Introduction

Water is a fundamental resource for the sustainability of agricultural production and global food security, while its growing scarcity represents one of the major challenges of the 21st century [1]. Intensifying demand, driven by population growth, the expansion of agricultural activities, and climate change, has increased pressure on water resources, compromising their availability in various regions of the world. It is estimated that approximately 47% of the global population lives in areas subject to water scarcity for at least one month a year, a scenario likely to intensify given the projected 20% to 30% increase in global water demand by 2050 [2]. In Brazil, although the country holds approximately 12% of the world’s freshwater reserves, marked inequality in spatial distribution results in significant regional disparities. This situation is particularly critical in the Brazilian semiarid region, where limited water availability and high spatial and temporal rainfall variability underscore the need for adaptive strategies capable of enhancing water-use efficiency and promoting the sustainability of agricultural production [3, 4].

Semi-arid regions, concentrated primarily in Africa, Asia, and Latin America, cover approximately 15% of the Earth’s land surface and support about 15% of the global population. Characterized by limited rainfall and high evaporative demand, these regions are particularly vulnerable to climate variability, with direct implications for water availability, agricultural production, and livelihoods [5, 6]. Within this context, the Brazilian semi-arid region is one of the most densely populated semi-arid areas in the world and exhibits high spatiotemporal rainfall irregularity, with average annual precipitation generally below 800 mm, a condition that restricts water availability and imposes significant limitations on agricultural production and regional socioeconomic development [7, 8].

Aquaculture has established itself as a key driver of the expansion in global aquatic resource production, reaching 235 million tonnes in 2024, with over 100 million tonnes consisting of farmed aquatic animals. In addition to contributing approximately 15% of the global supply of animal protein, the sector holds significant economic and social importance; roughly 36% of aquatic animal production enters international trade, generating around US$ 186 billion, while more than 65 million people work directly in the primary sector and over 600 million rely on these activities for their livelihoods [9]. In Brazil, although aquaculture is also practiced in estuarine zones, its recent expansion has been particularly notable in the rural areas of the Northeast, driven by the availability of reservoirs and climatic conditions favorable to farming high-value species [10]. In 2024, tilapia (Oreochromis niloticus) and whiteleg shrimp (Penaeus vannamei) stood out as the country’s leading aquaculture products, with production volumes of 499,364 and 146,832 tonnes, respectively. The Northeast region accounted for 99.75% of national shrimp production, led by the states of Ceará and Rio Grande do Norte, and 17.05% of Brazilian tilapia production, underscoring the activity’s strategic importance for income generation and economic revitalization within the region’s rural communities [10, 11].

Thus, the expansion of aquatic animal aquaculture entails significant environmental challenges, primarily due to the large volumes of effluent generated. Intensive tilapia farming, for instance, releases 100 to 200 L of effluent per kilogram of fish produced, whereas shrimp farming generates 500 to 1,000 L per kilogram. These nutrient-rich effluents, containing nitrogen and phosphorus derived from feed residues and metabolic waste, can contaminate and cause eutrophication in aquatic ecosystems if discharged without treatment [12].

On the other hand, phosphorus and nitrogen are essential nutrients for crop growth and production, supplied to the soil primarily through mineral and/or organic fertilizers. It is estimated that phosphate rock reserves, the main raw material used to manufacture phosphate fertilizers, could be depleted within the next 300 years [13, 14]. In contrast, nitrogen fertilizer production is not constrained by the availability of mineral reserves, as it relies on atmospheric N₂ and hydrogen derived from natural gas via the Haber–Bosch process. However, this process entails a high demand for fossil resources; producing 1 tonne of anhydrous NH₃ consumes approximately 949 m³ of natural gas and results in the emission of about 1.6 tonnes of CO₂, highlighting the significant carbon footprint associated with conventional nitrogen fertilizer manufacturing [15].

In this context, utilizing the water and nutrients found in aquaculture effluents for fertigation offers multiple benefits, including reduced contamination and eutrophication of water bodies, improved resource efficiency amidst water scarcity, and the provision of nitrogen with a lower carbon footprint, as well as phosphorus. without direct reliance on mineral reserves. Beyond alleviating pressure on conventional water sources, this practice supplies essential nutrients to crops and reduces dependence on synthetic fertilizers [16]. In Brazil’s semi-arid regions, where water scarcity is particularly acute, integrating aquaculture and agriculture can foster more resilient production systems [17].

Consequently, climate change, coupled with intensifying droughts and desertification, has heightened the need for crops capable of maintaining production under high temperatures, low water availability, and low-fertility soils. In this scenario, forage cactus stands out due to its high adaptability to arid and semi-arid environments, potentially contributing to the sustainability of agricultural systems and the restoration of degraded areas under adverse climatic conditions [18]. Among cultivated species, Opuntia cochenillifera (Nopalea cochenillifera) demonstrates high water-use efficiency and broad adaptability to low-fertility soils, characteristics that underscore its economic importance in semi-arid regions. Traditionally used as animal feed, especially during droughts, this crop also has applications in the bioenergy, human food, and pharmaceutical sectors [19]. In Northeastern Brazil, where forage cactus occupies approximately 500,000 hectares and plays a strategic role in sustaining livestock farming, the use of alternative water sources has enhanced its productive potential, helping the region achieve some of the world’s highest levels of fresh and dry biomass production for the crop [18].

Recent studies demonstrate that fertigation with effluents can be a promising strategy for cultivating forage cactus in the Brazilian semi-arid region, if application is properly managed to balance agronomic performance and soil quality. In O. cochenillifera (N. cochenillifera) the use of aquaculture effluent influences morphometric and productive characteristics, such as plant height and width; cladode number, dimensions, and thickness; occupied area; phytomass volume; and biomass production, with the substitution of up to 25% of the irrigation water with aquaculture effluent maintaining productivity levels comparable to the control condition [17]. Concurrently, this effluent proportion proved favorable for soil functionality, affecting physical-hydraulic and chemical attributes, including microporosity, total organic carbon, and cation availability, while other proportions influenced flocculation, macroporosity, and aggregate stability [20]. From a chemical perspective, however, increasing the effluent share in irrigation water intensifies changes related to electrical conductivity and the concentrations of bases, sodium, and micronutrients; this indicates that higher doses require stricter control to prevent cumulative adverse effects on soil quality. In this regard, the 25% effluent proportion also resulted in the least chemical alteration after prolonged cultivation [21]. Complementary evidence obtained using treated produced water from oil extraction reinforces O. cochenillifera (N. cochenillifera) ability to thrive in reuse systems and to extract and accumulate metals, thereby expanding its potential for production systems adapted to the environmental conditions of the semi-arid region [22].

In this context, valuing the water and nutrients present in aquaculture effluents represents a promising strategy for integrating the aquaculture sector into circular economy principles by transforming waste into agricultural and environmental assets. This approach also aligns directly with several Sustainable Development Goals (SDGs) of the 2030 Agenda: SDG 2 (Zero Hunger and Sustainable Agriculture), by fostering agricultural systems that are more resilient and capable of sustaining forage production under water-scarce conditions; SDG 6 (Clean Water and Sanitation), by expanding water reuse and reducing effluent discharge into the environment; SDG 12 (Responsible Consumption and Production), by promoting nutrient recovery and recirculation; SDG 13 (Climate Action), by reducing reliance on fertilizers associated with higher carbon emissions; SDG 15 (Life on Land), by supporting practices compatible with sustainable soil management and the reduction of environmental degradation; and SDG 8 (Decent Work and Economic Growth), by contributing to production diversification, income generation, and the strengthening of economic activities in rural communities [23, 24].

Despite growing interest in fertigation with wastewater as a strategy to enhance the sustainability of forage cactus production in semi-arid regions, available evidence focuses primarily on the species Opuntia ficus-indica, whereas studies specifically addressing Opuntia cochenillifera (Nopalea cochenillifera) are comparatively scarce. This knowledge gap is particularly pronounced regarding the use of aquaculture effluent and the conduct of long-term experiments capable of capturing cumulative and seasonal effects on the soil-plant system across successive cultivation cycles. There is limited information on how the continuous application of aquaculture effluent triggers potential toxicity responses that could ultimately alter the morphometric performance and productivity of Opuntia cochenillifera (Nopalea cochenillifera). Therefore, we hypothesized that the forage cactus’s response to aquaculture effluent depends on the application ratio and that a mixture containing 25% aquaculture effluent and 75% fresh water would provide the most favorable balance between nutrient supply and potential ionic constraints, thereby promoting superior morphometric and productive performance, even under continuous fertigation. To test this hypothesis, the present study evaluated the morphometric and productive responses of Opuntia cochenillifera (Nopalea cochenillifera) over a 26-month period spanning successive cultivation cycles.

