Operational optimization at the Nenäinniemi wastewater treatment plant's tertiary disc filter phosphorus removal installation to reduce chemical consumption Open Access

Municipal wastewater treatment plants (WWTPs) have the general driver of reducing operational costs while fulfilling effluent water quality requirements. Through optimization, processes become more efficient and consume fewer resources with the same or better level of treatment. Some optimization examples are the installation of more energy-efficient equipment or fine-tuning the control of the treatment processes at the plant. Usually, process optimization involves some form of adjustment of the control parameters. Adjustments can be carried out on the aeration regime to control the oxygen content in the activated sludge process or on the return sludge flow for sludge age control. Implementing advanced process control strategies has been a powerful and common approach to improve performances, such as dissolved oxygen (DO), nitrate recirculation, external carbon dosage and chemical dosing (Yuan et al. 2019) in wastewater treatment. Instrumentation control and automation (ICA) in the wastewater management sector has shown reductions in electricity usage by 10–30% (Olsson et al. 2005, 2014). The chemical usage for some applications was reduced by 25% with ICA (Ganigué et al. 2018).

There are examples where ICA has been implemented in municipal wastewater treatment in combination with coagulation and flocculation processes. Some WWTPs in Germany and Sweden have reported using orthophosphate (⁠–P) and pH analysers after coagulant addition or in the final effluent to control chemical dosing (Ingildsen 2002; Ratnaweera 2004). Mathematical models, neural networks or fuzzy logic have also been tested; however, flow-proportional dosing seems to be the dominant form of dosing control (Ratnaweera 2004; Nadeem et al. 2022). Feedback (FB) proportional–integral–derivative (PID) controllers have been used extensively in industrial processes for 100 years, but they are rarely used in wastewater treatment for coagulation dosing control. The main issue of using FB control seems to be the long and varying retention time in the separation stage (Olsson et al. 1973; Ratnaweera 2004; Ratnaweera & Fettig 2015). However, FB control of the iron dose in an activated sludge process with co-precipitation and tertiary settling has been proven to work to meet effluent requirements (Craig et al. 2014).

Feedforward (FF) and FB control, with turbidity sensors, has been used in chemically enhanced primary treatment with disc filtration to control the polymer dosing and operate at a consistent effluent turbidity set point (Väänänen et al. 2017). In principle, FB control seems possible in other applications with disc filters and chemical pre-treatment, such as tertiary treatment for phosphorus removal. The combination of online phosphorus analysers and turbidity sensors with FF and FB control loops should allow for lower chemical usage with a low-risk level. The FB controller allows for simple operation at the required effluent TP and turbidity concentrations. Effluent quality is continuously monitored and the dosing is adjusted automatically without the operator's involvement. The need to develop, calibrate and recalibrate coagulation/flocculation models is avoided. The additional use of a FF control loop can dampen the influences of disturbance variables (Liu et al. 2019) and can be implemented as a complement if needed.

Common activated sludge processes generally have a specific energy consumption of 300–600 Wh/m³ (Silva & Rosa 2015; Gu et al. 2017). The specific energy consumption for the tertiary disc filter applications with coagulation and flocculation is in the range of 30–50 Wh/m³. The dosing, rapid mixing and the coagulation and flocculation stirring stand for about 28% of the tertiary disc filter energy usage. The operation of the disc filter including the backwash (BW) and rotation is responsible for the remaining 72% (Kängsepp et al. 2016). In the long term, media fouling leads the filters to BW more often, which results in a significant increase in their energy consumption. To lower the disc filter energy usage in the tertiary treatment, fouling needs to be controlled. Fouling arises mainly from extracellular substances and inorganic compounds that build up on the surface and in the pores of the media (Iorhemen et al. 2017; Meng et al. 2017), reducing permeability over time. For disc filters, as well as for other filtration technologies, regular chemical media cleaning is needed to maintain the permeability over time. From a resource-saving perspective, conducting chemical cleaning is a trade-off between chemical costs and increasing operational costs. Effluent water quality can also be impacted by severe fouling. If filters are fouled, they can treat a lower water volume than originally designed for and untreated secondary effluent could bypass treatment and be discharged directly to the recipient.

