Efficiency of four different seeded plants and native vegetation as cover crops in the control of soil and carbon losses by water erosion in olive orchards

2.1 Experimental plots

The experiment was conducted over 2 years in the growing seasons 2011–2012 and 2012–2013, in six rectangular experimental plots of 5 × 10 m² located on a 20% slope in the IFAPA (Andalusian Institute for Research and Training in Agriculture, Fishing and Food) “Alameda del Obispo” Experimental Station in Cordoba (Spain; 37°51′39″N, 4°48′4″W), in a 6 × 5 m² plantation pattern of olive trees (Figure 1). The runoff and sediment were collected in the channel at the bottom of the plot, draining into a tipping bucket gauge (Edwards, Jackon, & Fleming, 1974).

The soil in the experimental plots, whose main physicochemical properties are shown in Table 1, has a loam textural class and belongs to the Typic Calcixerept subgroup (Soil Survey Staff, 2010).

Six treatments were established. Conventional Tillage (CT) was compared with sown or spontaneous cover crops as soil management systems. Four species were seeded—Brachypodium distachyon (BD), Sinapis alba (SA), Vicia sativa (VS), and Vicia ervilia (VE)—and compared with a treatment of spontaneous weeds (SW).

For the sown cover crops, the species belonging to the gramineous family (B. distachyon L. Beauv. (Poaceae)) was chosen because of its good ground protection as was tested by other authors (García-Díaz et al., 2017; Ruiz-Colmenero et al., 2011; Ruiz-Colmenero et al., 2013; Sastre et al., 2017). Moreover, it is not very competitive with olive trees and is easy to control, making it suitable for cover crops. S. alba L. subsp. mairei (H. Lindb. Fil.) Maire was chosen as a cruciferous plant, because it is common in olive orchards. This plant is also useful for its phytosanitary potential, especially Verticillium dahliae (Jurado-Bello et al., 2013) currently under research. Leguminous crops are suitable for their potential fixing of atmospheric nitrogen and are highly effective for green manures (Stagnari, Maggio, Galieni, & Pisante, 2017). Two legumes commonly used as cover crop species in Mediterranean areas, V. sativa L. and V. ervilia L. (Gómez-Muñoz, Hatch, Bol, & García-Ruiz, 2014; Sastre et al., 2017), were chosen for this work. BD and the leguminous crops were sown at a rate of 100 kg ha⁻¹ of cover. SA was sown and slightly buried following the procedures established in previous field studies (Alcántara et al., 2009) at a rate of 10 kg ha⁻¹ of cover. The sowing dates depended on yearly weather conditions, but all the species were sown manually in autumn during October and November.

One plot was dedicated to the spontaneous vegetation (SW), keeping the natural flora of the area. The dominant species identified were Calendula, Bromus, and Hordeum. Later, in spring, after irrigation produced by simulated rainfall, other species appeared such as Avena barbata, Erodium malacoides, Convolvulus arvensis, Crepis vesicaria, and some mallows.

The tillage plot was managed with a rototiller at 20-cm depth three times yearly. In the other plots, the weeds in the olive row strip (2 m wide) were controlled by systemic herbicide. In Figure 1, shaded stripes represent surface occupied by cover crop.

2.2 Rainfall simulation procedures

In order to compare water, soil, and nutrient losses between treatments, a rainfall simulator was used. The simulator consisted of a metal framework forming a rectangle and six sectorial sprinklers (VYR-802), located at the contour of the plots at 3 m above the soil surface. There were three sprinkler heads per side (Figure 1). Preliminary tests were carried out to select rainfall intensity and to measure rainfall uniformity. In each series, two rainfall intensities were used: between 36.2 and 41.4 mm hr⁻¹, which can be considered as high-intensity rainfall (HIR) for the area, and between 15.6 and 20.7 mm hr⁻¹, as a medium-intensity rainfall (MIR). The return period in the zone for rainfall of 40 mm hr⁻¹ during 1.5 hr is 18 years and for a rainfall of 15 mm hr⁻¹ during 3 hr is 4.4 years (Ayuso, J.L., personal communication). HIR was obtained with 250-kPa pressure and 4-mm nozzles, and MIR was achieved with 200-kPa pressure and 3-mm nozzles.

