Modeling Sediment Dynamics: Effect of Land Use, Topography, and Land Management in the Wami-Ruvu Basin, Tanzania

Abstract

Soil erosion and soil conservation have been major issues in Tanzania as it has been the case of many other tropical countries. Policy makers have identified soil erosion as a critical problem since the 1920s. However, it has been difficult to obtain reliable data on the type, extent, and current rates of soil erosion and sedimentation. The limitation of such information has delayed the current and future interventions for soil and water conservation in critical areas throughout Tanzania. The main objective of this study was to test the sediment prediction capability of the Water Erosion Prediction Project (WEPP) model on tropical watersheds and also identify erosion hotspot areas. Simulation of this initial study in Tanzania focused on secondary data. There was insufficient information about on-site soil properties, daily rainfall and temperature records, and initial condition of land use/cover data to define crop/plant growth and tillage practices. Runoff also varied with soil type in all four watersheds. The highest and lowest total average annual soil loss rate was estimated in Mfizigo Juu, 45.09 kg/m² and Kibungo chini, 0.45 kg/m², respectively. The cultivated land contributed to more than 81 % of soil loss and 86 % of sediment yield in all four scenarios. The overall spatial result maps indicated WEPP model can help water resources managers to implement necessary precaution measures to prevent sediment yield and soil erosion.

1 Introduction

Water resources management in recent years is facing many challenges including pressure on water demand. A number of studies have insisted the engagement of multisectoral natural resources management as cross-cutting agenda. Most of these studies focused on assessing quality and quantity of water available without linking to effects of land use, topography, and land management. Land degradation leads to the reduction or loss of biological or economic productivity of land caused by deterioration of physical, chemical, and biological or economic properties of soil. According to Hillel (1991) , large-scale degradation of land resources has been reported from many parts of the world in different figures depending on the variation of causing factors. The economic impact of land degradation is extremely severe in densely populated areas of South Asia and sub-Saharan Africa that account for 70% of the total degraded land of the world (Dregne and Chou 1994) . It has been observed that about 75% of soils in hilly areas are the most susceptible due to sheet, rill, and gully erosions (Hasan and Alam 2006) . In addition, human alterations of land use have caused erosion rates to increase in many areas of the world, resulting in significant land and environmental degradation. However, in East African highlands, soil loss rate is reported to exceed the tolerable recommended limit (10–12 t/ha/year) by 50 t/ha/year (Kimaro et al. 2008) . Since in Tanzania information and resources are limited, education to raise awareness on soil conservation seems to be more important to start with (Rapp et al. 1973) . In addition, policy makers have been asking for quantification of erosion rates at local, regional, and global levels in order to develop environmental and land use management plans which will consider both on-site and off-site impacts of erosion (de Vente et al. 2008) . However, because of limited resources, the Tanzanian government decided to concentrate only on land use management rather than implementing extensive projects for soil erosion prevention and conservation (Rappet al. 1973) .

Tanzanian water resources are managed at the basin level by implementing the concept of integrated water resources management (IWRM). IWRM practices vision ensures that water resources in a basin are sustainably managed for socioeconomic and environmental needs. The Wami-Ruvu basin is one of the nine basins in the country. Currently, water quality management at the Wami-Ruvu basin is facing challenges related to land use and human activities as a result of rapid population increase. The Morogoro District Council that occupies the Upper Ruvu catchment had a population of about 304,019 in 2011 with an average population density of 25 people/km² in 2000 at the Upper Ruvu catchment (Mt. Uluguru). The catchment accommodated 60% of the total population with a population density of 250–300 people/km² (URT 2011). Ruvu River catchment has been affected by agricultural activities toward Uluguru Mountains as a result of population increase.

Wetlands available in lower Ruvu plains have been affected by increase of sediments (Ngoye and Machiwa 2004) . In addition, observed increase of turbidity from 130 NTU in 1992 to 185 NTU in 2002 in the Ruvu River subbasin was a result of the increase in agricultural activities (Yanda and Munishi 2007) . Deforestation on the other hand is another major problem facing Uluguru, Ukagulu, Nguru, and Mgeta mountain forests in Ruvu subbasins derived from demand for timber cutting, collection of firewood, and land clearing for agricultural purposes and timber production. However, the forests and woodland cover in the Uluguru Mountains has decreased by 12.7 and 59.0% in 1995 and 2000, respectively (Yanda and Munishi 2007) .