Materials and Methods

Study Area and Experimental Design

The study was conducted at the Water Reuse Experimental Unit (WREU) of the Universidade Federal Rural do Semi-Árido (UFERSA), located in Mossoró, Rio Grande do Norte, Brazil (UTM coordinates 686,313.70 m E and 9,424,039.88 m N, Zone 24 S, referenced to the SIRGAS 2000 datum; altitude 67 m). Figure 1 presents the cartographic representation of the Water Reuse Experimental Unit (UERA), the UFERSA Aquaculture Sector, and the forage cactus cultivation area. The map was created using the QGIS Geographic Information System, version 3.40.4 [25], employing the Universal Transverse Mercator (UTM) coordinate system, SIRGAS 2000 datum, Zone 24 S (EPSG: 31984). The cartographic base was generated from vector geospatial data in shapefile (.shp) format provided by the Instituto Brasileiro de Geografia e Estatística [26]. Google Earth Pro, version 7.x [27], was used solely as an auxiliary spatial reference for the preliminary delineation of the experimental area and the generation of KML vector files, which were subsequently converted to shapefile format in QGIS.

Fig. 1
Fig. 1
Full size image

Location of the aquaculture sector and water reuse experimental unit at the Federal Rural University of the Semi-Arid Region (UFERSA), Mossoró, Rio Grande do Norte, Brazil

The study comprised two consecutive experimental cycles (A1 and A2). The first cycle (A1), conducted from October 20, 2021, to October 20, 2022, generated data previously reported in Pedrosa’s thesis [28] and subsequently published in [17]. The second cycle (A2), conducted from October 20, 2022, to December 20, 2023, generated the original dataset analyzed in the present study, allowing assessment of the spatiotemporal persistence and cumulative effects of the treatments.

The experimental area is located in a hot semi-arid climate region, classified as BSh under the Köppen-Geiger system [29]. Mossoró - RN has an average elevation of 55 m, an average annual rainfall of 794 mm (with rainfall concentrated between February and April), and an average annual temperature of 26.5 °C. The relative humidity is approximately 68.9%, characterizing an environment with high evapotranspiration and irregular rainfall typical of the Brazilian semi-arid region [30].

The region experiences maximum temperatures that can exceed 35 °C during the hottest months, while minimum temperatures rarely drop below 22 °C, reinforcing the predominance of a hot and dry climate. Annual potential evaporation reaches values above 2,000 mm [31]. The soil at the experimental site was classified as an Argissolo Vermelho-Amarelo Eutrófico, according to the Brazilian Soil Classification System, published by the Empresa Brasileira de Pesquisa Agropecuária (Embrapa) [32], with its corresponding classification under the United States Department of Agriculture Soil Taxonomy system established according to Soil Survey Staff [33].

For the experiment, a total area of approximately 62.36 m² (7.55 m wide × 8.26 m long) was demarcated, where 25 experimental plots of 1 m² each were distributed, spaced 0.80 m apart (Fig. 2). The area was protected by an external border, with a 0.80 m distance from the plots, and internal walkways 0.70 m wide for circulation [17, 21].

A randomized block design was adopted, in a split-plot scheme over time. Five proportions of aquaculture effluent (EA) and supply water (AA) were distributed in the plots, aiming at reducing salinity levels, according to technical protocols for the reuse of effluents in agriculture [34]. The treatments consisted of: 0 A (100% AA – control), 25 A (75% AA + 25% EA), 50 A (50% AA + 50% EA), 75 A (25% AA + 75% EA) and 100 A (100% EA). In the subplots, the two annual cultivation cycles (A1 and A2) were considered, corresponding to the morphometric and productive evaluations of the forage palm. The spatial distribution of treatments in the field was randomized within each block, as illustrated in Fig. 2.

Fig. 2
Fig. 2
Full size image

Schematic representation of the experimental layout and irrigation system. The diagram shows: (1) a 1.0 m³ supply-water reservoir; (2) a 0.5-hp motor pump; (3) a 130-mesh screen filter; (4) a pressure gauge (0–400 kPa); (5) the main line; (6) branch lines; (7) lateral lines equipped with non-pressure-compensating emitters (1.6 L h− 1); (8) a 1-m² experimental plot; (9) the border area surrounding the 25 experimental plots; (10) the aquaculture-effluent reservoir; (11) a 13-L watering can used to apply the different proportions of aquaculture effluent; (B) Randomized blocks, and (0A2, 25A2, 50A2, 75A2, and 100A2) aquaculture effluent proportions used for fertirrigation

Preparation and Planting of Plant Material

Cladodes of Opuntia cochenillifera (L.), a synonym of Nopalea cochenillifera (L.) Salm-Dyck commonly adopted in Brazilian scientific literature, were selected from mother plants cultivated under controlled conditions at the Empresa de Pesquisa Agropecuária do Rio Grande do Norte (EMPARN), located in Apodi, Rio Grande do Norte, Brazil (UTM: 631,203.20 m E, 9,377,889.77 m N, Zone 24 S, WGS 84; 130 m above sea level). Planting was carried out on October 20, 2021, using five cladodes per plot. The cladodes were uniformly established within the central area of each experimental plot at 20-cm spacing and oriented along the east–west axis to enhance interception of photosynthetically active radiation, following the procedure described by [17].

The establishment protocol followed technical criteria to ensure uniform plant growth. The cladodes, previously cut and subjected to a 13-day healing period in a shaded environment, were buried halfway into the soil with the basal end facing downward. The adopted spatial arrangement (1.7 m between rows and 0.2 m between plants) resulted in a population density of 29,412 plants per hectare, optimizing both light interception and water-use efficiency, as described by [17].

At 22 months after planting, supplemental fertilization was performed by uniformly applying 20 t ha− 1 of stabilized cattle manure to all experimental plots, including the control, to supplement nutrient supply and standardize organic fertilization across treatments. The manure was obtained from the Cattle Production Sector of UFERSA and incorporated into furrows between plant rows at a depth of 15 cm, limiting direct contact with the plant bases. A composite sample of the stabilized manure was collected from this unit by combining five individual samples taken from the composting pile. This material was characterized in triplicate at the Soil, Water, and Plant The manure was characterized at LASAP/UFERSA following methodologies recommended by EMBRAPA (Teixeira et al., 2017). Total N (15.46 ± 0.30 g kg− 1) was determined via the Kjeldahl method following high-temperature sulfuric acid digestion in the presence of catalysts. Available P (1.34 ± 0.10 g kg− 1), K⁺ (2.29 ± 0.25 g kg− 1), and Na⁺ (0.030 ± 0.005 g kg− 1) were extracted using the Mehlich-1 solution. Phosphorus was quantified spectrophotometrically using ammonium molybdate and ascorbic acid, while K⁺ and Na⁺ were measured by flame photometry. Calcium (6.76 ± 0.39 g kg− 1) and Mg²⁺ (2.10 ± 0.13 g kg− 1) were extracted with KCl. The sum of Ca²⁺ and Mg²⁺ and Ca²⁺ alone were determined by EDTA titration using Eriochrome Black T and calcon indicators, respectively, with the Mg²⁺ content calculated by difference. Finally, Fe (48.33 ± 29.45 mg kg− 1), Mn (65.00 ± 4.71 mg kg− 1), Cu (6.67 ± 0.47 mg kg− 1), and Zn (11.67 ± 2.87 mg kg− 1) were extracted with Mehlich-1 and quantified by inductively coupled plasma optical emission spectrometry (ICP-OES). The cattle manure was obtained from heifers belonging to the UFERSA herd, which were kept in an intensive rearing system and fed a concentrate based on corn, soybean, and wheat bran, with Echinochloa polystachya used as the roughage source, as described by [35].

This input, sourced from the UFERSA Cattle Farming Sector, was carefully incorporated into furrows located 20 cm from the base of the plants and at a depth of 15 cm, following the recommendations of [36] for crops under semi-arid conditions. This strategy aimed to optimize nutrient availability while minimizing losses through volatilization and avoiding direct contact with plant tissues, as recommended by [37].

