Last Updated 1 week ago

Predictive Hydrological Models and Zero Flow

---

---

From the numerous models developed to address the zero flow in the Ruaha Park and to simulate the water balance on the Usangu Rice Plain, three hydrological models are selected for further assessment.

A critical drainage flow from the Usangu Rice Plain, a treshold value which should not be reached in order to avoid zero flow at the Park HQ, will be a useful water management tool in any hydrological model for the Ruaha River.

The three selected models attempt to provide such a critical treshold flow. One of the models also facilitates estimates for the duration of zero flow periods by simulating different dry season abstraction levels and by simulating the impact of proposed zero flow mitigation investments. These investments include construction of a weir at the outflow of the Ihefu Swamp and others.

1. The “Sustainable Management of the Usangu Wetland Catchment Program” (SMUWC), a four million dollar study financed by the British DFID²⁷ and implemented around the millenium, produced the “Usangu Basin Model”. The model is developed by the Water Resources Engineering Programme of the University of Dar es Salaam, hired under the SMUWC program. According to the SMUWC Program Report of 2001²⁸, the Model has been installed on the computer of the Rufiji Basin Water Office in Iringa, which authority are responsible for the water management of the Ruaha River.

2. The second model (2006) assumes an open channel flow model²⁹ from the Usangu Rice Plain drainage pour point at Nyaluhanga through the Eastern Wetland (Ihefu Swamp) to the N’giriama outlet. This approach differs from the previous model in which the Eastern Wetland has a water budget based on a basin with a flat water surface rather than a river with considerable surface water slopes flowing through the wetland as an open channel. This model predicts not only the flow in the Ruaha at N’giriama which is the outlet of the Eastern Wetland but also downstream at the Park Headquarter (Msembe) and at Mtera where the Ruaha flows into the Mtera Reservoir (Map 1).

3. The third model (2016)³⁰ financed under the World Bank REGROW project, is based on the lumped conceptual hydrologic model HBV (Bergström, 1995)³¹ coupled with a river routing model to be suitable for the purpose of modeling a spatially distributed watershed system.

According to the authors of the last model “Previous modeling studies show partially contradictory methods and findings. For example, significant differences were found in the “water level‐surface area‐storage” relationship of the Ihefu (eastern) wetland. Also, rating curves of “water level‐discharge” at N’giriama vary between studies.” They conclude that “The probable reason is that the Usangu plains and the upper Rufiji River are extremely complex from a hydrological perspective as well from a water use perspective, and suffer data scarcity, which partly explains the contradicting results from studies over the past 20 years.”

Treshold Drainage Usangu Rice Plain Preventing ZF at Park HQ

The third model uses zero flow predictions for the Ihefu Swamp outlet while the low flow estimate of the rating curve at the Msembe Gauge Station is not considered adequately reliable³² . The N’giriama (Ihefu Swamp) Outlet flow status is considered..”more reliable and probably more representative to the actually experienced period of ‘dry river’ in the Ruaha National Park.” It is noted that the “observed” zero flow days at N’giriama are based on “satellite observation” while the simulated water levels are based on water level data communicated by C. Birkett. The correlation coefficient of the by this study, predicted N’giriama and measured/observed Park HQ zero day periods, based on the flow database of the Ministry and the Park Staff respectively in the period 1994-2016 is 0.04 and 0.21. Whether the N’giriama zero overflow is more representative for the actually experienced period of dry river at the Park HQ is doubted.

From the models the following treshold flow values in the Ruaha are extracted.

Table 1.: Comparison Treshold Drainage Levels Usangu for 3 selected Models

Drainage Usangu at	SMUWC1)	Open Channel Model2)	HBV_DS Model3)
Inflow Eastern Wetland (Ihefu)	lowering abstraction “Usangu Rice” with 7 cumecs	4cumecs (drainage)+ drainage swamp	?
Outflow Ihefu Swamp	>7cumecs from Usangu in dry season results in >0.3 cumecs (1999-2000)	Outflow 0.5 cumecs which equals loss Ruaha section N’giriama to Park HQ	min. 2 cumecs used for assessing impact on ZF Park HQ
Park HQ (Msembe)	no zero flow if drainage >7 cumecs + reference drainage (?)	no zero flow with >0.5 cumecs outflow N’giriama or > 4 cumecs from Usangu Rice Plain	It is assumed that this overflow at Ihefu relates to no zero flow at park

Based on the SMUWC model, the WWF (2010) study, “Environmental Flow Assessment”³⁴ recommends a minimum inflow for the Eastern Wetland (Ihefu) of 5.52-6.81 cumecs (Drainage from Usangu Rice Plain). This minimum flow is to sustain an outflow at N’giriama of 1-2 cumecs and “minimum” (flow) at Park HQ (Msembe) and Muhuwa (Haussman’s Bridge ? 50km u/s Msembe) of 1 cumecs. It is assumed that considering the losses along the N’giriama – Msembe section of the Ruaha River that more than 6 cumecs of inflow/drainage Usangu Rice Plain is required according to the study.

The Struggle for Meaningful Predictions

All attempts to model the low flow hydrology of the Usangu Rice Plain are handicapped by the quality of the low flow data, the missing and dubious data series available for the inflow, drainage and flow at Park HQ. The quality for the low flow data, in the Ruaha River which matter most for the analysis of Zero Flow at the Ruaha National Park is even more discouraging.

