The status of soil erosion in the Upper Blue Nile Basin: Identification of hot spot Areas and Evaluation of Best Management Practices in the Toba Watershed


 Land degradation caused by soil erosion has become the most serious problem in the Ethiopian highlands. Quantifying the spatial variations of soil loss with a strong evidence helps to prioritize the watersheds for the implementation of different management practices. The study was carried out in the Toba Watershed of the Upper Blue Nile Basin in Ethiopia. Its objective was to evaluate the rate of soil erosion and identify the hotspots with high risk of soil erosion for watershed management planning. Then, Soil and Water Assessment Tool (SWAT) was used to evaluate the effectiveness of best management practices (BMP) in reducing soil loss. The performance of SWAT in simulating streamflow and sediment yield was evaluated through sensitivity analysis, uncertainty, calibration and validation process. Statistically, the calibrated and validated sediment yields (SY) against the observed sediment data were reasonably accurate (R2 = 0.67, 0.65, NSE = 0.66, 0.64, PBIAS=-8.4%, 9.8% respectively). The annual SY in Toba watershed varies from 0.09 t ha− 1 yr− 1 to 44.8 t ha− 1 yr− 1 with an average SY of 22.7 t ha− 1 yr− 1. To prioritize the SY of the watershed, the annual severity of SY was divided into six classes: very low, low, moderate, high, very high and severe. The study also showed that SY in most watersheds (about 53.8%) were higher than the average. Cultivation on steep slopes leads to the highest SY, while forested areas have lower SY contribution. five management scenarios were evaluated using the Calibrated model. Seventeen sub-basins with SY exceeding the tolerable erosion of Ethiopia (t ha− 1 yr− 1) were considered for the analysis of the BMP scenario. The results show that reforestation combined with vegetative strips was the most effective for soil erosion control (87.8% reduction) followed by the combination of soil/stone bund and vegetative strips (83.7% reduction). Overall, the results of this study provided important data for watershed management and are very useful to ensure the sustainable management of land and natural resources at watershed level.

sediment yields (SY) against the observed sediment data were reasonably accurate 23 (R 2 =0.67, 0.65, NSE=0.66, 0.64, PBIAS=-8.4%, 9.8% respectively). The annual SY in 24 Toba watershed varies from 0.09 t ha -1 yr -1 to 44.8 t ha -1 yr -1 with an average SY of 22.7 25 t ha -1 yr -1 . To prioritize the SY of the watershed, the annual severity of SY was divided 26 into six classes: very low, low, moderate, high, very high and severe. The study also 27 showed that SY in most watersheds (about 53.8%) were higher than the average. 28 Cultivation on steep slopes leads to the highest SY, while forested areas have lower SY 29 contribution. five management scenarios were evaluated using the Calibrated model. 30 Seventeen sub-basins with SY exceeding the tolerable erosion of Ethiopia (t ha -1 yr -1 ) 31 were considered for the analysis of the BMP scenario. The results show that 32 reforestation combined with vegetative strips was the most effective for soil erosion 33 1. Introduction 39

General Background 40
Regardless of the endowed diverse natural resources, Ethiopia is experiencing severe 41 land and environmental degradation that has been a serious causes of low productivity 42 resulting in a widespread poverty and food insecurity. Agricultural productivity in 43 Ethiopian highlands are highly affected by pervasive land degradations [1-3]. Land 44 degradation due to soil erosion in the highlands are due to the intermingling factors like 45 lack of effective watershed management practices, increased agricultural activities on 46 steep slopes, land use/land cover change, heavy rainfall, climate variability and mixed 47 crop-livestock farming systems [4,5]. 48 In Ethiopia, severe soil erosion risks are strongly linked with population density [3]. The mean annual rainfall in the catchment varies from 1497 mm in the south western 130 and 2500 mm in the northeastern part of the watershed. The watershed is characterized 131 by humid tropical climate with heavy rainfall. The maximum and minimum 132 temperature in Toba ranges from 18 to 36 °C and 6.5 to 17 °C [20]. 133

