The trial was conducted at the Professor Ignacio Salcedo Research Station of the Instituto Nacional do Semiárido (INSA), located in the municipality of Campina Grande, state of Paraíba, Brazil (07º14’00” S, 35º57’00” W; 491 m above sea level). The climate is characterized as As’ or dry tropical. The average rainfall is 503 mm annually, and the soil is classified as Solonetz (WRB, 2015); (Santos et al., 2018). The experiment was conducted from 18 April 2022 to 9 December 2023. Figure 1 shows soil water balance (Camargo & Camargo, 2000) and weather data during the trial period.

Gliricidia fodder bank used in the trial was grown in 2017 with no fertilization. The fodder bank area is comprised of 0.15 ha, with 500 trees planted after producing seedlings in the greenhouse of Semiarid National Institute of Brazil (INSA). Seedlings were transplanted to the site at 50 cm height, spaced 1.5 m apart between rows and plants. Trees were assessed in the fodder bank regarding morphological traits (plant height and canopy diameter) from 14 February 2020 to 24 August 2021 and then harvested at 50 cm stubble height.

The experimental design was a randomized complete block with three replications (experimental plots measuring 20 m² formed by 16 plants each) and six treatments comprised 0%, 10%, 25%, 50%, 75%, and 100% of the recommended dose of bovine digesta based on soil chemical properties, bovine digesta chemical properties, and technical guidelines. For that, ten soil samples were taken from the trial site in a zigzag way at a 40-cm layer and homogenized to form a composite sample. After that, the composite sample was analyzed for chemical properties.

Soil chemical properties on 10 January 2022 were 1.00 and 58.65 mg/dm³ for P and K, respectively, besides 0.04, 0.93, 2.17, 3.15, and 0.26 cmolc/dm³ for Na, Ca²⁺, Mg²⁺, H⁺, and Al³⁺. Organic carbon concentration was 0.59%, pH was 5.22, the organic matter concentration was 1.02%, the effective cation exchange capacity was 3.55%, and the soil base saturation was 48.95%. On 7 January 2023, soil chemical properties were 11.4 and 132.9 mg/dm³ for P and K, respectively, besides 0.09, 2.30, 1.40, 1.60, and 0.10 cmolc/dm3 for Na, Ca²+, Mg²⁺, H⁺, and Al³⁺. Organic carbon concentration was 3.10%, pH was 5.94, the organic matter concentration was 2.30%, the effective cation exchange capacity was 4.23%, and the soil base saturation was 70.84%.

The bovine digesta was obtained from the Vera Cruz slaughterhouse in Campina Grange municipality, Paraíba state, Brazil. The bovine digesta fertilizer was obtained from the rumen emptying at the evisceration made by the employers, followed by sun drying. The top average temperatures in Campina Grande in 2022 (year 1) and 2023 (year 2) were 24.2 and 24.5 °C, respectively. The bovine digesta was analyzed in terms of concentrations of N (Horwitz, 2005), P, K, Ca, and Mg (Bezerra Neto & Barreto, 2011). P and K values were transformed into P₂O₅ and K₂O (Table 1) to calculate the necessary dose to fertilize gliricidia fodder banks. P, K, Ca, and Mg contents were analyzed in the Multipurpose Laboratory of INSA (LABINSA). In contrast, the N content was obtained in the Laboratory of Animal Feed and Nutrition of INSA (LANA-INSA).

According to the (Cavalcanti, 2008)handbook, the levels of N, P, and K required to fertilize the gliricidia fodder banks were 1, 30, and 20 kg/ha/yr of N, P₂O₅, and K₂O, respectively (1-30-20 of NPK). Thus, considering the soil’s chemical properties and the bovine digesta potential as fertilizer (Table 1), we calculated the entire bovine digesta level needed to address 1, 30, and 20 kg/ha/yr of N, P₂O₅, and K₂O using the rule of three, and found 12.50 t/ha/yr for that. This value (12.50 t/ha/yr) was converted into 25 kg per plot per year, considering the 20-m² plot size. The increasing doses were calculated from that: 0, 10, 25, 50, and 75% of the total recommended dose were applied at 0.00, 1.25, 3.125, 6.25, and 9.375 t/ha/yr, converted into 0.00, 2.50, 6.25, 12.50, and 18.75 kg per plot per year also considering the 20-m² plot size. The organic fertilizer was weighed on a digital scale with 1-g precision. Gliricidia plots were fertilized on 18 April 2022 and 13 February 2023, respectively. A top-dressing fertilization was performed, and the dehydrated bovine digesta was directly applied to the soil. We did not record stink issues during the trial, probably because of the rumen fluid evaporation (Edvan & Carneiro, 2011).

