Developing Sediment Yield Prediction Equations for Small Catchments in Tanzania - Advances in Data, Methods, Models and Their Applications in Geoscience - page 257

relationship developed from data collected from both dry and moderate climatic regions
combined has an improved performance according to R 2 of 0.732 as compared to 0.451 for
dry climatic zone alone. This could be partly due the fact that the number of data points
used to fit the regression is adequate for robust relationship as recommended by Statsoft
(2011). However, the author would like to recommend the use of a specific equation for
particular purpose/climate, especially Arusha region, as they have been developed.
Validation result for the developed prediction equation was much better than during model
training phase with R 2 of 0.873. This was attempted only to regions/climatic zone/region
where splitting of data into 70% and 30% for calibration and validation was possible, that is
Tabora. The number of observations used for this purpose was 7 (serial numbers 18 through
24 in Table 2.2.2).
Although the performances of the developed Regional Regression Relationships are
satisfactory, the author caution the reader, as supported by Morris and Fan (1998) that
these equations express only the general relationships between independent variables and
the sediment yield-fill and should therefore be used only for preliminary planning
purposes or as a rough check. Because these equations reflect regional average conditions,
the actual yields will tend to be higher (or much higher) than predicted in erosive areas
and lower than predicted in areas of undisturbed catchments. Local site-specific
conditions can influence sediment yield much more than drainage area or runoff, for
instance.
4. Conclusions and recommendations
4.1 Conclusions
This study uses readily available data on catchment area and reservoir sediment fill
and/or sediment yield to calibrate the prediction equations' parameters by regression
analysis approach. The influence of rainfall and/or runoff as important input variables
were indirectly captured by developing and grouping the equations with respect to their
climatic zones. The equations were validated and parameter uncertainty bounds
estimated. The data set was split into 70% and 30% proportions for calibration and
validation purpose, respectively. The measured and predicted reservoir sediment fill-
yield rates have satisfactory to good correlations with Coefficient of determination , R 2 ,
between 0.46 and 0.77 with a degree of freedom of, n , 11 to 41 at probability level of
significance, p , of 5%. R 2 of 0.87 at n equals 7 was achieved in one of the validation
experiments in moderate climatic zone. Although the performances of the developed
Regional Regression Relationship are satisfactory, the author would like to caution the
reader that these equations express only the general relationships between independent
variables and the sediment yield-fill and should therefore be used only for preliminary
planning purposes or as a rough check.
4.2 Recommendations
It should be noted that in this work the study area was ill-defined as the wet climatic zone of
Tanzania was not adequately represented. Notwithstanding a satisfactory performance
achieved, this chapter recommends both extending the data set to cover the wetter regions
and incorporating other parameters affecting sediment yield for processes studies and better