Twelve simulations for each plot and year were carried out from March to July in order to avoid many natural rain events, which occur mainly in autumn and winter, and to have a greater uniformity in the simulated rainfall due the low wind velocity recorded in this period. The simulations were grouped into two series: initially with a cover crop before mowing and later mowing them. Before mowing, 1.5-hr HIR and 3-hr MIR trials were performed. After mowing, 3-hr MIR and 2-hr HIR trials were run. There were three replications for each trial within 2–3 weeks depending on the atmospheric conditions. The plots were irrigated 2 days before each run to reach saturation.

2.3 Sampling and carbon analysis

The percentage of cover was estimated by measuring the corresponding fractions under a square grid of 1 m² divided into 100 equal squares (Agrela, Gil, Giráldez, Ordóñez, & González, 2003). The measurements in a 0 to 5 scale were integrated, and the final estimation was the average of 10 randomly selected points in every trial.

To check the initial soil moisture in the plots, soil samples were taken using an Edelman auger before each test. Four depth intervals were taken (0–5, 5–20, 20–40, and 40–60 cm) in three points per plot. The average gravimetric moisture in the soil profile at the beginning of the experiment is shown in the Supporting Information (Table S1). Some cylinder cores of a known volume were used to measure the bulk density.

Although previous tests had been carried out to check rainfall intensity and homogeneity, eight rain gauges were used during each simulation to measure the actual rainfall intensity. The Christiansen uniformity coefficient was used to evaluate the rain uniformity:

(1)

where X is the rainfall depth measured in each rainfall gauge, the average depth, and n the number of rainfall gauges.

Runoff generated and discharge over simulation time were measured with the flow gauge. Erosion was evaluated by a periodic runoff sampling of 1 L in volume at between 10 and 20 min, depending on the changes in flow. The samples of the runoff were weighed, oven dried, and weighed again to obtain the sediment dry mass. The sediment load in the runoff channel was also collected at the end of the simulation event.

The SOC loss was calculated by analysing the percentage of organic carbon in sediment samples. The determination of organic C was based on the Walkley–Black method (Sparks et al., 1996).

2.4 Data analysis

An analysis of variance was performed for runoff, ground coverage, and soil and carbon loss field data, and a comparison of means was carried out by the least significant difference test with p ≤ 0.05. Previously, the homogeneity of variance, the random distribution of residuals, and the normal distribution of errors had been tested. The experimental design of randomized blocks was considered, the blocks being each replication of the simulated rainfall on the plots in a trial.

2.5 Kinematic wave model

The proposed model is based on the approximation of the kinematic wave to the Saint-Venant expressions with a mass conservation equation for the water (Equation 2):

(2)

where q is the overland flow; t and x are temporal and spatial coordinates, respectively; h is the water depth; and r and f are rainfall and infiltration rates, respectively.

In this model, the momentum conservation equation is reduced to a relationship between flow rate and water depth (Equation 3) with the exponent α, representing the inverse of the resistance of the soil surface to water movement, and the coefficient m:

(3)

By substituting Equation 3 in Equation 2,

(4)

The term αmh^m − 1 characterizes the celerity of the kinematic wave.