Dry matter yield (DMY) evaluations and morphological traits were collected from 18 April 2022 to 9 December 2023, totaling four harvests. Two were performed in rainy seasons, and the others were made in dry seasons. The harvest frequency was 120 days during the rainy and 180 days during the dry seasons. The cutback to attain uniform height prior to the evaluations was made on 18 April 2022, while the first harvest occurred on 16 August 2022, the second one on 12 February 2023, the third was performed on 12 June 2023, and the last was made on 9 December 2023. The regrowth intervals from the first and third harvests were classified as “rainy season”, while the regrowth periods of the second and fourth cuts were the “dry season”. This classification followed the soil water balance observed in the trial (Figure 1). For all evaluated plants, the cut intensity was 50 cm of stubble height. All biomass was reaped from the fodder bank at each harvest time. Plots were 20 m2, and the net plot area comprised four plants centrally positioned on each plot (2.25 m2).

The total DMY, including the DMY in leaves and stems, was estimated based on the dry matter (DM) concentration and the four plants’ weight within the net plot area. A 1-kg aliquot was separated from the cut fodder to be fractionated into leaves and stems, which were dried in a forced-air oven at 55 °C for 72 h. The DM content (934.01) was determined according to the AOAC methods (Horwitz, 2005). In addition, parameters of Shukla’s stability variance for DMY (Shukla, 1972) have been evaluated owing to environmental variations over seasons and years. From the DMY of leaves and stems, we could calculate the leaf/stem ratio by dividing the leaf biomass by the stem biomass. Plant height and canopy diameter were assessed using a 2-m long stick graduate in centimeters.

Leaf samples were collected, identified, and dried in a forced-air oven at 55 °C for 72 hours to analyze the chemical composition. The analyses were performed in the Laboratory of Animal Feed and Nutrition of INSA (LANA-INSA). Concentrations of DM, ashes, crude protein (CP), and ether extract (EE) were analyzed using AOAC methods (A.O.A.C., 2005). Concentrations of neutral detergent fiber (NDF) and acid detergent fiber (ADF) were determined according to (Soest et al., 1991). The NDF was analyzed without thermostable amylase, including residual ashes and proteins. All chemical variables were expressed as percentages. The non-fiber carbohydrates (NFC) were determined according to(Sniffen et al., 1992) using the formula:

Furthermore, the total digestible nutrient content (TDN) was calculated using the ADF concentration and the following formula proposed by Patterson (2000):

NFC = 100 - (NDF + CP + ashes + EE)

Furthermore, the total digestible nutrient content (TDN) was calculated using the ADF concentration and the following formula proposed by (Patterson et al., 2000):

TDN = 88.90 - (7.79 × ADF)

Data were subjected to the normality test of residuals (covtest residual panel) and analysis of variance (ANOVA) using the PROC MIXED of SAS On Demand for Academics (Inc, 2014). The effects of season, bovine digesta level, and the season-bovine digesta interaction were fixed, while the block and evaluation year effects and their interactions were random. Bovine digesta levels within each season of the year were subjected to regression analyses (p<0.05) using the PROC GLM of SAS OnDemand for Academics. The seasons were analyzed as repeated measures in time. The mathematical model was:

Y_ijkl = µ + Xα_i + Xβ_j + Zτ_k + Zδ_l + X(αβ)_ij + Z(ατ)_ik + Z(βτ)_jk + Z(αβτ)_ijk + Z(αδ)_il + Z(βδ)_jl + Z(αβδ)_ijl + Z(τδ)_kl +e_ijkl

Where the Y_ijk is the dependent variable; µ is the overall mean; Xα_i is the fixed effect of bovine digesta level (1 to 6); Xβ_j is the fixed effect of season of the year (1 to 2); Z_τk is the block random effect (1 to 3); X(αβ)_ij is the season-bovine digesta interaction effect; Z(ατ)_ik, Z(βτ)_jk, and Z(αβτ)_ijk are the interaction effects of block with bovine digesta level and season of the year; Z(αδ)_il, Z(βδ)_jl, and Z(αβδ)_ijl are the interaction effects of evaluation year with bovine digesta level and season of the year; Z(τδ)_kl is the interaction between random effects; and e_ijk is the residual error.

When the F-test was significant (p<0.05), means of year’s seasons were compared using the probability of difference (“pdiff”) adjusted by the Tukey test. Means of the bovine digesta levels were compared using linear and quadratic effects of polynomial orthogonal contrasts, in addition to the regression analysis to formulate the equations. All tests were considered significant at 5% of error probability (p<0.05).

Parameters of stability variance (σ_i²) for DMY were calculated considering the bovine digesta levels and the four harvests over the two evaluation years. The model of Shukla’s stability variance (Shukla, 1972) was applied using the PROC MIXED procedure of SAS OnDemand for Academics, according to the statements proposed by (Piepho, 1999). In this analysis, lower variance indicates better stability (Reckling et al., 2021).

There was an interaction effect of season of the year and bovine digesta level on DMY per harvest and year (p<0.05). Considering the seasons within each bovine digesta level, higher DMY values were recorded in the rainy period. The DMY increased linearly regardless of the season of the year (Table 2). No quadratic effects of bovine digesta levels were observed on the gliricidia biomass production (p>0.05).

Furthermore, increasing values of parameters of Shukla’s stability variance were found as the bovine digesta levels enhanced. Thus, gliricidia fertilized with lower bovine digesta levels were more stable for DM production than those treated with higher levels (Table 2).

