MF–NF Treatment Train for Pig Manure: Nutrient Recovery and Reuse of Product Water

The livestock industry negatively impacts the environment by producing high organic and mineral loaded manure and wastewater. On the contrary, manure is also considered as the major focal point of resource recovery. The microfiltration (MF) process in manure treatment is well known for being the least complex and highly energy efficient. However, the major fraction of the dissolve nutrients easily bypasses the MF membranes. In this research work, we reported the efficiency of using MF–nanofiltration (NF) treatment train in a dead-end filtration system for the treatment of raw manure. The objectives were to produce nutrient rich separate streams in reduced volumes and a particle and pathogen-free product water. MF removed TSS above 98% and the COD and phosphorus (P) retention were noticed above 60 and 80%, respectively, within a reduced MF concentrate volume, which accounted for 40% of the initial feed volume. The NF of MF permeate by NF270 showed most promising results by concentrating overall 50 and 70% of the total nitrogen (TN) and potassium (K) within a reduced NF concentrate volume, which accounted for 30% of the initial MF feed volume. Finally, the MF–NF treatment train of raw pig manure could produce a particle-free product water that can be reused in farms to wash barns, to irrigate nearby cultures, or can be applied to specific fields based on the demand.


Introduction
The livestock industry negatively impacts the environment by producing high organic and mineral loaded manure and wastewater [1,2]. Manure contributes to environmental pollution by releasing ammonia and nitrous oxide into the atmosphere [3], by leaching nitrate mainly into ground water [4], and by increasing the soil acidification as well [5]. Excess nitrogen (N) and phosphorus (P) that are released due to manure application degrades the overall aquatic ecosystem [6]. This forced the European Community to implement nitrate directive guidelines to control the groundwater nitrate pollution [7].
On the contrary, manure is highly popular in agricultural applications for containing plant essential nutrients [8], in biopolymer production due to substantial volatile fatty acids concentrations [9], and most importantly in biogas production for energy recovery as well [10]. Therefore, the nutrient recovery techniques from manure are in high demand, despite the presence of high solid contents, organic materials, and its potential hazardous properties [11].
The previously mentioned nutrient recovery processes from manure, such as hydrogel application [12], calcium phosphate precipitation [13], ammonia stripping [14], and struvite precipitation [15] have proven to be very complex and required high chemical and energy inputs. Additionally, the drawbacks of using conventional mechanical processes to treat manure such as sedimentation, centrifugation, and pressurized filtrations have been well described previously [16]. Mechanical processes such as sedimentation and centrifugation

Membrane Characteristics
The raw manure samples were initially filtered by using 0.45 µm pore sized MF membranes to eliminate the suspended solids. The MF membrane characteristics are mentioned in a previous research study by Wei, Laborie, Aim, and Amy [22]. Three different loose NF membranes, NF270, HC50, and NTR7450 were further used for posttreatment of the MF permeate. The NF membrane characteristics are listed in Table 1.  [23] * sulfonated polyethersulfone (SPES); *" molecular weight cut off (MWCO). Manure samples were initially sieved through a 1 mm sieve to eliminate the particles (≥1 mm). The samples were then prefiltered in a dead-end stirred cell membrane filtration system, manufactured by Merck KGaA Germany (Supporting Information Figure S1), by using 0.45 µm pore sized MF membranes. The internal membrane diameter was 14 cm and the effective membrane area was calculated to be 154 cm 2 in the filtration cell. An initial feed volume of 500 mL was introduced in the feed tank for MF experiments. The filtrations were then performed by applying 1 bar pressure (N 2 gas, air liquid) and the rotational speed was maintained at 400 rpm. Consequently, 300 mL of permeate were collected in a sterile vial. The filtration was repeated twice for each manure sample to collect a total of 600 mL of MF permeate. The temperature was 25 • C ± 1 during the prefiltration experiments. The MF permeate samples were further analyzed for retention calculation.

Nanofiltration
An MF permeate volume of 200 mL each was used as the feed volume for the following NF experiments which were done using the same stirred cell dead-end filtration set up as mentioned in Section 2.3.1. Consequently, 100 mL of permeate were collected in a sterile vial after each NF experiment. The NF experiments were performed at 6.5 bar as the system could sustain maximum of 7 bar pressure. The rest of the filtration conditions were kept the same as the MF experiments. Similarly, pure water (Merck Millipore, Darmstadt, Germany) permeability (PWP) was measured at 6.5 bar pressure before and after each NF experiment. The NF permeate samples were further analyzed for retention calculation.

