Implementation of P-Reactive Layer for Improving Urban Water Quality: Kinetic Studies, Dimensioning and Economic Analysis

Abstract

Urbanization and climate change affecting water quality are the most critical problems that humanity has to encounter globally. Undoubtedly, urban water bodies are heavily contaminated by phosphorus (P). This study aims to identify the mechanisms and efficiency of the P sorption process for selected reactive materials (Autoclaved Aerated Concrete (AAC), Filtralite^® Nature P, lightweight expanded clay aggregate (Leca^®), limestone, opoka, and zeolite) with surface water as adsorbate and dimension of P-reactive reactive layer supported with economic analysis. Four kinetic models were used to know the sorption mechanism: pseudo-first order, pseudo-second order, Elovich, and intra-particle diffusion model. Calculating the P-reactive layer was based on dimensioning rain retention spaces standards. The pseudo-second model provided the best description of the adsorption kinetics of most materials. The sorption properties obtained after 72 h showed the reduction of 83, 81, 59, 53, 37, and 36% for AAC, opoka, Filtralite^® Nature P; limestone, Leca^®, and zeolite, respectively. Depending on the volume, the P-reactive layer can remove 29–77 or 61–163 g of P-PO₄. The unit cost of removing P-PO₄ by the P-reactive layer range from 49.57 to 85.53 €/P-PO₄ g. For these reasons, reactive materials seem to be an effective way of removing P from the urban water environment worldwide from both environmental and economic points of view.

1. Introduction

Urbanization and climate change are the crucial problems people face worldwide [1]. Moreover, it is predicted that the population in urban areas will increase to 67% by 2050 [2]. It is also widely estimated that the process of urbanization harms water resources that generally influence the degradation of the environment [1,3,4,5]. Climate change significantly affects water quality and quantity through higher temperatures and reduction of freshwater flow [3]. On the other hand, the rainfall becomes more variable and less frequent [6,7] and more contaminants get into surface and ground waters, intensifying degradation symptoms [3,8,9]. One of the contaminations which significantly affect the water quality, even in limited concentration, is phosphorus (P) [8,10]. From one side, it is an essential element of life support, part of DNA and RNA, ensuring plant growth and all biochemical functions of every living organism [8]. On the other hand, the surplus of P in the water environment leads to cyanobacterial harmful algal blooms and water eutrophication, a global problem affecting almost all water resources worldwide [8,11,12]. P is regarded as the main factor in managing eutrophication [13]. There is no precisely defined concentration of P above which the eutrophication process occurs. By European Council Directive 91/271/EEC [14], when the P concentration exceeds 0.1 mg/L, the river water is enriched with P. The exact value is stated by previous research suggesting that the P concentration above 0.1 mg/L promotes eutrophication [15,16,17,18]. Moreover, the United States Environmental Protection Agency proposed a lower value. A mean total phosphorus (TP) concentration should not exceed 10 µgP/L to keep lakes and reservoirs clear of eutrophication [19].

Urban areas are rich in P from atmospheric deposition, fertilization, animal wastes, soil erosion, or detergents [20]. The TP concentration in urban areas in stormwater runoff may range from 0.08 to even 17.9 mg/L [21,22,23]. The eutrophication impacts on fish deaths and reduction of biodiversity; toxins excretion that may seriously affect the health of fish, birds, and mammals; increased turbidity; and unpleasant odors, slimes, and foam formation, diminishing the aesthetic value of waters. Such consequences are well known, examined and proven [8,12,24]. However, in addition to environmental damage, eutrophication also causes significant economic losses [24]. Pretty et al. [25] estimated the annual damage costs of freshwater eutrophication between USD 105 to 160 million in England and Wales. The cost of the Baltic Sea eutrophication is valued at EUR 12–13 per inhabitant [26]. Considering that the sea basin is inhabited by around 85 million people [26], this gives a considerable amount of EUR 1.02–1.105 billion. In the U.S., the total annual cost of freshwater eutrophication was estimated at USD 2.2 billion in 2009 [27]. This cost also included economic losses of lakefront property values ranging from USD 0.3 to 2.8 billion per year and recreational use of the area amounting to USD 0.37–1.16 billion per year. For China lake Tai and its catchment, the losses caused by eutrophication and algal bloom were estimated to be USD 6.5 billion [28]. Due to excess P loading, the restoration cost of the grand lake St. Mary in Ohio is estimated to be USD 102 million. The excess P load in the lake has been estimated at USD 1561/kg P [29].

