Dispersion is essential in crop residue application

Site description and conditions

Majuro Atoll, the capital of the Republic of the Marshall Islands, is located in the Pacific Ocean near the equator. The maximum and minimum monthly average temperatures are 29.4 to 30.2°C and 24.7 to 25.0°C, respectively. The average monthly precipitation is 169 to 356 mm, and the average annual precipitation is 3, 365 mm (CLIMATE-DATA.ORG). Coral sand, the soil of Majuro atoll, has relatively high organic matter levels in top soil due to refractory organic matter (organic carbon of 46.9 g kg⁻¹ at 0–15 cm and 10.8 g kg⁻¹ at 15–45 cm) (Deenik & Yost, 2006), and high percolation rates (1.4–3.5 × 10⁻³ m s⁻¹) (Hunt & Peterson, 1980). Thus, the climate is stable, the soil has high permeability with no waterlogging, and frequent rainfall keeps the soil moisture close to field capacity, so applied organic matter is almost decomposed in the crop period. It is known that the residual effect of organic resources is small for sandy soils in a warm and humid climate (Chivenge et al., 2011).

The experimental field was located at Laura Farm (7°8"34"N, 171°2"9"E), which belongs to the Ministry of Resources and Development. The field size was 12.5 × 7.2 m. The plot size was 1.2 × 6.0 m, and the experimental field consisted of two sets of six plots. The government forbids the use of synthetic fertilizer and chemicals on this atoll in order to protect the underground aquifer; therefore, farmers use copra cake as fertilizer. Soil water-soluble NO₃-N (0–5 cm layer; 1:2.5 soil:water extraction) was 4 µg g soil⁻¹, as assessed by a nitrate ion meter (LAQUAtwin B-742, Horiba, Tokyo). No significant residual effect has been found in this field, although corn had been cropped three crops without fertilizer nor any chemicals from November 26, 2013 to July 28, 2014 (see raw data (Oda, 2018)).

Materials and crop management

Corn residue to be used as the material for the experiment was harvested on July 28, 2014, just after the harvest of the previous crop of corn in the experimental field. The residue was divided into 12 bundles of the same fresh weights of 4.7 Mg ha⁻¹ (2.1 Mg ha⁻¹ in dry weight) and applied to the plots by different methods (described later). Corn (Zea mays L.) was planted in two rows at 0.5 m row spacing at a population of 3.3 plants m⁻² on August 4, 2014, and harvested on October 21, 2014 (78 days after seeding). Irrigation, fertilizer and chemical applications were not performed. Hand weeding was performed 2 and 5 weeks after seeding and the weed residue was left on the soil surface of the plot.

Treatments

An L₁₂ orthogonal array design was used in the experiment (Taguchi, 1986). This design was used to ensure robust experimental results under a wide range of conditions. Only the main effects were seen in the experiments, and all the factor levels had six replications. The 12 experiments consisted of four factors (fungi propagation, cutting, dispersion and incorporation) that were randomly assigned to the experimental plots (Table 1). The four different treatments were established as follows. In the fungi propagation factor, the residue was stored in a compost house for 7 days until propagating fungi, and the remainder of the residue stored beside the field. For the cutting factor, the residue was cut into 10 cm pieces. For the dispersion factor, the residue was scattered evenly on the plots, and for the non-dispersion, the residue was placed on the center line of the plot in a swath approximately 30 cm wide. For the incorporation factor, I removed approximately 5 cm thickness of soil of the plots, then placed the residue, and covered the residue with the soil (Figure 1).

Figure 1. Photograph of the condition of the experimental field.

The 12 experiments consisted of four factors (fungi propagation, cutting, dispersion and incorporation), which were randomly assigned to the experimental plots.

Validation of the linearity of input and output

Strong linearity must be found between returned residue quantity and the yield if the adopted factors have a robust performance (Taguchi, 1986). For validation purposes, I returned the crop residues into the same plot using the combination of the two most effective factors (dispersion and incorporation). The plot effect can confound the yields in this plot design since the quantity (0.4–2.0 Mg ha⁻¹ in dry weight) of applied residues was proportional to the preceding yield in each plot. However, this conflict was considered to be small because the plot effect was very small in this field (see raw data (Oda, 2018)) and the residual effect of corn was small. I seeded corn on October 28, 2014 and harvested it on January 9, 2015 (73 days after seeding).

Determination of yields and statistical analyses

I got precipitation data from the automatic weather station at the Laura Farm. I weighed the fresh weight of the whole crop and the kernels in each plot and checked the kernel/whole ratio was stable (R² = 0.9983). The aboveground dry matter (DM) was calculated by multiplying the fresh weight to an aboveground DM/fresh-kernel-weight ratio (2.79) of an existing study (Miura & Watanabe, 2002). The aboveground residue DM was calculated by multiplying the DM/fresh ratio (0.46) of air-dried residue sample. Statistical Effect size analysis was conducted using the effect size of Cohen’s d (Cohen, 1992) by the MS Excel 2016.

d: Effect size, M: Mean, SD: Standard deviation

A significance is affected by the sample size; however, Cohen's d is not affected by sample sizes and can evaluate the true effect. The p values were calculated with an unpaired, one-sided, unequal variances t-test. I evaluated the linearity of the input and the output using the coefficient of determination (R²) in a simple linear regression.

Climate conditions

There were no irregular climate conditions during the experimental periods. The number of precipitation days for each 78-day and 73-day crop period were 58 and 54 days, respectively, and total precipitation was 1111 and 542 mm, respectively.

Effect size of the application methods

Table 2 shows the results of the Effect size analyses. The effect sizes of Cohen’s d < 0.2, 0.5, 0.8 and 1.2, and d > 2.0 correspond to small, medium, large, very large and huge, respectively (Cohen, 1992; Sawilowsky, 2009).

Dispersion had a larger effect (p = 0.045, Cohen’s d = 1.2) than incorporation (p = 0.223, Cohen’s d = 0.7). Fungi propagation and cutting had no effect (Cohen’s d = 0.0–0.1). I calculated the effects of dispersion and incorporation, assuming the effects of cutting and fungi propagation to be negligible. The interaction was huge (p = 0.005, Cohen’s d = 4.9), while the main effect of dispersion (Cohen’s d = 0.3) and incorporation (Cohen’s d = −0.2) were medium and small in effect size, respectively (Table 3).

Linearity of input and output

Under the combination of dispersion and incorporation, input (the quantity of applied residue) was proportional to the output (harvested aboveground biomass) (Figure 2; R² = 0.8963).

Figure 2. Biomass output based on the quantity of corn residue input.

Crop residues were returned into the same plot using the combination of the two most effective factors (dispersion and incorporation).