The main objectives of this study were, first, to optimize the chemical dosing at the tertiary stage of the Nenäinniemi wastewater treatment plant (165,000 PE) by implementing online instrumentation for phosphorus and turbidity measurement and online FF and PI FB dosing control. Additionally, this paper investigates disc filter energy usage and proposes how to control long-term fouling and chemical cleaning in order to reduce even further the operational costs of the tertiary treatment stage.

The Nenäinniemi WWTP has quarterly effluent permits on TP, TSS, BOD₇, NH₄-N and COD_cr with no requirement for nitrogen removal. The treatment plant has the following environmental permit set by the Finnish Ministry of the Environment.

• BOD₇ < 10 mg/L and 96% reduction.

• Total phosphorus < 0.3 mg/L and 96% reduction.

• Ammonium nitrogen < 4 mg/L and 80% ammonium nitrogen reduction.

• TSS < 10 mg/L and 90% reduction.

• COD_Cr < 80 mg/L and 90% reduction.

• Pathogens 90% reduction.

Despite the limit prescribed on TP by law, the treatment plant management has set a production target for the final effluent TP concentrations to be around 0.1 mg TP/L in order to further reduce eutrophication potential in the receiving water body.

The wastewater is fed into the tertiary stage of the treatment plant through a common rapid mixing zone (3–5 s retention time) where the coagulant is injected. The flow is then divided into two treatment lines, each consisting of one coagulation and one flocculation basin with top-mounted stirrers. The polymer is injected into the turbulent overflow zone between the coagulation and flocculation basins. The average hydraulic retention time, during dry weather daily flow variations, was 2–4 min for the coagulation stage and 8–12 min for the flocculation stage.

Two Phosphax Sigma P-analysers (LPV341.99.10000, Hach, USA) and Sigmatax 2 sampling preparation units (LXV215, Hach, USA) were used to measure TP and ^—P in the tertiary plant influent (secondary clarifier effluent) and disc filter effluent. Solitax SC sensors (LXG 423.99.00100, Hach, USA) were used to measure disc filter influent and disc filter effluent turbidity and TSS (Figure 1). SC200 control units were used with the turbidity sensors (LXV404.99.00551) (Hach, USA). The flow to the tertiary treatment stage was measured prior to the coagulant injection point by the WWTP's main control system.

The PLC was a Siemens S7-400 (CPU 412-2 PN) series, and the communication was with Siemens WinCC (WinCC V7.4 + SP1, Siemens AG, Germany)).

In parallel with the online monitoring, daily composite sampling and analysis were performed according to Nenäinniemi WWTP ordinary sampling routines.

The Nenäinniemi disc filter installation is designed hydraulically with bypass weirs prior to the disc filters. If the water level prior to the disc filters increases above these weirs, part of the wastewater leaving the treatment plant will not undergo tertiary treatment. To comply with the effluent requirements, bypass is to be avoided.

The above flow calculation was scaled to a 4–20 mA output signal in the PLC and sent to the dosing pumps. The dosing pumps were calibrated for the corresponding flow signals according to standard routines for signal transfer and dosing pump calibration. The coagulant product has an active Al³⁺ content of 9.3% and a density of 1,360 g/L. The powder polymer used is assumed to be 100% active polymer and the stock solution concentration was 1.5 g/L.

PV = process variable

u (t) = PI controller output (0-1)

Initially the effluent TP set point was 0.18mg/L then lowered to 0.15 and finally to the targeted 0.1mg/L. The maximum total coagulant dose FF + FB was limited to 1.2 mgAl³⁺/L.

The applied FB polymer dose was determined by the Siemens software PI controller output (0–1) multiplied by the FB maximum polymer dose of 0.4mg/L. In this case, the controller gain (K_c) was set to −0.2 (%/%) (reverse action, fixed). The integration time constant (τ_I) was first set to 720–0.16*Qin (m³/h) and re-tuned from May 3 (2021) to −1.5 (%/%) and 6,800–0.63*Qin (m3/h). A variable integration time constant was applied to create an adaptive controller with gain scheduling to have an appropriate integration time constant (Leith & Leithead 2000) within the flow interval to the tertiary treatment stage.