Solutions that counteract climate change in cities and mitigate the effects of urbanization are nature-based solutions NBS [1]. The main aim of implementing NBS is to support and be inspired by nature to provide environmental, social, and economic benefits that help build resilience [30]. Such solutions also play an essential role in delivering water services in urban areas, increasing water retention, micro climate regulation, rainwater drainage, surface runoff treatment, and recreational values [1,31].

Although NBS supports various benefits, P accumulation is not the primary treatment target. Some of the previously tested bio-/infiltration facilitates are effective in P removal. In the case of two grassed bioretentions located in North California, the average TP reductions were close to 60% for a year monitoring period [32]. A bioinfiltration rain garden in Pennsylvania removes the phosphate (P-PO₄) concentration from 0.21–0.25 mg/L in pond water to 0.03 mg/L in the pore water at the bottom of the infiltration bed [33]. On the other hand, Dietz and Clausen [34] observed no removal of TP concentrations between the inlet and the underdrains of two infiltrations rain gardens. Furthermore, the bioswale monitored by Jiang et al. [35] characterized by a P removal ratio, decreased even to −51%. In summarizing, in the case of infiltration facilities, the P-removal rate is variable, low, or even lacking.

A new approach to enhancing NBS in water remediation in urban areas is using reactive materials (RMs). The combination of NBS with RMs provides a more effective P removal in urban areas and valuable ecosystem services [7,20,35,36,37,38].

Removal of P from surface water or other liquids is challenging and generally obtained by physical, chemical, or biological methods [39]. Chemical precipitation and adsorption are used to reduce P concentration when employing RMs [11,39,40,41]. The materials based on metals oxides or hydroxides are the most successful and often used for P uptake [11,39]. Furthermore, the adsorption process is also regarded as one of the most economical P removal methods [24].

Previously, RMs were used as an additional filter to treat wastewater at an on-site wastewater treatment plant [42]. The range of RM applications has been spreading to surface water [43,44], agriculture runoff [45], or green roof leakage [46,47].

Some attempts have previously been made to apply RMs into NBS as a method to enhance P removal efficiency in urban areas. Some studies have been based on batch tests using artificial P solution that promotes P sorption [48] and laboratory-scale experiments [49,50]. There are some applications of implementation RMs in full-scale working systems, but they are still limited [35,37,51,52]. Moreover, to the best of my knowledge, there was no attempt to implement RMs in the form of a P-reactive layer in infiltration water reservoirs in urban areas.

This study aims to: (i) identify the mechanisms and efficiency of the P sorption process for selected reactive materials with surface water as an adsorbate; (ii) calculate the dimension of P-reactive infiltration layer supported with economic analysis.

The following research hypothesis was put forward in this work: RMs remove P from the water environment in urban areas.

The novelty of this study is in assessing the sorption properties of RMs in a batch test experiment with strongly polluted surface water imitating polluted urban water bodies. Most of the studies which have estimated the sorption properties of RMs has been conducted with synthetic P solution. The main aim of these studies is to check if the RM is attractive and efficient in P removal from aquatic solutions or not. Still, such theoretical values do not give a practical sorption capacity in the case of treated natural surface water. On the other hand, it has to be remembered that the cost of a pilot scale system implemented in field conditions is significant and, for that reason, it is better to perform laboratory tests before field scale application. This study is an attempt to fill the gap between batch test studies with an artificial P solution and pilot scale implementations.

