Stochastic methods in urban transportation planning: a case study of Kisumu
View/ Open
Date
1993-09Author
Matheri, George PK
Type
ThesisLanguage
enMetadata
Show full item recordAbstract
There are many urban planning problems in developing
countries. Unlike most Western Cities, Developing Country Cities
have grown rapidly and explosively in this century. Analysis of
cities in developing countries is more complex due to the
co-existence of different kinds and levels of technology with a
wider variety of urban structural forms.
even more difficult.
These make modeling
Kisumu is currently the third largest town in Kenya after
Nairobi and Mombasa. The town has experienced fast growth in
area population and commercial entrepreneurship. Its physical
structure continues to experience fast growth receiving new
residential, commercial and industrial premises. This trend
reveals a complete re-orientation of the town's settlement pattern
with the possibility for a future urban form accompanied by a
demand for transport probably unforeseen to date.
Transportation studies for Kisumu in the past by the
Department of Civil Engineering, University of Nairobi utilized
regression analysis and the gravity model techniques. The
regression analysis is disadvantaged for being static and
deterministic in nature. The gravity model has a probabilistic
term in its formulation but it still has the static qualities
common with most models of urban transportation planning. This is
not a good experience for developing country cities which often
experience random as well as sporadic growth. Probabilistic
models are strong in specifying random components and stochastic
models such as the Markov chains are models of multivariate time
series , explicity dynamic and geared towards analyzing the
asymptotic behaviour of a process in a system.
In this pioneering study in Kisumu, Markov Chain analysis was
applied to raw inter-zonal travel data observed in the town in
1982. The zonal trip interchange matrices were observed to have a
purely Markovian property. Despite difficulties in defining the
discrete-time-space",it was found that long term transition
probability matrices derived represented Ergodic Markov Chains
which were irreducible and a periodic. In other words, inter-zonal
travel is a real life process which does not terminate in time.
The first return periods calculated from the steady state
transition probabilities varied widely with the shortest period
always coinciding with zone 22 (East ward) in the Old Town which
was a built-up area with little room for expansion, while the
longest period always coincided with zone 6 (Swahili-Mkendwa) on
the hilly part of the municipality with little building activity
recorded during the survey. The first return periods for homebased
trips had the same range with a little difference shown at
the lower limit (51 to 5882 years for home-based work trips and 49
to 5882 years for home based non-work trips). The first return
periods for non-home-based work trips had a shorter range (48 to
3846 years) while the range for total trips took an average of all
the categories combined as expected in the analysis (50 to 5263
years) .
It was concluded that the method was more versatile since
trips were allocated to each zone on proportion to its population
growth. The method avoided the cumbersome regression analysis
while it simulated the long term travel pattern in Kisumu using
data from field observations. Another conclusion was that trip
distribution should be done at the target forecast date using
correspondingly forecasted socio-economic factors so as to take
care of any anticipated changes in the urban spatial form,
normally ignored in conventional transportation planning methods
where trips are distributed only at the base year. The Markov
model was completely dynamic and it eliminated the static
component of forecasting in conventional urban transportation
planning models.
Further work was recommended 1n order to test for the
authenticity of the discrete-time period adopted and the general
applicability of Markov model in developing countries. It was
also recommended that the influence of other factors besides
accessibility to employment in the assignment of growth to the
urban zones be investigated .
Citation
Masters thesis University of Nairobi (1993)Publisher
University of Nairobi Department of Civil Engineering
Description
degree of Master of Science in Civil Engineering