Boundary and initial conditions were null water depth:

(5)

The method of characteristics (Courant & Hilbert, 1962) was used to solve Equations 4 and 5 obtaining two ordinary differential equations that express the variation of the water depth with respect to time (Equation 6),

(6)

along the characteristic curves (Equation 7),

(7)

Ponding time, t_p, the time elapsed until runoff starts from the beginning of the rain, can be expressed (Smith & Parlange, 1978) as

(8)

K_s is the saturated hydraulic conductivity and S the sorptivity of the soil. This parameter is related to the effective net capillary drive, G (Smith & Parlange, 1978), by the initial and saturated water contents, θ_i, and θ_s, respectively,

(9)

The infiltration rate, f, can be expressed, according to Smith and Parlange (1978) as

(10)

The characteristic curve equation is found by integration of Equation 7,

(11)

The integral of Equation 11 was numerically evaluated with the Gauss–Legendre scheme, so water depth, h, could be obtained for each place, x, and time, t. Then, overland flow exiting the plot q(t) was calculated with Equation 3. The parameters (K_s, S²/2, α, and m) were estimated by optimization using the Rosenbrock algorithm (Press, Teukolsky, Vetterling, & Flannery, 2007). The Nash and Sutcliffe efficiency (E_NS; Ritter & Muñoz-Carpena, 2013) was calculated as an index of the goodness of the fit of the flow rate, q:

(12)

where and are the observed and calculated flow values, respectively; and is the average for the observed flows in each trial.

3.1 Runoff, soil, and SOC losses

Average intensities of 38.8 (±2.3) and 18.1 (±1.6) mm hr⁻¹, for HIR and MIR, respectively, were obtained. No significant differences in rainfall intensities were found between treatments in any series, so that the runoff yield and soil and carbon losses were comparable for the same rainfall rate type between treatments. The average values of the Christiansen uniformity coefficient were 91.3 for HIR and 86.2 for MIR. Detailed rainfall rates and uniformity coefficients are available in the Supporting Information (Table S2).

The average runoff values ​​obtained from the three replications in each series are shown in Table 2a. Significant differences with CT were found in almost all the series, except under freshly tilled soil, where water infiltration was improved over previous consolidated conditions. This happened in the first series of trials in the first year, in which SW and VS produced a large runoff volume but still lower than the corresponding value for the tilled plot. Nevertheless, the infiltration benefit of tillage was only temporary. At the end of the trials after 2 years, all treatments, compared with the tillage system, of seeded plants reduced the runoff generated by over 70%. With respect to the SW, SA yielded less runoff volume than did other plant types. The extensive root system of the crucifers enhanced infiltration, and the good biomass production of the SA provided high surface coverage (Alcántara et al., 2009; Alcántara, Pujadas, & Saavedra, 2011; Wolfe, 2000). Crucifer species usually increase saturated hydraulic conductivity in comparison with shallow root species (Ola, Dodd, & Quinton, 2015), facilitating water percolation (Archer, Quinton, & Hess, 2002). These results indicate the water conservation potential of cover crops in the spring and summer.

Soil losses were generally higher in the trials yielding a large runoff volume. Figure 2a shows the relationship between total runoff volume and soil loss in every simulation. The differences observed between trials were greater in the total mass of sediment than in the total runoff volume generated, the CT plot being more erodible than plots of other treatments. The reduction in soil losses in the different treatments with respect to CT was over 90% (Table 2b). In the first year, the compaction produced in SW plot without any seeded species to reduce water consumption increased the runoff; nevertheless, the sediment concentration in the flow was lower than in CT. Not considering the tilled plot, SA soil losses were statistically lesser than VS and BD. These species suffered growth problems during the first year, which reduced soil protection and increased the losses especially under the HIR.

Cover crops, either sown or spontaneous, were more effective in reducing soil losses than was runoff volume, similarly to results provided by Shi et al. (2013). The characteristics of the plant species are decisive for the runoff yield as was already reported by De Baets et al. (2009). All the cover crops were very effective at reducing losses from MIR, with BD and SA decreasing erosion almost completely with this rainfall rate. The average concentration of sediments in runoff was below a threshold value of 5 kg m⁻³ in almost all treatments except tillage as estimated from the data in Table 2.