At 12 months of cultivation, two plants located at the edges of each plot were removed, as described by [17], leaving three central plants per plot for the morphometric and productive evaluations conducted in this study. It should be noted that, in the second cycle, only the original plants established on October 20, 2021, were evaluated, without the use of regrowth or the replanting of new cladodes, which ensured uniformity in plant age and allowed for the assessment of the cumulative effects of the treatments throughout the experimental period.

Effluent Application Protocol and Irrigation Water Quality Monitoring

Irrigation management followed a weekly alternating regime, with supply water applied in the first week and the corresponding effluent proportions in the subsequent week. Irrigation was suspended whenever rainfall exceeded 20 mm in the 24 h preceding the scheduled application. Effluent was collected from the Aquaculture Sector of UFERSA (686251.49 m E, 9424273.81 m N; SIRGAS 2000/UTM zone 24 S, EPSG: 31984), where experimental ponds operated under freshwater, brackish, and saline conditions. Cultured organisms included pirá (Conorhynchos conirostris), tambaqui (Colossoma macropomum), tambatinga (Colossoma macropomum × Piaractus brachypomus), Nile tilapia (Oreochromis niloticus), Zanzibar tilapia (Oreochromis urolepis hornorum), and shrimp (Litopenaeus vannamei), fed diets tailored to each species and developmental stage. No medications were administered, whereas lime, sodium hydroxide, and chlorine were used for tank disinfection. Pond water originated from a well on the East Campus, with salinity increased by adding sodium chloride or diluted mother liquor from saltworks in Mossoró. The well draws from the Jandaíra aquifer, an unconfined, heterogeneous, and anisotropic system with karstic groundwater flow. This aquifer exhibits variable karstification and water quality and receives recharge from rainfall and through the Barreiras Formation, which acts as a transfer aquifer [17, 21, 38]. While the supply water was provided by the Companhia de Águas e Esgotos do Rio Grande do Norte (CAERN), sourced from a mix of deep groundwater wells in Mossoró, Rio Grande do Norte State, and surface water from the Armando Ribeiro Gonçalves Dam in Açu.

For irrigation using supply water or aquaculture effluent in various proportions, the following water balance was employed:

$$\:\text{E}{\text{T}}_{\text{c}}-{\text{E}}_{\text{p}}-\text{E}\text{C}\pm\:{\Delta\:}\text{S}$$
(1)

Where: ETc - Crop evapotranspiration; \(\:{\text{E}}_{\text{p}}\) - Effective precipitation, EC - Eventual contribuição por ascensão capilar; e \(\:{\Delta\:}\text{S}\) - Change in soil water storage.

The irrigation depth was calculated based on crop evapotranspiration (ETc), determined using the FAO-56 method [39], with daily meteorological data from the Campbell automatic weather station installed on UFERSA’s East Campus, located approximately 933 m from the experimental site. The Penman-Monteith equation was applied to compute reference evapotranspiration-ET₀ (Eq. 2).

$$\:{\text{E}\text{T}}_{0}=\frac{\text{0,408}\times\:\varDelta\:\times\:\left({\text{R}}_{\text{n}}-\text{G}\right)+{\upgamma\:}\times\:\frac{900}{\text{T}+273}\times\:{\text{U}}_{2}\times\:({\text{e}}_{\text{s}}-{\text{e}}_{\text{a}})}{\varDelta\:+{\upgamma\:}\times\:(1+\text{0,34}\times\:{\text{U}}_{2})}$$
(2)

Where: ET₀ - Reference evapotranspiration (mm d− 1); Rₙ - Net radiation at the surface (MJ m− 2 d− 1); G - Soil heat flux (MJ m− 2 d− 1); T - Air temperature at 2 m height (°C); U₂ - Wind speed at 2 m height (m s− 1); eₛ - Saturation vapor pressure (kPa); eₐ - Actual vapor pressure of the air (kPa); (eₛ - eₐ) - Vapor pressure deficit (kPa); Δ - Slope of the saturation vapor pressure curve (kPa °C− 1); and γ - Psychrometric constant (kPa °C− 1).

Figure 3 illustrates the monthly temporal variation in the monitored climatic parameters, mean air temperature, accumulated solar radiation, accumulated rainfall, and mean reference evapotranspiration, over the 26-month experimental period spanning successive cultivation cycles.

Fig. 3
Fig. 3
Full size image

Monthly variation in the monitored climatic parameters: mean air temperature, accumulated solar radiation, accumulated rainfall, and mean reference evapotranspiration (ET0)

A fixed crop coefficient (Kc) of 0.70 was adopted, as established by [40] for forage cactus under semi-arid conditions. The gross irrigation depth was adjusted using a simplified water balance, considering effective rainfall and irrigation as inputs and ETc as the output (Eq. 3).

$$\:{\text{E}\text{T}}_{\text{c}}\:=\:{\text{E}\text{T}}_{0\:}\times\:\:{\text{K}}_{\text{c}}$$
(3)

Where: ETc - Crop evapotranspiration (mm d− 1); and Kc - Crop coefficient (dimensionless).

Irrigation application efficiency (Ea) was estimated based on the distribution uniformity coefficient (DU), according to [41], accounting for a 10% margin for operational water losses, as shown in Eqs. 4 and 5.

$$\:\text{C}\text{U}\text{D}\:=\:100\times\:\:\left(\frac{{\text{q}}_{25}}{\stackrel{-}{\text{q}}}\right)$$
(4)
$$\:{\text{E}}_{\text{a}}\:=\:0.90\times\:\:\left(\frac{\text{C}\text{U}\text{D}}{100}\right)$$
(5)

Where: \(\:{\text{E}}_{\text{a}}\) - Irrigation application efficiency (decimal); DU - Water application uniformity (%); \(\:{\text{q}}_{25}\) - mean discharge of the 25% of emitters with the lowest discharge rates (L h− 1); and \(\:\stackrel{-}{\text{q}}\) - Mean discharge of all evaluated emitters (L h− 1).

DU and \(\:{\text{E}}_{\text{a}}\) remained consistently high throughout the evaluation period. DU and Ea values were 91.17% and 0.82 on 21 October 2022, 96.70% and 0.87 on 23 December 2022, 94.96% and 0.85 on 10 March 2023, 93.33% and 0.84 on 13 May 2023, 96.72% and 0.87 on 3 August 2023, 97.37% and 0.88 on 22 September 2023, and 97.15% and 0.87 on 7 November 2023, respectively.

The gross irrigation depth (GID), estimated using Eq. (6), was used to determine the volumes of supply water (AA) and aquaculture effluent (EA) for each experimental plot; application was performed manually using a 13 L watering can, with volumes precisely measured using a 1 L graduated cylinder. To this end, GDI values expressed in millimeters were converted to volumetric units, based on the equivalence of 1 mm to 1 L m− 2, and subsequently multiplied by the experimental plot area (1 m2) to obtain the total application volume in liters per plot. This volume was then divided according to the five treatments: 100% AA, 75% AA + 25% EA, 50% AA + 50% EA, 25% AA + 75% EA, and 100% EA.

$$\:\text{G}\text{I}\text{D}\:=\frac{\text{E}{\text{T}}_{\text{c}}\times\:{\text{K}}_{\text{s}}\times\:{\text{K}}_{\text{l}}}{{\text{E}}_{\text{a}}}\:$$
(6)

Where: GID - gross irrigation depth (mm); Ks – Soil moisture coefficient (decimal); Kl – Irrigation location coefficient (decimal); and \(\:{\text{E}}_{\text{a}}\) - Irrigation application efficiency (decimal).

Regarding the use of the drip irrigation system exclusively for the application of supply water, the gross irrigation depth (GID) was converted into irrigation time (IT) using Eq. (7).

$$\:\text{I}\text{T}\:=60\times\:\left(\frac{\text{G}\text{I}\text{D}\times\:{\text{S}}_{\text{l}}\times\:{\text{S}}_{\text{d}}}{\stackrel{-}{\text{q}}}\right)\:$$
(7)

Where: IT - irrigation time (minutes); GID - Gross irrigation depth (mm); Sl – Spacing between drip lines (meters); Sd – Spacing between drippers along the lines (meters); and \(\:\stackrel{-}{\text{q}}\) - Mean discharge of all evaluated emitters (L h− 1).

