It is generally accepted that chemical, physical, and biological processes intervene in the remediation of waste and soils contaminated with hydrocarbons (Volke and Velasco 2002; Riser-Robert, 2019). The use of bioremediation methods has aroused a considerable research interest and currently provides advantages over other treatment processes (Kumar et al., 2011; Tyagi et al., 2011; Thapa et al., 2012; Varjani, 2017; Yuniati, 2017; de Oliveira et al., 2018). Among the bioremediation methods, biopiles are an ex situ technology where the stimulation of microbial activity is achieved through the direct addition of nutrients and texturizers (agro-industrial organic wastes), as well as aeration and humidity control (Velasco and Volke, 2003; Prakash et al., 2015; EPA, 2017; Martínez et al., 2017; Ossai et al., 2020).

The removal efficiency of biopiles is influenced by a considerable number of variables, which can be grouped into three categories: the characteristics of the soil and texturizers, the characteristics of the pollutant, and climatic conditions. Among the essential factors are the hydrocarbon concentration, the texturizer content, the moisture content, and the C:N:P:K ratio (Velasco and Volke, 2003; Prakash et al., 2015; EPA, 2017; Yuniati, 2017; Al-Hawash et al., 2018).

Moreover, the response surface methodology is widely known as an experimental strategy that allows finding the best-operating conditions for a process. Particularly, the Box-Behnken experimental design is an excellent economic alternative because it requires a minimum of experimental runs to study three or more factors at three levels (Gutiérrez and de la Vara, 2012; Montgomery, 2017). During the past decade, the Box-Behnken design has been extensively applied to hydrocarbon bioremediation research (Agarry and Ogunleye, 2012; Martínez et al., 2015; Khaled et al., 2016; Shuo et al., 2019; Zhang et al., 2020).

Although these procedures guarantee an individual optimum, when more than one response variable is of interest in the arrangement, the global optimum is generally not the combination of the optimizations performed by the independent variables. Therefore, it is necessary to find a compromise solution where all the variables have a satisfactory level. This solution is recognized as a simultaneous optimum (Montgomery, 2017).

Statistical data report that, in the province of Cienfuegos, Cuba, there is still confined oiled sludge resulting from the territorial companies' collection, which amounts to 2 993 740 kg (CITMA, n.d.; Castro et al., 2022a). In this vein, the Environmental Studies Center of Cienfuegos (CEAC), the Institute of Marine Science (ICIMAR), and the Soil Research Institute (IIS) are three Cuban Institutions that have developed the project Evaluation of the ecotechnology of biopiles for the bioremediation of oily waste using local texturizers (BIOPILA).

The BIOPILA project aims to address an innovative and resilient technique to cope with the environmental problem of soil polluted with hydrocarbons in Cuba, specifically with the residues generated by the local industry which remain confined without a solution (Castro et al., 2022a). Previous research carried out within the framework of the aforementioned project has shown, at the bench scale, the relevance of biopile ecotechnology for the degradation of soils polluted with hydrocarbons (Gutiérrez et al., 2021; Castro et al., 2022b, Gutiérrez, et al., 2022; Castro et al., 2022c). Significant removal rates were reported for oils and grease concentrations (between 50,05 and 66,08%). For TPH, values between 40,00 and 59,12% were reported for all experimental treatments tested. However, neither the way in which the biopile design variables reported by Gutiérrez et al. (2020) should be operated nor the abiotic factors described by Casals et al. (2020) were sufficiently clear in terms of their influence on the bioremediation efficiency of waste and soil polluted with hydrocarbons. Hence, as part of the research-development BIOPILA project, this paper focuses on determining the best operating variable ranges of biopile ecotechnology at the bench scale. Moreover, the simultaneous optimization of biopile hydrocarbons removal was implemented as a prototype to obtain a compromise solution that maximized the mass of TPH removed while keeping the pollutant concentration under the Cuban disposal regulations.

Some bench-scale experiments were carried out with biopile ecotechnology to bioremediate hydrocarbons. This technique has proven to be more efficient than others in removing hydrocarbons from sludges (Shahryar, 2017). Moreover, it has the advantage that the microorganisms are already adapted to the site conditions (Casale et al., 2018).

