Full Length Research Paper
ABSTRACT
INTRODUCTION
The Sahel region is considered one of the most vulnerable ecosystems in the world and the majority of the population is highly dependent on soil resources to ensure adequate crop production or to maintain their crop production and livelihoods (Mohamed et al., 2001). Also, it is characterised by erratic precipitation and water shortage due to semi-arid climatic conditions. Speikerman et al. (2015) showed that the degradation of environmental condition was accelerated by droughts and an overall decrease in precipitation in the Sahel.
Soil erosion is one of the principal environmental problems that affect food production in the Sahel (Gandah et al., 2003); it is one of the manifestations of desertification processes. Cobo et al. (2010) explained that soil erosion is one of the most ordinary and extensive forms of soil degradation and is closely related to processes of nutrient reduction and desertification in semi-arid regions. It affects agricultural production, contributes to the contamination and quality of water resources. It further decreases the soil fertility content, the fine grained soil content and the water holding capacity and the depth of the top soils. Michels et al. (1995) reported that erosion processes have a significant impact on land productivity in the West African Sahel, a semi-arid region where both wind and water take action as erosive forces.
Soil texture is influenced by drainage condition, permeability and water holding capacity. It directly affects the porosity of the soil and its long-term soil fertility. According to Wischmeier and Smith (1978), soil texture determines the soil erodibility and affects the risk of soil erosion. Therefore, the knowledge of soil texture is essential for studying soil erosion. The TreeNet Model (Salford Systems Implementation, cf. Friedman, 1999) was used to estimate the spatial distribution of soil texture. Such techniques have been successfully used by Henderson et al. (2005), Ramadam et al. (2005) and Wilcke et al. (2008). However, the approach employed in this paper is based on a data mining technique developed by Friedman (1999).
To combat desertification, it is indispensable to understand soil erosion processes in the Sahel region. This paper assesses soil erosion research in the Sahel region and used the Tillabery landscape as a case study. In achieving this, it used the RUSLE and USPED models. The RUSLE model was originally applied in the USA, and has proven itself as a valuable model for estimation of soil erosion loss in other regions of the world (Wischmeier and Smith, 1978; Renard et al., 1997; Millward and Mersey, 1999; Angima et al., 2003) This model predict the average annual soil loss from rill and sheet erosion. The USPED is an empirical method which determines the spatial distribution of erosion and deposition (Mitasova et al., 1996) and it is one of the most commonly applied models to estimate soil erosion and deposition (Warren et al., 2005; L?pez-Vicente and Navas, 2010; Zhang et al., 2011) In this paper, RUSLE and USPED models have shown a realistic evolution of soil loss distribution during all the study period.
MATERIALS AND METHODS
Research area
The study area is located as shown in Figure 1 within the Department of Tillabéry and includes parts of the wide valley of the Niger River. The annual precipitation is between 250 and 400 mm. The soil in this area is very infertile and poses enormous challenges for agricultural production. This area was selected as investigation area due to the fact that desertification is the most serious environmental problem and it is located at the core of the Sahel. To compound it, the depth and width of the Niger River is continuously on a decrease, an issue associated with the moving of sand.
The traditional cereals, millet and sorghum, represent about 85% of the total food crop production requirements of the population in the study area. Livestock equally plays an important role in the economy. Therefore, cattle breeding and crop production are always competing for the fertile pieces of land available. This has spawned a series of conflicts between the various stakeholders – nomads and sedentary people.
Wind is another climatic factor influencing land degradation. Under the influence of high-speed winds, the process of wind erosion takes place. The combination of wind effect and high soil temperatures seriously affects the establishment of vegetation cover. Wind speeds exceeding 100 km/h have been observed (Sivakumar et al., 1993).