2. Materials and Methods

2.1. Reactive Materials

In this study, six different P RMs were used in the tests: Autoclaved Aerated Concrete (AAC), Filtralite^® Nature P, lightweight expanded clay aggregate (Leca^®), limestone, opoka, and zeolite. AAC is a lightweight popular material used in civil engineering made of quartz sand, lime, or cement and water as a binding agent. Filtralite^® Nature P is high-quality filter media manufactured from expanded clay material heating above 1000 °C, used for filtration and purification with an active surface of dolomite addition (www.filtralite.com). The Leca^® is a lightweight construction material used in civil engineering, geoengineering, agriculture, horticulture, and environmental protection (www.leca.pl). Limestone is a sedimentary rock composed mainly of the skeletal fragments of marine organisms. Opoka is a calcium silicate sedimentary rock heated at a temperature of 900 °C. Zeolite is a hydrated aluminosilicate mineral that contains alkali and alkaline earth metals. The photographs of tested materials are presented in Figure 1.

The mineral composition and physical properties of RMs are set in

Table 1. Main chemical composition and physical properties of tester RMs.

Specification	Tested Reactive Materials
	AAC	Filtralite®Nature P	Leca®	Limestone	Opoka	Zeolite
Main chemical composition [%]
SiO2	57.24	63	66.57	0.61	55.11	47.8
Al2O3	1.96	17	16.57	-	5.65	6.07
CaO	24.62	2	2.46	29.3	23.86	2.84
SO3	1.35	-	-	-	-	-
Fe2O3	1.03	7	7.1	0.31	-	-
MgO	0.52	-	1.99	6.79	-	-
K2O	0.48	4	2.69	-	1.04	2.19
Fe	-	-	-	0.14	2.1	1.07
Main physical properties of tested RMs
Grain size [mm]	1–2	0.5–4.0 *	0–4 **	1–2	1–2	1–2
pH [-]	9.0	≈12.0 *	7.2	8.8	10.8	8.5
Porosity [%]	83.7	60.0 *	42.0	40.0	38.0	50.0
Bulk density [g/cm3]	0.30	0.37 *	0.51 **	1.50	0.73	0.90

* producent information; Filtralite^® Nature P is enriched with natural dolomite (https://filtralite.com/en/node/104, accessed on 20 May 2022). ** producent information; LECA^® (https://leca.pl/produkty/zaprawy-i-betony/lecar-keramzyt-0-4-r/, accessed on 20 May 2022).

. The physical properties were determined under the following standards: particle size distribution PN EN 933-1:2012 and PN-ISO 11277:2005, bulk density PN EN 1097-3:2000, porosity PN-EN 1936:2010.

2.2. Batch Kinetic Tests

Surface water from a small, low-land river was used for kinetic studies with various contact times ranging from 5 min to 4320 min (72 h). The 1.0 g of RM was mixed with 100 mL of surface water at a speed of 70 rpm at a stable temperature of 20 °C. The experiment was performed in three replications. The surface water was analyzed for P-PO₄ (mg/L), biological oxygen demand (BOD) (mg O₂/L), total solids (mg/L), pH, color [PtCo], electric conductivity (EC) (µS/cm), total suspended solids (TDS) [mg/L], and turbidity (NTU). The P-PO₄ concentrations were measured with the ammonium-molybdate method in a flow injection analysis using FIAstar 5000 analyzer in the range of 5–1000 µg P-PO₄·dm⁻³. BOD was measured by the respirometry method using the OxiTop kit. Total solids were measured by an indirect method by spectrophotometrically determining the unfiltered suspension in the range of 0–750 mg/L. The pH was measured with an Elmetron model CPR-411 pH meter. Color was measured spectrophotometrically in the range of 0–500 PtCo. The EC and TSD were measured with an Ultimar conductometer and turbidity with TurbiDirect. The characteristic of surface water used in this study is set in

Table 2. Qualitative characteristics of surface water used in kinetic tests.

P-PO4 [µg/dm3]	BOD [mg O2/dm3]	Total Solids [mg/dm3]	pH [-]	Color [PtCo]	EC [µS/cm]	TDS [mg/L]	Turbidity [NTU]
186.594	4.0	24.0	7.84	150	1036	518	10.5

.