An interaction effect (bovine digesta level × season of the year) was found in the leaf DMY (p<0.05). A higher leaf biomass was harvested in the rainy than in the dry period within all bovine digesta levels (Table 3). Moreover, the leaf DMY enhanced linearly as the bovine digesta level was increased in both seasons. An interaction effect was also observed in the stem DMY (p<0.05), with positive linear responses to the bovine digesta fertilizer in both seasons too. Stem biomass production was higher in the rainy season than in the dry season in all tested bovine digesta levels.

The interaction between the year’s season and bovine digesta level affected gliricidia’s leaf/stem ratio (p<0.05). Within each bovine digesta level, a significant difference between rainy and dry periods was observed only in the 3.12 t/ha level (Table 4). There were reductions in the leaf/stem ratio as the bovine digesta was increased in both seasons under linear and quadratic effects. In the rainy season, the control treatment led to a leaf/stem ratio of 1.51, while the top fertilization level led to a 1.08 value. In the dry period, the mean decreased from 1.46 (control treatment) to 1.08 (12.50 t/ha level).

Gliricidia’s canopy diameter was affected by the interaction between the year’s season and bovine digesta level (p<0.05). Wider canopies were found in gliricidia trees harvested in the rainy than in the dry period in all tested organic fertilizer levels (Table 5). Canopies were more expansive as the bovine digesta levels increased in the rainy season under a linear effect. In the dry season, only a quadratic effect was found. This time, gliricidia canopies diminished from 0 (control) to 3.12 t/ha level. Then, progressive gains were recorded until the top-recommended dose (12.50 t/ha). Furthermore, an interaction effect was observed in the plant height. Within levels 0.00 (control), 1.25, and 9.37 t/ha of rumen digesta, there was no difference in plant size comparing the rainy to the dry period, contrary to what was observed at 3.12, 6.25, and 12.50 t/ha. Plants grew linearly as the bovine digesta level increased in both seasons (Table 5).

Isolated effects of season and bovine digesta levels were observed in the DM content (p<0.05), but no interaction was found (p>0.05; Table 6). Higher DM content in the leaves was recorded in the rainy than in the dry season, regardless of the tested levels. Linear DM content reductions were found as the bovine digesta level increased (from 27.08 to 24.55% on average).

We could observe an interaction effect of season and bovine digesta levels on ash contents in gliricidia leaves (p<0.05). Results were variable when comparing the seasons within each level of organic fertilizer; that is, they were higher in rainy season for some levels but were higher in the dry season for others (Table 6). Higher ash content in leaves was recorded in the rainy than in the dry season for the control treatment. The opposite was recorded in the other bovine digesta treatments, except for the 6.25 t/ha level, for which we did not find a significant difference. Within the rainy season, no linear or quadratic effect was observed, and no differences between the means were reported. Within the dry season, we found a linear ash content increment from 6.38 to 7.54% of DM. 

Only the season of the year changed the NDF concentration in gliricidia leaves (p<0.05), so the organic fertilizer did not affect this variable (p>0.05; Table 7). A higher NDF content was found in the dry than in the rainy season. Conversely, the year’s season and bovine digesta levels had an interaction effect on the ADF content, but no linear or quadratic effects were found. Higher ADF content was found in rainy than in dry season within 0.00 (control), 3.12 and 6.25 t/ha bovine digesta levels, while no significant differences were observed in the other levels (1.25, 9.37, and 12.50 t/ha). Within the rainy season, the highest leaf ADF content was observed in the 6.25 t/ha level and the lowest in the top level (12.50 t/ha). Within the dry season, the highest ADF content was found in 1.25 and 6.25 t/ha levels, while the lowest value was recorded in the 3.12 t/ha level.

Bovine digesta levels linearly increased the CP content in gliricidia leaves (p<0.05; Table 8). Conversely, no season effect was found. Consistent gains in the leaf CP concentration were observed due to the fertilizer-increasing levels in both rainy and dry seasons.

In addition, isolated effects of season and bovine digesta levels were observed on the NFC concentration in gliricidia leaves (p<0.05; Table 8). NFC content was higher in the rainy than in the dry season, regardless of the fertilizer level. Moreover, the leaf NFC concentration was higher in the 1.25 and 3.12 than in the 6.25 and 9.37 t/ha levels. The lowest NFC content was recorded in the 9.37 t/ha level, regardless of the season evaluated. No linear or quadratic effects were observed on the NFC leaf content in the rainy season, but both effects were found in the dry one. In this case, the NFC decreased from 0.00 (control) to 9.37 t/ha but increased from 9.37 to 12.50 t/ha of bovine digesta.

The bovine digesta levels affected the TDN concentration within both evaluation seasons (p<0.05). An interaction effect was observed, but linear or quadratic effects were not, thus we did not find regression responses. During the rainy season, the 1.25, 9.37, and 12.50 t/ha levels led to higher TDN contents in leaves. An intermediate value was found at the 3.12 t/ha level, and the lowest concentrations were found at 0.00 and 6.25 t/ha levels. Within the dry season, the 3.12 and 12.50 t/ha levels led to the highest TDN, while the 1.25 t/ha treatment led to the lowest TDN content (Table 8).