By iteration, an appropriate set point value of effluent TSS or turbidity was identified. The goal was to find a TSS/turbidity set point value that matches the effluent TP set point. Initially when the effluent TP set point was 0.18mg/L an effluent that was more turbid and contained a larger fraction of particulate P was allowed. With a lower effluent TP set point less particulate P was allowed in the effluent and subsequently the effluent turbidity set point was set lower. The TSS set point was initially 3.7, then increased slightly to 4, then reduced to 3.8 and finally 2.7mg/L as online turbidity measurements were scaled to measure TSS. The FB turbidity sensor was replaced due to a noisy signal on July 1 2021. The new sensor was calibrated to measure turbidity and the set point was fixed at 1.7 NTU.

The maximum allowed total polymer dose (FF + FB) was initially limited to 0.6mg/L and after tuning the control parameters, the maximum allowed total polymer dose was lowered to 0.4mg/L.

The disc filter installation consists of six filters, each equipped with a BW pump. Each pump is rated at 22 kW and produces BW water for the rinsing of the accumulated particles on the media. The daily total BW pump operational time for each of the six units was logged with the daily flow to the tertiary treatment stage. The specific energy (Wh/m³) was calculated with the logged BW pump daily operational time multiplied by the rated power of the BW pump divided by the daily flow.

Chemical cleaning of the six disc filters was conducted with diluted hydrochloric acid (HCl, 7–8% v/v) and sodium hypochlorite (NaClO, 3–4% v/v). The cleaning was semiautomatic including a pump and a chemical cleaning spray bar placed inside the filters and a preprogrammed cleaning sequence in the filtering software. One CIP of the six-disc filters consumes 100 kg of hydrochloric acid (HCl, 30%) and 100 kg of sodium hypochlorite (NaClO 15%). The price for the cleaning chemicals was 0.33€/kg for HCl (30%) and 0.65€/kg NaClO (15%) (internal communication).

For the economical break-even point between chemical cleanings the electricity price was 100€/MWh (Nord Pool). The daily average flow of 38,000 m³ was used in the economy and energy calculations. The calculations were compared with the online data from 2020 and 2021 for a second estimate.

For the return on investment the cost for the turbidity and phosphorus instrumentation is 95,000€. The price for PACl coagulant was assumed 4€/kg Al³⁺ and for the polymer, 3€/kg polymer. The calculations were further based on the yearly treated flow (13.9 million m³) and the average arithmetic coagulant and polymer dose for 2020 and 2021.

The novel FF and PI FB dosing control system and energy usage were evaluated during approximately 1 year of operation. We compared online and laboratory data from two scenarios: one with manual control (2020) and the other with automatic control (2021). The comparison objective is to reveal the possible savings in the amount of coagulant and flocculant added for the removal of TP and TSS in a disc filter system.

In 2020, influent laboratory TSS data show two periods from mid-January to the end of March and one from the end of September to mid-November with increased TSS concentrations. The common phenomenon of bulking sludge under cold-activated sludge conditions could be one reason (Bakos et al. 2022; Safder et al. 2022). The rise in flows (Figure 4(c) and 4(d)) with increased surface loading on the secondary settling could be another reason that contributed to the increasing TSS concentrations. During these periods, the influent TSS to the tertiary treatment stage increased to approximately 60mg/L with some peaks up to 85mg/L. In 2021, two periods of upset conditions could also be observed: one at the end of March and one at the end of December. The influent TSS during these periods was between 30 and 50mg/L, lower than the peaks observed in 2020 (Figure 4(a) and (b)). Online turbidity measurements show a similar trend as the laboratory TSS measurements, which was considered important for control purposes.

For both 2020 and 2021, a consistent increase in flow could be observed around February–March due to snow melting. A minor increase in mid-November due to rainfall can also be observed together with some peaks during summer due to thunderstorms. The snow melting period in 2021 led to flows up to 70,000 m³/day, which resulted in some bypass of the disc filters. The bypassed water was mixed with treated effluent and it gave rise to unrepresentative measurements from a dosing control point of view. This issue was addressed on April 28th by reprogramming the dosing control algorithm to a fixed flow proportional dosing set point during bypass.

The arithmetic average coagulant dose for 2021 with the online dosing control system in operation was 0.3 mgAl³⁺/L and polymer 0.18mg/L. The arithmetic average effluent TP concentration was 0.11mg/L, and the effluent TSS was 3.7mg/L (laboratory composite samples). The online analysers measured 0.11 and 3.2mg/L TP and TSS, respectively (Figure 5(b)); thus laboratory and online results showed good agreement also in 2021. The average coagulant and polymer doses were 70 and 36% lower, respectively, than those doses in 2020. The relative dosing of Al³⁺ to the amount of removed TP was higher in 2020 than 2021 both as a yearly average and for the summer months of April–September. For the polymer, the relative amount of polymer dosed to the amount TSS removed was similar to the yearly average but was for April to September higher in 2020 than 2021.