Figure 1 presents the heuristic diagram of the experimental protocol used in the bench-scale experimentation phase. The operating variables of interest for this research only comprised those related to the Evaluation of hydrocarbon degradation stage (highlighted in bold in Figure 1).

Figure 1 Source: Authors

Under methodological convergence, several engineering tools were used, such as the design of experiments, mass balance, analytical chemistry, engineering design, and statistical analysis. The experimental polygon consisted of an industrial warehouse, roofed, with a concrete floor and provided with open windows, where 15 experimental units were settled within identical polyethylene basins that were 1,10 m long and 0,95 m wide. Each basin acted as a container to avoid mixing materials and leakage, separating each experimental unit from the surroundings.

Selection, collection, and characterization of materials

The critical materials used for the experimentation are detailed below:

A mix of solid waste and soils contaminated with hydrocarbons, collected from the bottom of a settler.

Firstly, the oily sludge was retrieved during the cleaning procedures in the waste treatment plant of an oil refinery.

Secondly, a typical greyish-brown soil from the central region of Cuba was used (density = 1 190 kg.m^-3). The soil was classified as silty clay sand, with low plasticity and moderate permeability. It was sieved using an industrial mesh for particles smaller than 4,75 mm, according to Cuban standards based on ASTM norms (ASTM International, 2002, 2017).

For the mixture of texturizers, the best candidates for different treatments were selected, which had been tested in previous stages of the BIOPILA project (Casals et al., 2020; Gutiérrez et al., 2020; Castro et al., 2022c; Castro et al., 2022d). On the one hand, the candidate which reported the lowest TPH concentrations at 240 days, the highest degradation kinetic rate, and the shortest half-life (t_1/2) among all the considered treatments was used. On the other hand, the treatment that maintained the necessary water content requirements in the biopiles was also included in the mixture, with values of up to 35% moisture, increasing the soil porosity and contributing to soil moisture retention according to the criteria of Zang et al. (2020).

Finally, it is remarkable that the flora stimulated during the technological process was composed of the individual contributions of the microorganisms in each component of the biopile mixture (oily sludge, soil, and texturizers). The kinds and initial concentrations of the main groups of microorganisms identified in the mixture were total heterotrophic bacteria, filamentous fungi and yeasts, and hydrocarbon degraders, in the orders of 10⁶ CFU.g^-1, 10⁴ CFU.g^-1, and 10⁵ CFU.g^-1, respectively. The concentrations above ensured a total heterotrophic bacteria concentration greater than 10³ CFU.g^-1, in line with the initial requirements established for biopile efficacy (EPA, 2017).

Interpretation of the experimental design analysis results

Using the methods described above, the refined models' ANOVA for both response variables can be observed in Tables 2 and 3.

Table 2

Source: Authors

In the individual model of removed mass, the initial TPH concentration (A) was significant. It was also corroborated that it was the factor with the most influence on the response variability. Furthermore, the initial TPH concentration (A) is also significant in the individual model of final TPH concentrations. It is also noteworthy that the effects of factors A and C are greater than that of factor B, given that, when the former are increased, they further modify the variability of the response.

Table 3

Source: Authors

Although the first-order effects of the texturizer (B) and moisture (C) contents did not turn out to be significant, they were maintained in the models given the need to obtain a hierarchical model. Indeed, it is recognized that, in the Response Surface Methodology, this kind of model is preferred (Gutiérrez and de la Vara, 2012; Montgomery, 2017). Therefore, both terms were maintained in both models, as the BC interaction resulted to be significant. On the other hand, the quadratic interaction BB was also kept in both models because its elimination implied a decrease in the R² _aj.

The fitted mathematical models that describe the behavior of the mass of TPH removed and the final TPH concentration are presented in Equations (2) and (3), respectively.

where:

MR: mass of TPH removed in the experimental unit (kg)

TPH_t = ₂₄₀: TPH concentration in the experimental unit at 240 days (mg.kg^-1) (final concentration)

A: Initial TPH concentration (mg.kg^-1)

B: Texturizer content (% mass)

C: Moisture content (% mass)