Different studies have been conducted to explain the processes of erosion, identify the major factors influencing the processes, and also develop models suitable to quantify the processes. Most of these models have been “on-site impact oriented” by identifying loss of soil from a field. The breakdown of soil structure, and the decline of organic matter and nutrients, results in a decline in soil fertility and a reduced food security and vegetation cover. Furthermore, “the off-site effects” of erosion include sedimentation problems in river channels, increased flood risk, and reduced lifetime of reservoirs (de Vente et al. 2008) . Most available studies and modeling have limitations in the applicability and adaptability to tropical largerwatersheds (Ndomba 2007, 2010) . The challenge to slopes and topography is reported to affect sediment loading as compared to total watershed sediment yield. Although several soil erosion modeling studies have been adapted in larger tropical watersheds, validation with field measurements gave unreliable estimates (Ndomba 2010) . Therefore, this study aims at contributing to the development of a sediment assessment model/framework usingWater Erosion Prediction Project (WEPP) in Upper Ruvu subbasin. This study gives insight into the water quality problems related to sediment loading in streams to Kibungo catchment. It specifically aims to estimate the quantity of sediment yield delivered from the upland areas and identify areas that will benefit most from soil conservation practices. Quantification of sediment yield and upland/hillslopes and catchment runoff using the WEPP model will unfold information required for the ongoing effort to enhance riverwater quality management as an integrated component of water resources management of theWami-Ruvu basin.

2 Study Area

Wami-Ruvu Basin covers approximately 66,820 km² with three subbasins which are theWami River subbasin 43,946 km², Ruvu catchment 18,078 km², and coastal rivers 4,796 km² (Fig 10.1). This study focused on the Upper Ruvu River subbasin where there is an ongoing implementation of the Kidunda dam for Dar es Salaam city and coast region water supply. The subbasin is located between latitudes 6°05′S and 7°45′S and longitudes 37°15′E and 39°00′E. The rivers that flow in this subbasin are Mvuha, Mfizigo, Ruvu, and Mgeta whose headwaters are in the Uluguru and Mgeta mountains 2,634m above sea (WRBO 2010). Ruvu is the main river fed by its tributaries Mgeta, Mfizigo, Mvuha, and other small streams. The regime of the Ruvu River reflects the trend of the wet and dry seasons. According to the basin office and Ministry of Water report, the river flows at the Kidunda and Mikula stations decrease from about 60 m³/s in May to around 25 m³/s in October or September. After this month, it rises slowly reaching 70 m³/s in December. In January and February, the flows arrive at 60 and 50 m³/s, respectively. The highest monthly average flow is reached in April at around 160 m³/s. The lowest value of about 5 m³/s has been reached in October. The mean annual flow is approximately 66 m³/s (SP Studio Pietrangel Consulting Engineers 2010). To fulfill the objectives, this study focusedon the Kibungo catchment. For a detailed study, as required by WEPP model , the subbasin was delineated further to get four catchments which are Mgeta, Kibungo, Ngerengere, and Ruvu. The catchments were selected considering the type of regulation (natural or modified) and the region geographical importance.

Fig. 10.1

Location of the study area

3 WEPP Model Overview

The WEPP is aimed at developing process-based prediction technology to replace the universal soil loss equation (USLE). The WEPP model operates on a continuous daily basis by using mainly physically based equations (Baigorria and Romero 2007) . It describes hydrologic and sediment generation and transport processes at the hillslope and in-stream scales (Baigorria and Romero 2007) . The basic WEPP hillslope model components are weather generation (climate), surface hydrology, hydraulics of overland flow, hillslope erosion, water balance, plant growth, residue management and decomposition, soil disturbance by tillage, and irrigation (Foster and Lane 1987) . With limitations in availability of data to this study, main input components are weather generation, surface hydrology, hydraulics of overland flow, hillslope erosion, residue management and decomposition, and soil disturbances. A unique aspect of theWEPP technology is the separation of the erosion processes into rill detachment (as a function of excess flow shear stress) and interrill detachment. Additionally, the model simulates sediment transport and deposition, and off-site sediment particle size distribution. These items allow better assessment of soil erosion at a site, and subsequent sediment transport to channels and impoundments in catchments (Foster and Lane 1987) .

WEPP has been tested and applied in different geographic locations across the world out of the USA. In Peru (Baigorria and Romero 2007) , the modelwas validated using three different-sized runoff plots, at four locations under natural rainfall events. According to Baigorria and Romero (2007) , all climatic characteristics, soil physical parameters, and topographical and management characteristics were determined in the field and laboratory. The measured runoff and erosion from agricultural fields were low compared to predicted levels. A poor relationship between runoff and sediment yield as well as rainfall and runoff was observed, and this poor relationship was observed because of the dynamic change in soil properties during rainfall events due to sealing (Baigorria and Romero 2007) .

