PhD ThesisMaize (Zea mays L.), rice (Oryza sativa), and sorghum (Sorghum bicolor L. Moench) are major staple food crops to the most population in Tanzania. The three crops provide the primary source of livelihood for the majority of rural farming households. Of the three crops, maize is the most important, accounting for about 20% of the total agricultural GDP, followed by rice. Sorghum plays an important role in fighting hunger and food insecurity in central Tanzania, particularly in Dodoma and Singida regions. Unfortunately, like any other crops, some uncertainty exists about the future productivity and profitability of these important food crops. Such uncertainty hinders the implementation of different agricultural policies, plans and strategies set to achieve an agriculture revolution, hence impacting the decision of investment in agricultural technologies. The inadequacy of accurate and timely information on productivity and profitability of crops have a tremendous impact on farmers' decisions, as well as on policy and planning. Hence, a complete model to help in forecasting and economic analyses of crucial crop sub-sectors while including their stochastic nature is essential. In this regard, stochastic risk analysis models were developed and demonstrated to analyse risk and uncertainty in forecasting and economic analyses of major cereal crops in Tanzania. Most of the available models in economic analyses and forecasting yields, prices, and net returns of agricultural systems are deterministic. These models ignore the inherent risk of random variables and provide only a point estimate for the key output variables (KOVs) instead of values with probability distributions. Therefore, this study was conducted to address three specific objectives. The first objective was to develop and demonstrate a stochastic simulation model for analysing the future viability of main cereals crops in semi-arid and sub-humid areas of Tanzania. For this reason, a Maize-Sorghum-Rice Simulation Model (MASORISIM) was developed to simultaneously forecast yields, prices, and probable net returns for maize, sorghum, and rice as probability distributions. It utilizes deviations from historical yields and prices (2008 – 2018) to forecast random variables for seven years from 2019 – 2025. Since the analysis involved yields and prices of three crops, a multivariate probability distribution was built in the model to incorporate correlations of the variables and control their heteroscedasticity. The forecasting results on crop yield show an increasing trend for maize and rice with a marginal increase for sorghum in the Dodoma region by 2025. Likewise, the yield for rice is expected to rise in Morogoro with a slight increase for maize and a decreasing trend for sorghum during the same period. Meanwhile, the prices for the three crops all are projected to increase in the two regions. The results on economic feasibility using NPV values revealed a high probability of success for all crops in both regions except maize in Morogoro. The results for maize in Morogoro presented a 2.93% probability of negative NPV. Of the three crops, maize indicated the highest relative risk associated with NPV for both regions and was relatively higher in Morogoro (55.1%) than in Dodoma (34.2%). Although the results on production indicate increasing trends for the crops, the increase is relatively small, particularly in Morogoro, which is one of the food basket regions in the country. The second specific objective of the study was to develop and illustrate a bio-economic simulation model for analysing the economic feasibility of improved management practices on maize production in the Wami Basin of Tanzania. The bio-economic simulation model is an integrated decision support system (IDSS) developed to link data from two biophysical models, namely APSIM and DSSAT and econometric model (Simetar) for comprehensive decision-making. Under this objective, the economic feasibility of two farm management practices was analysed. These practices included the application of 40 kg N/ha and adjustment of plant population at a rate of 33 000 plants/ha from the current rate of 18 000 to 20 000 plants/ha. The simulated yield from the two crop models was then entered into the bio-economic IDSS model along with output prices, and cost for each option to simulate the probable economic net returns to farmers. The APSIM and DSSAT crop models were used in this study because the two models are capable of simulating yield as a function of the soil-plant-atmosphere conditions with and without the proposed farm management practices. However, crop models normally simulate yields and cannot simulate other variables like prices and costs of management alternatives to inform economic decisions. The bio-economic simulation model, therefore, was built to bridge the gap. The results on the economic viability show that the application of 40 kg N/ha was more profitable than the plant population of 33 000 plants/ha having a zero probability of negative returns. Both APSIM and DSSAT models suggest that when plant population is adjusted from current average of 20,000 plants/ha to 33 000 plants/ha, there is 16% and 27% probability of negative returns in semi-arid part, with a 14% and a 30% probability in sub-humid area. However, the net return for farms supplemented with the two management options (40 kg N/ha and the 33 000 plants/ha) has a slight difference from the farms with additional of 40 kg N/ha alone. However, the results suggest that the application of either fertilizer alone reduces the risks associated with the annual mean returns. The increase in plant population at a rate of 33 000 plants/ha without application of 40 kg N/ha has a high probability of economic failure. The third specific objective was to demonstrate user-friendly Monte Carlo simulation procedures to simulate the economic viability of different rice farming system in Tanzania. Production data for three seasons were used to demonstrate how panel survey data can be made stochastic to include risk available in the data. In this analysis, the rice farming systems entailing traditional and improved practices were compared by considering the risk associated with each system, and the best farming system was identified. The systems were categorized based on the type of seeds used (local or improved), application of fertilizers, and application of the systems of rice intensification (SRI) practices (partially or fully). The results of the economic analysis show a high probability of success for rice farmers using all the recommended SRI principles. Moreover, rice farms that partially applied the SRI principles did not realize better returns compared to their counterpart farmers that fully adopt the SRI package. Rice farms that applied fertilizers plus improved seeds were also better-off compared to rice farms under traditional practices. The study revealed that farmers who use SRI partially and fully had 2% and zero probabilities of negative annual net cash income (NCI), respectively. Meanwhile, farmers using fertilizers and improved varieties had a 21% probability of negative NCI. The farmers using improved and local rice varieties had 60% and 66% probabilities of negative returns, respectively. With high dependence on rain-fed farming, production of main cereal crops is likely to face a high degree of risk and uncertainty threatening incomes, livelihoods, and food availability to poor households. However, there is a high chance that such households will be better-off if improved technologies like the application of recommended fertilizers and SRI are properly applied. Nonetheless, the adjustment in plant population has demonstrated a slightly impact on both yield and economic returns, particularly under rain-fed production system. With evidence from crop models like APSIM and DSSAT, bio-economic integrated studies are, however, needed to explore the potential of more crop management practices and technologies for better decision-making. This study forms a basis for more studies that include risks and uncertainty to improved decision marking for farmers, government, and stakeholders in the agricultural sector. The methodology used in this study can be expanded to include more zones and other non-cereals crops and livestock farming systems.