The experiment was carried out at the Education, Research and Extension Unit (UEPE) of the Federal Institute of Science and Technology of Ceará (IFCE), Limoeiro do Norte Campus, in Chapada do Apodi, state of Ceará. The experimental area has flat relief and it is characterized by the predominance of Cambisol soils (Santos et al. 2006) at geographic coordinates 05°10’53” S and 38°00’43” W and altitude of 146 m. The climate of the region is hot and semi-arid, classified as BSw’h, according to Köppen (1936).
The experiment was conducted in the years 2017, 2018 and 2019. The cumulative rainfall, air relative humidity and average temperature during the evaluation years was 590.4 mm, 68.45% and 27.48°C (2017); 805.75 mm, 74.86% and 27.08°C (2018); and 716.32 mm, 76.56% and 26.96°C (2019), respectively. For climate characterization, data from a meteorological station at approximately 150 m far from the experimental area were used (Fig. 01).
Thinning was carried out in the dry season of 2015, using a tractor with front blade, suppressing 100% vegetation present in the cultivable area, pushing the material to the edges. Before thinning, the area consisted of an arboreal caatinga (Araújo Filho 2013) in secondary succession stage.
The experimental area consisted of nine systems arranged in strips of Caatinga trees (woody component) with dimensions of 6.0 x 100.0 m, in the north-south direction. Enrichment was carried out by planting Opuntia stricta and Nopalea cochenillifera at a spacing of 1.5 m between rows and 0.20 m between plants, forming the understory, between tree strips during the rainy season of 2016. We decided to use these cultivars due to satisfactory adaptation to soil and climate conditions of the study region, described in several works (Barbosa et al. 2017; Pereira et al. 2015; Silva et al. 2015; Pinheiro et al. 2014).
After a phytosociological survey, the following tree species were identified: Cordia goeldiana (Freijó), Mimosa caesalpiniaefolia (Sabiá), Cenostigma pyramidale (Catingueira), Commiphora leptophloeos (Imburana de espinho) and Mimosa tenuiflora (Jurema negra). Freijó showed the highest relative density (81.59%) followed by catingueira (7.87%). The mean height of the woody component was 3.98 ± 1.30 m.
Woody covers corresponding to 46.15, 30.00 and 17.64% were obtained considering 6.0 m strips of Caatinga trees, assigning spacings of 07, 14 and 28 m (understory), respectively (Fig. 2). Densities of 1,418 (46.15%), 925 (30.00%) and 524 (17.64%) trees per hectare were obtained, respectively.
The treatments were: cactus pear grown under 46.15% woody cover (ICS46), cactus pear grown under 30.00% woody cover (ICS30) and cactus pear grown under 17.64% woody cover (ICS18) (Fig. 2). The experimental design was completely randomized, in which the ICS18, ICS30 and ICS46 systems were randomly assigned among the three replications, totaling nine plots. Within each plot, the two palm cultivars Opuntia stricta and Nopalea cochenillifera were randomized, totaling eighteen subplots. In each subplot, the three positions (center, east and west) were considered, totaling fifty-four sub-subplots.
Levels were determined as recommended by Araújo Filho (1992), who established a 30% woody cover in savannah without compromising the growth of the natural herbaceous layer. Thus, it was decided to evaluate woody covers in strips, establishing upper, lower and equivalent levels of 30%.
Before planting the crops, a soil sample was taken from the 0–20 cm layer for further chemical analysis in the soil laboratory of IFCE, Limoeiro do Norte campus. Values of the soil chemical properties for calcium (93.33, mmolc. dm³), magnesium (18.7, mmolc. dm³), potassium (6.32, mg. dm³) and organic matter (31.76, g. kg) were classified as very good, while those of phosphorus (6.03 mg. dm³) were classified as low, according to Ribeiro et al. (1999).
Phosphate fertilization was carried out every year, using the fertilizer single-ammonium-phosphate - MAP (44% P2O5 and 10% N), in the level of 200 kg ha− 1 year− 1, according to the demand for phosphorus found in soil analysis.
The following evaluations were carried out: mean canopy height (H, cm), with a graduated-retractable stick, by sampling six spots per position, number of cladodes per linear meter (NC, cladodes / linear meter), by counting 1st order cladodes within 1.0 linear meter and total forage biomass (TFB, kg ha− 1 year− 1), estimated from the harvest of two plants per linear meter by cutting right after the 1st order cladode. After harvesting, the material was packed in plastic bags and sent to the laboratory, where it was weighed, then samples were chopped into cubes of approximately 0.05 m² to determine the pre-dry matter, in a forced air oven at 65ºC.
Evaluations of H, NC and TFB were made in the end of the rainy season (August). It was considered the positions: center (central region of the plot), east (lateral position with shade in the early morning) and west (lateral position with shade in the early afternoon) of each plot, in order to characterize the effect of the arboreal component in the understory at different spots within the system, because, according to Santos et al. (2016), crops in the understory of integrated systems are subjected to morphological and structural changes, especially in places close to the tree component.
At the end of the growth period of cactus pear Opuntia stricta and Nopalea cochenillifera, a standardization cut was performed, removing all biomass present after the 1st order cladode, with the aim of standardizing the crops for the next growth period (rainy season).
Data normality was tested by the Cramer-von Mises test. Data were analyzed using the mixed model procedure, using the following statistical model:
$${Y}_{ijklm}= \mu + {S}_{j}+ {\alpha }_{ij}+ {C}_{k}+ {\left(SC\right)}_{jk}+ {\beta }_{ijk}+ {P}_{l}+ {\left(SP\right)}_{jl}+ {\left(CP\right)}_{kl}+ {\left(SCP\right)}_{jkl}+ {\epsilon }_{ijkl}$$
Where:
\({Y}_{ijklm}\) : value of the ith experimental unit subjected to the jth shading level, in the kth crop, and lth position;
\(\mu\) : fixed effect of the general mean;
\({S}_{j}\) : fixed effect of the shading level;
\({\alpha }_{ij}\) : random effect of the ith replication in the jth shading level, where \({\alpha }_{ij}\) assumes iid N(0, σA2);
\({C}_{k}\) : fixed effect of the crop;
\({\left(SC\right)}_{jk}\) : fixed effect of the interaction between shading level and crop;
\({\beta }_{ijk}\) : random effect of the kth crop in the ijth shading level, where \({\beta }_{ijk}\) assumes iid N(0, σB2);
\({P}_{l}\) : fixed effect of the position;
\({\left(SP\right)}_{jl}\) : fixed effect of the interaction between shading level and position;
\({\left(CP\right)}_{kl}\) : fixed effect of the interaction between crop and position;
\({\left(SCP\right)}_{jkl}\) : fixed effect of the interaction between shading level, crop and position;
\({\epsilon }_{ijkl}\) : random effect associated to the ith replication, jth shading level, kth crop, and lth position where \({\epsilon }_{ijkl}\) assumes ~ N(0, σ2);
Data were analyzed using the SAS procedure PROC MIXED (SAS Institute. Inc., 2015). The corrected Akaike Information Criterion was used to select the covariance matrix structure for the random effect of year of evaluation. The fixed effects of level of shading, crop, positions and their interactions were considered significant at a p-value < 0.05. When significant differences were detected, mean values were compared by Tukey-Kramer test at a p-value < 0.05.