Cuoiodepur WWTP treats vegetable tannery wastewater produced in the Tuscan leather-industry district of Santa Croce sull’Arno (PI). Municipal wastewaters from the surrounding area also are treated in the plant. The industrial stream is characterised by high organic carbon and nitrogen load, high salinity, sulphur compounds, biorefractory and nitrification inhibiting compounds, such as syntetic or natural tannins used in the production line. Municipal flowrate accounts for almost 40% of the total influent flowrate but its contribution in terms of carbon and nitrogen incoming load is less than 2%. In addition, the plant is subject to strong temporal variations in terms of industrial contaminant loads due to seasonal and market-related production fluctuations. Cuoiodepur water treatment line is presented in figure 1.
Ammonia oxidising bacteria (AOB) are reported to be very sensitive to salinity fluctuations and inbiting conditions in tannery wastewaters and low maximum growth rates have been observed (Moussa et al., 2006; Szpyrkowicz and Kaul, 2004; Munz et al., 2008). Salinity and inhibitory compounds, such as tannins are tipically ackowledged as the main disturbing agents for AOB activity. Thereby, in Cuoiodepur WWTP, the conventional activated sludge unit is operated at very high sludge retention time (SRT = 55 days, on average) in order to promote nitrification even in case of significant nitrifiers inhibition. High-SRT operation is also conducive for the degradation of the biorefractory organic fractions, thereby maximising the overall carbon removal. The most critical conditions for nitrification instability occur at industrial activity resume after long period of inactivity, i.e. typically after summer break or Easter and Christmas holidays. Indeed, significant nitrogen and inhibiting agents loads are faced by nitrifiers after weeks of moderate load conditions. After Christmas holidays, low temperature futher decrease activity of nitrifiers and potential nitrification instability could be faced.
In order to ensure efficient and stable nitrification, AOB response to such remarkable dynamic conditions should be capable of maintaining sufficient system nitrification capacity, even during fast transition phases. If such a minimum condition is not ensured, AOB washout and ammonia accumulation in the effluent might occurs. Thereby, it is primary interest of the industrial plant operators to be able to monitor nitrification stability and prevent events of ammonia discharge over the regulation limits.
Online Differential Titrimeter
The ODT is an innovative instrument for real-time estimation of the Ammonium Oxidation Rate (AOR) in WRRFs. The ODT consists of two identical 2-L jacketed CSTR, reactor 1 (R1) and reactor 2 (R2), equipped with pumping system for reagents and sludge dosing and sensors for DO, pH and T monitoring (Figure 2).
Operational conditions and data acquisition are controlled by a centralised PLC. Activated sludge is continuously withdrawn from the nitrification tank of Cuoiodepur WWTP, pumped in the two reactors and then returned back to the plant unit. The hydraulic retention time (HRT) is set to 1 hour. Withdrawing point of the mixed liquor sludge is located at the ending part of the nitrification tank, so that only residual ammonia and biodegradable organic matter are present.
Non-limiting DO concentrations (5±1 mgO2 L-1) in the two reactors are achieved by controlled dosage of hydrogen peroxide (0.65 M H2O2); temperature is controlled by continous recirculation of tempered water in the reactor’s jacket and set at the actual temperature observed in the plant’s nitrification tank by mean of a thermostat. Both reactors are continuously fed with ammonium solution at concentrations that ensure non-limiting and non-inhibiting conditions for the biological activity of nitrifiers (ca. 30 mgN_NH4+ L-1, in both reactors). Nitrification is inhibited in reactor 1 by continuously dosing Alliltiourea (ca. 30 mgATU L-1). The system PLC allows for continuous data acquisition on operational condition as well as solutions dosing (ammonia, hydrogen peroxide and acid/base), in the two reactors. Provided that operational conditions are identical in the two reactors and that constant pH is maintained, the differential dosage of the base solution is proportional to the acididity production due to nitrification only (occurring in R2 and not in R1) and it enables to calculate the AORMAX of the non-inhibited reactor (R2). Real-time AORMAX is straightfarwardly calculated by such a diffential dosage through system algoritms. AORMAX was monitored over a nine-month period, from April to December 2019. January, February and March 2019 data were not considered, since important modifications of ODT hardware and software were conducted and AORMAX data set was considered not reliable. AORMAX values were corrected to the temperature of 20°C through Arrhenius equation (θ=1.07), in order to easily compare outcomes over the year.
When needed, the ODT equipment can also be used in batch-mode, as an offline respirometer/titrimeter, as presented in the following sections.
A plant-wide model of the biological unit of Cuoiodepur WWTP was implemented in SUMO software (Dynamita, Canada). Thereby, the effluent of the primary sedimentation tank was considered as industrial influent in the model. The library model Sumo2S (Hauduc et al., 2017) was adopted in order to include processes related to the sulphur cycle. The model was partially modified in order to better suit with some peculiar characteristics observed in activated sludge processed applied to tannery wastewaters. The main modifications were based on the evidence that in WWTP operating at very high SRT (>50 days), as it is the case of Cuoiodepur WWTP, volatile suspended solids (VSS) and total suspended solids (TSS) concentration are often overestimated applying conventional ASM models (Lubello et al., 2009). This is due to the fact that complex particulate compounds, typically not reactive at normal SRT, are ultimately (though slowly) hydrolysed or degraded. Thereby, a low-rate conversion process was introduced for the conversion of the variable XU (unbiodegradable particulate COD, non-reactive by default) into the variable XB (slowly biodegradable particulate COD). Also, a low-rate hydolysis process was introduced for the hydrolysis of the variable XI (inert suspended solids, non-reactive by default) into the variable SI (inert dissolved solids). Both the new processes were modelled with first order kinetics, introducing the conversion/hydrolisis rates, Kconv_XU and Khyd_XI for the two processes, respectively (Lubello et al., 2009).