In Africa, with little modifications, WEPP has been used in mountainous catchments. For instance, WEPP has been tested in the Anjani catchment, Ethiopian highlands (Zeleke 2001) . The emphasis was on the new standalone program to create a climate input file for WEPP using standard weather data sets called breakpoint climate data generator (BPCDG). However, the final results overpredicted runoff and underestimated soil loss. In addition to that, validation of the model was done in Kenya at Amala and Nyangore upstream watersheds of Mara River basin (Defersha and Melesse 2012; Defersha et al. 2012) . Simulated runoffwas reasonably compared to observed results. However, sediment yields and erosion were fairly simulated for different land use areas as expected (Deferesha and Melesse 2012; Defersha et al. 2012). In North Africa, the analysis of model performance on sediment yield and runoff predication was conducted on Mediterranean cultivated Kamach catchment, Tunisia. In the Minnesota River basin, WEPP was applied to estimate sediment load as well as evaluate the effect of subsurface drainage using tiles on runoff and sediment transport (Maalim and Melesse 2013; Mallim et al. 2013). .

3.1 Predicting Runoff (Interflow and Base Flow)

Hydrology within the watershed reflects on the effects of water balance, channel hydrology, and soil effects. WEPP mathematical calculations of channel hydrology and water balance are the result of infiltration, evapotranspiration, soil water percolation, canopy rainfall interception, and surface depressional storage. The model uses the Green-Ampt Mein–Larson approach to simulate the temporal changes in infiltration rate during the rainstorm (Ascough et al. 1997) . Runoff which is rainfall excess occurs when rainfall rate exceeds infiltration rate. This is assumed to start after the depression storage is filled. The differential form of mathematical equation for soil matrix of infinite depth is

$$i={{k}_{e}}+(1+\frac{(\phi -{{\theta }_{{}^\circ }}){{\varphi}_{c}}}{l})$$

(10.1)

where

i :

Actual infiltration rate (m s⁻¹)

k _e :

Effective hydraulic conductivity of the wetted zone (m s⁻¹)

${{\theta }_{{}^\circ }}$ :

Initial saturation (m³ m⁻³)

$\phi$ :

Effective porosity (m³ m⁻³)

${\varphi _c}$ :

Effective capillary tension or wetting front suction potential (m)

l :

Cumulative infiltration (m)

Water balance is based on a component of the simulator for water resources in rural basins (SWRRB) model with some modifications for improving estimation of percolation and soil evaporation parameters (Ascough et al. 1997) . The distribution of water through soil layers is based on evapotranspiration percolation models and storage routing techniques. If the potential surface storage depression is completely satisfied, the positive difference between the net rainfall intensity at the ground surface and the infiltration rate becomes the input to the overland flow calculation (Defersha et al. 2012) . The basic equations which describe the movement of water are based on the laws of mass and momentum conservation:

$$ \frac{{\partial A}}{{\partial t}} + \frac{{\partial Q}}{{\partial x}} = r(t) - i(t) = q(x,t) $$

(10.2)

and

$$ Q =\propto P{R^{m - 1}} $$

(10.3)

with

$$\propto= {\rm{C}}\sqrt {\rm{s}}$$

(10.4)

where

A :

Cross-sectional area (m²)

t :

Time (s)

Q :

Discharge (m³/s)

x :

Down slope distance (m)

r :

Rainfall intensity (m s⁻¹)

i :

Local infiltration rate (m s⁻¹)

q :

Lateral inflow rate (m s⁻¹)

R :

Hydraulic radius (m)

P :

Wetted perimeter (m)

m :

Depth-discharge exponent Chezy: m = 3/2, Manning: m = 5/3

α :

Depth-discharge coefficient(m^1/2 s⁻¹)

s :

Average slope (mm⁻¹)

In WEPP, the overland flow is conceptualized as plane runoff which means that A is substituted by the average flow depth h (expressed in m) . Equations 10.5 and 10.6 are solved analytically by the methods of characteristics which require the rewriting of these equations as differential equations on characteristics curve on the x–t plane :

$$ \frac{{dh}}{{dt}} = v(t) $$

(10.5)

and

$$ \frac{{dx}}{{dt}} = \alpha mh{(t)^{m - 1}} $$

(10.6)

where

h :

Flow depth (m)

v :

Runoff or rainfall excess (m s⁻¹)

These equations are solved together with the infiltration calculations by using a Rungge–Kutta iteration scheme with as spatial resolution of one hundredth of the total hill slope length and a time step of 1 min.

