The experiment was conducted in two sites of southern Mali belonging to Soudanian Agro-ecological zone. The first one is located at a technology park in Flola village, district of Bougouni. The second site is at a technology park in Mpessoba village, district of Koutiala. The technology park of Flola is at 11˚42'N latitude, 7˚64'W longitude and 350 m altitude and the technology park of Mpessoba is at 12˚67'N latitude, 5˚71'W longitude and 346 m altitude. The two experimental sites are represented in Figure 1.

The average annual rainfall over the last 46 years (1971-2017), was 857 mm in Koutiala (40 km from Mpessoba village toward north) and 1095 mm for Bougouni (15 km from Flola village toward south) and has an irregular spatio-temporal distribution. Over 37 years (1971-2008), low temperatures occurred between December and February with monthly averages of 16.8˚C and 16˚C, and high temperatures between April and May, with monthly averages of 38˚C and 37˚C, respectively for Koutiala and Bougouni (Figure 2(a)). The daily evapotranspiration was 6 - 7 mm·day⁻¹ in the dry season and 4 mm·day⁻¹ during the rainy season.

Rainfall follows a uni-modal pattern with maximum events occurring in July and August. Enough rain for crop planting without prolonged dry spells that could hurt seedlings after sowing occurs in May and ends of rainy season in October [25]. Monthly rainfall, maximum and minimum temperatures and potential evapotranspiration (PET) were collected from the National Meteorological Service of Mali in Bougouni and Koutiala during the study period (Figure 2(b)).

Dominant soil types in the study areas are classified as leached tropical ferruginous soils with spots and concretions [26] ; Arenosols, Lixisols and Acrisols [27] and Alfisols according to Soil Taxonomy [28], with many Paleustalfs and frequent Plinthustalfs. The Alfisol soil order indicates that the soils are constrained by both small amounts of nutrients and a low capacity to retain nutrients due to the chemical constituents [21]. Soils are characterized by light textured sandy loam topsoil (0 - 20 cm) with 8% clay covering heavy textured subsoil of 20% - 28% clay. Average bulk density is 1.5 g·cm³ at 0 - 20 cm soil horizon. These soils are inherently fragile with weak water retention capacity and poor in plant nutrients [12]. Soil pH varied from moderately acid (pH₂O = 6.5) to acid (pH₂O = 5.5), Organic Carbon (5 - 6 g·kg⁻¹), Nitrogen (<0.5 g·kg⁻¹), available P (<0.07 g·kg⁻¹), exchangeable K (<72 mg·kg⁻¹), Cation Exchange Capacity (CEC) (4 - 6 cmol·kg⁻¹) and a base saturation ratio 75%.

The main farming system in Bougouni and Koutiala is a crop-livestock based rotation system of Cotton (Gossypium hirsutum, L.), Maize (Zea mais), and Sorghum (Sorghum bicolor).

Tested varieties were Sotubaka (improved maize variety) and Pablo (sorghum hybrid).

Rotation head is cotton followed by maize and sorghum allowing cereals to benefit from the cotton residual fertilizer effects. Annual staple crops, sorghum and maize, were planted in the middle and bottom of the catena. Cotton was also planted in the bottom while grazing area was in the top of the catena. Smallholder farmers used extensively animal manure collected in farmyard or sometime compost to improve crop productivity for food (grain) and straw for feed. The rainy season covers May to November for the main cropping period. Income from cash crops and livestock sales was partially used to cover farm inputs and other household needs [29].

For both sites, maize and sorghum were planted from 15-25 June in 2016 and from 18-30 June in 2017 respectively in CRT and NoCRT plots. Planting density was 0.40 m within hills on the row and 0.75 m between rows for maize. For sorghum, density was 0.5 m within hills on the row and 0.75 m between rows. Elementary plot sizes were 34 m length and 7.5 m width for 255 m². One row at each border of the plot was discarded to determine the net plot sizes leading to 34 m length and 6 m width i.e. 204 m². Seedlings were thinned to two plants per hill 15 days after emergence for targeted populations of 66,666 (Maize) and 53,333 (Sorghum) plants·ha⁻¹ which are the density advised by extension services for the area. Thinning was done two weeks after emergence. Base fertilizer was uniformly applied to each treatment (CRT and NoCRT plots) at the rate of 100 kg·ha⁻¹ of NPK (15–15–15) at planting time for both crops. Thirty days after germination, 75 kg·ha⁻¹ of urea for Maize and 50 kg·ha⁻¹ of urea (46% of nitrogen) for sorghum were applied followed by hand hoeing. A second dose of 75 kg·ha⁻¹ of urea was applied 45 days after germination on maize. Fertilizer was buried 5 cm below and 5 cm away from plants on the row banks.

