MODELLING CONTROL OF POLLUTANTS LOADS IN LAKE KIVU USING BINARY LOGISTIC REGRESSION METHODS

Abstract

High concentration of pollutants in water surfaces causes catastrophes which kill aquatic life,
kill people and affect livelihoods of those living in the catchment of the polluted water surfaces.
In this study, a logistic binary regression model was applied to existing data of pollutant
concentrations in rivers in flowing into Lake Kivu to compute estimations of probabilities that
the rivers pollute the Lake. Rivers that discharge large amounts of pollutants in the Lake were
identified. The goodness-of-fit test statistic established that the formulated model could be
used to predict the status of pollutants load in the Lakes. The amount of Total Suspended
Solids, Soluble Reactive Phosphorus, Nitrates, and Dissolved Silica are determinant factors to be
used to predict the status of pollutant loads in Lake Kivu. The model was then used to simulate
scenarios to reduce and control pollutant loads using methods of Infiltration Basin, Terrace
System, and Constructed Wetlands.
It is established that the use of any of the above control methods will drastically reduce the
pollutants load to allowable levels in all of the five identified rivers on the Rwanda side and five
out of the eight identified rivers on the DRC side. The constructed wetlands method is the best
option to reduce the level of pollutants followed by the Terrace System. The least is the
Infiltration Basin method. However, since the constructed wetlands method is expensive and
has many limitations, the Terrace System could be used for all the rivers on Rwanda side and
five on the DRC side. It fails for three rivers on DRC side. These results are in excellent
agreement with former studies elsewhere.