Global sensitivity and uncertainty analysis of a microalgae model for wastewater treatment

https://doi.org/10.1016/j.scitotenv.2021.150504Get rights and content

Highlights

  • A global sensitivity and uncertainty analysis of a microalgae model was conducted.

  • Eleven factors out of 34 were identified as influential and were calibrated offline/online.

  • Four different experimental periods were properly represented by a single set of model factors.

  • The uncertainty bandwidth revealed a need to apply offline calibration procedures.

Abstract

The results of a global sensitivity and uncertainty analysis of a microalgae model applied to a Membrane Photobioreactor (MPBR) pilot plant were assessed. The main goals of this study were: (I) to identify the sensitivity factors of the model through the Morris screening method, i.e. the most influential factors; (II) to calibrate the influential factors online or offline; and (III) to assess the model's uncertainty. Four experimental periods were evaluated, which encompassed a wide range of environmental and operational conditions. Eleven influential factors (e.g. maximum specific growth rate, light intensity and maximum temperature) were identified in the model from a set of 34 kinetic parameters (input factors). These influential factors were preferably calibrated offline and alternatively online. Offline/online calibration provided a unique set of model factor values that were used to match the model results with experimental data for the four experimental periods. A dynamic optimization of these influential factors was conducted, resulting in an enhanced set of values for each period. Model uncertainty was assessed using the uncertainty bands and three uncertainty indices: p-factor, r-factor and ARIL. Uncertainty was dependent on both the number of influential factors identified in each period and the model output analyzed (i.e. biomass, ammonium and phosphate concentration). The uncertainty results revealed a need to apply offline calibration methods to improve model performance.

Introduction

Microalgae-based wastewater treatment represents a promising biological system to treat different wastewater sources in a way that can transform conventional wastewater treatment plants (WWTPs) into water resource recovery facilities (WRRFs) (Seco et al., 2018). Photoautotrophic microalgae use light energy, inorganic carbon and nutrients (inorganic nitrogen and phosphorus) for growth. Solar energy and nutrients are harvested in form of microalgae biomass while inorganic carbon is biofixed. Microalgae-based wastewater treatment can reduce treatment costs, generate clean water and reduce the environmental impact of the process (Seco et al., 2018).

An in-depth knowledge of the processes involved in microalgae metabolism is required to better understand how to operate microalgae-based technologies, how to optimize processes associated, how to improve reactor design and how to select the best control strategies to enhance pollutant removal efficiency. Microalgae and traditional activated sludge systems are intrinsically complex, since both depend on environmental variables such as temperature, pH, substrate availability, etc. However, it should be noted that photoautotrophic microalgae metabolism is not only affected by the environmental factors that influence activated sludge but also by seasonal and daily fluctuations in light intensity (González-Camejo et al., 2018). The correct operation of microalgae-based wastewater treatments thus demands a robust, feasible and efficient tool to forecast the culture development and its compliance with increasingly stringent regulations. Mathematical models can help to study the main processes and variables that influence algal metabolism in different culture media, including municipal wastewater.

An array of mathematical models for predicting microalgae growth has been developed in the last ten years (Costache et al., 2013; Eze et al., 2018; Ndiaye et al., 2018; Ruiz et al., 2013; Solimeno et al., 2015, Solimeno et al., 2017; Wágner et al., 2016). This process cannot be considered a well-characterized system, since some model factors are uncertain and speciation-dependent. The ammonium semi-saturation constant has been reported to range from 0.1 to 31.5 g N m−3 (Aslan and Kapdan, 2006; Solimeno et al., 2017), and is a perfect example of the intrinsic variability and uncertainty of model factors, so that the application of these models requires a great number of assumptions regarding the simplification of biological processes and model factors. These assumptions are sources of uncertainty that could propagate through the model thus generating uncertainty in the model outputs. The resulting uncertainty in the model results could lead to misleading decisions during process design and/or optimization. Hence, performing a global sensitivity and uncertainty analysis (GSA and UA, respectively) would help to deal with these issues by analysing and understanding model performance. GSA involves identifying the most important model factors to be calibrated, while UA entails determining the model output uncertainty derived from uncertain model input factors (Rajabi et al., 2020). GSA and UA should be performed concurrently, as both are essential parts of the model development process in design optimization, reliability analysis, and data-worth analysis (Rajabi et al., 2020). To the best of the authors' knowledge, both GSA and UA have not been performed concurrently in mathematical models for wastewater treatment with microalgae. Therefore, no information is available on the microalgae models' most influential factors and the variability of the uncertainty of model output.

