3. Results and discussions
3.1 Model calibration and validation results
In Rim sub-watershed, KTC(h) values of 350 and KTC(l) 75 delivered optimal model performance, whereas a KTC (h) value of 250 and KTC(l) value 25 were optimal in Asanat and Debre Yakob sub-watersheds. The observed and predicted sediment yields were 35.6 and 36.1 t ha-1 y-1 for Asanat, 24.4 and 25.1 t ha-1 y-1 for Debre Yakob and 31.7 and 32.5 t ha-1 y-1 for Rim respectively, with corresponding NSE values of 0.81 in Asanat, 0.56 in Debre Yakob and 0.72 in Rim at optimal values of KTC(h) (Figure 2). Generally the model over-predicted sediment yield in all sub-watersheds (Table 3). This may be partially attributed to considerable effects of SWC measures on soil erosion and transport capacity (Haregeweyn et al., 2013; A. Van Rompaey et al., 2005). The majority of cultivated lands of the study sub-watersheds are treated with bund structures. Although bunds had variable spacing, a routing algorithm was used for all bunds by creating a specific parcel map layer for all bunds. The model showed relatively higher performance in Asanat sub-watershed. This is likely due to a well-defined parcel map layer, and LS (i.j) and flow routing algorithm as a result of narrow and uniform spacing of bunds compared to Rim and Debreyakob. The parcel connectivity in each pixel, Ptef and P-factor, which together adjust the effect of bunds on erosion, accurately represent the situation on the ground in Asanat.
Fig. 2 approximately here
Table 3 approximately here
3.2 The effect of SWC measures on rate and patterns of soil erosion
The net soil erosion maps as calculated by WATEM/SEDEM for the three study sub-watersheds are given in Figures 5.3-5.5. The mean annual soil erosion rate calculated from the sum of mean annual sediment production and sediment deposition simulated by WATEM/SEDEM for the study watersheds are given in Table 4. The simulated soil erosion indicates high spatial variation in all study sub-watersheds.
Table 4 approximately here
The results of the model with SWC scenarios show a progressive decrease in soil erosion, indicating a considerable effect of SWC measures. When comparing the present-day situation (scenario I) with a situation without SWC measures (scenario V), simulated soil erosion is more than 57% lower in Asanat, 65% in Debre Yakob and 53% in Rim sub-watersheds. In scenario II (bunds and contour cultivation), erosion rates were further decreased by 8% in Asanat, 10% in Debre Yakob and 15% in Rim with respect to current conditions. In scenario III, 5% less erosion was observed in Asanat, 12% less in Debre Yakob and 14% less in Rim compared to scenario II. The largest reduction of soil erosion was simulated for scenario IV in all study sub-watersheds (Table 4). In scenario IV (combination of bunds, contour cultivation, strip cropping and grass strips), soil erosion was decreased by 128, 164 and 180% in Asanat, Debre Yakob and Rim, respectively, when compared to scenario V (no SWC measures) and by 45, 61 and 83 % when compared to scenario I (present situation).
Fig. 3 approximately here
Soil erosion was reduced by a higher percentage in Debre Yakob than for Asanat and Rim when comparing the present day situation to a scenario where sub-catchments would be untreated (scenario V). This may be due to a larger coverage of bund structures and other conservation measures such as traditional ditches, diversions and check dams in Debreyakob. Relatively, contour cultivation, strip cropping and grass strips were more effective in reducing soil erosion in Rim compared to Debre Yakob and Asanat (Table 4). This is most likely due to the topographic characteristics in Rim. On steeper slopes, agronomic measures such as contour cultivation and strip cultivation are less effective in reducing soil erosion. Although soil erosion and/or sediment yield is reduced in the present-day situation (scenario I) as compared with a ‘no SWC measures’ situation (scenario V), in most parts of the cultivated lands of the study sub-watersheds still high rates of erosion were simulated. This emphasises the requirement to combine various conservation strategies to reduce soil erosion and sediment delivery in the study sub-watersheds. From the analysis of the present-day situation, the erosion map suggests that areas with greatest soil erosion are concentrated on locations with steep slopes and/or areas with poor bund structures (smaller dimensions and wider spacing). Lower erosion rates in intervention scenarios II-IV correspond to areas treated with effective bund structures, including upgrading of the stability of bund structures. Extremely high erosion rates (>66 t ha-1) were observed over large parts of cultivated lands in Asanat whereas higher deposition areas were concentrated in Debre Yakob and in the downstream part of Rim (Figure 3-5).