Calculated Parameters
The permeate flux (J) was determined by the ratio of the permeate flow rate (Q p ) to effective membrane area (A m ) [18]: The pure water permeability (PWP) was calculated by the ratio of the pure water flux (J w ) to the applied pressure (TMP): The retention calculation was performed by following equation [18]: where, R is the calculated retention in percent (%). C p and C f are the permeate and feed concentration of any parameter at a given recovery. The permeate volume recovery was calculated according to equation [18]: where, Rec is the calculated recovery in percent (%). V p and V f are the permeate and feed volume at a given time.

Chemical Characterization of Manure
Pig manure characteristics vary strongly depending on various parameters such as pig feed, manure storage conditions (site location, storage duration, temperature, etc.), and manure sampling methods [28]. Chemical characteristics of pig manure slurry of three investigated sites are given in Table 2. A standard deviation of maximum 5% within the measured values was observed. In general, TSS is the combination of suspended solids that can be degraded by microorganisms and non-degradable suspended solids. The TSS ranged within 3-5 gL −1 in all three manure samples. Similar results were shown in previous literatures for raw manure [29,30]. TSS is also considered as one main contributor of tCOD. This could be noticed in the close ratios between tCOD and TSS among the manure samples [31]. Relatively lower NH 4 + -N concentration was found in a sample of site 2. This could be attributed to various facts such as (i) conversion of ammonium to ammonia due to longer storage may lead to further valorization or evaporation [32] and (ii) different pig feed and growth stage as well [33]. The Tot-P concentration was lowest in the sample of site 2, followed by site 1 and site 3. Only 19% of the Tot-P was dissolved in the sample of site 2. The number was raised to 55 and 44% for the samples of sites 1 and 3, respectively. This reflects the dominant presence of P in the solid fraction of manure.
The DTN in pig manure is the sum of dissolved organic N, dissolved NH 4 + -N, and dissolved nitrate N, although nitrate is not typically present in manure [34]. However, parts of the organic N convert into NH 4 + -N, which then further converts into ammonia and contributes into total gradual loss of DTN [32]. This might lead to the lowest DTN value for sampling site 2 as the manure storage timing was the longest. Similar DTN values in pig manure samples were mentioned previously [35]. DOC is generally metabolized through volatile fatty acid intermediates [36]. Hence, the acetic acid and DOC concentrations were strongly correlated in manure samples. It was found that the organic carbon from acetic acid made up 44, 30, and 24% of the DOC in the manure samples of sites 1, 2, and 3, respectively.

Microfiltration of Manure
MF was performed to achieve the solid-liquid separation. Feed samples were collected before the start of each MF experiment. Permeate and concentrate samples were collected after 60% recovery was achieved. The recovery was calculated by following Equation (4). The samples were then analyzed to probe the TSS, Tot-P, COD, and NH 4 + -N retention by using Equation (3). The MF retention results are displayed in Figure 1.
The DTN in pig manure is the sum of dissolved organic N, dissolved NH4 + -N, and dissolved nitrate N, although nitrate is not typically present in manure [34]. However, parts of the organic N convert into NH4 + -N, which then further converts into ammonia and contributes into total gradual loss of DTN [32]. This might lead to the lowest DTN value for sampling site 2 as the manure storage timing was the longest. Similar DTN values in pig manure samples were mentioned previously [35]. DOC is generally metabolized through volatile fatty acid intermediates [36]. Hence, the acetic acid and DOC concentrations were strongly correlated in manure samples. It was found that the organic carbon from acetic acid made up 44, 30, and 24% of the DOC in the manure samples of sites 1, 2, and 3, respectively.

