The dongas studied in the subhumid zone precisely in the North of Benin in the district of Karimama, are distributed in the W National Park and on the land use areas. The district of Karimama (11˚26 - 12˚26 latitude N; 2˚17 - 3˚05 longitude E) is bounded by the Niger River at north, by the District of Banikoara the south-west, by the District of Kandi at south-east, by the District of Malanville at east and by the Republic of Burkina-Faso at west (Figure 1). The climate is Sudano-Sahelian with a single rainy season lasting 4 to 6 months and a dry season lasting 6 to 8 months. The average rainfall is 750 to 900 mm. The soils are of tropical ferruginous types; hydromorphic soils are found along the Niger River. Population density (49 inhabitants per km²) is higher than in Borgou (20 inhabitants per km²) and Alibori (28 inhabitants per km²) [11]. This high density partly explains the strong anthropogenic pressures on natural resources and particularly on land.

The experimental design used was a split plot (RCB) with two factors (topography and degree of donga degradation). The blocks consist of three plots installed from the Upstream to the Downstream of the donga. The experiment was carried out in two strata or sites (whole plot): the W National Park and the land use area. The three levels of slope (Upstream, Center and Downstream of the donga) are the modalities of the topography (main plot factor). The subplot factor, which is the degree of degradation of the dongas, includes two types of dongas that have been chosen according to the degree of erosion: shallow dongas (diameter between 10 and 50 m and depth less than 1 m) and advanced dongas

(diameter > 50 m and depth > 1 m). Two dongas were chosen per site: the beginner donga from Baru and the advanced donga from Kargui in the village area and the beginner donga DP₁ and the advanced donga DP₂ in the W Park.

In situ data (mean slope and permeability were collected in experimental plots within each donga [12] [13]. Erosion plots are repeated three times in each donga. A total of 36 deviation plots (9 plots/dongas) were considered.

The experimental plots were delimited by sheet metal (Figure 2). They consist of a rectangular plot (1 m × 2 m) extended downstream by a trapezoidal one [large base = 1 m; small base = 0.1 m (funnel hole) and height = 0.22 m]. The sheets are sunk 10 cm into the ground, so as to have an overflow of 20 cm. The center of the downstream width of the device is a collecting gutter in the form of a collector. Runoff water was collected through the gutter at the plot outlet. Each plot covers 2.12 m².

After each rainfall event, measurements are taken in sampled dongas to assess runoff and recorded soil losses. The rain episode common to all targeted sites and dongas was studied. In order to limit the biases related to the differences in recorded rainfall, in addition to the rainy episode of October 16, 2010 which was common to all sites and dongas, other episodes with approximate rainfall amounts or occurring in the same period when the vegetation cover is maximum, were selected and analyzed.

As it was impossible to store all the run-off volume, partitioning was carried out using two 25-litre tanks. The first was connected to the gutter by a plastic pipe (20 mm in diameter) with 20 identical holes (upper part). The second tank was connected to the first by a PVC pipe of 20 mm diameter. After each rainy episode, the volume of water runoff on the deviation plot is collected and measured. A homogenized sample is taken to determine the solid load. This solid load is determined by weighing 1.5 liters of homogenized runoff water and extrapolating the results to the m² surface area.

The runoff and soil loss data were subjected to a 3-criteria analysis of variance using GenStat-Release 4.2 software [14]. The effect of donga, site and topography factors and their interactions on runoff and erosion were analyzed at the 5% significance level. Homogeneous groups were determined by comparing means based on the Least Significant Difference (l.s.d).

Overall, the average slope was slightly higher in W Parkthan in the land use areas (3% versus 2.4%). Permeability varied from 0.36 cm∙h⁻¹ (plot 5 of the Kargui donga) to 26.64 cm∙h⁻¹ (plot 4 of the DP2 donga in W Park). Generally, the soils of the park are more permeable (10.8 ± 7.2 cm∙h⁻¹) than those of the land use areas (3.6 ± 3.2 cm∙h⁻¹).

Rainfall in 2010 indicates that the months of May, August and October are the wettest months. Similarly, 1043 mm of water spread over 7 months were recorded in 2010.

The number of rainy episodes varied from 30 (on the dongas of land use areas) to 32 (on the dongas of W Park). Thirty-nine episodes were examined with 59% in the land use areas (Baru: 21% and Kargui: 38%) and 41% in the dongas of Niger’s W Park (DP₁ and DP₂). The rainy episodes studied were: 1) October 16, 2010 (Land use areas with 25 mm of rain and W Park with 17 mm of rain), common to all sites and dongas sampled to compare runoff and soil loss data; 2) October 2 (with 45 mm of rain), October 6 (with 21 mm of rain) and October 11 (with 11 mm of rain) recorded on the dongas of W Park (DP₁ and DP₂) to study the effects of the level of degradation and topography; 3) those of July 30, 2010 in Kargui (with 27.5 mm of rain) and September 3, 2010 in Baru (with 27 mm of rain), then the episodes of equal rainfall of September 20, 2010 in Kargui and August 14, 2010 in Baru (i.e., the same rainfall).