3.2 Predicting Sediment Concentration

Hillslope Erosion (Rill and Interrill) The WEPP model divides erosion into two types: rill and interrill erosion. The movement of the sediment along the hillslope is described on the basis of the steady-state sediment continuity equation which is applied flow conditions (Elliot et al. 1995; Flanagan et al. 2007; Nearing et al. 1989) :

$$ \frac{{dG}}{{dx}} = {D_i} = {D_r} $$

(10.7)

where

G :

Sediment load (kg sec⁻¹ m⁻¹)

x :

Distance down slope (m)

D _r :

Rill erosion rate (kg sec⁻¹ m⁻¹)

D _i :

Interrill erosion rate (kg sec⁻¹ m⁻¹)

The interrill erosion is estimated from the equation

$$ {D_i} = {K_i}{I^2}{S_f}f(C) $$

(10.8)

where

D _i :

Detachment rate (kg sec⁻¹ m⁻²)

K _i :

Interrill soil erodibility parameter (kg sec⁻¹ m⁻⁴)

I :

Effective rainfall intensity (m sec⁻¹)

S _f :

Slope factor (m sec⁻¹)

f(c):

Function of canopy and residue

The erosion rate in rill erosion is a function of hydraulic shear and amount of sediment already in the flow (Elliot et al. 1995; Nearing et al. 1989) . Rill is estimated in WEPP model by:

$$ {D_r} = {K_r}(t - {t_c})(1 - \frac{G}{{{T_c}}}) $$

(10.9)

where

D _r :

Rill erosion rate (kg sec⁻¹ m⁻²)

K _r :

Rill soil erodibility parameter (sec m⁻¹)

t :

Hydraulic shear of water flowing in the rill (Pa)

t _c :

Critical shear below which no erosion occurs (Pa)

G :

Sediment transport rate (kg sec⁻¹ m⁻¹)

T _c :

Rill sediment transport capacity (kg sec⁻¹ m⁻¹)

WEPP model is based on modern hydrological and erosion science. It calculates runoff and erosion on daily basis (Baigorria and Romero 2007) . Its simulations can be enhanced by using digital sources of information through the linkage with geographic information systems (GIS). The GeoWEPP model is a geo-spatial erosion prediction model developed to incorporate advanced GIS features (ArcGIS software and its Spatial Analyst extension) to extract essential model input parameters from digital data sources.

GeoWEPP is designed to integrate four different data for accuracy of WEPP process-based model. The main inputs include topography, soil, land use, and climate information while the basic maps required should be in American Standard Code for Information Interchange (ASCII) formats exported by ArcGIS. The scales and resolution of the spatial inputs can vary according to the variable.

4 WEPP Application

In this study, data requirements were in accordance to GeoWEPP for ArcGIS 9.x FullVersion Manual for GeoWEPPVersion 2.2008. The GeoWEPP package includes two tools that further expand its utility. These are the topographic parameterization (TOPAZ) tool and Topwepp software products developed by the US Department of Agriculture/Agricultural Research Service (USDA-ARS). The TOPAZ generates hillslope profiles by parameterizing topographic data using a given digital elevation model (DEM). This process provided the needed input data for the subsequent delineation of a watershed, subcatchments, flow direction determination, and channel network generation. Topwepp uses grid-based information stored in the raster layers of the land cover; soil and land use management to execute the model runs and produces the output maps.

4.1 Topography

The DEM of 30 m resolution (Fig. 10.2b) was downloaded from theAdvanced Spaceborne Thermal Emission and Reflection Radiometer Global DEM (ASTER GDEM) website. TheASTER GDEM provides topographic information of the global terrain. A polygon representing study area was used to extract study area DEM from the original data set using the Spatial Analysts tool. The advantage of using this high resolution of 30m is its applicability for GeoWEPP modeling. Figure 10.2a shows the slope values of the study area in percent and existing rivers used to delineate catchments in GeoWEPP The projected DEM was changed to raster format and then converted to ASCII format using the ASCII conversion tool in ArcToolbox of ArcGIS.