To mitigate the risk of soil salinization and sodification, applications of supply water via the drip irrigation system were alternated with manual applications of varying proportions of aquaculture effluent. This management strategy likely promoted the periodic dilution and leaching of soluble salts within the root zone. Furthermore, the accumulated rainfall of 1,589 mm recorded over the 26-month experimental period may have further contributed to salt leaching, thereby helping to limit the accumulation of salts and exchangeable sodium in the soil.

Analysis of the experimental water balance indicated an accumulated rainfall of 1,589 mm over the two forage cactus cultivation cycles (A1 and A2). Of this total, 944 mm were recorded during the first cycle (A1) and 645 mm during the second cycle (A2), based on data from the UFERSA weather station. The supplemental water depth applied throughout the experiment totaled 449 mm, comprising 178 mm in the first cycle and 271 mm in the second. Supplemental irrigation was carried out exclusively using supply water via a drip irrigation system, whereas the different proportions of aquaculture effluent were applied manually according to the established treatments. Details regarding the irrigation depths and manual effluent applications for the second cultivation cycle (A2) are presented in Table 1.

Table 1 Details of irrigation using only supply water (SW) and manual applications of aquaculture effluent (AE) proportions
Full size table

Every two months, samples were collected from the five aquaculture effluent dilutions in the supply water to characterize their physico-chemical attributes that could potentially influence the production and development of spineless forage cactus. Analyses were performed at LASAP/UFERSA following the procedures described by EMBRAPA [42] (Teixeira et al. 2017) and Standard Methods for the Examination of Water and Wastewater [43]. The pH and electrical conductivity (EC) were measured directly using a pH meter and a conductivity meter, respectively. Iron (Fe), manganese (Mn), zinc (Zn), and copper (Cu) were determined in filtered samples by atomic absorption spectrometry, while potassium (K⁺) and sodium (Na⁺) were measured by flame photometry. Calcium (Ca²⁺) and magnesium (Mg²⁺) were quantified by EDTA complexometric titration using calcon and Eriochrome Black T indicators. Chloride (Cl−) was determined by titration with silver nitrate using potassium chromate as the indicator. Carbonate (CO₃²⁻) and bicarbonate (HCO₃⁻) were quantified by sulfuric acid titration using phenolphthalein and methyl orange, respectively. Kjeldahl nitrogen (N) was determined by steam distillation followed by sulfuric acid titration, whereas dissolved reactive phosphorus (P) was measured spectrophotometrically through molybdate reduction with ascorbic acid. The results are presented in Table 2.

Table 2 Arithmetic mean and sample standard deviation of the physicochemical variables of the aquaculture effluent monitored bimonthly during the second year of culture
Full size table

Characterization of Physical and Chemical Soil Attributes

Soil sampling was conducted three distinct times: prior to the application of treatments in October 2021; during the first annual assessment in October 2022 (with chemical parameters published by [17]; and during the final assessment in December 2023 (with consolidated results presented in this study). Data obtained by [17], particularly those regarding the initial soil assessment were used to support the interpretation of the results in the present work.

For the final assessment conducted in December 2023, soil samples were collected from each of the 25 experimental plots. Twenty-four hours prior to collection, the soil was pre-moistened using a 13 L watering that can facilitate sampling at the two established depths. In each plot, five individual samples were collected at distinct points within the 0–20 cm and 20–40 cm layers using a Dutch auger with a capacity of approximately 350 cm3 of soil. The five subsamples from the same depth were subsequently homogenized in a 10 L container, yielding a representative composite sample of approximately 500 g for each soil layer. This procedure was adopted to account for small-scale spatial heterogeneity within each experimental unit while simultaneously ensuring greater representativeness of the samples intended for chemical characterization. Subsequently, the composite samples were placed in 500 g aluminum containers, duly labeled with the sampling depth, applied treatment, and corresponding experimental block, and transported to the Semiarid Soil, Water, and Plant Analysis Laboratory (LASAPSA) at the UFERSA. In the laboratory, the samples were initially air-dried for 48 h and then oven-dried with forced air circulation at 60 °C for another 48 h to remove residual moisture. After air-drying, the soil samples were transferred to the preparation and storage room of the Multi-user Laboratory for Soil, Water, and Plant Analysis of the Semiarid Region (LASAPSA/UFERSA), where they were disaggregated and passed through a 2 mm sieve to obtain the air-dried fine earth fraction. This material was stored in 500 mL cylindrical polypropylene containers, labeled with the treatment, experimental block, and sampling depth. For both sampling depths, particle-size and chemical analyses were performed on the prepared fine earth fraction, while soil bulk density was determined using undisturbed samples (cylinders). Analytical procedures followed Embrapa standards [42], with an adaptation for organic carbon determination. Particle-size distribution was determined using the pipette method, following dispersion with sodium hexametaphosphate in distilled water and mechanical agitation on a Wagner-type shaker. Sand was separated by wet sieving through a 0.053 mm mesh, while clay (< 0.002 mm) was quantified via pipette sampling after sedimentation. Silt content was calculated as the remaining fraction after subtracting the sand and clay contents from the total fine earth fraction. Soil bulk density (BD) was calculated as the ratio of oven-dry soil mass to the internal volume of the sampling ring, expressed in g cm− 3. Soil pH and electrical conductivity (EC) were measured at a 1:2.5 soil-to-water ratio using a pH meter and a conductivity meter, respectively. Soil organic carbon (SOC) was determined using a wet oxidation procedure adapted from [44]. Samples were digested with potassium dichromate and sulfuric acid, and the residual dichromate was subsequently titrated with ferrous sulfate, using ferroin as an indicator. Organic matter (OM) content was estimated by multiplying SOC by 1.724. Exchangeable Ca²⁺ and Mg²⁺ were extracted with KCl and quantified by complexometric titration with EDTA. The sum of Ca²⁺ and Mg²⁺ was determined using the Eriochrome Black T indicator, while Ca²⁺ alone was determined using the calcon indicator; Mg²⁺ content was calculated by difference. Available P and extractable K⁺ and Na⁺ were determined in Mehlich-1 extracts. Phosphorus was quantified by spectrophotometry following color development with ammonium molybdate and ascorbic acid, while K⁺ and Na⁺ were determined by flame photometry. Potential acidity (H + Al) was determined by extraction with calcium acetate, followed by titration with NaOH using phenolphthalein as an indicator. Cation exchange capacity at pH 7.0 (CEC) was then calculated by adding potential acidity to the sum of exchangeable bases (Ca²⁺, Mg²⁺, K⁺, and Na⁺). The exchangeable sodium percentage (ESP) was calculated as 100 × exchangeable Na⁺/CEC, with Na⁺ and CEC expressed in the same charge-based units [42, 45]. Table 3 summarizes the physical and chemical soil properties before and after fertigation of spineless forage cactus with different proportions of aquaculture effluent.

Table 3 Changes in soil physicochemical properties before and after fertigation of spineless forage cactus with aquaculture effluent dilutions
Full size table

Processing, Growth Measurement, and Yield estimation of Forage Cactus

The morphogenic assessment of forage cactus was conducted after 26 months of cultivation. Growth variables included plant height (PH) and plant width (PW), measured using a measuring tape, as well as the number of cladodes (NC), determined by manual counting. Cladode length (CL), width (CD), and perimeter (CP) were also measured using a measuring tape, whereas cladode thickness (CT) was determined with a calibrated digital caliper (0.01 mm precision; MTX, model 316119; measurement range: 0–150 mm). These measurements allowed for the analysis of vegetative development in response to aquaculture effluent treatments over two production cycles.

For the morphometric and productivity analyses, two representative plants were selected from the usable area of each plot, based on established development, intact architecture, regular cladode emergence, and the absence of visible symptoms of pests, diseases, mechanical damage, or atypical development. The experiment was dismantled on December 20, 2023, following a standardized protocol for final forage cactus evaluation. Initially, secondary cladodes were selectively removed while preserving the basal cladode due to its connection with the root system. The final step involved separating the basal cladode along with its associated root system using a sterilized blade to ensure precision and prevent contamination.