The rainy episode of October 16 that occurred at all sites indicates runoff values that are significantly site and topography-dependent (Table 1). The effect of topography is more pronounced than that of site [(Vr = 12.85; P_Fi = 0.003) versus (Vr = 70.96; P_Fi = 0.014)]. Similarly, the level of degradation and topography influence soil loss. Indeed, while the rainy episode of October 6, 2010 shows a highly significant effect of topography (Table 1, Vr = 10.04, P_Fi = 0.007) in W Park, the episodes of July 30 and September 3, 2010 indicate a significant effect of the level of degradation of dongas (Vr = 48.6; P_Fi = 0.020) in the soil. Only the level of degradation has a significant effect on runoff in Park W following the rainy episode of October 11, 2010 (Vr = 47.8; P_Fi = 0.020).

Vr: Variance P_Fi: Fisher Probability (of significance); **: p <0.01; *: p < 0.05; ns: not significant. -: data not available.

The results of the Three-Criteria Analysis of Variance show no significant effect of Site * Donga, Site * Topography, Site * Donga * Topography interactions on runoff and soil loss (Table 2). These results indicate how complexity there is in the mechanisms explaining the processes of runoff and soil loss in the protected areas and land use areas. The average values of runoff vary from 4.10 mm to 7.4 mm and those of soil loss from 102.6 g∙m⁻² to 462 g∙m⁻² (Table 2).

P_Fi: FisherProbability (of signification); **: p < 0.01; *: p < 0.05; ns: non significatif; the average of runoff and soil erosion followed by the same letters are not statistically different at the 5% level.

For the rainy episode of October 16, 2010 common to all sites, the high runoff in the soil compared to the park would come from multifactorial interactions, including: 1) Higher permeability in the Park W (10.8 cm∙h⁻¹ ± 7.2 versus 3.6 ± 3.2 cm∙h⁻¹). Indeed, high soil permeability leads to high water filtration due to the large pore volume [15]. This increase in infiltration leads to a decrease in runoff; 2) The steeper slope in the park (3.0% ± 1.3% versus 2.4% ± 0.9%). There is evidence that runoff and soil loss increase with the length and degree of slope [16]. Indeed, according to the work of [4], runoff increases rapidly for small changes in slope inclination. However, land degradation is linked to poverty and human pressures, people’s attitudes and values, poor land management, land tenure, and drought [17] which result in overgrazing, unsustainable agricultural activities, overexploitation of land [10] [18]. Degradation has increased desertification, decreased land productivity, and caused the loss of arable land [19]. Overgrazing, unsustainable agricultural activities that do not respect the environment and overexploitation of land destroy the vegetation cover causing bare soils favorable to erosion and soil degradation. These are factors favoring soil erosion and degradation.

The fact that the park and the land use areas are statistically identical in terms of soil loss but well discriminated on runoff would be relative to the low availability of materials likely to be eroded on the land use areas. Indeed, the trampling of animals and humans compacts the soil materials, obviously limiting their transport through the runoff. Moreover, this compaction limits the detachment of soil particles under the effect of rain. The result is low soil loss despite the high runoff in the village areas. This phenomenon has already been highlighted by [20]. Of course, soil crusting is a common phenomenon in semi-arid regions characterized by sparse vegetation and high rainfall intensities. It also affects certain cultivated soils sensitive to threshing. Surface crusting and its effect potential on runoff in a Sahelian context has been widely described [21] [22]. Surface crusts have a strongly limiting action with respect to infiltration due to the impedance effect exerted by their low hydraulic conductivity as well as by the desaturation of the underlying soil that they cause; in effect, the equivalent conductivity of the crust-soil system tends towards that of the underlying soil [23] but with an unsaturated value thus reduced. The in-situ measurement of the hydrodynamic properties of crusty soils is delicate [24] and, since the crusting is rarely uniform, it is difficile to find crusty zones of the same size as the field measuring devices. Runoff from partially encrusted surfaces is measured using visual observation grids and statistical approaches [25]. The effect of a surface crust on the infiltration is largely explained via models adapting the principle of Green and Ampt [26] [27] [28] [29]. In this context of heavy rains, within the plots used for our work, some surface crusting is noted. However, they deserve more attention in order to explain the effects of crusting on runoff and soil loss.

Beginner dongas are distinguished from advanced dongas by the runoff essentially induced by a higher permeability (20.8 cm∙h⁻¹ ± 11.9 versus 9.2 ± 5.4 cm∙h⁻¹). Indeed, the slope and the covering are not very different. This high soil permeability, which is also due to the pore volume, allow high water infiltration. This increase in filtration leads to a decrease in runoff [30]. The high runoff in beginning dongas corresponds to a lower loss of soil. Beginning dongas behave like poorly developed soils for which runoff is high but soil loss is limited [31] [32]. This situation is believed to be due to the low availability of erosion-sensitive materials in beginning dongas. Indeed, the upstream crusting of the beating crust increases runoff while favoring the slow draining resistance of particles rearranged in a more compact manner [20]. This rearrangement would also serve as a resistance to detachment and availability of particles mobilized under the raindrops. Our observations are in agreement with this observation. Thus, the more advanced dongas are remarkable for the important loss of soil (making available the soil that can be dragged by the runoff water), and the less so are the beginner dongas.