In CRT and NoCRT plots, harvest was done from 2-30 October and from 7-27 October in 2016 and 2017, respectively for maize and sorghum, followed by 20 days of sun drying.

The erosion study was conducted in 2016 and 2017 in the technology park of Flola and Mpessoba, in Mali. Experiment plot was divided in two parts: the first one was under contour ridge tillage (CRT) and the second one with farmer’s practice (NoCRT) as a control. There were 4 experimental plots with 0.75 m width and 34 m length surrounded by an oblique galvanized iron sheets of 55 cm height inserted to a depth of 15 cm to prevent runoff to seep in or out from the plot, corresponding to four replicates for both in CRT and NoCRT plots.

In the paired CRT and NoCRT erosion plots, the surface runoff from each experimental plot was 1/10 diverted by a channel into a collection barrel of 200 liters capacity and an additional barrel for collecting another 1/10 of total runoff water. These barrels were placed in a pit of 1.5 m × 0.7 m × 1.3 m, covered by a metal sheet of 1.8 m × 1.0 m.

Runoff coefficient (Rco) was used to compare the influence of CRT and NoCRT and expressed as follow:

Rco = Rw ( mm ) Ra ( mm ) (1)

where Rw = Runoff water; Ra = Rainfall received;

Rw = Vw ( m 3 ) S ( m 2 ) (2)

where Vw = Volume of water generated; S = Water measurement area.

To measure the average sediment concentration in the runoff water and estimate soil loss in each of the four replicates of the CRT and NoCRT plots. Three water samples of one-liter each were taken in each collecting barrel to maximize accuracy of the operation. These samples were taken after runoff water has been stirred vigorously to better capture sediments in the barrel. Sediments and nutrients concentrations were measured by oven drying at 40˚C for 7 days to constant weight and chemical analysis (organic carbon, N, P, K) whereas concentrations were determined in the Soil Plant and Water Laboratory of Institut d’Economie Rurale (IER). Runoff, soil loss by erosion, and nutrients contents were compared between CRT and NoCRT plots using pairwise t test at 0.05 significant levels. Two rain gauges were installed in each research site. Time-domain refractometry probes were installed, 100 cm deep below soil surface at the middle of CRT and NoCRT plots to measure soil moisture during the whole cropping season. Before rain onset, dry soil moisture content was recorded immediately after trials installation. In order to characterize soil moisture during the cropping season, CRT and NoCRT plots were represented by soil moisture daily measured, at the months of July (beginning of the rainy season), August (at the middle of the rainy season) and September (at the end of the rainy season) in 2016 and 2017.

In each site (CRT and NoCRT) plots were treated the same way in sowing dates, crop species and other cropping operations, except ridging mode.

Soil Water Storage calculation was performed to quantify the water stored in each soil profile,

In each profile it was calculated as the sum of the soil water per depth interval through the profile.

On each site, 40 samples from horizon 0 - 20 cm were randomly collected in May 2016, on 1ha, using an Edelman Combination Auger (4 cm core) of 1.2 m length, mixed to form composite soil samples. Samples were air dried by spreading them on a plastic sheet at room temperature. Composite samples were made from the ones taken in an asterisk shape pattern in each site. Samples were analyzed for both physical and chemical properties. Particle size (soil texture) analysis was performed by the hydrometer method [30], pH was determined by the electrometric method in a soil solution with a soil/water ratio of 1:2.5. Soil organic was determined by the modified Walkley-Black wet oxidation method as outlined by [31]. Total nitrogen was determined by the modified Kjeldahl digestion method of [32], while bases, CEC and available P were determined as described in [33].

Crop yields were measured in central rows while discarding the two border rows on each side of the plot. At harvest, total panicles and cobs, grain and stems dry weights were recorded in the central rows and data extrapolated from the subplot size to hectare. Paired CRT and NoCRT plots data were analyzed as a simple trial in a four-block experimental design to determine the global significance of runoff volume, soil erosion, nutrient losses and crops yields using STATBOX 7.4.4. Newman-Keuls test was used to separate means for significant differences between treatments. Treatments effects were considered significant at P < 0.05.

Total cropping season rainfalls were 730 and 954 mm in 2016 and 635 and 945 mm in 2017 for Koutiala and Bougouni, respectively. Maximum rainfalls of 264 and 391 mm were received in July in 2017, while 260 and 269 mm were observed in August 2016, for Koutiala and Bougouni, respectively. July 2017 accounted for 42% and 41% of the total amount of rainfall in Koutiala and Bougouni, respectively. In August 2016, rain amount was +79 and +65% higher than those of 2017 for the same month, for Koutiala and Bougouni respectively. These increases represented 36% for Koutiala and 28% for Bougouni. In May, at the beginning of the rainy season, on both sites, rainfall was less than 85 mm (Figure 2(b)).