Cladode measurements were used to estimate variables describing the growth and productivity of forage cactus. Cladode area (CA) was estimated to use an elliptical approximation based on cladode length (CL) and width (CD) (Eq. 8). The mean CA per plant was then multiplied by the total number of cladodes (NC) to obtain the total cladode area per plant (PA; Eq. 9). Following the geometric approach described by [46], the fresh shoot phytomass volume (FV) was estimated by multiplying PA by the mean cladode thickness (CT). The resulting volume per plant was extrapolated to a per-hectare basis using the planting density (ni ≈ 29,412 plants ha⁻¹) (Eq. 10). Finally, fresh shoot biomass productivity (FW) was calculated by multiplying the mean fresh shoot biomass per plant (WFMS) by the planting density (Eq. 11).

$$\:\text{C}\text{A}=\frac{{\uppi\:}\times\:\text{C}\text{D}\times\:\text{C}\text{L}\:}{4}$$
(8)
$$\:\text{P}\text{A}=\frac{\text{C}\text{A}\:\times\:\:\:\text{N}\text{C}}{\text{10,000}}$$
(9)
$$\:\text{F}\text{V}\:=\:\text{C}\text{A}\times\:\text{N}\text{C}\times\:\text{C}\text{T}\times\:\text{n}\text{i}$$
(10)
$$\:\text{F}\text{W}\:=\:\text{W}\text{F}\text{M}\text{S}\:\times\:\:\text{n}\text{i}$$
(11)

Where: CD - Cladode width (m); CL - Cladode length (m); NC: Number of cladodes per plant; CT - Cladode thickness (m); CA - Individual cladode area (cm² plant− 1); and PA - Total photosynthetic area per plant (m²); FV - Fresh phytomass volume (m³ ha− 1); FW - Fresh weight (t ha− 1); WFMS - Fresh weight per plant (g); and ni - Plant density (29,412 plants ha− 1).

Statistical Analysis

The experiment was analyzed using a split-plot design over time, within a randomized complete block design with five blocks. The five proportions of aquaculture effluent (0 A, 25 A, 50 A, 75 A, and 100 A) served as the main-plot factor, while the two cultivation cycles (A1 and A2) served as the sub-plot factor, as both cycles were evaluated on the same plots. The two plants assessed per plot were treated as subsamples and summarized by their mean; thus, the plot served as the experimental unit (n = 50 observations: plot × cycle).

Univariate

Each variable was analyzed using a linear mixed model, with block, effluent proportion, cycle, and the proportion × cycle interaction as fixed effects, and the experimental plot as a random effect, accounting for the dependency between the two cycles evaluated within the same plot. F-tests were performed using Kenward–Roger degrees of freedom [47, 48]. The proportion factor was tested against the between-plot error (error a; block × proportion; 4 and 16 degrees of freedom). The cycle factor and the interaction were tested against the within-plot error (error b; 1 and 20 degrees of freedom, and 4 and 20 degrees of freedom, respectively).

Residual normality was assessed using the Shapiro–Wilk test, supplemented by an examination of skewness, kurtosis, and residual histograms. Homogeneity of variances across the ten proportion × cycle groups were evaluated using Levene’s test (median-centered). Variables that met both assumptions (P ≥ 0.05) were analyzed via the F-test from the mixed model. For these variables, each proportion was compared against the control (0 A) within each cycle using Dunnett’s test. The response trend to the proportions was evaluated using linear and quadratic orthogonal polynomial contrasts; regression equations were fitted only if a given contrast proved significant. When the proportion × cycle interaction was significant, proportions were compared within each cycle, and cycles were compared within each proportion, using Tukey’s test.

The variable that failed to meet the normality assumption (FV) was analyzed using the Aligned Rank Transform ANOVA (ART-ANOVA), following the same model structure [49], with subsequent comparisons performed via ART-C contrasts [50]. Results are presented as estimated marginal means ± standard error [51]. Differences relative to the control are expressed as percentages, with a 95% confidence interval. A significance level of 5% was adopted, and analyses were performed in R 4.3.3 [52].

Multivariate

Multivariate analyses were performed using plot means for each cycle (mean of two plants; n = 50), with no missing data. Plant-occupied area (AOP) and fresh mass per hectare (FW) were excluded, as they are constant multiples of PA and WFMS, respectively (r = 1.000).

The relationship between pairs of variables was assessed using Pearson’s correlation coefficient (r). With n = 50, |r| values ≥ 0.28 are significant at the 5% level, and |r| values ≥ 0.36 are significant at the 1% level. Since the two cycles from the same plot are not independent, these P-values were used for descriptive purposes.

Principal component analysis (PCA) was conducted on the correlation matrix of the eight directly measured variables: PH, PW, NC, CT, CL, CD, CP, and WFMS (a matrix of 50 observations × 8 variables). Each variable was standardized to a mean of zero and a variance of one, as the variables are expressed in different units. Variables calculated from these measurements (CA, PA, and FV) were included as supplementary variables; their correlations with the components were calculated after the extraction process, without them influencing the components themselves.

Two criteria were used to assess the suitability of the data for PCA: (i) The Kaiser–Meyer–Olkin (KMO) index [53], which ranges from 0 to 1 and indicates the proportion of variance shared among the variables; data were considered suitable if the overall KMO was ≥ 0.60 and the individual measure of sampling adequacy (MSA) for each variable was ≥ 0.50. (ii) Bartlett’s test of sphericity [54], which tests the null hypothesis that the variables are uncorrelated; a P-value < 0.05 indicates sufficient correlation for PCA. Regarding the number of components, those that met two criteria simultaneously were retained: an eigenvalue greater than 1 (Kaiser’s criterion) and an eigenvalue greater than the 95th percentile of the eigenvalues obtained from 5,000 random datasets with the same number of observations and variables, a method known as Horn’s parallel analysis [55].

Loadings were calculated as the correlations between each variable and each component, without rotation. A variable was considered associated with a component when its absolute loading (|loading|) was ≥ 0.60, which implies that the component explains at least 36% of the variance in that variable [56]. Communality, the sum of the squared loadings on PC1 and PC2, indicates the proportion of the variance in each variable that is accounted for by the two retained components.

Furthermore, the effects of the effluent proportion, the cycle, and their interaction on the set of eight standardized variables were tested using permutational multivariate analysis of variance (PERMANOVA) [57], employing Euclidean distance and 9,999 permutations restricted to a split-plot design. The proportion and the interaction were tested by permuting whole plots within blocks, with the proportion tested against the “error a” term (block × proportion). The cycle was tested by permuting cycles within each plot, against the “error b” term. Effects with P < 0.05 were considered significant, and the R² value for each source indicates the proportion of the total multivariate variance explained by that source.

Finally, the ten treatments (proportion × cycle) were clustered using Ward’s method, based on Euclidean distances calculated from the standardized means of the eight measured variables. The quality of the representation was evaluated using the cophenetic correlation, which compares the distances in the dendrogram with the original distances; values above 0.70 indicate a good representation. It should be noted that the clustering was used solely as a descriptive summary.

The analyses were performed in R 4.3.3 [52], using the packages psych [54] and vegan [58].

Results and discussion

Quality of fertigation water and tested soil

As shown in Table 2, the mean values of the physicochemical variables increased with the rising proportion of aquaculture effluent. The observed variability, expressed by standard deviations, may be associated with the operational and seasonal conditions of the culture system, including daily feed input, organism excretion, tank cleaning, and the use of water with varying salinity levels. Although no medications are used in the aquaculture management, lime, caustic soda, and chlorine are employed for tank cleaning and disinfection, while salinity adjustment is achieved using sodium chloride and effluent from solar saltworks in the Mossoró region [38]. Furthermore, seasonality can alter effluent composition, resulting in greater dilution during the rainy season and evapoconcentration during the dry season.

Regarding pH, all evaluated proportions remained within the 6.5 to 8.4 range, which is considered suitable by the Food and Agriculture Organization of the United Nations (FAO) for the soil–plant system [59]. However, from the perspective of drip irrigation systems, pH values between 7.0 and 8.0 (for 25A2, 50A2, 75A2, and 100A2) and above 8.0 (for 0A2) are associated with moderate and severe risks of emitter clogging, respectively. This risk is particularly significant in waters with high Ca2+ or Mg2+ content or high hardness, as a pH above 7.5 can promote carbonate precipitation, leading to deposits on filter elements, piping, and drippers [60].