Soil sampling and laboratory methods
To investigate the amount and spatial distribution of the soil structure and soil erosion, a stratified random sample was established based on two-stage sampling plans with topography and land use as the stratifying variables. A topographic wetness index was derived from the Shuttle Radar Topography (SRTM) at a resolution of 90 m with two classes. This index shows areas of potential soil moisture (with value 1) and dry areas (with value 2). The wetness index was combined with land use derived from a Landsat image of 2001. The result displays 12 environmental units within the study area (Figure 2). Two environmental units within the water bodies were not considered because they do not show a classical category of soil attributes. In addition, 25 soil samples were randomly selected within each of the 10 environmental units (250 soil samples). For each point, geographical coordinates were taken using a GPS navigation unit.
The 250 soil samples (soil samples depth 0-30 cm) were collected and tested by feel soil method (Thein, 1979) during rainy season (September, 2012). Four soil samples within the 10 environmental units (40 soil samples) were submitted to laboratory analysis. The following elements were analysed: particle size percentage, USDA classes by pipette method (Gee and Bauder, 1986), organic carbon by combustion method (Read, 1921).
Land forms
A couple of satellite data were used to derive the land form parameters for soil erosion and soil texture regionalisation in the study area. The Shuttle Radar Topography Mission (SRTM) DEM at 90 m spatial resolution was obtained from the University of Maryland (Earth Science Data Interface at Global Land Cover Facility) and an Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) DEM at 30 m spatial resolution was obtained from the National Aeronautic and Space Administration (NASA).
Spatial prediction of soil texture using TreeNet Model
A statistical model was adapted to predict the spatial distribution of soil texture from terrain parameters. For this task, the field survey x- and y-coordinate data together with the z-field representing texture information from 250 observation points were used. SAGA GIS (System for Automated Geoscientific Analyses, Conrad, 2007) was used to overlay terrain attributes derived from digital elevation models. Spatial prediction of soil texture landscape was developed using the TreeNet model (Salford machine implementation, cf. Friedman, 1999).
The dependent variable is categorical (soil texture) and the independent are terrain attributes. A predictive map of soil texture was obtained under SAGA GIS through the application of a spatial interpolation method (inverse squared distance). Inverse squared distance (IDW) was selected based on the error distribution histograms showing that the IDW interpolation method has a superior performance than Kriging and Splines methods. Additionally, from the visualisation point of view, IDW presented the best result.
Soil erodibility (K) and cover management factor (C) calculation
In this study, K was expressed by two empirical equations. A calculating relation was proposed by Renard et al. (1997) in which they applied the percentage of the respective soil texture classes and geometric mean diameter to arrive at the K (without considering organic matter data). This is expressed in this equation.
Where: K: is the soil erodibility (t ha h / ha MJ mm); : is the geometric mean weight diameter of the primary soil particles (mm) expressed by:
Where: is the primary particle size fraction in percent, is the number of size classes in which the distribution curve has been divided and is the arithmetic mean of the particle size limits of that size. The results exposed that the erodibility ranges from 0.05 to 0.40 t ha h/ha MJ mm.
Torri et al. (2002) used a different approach by calculating the percentage of silt plus very fine sand, percentage of sand, percentage of organic matter and soil structure to arrive at the K (with considering organic matter data).
Where: K is the erodibility factor (Mg ha h/MJ ha mm), is the geometric mean weight diameter of the primary soil particles (mm), is organic matter content (%) and is soil clay content (fraction). After assigning the appropriate equation, the TreeNet was applied. The results show that the equation of the k with organic matter has lowest prediction. This implied the Receiver Operating Characteristic (ROC) integral was less than 0.70 for all the classes due to low amount of soil samples (40 samples). Ließ et al. (2011) explained that “poor model performance is most probably caused by the small size of the dataset”. In addition, model performance might be related to the size of the study area (Ziadat, 2005). Finally, both k factors were computed using Co-kriging model with wetness index as independent variable to produce a Co-kriged map of K factors results. Wetness index is selected as independent variable because it is the first important variable for both k factors.
Land use/land cover from Landsat images of 1973, 1989, 2001 and 2007 were converted to C. The agricultural land is assigned to a C of 0.4 as reported by Morgan (1995). The C varies between 0 (for example river), indicating that no erosion occurs, to 1, expressing the maximum of erosion (for example bare areas) and takes under consideration both cover and management variables. For shrub areas, a C of 0.01 was used as reported by Roose and De Noni (2004). For Plateau vegetation, a C of 0.1 was selected, which is also proposed by Roose and De Noni (2004).