The effluent water quality after the implementation of the proposed control algorithms improved by 21% regarding TP concentration and 34% regarding TSS concentration on a yearly average. The yearly savings in chemicals were estimated to be approximately 1.4 tons of polymer and 100 tons of coagulant product based on average daily flow calculations. In 2020 from February to December, 3.34 tons of polymer and 126 tons of coagulant were dosed. In 2021 from February to December, 2.22 tons of polymer and 42 tons of coagulant were dosed. Thus, the online data show a saving of 1.12 tons of polymer and 85 tons of coagulant for the same period February–December 2020 and 2021.

The required coagulant and polymer dose to achieve an effluent TP similar to the one in 2020 of 0.10–0.15mg/L was approximately 0.1–0.5mg/L (Al³⁺ or polymer). To operate at a TP concentration of 0.1mg/L in the disc filter effluent, the molar ratio of total dosed Al³⁺ to influent online –P ranged from 0.6 to 6 (Figure 5(c)).

In general, higher dosing of both coagulant and polymer was observed to be required during the early summer. From May to mid-July, the dosing was highest at approximately 0.35–0.4mg/L polymer and 0.5mg/L Al³⁺ (or 5–6 in online molar ratio). The lowest chemical dosing was generally observed in February and July, but periods ending in April and beginning in June with a dose of approximately 0.1 mgAl³⁺/L for the coagulant and 0.1mg/L polymer could also be observed (Figure 5(b)). After the end of August, the FF control loop was retuned allowing for some overdosing lest low chemical dosing occurred. To achieve overdosing, the controller was tuned to allow for a higher dose than necessary. Effluent water quality could be below the set point without controller adjustment of the applied doses.

The increase in chemical dosing from May to mid-July could, to some extent, be explained by the increased influent –P concentrations during summer. However, the lower dose in July and August with similar influent –P suggests there be other factors involved. Other parameters that are known to influence coagulation processes and dosing requirements such as alkalinity and pH (Henze et al. 2002) were measured between 1.5 and 2 mmolCaCO₃/L and 6.9–7.2, respectively (data not shown) throughout 2021. The wastewater temperature was rather consistent during the summer, 15–17 °C (data not shown). These variations in alkalinity, pH and temperature were minor and would not influence the coagulation process to such a large extent explaining the increase in the chemical dosing. Another factor that was suspected to influence the coagulation process was maintenance. The maintenance was done in August about the same time when the dosing requirement was lower. During maintenance, treatment lines are taken out of operation and the loading of the existing treatment lines is higher. This was suspected to influence the water matrix so that the dosing requirements changed. The plant observed a similar behaviour during the maintenance period in 2020. The effluent water quality worsened and, therefore, the coagulant dosing was increased to 1.8 mg Al³⁺/L (Figure 5(a)). The problem is very complex, as many factors influence coagulation-flocculation processes (Nadeem et al. 2022). Future investigations are required to find proof that there is a link between maintenance and dosing requirements.

As indicated previously, the purpose of the FF control loops was to set a base chemical dose, generally not enough to obtain effluent water quality targets. The FB control loop supplemented the base chemical dose with additional amounts of coagulant and polymer to achieve the desired effluent concentrations of TP or TSS.

The appropriate molar ratio for the FF coagulant control loop for the first tuning stage was 0.6 mol Al³⁺ dosed/influent ^—P concentration. The FF polymer control loop was set for the first strategy to add approximately 0.06mg/L polymer (Figure 6(b)). An offset issue between effluent turbidity and TP was observed during a 2-week period from July 25 onwards. This situation was new and had not been observed during the previous 5 months of operation. The offset was characterized by the effluent TP concentration increasing to approximately 0.2mg/L, while both influent and effluent turbidity remained very low, below the set point (<1.8 NTU), and this situation was of some concern. Even though the FB polymer control loop was active from time to time during this period lowering the effluent TP, it was considered necessary for operational and effluent permit reasons to deal with this offset. This type of offset between turbidity and effluent TP resulted in zero FF and FB polymer dosing and caused instability in the process. Coagulant dosing alone created flocs that were not retained on the woven media, resulting in worsened effluent water quality.