Compared to 0A2, increasing the proportion of aquaculture effluent raised the EC and, consequently, the saline effect on water availability to the crops by factors of 3.29, 5.64, 6.55, and 10.08 for 25A2, 50A2, 75A2, and 100A2, respectively. According to [59], 0A2, 25A2, and 50A2 fell within the slight-to-moderate restriction range, whereas 75A2 and 100A2 presented severe restrictions for the soil-plant system. Similarly, the risk of emitter clogging increased with the effluent proportion, shifting from low for 0A2 to moderate for 25A2, 50A2, and 75A2, and severe for 100A2, based on the criteria established by [60].

Regarding trace metals, the mean concentrations of Fe, Mn, Zn, and Cu across the different proportions of aquaculture effluent remained below the FAO-recommended limits for irrigation water of 5.0, 0.2, 2.0, and 0.2 mg L⁻¹, respectively [59], indicating a low risk of plant toxicity, primarily due to the slightly acidic to slightly alkaline soil conditions observed in the fertigated plots (Table 3).

Furthermore, mean Fe and Mn concentrations were generally below the 0.20 and 0.10 mg L⁻¹ thresholds associated with a low risk of emitter clogging [60], except for Fe in the 100A2 treatment, which showed a moderate risk. The elevated levels of Fe, Mn, Zn, and Cu in the aquaculture effluent may be primarily attributed to mineral supplementation in the feed, which includes iron sulfate, manganese sulfate, manganese monoxide, zinc oxide, and copper sulfate. A portion of these micronutrients is not assimilated by the cultured organisms and may be released into the water through excretion, feces, and uneaten feed residues.

Mean K+ levels across all aquacultures effluent ratios exceeded the standard limit for irrigation water established by the FAO [59]. The elevated potassium concentration in the aquaculture effluent may be attributed, in part, to the composition of the fish and shrimp feeds used, specifically the inclusion of plant-based ingredients such as soybean, wheat, and rice meals, as well as corn and other cereals, which are naturally rich in this element.

Na⁺ toxicity was classified as mild to moderate across all aquaculture’s effluent ratios, based on the criteria set by [59]. At high concentrations, Na⁺ can compromise membrane and protein integrity and affect cell division and expansion, primary and secondary metabolism, and mineral nutrient homeostasis [61].

Comparing the proportions of aquaculture effluent revealed a higher input of Mg²⁺ in the 75A2 and 100A2 treatments. This increase is physiologically significant, as Ca²⁺ and Mg²⁺ can exhibit antagonistic interactions in plant tissues, and plant growth and development depend on maintaining a homeostatic balance between these cations [61]. Regarding the risk of emitter clogging, Ca²⁺ posed a low risk in treatments 0A2, 25A2, 50A2, and 75A2, rising to moderate in 100A2. For Mg²⁺, the risk was low in 0A2, moderate in 25A2, and severe in treatments 50A2, 75A2, and 100A2 [62].

The increase in Na⁺, Ca²⁺, and Mg²⁺ in the aquaculture effluent may be associated with both feed composition and pond management practices. Ca²⁺ can be introduced via calcium compounds in feed formulations and the use of lime during cleaning; Na⁺, via sodium chloride in the feed, caustic soda used for sanitation, and, primarily, the addition of sodium chloride to adjust salinity. Mg²⁺, in turn, can be supplied by feed ingredients and mineral supplements, as well as by effluent from saltworks that is rich in this element.

Cl⁻ was the variable that showed the greatest increase with the rising proportion of aquaculture effluent. According to [59], the 0A2 treatment posed no toxicity risk, whereas the other proportions were classified as posing a severe risk. In excess, Cl⁻ can accumulate in plant tissues and impair photosynthesis, nutritional balance, growth, and crop quality; it can also induce chlorosis and leaf necrosis and, under certain conditions, increase the phytoavailability of Cd [63]. This enrichment may be primarily associated with the use of NaCl in feed formulation and salinity adjustment, the addition of solar saltworks effluent, and, secondarily, the use of chlorinated compounds for tank sanitation, which can generate Cl⁻ following disinfection reactions.

Mean CO3²⁻ and HCO3⁻ concentrations across all aquacultures effluent ratios remained below the standard limits of 1 and 10 meq L− 1, respectively, recommended by the FAO for irrigation water. Although critical concentrations were not observed, elevated levels of these anions can increase alkalinity and promote the precipitation of Ca²⁺ and Mg²⁺ as carbonates, thereby reducing their availability in the soil solution and proportionally increasing the relative proportion of Na⁺, with potential implications for sodicity and ionic balance [59]. Furthermore, high HCO3⁻ concentrations, typical of calcareous soils, can induce Fe deficiency and impair plant mineral nutrition [64].

The N and P inputs provided by the 25A2, 50A2, 75A2, and 100A2 ratios endow the aquaculture effluent with fertigation potential, as these macronutrients support essential physiological processes. N is a constituent of amino acids, proteins, chlorophyll, and components of the photosynthetic apparatus, whereas P plays a role in energy transfer via ATP, the formation of nucleic acids and phospholipids, and the regulation of photosynthesis and plant metabolism [65]. Recovering these macronutrients within the soil-plant system can partially reduce reliance on synthetic fertilizers and the emissions associated with their production chains, while simultaneously decreasing N and P discharge into receiving water bodies and, consequently, the risk of eutrophication [66].

Apparent changes in particle-size distribution should not be interpreted as evidence of pedogenic modification of soil texture during the 26-month experimental period of fertigation with different proportions of aquaculture effluent (Table 3). Instead, these changes may reflect the short-term redistribution of fine particles within the surface layer, inherent spatial variability, and potentially, the external input of aeolian material. In semi-arid environments, atmospheric dust deposition can introduce substantial quantities of fine mineral particles into surface soils, progressively altering their particle-size distribution [67].

This mechanism is particularly relevant in arid and semi-arid landscapes, where exposed soil surfaces can act as significant sources of wind-eroded material. Fine silt and clay fractions are especially susceptible to long-range atmospheric transport and, once deposited, can contribute significantly to the development, composition, and textural characteristics of the receiving soils [68]. These processes may have been further influenced by the marked seasonal alternation between dry periods, characterized by high atmospheric evaporative demand, and concentrated rainy periods (Fig. 3).

During dry periods, evapoconcentration may have favored the accumulation of soluble ions and exchangeable Na, whereas subsequent rainfall likely reduced electrolyte concentrations in the soil solution. Given a high ESP, such dilution can intensify clay dispersion and particle mobilization, with potential consequences for pore continuity and soil structural stability [69,70,71]. This interpretation is consistent with the increase in soil bulk density observed following fertigation (Table 3).

Statistical Assumptions of the Morphometric and Productive Variables of Forage Cactus

As shown in Table 4; Fig. 4, the residuals for 10 of the 11 variables followed a normal distribution (Shapiro–Wilk, P = 0.098–0.993; skewness between − 0.41 and 0.63), and variances were homogeneous across the proportion × cycle groups for all variables (Levene, P = 0.33–0.99). Only the fresh phytomass volume (FV) deviated from normality (W = 0.948; P = 0.028; skewness = 0.78; kurtosis = 2.35) and was, therefore, analyzed using ART-ANOVA.

Table 4 Normality and homogeneity of variance of the residuals of the split-plot mixed model (plot × cycle means, n = 50)
Full size table
Fig. 4
Fig. 4
Full size image

Distribution of residuals from the mixed-effects model for a split-plot design for each variable (plot means × cycle, n = 50): histogram, kernel density estimate (KDE), and a reference line at zero (dashed). The panels display the Shapiro–Wilk statistic (W) and P-value, skewness, kurtosis, and the coefficients of variation for errors a and b. Green border: normal residuals (P ≥ 0.05); red border: non-normal residuals

Univariate Statistical Analysis of Morphometric and Yield Variables in Forage Cactus

The proportion of effluent did not affect any of the 11 variables evaluated (P = 0.17–0.83; Table 5). The cultivation cycle significantly affected plant height (PH), plant width (PW), number of cladodes (NC), photosynthetic area (PA), FV, and fresh mass per plant (WFMS) (P < 0.001; Table 5).

Table 5 Analysis of variance of the split-plot in time (F and P values) and coefficients of variation
Full size table

These variables were higher in A2 than in A1 across all effluent proportions (Table 6). Conversely, cladode dimensions did not differ between cycles (P ≥ 0.33): thickness (CT), length (CL), diameter (CD), perimeter (CP), and cladode area (CA). These dimensions also showed less variation among plots (error coefficient of variation, CVa = 8.4–17.9%) compared to NC, PA, FV, and WFMS (CVa = 59.3–71.6%).