Processing for soil erosion
The RUSLE model is supported by five components: rainfall erosivity factor (R), soil erodibility factor (K), topographic factor (LS), cover management factor (C), and support practice factor (P). The average annual soil loss (A, t/ha year) is computed by multiplying these factors.
The spatial patterns of soil erosion and deposition were simulated with the USPED model (Mitasova et al., 1996) to demonstrate the magnitude and spatial variability of soil loss and accumulation and to identify the most severely affected areas in the Tillabéry landscape using land use for the four years 1973, 1989, 2001 and 2007. To determine C factors, the precipitation data, from 1973 to 2007 were taken into consideration to calculate R factor values. K factor without considering organic matter is the same as used in the RUSLE model. However, the topographic component was computed by combining the Profile and the Tangential curvature, Upslope contributing area, Aspect and Slope. In the model output, erosion and deposition areas are shown by negative and positive signs, respectively. The maps obtained using USPED model were divided in the following classes as suggested by De Rosa (2005): -5 to -0.1 (t/ha year): erosion; -0.1 to 0.1 (t/ha year): stability; 0.1 to 5 (t/ha year): deposition or sedimentation.
RESULTS AND DISCUSSION
Soil texture
Influence of independent parameters on soil texture processes (rankings of important variables)
The quantification of the relationships between topographical elements and soil texture was done in this part. Using TreeNet Model technique, an estimate of important topographical parameters in relation to different soil structure elements is presented in Figure 3. The results show that there is a relatively strong relation between soil structure distribution and topographical parameters.
The global picture is the following: Altitude (100%) has the highest value for soil structure models. Watershed (88%), channel network (76%), profile curvature (75%), plan curvature (74 %), wetness index (72%), analytical
hillshading (72%), curvature (70%), aspect (66%), and LS-factor (6 %) are apparently important as well. Other variables achieved score values of 6 to 56% and their influence cannot be neglected. Sand content displayed the highest score with altitude (100%) and channel network (94%), whereas loam content had the highest score with altitude (100%) and plan curvature (83%). Wilcke et al. (2008) argued that there is a strong dependence of soil texture on altitude, articulated by a good positive correlation between altitude and sand content but a negative correlation concerning clay content. Altitude above channel network was the most important predictor for sandy loam, loam and clay loam in the study area due to erosion and deposition processes.
Model performance evaluation and cross validation
Receiver operating characteristic (ROC) curve was used
for characterizing the model performance. The value of ROC is between 0 and 1. In a worse scenario, the ROC value cannot be less than 0.5 and the best could be close to 1 (Lasko et al., 2005). For the ROC analysis, the true positive rate is plotted on the Y axis and the false positive rate is plotted on the X axis. Figure 4 illustrates the error tradeoffs available with a given model and describes the predictive behavior of the six classifiers, independent of class distributions. The most reliable soil texture predictor was loam and sand (0.981 and 0.983, respectively).
The prediction errors for the spatial texture of the topsoil in the study area were quantified by a ten-cross validation method (Table 1). The model had an accuracy classification rate of 72.80% by ten-fold cross validation. The result from cross validation is unsuitable because there were not enough data sets but with large data sets, the model validation becomes particularly impressive.
Soil texture map
The scoring process of the entire data set of Tillabéry landscape allowed constructing a map showing the spatially distributed soil texture potential for each class. Soil texture plays a very important role not only for soil fertility, but also for soil stability, water retention capacity and soil biodiversity. Figure 5 shows how the soil texture is spatially distributed in the study area. Soil texture estimation provides a better view of erosion and land degradation processes. The map reflects the high importance of topographic parameters for soil texture distribution. Also the model detected for the following terrain attributes a very high prediction potential for soil texture: Altitude above channel network, watershed, channel network, profile curvature, convergence index, plan curvature, wetness index, analytical hillshading, curvature, slope, aspect and LS-factor.