Fig. 4 approximately here
Comparison of these model results with the range of soil erosion rates reported for cultivated lands treated with SWC measures shows generally good agreement. The model predicts reasonably acceptable ranges of soil erosion as compared to annual soil loss observed in treated farm plots in previous studies in similar areas in the Ethiopian highlands (Bewket & Sterk, 2003; Mitiku et al., 2006). Herweg and Ludi (1999) estimated an average soil loss reduction of 40% by graded soil bunds and 50% reduction with fanyajuu bunds in Anjeni, Ethiopian highlands. Another study in Tigray, Northern Ethiopia by Vancampenhout et al. (2006) found that stone bunds trapped 64% of soil otherwise lossed by soil erosion.
Fig. 5 approximately here
Simulation of the effect of physical SWC measures (bunds and diversion channels) on sediment yield with the Soil and Water Assessment Tool (SWAT) in the upper blue Nile basin by Lemann et al. (2016a) estimated an average sediment yield reduction of 54% while Dagnew et al. (2015) found a 57% decrease in suspended sediment concentration (SSC) at Debremewi sub-watershed in NW Ethiopian highlands. A similar study in Kenya reported by Hessel and Tenge (2008) show that LISEM-simulated physical SWC scenarios decreased erosion by 60% in an agricultural catchment. Subhatu et al. (2017) estimated 32-37 t ha-1 y-1 soil loss using the USLE in treated catchment of Minichet, North Ethiopian highlands.
The estimate made on impacts of SWC measures in this study also agrees well with another model-based soil erosion estimation for treated catchments by Hessel, Messing, Liding, Ritsema, and Stolte (2003) where soil loss was decreased by 60% by simulating the impacts of SWC measures using LISEM (Limburg Soil Erosion Model). Nyssen et al. (2007) estimated a 0.32 conservation practice P-factor for bund structures in USLE and estimated soil loss rates of 58 t ha-1y-1 in the Tigray region of Northern Ethiopia. In a related study, Nyssen et al. (2006) investigated the effects of SWC measures using the WOFOST and LISEM models for Tigray and found a 68% reduction of soil erosion due to bund structures. In their assessment of landscape susceptibility to soil erosion using a GIS-based approach in North Ethiopia, Tamene, Adimassu, Aynekulu, and Yaekob (2017) predicted a mean annual soil loss of 45 t ha-1y-1 using RUSLE for treated cultivated lands.
The large variation in predicted erosion rates across the study sub-watersheds reflects the high spatial variation of factors potentially influencing soil erosion. The effect of SWC measures on soil erosion was not uniform for the same land use types and slope classes. This emphasises that the effectiveness of SWC measures on controlling erosion depends on biophysical factors such as topography, land use and geology, etc. Furthermore, the spatial pattern and type of land use are relevant to erosion because changes in land use can alter the efficiency of SWC measures to control soil erosion within sub-watersheds (Desmet & Govers, 1995). Roads, field boundaries and other landscape structures also affect the efficiency of SWC measures to prevent soil erosion and sedimentation between various land units (A. J. Van Rompaey et al., 2001). This effect is well accounted in WATEM/SEDEM by incorporating a parcel map (A. J. Van Rompaey et al., 2001). The spatial variation of various conservation scenarios clearly indicates the importance of landscape modification by the use of physical SWC measures on soil erosion. The primary purpose of SWC measures is to divide the natural length of the hill slope into smaller sections so that runoff and soil erosion are reduced (Troeh, Hobbs, & Donahue, 1980) and this process is determined mainly by topographic characteristics and land use (Meshesha, Tsunekawa, Tsubo, & Haregeweyn, 2012). Meshesha et al. (2012); Nyssen et al. (2007) reported high variation in soil loss rates in plot experiments and catchment scale modelling, confirming the strong spatial variability and scale dependency of soil erosion processes due to various attributes of catchment areas.
Previous studies in the Ethiopian highlands suggest that bund structures reduce soil erosion; however, the effectiveness of these measures can be improved by integrated use of physical, agronomic and vegetative conservation strategies at sub-catchment level (Betrie et al., 2011; Dubale, 2001). Nyssen et al. (2007) indicated that the use of one or a combination of SWC measures depends on the objective and economic viability of conservation strategies. According to the plot experiments of Amare et al. (2014), the combined use of soil bund structures with Tephrosia plantation (a biological SWC measures) in the North-western Ethiopian highlands on average decreased soil loss by 71 to 26 t ha-1 y-1. An additional benefit of the biological SWC measures reported by Amare et al. (2014) was that 2.8 t ha-1 y-1 of dried forage was obtained from elephant grass grown on bund structures. Thus, the biomass obtained could compensate the land taken out of production by physical structures (8-10%) and could alleviate the shortage of animal feed (Adimassu et al., 2012). In addition, the soil organic matter content is enhanced by integrated use of biological conservation strategies and bunds (Amare et al., 2014).