The proportion × cycle interaction was significant only for WFMS (P = 0.027) and FV (ART-ANOVA, P = 0.017; Table 5), indicating that the increase from A1 to A2 was not uniform across all proportions. For WFMS, the increase from A1 to A2 was significant for 0 A, 25 A, and 100 A, but not for 50 A and 75 A. Furthermore, for WFMS in cycle A2, the 25 A treatment (3,996 ± 457 g plant⁻¹) outperformed the 50 A treatment (1,979 ± 457 g plant⁻¹), though neither differed from the control (2,879 ± 457 g plant⁻¹; Table 6). For FV, A2 values were higher than those of A1 across all proportions, except for 75 A, and there were no differences between proportions within either cycle (Table 6).

Table 6 Means ± standard error of morphological and productive variables of spineless forage cactus under aquaculture effluent proportions in two cultivation cycles (n = 5 plots, two plants per plot)
Full size table

These responses occurred under substantial rainfall, totaling 1,589 mm over the experimental period, distributed as 944 mm for cycle A1 and 645 mm for cycle A2, alongside 449 mm of supplemental irrigation (178 mm and 271 mm, respectively). Precipitation accounted for approximately 78% of the total gross water input, underscoring its significance in the experimental water balance (Fig. 3e Table 1). Differences in precipitation and irrigation depths between the cycles, combined with plant development, may have contributed to the temporal variation observed in productive performance.

The proportion of effluent did not affect any of the 11 variables evaluated (P = 0.17–0.83; Table 5), and no proportion differed from the control in any of the cycles (Dunnett, P ≥ 0.19; for FV and ART-C, P ≥ 0.35; Table 7). Furthermore, there was no trend toward an increase or decrease associated with the proportions: no linear or quadratic contrasts were significant in any cycle (Table 8). Consequently, no regression equations were fitted.

Table 7 Comparison of each aquaculture effluent proportion with the control (0 A) within each cultivation cycle
Full size table
Table 8 Orthogonal polynomial contrasts of effluent proportion (0–100%) within each cultivation cycle (P values)
Full size table

The maintenance of the morphometric and productive performance of forage cactus when supplied with varying proportions of aquaculture effluent may be linked to a set of morphophysiological adaptations that favor its survival and growth in semi-arid environments, most notably Crassulacean Acid Metabolism (CAM). This metabolic pathway is characterized by the predominance of nocturnal CO₂ assimilation, which occurs when temperatures and the atmospheric vapor pressure deficit are generally lower, thereby reducing water loss through transpiration [72, 73].

Multivariate Statistical Analysis of Morphometric and Productivity Variables in Forage Cactus

In the heatmap presented in Fig. 5, it was observed that the variables formed two groups. The first group comprised plant size and yield: NC, PH, PW, and WFMS showed strong positive correlations with one another (r = 0.82 to 0.91), meaning that plants with more cladodes were taller, wider, and heavier. The second group consisted of cladode dimensions: CD–CP (r = 0.96), CL–CD (r = 0.77), and CL–CP (r = 0.72). There was little correlation between the two groups; specifically, NC showed virtually no association with the size of individual cladodes (r = 0.14 to 0.28; P > 0.05). CT exhibited weak, negative correlations with PH (r = − 0.34) and NC (r = − 0.29) and showed no correlation with the other cladode dimensions (|r| ≤ 0.27).

Fig. 5
Fig. 5
Full size image

Pearson correlation matrix among morphological and productive variables of spineless forage cactus (plot × cycle means, n = 50). ∗ and ∗∗: P < 0.05 and P < 0.01. CA, PA and FV are calculated from the measured variables

Overall, the morphometric and productive data for the forage cactus met both criteria for suitability for PCA. The global KMO was 0.77 [53], which is above the minimum threshold of 0.60. The individual index values for each variable (MSA) were PH = 0.91; CT = 0.85; WFMS = 0.81; PW = 0.80; CP = 0.73; NC = 0.72; CL = 0.72; and CD = 0.67, all of which are above the minimum threshold of 0.50. Bartlett’s test was significant (χ² = 436.6; 28 df; P < 0.001), leading to the rejection of the hypothesis that the variables are uncorrelated [54]. Therefore, the eight variables share sufficient correlation to be summarized into principal components.

Thus, two components met the two retention criteria (Table 9): PC1 (eigenvalue 4.55, which is above 1 and the parallel analysis threshold of 1.87) and PC2 (eigenvalue 1.85, which is above 1 and the threshold of 1.53) [55]. Together, they accounted for 80.0% of the variation across the eight variables: 56.9% in PC1 and 23.1% in PC2.

Table 9 Eigenvalues and explained variance of the principal component analysis (plot × cycle means, n = 50)
Full size table

The communalities were high (0.80 to 0.95) for all variables, except for CT (0.16), indicating that the two components represent all variables well (Table 10).

Table 10 Loadings of the variables on the first two principal components (unrotated)
Full size table

PC1 represented plant size and yield. The highest loadings [56] were for PH (0.90), PW (0.87), WFMS (0.86), and NC (0.82), followed by CD and CP (0.73) (Table 10; Fig. 6). High PC1 values indicate larger plants with more cladodes and greater fresh mass. PC2 represented cladode size in opposition to the number of cladodes: CL (0.69), CD (0.63), and CP (0.61) had positive loadings, while NC had a negative loading (-0.53). Thus, plants with more cladodes tended to have relatively smaller cladodes. CT was not associated with either of the two components (|loading| ≤ 0.39) and was the only variable related to the third component, which was not retained (eigenvalue of 0.91). This indicates that CT varied independently of the other variables. Among the supplementary variables, PA (0.89) and FV (0.85) aligned with PC1, while CA aligned with PC2 (0.71).

Fig. 6
Fig. 6
Full size image

Biplot of the principal component analysis of the eight measured variables (plot × cycle means, n = 50). Solid arrows, measured variables; dashed arrows, supplementary (calculated) variables. Small symbols, plot × cycle observations; large labelled symbols, treatment means (circles, cycle A1; triangles, cycle A2). Colours indicate the effluent proportion. Ellipses: 95% confidence regions for cycles A1 (dashed) and A2 (solid)

Including these three variables in the component calculation did not alter the structure (two components, accounting for 60.0% and 23.6% of the variation) or the interpretation. In the biplot (Fig. 6), the plots from cycle A2 shifted toward positive PC1 values compared to those from cycle A1, indicating larger and more productive plants in the second cycle. The effluent proportions overlapped within each cycle, with no apparent separation.

PERMANOVA confirmed the pattern observed in the PCA (Table 11). The cycle affected the set of eight variables (pseudo-F = 11.49; P < 0.001; R² = 0.134), explaining 13.4% of the total multivariate variation [57]. The proportion of effluent had no effect (pseudo-F = 0.79; P = 0.647; R² = 0.074), nor did the interaction between proportion and cycle (pseudo-F = 0.74; P = 0.494; R² = 0.035). Blocks accounted for 14.6% of the variation, while the variation between plots (error a) accounted for 37.8%, the largest share of the total.

Table 11 Permutational multivariate analysis of variance (PERMANOVA) of the eight measured variables in the split-plot structure
Full size table

These results are consistent with the univariate analysis, in which the proportion of effluent did not affect any variable, while the cycle affected the growth and mass variables. The interaction detected in the univariate analysis for WFMS and FV was not reflected across the entire set of variables, as it was restricted to these two variables and accounted for only 3.5% of the multivariate variation.

Finally, the dendrogram (Fig. 7) separated the ten treatments into two groups, one containing the five treatments from cycle A1 and the other containing the five from cycle A2, consistent with the cycle effect detected by PERMANOVA [57]. The cophenetic correlation was 0.67, slightly below 0.70, indicating that the dendrogram represents the original distances only reasonably well. For this reason, and because it is a descriptive technique, the proximity between treatments within each cycle was not interpreted as indicating equality or difference between effluent proportions; this aspect was instead evaluated via PERMANOVA, which did not detect any effect of the proportion.