Microfiltration of Manure
MF was performed to achieve the solid-liquid separation. Feed samples were collected before the start of each MF experiment. Permeate and concentrate samples were collected after 60% recovery was achieved. The recovery was calculated by following Equation (4). The samples were then analyzed to probe the TSS, Tot-P, COD, and NH4 + -N retention by using Equation (3). The MF retention results are displayed in Figure 1. TSS retention of manure from all three sampling sites was above 98% (Figure 1). TSSfree MF permeate could visibly be observed as a transparent liquid compared to the MF TSS retention of manure from all three sampling sites was above 98% (Figure 1). TSSfree MF permeate could visibly be observed as a transparent liquid compared to the MF feed ( Figure 2). The Tot-P retention remained above 80%. Short filtration duration may lead to retained dissolved P as well. The COD retention of 80% was highest in site 3. In the other sites, the retention remained within 50-60%. Low NH 4 + -N retention was found in all sampling sites. The overall NH 4 + -N retention stayed within 15-20%. The efficiency of MF on TSS, colloids, and bacteria removal from mixed liquor and the effluents from biological reactors treating manure is well known [37,38]. Different studies also showed the complete removal of TSS from manure by using polymeric MF as well [39].
Higher retention of Tot-P could be attributed to its linkage with the particles between 0.45 and 10 µm in pig manure [40]. Consequently, Christensen et al. (2009), quantified that more than 70% of the Tot-P in pig manure slurry is associated with particles or colloids [41]. Therefore, it can be stated that the high TSS retention by MF enhanced the Tot-P retention as well. tCOD retention of manure by MF is closely associated with the retention of the particulate organic matter content [39,42]. Hence, the manure sample of site 1 with the lowest TSS content (Table 2) presented the lowest tCOD retention by MF. Similarly, the tCOD retention of site 3 manure, which contains 56% higher TSS than site 1, resulted in 30% higher tCOD retention by MF as well. Previous studies have mentioned the dominant (>88% of TN) presence of NH 4 + -N in the liquid fraction of pig manure [41,43]. Consequently, 5-10% mineralization of the organic N during manure storage also supports the above findings [44,45]. Hence, the 15-20% retention of NH 4 + -N in the present study can be associated with the mineralized fraction retention by MF. In addition, MF was Membranes 2022, 12, 165 6 of 12 also expected to remove antibiotic resistance bacteria and pathogens from raw manure as well [46,47]. feed ( Figure 2). The Tot-P retention remained above 80%. Short filtration duration may lead to retained dissolved P as well. The COD retention of 80% was highest in site 3. In the other sites, the retention remained within 50-60%. Low NH4 + -N retention was found in all sampling sites. The overall NH4 + -N retention stayed within 15-20%.
The efficiency of MF on TSS, colloids, and bacteria removal from mixed liquor and the effluents from biological reactors treating manure is well known [37,38]. Different studies also showed the complete removal of TSS from manure by using polymeric MF as well [39]. Higher retention of Tot-P could be attributed to its linkage with the particles between 0.45 and 10 μm in pig manure [40]. Consequently, Christensen et al. (2009), quantified that more than 70% of the Tot-P in pig manure slurry is associated with particles or colloids [41]. Therefore, it can be stated that the high TSS retention by MF enhanced the Tot-P retention as well. tCOD retention of manure by MF is closely associated with the retention of the particulate organic matter content [39,42]. Hence, the manure sample of site 1 with the lowest TSS content (Table 2) presented the lowest tCOD retention by MF. Similarly, the tCOD retention of site 3 manure, which contains 56% higher TSS than site 1, resulted in 30% higher tCOD retention by MF as well. Previous studies have mentioned the dominant (>88% of TN) presence of NH4 + -N in the liquid fraction of pig manure [41,43]. Consequently, 5-10% mineralization of the organic N during manure storage also supports the above findings [44,45]. Hence, the 15-20% retention of NH4 + -N in the present study can be associated with the mineralized fraction retention by MF. In addition, MF was also expected to remove antibiotic resistance bacteria and pathogens from raw manure as well [46,47].

Nanofiltration of MF Permeate
The objectives of performing the NF of MF permeate were to concentrate the dissolved nutrients (e.g., NH 4 + -N, K, and P) and to produce a permeate stream that can be reused. The permeate and concentrate samples were collected after 50% recovery was achieved. The recovery was calculated by following Equation (4). The samples were analyzed to calculate the retention by following Equation (3).