Fig. 10.2

(a) Slope map and (b) population increase toward high elevation in Kibungo subwatershed

4.2 Soil Data

The geology of the catchment is influenced by the Precambian Usagarian system that has suffered different plutonic histories and Neogene. The area contains Jurrassic, Karoo, Neogene, and Quaternary strata in some parts of the catchment. There are different types of soils in the upper river basin that vary in texture from sand to clay. Soil map of the study area was downloaded from the Harmonized World Soil Database (HWSD) from the Food and Agriculture Organization (FAO) with a scale of 1:2,000,000. The data were in tagged image file format (TIFF) converted to grid and reprojected in similar cell size and resolution of DEM and then converted to shape file. The different soil types in the upper river basin were classified as per the Soil Terrain Database of East Africa (SOTER) classification. The main soil orders are Fluvisols, Cambisols, Leptosols, Acrisols, Ferralsols, and Vertisols. Three essential soil files required in the model (anASCII and two text files) were generated. The shape file was converted to raster format then converted to ASCII format. The soil data, the soilmap.txt, and soilmapdb.txt were created with the map unit key and the map unit symbol corresponding to the raster value and description, respectively.

4.3 Land Cover Data

The land cover in the study area is characterized by various types of natural vegetation. Cultivation is the main land use activity in the catchment. Human settlements in the most steep area range from small towns to villages (Ngoye and Machiwa 2004) . A high percentage change of land cover from 1990–2000 caused by agriculture and settlementwas observed (Fig. 10.3). The 1990 land use/land cover used in simulation considered as presettlement period to give snapshot erosion effects at that period. The shape files converted to raster format making sure the projection and cell size are similar with DEM and then exporting the attribute table values to text file. The raster also was converted to ASCII format to be used in modeling.

Fig. 10.3

Land use changes in the Upper Ruvu catchment

4.4 Climate Data

Monthly climate data for the study area were obtained from Tanzania Meteorological Agency (TMA) and Morogoro Weather stations as recorded by Wami Ruvu Basin Water Office (WRBWO). For this study, a total of ten available meteorological/weather stations were selected, namely, Matombo Primary School, Hobwe, Morning side, Mikula, Kisaki, Ng’ese Utari, Duthumi Singisa, Mtamba, and Ruvu at Kibungo (Fig. 10.4). Stations selected were those within or near the study area and with 30 or more years (between 1950 and 2005) of monthly and some with daily data acquired for further processing (Fig. 10.5). Based on information from these gauges, the mean annual rainfall varies from 900 to 1,300mm and daily temperature ranges between 22 and 33 °C.

Fig. 10.4

Rainfall variation with elevation: upstream and downstream of the Kibungo subwatershed

Fig. 10.5

Rainfall variation in the Upper Ruvu catchment as recorded in eight weather stations (1950–2005) average: a annual, b monthly

Maximum and minimum daily temperatures and precipitation depths are required as model inputs. The data were analyzed by the Parameter-elevation Regressions on Independent Slopes Model (PRISM) tool which allows modification of an existing WEPP climate parameter file—the files WEPP uses to generate the climate events for a simulation. PRISM allows this modification to the WEPP climate parameter files so that it can more closely match the climate found in area of interest (Minkowski and Reschler 2008) . The information in PRISM files are monthly temperature, precipitation, and wet days. Also, there is climate station name, elevation, and its location (latitude and longitude). For the purpose of this study, average monthly rainfall measurements from all stations were converted to inches to allow comparable modification process.

5 Experimental Procedures

Field measurements for water quality are important to determine whether significant changes occurred with time. The quality of data depends on sampling protocols including methods, time interval, documentation, and purpose of field measurements. The accurate measurement and calculation of suspended transport depend on the time and sampling procedures used. In theWami-Ruvu basin, there is limited continuous sediment data of its catchments. The data available are event based and most were taken during rainy season. In the case of this study, the field measurement was done for the purpose of getting a snapshot of total suspended solids (TSS) in dry season in the area where secondary streams converges in most hill slopes. Although the results are on a weekly basis, the data will assist in proposing sediment sampling locations for monitoring.

5.1 Hydrological Data

Sediment sampling locations were set up in order to obtain insight into the spatial variability of suspended solids concentration (SSC) in the river systems. Continuous/throughout the year accessibility to sampling locations was considered in line with assessment of the spatial variation in sediment response. The sampling points were selected as much as possible on bridges along the roads that cover the study area allowing easy access to collect river water samples (with suspended sediment) and to carry out streamflow velocity measurements. Seven sampling sites of Ruvu at Kibungo, Mgeta at Duthumi, Mgeta at Mgeta, Mfizigo at Lanzi, Mfizigo at Kibangile, Mvuha at Tulo, and Mvuha at Ngangama were selected along the Upper Ruvu River and its major tributaries.