The P-PO₄ concentrations in batch tests were measured the same way as surface water. All samples were filtered by a syringe filter of 0.45 µm pore size.

The sorption capacity (S) was calculated from the formula:

$$S=\frac{\left(C_0-C_e\right)·V}{M},$$

(1)

where, S—sorption capacity (mg/g); V is the volume of solution (dm³); M is the mass of material (g), and C₀ and C_e are the initial and final (equilibrium) concentrations (mg/L).

2.3. Statistical Analysis

The calculated P-PO₄ sorption of tested RMs and initial P-PO₄ loads were compared.

For this reason, the normality was assessed using the Shapiro–Wilk test. A lack of normal distribution for all tested materials causes the rejection of the zero hypothesis that all group averages are the same and the necessity to use non-parametric tests. The values were subjected to analysis using the Kruskal–Wallis test. As the test was statistically significant, multiple ranks mean comparisons were used for all trials, which is the post hoc equivalent of the nonparametric tests. All statistical analyses were carried out using STATISTICA 13.3 software by StatSoft (Hamburg, Germany).

2.4. Kinetic Models

To better understand the mechanism of adsorption kinetic, four different kinetic models were used: pseudo-first (PFO) and pseudo-second kinetic order (PSO) models, Elovich, and intra-particle diffusion (IPD) models.

The equalizations of the following models are presented below:

Pseudo-first order kinetic model proposed by Lagergren [51]:

$$log\left(q_e-q_t\right)=log\left(q_e\right)-\frac{k_1}{2.303},$$

(2)

where, q_t (mg/g) represents the amount of P adsorbed at any time t (min), k₁ (min⁻¹) is a constant rate of sorption of the pseudo-first kinetic model and q_e is the amount adsorbed at equilibrium (mg/g).

Pseudo-second order kinetic model also called the Ho model [52]:

$$\frac{t}{q_t}=\frac{1}{k_2{q}_e^2}+\frac{1}{q_e}t,$$

(3)

where, q_t (mg/g) represents the amount adsorbed at any time t (min), k₂ (g/mg·min) is a constant rate of sorption of the pseudo-second kinetic model and q_e is the amount adsorbed at equilibrium (mg/g).

Elovich model [53]:

$$q_t=\frac{1}{\beta }ln\left(\alpha \beta \right)+\frac{1}{\beta }ln\left(t\right),$$

(4)

where, q_t (mg/g) represents the amount adsorbed at any time t (min). α (mg/g·min) is the initial sorption rate, and parameter β (g/mg) is related to the extent of surface coverage and activation energy for chemisorption.

Intra-particle diffusion model [54]:

$$q_t=k_d{t}^{\frac{1}{2}}+C,$$

(5)

where, q_t (mg/g) represents the amount adsorbed at any time t (min), k_d (mg/g·min^0.5) is a constant rate of sorption of the intra-particle diffusion model. C is a boundary layer thickness. The higher the value, the greater the effect.

All constants were calculated based on slopes and intercepts of a linear formula of each tested kinetic model.

2.5. Reservoir Dimension

The dimension of infiltration reservoir was based on the ATV-A117 [55].

The calculation was made for part of the Warsaw University of Life Sciences campus (3.38 ha) in Warsaw, Poland. The percentage type of surface coverage of the part of campus and runoff coefficients are presented in

Table 3. Percentage type of surface coverage with runoff coefficient.

Type	Roofs	Parking Surface	Roads	Pedestrian Walks	Green Areas
Surface coverage, Fi [%]	12.29	14.98	5.92	4.02	62.80
Runoff coefficient, $\psi$[-]	0.8	0.7	0.7	0.5	0.1

. Based on such data, the average surface runoff coefficient ($\psi$_av) for the analyzed area was calculated:

$${\psi }_{av}=\frac{{{\displaystyle \sum }}_{i=1}^{i=n}{\psi }_i·F_i}{{{\displaystyle \sum }}_{i=1}^{i=n}{F}_i},$$

(6)

where, $\psi$_i is the runoff coefficient for a type of surface [-], F_i is the area of surface coverage type [m²], i = 1, 2, 3, …, and n is a number of surface coverage.

