| Summary: | This study investigated contemporary floodplain sedimentation, interactions between sediment, vegetation, and agricultural land use, and the potential utility for a Bayesian Network Decision Support System (BNDSS) to assist farmers in making better decisions concerning agricultural land use. The research was performed around Bara Bania Mouza (village) under Daulotpur Uazila in Manikgong district of the Brahmaputra-Jamuna floodplain in Bangladesh. This area was selected because it is representative of the young and active floodplain, where the land is flooded and receives overbank sediment deposition every year. The research employed exploratory data analysis and Bayesian approaches to identify and investigate causal relationships among the variables and so support probabilistic inferences. The study investigated two distinctly different types of monsoonal flood: a bonna (an abnormally large flood that occurred in 2007) and a barsho (a normal flood that occurred in 2008). Data on landforms, flood hydraulics, sediment dynamics (suspended sediment concentrations and sediment accumulation rates), and vegetation, rain-fed flooding, land use and farmers knowledge on soil suitability and cropping were collected through field surveys. The results establish how flow and sediment dynamics contrast as a function of landform and demonstrate that the thickness and calibre of deposited sediment strongly influence farmers' decisions on which and how many crops to cultivate on a given plot. Natural vegetation (e.g. sun grass) and certain agricultural crops were shown to have huge potential for use in slowing floodwater and trapping coarse grain sediment particles in buffer stripes. Marked contrasts were also observed between the characteristics of sediment deposited by rain-fed and river water flooding. Questionnaires and semi-structured interviews revealed that although farmers have profound knowledge on soil types and crop associations their methods are crude and little or no science is involved in the investigation of soil and sediment properties. Despite this, farmers' estimates of soil properties proved to be reasonably accurate with the estimate of particle size differing by only <15% from the results of laboratory particle size analysis. This suggests that the farmers' methods do give reliable indications of key soil attributes, but that they could be improved if scientific information was integrated with their local knowledge. A Bayesian approach provides a means of achieving this and the BNDSS developed in this study was found to produce good results when compared to field observations and backward propagation indicated that for better decision making it is crucial to consider both physical and socioeconomic variables. The findings of the research reported in this thesis show that sedimentation has major impacts on agricultural land use dynamics in the Brahmaputra-Jamuna floodplain and that both natural vegetation and agricultural crops significantly influence sediment movement and the way that deposition is distributed over the floodplain. In a wider context, flood, sediment, vegetation and agricultural land use dynamics are controlled by complex set of both physical and human phenomena that are challenging to describe, integrate, analyse and interpret in a single study. In light of this, it is not surprising that the findings presented in this thesis highlight important gaps in knowledge that need to be addressed through further research.
|
|---|