Reasonable coefficients of determination were obtained from Equations (2) and (3), R² _MR = 95,54% and R² _TPHt=₂₄₀ = 86,41%. Although these results are barely lower than the R² reported in similar research (Agarry and Ogunleye, 2012; Zhang et al., 2020), it could be inferred that the models, adjusted in this way, explain a considerable percentage of the total variability. On the other hand, according to the criteria given by Gutiérrez and de la Vara (2012), the adjusted coefficient of determination (R² _aj) is more suitable for comparing the models of Equations (2) and (3). Regarding this issue, it yields values of R² _ajMR = 91,07% and R²._TPHf= 72,81%; in both cases, the required 70%, in agreement with the criterium of Montgomery (2017), is appreciably exceeded. Consequently, Table 4 shows the final TPH concentration values obtained through laboratory tests of the samples obtained from monitoring and the values adjusted by the model of both response variables for each treatment and levels assigned to the three factors under study.

Table 4

Source: Authors

By comparing the observed and adjusted values in Table 4, a similar behavior could be appreciated in the 15 treatments, with no sudden changes and variations, which is consistent with Agarry and Ogunleye (2012) and Shuo et al. (2019).

Figures 2 and 3 illustrate the estimated response surface mesh for both response variables under analysis.

Figure 2 Source: Authors

As shown in Figure 2, the highest mass of TPH removed (between 1,17 and 1,33 kg) is found in the region comprised simultaneously by the low and high levels of moisture and texturizer content, along with the high level of the initial TPH concentration.

Figure 3 Source: Authors

On the other hand, in Figure 3, it can be observed that the region with lower final concentrations of TPH (up to 5 400 mg.kg^-1) was found where there was a lower initial concentration of TPH and lower texturizer content but higher moisture content.

Table 5 shows the values of the studied factors for which the desirability was maximum (1), according to the results of the simultaneous optimization.

Table 5

Source: Authors

For the levels associated with desirability, a significant TPH mass of 1,19 kg was removed, representing 79,58% of the initial TPH concentration. Likewise, a final TPH concentration of 8 000 mg/kg was obtained, which reasonably complies with the final disposal standards of Cuba.

Figure 4 shows the estimated response surface plot for the desirability function, and Figure 5 shows the contour of the estimated response surface for the same function. In both cases, the moisture content factor was kept at its optimum value (25,95 %).

The optimum moisture content where the desirability was maximum (approximately 26%) is under the levels of moisture percentage by weight (12-30%) recommended by EPA (2017) for the design of biopiles.

In addition, the optimum also fits the qualitative criteria of manipulation, as evidenced during the experimental phase, because the experimental units with a high moisture content had severe difficulties in their handling. Specifically, these biopiles (30% of moisture content) exhibited marked differences from the ones in the medium and low levels regarding moisture content, thus producing puddling of the mass and a more significant generation of leachates, with a proven transfer of the pollutant to the aqueous medium.

Figure 4 Source Authors

Figure 5 Source: Authors

By interpreting Figures 4 and 5, it can be confirmed how the values of the factors shown in Table 5 could achieve optimum desirability. In addition, it was observed that the desirability function is maintained around the maximum for texturizer content values between 6 and 10%, keeping the other two studied factors at their optimal values.

Additionally, the models' assumptions can be graphically verified for both response variables, as shown in the Appendices (Figures 6 and 7).

It was remarkable that the statistical analyses and the closeness of the experimental results and model predictions show the reliability of the regression model. The highest mass of TPH removed was obtained in the region defined by the high initial concentration of TPH and the levels from low to high regarding the texturizer and moisture contents. On the other hand, the lowest values for the final TPH concentrations were found in the region defined by the high moisture content factor and the low levels of both the initial TPH concentration and the texturizer content.

It was crucial to obtain a simultaneous optimum with initial concentration values of TPH of 39 278 mg.kg^-1, a texturizer content of 6,45%, and a moisture content of 25,95%. For these values, it was possible to maximize the mass of TPH removed and maintain the final TPH concentration of 8 000 mg.kg^-1, lower than that stipulated by the Cuban disposal standard. The mass of TPH removed remained in the range of the compromise solution for texturizer content values between 6 and 10%. At the same time, the other two factors studied were kept at their optimal values.

Finally, this research responded to a scientific demand for the continuity of the subsequent stage of the research-development BIOPILA project. The results are encouraging and should be validated during the ecotechnology scale-up processes, i.e., into a pilot scale.