Modelling of nitrates in River Nzoia using MIKE 11

Lake Victoria, one of the largest freshwater lakes in the world supporting the lives of about 30 million people in Kenya, Uganda and Tanzania (Khan et al. 2011), is under threat from pollution. Eutrophication is a serious problem in Lake Victoria as characterised by the presence of water hyacinth. Nitrogen and phosphorus are the most important nutrients responsible for eutrophication in lakes and streams. In lakes, phosphorus is the limiting nutrient and therefore is the centre of interest in the control of eutrophication.

River Nzoia is the largest river on the Kenyan side of Lake Victoria basin and therefore any change in water quality (WQ) of the river will ultimately affect the lake. It is the source of water for industrial, domestic and agricultural activities in the Nzoia catchment. The river receives both non-point source pollution from the nutrient-rich agricultural run-off from the catchment and point sources from the agro-based industries and municipal wastewater treatment plants. According to Nyongesa (2005), River Nzoia contributes the largest quantities of nitrogen and phosphorus as compared to other rivers draining the Kenyan portion of L. Victoria. Carrying out compressive assessment require huge resources which are not available in most countries in Africa. Models are cost-effective in pollution monitoring and prediction of future WQ of rivers and streams (Deksissa et al. 2004) and they can be used in the study of eutrophication and assessment of the effects of point and diffuse pollution (Erturk 2010).

One-dimensional models, such as MIKE 11 and QUAL2E/K, are commonly used in river systems while two- and three-dimensional models are typically used in reservoirs, lakes, and estuaries (Razdar et al. 2011). Cox (2003) suggested that complex models like MIKE-11 and QUAL2E should be considered because they include a large number of processes that could be used in research not only for oxygen but also other chemical properties of water. MIKE 11 model was used in this study due to its hydrodynamic component which makes it suitable for rivers which are unsteady like River Nzoia. The model was used by Chibole (2013) and Kanda et al. (2015) for simulation of dissolved oxygen and biochemical oxygen demand for River Sosiani (a tributary of River Nzoia) and River Nzoia respectively and was found to be reliable in WQ modelling. Using MIKE 11, this study aimed at investigating the behaviour of nitrates in River Nzoia under the existing agricultural practices and to establish the relationship between nitrogen and phosphorus.

The predominant type of land use (Figure 1) is rain-fed herbaceous crops mainly sugarcane plantations owned by the out-growers of the two major sugar industries of Mumias and Nzoia. The irrigated herbaceous crops are found around the sugar factories forming the nucleus of the sugarcane farms owned by the sugar factories.

In this study, five sampling sites (Figure 1) were used for the measurement of the WQ parameters and discharge. The sampling was done two times each for the months of March and April and the average values used for the simulations. Monthly WQ and discharge data for 2009 were used to calibrate the model while the data collected in March 2013 was used for validation. The sampling points are described in Table 1.

The Hydrodynamic module (HD) of MIKE 11 (DHI 2009) was used to simulate flow in the river. The model was calibrated and used in the advection-dispersion (AD) module. WQ module was developed and integrated with the AD module for simulating the river's WQ model. The HD and AD model were calibrated as described in Kanda et al. (2015). The main calibration parameters were bed resistance for HD module, and dispersion coefficient and dispersion exponent for the AD module since electrical conductivity which was used for calibration is a conservative parameter and therefore decay is zero. For WQ module the constants shown in Table 2 were used where most of them were default values in the model while others were from the reference Manual (DHI 2009). The WQ module was implemented using predefined Ecolab templates.

The model calibration for the WQ module was done using monthly data of 2009. Model performance was assessed using the coefficient of determination (R²), Nash-Sutcliffe Efficiency (EF) and Percent Bias (PBIAS). After calibration, the model was validated using data collected in March 2013 and the validated model was used to simulate the nitrate concentrations for the month of April 2013. The performance of the model is indicated in Table 3.

In order to assess the role of rainfall in the concentration of nitrates in the river, the variations in the concentrations for wet months and dry months were analysed between 2009 and 2013. Correlation analysis for the sets of data indicated weak relationship (r = 0.45). The ANOVA results showed significant variation (p < 0.05) which confirm the observed variations in the values. This was consistent with studies by Mokaya et al. (2004) which indicated a strong positive relationship between nitrate and discharge. Nzoia basin is characterised by large maize plantations in the upper parts and sugar belts in the middle region with a substantial application of fertilisers, like Di-Ammonium Phosphate (DAP), Urea, and Calcium Ammonium Nitrate (CAN), which are important sources of nitrogen. Thus, the high concentrations of nitrates can be explained by its accumulation in soils during dry periods which are later mobilised and transported by higher rainfall, soil moisture, runoff, and base flow generation to the streams (Burgos 2012). A Study by Omwoma et al. (2012) established that retention ponds dug in the sugarcane plantations of Nzoia basin had high concentrations of nitrates and therefore it shows that nutrients such as nitrogen leach from these plantations and eventually end up in River Nzoia through the various streams. Crops such as maize and sugarcane which form the bulk of crops in the Nzoia catchment, leak substantial amounts of nitrates compared to perennial crops (Randall & Mulla 2001) since the latter, due to their extensive root system, take up nutrients more efficiently and therefore require reduced amount of nutrients through fertilizer application hence reduced fertilizer run-off.

From Figure 3, 51 percent of the variation in phosphorus can be explained by variation in nitrogen. This is because, apart from agricultural activities like fertilisers which contribute to both nutrients, nitrogen contribution from other sources such as atmospheric deposition can be significant. According to studies by Pieterse et al. (2003), fertilisers contributed virtually all the phosphorus into the river and about 90 percent of nitrogen with atmospheric deposition contributing the rest. However, Mayer et al. (2002) found that up to about 40 percent of riverine nitrate originated from atmospheric deposition.

This study assessed the concentration of nitrates in River Nzoia using MIKE 11 which is an unsteady –state one-dimensional model. Nitrates showed significant seasonal variation with low concentrations being observed during low flow period but high concentrations during the wet months. This suggested that significant contributions of nitrates from the catchment could be attributed to run-off. The study also noted elevated concentrations of nitrates as a result of discharge from industrial wastes from MSF and wastewater stabilisation ponds in Webuye.

Owing to the importance of nutrients in the eutrophication process with phosphorus being the limiting nutrient, the relationship between TN and TP was analysed and found to have a positive correlation. This indicated that P and N have similar factors affecting their concentrations in a river. Phosphorus and nitrogen originate mostly from agricultural activities. The Nzoia catchment is characterised by agricultural activities with the growing of sugarcane dominating the middle catchment where the study was carried out while the upper part of the catchment has extensive plantations of maize (Trans- Nzoia region) and wheat and maize in Uasin-Gishu areas. These crops require the intensive application of fertilisers such as DAP, CAN, and Urea. The study demonstrated that mathematical models such as MIKE 11 are good in the modelling of nutrients such as nitrates and thus aid in monitoring and management of pollution in rivers. Further studies could be carried out using watershed models in order to quantify the non-point source component and thereby help in the holistic management of the river.

The authors gratefully thank the National Commission for Science, Technology and Innovation of Kenya (NACOSTI) and the German Academic Exchange Service (DAAD) for funding this research.