SOC losses were also higher in treatments with larger soil losses (Table 2c). Occasionally, the low content of organic matter in tilled soils (Nieto, Castro, & Fernández-Ondoño, 2013) can induce a smaller loss than can other initially richer soils, mainly with not very erosive rainfall events (Almagro et al., 2016; Martínez-Mena et al., 2012). Figure 2b represents the percentage of organic carbon and soil losses. The organic carbon concentrations show a descending trend with increasing erosion, and, consequently, with increasing runoff. The organic carbon contents in the sediments from the CT plot were generally lower than in the treatments with cover crop.

Table 2b and 2c presents the overall differences for SOC and soil loss between treatments. Our soil loss reductions agree with those obtained by Francia, Martínez-Raya, and Ruiz-Gutiérrez (2000) and Gómez, Sobrinho, Giráldez, and Fereres (2009), who obtained 92% and 93% reductions, respectively, measured with respect to a bare soil treatment. Comparing different types of seeded and spontaneous cover crops to tillage in grape vines and olive trees in southern Europe, Gómez et al. (2011) detected large reductions in soil loss, SOC, and nutrients. In a study conducted in eight olive orchards in Andalusia (Spain), Espejo-Pérez, Rodríguez-Lizana, Ordóñez, and Giráldez (2013) compared tillage and cover crop systems in microplots. Their results gave an average reduction of 76% using cover crops. Sastre et al. (2017) found that cover crops reduced the soil loss by between 40% and 80%, the reduction being higher with gramineous than leguminous plants. This was the opposite to the findings in which BD gave worse results than did those obtained by Sastre et al. (2017) because of establishment problems.

Studies in vineyards, performed with real rainfall, indicated high differences in runoff and soil loss due to soil management. Biddoccu, Ferraris, Opsi, and Cavallo (2016) in a long-term experiment obtained a reduction of over 37% in runoff when comparing grass cover and CT; reduced tillage and grass cover decreased by nearly three times the mean soil losses measured in the tilled plot. Reductions of between 40% and 75% were recorded by Novara et al. (2011) with different cover crops; gramineous and leguminous species and mixes were used in an experiment in which crucifer cover crops were not considered. Erosion, compared with tillage, was reduced by 87% by B. distachyon, and the SOC losses by 67% in the experiment of Ruiz-Colmenero et al. (2013); and they also obtained a higher SOC concentration in the sediments from the cover crop plots. In an experiment carried out in an almond orchard, Almagro et al. (2016) obtained a 67% reduction in SOC loss in sediments by using reduced tillage and green manure.

Some authors indicate that tillage could decrease soil organic matter by between 30% and 50% in a few years (Robert et al., 2004), and that losses could reach 60% (Jones, Yli-Halla, Demetriades, Leifeld, & Robert, 2004). In our trials, the SOC loss was decreased by over 90% with the use of cover crops, slightly higher results than those of Márquez-García, González-Sánchez, Castro-García, and Ordóñez-Fernández (2013), who obtained a reduction of 80.5% of soil loss and 67.7% of SOC loss, in 4 years. The reductions in this study agree with those obtained by Gómez, Guzmán, et al. (2009), who used a cover crop of Lolium in a 4-year experiment carried out in bigger plots and on an 11% slope under natural rainfall.

3.2 Percentage of cover

The average percentages of ground cover over each trial, year, and total in the study period are shown in Table 3. After mechanical mowing, the surface cover may even slightly increase, due to the standing stem-covered area being smaller than the one protected by freshly mowed plants. Nevertheless, those differences decrease as the residues gradually decay. In some cases, such as in SW, some differences increased due to the emergence of weeds in summer. Parallel to the mowing performed in the plots with cover crop, tillage was conducted in the CT plot, which had also been tilled before carrying out the tests each year. In this management system, the usual practice consists of at least two tillage passes per year.