DTN and NH 4 + -N Retention
DTN and NH 4 + -N retention by NF membranes are shown in Figure 3. The DTN retention of site 1 manure remained within a range of 40-50% by all NF membranes. However, the retention went above 60% for the other two sites. Interestingly, the DTN feed concentration of site 1 was 15 and 30% higher than that of sites 2 and 3, respectively. The individual membranes presented similar retention trends in all sampling sites. The retention by NF270 was the highest, followed by HC50 and NTR7450 membranes, respectively.
The NH 4 + -N contributed approximately 60-70% to the DTN. The overall NH 4 + -N retention was nearly 10-20% lower than DTN retention in all sampling sites. The retention trend was mainly unchanged. The NH 4 + -N retention differences among the membranes for site 2 remained within 5% only. However, the differences between NH 4 + -N retention by NF270 and NTR7450 membranes was found as high as 30 and 15% in sites 1 and 3, respectively. This can be attributed to their respective surface charges and pore sizes as well.
Pig manure can be viewed as a mixed salt solution (Table 2). Therefore, the NH 4 + -N retention is assumed to be affected by the retention of the other competing ions as well (Supporting Information, Figure S2). A previous study showed that the increase in competing ions concentration in the feed can reduce the examined ion retention [48]. The NH 4 + -N retention by NF membranes is mainly influenced by charge interaction. It was already found that at pH7, NF270 retained 87.5% NH 4 + -N, when the filtration was performed under critical pressure. The study stressed the coupling of positively charged NH 4 + -N ions with the negatively charged active layer of the membrane [49]. Whereas, filtering dairy manure digested by NF270 resulted in 30-36% NH 4 + -N retention [11].
achieved. The recovery was calculated by following Equation (4). The samples were ana-lyzed to calculate the retention by following Equation (3). 3.3.1. DTN and NH4 + -N Retention DTN and NH4 + -N retention by NF membranes are shown in Figure 3. The DTN retention of site 1 manure remained within a range of 40-50% by all NF membranes. However, the retention went above 60% for the other two sites. Interestingly, the DTN feed concentration of site 1 was 15 and 30% higher than that of sites 2 and 3, respectively. The individual membranes presented similar retention trends in all sampling sites. The retention by NF270 was the highest, followed by HC50 and NTR7450 membranes, respectively. The NH4 + -N contributed approximately 60-70% to the DTN. The overall NH4 + -N retention was nearly 10-20% lower than DTN retention in all sampling sites. The retention trend was mainly unchanged. The NH4 + -N retention differences among the membranes for site 2 remained within 5% only. However, the differences between NH4 + -N retention by NF270 and NTR7450 membranes was found as high as 30 and 15% in sites 1 and 3, respectively. This can be attributed to their respective surface charges and pore sizes as well.
Pig manure can be viewed as a mixed salt solution (Table 2). Therefore, the NH4 + -N retention is assumed to be affected by the retention of the other competing ions as well (Supporting Information, Figure S2). A previous study showed that the increase in competing ions concentration in the feed can reduce the examined ion retention [48]. The NH4 + -N retention by NF membranes is mainly influenced by charge interaction. It was already found that at pH7, NF270 retained 87.5% NH4 + -N, when the filtration was performed under critical pressure. The study stressed the coupling of positively charged NH4 + -N ions with the negatively charged active layer of the membrane [49]. Whereas, filtering dairy manure digested by NF270 resulted in 30-36% NH4 + -N retention [11]. In addition, Hurtado and Cancino-Madariaga [50] also observed that higher NH 4 + -N feed concentration resulted in lower retention by NF membranes. It is assumed that the increased feed concentration enhanced NH 4 + -N flux through negatively charged NF membranes [51] due to the reduction in the Donnan effect and the neutralization of the membrane surface [52].