At each station, flow velocity measurements were done by the Acoustic Digital Current (ADC) meter (wading method) and the Q-liner instrument as shown in Figs. 10.6 and 10.7. Flow velocity measurements were normally taken at 60% of the water depth (0.6D) at regular intervals along the cross section in order to establish a stage discharge rating curve. For suspended sediment loading analysis, water sampling at each station of the river was demarcated into 3–5 sections in which 3–5 samples for analysis of suspended solids were taken by using a D-48 sediment sampler and D-74 integrating suspended handline sampler into a labeled container for laboratory analysis.

Fig. 10.6

D-48 and D-74 sediment sampler used during field research

Fig. 10.7

Field sampling photos and source of the diagram

5.2 Suspended Sediment Load

Analysis of suspended sediment load in water samples was carried out at the Soil Laboratory of Sokoine University of Agriculture in Tanzania. Filtration of water samples for suspended solids was done by using vacuum pressure pump fitted with glass fiber of 0.45-μm-diameter membrane filters. The membrane filters were initially dried in the oven at 70 °C for 24 h and weighed (in grams) using a sensitive balance. The water samples were filtered, and then the wet filters were dried in an oven at 103–105 °C for 1 h. The weights in grams of the filters with dried residue were noted. After the laboratory analysis, the amount of suspended solids in each sample was calculated using the formula

$$ T = \frac{{(A - B)*1,000}}{C} $$

(10.10)

where

T :

Total suspended solids (mg/l)

A :

Weight of filter with dry residue in (mg)

B :

Dry weight of filter in (mg)

C :

Sample volume (ml)

The total suspended load (milligram per second) was calculated by multiplying by river flow at crossing area in cubic meter per second and then changed to kilogram per second by multiplying by 1,000.

6 Results and Discussion

6.1 Simulation Outputs

The model was simulated in four subwatersheds delineated by the model. Subwatershed delineation was done to select a channel cell as an outlet point. The number of hillslopes and streams were calculated and checked before simulation to compare with model requirements. The names of subwatersheds were given to relate to the rivers/streams drained in a particular subwatershed.A total number of four subwatersheds, namely Mfizigo, Msumbizi, Mvuha, and Kibungo, were generated. GeoWEPP results of runoff, soil loss, sediment deposition from hillslopes, and channels are displayed as text files and sediment yield is visualized as a map showing hotspot areas by subwatersheds that are very vulnerable to soil erosion. The discussion is based on the effects of land cover change, topography, and land use management.

6.1.1 Average Annual Runoff

Effect of Topography and Land Use/Land Cover:

Average annual runoff volume on cropland, open woodland, and grassland is shown to be greatest in most of all subwatersheds as shown in Fig. 10.8. In these areas, the surface has been paved or soil is no longer retaining water which leads to rainfall to be converted to runoff. The low runoff indicated in land use/cover is characterized by shrubs/bushland and natural forest. The model simulated high amount of runoff in Mfizigo Juu (11,247 m³/ha/year) and Mvuha (9,293 m³/ha/year) as it is caused by higher elevation and substantial land use/land cover contributing factors (Fig. 10.8). Mfizigo Juu is characterized by 41% of woodland at Tegetero, Kibogwa, and Kinole wards which makes runoff to be high as related to land use/land cover. Although there are patches of natural forest that can be found in Kibungo Juu and Mkuyuni wards, still high rainfall caused high runoff volume. Most of Mvuha subwatershed has been converted into agriculture and grassland in larger areas of the Kasanga, Kolero, and Mvuha wards. However, being in lower elevation makes the area to be a flooding area of runoff volume from the highland. Figure 10.8 shows Msumbizi and Kibungo chini subwatersheds with low average annual runoff volume. Msumbizi area, which covers Tununguo and Kiroka wards, is more characterized by woodland (58 %) and natural forest (16 %), while Kibungo chini area is characterized by open woodland with cultivation (94 %). The average annual runoff volumes at Msumbizi and Kibungo chini are 5,689 m³/year and 4,578 m³/year, respectively. Moreover, runoff depth was estimated under four scenarios categorized considering two levels of topography steep slopes and lowland areas. These four scenarios considering soil types of the high coverage percentage in the area such as Ferralic Cambisols and Humic Acrisols are the major soil types in the highland areas (Mfizigo Juu) while Rhodic Ferralsols and Eutric Leptosols are found in lowland areas (Kibungochini). Simulation results showed some variations with land use/land cover both in highland steep slope and lowland slope areas. Results showed that cultivated land to have the maximum average annual runoff depth of 1,135 mm. The minimum average runoff depth was 51mm represented by bushland and natural forest areas. A summary of results of the average annual runoff depths for all scenarios as related to land use/land cover are shown in Figs. 10.9–10.12.

