- author = "Moritz Jan Hildemann",
- title = "Spatially explicit land use optimization under uncertainty",
- school = "University of Muenster",
- year = "2023",
- address = "Germany",
- keywords = "genetic algorithms, genetic programming, Geoinformatik, SWC-measure, LandClim Simulation, ForSys, LANDFIRE, GIZ GmbH, Central Statistical Agency Ethiopia",
- publisher = "Westfaelische Wilhelms-Universitaet Muenster",
-
URL = "
http://www.cs.ucl.ac.uk/staff/W.Langdon/ftp/papers/MoritzHildemann_PhD_thesis.pdf",
-
URL = "
https://books.google.co.uk/books?id=87qv0AEACAAJ",
- size = "180 pages",
-
abstract = "Land is a scarce resource, and the competition for
using it continues to increase: Urban areas are growing
worldwide while the demand for agricultural products is
rising with rapid world population growth. These
developments combined are leading to devastating
environmental impacts. Consequently, many current and
future systems depend on efficient land use planning.
Land use allocation, which is defined as the process of
allocating different activities or uses to particular
areal units within a region, contributes to more
efficient land use. Spatial optimization is a method to
computationally address spatially explicit land use
allocation problems. Spatial data is essential to
define and solve problems, but it contains (spatial)
uncertainty that complicates solving such land use
problems: The uncertainties in the spatial data
propagate via the decision variables (the land use
configurations), objectives (what we want to optimize,
e.g. minimize financial costs and/or maximize
biodiversity), and constraints (conditions that the
land use configurations have to meet), to the
optimization outputs. Considering spatial data
uncertainty in land use allocation optimization may
help better assess the reliability and robustness of
optimization outputs.
Yet, land use allocation optimization studies considering uncertainty in input data are rare, and no research has been conducted on spatially explicit land use allocation problems.
This thesis aims to quantify and use uncertainty in spatially explicit land use allocation optimization problems. In this thesis, methods are proposed and advanced to use probability distributions of uncertain spatial data for the evaluation of land allocation solutions and show the potential effects of uncertain data on the optimization outputs. Furthermore, methods are suggested to analyze the objective values and the land use configurations associated with optimal solutions. The developed methods help to identify robust solutions to several land use problems that are exposed to uncertainty. This thesis advances methods to reduce computational costs associated with optimization and uncertainty analysis to accelerate research in this complex yet important research domain.
After laying out the main principles, Chapter 3 quantifies the propagated uncertainty from spatial input data to outputs of a multi-objective land use allocation, i.e. Pareto fronts that include all optimal trade-off solutions. Hereto, we developed a new method to estimate the lower and upper bounds of Pareto fronts. The method allows demonstrating the potentially very large effect of spatial input uncertainty on land use allocation. The results show that uncertain spatial input data to land use allocation optimization results in irregularly shaped Pareto fronts.
In addition to quantifying the propagated uncertainty, this thesis compares solutions with stochastic objective values during the optimization procedure (Chapter 4). This approach allows for evolving optimal solutions with uncertain objective values. We applied it to an allocation optimization for soil and water conservation measures in Ethiopia to minimize soil loss rates and labour requirements. In this problem, uncertain spatial soil and precipitation data resulted in uncertain objective values for labour requirements and soil losses. Upon in-depth analysis of common patterns in optimal solutions, we concluded that the results can help determine optimal intermediate construction stages of the conservation measures in addition to optimal final construction stages. Our findings showed that the modeling and the consideration of spatial data uncertainty play a crucial role in identifying optimal solutions.
Chapter 5 addresses a forest treatment allocation optimization to reduce wildfire risk. The variable that is necessary to evaluate solutions was deeply uncertain since it is based upon data that can not be validated. Consequently, the trustworthiness of generated optimal solutions is questionable. We addressed the problem by generating multiple possible scenarios of the uncertain variable. The combination of a trade-off analysis with a spatial configuration analysis resulted in robust solutions to all scenarios.
To promote the uncertainty analysis of spatially explicit land use allocation problems that are generally computationally expensive, we propose a new search algorithm that drastically reduces computation time (Chapter 6). Moreover, the improvement in computation time did not reduce the algorithm performance in finding optimal solutions; often, it even increased.
We proved that uncertainty in spatial data can drastically affect land use allocations and their objective values. Instead of ignoring uncertainty, approaches should quantify and address uncertainty within optimization procedures. This thesis focuses on methods that directly address uncertainty; however, it also proposes approaches that indirectly help to address uncertainty by increasing the computational efficiency of spatially explicit land use allocation under uncertainty. Due to current trends of better data availability and increasing computational capacities, we expect a future state in which the most constraining matter for land use optimization under uncertainty is the conceptual knowledge of the modeled topic and the methods to address the specific problems. Moreover, we expect that the uncertainty will not necessarily be reduced with more data and new models: even though new technologies might reduce the uncertainty in data measures, more data with quantified uncertainty will be available. We highly recommend making use of spatial data with quantified uncertainty to provide a more robust foundation for decision-making in land use planning.",
-
notes = "Also know as \cite{hildemann2023spatially}
In english
Supervisor: Judith Verstegen",