From the end of August 2021 onwards (Figure 6(a)), the FF coagulant control loop was tuned for increased process stability. The FF coagulant control loop molar ratio was increased to 2.5 mol Al³⁺ dosed/influent ^—P concentration. This resulted in a mostly inactive FB coagulant control loop. This situation was deliberately causing some overdosing but the results show that the process was stabilized. The increased dosing also solved two operational issues with clogged coagulant pipes due to low pipe flows and zero dosing during the night under low influent flow conditions that had been identified when the first dosing strategy was evaluated. These two operational issues were most likely contributing to some of the instability that was observed.

The FF polymer dosing control loop for the second stage was adjusted to increase the minimum dose from 0.06 to 0.15 mg polymer/L to increase the process stability (Figure 6(b)). With this setting, effluent turbidity was below the set point. The FB polymer control loop was activated periodically when effluent quality worsened and was above the set point (Figure 6(b)). The FB polymer control loop added approximately up to another 0.3 mg polymer/L in shorter periods to maintain effluent water quality at the turbidity set point.

To have a fully adaptive system and avoid overdosing, some auto-tuning of the FF control parameters can be implemented. One can, for example, add a correction factor on the applied FF dose based on the measured difference between measured effluent TP and effluent TP set point similar to a method described by Bertanza et al. (2021). Another method to improve the controller to avoid overdosing would be to rescale the FB controller range of variation output to also allow negative values. This would allow the FB control loop to reduce the influent coagulant and polymer concentrations imposed by the FF controller when it is excessive. Auto-tuning and optimization of the controller is a topic for future research.

The main conclusion is that with an active FB control loop (lower contribution of the FF control loop), the target effluent water quality could be achieved on average. Effluent water quality was oscillating around the set point for both effluent TP and TSS, minimizing the amount of chemical usage. With the more aggressive FF controller action, the effluent water quality was generally below the target TP and turbidity (Figure 6(a) and 6(b)). However, the result indicates that the process was more consistent. Despite some overdosing, a more active FF control resulted in an effluent water quality below the set point, while the coagulant and polymer consumption remained on average significantly lower (around 65% compared to 2020).

The required molar ratio on a daily average showed variation that, to some extent, indicated to be seasonal ranging between 2.5 and 20 mol Al³⁺/mol influent –P(Lab.) (0.6–6 if online –P) to achieve 0.1mg/L TP in the effluent. In the work presented by Langer & Schermann (2013) for disc filters, a molar ratio of around 15–30 mol Al³⁺/mol influent –P was used and Fundneider et al. (2020) reported molar ratios in the range of 5–20 to achieve similar 0.1mg/L effluent TP concentrations in advanced tertiary treatment and cloth-filtration. Bertanza et al. (2021) also reported large variations by a factor up to 10 in the required relative dosing between coagulant and influent –P to operate at a target effluent –P set point in a full-scale coagulation/flocculation and lamella sedimentation tertiary treatment stage. One can expect that the optimal relative dosing is variable and that some form of FB control strategy is needed to be able to optimize the chemical used in a tertiary treatment stage from the results presented in this paper the FB control strategy should preferably be based on real-time online measurements.

The results from this paper conclude that FF control strategies can be used to improve controller and process performance in disc filter installations for the tertiary P removal. For the highest possible savings in chemical usage, it is most advantageous to use a FF control loop that contributes less to the final chemical dose and allows the FB control loop to be active most of the time. This will result in effluent water quality oscillating around the set point with the average effluent concentrations complying with requirements and with minimal chemical usage. With a slightly more active FF control loop, the process stability can be improved. The challenge is to find the balance for the two controller loops by a few iterations and analyse the FB and FF controller actions step by step.

For a disc filter installation, the major energy-consuming component is the BW (Kängsepp et al. 2016). Reducing the operational time of BW pumps is an important contribution to lower operational costs (Nunes et al. 2013; Kängsepp et al. 2016).

The operational time of the BW in the short timeframe (minutes) is determined by the solid loading. This means that the operational time of the BW pumps increases with increasing flow, increasing coagulant dose and/or increasing influent TSS concentrations or a combination. In the longer timeframe though (days–weeks), the fouling of the media will become significant for the BW operational time. Therefore, an effective measure to reduce energy usage in a disc filter installation is to control media fouling. The fouling is caused by the buildup of inorganic scaling and organic deposits around and within the structure of the media and is of similar nature as membrane fouling that is extensively described in the literature.