The mean annual minimum temperature was 16˚C and 15.5˚C and maximum temperature was 38.8˚C and 39.5˚C, for Koutiala and Bougouni, respectively.

Lowest mean annual potential evapotranspiration (PET) of 3.65 and 3.20 was recorded from June to September for Koutiala and Bougouni, respectively, while the other months of the year showed values varying from 5 to 6 mm.

Table 1 presents surface soil horizon data on granulometry and chemical properties. Granulometric composition of Bougouni and Koutiala soils was closer. However, soil in Koutiala was slightly sandy than that of Bougouni (+8%), less silty (−2%) and less clayed (−6%). Soil pH (water and KCl) of the study sites were globally slightly acid. The Ca, K and CEC values in Bougouni were at least +40% higher than those of Koutiala while Mg value was +20% higher, compared to the same site. Phosphorus level was low but slightly higher in Bougouni (+7%) than Koutiala. Soil of both sites showed very low values in organic matter and nitrogen.

Rainfall analysis revealed not only inter-annual rainfall variabilities (635 - 1437 mm) but also intra-annual rainfall distributions. This situation is likely to negatively impact crop production as reported by [34] when studying the relation between climate and soil productivity in Sudanian and Sahelian zones of Africa where amount of rainfall distribution in a given year determines crop productivity level [35] [36]. In the same way, [37] mentioned that climate variability was among the main impediment to the realization of the first Millennium Development Goal of reducing poverty and food insecurity through increase of agricultural production in developing countries. Mean temperatures and PET values during the study period were very close to those from 37 years data analysis.

Soils were predominately loamy and sandy (Table 1) indicating a strong susceptibility of surface soil layers to compaction which can lead to severe runoff and erosion, mainly when combined with greater than average rainfall [38] [39]. In both sites, the highest organic carbon content was less than 6 g·kg⁻¹·ha⁻¹ and in line with the low values of nitrogen, phosphorus and CEC, indicating that they had low soil fertility and water-holding capacity. Findings reported in this study agree well with those claimed by several authors [6] [18] [40] [41] [42] for tropical ferruginous soils of West Africa. Findings of [42] showed that in soil surface layers of West Africa, organic carbon content is dependent on soil texture, explaining that the lower the clay and fine silt contents, the lower the soil carbon content. In the same way, the dominant clay type, although low in these soils is kaolinite (1:1 low activity clay) indicating the necessity to apply fine elements such organic matter to improve soils CEC and storage capacity for nutrients exchange [18] [42] [43]. This observation agrees with what of [44] for sandy soils who reported that an increase of 1 g·kg⁻¹ of organic carbon leads to an increase of 4.3 mol·kg⁻¹ of CEC.

For all rainfall events measured, runoff was always greater in the NoCRT plots than the CRT ones. The greater rainwater loss was probably due to high runoff and low infiltration rates during intensive rainfall peaks (Figure 3 and Figure 4). These peaks may be due to soil saturation leading to infiltration reduction rate. In assumption, a light rain event is associated with less runoff than a heavy rain event, which may release a greater volume of water over a short period of time. In this study, runoff peaks were observed in July and August, the rainiest period of the cropping season in both sites, explaining that conditions for runoff to occur are important. In fact, wet soils generally have lower infiltration rates than dry soils since pore spaces are already filled with water due to previous rain events. Soil type also, may play a role in runoff severity because number and pores sizes are reduced in a clay type soil which swells when wet. This situation results in weak infiltration rate and consequently to runoff. Soil in Bougouni has more clay and silt, when saturated and then receives considerable quantity of rainfall, may explain the highest runoff rate compared to Koutiala site. The sandy nature of the parent materials of the studied soils also emphasizes runoff, in agreement with [45] who reported that in most tropical soils in Africa, even not originally sandy, were intensively washed by runoff and leaching that transform their texture to coarse after heavy rainfalls. These observations agreed with those of several authors [5] [21] [46] - [52]. In agreement with findings of this study, it has been reported by many authors [20] [24] [53] that CRT was beneficial for reducing runoff and soil loss, as well as for increasing crop yield and was a holistic landscape approach to managing water and capturing precipitation in farmer’s fields.

