved or hidden. These hidden states represent the regions where weed counts are relatively homogeneous. Analysis of these data involves the estimation of the number of hidden states, Poisson means and covariances. Parameter estimation is done using a modified EM algorithm for maximum likelihood estimation. <p>We extend the univariate Markov-dependent Poisson finite mixture model to the multivariate Poisson case (bivariate and trivariate) to model counts of two or three species. Also, we contribute to the hidden Markov model research area by developing Splus/R codes for the analysis of the multivariate Poisson hidden Markov model. Splus/R codes are written for the estimation of multivariate Poisson hidden Markov model using the EM algorithm and the forward-backward procedure and the bootstrap estimation of standard errors. The estimated parameters are used to calculate the goodness of fit measures of the models.<p>Results suggest that the multivariate Poisson hidden Markov model, with five states and an independent covariance structure, gives a reasonable fit to this dataset. Since this model deals with overdispersion and spatial information, it will help to get an insight about weed distribution for herbicide applications. This model may lead researchers to find other factors such as soil moisture, fertilizer level, etc., to determine the states, which govern the distribution of the weed counts.

Multivariate count data are found in a variety of fields. For modeling such data, one may consider the multivariate Poisson distribution. Overdispersion is a problem when modeling the data with the multivariate Poisson distribution. Therefore, in this thesis we propose a new multivariate Poisson hidden Markov model based on the extension of independent multivariate Poisson finite mixture models, as a solution to this problem. This model, which can take into account the spatial nature of weed counts, is applied to weed species counts in an agricultural field. The distribution of counts depends on the underlying sequence of states, which are unobserved or hidden. These hidden states represent the regions where weed counts are relatively homogeneous. Analysis of these data involves the estimation of the number of hidden states, Poisson means and covariances. Parameter estimation is done using a modified EM algorithm for maximum likelihood estimation. <p>We extend the univariate Markov-dependent Poisson finite mixture model to the multivariate Poisson case (bivariate and trivariate) to model counts of two or three species. Also, we contribute to the hidden Markov model research area by developing Splus/R codes for the analysis of the multivariate Poisson hidden Markov model. Splus/R codes are written for the estimation of multivariate Poisson hidden Markov model using the EM algorithm and the forward-backward procedure and the bootstrap estimation of standard errors. The estimated parameters are used to calculate the goodness of fit measures of the models.<p>Results suggest that the multivariate Poisson hidden Markov model, with five states and an independent covariance structure, gives a reasonable fit to this dataset. Since this model deals with overdispersion and spatial information, it will help to get an insight about weed distribution for herbicide applications. This model may lead researchers to find other factors such as soil moisture, fertilizer level, etc., to determine the states, which govern the distribution of the weed counts.

No description available.

The reported requirement for a higher weeding effort due to increased weed infestations under conservation agriculture (CA) relative to conventional mouldboard plough tillage is perceived by both smallholder farmers and extension workers as the main limiting factor to the widespread adoption of CA by smallholder farmers in southern Africa. However, proponents of CA argue that weeds are only a problem under CA in the initial two years and decline afterwards resulting in reduced labour requirements for weeding under CA. They further posit that weeds are only major problem where minimum tillage (MT) is adopted without crop residue mulching and diverse crop rotations. This thesis explores the effect of time under CA on weed population dynamics and crop growth under the recommended CA practices and actual smallholder farmer practice in semi-arid Zimbabwe. Assessment of weed and crop growth on a long-term CA experiment at Matopos Research Station revealed that the MT systems of planting basins and ripper tine were associated with higher early season weed density and biomass than conventional early summer mouldboard tillage (CONV) in both the fifth (cowpea phase) and sixth (sorghum phase) years of CA. This increased weed infestation within the first four weeks after planting in CA necessitated early weeding to provide a clean seedbed and avert significant crop yield loss. Maize mulching only suppressed early season weed growth in sorghum mostly at a mulch rate of 8 t ha-1 which is not a mulching rate that is attainable on most smallholder farms. However, the lower maize residue mulch rate of 4 t ha-1 was consistently associated with increased weed emergence and growth as from the middle of the cropping season in both crop species. The increased weed infestations under the mulch were probably due to the creation of ‘safe sites’ with moist conditions and moderate temperatures. The high weed growth under the mulch contributed to the low sorghum grain yield obtained under mulched plots. In addition, maize mulching was also associated with a less diverse weed community that was dominated by the competitive Setaria spp. and difficult to hoe weed Eleusine indica (L.) Gaertn. However, the weed community under CA was similar to that under CONV tillage with no evidence of a shift to the more difficult to control weed species. The increased early season weed growth and high weed pressure under CA meant that it was still necessary to hoe weed four times within the cropping season to reduce weed infestations and improve crop growth even after four years of recommended CA practices. Early and frequent weeding was effective in reducing weed growth of most species including Setaria spp. and E. indica demonstrating that on smallholder farms where labour is available hoe weeding can provide adequate weed control. The wider spacing recommended for use in CA contributed to the low cowpea and sorghum grain yields obtained under CA compared to CONV tillage. On smallholder farms in Masvingo District, the MT system of planting basin (PB) was the only conservation farming (CF) component adopted by farmers. There was no difference in the total seedling density of the soil weed seed bank and density of emerged weeds in the field in PB and conventional mouldboard ploughing done at first effective rains (CONV tillage). However, the first weeding in PB was done at least 15 days earlier (P < 0.05) than in CONV tillage suggesting high early season weed growth in PB relative to CONV tillage. As weed density did not decline with time in PB, weed management did not differ with increase in years under PB. Shortage of inputs such as seed and fertiliser was identified by smallholder farmers as the most limiting factor in PB crop production with the area under PB was equivalent to the seed and fertiliser provided by CARE International for most farmers. On this small area, weeds could be managed by available family labour. Double the maize grain yield was obtained in PB (mean: 2856 kg ha-1) due to improved weed management and soil fertility. However, the use of poorly stored composts was found to introduce weeds into some PB fields. The findings of this study demonstrated that weed pressure was still high and weed management were still a challenge under the practice recommended to smallholder farmers in Zimbabwe even in the sixth year of practice. There is, therefore, a need for research on the economic feasibility of using herbicides, intercropping and optimal crop density to ameliorate the high weed pressure under CA. / Thesis (PhD)--University of Pretoria, 2013. / Plant Production and Soil Science / unrestricted