3.3 Impacts of SWC measures on sediment connectivity and yield
This study illustrates that sediment deposition and yield were highly variable within study sub-watersheds. This is most likely due to the effect of SWC measures as well as biophysical characteristics such as topography, land use and soil types (Grum et al., 2017; Mekonnen et al., 2015). However, previous studies indicated that specific sediment yield decreases with increase in catchment area (de Vente, Poesen, Arabkhedri, & Verstraeten, 2007; Descheemaeker et al., 2006). According to Grum et al. (2017); Haregeweyn et al. (2008), the lower rate of specific sediment yield in larger watersheds is due to increased sedimentation processes and sinks at obstructions in the lower reaches of larger watersheds. Even though the topography is less steep, the SDR was very high in Rim sub-watershed reaching up to 56% (Table 5). This might be due to natural ditches, stream bank erosion and gullies which increase sediment connectivity. Rim sub-watershed is severely affected by landslides and gully erosion (Jemberu et al., 2018). In line with this, Verstraeten et al. (2002) reported that areas where surface runoff is concentrated in ditches and gullies facilitate sediment delivery and high TC and SDR. A large part of landscape becomes connected to the stream by continuous paths of concentrated overland flow and channel flow from gullies (Gallart, Llorens, & Latron, 1994).
This study demonstrates that the measurement of sediment yield at the outlet of catchment areas, taking into account the spatial distribution of total erosion and deposition processes, can be a good indicator of erosion taking place within the upland areas. However, in catchment areas with low TC and SDR, the measurement of sediment yield at the outlets of the catchment areas can be a poor indicator for erosion processes in the upland catchment areas. Overall connectivity within catchment areas varies with all sediment production, transfer and delivery processes that occur within it (Borselli et al., 2008). Therefore, catchment areas with high soil erosion are not necessary areas contributing most sediment to rivers (Stall, 1985; Steegen et al., 2000; Verstraeten et al., 2002). According to Cammeraat (2002), surface roughness, vegetation cover and rain intensity influence sediment production, transport from the upper part of the catchments and delivery to the river channels. The differences in sediment connectivity and yield between study sub-watersheds were apparently due to the effects of conservation measures and topographic characteristics. In addition the morphology and channel incision also controlled sediment connectivity and yield. The impact of a given type of impediment of sediment flow depends upon its size and position in the catchment area (Fryirs, Brierley, Preston, & Kasai, 2007). Moreover, the sediment detachment, transfer and delivery to the river channels not only dependent on overall biophysical characteristics of catchment areas such as land cover, topography and geology, but also on the effects of watershed development activities (Grauso, Pasanisi, & Tebano, 2018).
Table 5 approximately here
The large heterogeneities in TC and SDR in the study sub-watersheds may reflect topographic characteristics and the spatial pattern and effect of conservation measures implemented in the study sub-watersheds. In line with this, Einstein (1950); Verstraeten et al. (2002) reported that sediment flows are highly variable with the topographic characteristics and land use types. Earlier work by Einstein (1950); Stall (1985) showed that factors influencing soil loss and TC across catchment areas have implications for sediment connectivity and sediment yield. Marchamalo et al. (2016), in their study of sediment connectivity as a framework for understanding sediment transfer at multiple scales in SE Spain, emphasises that land use and SWC measures have a clear impact on sediment connectivity, by affecting the link between the sediment produced on the upper side of the catchment and transporting to streams. The Water Availability in Semi-Arid Environments with Sediment Dynamics Component (WASA-SED) model simulation by Medeiros, Güntner, Francke, Mamede, and Carlos de Araújo (2010) in Brazil showed that the spatial pattern of sediment connectivity within catchment changes as a function of landscape and land use. The high variation in sediment deposit and sediment yield could also be attributed to the effects of SWC measures, topography, land use and geology on transport in a variety of other studies (Baartman, Masselink, Keesstra, & Temme, 2013; de Vente et al., 2008; Marchamalo et al., 2016; Stall, 1985; Verstraeten et al., 2002).