Mean runoff coefficient was significantly higher in the NoCRT plots cultivated up and down the slope which was farmer’s practice compared to the CRT plots where ridges followed the contour line. This practice of up and down ridging does not create any kind of resistance to runoff flow, facilitating faster flow of excess water leading to high runoff amounts. This water loss is detrimental to agricultural production because, it reduces availability for crops and can result in severe moisture shortage during dry spells occurrences which are the cropping season characteristic in the area. These results agreed with other findings [5] [18] [53] who have pointed out significant water losses due to runoff in semi-arid lands. Runoff reduction was 46% in our study and agrees with the findings of [21] who reported that the main roles of CRT were capturing and recycling precipitations in treated fields. Also, CRT assists to the evacuation of excessive rainfall and surface fluxes destructive that can trickle into the fields as its application reduced rain runoff from 22% to 61%.

Mean runoff coefficient was higher in 2016 than 2017 for both sites. This situation can be attributed to rainfall difference between year and rain erosivity, in agreement with those reported by [53] when studying runoff in southern Mali.

Mean soil loss was two folds greater in NoCRT plots than CRT ones where the difference was statistically significant (p = 0.02). Similar observations were reported [54] in Sri lanka and in southern Mali [5]. CRT reduced soil erosion in cultivated land by reducing the erosive power of runoff intercepting and its speed [55]. These results support CRT use in cultivated fields to ensure soil conservation. Erosion varied greatly between the two experimental sites. These findings agree with those of many scientists [56] [57] who have widely reported the spatial variation of erosion when studying soil erosion and restoration in Mali and Burkina watersheds, in the semi-arid areas of West Africa. The fact that erosion was greater in Bougouni than in Koutiala can be attributed to 39% higher rainfall amount obtained in Bougouni. Soil loss was also greater in 2016 than 2017 but no year showed erosion above the tolerable limit of 2.5 t·ha⁻¹ and 12.5 t·ha⁻¹ reported by several researchers [57] [58] [59] [60]. This can be attributed to rainfall amount and aggressivity for Koutiala and Bougouni as mentioned by [61] in similar agro-ecological zone of Burkina Faso. Also, soils of the studied area were both sandy and low in organic matter (Table 1) resulting to weak surface structure, then leading to runoff and erosion by surface sealing as similarly reported by [39] for sandy soil of United Kingdom. Our results support the recommendation of organic manure to improve soil surface structure for saving agricultural land from erosion. Globally, the use of CRT reduces runoff velocity and rainfall energy resulting in runoff and soil loss decrease in cultivated lands as supported by [62] who mentioned that a small reduction in runoff velocity can substantially reduce the amount of transported material. These results corroborated with the findings of many researchers [5] [21] [56] [57] [60] [63] [64].

In NoCRT plots, a large amount of organic carbon, nitrogen, phosphorous and exchangeable bases, vital plant nutrients, were annually lost through eroded sediments, compared to the CRT ones. This situation can be explained by the serial disposal of narrowly spaced ridges (0.6 - 0.70 m) and furrow between ridges, which allow rainwater to be retained where it falls, resulting in better infiltration and remarkable slowing down or stopping runoff, erosion and nutrients losses. Higher nutrient loss in farmer’s practice (NoCRT) and reduced nutrient loss in CRT plots in the fields on gentle slope (1% - 3%) is in line with earlier similar studies on CRT [57] [60] [63] - [68]. Carbon, nitrogen and phosphorus were at least two folds greater in the NoCRT plots than the CRT ones. This can be explained through the work of [67] who reported that about 95% of the soil nitrogen and 25% to 50% of the phosphorus are contained in the soil organic matter and a large amount of it is found in the soil surface as decaying leaves, stems and other fine organic particles, facilitating erosion and nutrients losses. This relation among phosphorus, carbon and nitrogen were well explained [69] who mentioned that the loss of soil organic matter was often followed by a corresponding loss of nitrogen. This situation is detrimental for agriculture since several soil quality indicators such as infiltration, water retention capacity, aggregate stability, cation exchange capacity (CEC), nutrient availability to plants are tightly related to organic matter and well documented [6] [52] [70] [71] [72]. Moreover, organic carbon lost from agricultural land could be released as inorganic carbon and increase quantity of greenhouse in the atmosphere. Otherwise, concerning phosphorus, farmers in the area annually apply cotton complex fertilizer composed of nitrogen, phosphorus, potassium, sulfur and boron corresponding to NPKSB (14-22-12-8-1) at the rate of 150 kg·ha⁻¹, buried in the soil surface using animal traction. Therefore, increasing runoff may increase phosphorus amount in eroded soil since the concentration is high in the upper soil layer as reported by [73] who concluded that 80 percent of eroded P were transported in surface runoff from most cultivated land during flow events. Similar findings were mentioned by several authors [74] [75] [76] [77] when reporting on factors affecting nutrient losses.