Input Data 134
The application of SWAT model to evaluate the spatial distribution of soil loss and 135 quantify effectiveness of the BMPs requires the integration of spatial and temporal data 136 with the application of different management practices. The spatial datasets used 137 include: Digital Elevation Model (DEM), land use/land cover and soil data (Table 1). 138 Whereas the temporal data includes weather data, streamflow and sediment data. The spatial maps of the Toba watershed landscape attributes are presented in Figure 2.  is given in Eq (1) below: 164 Where: SWt is the final soil water content (mm), SWo is the initial water content (mm), Where: Sed is the sediment yield from a given HRU on storm basis (ton/day), Qsurf is 174 surface runoff volume (mm/ha), qpeak is peak surface runoff (m 3 /s), areahru is the area 175 of hydrologic response unit (ha), KUSLE is the soil erodibility factor (MgMJ -1 mm -1 ), 176 PUSLE is soil erosion control protection factors, LSUSLE is topography factor, CUSLE is 177 crop management factor, CFRGUSLE is coarse fragment factor.   Table 2. increase soil cover, helping to ensure the soil and water conservation [12]. In this 233 study, the reforestation of grasslands, shrublands and cropland that are on slopes 234 greater than 16% was applied by introducing land use/land cover in the land use 235 update of the watershed data. We considered this scenario to restore forests that 236 have been destroyed. Converting all crop land to forest land is not feasible. In this 237 regard, only 5% of the crop land was considered for reforestation.
Where Qobs is the observed variable, Qsim is the model simulated output, Q obs is the 261 mean of the observation and Q sim is the mean of the simulated output and n is the total 262 number of observations. 263 B. Nash Sutcliff efficiency, NSE 264 The use of deterministic approach that results in a single set of parameters as best 268 simulation is an outdated approach in calibration as it doesn't recognize the errors and 269 uncertainties in the modelling works. Consequently, any model calibration must include 270 the analysis of the uncertainty with propagations of parameter uncertainties [29] in 271 addition to the statistics R 2 , NSE and PBIAS. 272 Parameter uncertainty in SUFI-2 expressed as ranges accounts for all sources of 273 uncertainty from conceptual model, parameters, measured data and uncertainty in 274 driving variables [29]. Two statistics, P-factor and R-factor were used to quantify the 275 fit between the simulation result expressed as 95% prediction uncertainty (95PPU) and 276 the observation. the degree to which all uncertainties are accounted for is designated by 277 P-factor whereas, R-factor is the average thickness of the 95PPU envelop (30). For P-278 factor, the value of greater than 70% and R-factor of around 1 could be acceptable for 279 stream flow whereas, smaller value of P-factor and a larger value of R-factor could be 280 acceptable for sediment. 281

Sensitivity Analysis, Calibration and Validation 283
The relative sensitivity analysis for streamflow and sediment were carried out on the 284 monthly time-scale at subbasin 11 where the gauging station is located. The parameter 285 sensitivity and rankings with the significance of the relative sensitivity are determined 286 using t-stat and p-value. The lower p-stat and larger absolute t-stat value indicate the 287 most significant parameter. Using the p-value and t-stat, Global sensitivity using Latin 288 hypercube 'one-at-a-time' regression system was used to evaluate the relative 289 sensitivity. The sensitive streamflow and sediment in Toba watershed are described in 290 Table 3. From Table 3 Figure 4). However, the general prediction of the model is good enough to simulate 332 the streamflow except the peak flow in most of the calibration and validation years.

Prioritizations of Toba watershed to sediment yields 344
Soil erosion by water has become the responsible factor for the degradation of the fertile 345 top soil from agricultural lands. This is a great challenge for agricultural productivity 346 in highland parts of Ethiopia where agriculture is the dominant activity of the 347 community. Toba watershed is one of the highland watersheds where soil erosion has 348 become the challenging problem for agricultural activity. The annual sediment yield in 349 the watershed ranges from 0.09 t ha -1 yr -1 to 44.8 t ha -1 yr -1 with an average sediment 350 yield of 22.7 t ha -1 yr -1 . The annual SY of the watershed was classified into six severity 351 classes: very low (0-5 t ha -1 yr -1 ), low (5-10 t ha -1 yr -1 ), moderate (10-18 t ha -1 yr -1 ), 352 high (18-30 t ha -1 yr -1 ), very high (30-40 t ha -1 yr -1 ) and severe (>40 t ha -1 yr -1 ) ( Table  353 5). The very low and low class represents the level of erosion was less than the rate of 354 soil formation very high and severe classes showed that the SY is higher than the 355 average SY. 356 The spatial distribution of the sediment sources shows that, very low and low SY (<11 358 t ha -1 yr -1 ) in the watershed was generated from sub-basin 18, 19 and 14 ( Figure 6). The estimated annual average rate of SY in Toba watershed was 22.7 t ha -1 yr -1 . This 372 was higher than the tolerable soil loss (2-18 t ha -1 yr -1 ) from agricultural lands in 373