Therefore, the result suggests that topographic elements were important in determining the spatial distribution of soil texture. This can be explained by the fact that all terrain attributes considered in this investigation influenced soil texture due to erosion, transport and deposition processes. This study has demonstrated that loam (63.91%) was the dominant soil texture, suggesting that the study area tends to have high soil loss (Figure 6). The map shows that silt loam was found close to the Niger River and seasonal water bodies.
This result shows that the spatial distribution of soil texture in the study area depends on the wind regime. Sandy soil distribution is indicated by the activity of aeolian processes and act with dry wind direction (October - April) that move northeast and during the rainy season wind direction (May - July) that move westwards through the Sahel in the same time. In addition, the upper part of the study area is mostly dominated by loamy soils under the influence of aeolian processes. In general, the study showed that the spatial distributions of soil texture are influenced by the topographic elements and aeolian activities in the Sahel region.
Spatial distribution of soil loss using RUSLE
The spatial distribution maps of sheet and rill erosion in Tillabéry landscape from 1973 to 2007 are shown in Figures 7, 8, 9 and 10 based on K factor without organic matter and soil loss based on K factor with organic matter. The results are displayed in a 90 × 90 meter grid cells.
The intensity of erosion is classified in different classifications. Wischmeier and Smith (1978) reported that the maximum tolerable soil loss (12 t/ha year) is described as “the maximum level of soil erosion that will permit a level of crop productivity to be sustained economically and indefinitely”. In this study, the K factor, LS factor and P factor values were constant, and R and C factor values were changed according to scenario settings. The results of this analysis demonstrated that rill and sheet erosion during the study periods increased alongside the K factor models.
Mean annual soil loss values are presented in Figures 7, 8, 9 and 10 from 1973 to 2007 and they range from 0 to 140 t/ha year. The highest values (greater than 50 t/ha year) of soil loss were recorded in an infinitesimal amount in the centre of the study area in 1973 and increased with time in the study area for both models. It should be noted that the zone of intensive rill and sheet erosion is situated in the study area with loamy texture soils, bare soil areas and with a slope greater than 7% (Figure 11). Analysis of the RUSLE model series (1973, 1989, 2001 and 2007) show that rill and sheet erosion increased moderately (30 - 50 t/ha/year) and this increase was closely related to an increase in bare areas in the study area.
These results clearly show the existence of interactions between the slope, land cover and soil texture. Amounts of soil loss are higher with the first model because of high values in K factor. Both models indicate low erosion risk close to the river and southwestern part of the study area, but show high erosion risk in the uplands. By comparison of both models, it can be noted that the rill and sheet erosion areas reveal high values for both methods (50 - 140 t/ha/year), which implies that rill and sheet erosion play an important role in land degradation and desertification in the Sahel region (Figures 12 and 13).
On the one hand, sheet erosion is more concentrated in the north east of the study area, dominated by agricultural lands and grasslands. This can be explained by the fact that in the Sahel region, sheet erosion is associated with animal and agricultural production activities which accentuates compaction of the soil and the destruction of vegetation cover. On the another hand, the spatial distribution of rill erosion is more concentrated around the centre and close to major water bodies This is due to the fact that rill erosion in the Sahel region is connected with areas with high altitude (centre of the study area) and with silt loam soil type (close to major water bodies).
Comparison of soil erosion scenarios under rainfall options between two K factors
The rainfall erosivity for the year 2070 in Tillabéry area using future precipitation predicted from Had CM3 mode under A2 emission scenario (Alcamo et al., 2003; IPCC, 2000) was used. It gives useful information on the erosion risk in a changed climate condition. Figures 14 and 15 illustrate the expected future trend of soil erosion for the Tillabéry landscape up to 2070.