The greatest losses usually corresponded to lower coverage values, which occur mainly in the CT plot. Similar results were provided by Podwojewski, Janeau, Valentin, Lorentz, and Chaplot (2011) using a rainfall simulator in a plot of 1 m² in size. The mulching generated after mowing also kept the soil protected and reduced soil and SOC losses as reported by Gholami, Sadeghi, and Homaee (2013) with straw mulching, and by Prosdocimi et al. (2016) in a vineyard. In a recent study in an apricot orchard, Keesstra et al. (2016) observed the efficiency of soil protection with plant cover and chipped branches, in comparison with bare or tilled soil, reducing runoff and soil losses. Cerdà et al. (2016) used a 60% mulch coverage to reduce runoff and erosion in persimmon orchards. A similar ground cover was reached by all cover crops from our experiment in almost all the tests run.

Ramírez-García, Gabriel, Alonso-Ayuso, and Quemada (2015) reported more extensive ground cover achieved with grasses than with S. alba and V. sativa in the developing stage (until mowing). However, the grass used in our experiment (BD) had growth problems in the first year, as was also reported by Saavedra and Alcántara (2011), and thus gave a lesser soil protection. In the trial with HIR before mowing in the first year, BD and VS had higher soil loss than had other cover crops due to the lower ground coverage provided (Table 3). In the second year, all the seeded species were well established, and the soil loss was decreased before mowing. Legumes are creeping plants that allowed the extension of their shoot system over the rows strips surface, increasing the coverage percentage in the second year.

The ground coverage values obtained with any cover crop, either before or after mowing, exceeded 30%, a basic threshold that is internationally considered in conservation agriculture to keep the soil protected (Conservation Tillage Information Center, 1990). This is also in accordance with the studies reviewed by González-Sánchez, Veroz-González, Blanco-Roldán, Márquez-García, and Carbonell-Bojollo (2015) in Spain.

3.3 Estimation of soil and surface characteristics through the application of the kinematic wave model

The model was fitted to the HIR before the mowing field data hydrographs obtained in the first series. Table 4 shows the estimated value of the parameters (K_s, S²/2, m, and α) and the Nash and Sutcliffe efficiency (E_NS) of the respective fits. One hydrograph chosen from every HIR before mowing treatment is shown in Figure 3. In general, the model fitted the runoff experimental data well, mainly in the SW plot, in which a steady state was reached (Figure 3).

The field saturated hydraulic conductivity values remained within a very narrow range, close to the average value of the Rawls, Brakensiek, and Saxton (1982) table for the loam textural class, namely, 13.0 mm hr⁻¹. Apparently, the treatments did not cause any great changes in soil structure. The average sorptivity values were smaller in the VS, VE, and SW treatments than in the tilled and other cover crop treatments. With the conversion of the data into effective net capillary drive, G, with the information on the initial moisture content, with the bulk density value, 1.58 Mg m⁻³, and the saturated water content, θ_s = 0.435 m³ m⁻³, the resulting values were, again, close to the average value of the corresponding textural class in the Rawls et al. (1982) table, that is, 108 mm. The effective net capillary rate of the two Vicia and SW treatments was lower than that of the other three, possibly due to the initially drier conditions on the surface. The values of the exponent m were close to the 5/3 value corresponding to the Manning equation, as is usual in these runoff plot trials. Nevertheless, the exponent α, which is the inverse of the resistance to the surface water flow, was clearly lower in all the cover treatments than in the CT treatment, indicating the influence of the vegetation on retarding the runoff water flow. Possibly in a middle- or long-term cover crop treatment, the potentially beneficial effects on soil structure could be reflected in the above-mentioned soil properties estimated with the rainfall–runoff model. At any case, the model has been effective in the estimation of the hydraulic conductivity at saturation and of the effective net capillary drive.

Experimental ponding time and runoff in each run has been compared with the value obtained by Equation 8 with a good agreement as shown in Figure 4a, b, which supports the use of the proposed model as an efficient tool for interpreting runoff trials in controlled plots.