The specific energy is lowest a few days after a chemical clean and is the result of chemical residues. After the residues from the cleaning chemicals are rinsed off completely, the specific energy usage is initially showing a slow increase. After about 15 days of operation on average, the specific energy increases at a significantly faster rate. The results indicate that long-term fouling generally follows a two-stage pattern (Figure 7(c)). In this study, the daily increase rate during stage one was approximately 0.1–0.4 Wh/m³/day from day 0 to approximately day 15 after a chemical cleaning. This corresponded to a total treated volume of 0.5–0.7 million m³ but could be up to 1.5 million m³ during summer. For stage two, from day 15 to approximately >25 days after a chemical cleaning, the rate in specific energy consumption increased to 3–6 Wh/m³/day corresponding to the treated total volume of about >0.7–1 million m³ or a bit more. A similar two-stage fouling pattern was observed previously in a study by Langer & Schermann (2013).

The breakeven point for a chemical cleaning to pay off was calculated to be approximately every 20 days (a bit longer over the summer periods). It was calculated that the additional cost per day for the increase in electricity consumption due to fouling was about 1€/day during the first 15 days and 15€/day from day 16 onwards. This interval can vary depending on electricity prices and the price for chemicals but the results show that initially, the fouling is slow with a minor increase in energy usage to rather rapidly change and that it is early in this stage of rapid increase it is beneficial to do the chemical cleaning from a cost perspective.

Potentially, energy usage can be further reduced. The disc filter installation at the site operates at a varying head loss of approximately 70–150 mm, this is governed by the control software. Head loss or transmembrane pressure influences the flux (Field et al. 1995), and the energy required to run the BW pump is head loss-independent. Operating at a higher head loss would increase the flux and lower the specific energy requirement. Currently, all filters are in use with the overlapping operation, they are also generally not over-loaded, operating at a relatively low average head loss. From an energy consumption perspective, it would be more beneficial to take some of the units out of operation and increase the load of the remaining units. This would increase the average head loss and lower the specific energy usage. However, in practice, this is not easily implemented.

The return on investment is approximately 2 years. The result also suggests that the dosing could have been lower from the end of August to mid-December, but for the discussed reasons, the FF controller settings did not allow for lower dosing. A lower dose would impact the investment calculations somewhat. However, the return on investment is short.

Chemical usage at the tertiary TP removal installation of the Nenäinniemi WWTP was lowered by around 85–100 tons of coagulant and 1.1–1.4 tons of polymer by using online turbidity sensors and phosphorus analysers and implementing simple FF and PI FB dosing control loops. Improvement in chemical usage related to the incoming TP and TSS was also observed. Despite the lower chemical dosing, the effluent water quality improved by 20–30% regarding TP and TSS concentration. Therefore, implementation of the required ICA to control the chemical dosing in a tertiary disc filter TP removal installation can render substantial savings in chemical use and improved effluent water quality with a payback time of 2 years for a medium-sized treatment plant.

To further improve energy usage, proper maintenance of the disc filter installation is of great importance. The optimal interval between chemical cleanings of the filter media from an energy and economical point of view was approximately 20 days long.

To have the lowest chemical consumption, the FF control loops for coagulant and polymer dosing should be tuned to dose low amounts of chemicals during normal operation and aim at having an active FB control loop most of the time. With the wastewater characteristics found in this study, an FF control loop dosing approximately 1 mol Al³⁺/per mol influent –P and 0.1mg/L polymer seems to be a good starting point. The FB PI control loops adjust the necessary dosing to maintain effluent water quality around the set point. Generally, it is also initially preferable to have a more feedback-oriented controller action to be able to tune the FF control loop from the operational data. A strategy with a more active FF control loop and minor overdosing can be a risk for operational stability with the lowest chemical use. Combining both FF and PI FB control loops is highly recommended.

This work is supported by the funding from the cosponsors Hydrotech Veolia Water Technologies AB, Jyväskylän Seudun Puhdistamo Oy and Insta Group Oy.

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.

Data cannot be made publicly available; readers should contact the corresponding author for details.

The authors declare there is no conflict.