Figures 14 and 15 showed the spatial development of soil erosion up to the year 2070 under two scenarios (Tillabéry I and II). Tillabéry I is a simulated change in soil erosion risk for the scenarios based on rainfall erosivity till the year 2070 in the Tillabéry area using future precipitation predicted from Had CM3 model under A2 emission scenario based on K factor with organic matter. Tillabéry II depicts a simulated change in soil erosion risk for the same area based on rainfall erosivity until 2070 in the Tillabéry area using future precipitation predicted from Had CM3 model under A2 emission scenario (Alcamo et al., 2003; IPCC, 2000) based on K factor without organic matter.
In both scenarios (Tillabéry I and II), an increase in soil erosion is observed. In Tillabéry I, no risk of erosion is in the flat zones of the Niger River Basin. In Tillabéry II, the higher erosion risk (>50 t/ha/year) is visible. As such, these changes affect the rest of the classes. The spatial distributions of soil erosion scenarios between the two scenarios show a close relationship in general (Figures 14 and 15). The areas with a high erosion value with Tillabéry I may likely have a high erosion value with Tillabéry II. The spatial distribution of areas with high erosion values with Tillabéry II show a concentration in the central parts of the Tillabéry landscape, whereas the spatial distribution of areas prone to erosion with Tillabéry I are concentrated in the north-eastern part of the study area, precisely close to the Niger River.
Spatial patterns of soil erosion and soil deposition simulated by USPED Model
The results show that high soil erosion was observed and predicted on convex landscape positions and soil deposition was observed and predicted on concave landscape position and confirm the results obtained from the RUSLE analysis (Figure 16). During the four different years (1973, 1989, 2001 and 2007), simulated soil erosion/deposition maps show a low amplitude of simulated values. Indeed, the soil loss values range between 0.1 and 5 t/ha year and deposition values show similar amplitudes in the study area. These values can be elucidated by the specific terrain parameter of the Tillabéry landscape, where areas that have a low slope gradient, produce low soil loss rates and low accumulation.
The soil erosion/deposition maps for 1973 show that most of the study area was very stable. Soil loss was very minimal (0 - 0.1 t/ha/year). Deposition rate was also very low (0 – 0.1 t/ha/year) (Figure 18). Erosion had a significantly lower spatial coverage in 1973. The increase in the erosion endangered zone was caused by a reduction of shrub and grassland, which accentuated desertification and affected landscape stability. This means that rill and sheet erosion were amongst the major contributory factors to land degradation, decreasing soil productivity and an eventual deterioration in the soil and water quality in the Tillabéry landscape and calls for soil conservation and management methods to be applied. The USPED model highlights areas that were potentially affected by water erosion. So erosion areas could be well-known in a short period of time using this model.
The deposition zone is limited to the Niger River prone to accumulation and close to the runoff network as a result of the change in slope gradients. Figures 17 and 18 show a clear correlation between erosion affected zone and bare areas with slope gradients greater that 10%. There are areas in the center part of the study area in 1989, 2001 and 2007 which show high erosion rates, but in 1973 low erosion rates. This could be explained by the fact that during this period, these areas were generally more densely covered by shrubs. Compared with 1973, the stable areas decreased by 100,431 ha in 2007 (Figure 18). An increase in sediments was observed during all the study period due to climatic changes, development of agricultural lands and bare areas. These conversions have intensified the issue of desertification in the study area.
Decreasing soil erosion in the study area is a more cost efficient land management strategy than to deal with removing sediments from the Niger River, which is extremely costly and difficult. Investigating soil erosion/deposition in the Tillabéry landscape can actually be seen as a step to gain information about the spatial distribution of soil erosion and sediment deposition within the area. In addition, the figures could be applied to identify appropriate areas for implementing land conserva-tion assessment. Also, the results show that slope and land cover are the most important factors influencing soil erosion in the study area, which point out at the same time that sound land management practices have a controlling effect on soil loss.
CONCLUSION AND FUTURE RESEARCH NEEDS
The relationship between a set of topographic attributes, soil texture and different K factors for soil erosion assessment were investigated by using the TreeNet model in order to improve our understanding regarding soil degradation in the Sahel region and to build up a soil model for the study area. A set of topographic attributes (independent variables) were investigated, which allowed for a better understanding of the major driving factors that influence the spatial distribution of soil texture and erosion processes in the Sahel region. The results show that the altitude above the channel network and the wetness index are the most important factors that play a dominant role in the spatial distribution of soil texture because of their direct impact on runoff and transportation processes.
