Frankfurt Hydrology Paper
Expert-based Bayesian Network modeling for environmental management
Sina Kai Frank
- Bayesian Networks are computer-based environmental models that are frequently used to support decision-making under uncertainty. Under data scarce conditions, Bayesian Networks can be developed, parameterized, and run based on expert knowledge only. However, the efficiency of expert-based Bayesian Network modeling is limited by the difficulty in deriving model inputs in the time available during expert workshops. This thesis therefore aimed at developing a simple and robust method for deriving conditional probability tables from expert estimates in a time-efficient way. The design and application of this new elicitation and conversion method is demonstrated using a case study in Xinjiang, Northwest China. The key characteristics of this method are its time-efficiency and the approach to use different conversion tables based on varying levels of confidence. Although the method has its limitations, e.g. it can only be applied for variables with one conditioning variable; it provides the opportunity to support the parameterization of Bayesian Networks which would otherwise remain half-finished due to time constraints. In addition, a case study in the Murray-Darling Basin, Australia, is used to compare Bayesian Network types and software to improve the presentation clarity of large Bayesian Networks. Both case studies aimed at gaining insights on how to improve the applicability of Bayesian Networks to support environmental management.
Global estimation of monthly irrigated and rainfed crop areas on a 5 arc-minute grid
Felix Theodor Portmann
- Agriculture of crops provides more than 85% of the energy in human diet, while also securing income of more than 2.6 billion people. To investigate past, present and future changes in the domain of food security, water resources and water use, nutrient cycles, and land management it is required to know the agricultural land use, in particular which crop grows where and when. The current global land use or land cover data sets are based on remote sensing and agricultural census statistics. In general, these only contain one or very few classes of agricultural land use. When crop-specific areas are given, no distinction of irrigated and rainfed areas is made, whereas it is necessary to distinguish rainfed and irrigated crops, because crop productivity and water use differ significantly between them.
To support global-scale assessments that are sensitive to agricultural land use, the global data set of Monthly Irrigated and Rainfed Crop Areas around the year 2000 (MIRCA2000) was developed by the author. With a spatial resolution of 5 arc-minutes (approximately 9.2 km at the equator), MIRCA2000 provides for the first time, spatially explicit irrigated and rainfed crop areas separately for each of the 26 crop classes for each month of the year, and includes multi-cropping. The data set covers all major food crops as well as cotton, while the remaining crops are grouped into three categories (perennial, annual and fodder grasses). Also for the first time, crop calendars on national or sub-national level were consistently linked to annual values of harvested area at the 5 arc-minutes grid cell level, such that monthly growing areas could be computed that are representative for the time period 1998 to 2002.
The downscaling algorithm maximizes the consistency to the grid-based input data of cropland extent [Ramankutty et al., 2008], crop-specific total annual harvested area [Monfreda et al., 2008], and area equipped for irrigation [Siebert et al., 2007]. In addition to the methodology, this dissertation describes differences to other datasets and standard scaling methods, as well as some applications. For quality assessment independent datasets and newly developed quality parameters are used, and scale effects are discussed.
Supplementary Appendices document crop calendars for irrigated and rainfed crops for each of the 402 spatial units (Appendix I), data sources of harvested area and of cropping periods for irrigated crops, country by country (Appendix K), as well as data quality parameters (Appendix L, including spreadsheet files).