Phosphorus and Potassium Retention
The P retention by NF membranes is presented in Figure 4A. The substantial amount of P retention by MF resulted in a low P feed concentration for NF. The overall P retention was at or above 70%. In particular, NF270 retained above 90 and 80% of the remaining P from manure samples of sites 1 and 3, respectively. These were 10-20% higher than the retention by HC50 and NTR7450 membranes. Interestingly, no such retention differences between the NF membranes were noticed in the case of manure from site 2. Low feed P concentration and even lower permeate P concentration, led the analyzer to yield the limit of quantification as the permeate concentration in this case.
The potassium retention by NF membranes is displayed in Figure 4B. No real correlation between the initial feed potassium concentration and the retention was observed. The overall potassium retention was found to be within 22-54%. A similar retention trend was noticed in all manure samples. The highest retention was achieved by NF270. The retention by NTR7450 was 15-20% lower than NF270 and 5-10% lower than HC50 as well. High organic matter retention by NF (Supporting Information, Figure S3) could visibly be noticed ( Figure 5) as well.
Previous literature reported 96.4-97.2% P retention, when filtering dairy manure digestate using NF270. They stressed the charge repulsion effect as the main retention mechanism [53]. It is a well-known fact that the dissolved fraction of P in manure is mostly found in the orthophosphate (PO 4 −3 ) form [41]. Ballet, Hafiane, and Dhahbi [54] also reported 99% retention of the divalent (HPO 4 −2 ) form of phosphate by NF NF200 membrane. However, they carried out the experiments in a single salt solution condition, which might show very high retention [48]. Therefore, the higher retention of the trivalent form of dissolved phosphate by negatively charged NF membranes justifies the current findings. Previous literature has reported the proportional relation between the potassium chloride (KCl) feed concentration and the corresponding retention by NF membranes. They reported the range of KCl retention to be between 25 and 45% while filtering KCl solution of 5-15 gL −1 feed concentration, using an NF270 membrane [55]. Probably, due to the higher complexity and the presence of different ions and compounds of the feed manure, the trend was not clearly observed in the present study. Masse et al. (2007) noticed that, at higher recovery, the K retention was slightly decreased when filtering pretreated pig manure, with reverse osmosis membranes [17]. Lastly, it is proven that NF retains extracellular antibiotic resistance genes from manure and digestate above 99.99% [56,57]. This facilitated the pathogen-free product water production as well.

Phosphorus and Potassium Retention
The P retention by NF membranes is presented in Figure 4A. The substantial amount of P retention by MF resulted in a low P feed concentration for NF. The overall P retention was at or above 70%. In particular, NF270 retained above 90 and 80% of the remaining P from manure samples of sites 1 and 3, respectively. These were 10-20% higher than the retention by HC50 and NTR7450 membranes. Interestingly, no such retention differences between the NF membranes were noticed in the case of manure from site 2. Low feed P concentration and even lower permeate P concentration, led the analyzer to yield the limit of quantification as the permeate concentration in this case.
The potassium retention by NF membranes is displayed in Figure 4B. No real correlation between the initial feed potassium concentration and the retention was observed. The overall potassium retention was found to be within 22-54%. A similar retention trend was noticed in all manure samples. The highest retention was achieved by NF270. The retention by NTR7450 was 15-20% lower than NF270 and 5-10% lower than HC50 as well. High organic matter retention by NF (Supporting Information, Figure S3) could visibly be noticed ( Figure 5) as well.  Previous literature reported 96.4-97.2% P retention, when filtering dairy manure digestate using NF270. They stressed the charge repulsion effect as the main retention mechanism [53]. It is a well-known fact that the dissolved fraction of P in manure is mostly found in the orthophosphate (PO4 −3 ) form [41]. Ballet, Hafiane, and Dhahbi [54] also reported 99% retention of the divalent (HPO4 −2 ) form of phosphate by NF NF200 membrane. However, they carried out the experiments in a single salt solution condition, which might show very high retention [48]. Therefore, the higher retention of the trivalent form of dissolved phosphate by negatively charged NF membranes justifies the current findings. Previous literature has reported the proportional relation between the potassium chloride (KCl) feed concentration and the corresponding retention by NF membranes. They re-A B
Membranes 2022, 12, x FOR PEER REVIEW 10 of 13 smallest for the lower rejection NTR7450 membranes. It is also evident that the NTR7450 showed the highest fouling, followed by HC50 membranes.

Conclusions
The main benefits of the raw manure treatment by a cascade of MF and NF are to produce nutrient rich separate streams in reduced volumes. MF can separate the particulate material and NF can further concentrate the dissolve nutrients. Finally, a particle and pathogen-free product water is generated, which can be further reused in farms. Hence, the overall advantages of using a MF-NF treatment train for raw manure treatment is given below: i.
MF retained phosphorus above 80% within a smaller MF concentrate volume, which accounted for 40% of the initial feed volume. Additionally, the MF permeate contained above 80% of the total nitrogen and most of the dissolve potassium. ii.
NF of the MF permeate by three different NF membranes showed a maximum of 50-70% potassium and NH4 + -N retention, respectively, within smaller NF concentrate volumes, which accounted for 30% of the initial feed volume of MF. Among all of the NF membranes, NF270 showed the most promising retention and was found to be the least prone to fouling. iii.
Finally, the MF-NF treatment train was able to produce a particle-free final product water, which accounted for 30% of the initial feed volume of MF. This has the potential to be reused in farms to wash barns, to irrigate nearby cultures, or can be applied to specific fields based on the demand.