Evaluation of Best management Practices 389
Usually, it is important to establish threshold value between tolerable and intolerable 390 level of soil erosion to minimize the risks of soil erosion. The rate of soil loss considered 391 as tolerable based on maintenance of crop production was reported from 1 to 11 t ha -1 392 yr -1 [36]. According to FAO [36], SY from 8.2% of the watershed is considered as a 393 tolerable rate of erosion. In Ethiopia, the tolerable rate of soil loss in different agro-394 ecological conditions were reported from 2 to 18 t ha -1 yr -1 [33]. In this study, sub-395 basins that generates SY more than 18 t ha -1 yr -1 which accounts for 72.9% were 396 considered for the BMP scenario analysis. From the total 25 sub basins, only 8 sub-397 basins generate the tolerable soil loss and the remining 17 sub-basins require urgent 398 actions for management. 399 The summary of implementing the individual BMPs and their combination in Toba 400 watershed was summarized in Table 6. The lowest SY reduction was reported as 36.1 % 401 during the implementations of filter strip (FS) whereas the highest reduction was 402 reported as 80.5% by the simulation of vegetative strip (VS) followed by soil/stone 403 bund (SB). Application of SB on steep slopes and reforestation of the hilly areas 404 reported SY reduction by 69.3% and 47.5% respectively. However, implementing the 405 combinations of the BMP scenarios improved SY reduction better. The highest 406 reduction in SY was attained by the combination of R and VS followed by SB and VS. 407 This finding suggests a reasonable reduction of SY requires implementation of 408 appropriate combinations of BMPs. Improved reduction of SY by combining the BMPs 409 is also reported by similar studies [13]. 410 Although the application of all BMPs have shown reasonable reduction of SY, the 412 simulation of all BMPs revealed considerable spatial variability (Figure 7). The intervention packages that can be linked to this strategic goal of enhancing socio-491 economic resilience should logically be targeting on improving and/or modernizing the 492 agricultural production system through intensification, among others. An important 493 consideration of these interventions is that they have to contribute to the realization of 494 eco-friendly or climate-smart agricultural production systems. 495

Conclusion 496
Soil erosion by water has become a challenge facing agricultural production in 497 agricultural watersheds. The increasing risks of soil erosion and related environmental 498 problems have driven the need for research to address sustainable land and water 499 resources management. This study attempted to examine the soil erosion status of Toba 500 watershed and identify hotspot areas for effective watershed management interventions 501 to reduce the risk of sediment generation. Considering different alternatives to 502 investigate the possible soil and water management practices is one of the conceivable 503 outcomes of the study for decision policy. 504 The estimated annual sediment yield varies from 0.09 t ha -1 yr -1 to 44.8 t ha -1 yr -1 with 505 an average sediment yield of 22.7 t ha -1 yr -1 . The highest SY was contributed by the 506 steep farmland. The severity of erosion at the very low, low and moderate severity 507 levels covering 27.1% of the watershed area was within the tolerable ranges of soil 508 erosion in Ethiopia (2 to 18 t ha -1 yr -1 ). Seventeen sub-basins, which represent about 509 72.9% of the watershed area, have been identified as critical areas that require 510 implementation of proper measures. 511 Regardless of the considerable SY by all scenarios, the simulation of the individual 512 BMPs in reducing SY over Toba watershed has varied appreciably. The application of 513 certain scenarios (FS and R) cannot reduce the risk of soil erosion below the tolerable 514 limit of the soil loss. However, the combination of the scenario is more pronounced and 515 desirable in SY reduction. Therefore, this finding suggests that a reasonable reduction 516 in SY requires the implementation of an appropriate combinations of BMPs. 517 Overall, the study demonstrated how prioritization of erosion hotspot areas can be used 518 to aid systematic watershed planning through the use of modelling, SWAT. Coordinated 519 development and management of land and water with the broader upstream and 520 downstream interests could help to achieve better implementation of best management 521 practice. Therefore, this study recommends, creating awareness of the risk of soil 522 degradation in order to persuade and ensure the long-term engagement of the 523 community and stake holders in management activities. 524 Declarations 525

Competing Interest 526
The authors declare that they have no competing interests. 527

Funding 528
This research received no external funding. 529

Author Contributions 530
Wakjira T Dibaba have developed the concept of the study, methodology, and formal 531 analysis. Dessalegn G Ebsa conducted field works, investigation. Both authors involved 532 in writing review and editing the manuscript. All authors have read and agreed to the 533 published version of the manuscript. 534