Spatial Planning, Urban Growth, and Flooding-Contrasting Urban Processes in Kigali and Kampala

PhD Research: Urban growth is a factor known to intensify local flooding. By orienting urban development, land use planning may contribute to reduce flood risk through regulatory constraints. Two case studies were developed to determine the extent to which such strategy may be effective: Kigali, Rwanda (where land use regulations are stringently applied) and Kampala, Uganda (with much less effective institutions but important infrastructure investments over the last decade). Both cities are mid-sized (one to two million inhabitants), they share a physical context of hilly terrain and low-lying flood prone valleys but with divergent policy and institutional organizations. Two main hypotheses were investigated based on the case studies. The relations between the physical system, through recurrent flooding, and the human settlement pattern were first explored. Urban growth is one cause of increased flooding but, in turn, flooding was thought to contribute to the urban pattern's evolution. Secondly, and based on this premise, a land use management system (with regulation a prominent component) was proposed as a flood risk mitigation strategy: these questions hinged around the feasibility of land use controls in the specific context of the cases (mid-to-large cities in Sub-Saharan Africa) and of their cumulative impact over the long run. Spatially explicit prospective simulations of urban growth, up to the year 2030, were developed for both Kampala and Kigali to understand the impacts of flooding and land use regulations; additionally, a set of scenarios for Kampala was specified to explore the potential feedback effect between exposure to recurrent flooding and urban development patterns. The main lessons derived from these simulations were: in Kampala, which has until the present expanded without strong land use controls, the implementation stringent land use regulations (envisioned already in their strategic plans) would likely result in a more compact growth; however, in Kigali, the land use plan may have the unintended consequence of promoting sprawling patterns. Kigali was revealed to be a smaller urban system than Kampala, with the transitory benefit of not being yet impacted by recurrent flooding due to the scale of processes configuring urban growth. As for Kampala, (1) while land use planning may reduce exposure to flooding, it is unlikely to impact runoff generation and (2) explicitly incorporating feedback between flooding and urban growth makes visible a difference introduced by land use planning: under trend (unplanned) expansion, exposure to flooding is unlikely to constrain urban growth; however, under the double restriction of recurrent flooding and land use controls, much less new development is exposed to flooding. The scenarios were carried out using a cellular automata of urban growth, specifically designed to be integrated with the flood model (implemented in OpenLISEM). Important characteristics of the model included: a continuous response variable, in the form of a land cover fraction value (built-up fraction for urban development but also vegetation, bare soil, and water fractions to complete the description of the landscape), a suitability based allocation procedure to mimic urban agents' locational preference, and the potential to explicitly account for several supply scenarios (which was especially important when considering the relation between population growth and densification in the scenarios). The suitability index was defined by a neighborhood effect, accounting for the immediate context of each potential development location, as well as ancillary variables representing accessibility, physical characteristics (slope and wetlands location), and informal settlements location. The cellular automata model was developed using the Upper Lubigi sub-catchment of Kampala. The model was then expanded, calibrated, and validated for the metropolitan areas of Kampala and Kigali. Calibration was based on the application of the Metropolis-Hastings algorithm to determine the relative importance of each factor in the suitability index and using the land cover maps to simulate potential supply. Simulations using 2000 (for Kigali) and 2001 (for Kampala) as baseline years were generated for a 15 year period; for each simulated time step, landscape metrics were calculated. An intermediate year for which independent land cover maps were available (2009 for Kigali, 2010 for Kampala) was used for validation. The calibration approach proved useful in producing patterns that better replicate the evolution of urban growth patterns, relative to random parameters and data. However, some degree of equifinality was discovered in the model, since the uncertainty introduced by parameters was found to be less important than the amount of information (relevant spatial determinants) when validating the model. The scenario assumptions on the presence of a feedback (for Kampala) and on the possible efficiency of land use controls were based on statistical analysis. The potential impact of flooding on urban growth was investigated using a structural equations model of Kampala, by making strong assumptions on the causal structure and performing confirmatory analysis to test whether data conform to such assumptions. The conclusion was, as expected, of a significant but weak restriction of flood impacts on urban growth patterns. Land cover maps were used to describe urban growth, the flood impacts were derived from the OpenLISEM model of Kampala. The effects of land use regulations on urban development patterns were calculated for Kigali using a difference-in-differences estimator. Kigali was chosen because already a stringent land use control system has been in place for the better part of two decades which, coupled with relatively rapid urban growth, provides a quasi-experimental setting. The conclusion was land use controls had in fact a statistically significant and strong impact on urban development. In synthesis, this dissertation has developed a spatially explicit methodological framework to simulate the relations between urban growth, land use planning, and recurrent flooding. It was applied to two case studies in Sub-Saharan Africa, the cities of Kampala and Kigali. The framework is based on the integration of a cellular automata of urban growth and a flood model to reproduce the processes configuring spatial patterns. Scenarios, in turn, were specified based on the result of spatial statistical analysis of the relation between the main variables being explored. The results underscore the importance of opportunities but also the pitfalls of land use regulation as a policy response for flood mitigation.