Fig. 10.8

Average annual runoff variations within the subwatersheds as estimated by the model

The average annual runoff volume as estimated by the model showed good correlation with subwatershed characteristics. It is indicated that the amount of runoff in Mfizigo and Mvuha is high caused by upland high slopes with open woodland land cover and some patches of agricultural land (Table 10.1). Runoff of Msumbizi was low, although the subwatershed is located at high slope but characterized by the natural forest land cover. Discharge at the outlet of watershed depends on drainage area, inflow, or outflow of groundwater to or from the surface area.

6.1.2 Soil Loss

Effect of Land Use/Land Cover:

Soil loss seems to be affected by land use/land cover, runoff, and topography in all subwatersheds. The areas with high slopes tend to increase soil loss in different land use/land cover. WEPP modelwatershed simulation predicated minimum average soil loss rate in grassland and natural forest areas. The minimum annual soil loss ranged from 0.33 to 9.14 kg/m². The reason for these estimates was the coverage area in watershed for the cultivated land use/land cover to be small compared towoodland and natural forest. The maximum average annual soil losswas inwoodland and cultivated land as compared to natural forest/bushland with minimum average soil loss 0.5 kg/m² in steep slope and 0.1 kg/m² in lowland areas (Table 10.1). However, although soil in woodland is covered, the area is classified as disturbed land cover with severe fire every year, leaving bare land and reducing its erosivity and increasing erodibility by water.

Table 10.1 Summary of variation in soil annual loss and sediment yield with land use/land cover under different scenarios

Moreover, soil loss seems to be high in areas with high slopes (scenarios 1 and 2) and low in low slope areas (scenarios 3 and 4). The reason for this much difference is the effect of a runoff volume. Reduction of runoff depth from 1,135mm (1.135 m) in scenarios 1 and 2–51mm (0.051 m) in scenarios 3 and 4 resulted in a decrease of soil loss from 45.5 to 10.7 kg/m² under cultivated land (Table 10.1). Even though there are variations of soil loss, yield, and runoff in all four scenarios, land use/land cover showed sensitivity in all factors considered in modeling.

Fig. 10.9

Variation of soil loss, sediment yield, and runoff with land use/land cover under scenario 1 in steep slope area as estimated by the model

Fig. 10.10

Variation of soil loss, sediment yield, and runoff with land use/land cover under scenario 2 in steep slope area as estimated by the model

Fig. 10.11

Variation of soil loss, sediment yield, and runoffwith land use/land cover under scenario 3 in low slope area as estimated by the model

Fig. 10.12

Variation of soil loss, sediment yield, and runoffwith land use/land cover under scenario 4 in lowland area asestimated by model

6.1.3 Sediment Yield

GeoWEPP model generated sediment yield map which indicates the area with tolerable yield (i.e., from light to dark green) and not tolerable yield (i.e., from light to dark red). Results given from the model simulation are in a tolerable maximum value (T-Value) of 1 t/ha/year. Results were converted to 12 t/ha/year, the EastAfrican highlands soil loss rate-tolerable limit as reported from previous studies (Kimaro et al. 2008), in order to get clear visualization of maps. Sediment yield was categorized in eight groups; four displayed in green color are below the tolerable soil losses. The other four categories are displayed in red color and are above the tolerable value which indicates areas with high level of yield as shown in (Figs. 10.13–10.16). Since sediment yield is a function of contributing factors including topography, land use/land cover, runoff, and land management practices, and the simulated results in four subwatersheds prove good relation with regard to these factors. .

The simulated average annual sediment yield at the Mfizigo Juu watershed outlet was 113,009,137.60 t/year and sediment delivery ratio for the watershed was 0.30. The contributing area was 81,113.82 ha which includes 1,023 hillslopes and 411 channels. Average annual sediment yield from Msimbizi subwatershed was 183,430.00 t/year and the predicted sediment delivery ratio for thewatershedwas 0.2. Simulation created 272 hillslopes and 109 channels in 21,217.66 ha of contributing area. Mvuha and Kibungo chini are subwatersheds at the downstream of Kibungo watershed. These areas showed 418,459,188 t/year and 270,658.9 t/year, respectively. Results showed that 461 hillslopes and 185 channels in Mvuha, and also 78 hillslopes and 31 channels in Kibungo Chini. Sediment delivery ratio in Mvuha was 0.66 from 36,871.54 ha contributing area while computed delivery ratio in Kibungo chini was 0.87 with a contributing area of 5,478 ha. As shown in Figs. 10.13–10.16, higher values of soil losses from high rainfall and high elevation are expected in the zones with red color, whereby the minimum losses are in the zone with green color.

