2.1. Experimental areas
The study was conducted for three years (2018, 2019, and 2020) in two sites, Rio Verde and Montividiu, both in Goiás State, Brazil (Figure 1). The areas are located in the representative regions of highly intensified and technological agriculture in the Cerrado biome.
Figure 1.
Prior to the installation of experiments, samples were collected for soil characterization (Table 1). Soil liming was carried out according to the technical recommendations for the crops.
Table 1.
According to Alvarez et al. (2013), the climate is classified as tropical Savannah with dry winters and rainy summers (Aw), with an average annual precipitation higher than 1.600 mm. Detailed rainfall data of the field studies during the experimental time is showed in the Figure 2.
Figure 2.
2.2. Treatments
The treatments were arranged in randomized strips of 30 m x 50 m in Montividiu, GO, totaling an area of 1,500 m², and 12 m x 80 m in Rio Verde, GO, totaling an area of 960 m². Before the experiment started the areas were composed of conventional soybean-maize system in Montividiu and degraded pasture in Rio Verde, for ten years. The treatments evaluated were soybean production systems (first crop) followed by cover crops (second crop): 1) mix of cover crops (Pennisetum glaucum, Crotalaria spectabilis, and Urochloa ruziziensis), 2) P. glaucum, 3) C. spectabilis, 4) U. ruziziensis, 5) U. brizantha cv. Marandu, and 6) U. brizantha BRS Paiaguás.
Cover crops were always implanted in a crop season prior to soybean sowing. Cover crops were sowed in the following dates: 03/13/2018 (Year 1), 03/09/2019 (Year 2), and 03/17/2020 (Year 3) in Rio Verde and 02/22/2018 (Year 1), 03/08/2019 (Year 2), and 03/11/2020 (Year 3) in Montividiu. Soybean sowing was carried out on 10/19/2018 (Year 1), 11/04/2019 (Year 2), and 10/17/2020 (Year 3) in Rio Verde. In Montividiu, soybean sowing was carried out on 10/27/2018 (Year 1), 10/25/2019 (Year 2), and 11/05/2020 (Year 3). Soybean harvest was carried out on 02/07/2019 (Year 1), 03/12/2020 (Year 2), and 02/19/2021 (Year 3) in Rio Verde and 03/07/2019 (Year 1), 05/03/2020 (Year 2), and 03/16/2021 (Year 3) in Montividiu. In all years and locations, pest, disease and weed management was carried out in accordance with technical recommendations for soybean cultivation.
2.3. Crop management during the experiment
Cover crops species were sown with 8 kg of pure viable seeds per hectare for Urochloa, 20 kg ha-1 for Crotalaria spectabilis, and 20 kg ha-1 for Pennisetum glaucum. The mix of cover crops was sown with 4 kg of pure viable seeds per hectare for U. ruziziensis, 10 kg ha-1 for Pennisetum glaucum, and 10 kg ha-1 for Crotalaria spectabilis (Calegari, 2019).
A mowing management was carried out on cover crops in order to improve soybean sowing conditions. In Rio Verde, forage plants were cut in July 2018 and 2019; in July 2020, the management was carried out using a knife roller. In Montividiu, plants were cut in July 2018 and handled using a knife roller in July 2019 and 2020.
Cover plants were dried on 09/29/2018 (Year 1), 09/05/2019 (Year 2) and 08/14/2020 (Year 3) in Rio Verde and 09/26/2018 (Year 1), 20/09/2019 (Year 2), and 30/09/2020 (Year 3) in Montividiu. Desiccations were carried out with the application of the herbicide glyphosate at a dose of 2.5 liters per hectare.
The varieties used for soybeans were Syn15640 IPRO - 444,444 plants/ha (year 1), Brasmax Bônus IPRO - 288,888 plants/ha (year 2), and Brasmax Foco IPRO - 311,110 plants/ha in Rio Verde. In Montividiu, ST 797 IPRO - 311,110 plants/ha (year 1) and Brasmax Bônus IPRO - 288,888 plants/ha were used (year 1 and 2).