Although, the mechanistic microalgae model proposed by Viruela et al. (2021) was validated using 4 experimental periods, which combine key environmental and operational conditions characteristic of a microalgae-based wastewater treatment, the uncertainty of model parameters could lead to uncertainty propagations on modeling results, reducing its practical application. Thus, this study tends to address data gaps related to uncertainty on microalgae-based wastewater treatment models, based on Viruela et al. (2021), by performing a GSA and UA. The Morris screening method was applied as GSA method to identify the most influential factors of the model, which were calibrated through offline (obtained from experimental assays) and online (variation of model parameters to match model predictions to experimental results) methodologies. For further enhancing the model performance, the calibrated values for the influential factors were dynamically optimized using online data. Model uncertainty was analyzed and quantified from Monte Carlo simulations and three uncertainty coefficients: the p-factor, the r-factor and the Average Relative Interval Length (ARIL). A calibration protocol was also recommended to reduce model uncertainty by means of prioritizing different calibration methodologies.

Hence, this work could be seen as the first study to simultaneously perform GSA and UA in the field of microalgae-based wastewater treatment modeling, while proposing a set of input factors to be calibrated by a given protocol.

Section snippets

The mathematical model

The model used in this work (Viruela et al., 2021) simulated microalgae growth from different phosphorus and nitrogen sources. Regarding phosphorus source, the microalgae had two different metabolic pathways: under phosphorus-replete conditions, microalgae uptake dissolved extracellular phosphate (SPO4) to support their vital metabolic functions and stored part of the excess in form of intracellular polyphosphate (XPP-ALG) while under phosphorus-starved conditions they consumed their XPP-ALG

SNHX output

SNHX concentration decreases due to microalgae uptake for growth and SNH3 stripping. Conversely, SNHX concentration increases due to microalgae lysis and endogenous respiration. Processes 1 (XALG growth on SNHX and SPO4), 3 (XALG growth on SNHX and XPP-ALG), 6 (XALG endogenous respiration), 7 (XALG lysis) and 11 (S[NH3] stripping) in Table 1 therefore affect SNHX concentration.

Fig. 1 gives the sensitivity measurements (μ* and σ) calculated from each input factor on the SNHX output for the 4

Conclusions

This paper presents a GSA, an offline/online calibration, a dynamic optimization, and a UA of a previously proposed and validated microalgae model. Eleven out of 34 influential factor were identified from the GSA. The four factors with the most important overall effect on the three outputs evaluated (SNHX, SPO4 and XALG) were μALG, qXPP, TMAX and IOPT. SNHX and XALG model outputs were influenced by kinetic input factors related to microalgae growth, while SPO4 model output was affected by X

CRediT authorship contribution statement

Stéphanie Aparicio: Analysis, Data curation, Writing-original draft; Rebecca Serna-García: Data curation, Revision and Supervision; Aurora Seco: Revision, Supervision and Funding acquisition; José Ferrer: Revision, Supervision and Funding acquisition; Luis Borrás-Falomir: Revision and Supervision; Ángel Robles: Data curation, Revision and Supervision and Funding acquisition.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgements

This research work was supported by the Spanish Ministry of Economy and Competitiveness (MINECO, Projects CTM2014-54980-C2-1-R, CTM2014-54980-C2-2-R, CTM2017-86751-C2-1-R and CTM2017-86751-C2-2-R) jointly with the European Regional Development Fund (ERDF), both of which are gratefully acknowledged. It was also supported by the Spanish Ministry of Education, Culture and Sport via a pre-doctoral FPU fellowship to author Stéphanie Aparicio (FPU/15/02595).

References (57)

  • X. Jin et al.

    Parameter and modeling uncertainty simulated by GLUE and a formal Bayesian method for a conceptual hydrological model

    J. Hydrol.

    (2010)
  • S. Kim et al.

    Removal of nitrogen and phosphorus by Chlorella sorokiniana cultured heterotrophically in ammonia and nitrate

    Int. Biodeterior. Biodegrad.

    (2013)
  • G. Mannina et al.

    Greenhouse gases from membrane bioreactors: mathematical modelling, sensitivity and uncertainty analysis

    Bioresour. Technol.

    (2017)
  • G. Mannina et al.

    Sensitivity and uncertainty analysis of an integrated ASM2d MBR model for wastewater treatment

    Chem. Eng. J.

    (2018)
  • H. Nagase et al.

    Uptake pathway and continuous removal of nitric oxide from flue gas using microalgae

    Biochem. Eng. J.

    (2001)
  • M. Ndiaye et al.

    CO2 gas–liquid mass transfer and kLa estimation: numerical investigation in the context of airlift photobioreactor scale-up

    Chem. Eng. Res. Des.

    (2018)
  • M. Pastore et al.