Soil texture output provided excellent predictive performance, confirming that the Tree Net model was an effective tool. But, the output of K factors provided a bad predictive performance (test data set) due to limited amount of soil sampling (low sample size). Therefore Co-Kriging was used in estimating the K factors in the case of limited available data.
RUSLE and USPED models have shown a realistic evolution of soil loss distribution during all the study period and confirm the need to continue the investigation of soil erosion processes by using data mining tools in the Sahel region. Puerta et al. (2008) reported that soil loss by soil erosion is often noted to play a major role in the desertification process. Both soil erosion scenarios (based on future precipitation predicted from Had CM3 mode under A2 emission scenario) output show greater soil erosion in 2070 in the study area. The scenario based on K factor without considering the organic matter proved to have the highest erodibility.
The present project confirms the importance of organic matter in influencing soil erosion prediction and also as a useful instrument in planning soil degradation control practices. Therefore, reducing soil erosion is more prudent than to deal with removing sediment and sandy soils from the Niger River, which is costly and difficult in the Sahel region.
Erosion models are prone to further research. Further studies must be focused on trying to improve the soil erosion processes by using data mining tools in the Sahel region. However, results indicate that the greater number of sampling sites and using higher quality DEMs derived from LIDAR/RADAR and TanDEM-X are needed in order to improve the result of the soil model.
CONFLICT OF INTERESTS
The author did not declare any conflict of interests.
REFERENCES
Alcamo J, Märker M, Flörke M, Vassolo S (2003). Water and climate: A global perspective. Kassel world water series 6, Center for environmental systems research, University of Kassel, Kurt-Wotters-Strasse 3, 34109 Kassel, Germany. | ||||
Angima SD, Stott DE, O`Neil MK, Ong CK, Weesies GA (2003). Soil erosion prediction using RUSLE for central Kenyan highland conditions, Agriculture, Ecosystems and Environment 97:295-308. Crossref |
||||
Ben Mohamed A, Van Duivenbooden N, Abdoussallam S, (2001). Impact of climate change on agricultural production in the Sahel- Part 1. Methodological approach and case study for millet in Niger. | ||||
Cobo JD, Dercon G, Cadisch G (2010). Nutrient balances in African land use systems across different spatial scales: A review of approaches, challenges and progress. Agr. Ecosyst. Environ. 136:1-15. Crossref |
||||
Conrad O (2007). SAGA – Entwurf, Funktionsumfang und Anwendung eines Systems für automatisierte geowissenschaftliche Analysen. Electronic Doktorat Dissertation, University of Göttingen. | ||||
De Rosa P (2005). Analisi e Confronti di modelli di erosion del suolo e trasporto di sediment tramite l`uso di sistemi G.I.S. Degree Thesis, Universitá degli studi di Perugia, Italia. | ||||
Friedman JH (1999). Greedy function approximation: a gradient boosting machine. Stanford: Statistics Dept., Stanford University, USA. | ||||
Friedman JH (1999). Stochastic gradient boosting. Stanford: statistics Dept., Stanford University, USA. | ||||
Gee GW, Bauder JW (1986). Particle-size analyisis. In R. Burt(ed), Soil survey field and laboratory methods manual, U.S Depart. of agriculture, Lincoln, Nebraska. Report no 51 pp. 53-54. | ||||
Henderson BL, Bui EN, Moran CJ, Simon DAP (2005). Australia-wide predictions of soil properties using decision trees.124, Geoderma pp. 383-398. Crossref |
||||