Supplementary Materials:
The following Supporting Information can be downloaded at: www.mdpi.com/xxx/s1, Figure S1: Stirred cell dead-end membrane filtration system; Figure S2: (A) COD and (B) DOC retention of MF permeate at 50% recovery by NF270, HC50, and NTR7450 membranes from all sampling sites. Pressure: 6.5 bar; stirring rate: 400 rpm; temperature: 25 °C; Figure  S3: Ion retention of MF permeate at 50% recovery by NF270, HC50, and NTR7450 membranes from all sampling sites. Pressure: 6.5 bar; stirring rate: 400 rpm; temperature: 25 °C; Figure S4: Flux during filtrations of NF270, HC50, and NTR7450 membranes while filtering pig manure from all sampling sites. Pressure: 6.5 bar; stirring rate: 400 rpm; temperature: 25 °C; Equation (S1): Normalized flux calculation.  The drop in normalized flux was mainly caused due to the combined effect of reversible and irreversible fouling but the PWP decline after the filtration experiments was associated principally with the irreversible fouling [58]. Since NF270 showed the least fouling, the low normalized fluxes in the experiments with these membranes are presumably associated with the concentration polarization (CP). Winter et al. (2017) noticed that higher CP played an important role for lower MWCO membranes, while filtering natural organic matter [59]. Hence, CP might largely affect the NF270 fluxes while the effect was smallest for the lower rejection NTR7450 membranes. It is also evident that the NTR7450 showed the highest fouling, followed by HC50 membranes.

Conclusions
The main benefits of the raw manure treatment by a cascade of MF and NF are to produce nutrient rich separate streams in reduced volumes. MF can separate the particulate material and NF can further concentrate the dissolve nutrients. Finally, a particle and pathogen-free product water is generated, which can be further reused in farms. Hence, the overall advantages of using a MF-NF treatment train for raw manure treatment is given below: i. MF retained phosphorus above 80% within a smaller MF concentrate volume, which accounted for 40% of the initial feed volume. Additionally, the MF permeate contained above 80% of the total nitrogen and most of the dissolve potassium. ii. NF of the MF permeate by three different NF membranes showed a maximum of 50-70% potassium and NH 4 + -N retention, respectively, within smaller NF concentrate volumes, which accounted for 30% of the initial feed volume of MF. Among all of the NF membranes, NF270 showed the most promising retention and was found to be the least prone to fouling. iii. Finally, the MF-NF treatment train was able to produce a particle-free final product water, which accounted for 30% of the initial feed volume of MF. This has the potential to be reused in farms to wash barns, to irrigate nearby cultures, or can be applied to specific fields based on the demand.  Figure S1: Stirred cell dead-end membrane filtration system; Figure S2: (A) COD and (B) DOC retention of MF permeate at 50% recovery by NF270, HC50, and NTR7450 membranes from all sampling sites. Pressure: 6.5 bar; stirring rate: 400 rpm; temperature: 25 • C; Figure S3: Ion retention of MF permeate at 50% recovery by NF270, HC50, and NTR7450 membranes from all sampling sites. Pressure: 6.5 bar; stirring rate: 400 rpm; temperature: 25 • C; Figure S4: Flux during filtrations of NF270, HC50, and NTR7450 membranes while filtering pig manure from all sampling sites. Pressure: 6.5 bar; stirring rate: 400 rpm; temperature: 25 • C; Equation (S1): Normalized flux calculation. Funding: This research was funded by the federal ministry of education and research (BMBF, Germany), in the framework of the project "KompaGG_N" (Grant number 02WQ1516D). We acknowledge support by the KIT-Publication Fund of the Karlsruhe Institute of Technology.

Data Availability Statement:
The data that support the findings of this study are available on request from the corresponding author.