    Application of photorespirometry to unravel algal kinetic parameters of nitrogen consumption in complex media

    Algal Res.

    (2020)
  • M.M. Rajabi et al.

    Uncertainty quantification and global sensitivity analysis of double-diffusive natural convection in a porous enclosure

    Int. J. Heat Mass Transf.

    (2020)
  • A. Robles et al.

    Global sensitivity analysis of a filtration model for submerged anaerobic membrane bioreactors (AnMBR)

    Bioresour. Technol.

    (2014)
  • A. Robles et al.

    Model-based automatic tuning of a filtration control system for submerged anaerobic membrane bioreactors (AnMBR)

    J. Memb. Sci.

    (2014)
  • M.V. Ruano et al.

    An improved sampling strategy based on trajectory design for application of the Morris method to systems with many input factors

    Environ. Model. Softw.

    (2012)
  • J. Ruiz et al.

    Performance of a flat panel reactor in the continuous culture of microalgae in urban wastewater: prediction from a batch experiment

    Bioresour. Technol.

    (2013)
  • A. Ruiz-Martinez et al.

    Mixed microalgae culture for ammonium removal in the absence of phosphorus: effect of phosphorus supplementation and process modeling

    Process Biochem.

    (2014)
  • A. Ruiz-Martínez et al.

    Effect of intracellular P content on phosphate removal in scenedesmus sp. experimental study and kinetic expression

    Bioresour. Technol.

    (2015)
  • G. Sin et al.

    Improving the Morris method for sensitivity analysis by scaling the elementary effects

    Comput. Aided Chem. Eng.

    (2009)
  • G. Sin et al.

    A critical comparison of systematic calibration protocols for activated sludge models: a SWOT analysis

    Water Res.

    (2005)
  • G. Sin et al.

    Global sensitivity analysis in wastewater treatment plant model applications: prioritizing sources of uncertainty

    Water Res.

    (2011)
  • A. Solimeno et al.

    Parameter sensitivity analysis of a mechanistic model to simulate microalgae growth

    Algal Res.

    (2016)
  • Cited by (7)

    • Microalgae-based wastewater treatment – Microalgae-bacteria consortia, multi-omics approaches and algal stress response

      2022, Science of the Total Environment
      Citation Excerpt :

      Menger-Krug et al. suggested a post-treatment of the effluent with activated carbon to remove unharvested biomass because harvest efficiency was set at 88 %, and the activated carbon treatment removes retained biomass components, nutrients, and residual pollutants via adsorption (Menger-Krug et al., 2012). Recent technologies such as computational modeling, artificial intelligence, and machine learning can be adapted to study the effluent characteristics, and assess the social and economic aspects of WWT (Aparicio et al., 2022; Ly et al., 2022). A search for specific niches where microalgae-based WWT is considered a better or the only available treatment option could open new opportunities for a decisive application.

    • Comprehensive evaluation of parameter importance and optimization based on the integrated sensitivity analysis system: A case study of the BTOP model in the upper Min River Basin, China

      2022, Journal of Hydrology
      Citation Excerpt :

      Unfortunately, it remains a challenging task due to the computational burden and the equifinality issue from calibrating many parameters simultaneously (Clark et al., 2011; Renard et al., 2010). A possible way to overcome these obstacles is to reduce parameter dimension through sensitivity analysis (SA) (Gan et al., 2014; Li et al., 2013), which is a practical parameter screening approach by identifying the model parameters that have a significant effect on the output variability (Aparicio et al., 2022; Pianosi et al., 2016; Saltelli et al., 2004). Benefiting from providing the total effect acquisition of input factors while maintaining computational efficiency, the global SA (GSA) based on the design of experiment (DoE), is more appropriate for qualitative and quantitative assessment of multi-parameter DHMs (Gan et al., 2014; Li et al., 2013; Liu et al., 2020; Xing et al., 2021).

    • Assessing and modeling nitrite inhibition in microalgae-bacteria consortia for wastewater treatment by means of photo-respirometric and chlorophyll fluorescence techniques

      2022, Science of the Total Environment
      Citation Excerpt :

      The inhibitory effect of NO2-N can be species-dependent. Microalgal model, to which NO2-N inhibition of microalgae growth has been added, was validated with different structures of microalgae communities (8% Chlorella and 92% Scenedesmus; 56% Chlorella and 44% Scenedesmus; 87% Chlorella and 13% Scenedesmus; and 96% Chlorella and 4% Scenedesmus) (Aparicio et al., 2022; Viruela et al., 2021). However, calibration procedure was performed with a community structure dominated by different Scenedesmus genera.

    View all citing articles on Scopus
    View full text