A new variable was introduced in the model to describe the rate of ammonium removal due to the nitrification only: AOR_AOB. Non-nitrifying biomass assimilation, N removal by anammox and precipitation processes were excluded.
YAOB is the growth yield factor for AOB [-]
iN is the specifc uptake for synthesis [gN gCODnew biomass-1]
XAOB is the concentration of AOB [gCOD m-3]
rAOB is the AOB growth rate [d-1]
rAOB is calculated by SUMO according to Monod-like saturation functions for substrates and AOB concentration.
The model layout of Cuoiodepur biological unit is presented in figure 3. The denitrification and oxidation units are simulated through 4 compartments in series in order to represent the plug-flow-like condition of the real plant. Wastage sludge was dynamically obtained, according to SRT set at 55 d. The secondary settler was considered as a reactive tank and simulated by a point-settler tank and an anoxic reactor, with a volume equal to 0.8 times the real settler volume, fed with settled sludge. A high recycle rate (10 times the Qind) was applied between point settler and reactive settler unit. Such a modification was supported by the evidence that a further aliquot of nitrate is consumed in Cuoiodepur sedimentation tank.
Five-year historical data, on influent and effluent quality as well as operational conditions, were processed and simulated in SUMO, on a monthly-average base. The reference period comprised the years 2015 to 2019. Prior to the dynamic simulation, yearly average values of the reference year 2015 were used as constant input for steady state calculation in order to define initial condition for state variables. Historical data produced during regular plant monitoring were provided by Cuoiodepur WWTP. The main analites of interest for modelling purpose were: COD, ammonia, nitrite, nitrate, total nitrogen, chloride, sulphide, hydrogen sulphide, volatile suspended solids (VSS) and total suspended solids (TSS). Beside municipal/industrial influent and secondary settler effluent, historical data were processed also for strategic plant units, such as primary sedimentation tank effluent and oxidation tank. COD and N fractionation of industrial and domestic influents were characterised according to chemico-physical analysis as well as respirometric techniques, on multiple 24-h composite samples collected over the year 2019. Specifically, the primary effluent was filtered at 0.45 and 0.1 µm in order to differentiate the particulate, soluble and colloidal fraction. Colloidal solids were considered as those retained at 0.1µ filtration, after 0.45µ filtration. Biodegradability of each fraction was then assessed through respirometric test in the ODT operated in batch-mode. The Respirometric tests were performed similarly to Munz et al., 2008. COD balance over the exogenous OUR returned the biodegradable share of the injected COD (non filtered, filtered at 0.45 µ and filtered at 0.1 µ). A similar procedure was adopted for the characterisation of the civil influent. According to personal communication with the plant process engineer, the main characteristics of industrial and municipal influents can be considered constant over the 5-year period of interest, since no significant variations in terms of industrial activity nor local population occurred. Thereby, influent fractionations were assumed constant throughout the years. Monthly average values of temperature and DO concentration in the aeration tank (terminal part) as well as influent, effluent and recirculation flowrates were also introduced as dynamic variables in the model. Specific measurement campaigns were performed in order to assess: (i) the actual flowrate of mixed liquor recirculation (QML), (ii) the nitrate load removed in the post denitrification unit and secondary tank, (iii) the DO concentration profile over the longitudinal axes of the nitrification tanks.
Chlorides are present at concentrations as high as 5-6 gCl-L-1 in the industrial influent (around 3 gCl-L-1, in the biological unit due to mixing with domestic influent) and are typically representative of the industrial activity, i.e. the higher the chlorides the higher the tanneries production. Chlorides were also selected as non-reactive compounds and their concentration was measured over the manhole entering the biological units as well as the primary sludge effluent, in the first measuring campaign. Points A, B and C as presented in figure 1, were monitored and chloride mass balance allowed for QML assessment. The second measuring campaign was performed over the post-treatment unit following the nitrification tank, comprising a post-denitrification and post-aeration tank, and the secondary tank effluent. Ammonia, nitrite and nitrate concentrations were measured and mass balance performed. Moreover, DO profile over the nitrificaton unit length were measured in the third measurement campaing.
ODT and Modelling Integration
Model calibration of the maximum growth rate of AOB (µAOB) was based on best fitting with the observed effluent ammonia concentration. Calibration was based on a trial and error iterative approach targeting least-square minimization. Oxygen half saturation coefficient for AOB was set to 0.5 mgO2 L-1 (Mannucci et al., 2020). First order hydrolisis/conversion constants for the new processes accounting for XU and XI reactions were set as follows: Kconv_XU = 0.014 d-1 and Khyd_XI = 0.016 d-1, in line with values reported in previous works (Lubello et al., 2009). Oxygen half saturation constant for the sulphur oxidizing bacteria was set to 0.1 mgO2 L-1 (Mora et al., 2016), instead of the default value of 1 mgO2 L-1. Kinetic and stoichiometric parameter referred to other biomasses were left at the default values.
Data set refferring to the period 2015-2018 was used for calibration; data set on 2019 was used for model validation instead.
AOR estimation through ODT were crossed with WRRF operational conditions and influent characteristics as well as model outcomes. Specifically, AOR profiles were correlated with: effluent chlorides (proportional to the industrial load), nitrogen load inlet, ammonium outlet and temperature. Results on AORMAX estimated from the ODT were critically compared to model outcomes.
The Maximun Nitrifying Capacity (MNC) expressed in tonN d-1 was estimated by multipling daily average values of AORMAX (at DO of 2 mgO2 L-1) and actual volumes of the oxidation tanks. The MNC profile were correlated with total nitrogen load inlet and ammonium outlet concentration.