The most important model variable in the combat with zero flow in the Ruaha Park, is the drainage flow from the Usangu Rice Plain measured at Nyaluhanga. Unfortunately the available data are inadequate for proper low flow hydrology analysis. The rating curve (January 1, 2000 to December 31, 2016) is based on 21 flow measurements between 0.2 and 6 cumecs with a R-squared of 39% and a RMSE of 3 cumecs in the above low flow range. The 13 low flow series for 2001 to 2013 from June 1 to the arrival of the rains, has 11 series (excluding 2008 and 2013) and which cover the complete low flow period up to the minimum or zero flow, of which seven have serious irregularities.

In addition the occurrence of zero drainage events observed during at least 10 of the 13 low flow periods in the section of the Ruaha River between Nyaluhanga and the Ihefu Swamp (eastern wetland) is not reflected by the flow data series from the gauge station at Nyaluhanga.

For more detailed description of the quality of the data and missing flow data visit this link.

The sum of the annual minimum inflow rates for the Ruaha and Mbarali Rivers on the Usangu Rice Plain, during years without zero flow (mostly before 1990) is as low as 4 to 5 cumecs. The sum of annual minimum smaller than 4 cumecs occur mostly after 1990 which are the years with periods of zero flow at the Park.

This combined minimum flow seems rather low especially because it will be further reduced by diversion, spill and riverine “losses” before it reaches the drainage station of Nyaluhanga considered as the Usangu Rice Drainage station. Due to the duration of the flow to reach the Park HQ the threshold inflow associated with the zero flow will pass the inflow gauging stations some 4 weeks before the date of the first observation of zero flow at the Park HQ. In most low flow seasons this will be some weeks before the annual minimum flow at the inflow stations is registered. During the low flow period the decline per month is around 1 cumecs which means that with 6 to 7 cumecs inflow on the Usangu Rice Plain, the critical inflow level, associated with zero flow at the Park HQ, is reached.

The remote sensing analysis³⁵ of the low flow season abstraction, spill and riverine/percolation “losses” for the Mbarali and Ruaha subcatchments, based on evapo(transpi)ration for three days in 2019 results in a “loss” of 3.5 cumecs on July 2 and gradually reducing that year to 1 cumecs on October 22. The resulting treshold drainage value associated with zero flow in the park which was first observed on November 20, 2019 (=Inflow Usangu some 1 month before occurrence Zero Flow Park HQ – minus “losses” on the Usangu Rice Plain) will be in the range of 4 cumecs which corresponds with the range predicted by the models with the lower level of critical drainage flow for prevention of zero flow.

The RS Ruaha Flow Monitor Approach avoids working with missing or poor low flow data

Although it is common practice in water management, to base interventions on assessed flow rates, the remote sensing technique developed for the Ruaha River (zero) flow monitoring avoids basing its conclusion on ireliable flow data which may introduce incorrect conclusions.

Even in the absence of reliable (low) flow rates the remote sensing monitor can still provide relevant alerts as is demonstrated in Table 2.

The analysed trial runs of the Ruaha Flow Monitor in 2019 to 2022 show that the alerts are adequately in time for the required management interventions to prevent or reduce zero flow periods.

The alert phases also start when most or all rice irrigation has been completed. In fact if late rice irrigations will increase the risk of zero flow the only conclusion should be that too much rice area has been cultivated. Based on rainfall patterns and intensities also the area for responsible rice cultivation can be forecasted.

The water authorities alerted by the monitor will have sufficient time to approach farmers and canal groups to correct uncontrolled abstraction and spill in their area. The salvaged spilled water will boost drainage from the plain for continuous flow in the Ruaha in the Park. In normal years adequate response to the Alerts by the authority can prevent zero flow while in extremely dry years it will reduce the period of zero flow.

Table 2.: Days to Zero Flow at Park HQ

Alert Levels	2018	2019	2020	2021	2022	2023	2024
Alert 1	131	170	119**	120	70	82	95
Alert 2	102	140	85**	75	35	52	55
Alert 3	69	37	47**	17	22	19	52
Dry Season Park*	June 29 to Dec 15	June 3 to Dec 2	Aug 12 Dec**	July 16 to Jan 21 (2022)	June 12 to Jan 23 (2023)	July 27 to Dec 2

The release of the Rainfall (IMERG v07) data expected from early 2023³⁶ is important for assessment of the role the rains play during the preceding rainy season and during the dry season on the occurrence and length of the 3 Alerts periods. Also the total water in the system (wet season plus dry season rains) which relates to the abstraction/spill levels but also to the intensity and spread of the rains, seem to matter on the length of the Alert phases as is concluded from the 5 years of trial runs for the flow monitor Table 2.

The IMERG version 6 for monthly rain data for the upper reaches of the Ruaha River which proved to correlate well with the occurrence of zero flow at Mtera, stopped being produced in September 2021.

The IMERG version 7 rainfall data are now available. The total precipitation for November and December of the rainy season preceding the zero flow at Mtera is now selected for the prediction in 2024. With the IMERG rain data available in April adequate time is available to alert the authority well in advance. The formula for prediction of 1st day of zero flow is:

DoY 1^st zero flow =0.294*(total rain in mm) + 161

For details on the IMERG data, visit NASA’s Global Precipitation Measurements project website