In Rio Verde, GO, fertilization was carried out every year during the cover crops seeding with 96 kg ha-1 of N, 40 kg ha-1 P2O5, and 40 kg ha-1 of K2O. In addition, a broadcast fertilization was carried out with 8 kg ha-1 of N, 80 kg ha-1 of P2O5, and 80 kg ha-1 of K2O, targeting the entire production system and always carried out before soybean sowing during the dry season. In Montividiu, GO, fertilization was carried out in all years during cover crops seeding with 108 kg ha-1 of N; in addition, a broadcast fertilization was carried out with 100 kg ha-1 of P2O5 and 120 kg ha-1 of K2O.
2.4. Plant and soil evaluations and analyses
The biomass was measured at 117 days, 121 days, and 113 days after sowing in 2018, 2019, and 2020, respectively, in Rio Verde. The same evaluation took place at 135 days, 121 days, and 116 days after sowing in 2018, 2019, and 2020, respectively, in Montividiu. Cover crops were cut close to the ground with a mechanical cutter in an area of 1 m², with 12 replications per treatment. Biomass was weighed to obtain fresh weight. Later, a subsample was removed and placed in a paper bag, weighed, and taken to a forced air circulation oven (55 °C) for 72 h. The dehydrated material was weighed to obtain the proportion of dry matter to later obtain the amount of biomass in kg of dry mass ha-1.
The speed of cover crops decomposition was evaluated after desiccation only in Rio Verde using the method of decomposition bags called “litter bags" (Silva et al. 1997; Espíndola et al. 1998). These bags are made of nylon with an internal area of 0.06 m² and filled with a known mass of cover crops at the time of desiccation. Eight bags were randomly distributed in each treatment for each collection during each time (30, 60, 90, and 120 days after desiccation - DAD). The material collected from litter bags, at each collection time, was dried in an oven with forced air circulation at 55 °C for 72 hours and weighed to determine the remaining biomass by the difference between initial and final mass. Afterwards, plant material samples were ground to determine the concentration of macronutrients (Malavolta et al. 1997).
The nematode population was evaluated 70 days after soybean sowing through soil and root collection following the methodology of Jenkins (1964) and Coolen (1972). In Montividiu, this assessment was carried out in the second and third years; in Rio Verde, in the first and third years.
The evaluation of soybean plant population was performed by counting plants at harvest in two three-meter-long soybean lines with 12 replications per treatment. Soybean productivity was determined by harvesting two lines of soybeans, three meters in length, also with 12 repetitions per treatment. A sample of grains from each replication was taken to determine the thousand-grain mass, and yield data were adjusted to a moisture content of 13%. The number of pods per plant was evaluated from five plants randomly collected in each sampled area.
2.5. Data analysis
All data analyses were performed using the R software v. 3.5.1 (R Core Team, 2018). Outliers were checked using function outlierTest from the package car (Lenth 2020). Normality of residuals and homogeneity of variance were verified using the Shapiro-Wilk and Bartlett tests, respectively.
Soybean grain yield data were analyzed using a linear model considering the effects of treatment, year, location, and the interaction between them. The significance of effects was determined by Analysis of Variance considering type 3 error. Means were obtained by least squares using the emmeans package (Lenth 2020) and compared by Tukey test (p<0.05).
For analysis of nematode population data, the function glm.nb from the MASS package (Venables and Ripley, 2002) was used due to the negative binomial distribution of nematode count values. The fixed effects of year, treatment, location, and the interaction between treatment and location were considered. Means were obtained using the least squares method and compared using Tukey test at 5% probability.
In order to analyze the speed of decomposition and understand the dynamics of nutrient release, data were submitted to analysis of non-linear models using the nlme package. The non-linear model used was: b = y - a * d, where b is a biomass at time d, y is the initial amount of biomass or nutrient, a is the rate of decomposition of the biomass, and d is the number of days (Thomas and Asakawa 1993). Treatments were considered as a random effect on the parameters y and a. Thus, the biomass half-life was calculated. It expresses the period required for half of the residues to decompose or for half of the nutrients contained in the residues to be released. The half-life was calculated using the formula: HL = 0.693/a (Thomas and Asakawa 1993). Analysis of variance was performed for the variables y, a, and HL considering the effects of treatment and year and their interaction and comparing means according to the other variables.