Fig. 10.13

Map of average annual sediment yield for Mfizigo subwatershed

Fig. 10.14

Map of average annual sediment yield for Mvuha subwatershed

Fig. 10.15

Map of average annual sediment yield for Msumbizi subwatershed

Fig. 10.16

Map of average annual sediment yield for Kibungo Chini subwatershed

6.2 Field Results

Total Suspended Solids and Average Annual runoff: During the field data collection campaign, the sediment load was observed to be higher at Mgeta at Duthumi (MD), Mvuha at Ngangama (MN), and Mvuha at Tulo (MT) as indicated in Fig. 10.17. The levels observed at MD, MN, and MT were 298.7, 243.0, and 351.8 kg/s, respectively. All these sampling locations are downstream of cultivated lands. MD in Mgeta River is located downstream of Mgeta, Singisa, Bwakila Juu, Bwakila Chini, Kolero, and Kisaki wards. MN is located downstream of hillslope cultivated lands of Mtombozi, Kisemu, Kasnga, and Mvuha wards. The MT is the location further downstream of MN at the same Mvuha River. MM is the starting location of Mgeta River at the mountain, which is not much degraded. Water can be clearly seen with less sediment.

Fig. 10.17

Selected spatial sediment sampling locations at Upper Ruvu catchment

ML, MK, and RK are downstream of Kibungo, Tawa, Kibogwa, Mfizigo, and Ruvu Rivers. These rivers pass through the forest reserve area while the cultivated upland is managed by agricultural management practices. Management practices in Kibungo subwatershed include contour farming, strip cropping, and mixed farming. Although the sampling interval was weekly in 4 weeks, the trends show differences in results. The first week sampling shows a high level of sediments in rivers because of some of final rains of the rainy season in June and early July. The second through fourth sample was taken during July and early August which is within the dry season of the area. With no incoming runoff, streams have resulted in unchanged levels of sediment as shown in Fig. 10.18.

Fig. 10.18

Sediments load and flows in the Upper Ruvu catchment rivers

It was also found that there is a very good correlation between stream flow and sediment load as shown in Fig. 10.19.

Fig. 10.19

Relationship of average measured sediment load variation with flow in Ruvu catchment

7 Conclusions

This study should not be seen as validation of the WEPP model in the Upper Ruvu catchment environment. The main objective of this study was to test the prediction capability of WEPP on tropical watersheds located in hillslope areas. The overall results indicated that the WEPP model can assist watershed-related management institutions to quickly generate conservation zones by accepting predictions of sediment yield runoff outputs in spatially distributed format. Further study on testing the model with different assumptions is needed. The results showed that hydrological outputs were quite well predicted.Average annual runoff depths predicted in all four subwatersheds during the research period at different rainfall events varied from 460mm in lowlands within a high forest area (Kibungo chini) to 1,120mm in hillslope area, within open woodland and cultivated areas (Mfizigo Juu) subwatershed. Establishment of four scenarios at different conditions also predicted maximum average annual runoff depth to be 1,135mm at scenarios 1 and 2, while minimum average annual runoff depth was 51mm in areas with the same conditions of soil types.

The maps locate only potential hotspots erosion areas with high sediments delivery which can help watershed and basin managers to implement necessary precautionary measures to minimize or prevent soil erosion. High hazard soil erosion spots appeared in high elevation areas of Tegetero, Kibogwa, Kibungo Juu, Tawa, Kinole, and Mkuyuni wards in Mfizigo subwatershed. Similarly, Mtombozi, Singisa, Kasanga, and Kolero wards in Mvuha subwatershed showed high vulnerability to soil loss. The watershed soil loss and sediment yield were correlated with runoff. The highest and lowest total average annual soil loss rates were estimated to be 45.09 kg/m² at Mfizigo Juu watershed and at 0.45 kg/m² Kibungo chini watersheds, respectively. Although average total soil loss varied with runoff changes, the simulation showed some effect of soil type and slope in different land use/land covers within subwatersheds. Model results show cultivated land contributes 81% of soil loss and 86% sediment yield in all four scenarios.

A detailed accurate prediction of sediment yield and runoff inWami-Ruvu basin is crucial for planning and development of watershed-based projects. According to Walling (1994) , sediment load represents a small percentage of the total land area eroded and converted to sediments but still is a good source of information for management of soil erosion and sedimentation within the basin. More complete experimental data sets such as plot scale information than this used in the study will enhance the applicability of the WEPP model. The limited measured data used in simulation was certainly not ideal for the model validation. However, the study leads the way to better understanding of model assumptions and data input needs.