224
J. U. Kitheka et al.
affected by changes in rainfall and river discharge levels. The mean monthly rainfall data were obtained from the Kenya Meteorological Department (KMD) rainfall stations located at Ahero, Kibos, Koru, Lumbwa, and Nyando. Rainfall data
were used to determine the relationship between rainfall and streamflow.
1 There is,
however, a huge gap in the sediment data because the only available data on TSSC
for RGS 1GD04 were for the period between 2000 and 2015.
3.3 Analysis of Land-Use/Land Cover Changes
Cloud-free Landsat images were downloaded from the United States Geological
Survey (USGS) website for the years 2000 and 2015. To obtain complete coverage
of the study area, two scenes, of paths 169 and 170, and raw satellite images were
subjected to atmospheric correction and then projected to UTM 37 using WGS 84
data. Bands 4, 3, 2 for these images, and 5, 3, 2 for OLI images, were used to
obtain a false color composite. Three subsets were created using the shape file of
the study area. Images were enhanced to facilitate image interpretation. To obtain
a rough idea of the number of classes in the study area, unsupervised clustering of
ISODATA was used. Training sites were developed from Google maps. Supervised
classification using maximum likelihood algorithm was used to classify the images.
The classified images were then subjected to post-classification smoothing using a
majority filter to remove the “salt and pepper” appearance and enhance cartographic
display. Post-classification change detection was used.
3.4 Methods of Data Analysis
Data were analyzed using both qualitative and quantitative methods. The frequency
distribution model (Gumbel Extreme Value) was used to determine the return periods
of various categories of streamflow of the Nyando River. In this analysis, the magnitude of floods was related to frequency of occurrence. The frequency and magnitude
of flood flows were determined using the Gumbel Extreme Value Probability Model
that also compares different return periods to predict future flows. Return periods
of rainfall and river discharges were also determined in this study. The inverse of
probability is generally expressed in percentage, given the estimated time interval
between events of a similar size or intensity. Trend analysis was undertaken to determine seasonal and annual variations of both rainfall and river discharges. The study
also employs the use of measures of central tendency in the analysis of data (Lane
2008). Various quantitative methods used in data analysis include regression and
correlation analyses and Analysis of Variance (ANOVA).
1 Data collected were coded, processed, and analyzed using Microsoft Excel.
J. U. Kitheka et al.
affected by changes in rainfall and river discharge levels. The mean monthly rainfall data were obtained from the Kenya Meteorological Department (KMD) rainfall stations located at Ahero, Kibos, Koru, Lumbwa, and Nyando. Rainfall data
were used to determine the relationship between rainfall and streamflow.
1 There is,
however, a huge gap in the sediment data because the only available data on TSSC
for RGS 1GD04 were for the period between 2000 and 2015.
3.3 Analysis of Land-Use/Land Cover Changes
Cloud-free Landsat images were downloaded from the United States Geological
Survey (USGS) website for the years 2000 and 2015. To obtain complete coverage
of the study area, two scenes, of paths 169 and 170, and raw satellite images were
subjected to atmospheric correction and then projected to UTM 37 using WGS 84
data. Bands 4, 3, 2 for these images, and 5, 3, 2 for OLI images, were used to
obtain a false color composite. Three subsets were created using the shape file of
the study area. Images were enhanced to facilitate image interpretation. To obtain
a rough idea of the number of classes in the study area, unsupervised clustering of
ISODATA was used. Training sites were developed from Google maps. Supervised
classification using maximum likelihood algorithm was used to classify the images.
The classified images were then subjected to post-classification smoothing using a
majority filter to remove the “salt and pepper” appearance and enhance cartographic
display. Post-classification change detection was used.
3.4 Methods of Data Analysis
Data were analyzed using both qualitative and quantitative methods. The frequency
distribution model (Gumbel Extreme Value) was used to determine the return periods
of various categories of streamflow of the Nyando River. In this analysis, the magnitude of floods was related to frequency of occurrence. The frequency and magnitude
of flood flows were determined using the Gumbel Extreme Value Probability Model
that also compares different return periods to predict future flows. Return periods
of rainfall and river discharges were also determined in this study. The inverse of
probability is generally expressed in percentage, given the estimated time interval
between events of a similar size or intensity. Trend analysis was undertaken to determine seasonal and annual variations of both rainfall and river discharges. The study
also employs the use of measures of central tendency in the analysis of data (Lane
2008). Various quantitative methods used in data analysis include regression and
correlation analyses and Analysis of Variance (ANOVA).
1 Data collected were coded, processed, and analyzed using Microsoft Excel.