Bemigisha, Jane 
  
1.2 The study Area 
The study was undertaken in the Lake Naivasha riparian area (also known as the “Lake Zone”) covering an area of 
30,000. It is located in the eastern rift valley of Kenya between 194000m E, ”32000mN, 215000mE, ”08000mN. The 
area is dominated by flat to gently sloping topography, except for the western offshore areas, which are moderately 
steep. Lake Naivasha is a fresh water crater lake, surprisingly in alkaline environs. This is believed to be a result of 
underground water inflows plus the inflow of Gilgil, and Malewa rivers. One spectacular characteristic of the lake is its 
oscillating nature over the past years, recorded between 1800s to 1990 ( John Goldson Associates, 1993; Stuttard et al 
1995). 
The vegetation forms a zonation along a topographic gradient starting from the lake dominated by a papyrus fringe, 
There are other macrophites to the immediate fringe of the lake. Other vegetation include shrublands, grasslands 
and woodlands. 
Most of the human activities are supported by the wetland through water supply and the ecological functions of the 
swamp. The characteristic species, C. Papyrus plays a vital role in the hydrological regime, modification of the water 
quality, and is habitat to a number of wildlife resources. Activities like agriculture (intensive irrigated horticulture; dairy 
and cattle ranching), tourism/recreation (popular black bass sport fishing; and sanctuary for about 350 bird species plus 
hippos, buffalo, Thompsons, Grants gazelles and impala) , and urbanisation ( Naivasha town ) are supported (e.g. John 
Goldson Associates 1993; Gaudet 1997; and Gaudet 1980). 
2 METHODOLOGY 
2.1 Landcover Mapping 
The indicator, area per habitat was derived through mapping of the landcover for the years of analysis, 1967, 1984, 
and 1995. Following ITC procedures (Groten 1995), the 1967 aerial photographs (panchromatic) of scale 1:50,000 were 
interpreted and both the tilt as well as flight error corrected using a pantograph. The corrected interpretation was 
digitised in a GIS (ILWIS 2.1). The 1984 aerial photographs (panchromatic) of scale 1: 12,000 were interpreted. In 
addition to the tilt /flight error correction, the scale was reduced to 1: 50,000 using the pantograph. The scale reduction 
was done to allow for the overlay with the map of 1967, which was at 1: 50,000. The corrected interpretation was 
digitised in a GIS (ILWIS 2.1). A satellite image TM of 1995 was interpreted and digizided on screen. This was 
verified with field data on vegetation structural classes. For each map, areas for landcover types were calculated using 
the HISTOGRAM function in ILWIS 2.1 
2.2 Change Detection 
A change analysis of the land cover types was undertaken to establish trends and this was done in a GIS (ILWIS 2.1, 
MAPCROSS function) for the years 1967 to 1984, and 1984 to 1995. In anticipation of obvious change between the 
year of study 1997 and the image date of 1995, the 1984/1995 change map results were evaluated using a rule based 
interpretation ( Mariumni, 1997) . The process is able to check the mapping inconsistencies based on established rules 
that take the form: “If condition then Inference’ 
Two criteria were used to develop the rules: 
1- A threshold area (a minimum of 0.01 ha) below which a change could be established was used with the assumption 
that an area that is realised as changed should be at least twice the pixel size (50 m) 
2- The expected change direction using own knowledge, for example a reasoning that a built up area may not change 
into papyrus between 1984 and 1995. 
A list of rules were identified and applied in assessing the change and scores of positive (--) and negative ( -) assigned to 
each land cover type that changed to or from papyrus respectively. 
2.3 Deriving model inputs 
2.3.1 Criteria 1: Risk based on distances to papyrus areas (pixels): It was assumed that the nearer the papyrus to 
the landcover the higher the risk. Equal chance was assumed for all the land cover types assuming no constraint. From 
the land cover map of 1995, individual landcover type maps were extracted using a map calculation (MCALC) function in a 
GIS (ILWIS 2.1) resulting in the respective cover type maps. For each land cover type map a DISTANCE function was 
used to characterise each pixel in the study area with the distance from the cover type. To get the potential risk for 
papyrus, a map calculation was done for each distance land cover type map to get the area of papyrus within. The risk 
index was developed from the range of the values in all the land cover maps by a normalisation process (Eastman et al. 
1995) using the following formula: 
  
  
166 International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B7. Amsterdam 2000.