Fig. 7
Fig. 7
Full size image

Hierarchical clustering (Ward’s method, Euclidean distance) of the ten treatments (effluent proportion × cultivation cycle), based on the eight standardized measured variables (cophenetic correlation = 0.67)

Therefore, the hypothesis of the present study was partially confirmed regarding the temporal dependence of certain production responses, but it was not supported concerning the statistical superiority of the 25% aquaculture effluent ratio. The study’s main finding is that, over a 26-month cultivation period, the partial or total replacement of the water supply with aquaculture effluent did not result in statistically detectable reductions in morphometric or production parameters compared to the control. Thus, this finding broadens the prospects for using aquaculture effluent as an alternative source of nutrients and water to produce forage cactus in semi-arid regions, although its long-term agronomic viability depends on further assessments regarding soil salt accumulation, solute mobility, and the crop’s physiological responses.

Study Limitations and Prospects for Future Research

This study presents certain limitations related to the environmental heterogeneity of the experimental area and the absence of physical, hydrological, and physiological indicators that would have allowed for a more precise understanding of the mechanisms underlying temporal changes in soil attributes, as well as the morphometric and productive responses of forage cactus irrigated with varying proportions of aquaculture effluent in the Brazilian semi-arid region. Although a randomized complete block design was employed to mitigate the effects of spatial heterogeneity, shading could not be entirely controlled throughout the cultivation cycles [74]. Trees and other structures located primarily to the north and east of the experimental site (Fig. 1) served as sources of spatially heterogeneous shading, with the position, duration, and intensity of this shade varying seasonally according to the sun’s path. Consequently, experimental units within the same block may have been exposed to different levels of solar radiation, with a portion of this variability remaining as an uncontrolled environmental factor. This is particularly significant for forage cactus, as radiation availability directly influences photosynthetic activity, cladode growth, and biomass accumulation. In Opuntia ficus-indica [75], observed that a 70% reduction in irradiance led to an approximately 17% decrease in biomass accumulation in newly formed cladodes, as well as an 11% reduction in their thickness. Similarly [76], linked shading to a reduction in CO₂ assimilation, relative water content, parenchyma thickness, and chlorophyll content. Therefore, spatial differences in light availability may have contributed to the observed variability in the attributes PH, NC, CT, CL, CD, CP, PW, FV, WFMS, CA, and PA, regardless of the proportions of aquaculture effluent applied.

To minimize edge effects, border plants were established around the experimental area, aiming to provide the peripheral units with growth conditions more like those observed within the interior of the cultivation site. This strategy can reduce differences associated with plant position, given that outer rows often intercept greater amounts of direct and diffuse photosynthetically active radiation compared to inner rows, a factor that influences growth independently of the applied treatments [77]. Nevertheless, the presence of these borders did not eliminate the heterogeneous shading caused by surrounding trees and structures (Fig. 1). Furthermore, the border plants themselves may have caused localized shading on adjacent experimental units, the magnitude of which was not quantified.

Other limitations relate to the absence of indicators capable of clarifying the mechanisms responsible for the temporal variability of the soil’s physical attributes. Parameters such as water-dispersible clay, aggregate stability, and hydraulic conductivity were not determined. Likewise, wind speed and direction, as well as the intensity and composition of aeolian deposition, were not monitored. Consequently, it was not possible to quantify the relative contribution of particle dispersion and redistribution processes, atmospheric sediment inputs, and the soil’s intrinsic spatial variability to the changes observed throughout the cultivation cycles. Furthermore, the absence of drainage lysimeters precluded the direct determination of percolated water volumes, the mobility of solutes within the soil profile, and the loss of nutrients and salts through leaching. This limitation constrained the development of more comprehensive water and solute balances, as well as the assessment of the transport, retention, and accumulation processes of constituents associated with the prolonged application of aquaculture effluent.

Regarding physiological aspects, the absence of biochemical and photosynthetic assessments limited the interpretation of the mechanisms involved in maintaining the morphometric and productive performance of the forage cactus. Indicators such as proline concentration, soluble sugar content, tissue water potential, stomatal conductance, CO₂ assimilation, chlorophyll a fluorescence, and nocturnal organic acid accumulation were not monitored. These variables could have helped in understanding processes such as osmotic adjustment, water conservation, stomatal regulation, and the maintenance of photosynthetic activity, all of which are associated with the acid metabolism of the forage cactus. Specifically, measuring proline and soluble carbohydrates would have allowed for the investigation of potential metabolic responses to changes in water availability and the chemical composition of the irrigation water. In this context, the lack of statistically significant differences in morphometric and productive attributes between the various proportions of aquaculture effluent and the control treatment does not rule out the possibility of physiological or biochemical alterations that remained undetected by the variables evaluated.

Given these limitations, it is recommended that future research adopt an integrated approach to monitoring the soil–water–plant–atmosphere system. Initially, priority should be given to experimental areas with greater uniformity in terms of solar exposure, duration of shading, and the spatial distribution of irradiance, which would allow for the quantification of how these conditions influence the growth and productivity of forage cactus. Concurrently, temporal monitoring of the soil’s physical, chemical, and hydraulic properties, combined with the characterization of aeolian deposition and spatial variability, could clarify the mechanisms underlying the changes observed throughout the cultivation period. The installation of drainage lysimeters, alongside the determination of electrical conductivity, ionic composition, and nutrient concentrations in the percolated water, would enable the quantification of drainage and leaching fluxes, thereby contributing to the assessment of risks related to salinization, sodification, and nutrient loss. Furthermore, the inclusion of biochemical and physiological indicators, specifically proline levels, soluble sugars, gas exchange, the photochemical efficiency of photosystem II (Fv/Fm), and the diurnal variation in cladode acidity, could shed light on whether the maintenance of productive performance is linked to the stability of the plant’s water status, osmotic regulation, and the preservation of CAM metabolism. Finally, long-term studies conducted under diverse edaphoclimatic conditions, utilizing integrated monitoring of these indicators, will be essential for understanding the cumulative effects of irrigation with aquaculture effluent. Such research will be vital for establishing management strategies that balance forage cactus productivity with the preservation of soil quality and the sustainable use of water resources in semi-arid regions [78,79,80].

Conclusions

The hypothesis that a 25% proportion of aquaculture effluent would maximize the morphometric and productive performance of Opuntia cochenillifera (Nopalea cochenillifera) was not statistically confirmed. Over the course of 26 months and two successive cultivation cycles, the effluent proportions did not show significant main effects on the morphometric and productive variables evaluated. However, the cultivation cycle significantly influenced plant height, plant width, number of cladodes, total photosynthetic area, volume of fresh phytomass, and fresh mass per plant. Significant interactions between the proportion and the cycle were detected for the volume of fresh phytomass and the fresh mass per plant, indicating that productive responses were dependent on the cultivation cycle.

Multivariate analysis reinforced these results, with the first two principal components explaining approximately 78% of the total variability. The number of cladodes and the total photosynthetic area showed a strong association with the volume of fresh phytomass, highlighting their relevance to biomass production. Principal component analysis revealed a clearer differentiation between cultivation cycles than between effluent proportions, indicating that temporal changes in plant development were more pronounced than differences resulting from the treatments in terms of overall morphometric and productive response.

During cycle A2, the effective depths of aquaculture effluent applied were 28, 56, 84, and 112 mm, representing, respectively, approximately 10.3%, 20.7%, 31.0%, and 41.3% of the total irrigation depth (271 mm). Therefore, the 100% effluent proportion in the manually applied mixture resulted in the highest level of substitution for the water supply among the treatments evaluated, with no statistically detectable reduction in productive performance compared to the control.

In general, aquaculture effluent partially replaced the water supply without any statistically detectable reductions in the morphometric and productive performance of the Opuntia cochenillifera (Nopalea cochenillifera), under the conditions evaluated. It should be noted that our results are specific to Opuntia cochenillifera (Nopalea cochenillifera), the edaphoclimatic conditions of the semi-arid region of Rio Grande do Norte, the recorded rainfall, the irrigation practices employed (alternating between drip irrigation with water supply and manual irrigation using a mixture of aquaculture effluent and water supply), and the fertilization management adopted. Although the treatment with 25% effluent showed numerically favorable productive responses during period A2, the treatment with 100% effluent achieved the highest level of replacement of the water supply. Therefore, further long-term research is required, incorporating the dynamics of soil salinity and sodicity, drainage monitoring, solute balance, and plant physiological responses, to assess the environmental implications and the feasibility of a broader application of aquaculture effluent reuse for forage production in semi-arid regions.