Lasko TA, Bhagwat JG, Zou KH, Ohno-Machado L (2005). The use of receiver operating characteristic curves in biomedical informatics. J. Biomed. Inf. 38:404-415. Crossref |
||||
IPCC (Intergovernmental Panel on Climate Change) (2000). Special report on emission scenarios. Intergovernmental Panel on Climate Change. Cambridge University Press. Cambridge. p.599. | ||||
Ließ M, Glaser B, Huwe B (2011). Uncertainty in the spatial prediction of soil texture comparison of regression tree and random forest models. Geodenna 170:70-79. | ||||
Michels K, Sivakumar MVK, Allison BE (1995). Wind erosion control using crop residue. I. Effects on soil flux and soil properties. Field Crops Res. 40:101-110. Crossref |
||||
Millward AA, Mersey JE (1999). Adapting the RUSLE to model soil erosion potential in a mountainous tropical watershed Catena. 32(2):109-129. Crossref |
||||
Mitasova H, Hofierka J, Zlocha M, Iverson LR (1996). Modeling topographic potential for erosion and deposition using GIS. Int. J. GIS. 10:629-642. | ||||
Morgan RPC (1995): Soil Erosion and Conservation. Addision-Westey. Edinburgh, UK. p.198. | ||||
Gandah M, Brouwer J, Hiernaux P, Van Duivenbooden N (2003). Fertility management and landscape position: farmers use of nutrient source in western Niger and possible improvements Nutrient Cycling in Agroecosystems 67:55-66. Crossref |
||||
Ramadam Z, Hopke PK, Johnson MJ, Scow KM (2005). Application of PLS and back-propagation neural networks for the estimation of soil properties. Chemon. Intell. Lab. Syst.75:23-30. Crossref |
||||
Read JW (1921). Rapid dry combustion method for the simultaneous determination of soil organic matter and organic carbon, J. Ind. Eng. Chem. pp. 304-308. | ||||
Renard KG, Foster GR, Weesies GA, McCool DK, Yoder DC (1997). Prediction soil erosion by water: a guide to conservation planning with the Revised Universal Soil Loss Equation. Agricultural Handbook, vol. 703. U.S. Department of Agriculture. p. 404. | ||||
Roose E, De Noni V (2004). Recherché sur l érosion hydrique en Afrique-revue et perspectives. Sécheresse 15:121-129. | ||||
Sivakumar MV, Maidoukia A, Stern RD (1993). Agroclimatologie de l Afrique de l Ouest Niger, Bulletin d'informations No 5, ICRISAT, Patancheru. India. pp.108. | ||||
Thein SJ (1979). A flow diagram for teaching texture by feel analysis. J. Agron. Edu. 8:53-55. | ||||
Torri D, Poesen J, Borselli L (2002). Corrigendum to predictability and uncertainty of the soil erodibility factor using global dataset, Catena 46:309-310. Crossref |
||||
Warren SD, Mitasova H, Hohmann MG, Landsberger S, Iskander FY, Ruzycki TS, Senseman GM (2005). Validation of a 3-D enhancement of the Universal Soil Loss Equation for prediction of soil erosion and sediment deposition, Catena 64:281-296. Crossref |
||||
Wilcke W, Yasin S, Schmitt A, Valarezo C, Zech W (2008). Soils along the altitudinal transect and in catchments. In: Beck, E., J. Bendix, I. Kottke, F. Makeschin, and R. Mosandl (Eds). Gradients in a Tropical Mountain Ecosystem of Ecuador-Ecological Studies, vol.198. Springer, Berlin, pp.75-87. Crossref |
||||
Wischmeier WH, Smith DD (1978). Predicting rainfall erosion loss. USDA. Agricultural research Service Handbook 537. | ||||
Zhang, SL, Zhang XY, Huffman T, Liu XB, Yang JY (2011). Influence of topography and land management on soil nutrients variability in Northeast China. Nutr. Cycling Agroecosyst. 89:427-438. Crossref |
||||
Ziadat FM (2005). Analyzing digital terrain attributes to predict soil attributes for a relatively large area. Soil Sci. Soc. Am. J. 69:1590-1599. Crossref |
Copyright © 2024 Author(s) retain the copyright of this article.
This article is published under the terms of the Creative Commons Attribution License 4.0