It was observed that higher quantities of carbon, nitrogen, calcium, potassium, phosphorus and magnesium were lost in eroded soil in 2016 compared to 2017. This could be attributed to higher and severe rainfall events in 2016, where, for instance, in Flola, 31 July and 07 August rains produced 60 mm and 70 mm corresponding to I₃₀ of 67- and 75-mm·h⁻¹, respectively. For Mpessoba, in 2016, 21 July and 07 August rains produced 76 and 65 mm corresponding to I₃₀ of 80 and 72 mm·h⁻¹, respectively. Additionally, in Flola, 32 rain events producing runoff were recorded in 2016 against 19 in 2017. For Mpessoba, the number of rain events was almost the same (21 in 2016 and 22 in 2017) but rain intensity and quantity were greater in 2016.

In Mpessoba and Flola, soil moisture was always higher in the CRT plots compared to the NoCRT ones. CRT technology is applied to reduce runoff, which therefore increases infiltration and soil moisture as demonstrated by the work of [18] who, for instance, reported an average moisture difference of 25% at 60 cm depth when comparing CRT and NoCRT, plots. In fact, CRT plots benefited from rain water captured as reported [21] who mentioned that CRT increased soil moisture in areas explored by plant roots by 16% to 64% compared to NoCRT plots. At the end of the season, soil moisture was at least 25% on CRT plots explaining a real water supply potential for the park land trees as reported by [21] when assessing the effects of CRT on soil water dynamic in Siguidolo and Fansirakoro, in Mali. This finding also corroborated with [78] regarding assessment of farmers’ perception on the effect of CRT in Cinzana (Mali) who showed that with the use of CRT, soil moisture was found to be better conserved and allowed field operations for an extended 7 to 10 days.

Water storage was always higher in CRT plots than the NoCRT ones in both sites. These observations were supported by [20] who reported 17% of water storage in CRT plots in the 80 - 160 cm profile horizons and 12.7% in the first 80 cm. Similar trend was mentioned [18] when studying the effect of CRT on soil moisture in a 60 cm soil profile in Sahelian area of Mali. They concluded that, the use of CRT can result in reducing soil erosion by reducing precipitation water runoff. It allows more time than the control for rainwater to infiltrate, therefore increasing water storage. This leads to better growth and higher yield during cropping seasons with unpredictable rainfall or low total rainfall as reported by many authors [18]. [19] [22] Also, this situation can be important for crops sown late in the growing season but also for trees to continue surviving during the dry and hot season when the maximum temperature may reach 46˚C.

Significantly lower maize and sorghum grain and straw yields were obtained under NoCRT plots, a traditional cropping system [79]. Part of this trend explanation was the high amounts of runoff and erosion, which resulted in low moisture and nutrient availability for plant growth. These observations corroborated with several researchers’ findings who reported that CRT allowed water accumulation which became significantly available for crop to accomplish its physiological processes of biomass accumulation and grain filling compared to e NoCRT plots [4] [5] [20] [22]. In related studies [64] [80] reported that CRT increases soil nutrients and available soil moisture for crop uptake and enhances crop growth and dry matter yield compared to the NoCRT plots.

Higher grain and straw biomass sorghum yields compared to the national average yield of 1000 kg·ha⁻¹ of grains, could be attributed to growth and genetic characteristics of hybrid crop. In fact, improved varieties have a greater ability to convert assimilates to grain and biomass as reported by [18] when studying the effect of CRT on improved and local cereals varieties in Sahelian area of Mali. However, sorghum hybrid grain yields obtained in this study were lower than those reported by [81] in Mali which averaged 3500 kg·ha⁻¹. One of the explanations of higher yield reported by [81] could be sites difference since their trials were implemented in the ICRISAT Research Center at Samanko where all fields operations were under control.

The average maize grain yield of 2600 kg·ha⁻¹ for the same variety reported by [82] in southern Mali fell within the range of 1729 to 2732 kg·ha⁻¹ obtained in this study.

This study, besides what has been mainly reported in Mali, combines determination of runoff, erosion, nutrient losses and crop yields at the same time. These data on soil and yield are strong decision-making tools for agricultural policies under rainfall conditions in Sub-Saharan Africa. Thus, farmers can increase crops grain yield by 50% and up to 87% for maize grain. For draught and fattening animals, there are gains of 116% and 70% for sorghum and maize straw yields, respectively. All these advantages were obtained when CRT was applied. Consequently, this study pleads for training of farmers, NGOs and extension agents on the contour ridge tillage technique for wide up scaling, targeting sustainable crop production through more water and nutrient conservation to mitigate recurrent drought.