Background
Area description
The area (see map) is situated in the tropics with annual rainfall reaching up to 2,000 mm and mean
maximum temperatures ranging between 23.8 °C and 27.7 °C. Placed on a modulated terrain with moderate
to very fertile soils, the farmland surrounding Kakamega Forest is characterized by small-scale, largely
subsistence-oriented farms. While sugarcane and tea are cultivated as cash crops in the northern and
southern parts respectively, maize and beans are cultivated as staple food throughout the region. A very
high population density of 643 people per km² (1999, determined for a 2 km-zone around Kakamega Forest)
coupled with an inheritance system, where the family ground is equally divided between the sons, has led to
increasing land scarcity and thus to an extremely structured and heterogeneous landscape.
Contents depicted
For a typification of the farmland, three different themes have been identified: land use, landscape structures,
and accessibility. In addition a synoptic farmland typology is considering all three themes. All typologies displayed
were derived by means of cluster analysis. Here, results of the satellite image classification and other spatially
explicit information derived from additional data sources served as input. The maps created allow the identifying of
areas of similar characteristics and thus reveal distinct spatial patterns. From these findings local planning could
benefit by adjusting planning activities to the specific needs and problems of people living in similar settings. In
order to further illustrate the differences between the ten farmland types, one hexagon has been selected per farmland
type. Here, through large-scale visualizations of present and future land use/cover situations, the typology is shifted from
the rather abstract presentation at hexagon level to a closer-to-reality depiction making it more comprehensible for e.g.
the local people to become aware of upcoming problems.
Eight scenarios were developed to describe the situation of rural livelihood as reflected by the satellite image classification
(2005) and possible future situations (2010, 2015 and 2020) assuming various long-term developments in yields and prices. Values
for prices and yields were derived from diverse (literature) sources and extrapolated into the future. Projections on population
growth were derived from statistical data on sublocation level and spatially distributed/gridded based on the location of houses
as analysed from the satellite imagery. Other socio-economic factors considered in the modelling include the amount of maize used
for home consumption, the number of harvests per year, the area making up a homestead and assumed declining household sizes (see
this document for details).
For the scenarios, seven map topics have been identified to describe the situation of rural livelihoods. These include the
production of the most important cash as well as staple crops per household (hh) and year. Additionally, the share of land
under cultivation of maize that is available for sale on the market and the total production of maize minus home consumption,
termed `maize production balance´, have been selected. These offer to observe potential developments regarding the most important
agricultural products of the area. Further scenarios focus on the income generated through sale of sugarcane, tea, and maize,
the latter deduced by the maize needed for home consumption, as well as on what additional maize yields could be achieved if unused
fallow land was used for the cultivation of maize (and beans).
Map visualization
The classification of the QuickBird satellite imagery resulted into 15 classes of land use/cover and a total of over 700,000
individual polygons. A direct representation of the classification result (possible in scales of 1 : 5,000 and 1 : 25,000, depending
on the purpose) would lead to map sizes of approx. 7.6 m by 5.4 m and 1.5 m by 1.1 m respectively. Therefore, it seems impossible to
meaningfully visualize the actual classification result for the complete area under investigation. In order to obtain an effective
overview, aggregation is needed, which also allows for spatial patterns exploration and thus the gaining of geospatial knowledge.
In all maps of the tool, the area under investigation is sub-divided into 1,324 hexagons forming the spatial basic unit of analysis.
For the 8 scenarios, 7 map topics, and 4 time steps, a total of 175 (3 x 7 x 8 for 2010-2020 plus 7 for 2005) thematic maps are
depicted (5 of them being identical), which are visualized as honeycomb choropleth maps. Depending on the map topic, colour codings
with distinctive colours (see typology and land use maps), or in case of quantitative information unipolar or bipolar colour schemes
are applied. On the page view all themes, all individual maps can be displayed in large after having chosen
a map theme (typology maps and a population map are available as well), a year, and in case of the scenarios also a map topic. On the
page compare scenarios, several maps can be viewed at the same time following the concept of small multiple
maps or map series. Maps for different years (choose a scenario and a topic), different scenarios (choose a topic and a year), as well
as different topics (choose a scenario and a year) can be juxtaposed. Here, the user is highly benefitting from the interactive and/or
dynamic visualization techniques enabling a switching between the many maps. The maps depicted on the investigate
typology page help to make the maps more useful for local planning, as the farmland types (each visually explained) might offer
the required spatial reference.
Technical realisation
This website was realised using a combination of different web technologies, including Scalable Vector Graphics (SVG), JavaScript
(i.e. ECMAScript), Cascading Style Sheets (CSS), Extensible Markup Language (XML), and Extensible HyperText Markup Language (XHTML,
version 1.0, strict). While all map geometries are created from a single SVG file, the information of the map content, i.e. the colour
coding of the hexagons, is dynamically loaded from XML files using JavaScript. Map legends are also designed using SVG, while image files
(e.g. of the large-scale visualizations) are in Portable Network Graphics (PNG) format. Finally, XHTML forms the frame for creating the
individual web pages. For correctly displaying the website, JavaScript needs to be enabled in the browser and an SVG viewer needs to be
available such as by default in case of Mozilla Firefox from version 3.6 onwards.
Further reading
Further information on the spatially-explicit results of the BIOTA East Africa research project can be obtained from the
GVISЯ Web pages. For more detailed information on the analysis of
the QuickBird satellite imagery, the spatial farmland typology and the scenarios of rural livelihoods see Lübker, T. (currently in prep.):
Object-based remote sensing for modelling scenarios of rural livelihoods in the highly structured farmland surrounding Kakamega Forest,
western Kenya (a PhD thesis supervised by Prof. Dr. M. Buchroithner (Dresden) and Prof. Dr. G. Schaab (Karlsruhe) to be handed in at
Dresden University of Technology, Faculty of Forest, Geo and Hydro Sciences).
Imprint
This website was prepared by Johannes Klein as part of his Bachelor thesis at Karlsruhe University of Applied Sciences, Faculty of Geomatics.
© by J. Klein, T. Lübker, and G. Schaab (Feb 2011). (Contact: Prof. Dr. Gertrud Schaab, Karlsruhe University of Applied Sciences, Faculty
of Geomatics, Moltkestr. 30, 76133 Karlsruhe, gertrud.schaab@hs-karlsruhe.de)