Based on the average surface runoff coefficient, the reduced area (F_red) was calculated:

$$F_{red}={\psi }_{av}·F,$$

(7)

where, $\psi$_av is the average surface runoff coefficient [-], and F is the total area [m²].

The stormwater outflow for the analyzed part of campus was calculated using the formula:

$$Q=F_{red}·q,$$

(8)

where, F_red is a reduced area [m²], q is a design rainfall [L/(s∙ha)].

The design rainfall was calculated according to empirical Formula (9) developed by Bogdanowicz and Stachy [56]

$$q=166.66\frac{h_{max}}{t},$$

(9)

where, h_max is the maximum rainfall amount [mm] calculated according to the Formula (10), and t is a rainfall duration time [min].

$$h_{max}=1.42·t^{0.33}+\alpha \left(R, t\right)·{(-\ln p)}^{0.584},$$

(10)

where, p is the probability of occurrence of excess rainfalls [%], α is the parameter dependent of a region of Poland (R), and duration time (t) that for the central region of Poland is defined as: $\alpha =4.693·ln\left(t+1\right)-1.249.$ The duration time t = 15 min and p = 50% were assumed for the purpose of calculations.

The design rainfall calculated using these assumptions was equal to q = 159.48 L/(s∙ha).

The efficiency of infiltration (Q_s) by the reservoir was calculated according to the formula [55]:

$$Q_s=0.5·k_f·\frac{A_{min}+A_{sides}}{2},$$

(11)

where, k_f is filtration coefficient [m/s], A_min and A_max are minimal and maximal areas through which the infiltration process occurs [m²]. The infiltration coefficient for the analyzing area is 4.7∙10⁻⁵ m/s [57].

Assuming the reservoir with a rectangular base area, the minimum infiltration area corresponds to the bottom of the reservoir. $A_{min}=a·b,$ is a minimal infiltration area [m²], A_sides is a sum of the sides of the reservoir [m²] defined as follows: $A_{sides}=2h·\sqrt{1+m^2}·\left(a+b+2mh\right),$where, a, b are dimensions of the bottom sides of the reservoir [m], h is the depth of the reservoir [m] and m is a slope of the sides [-].

Knowing the efficiency of infiltration and stormwater outflow, we are able to calculate the coefficient:

$$\mathsf{\eta }=\frac{Q_s}{Q},$$

(12)

For the assumed time t = 15 min, the coefficient $\mathsf{\eta }$ = 0.02, and the value of the rainfall duration for the reservoir BR is read from the ATV-A117 standard [55]. For $\mathsf{\eta }$ = 0.02 the BR = 1150 s.

The calculated volume of the reservoir (V_cal) is calculated according to the formula:

$$V_{cal}=BR·\frac{Q}{1000 } ,$$

(13)

where, BR is the value of rainfall duration for the reservoir read from the standard [s] [55], and Q is a stormwater outflow for the analyzed part of campus [L/s].

The real volume of the reservoir (V_real) is calculated according to the formula:

$$V_{real}=\frac{1}{2}·h\left[a·b+\left(a+2mh\right)·\left(b+2mh\right)\right],$$

(14)

where, a and b are dimensions of the bottom sides of the reservoir [m], h is the depth of the reservoir [m], and m is the slope of the sides [-]. The reservoir is appropriately calculated when the assumption $V_{cal}\le V_{real}$ is fulfilled.

The infiltration time T [s] during raining time t = 15 min, was calculated according to the formula:

$$T=\frac{Q·t}{Q_s},$$

(15)

where, q is a design rainfall [L/s], t is the rain duration time t = 15 min [s], and Q_s is the efficiency of infiltration [L/s].

2.6. Phosphorus Removal Efficiency in the Infiltration Time, T

The raw data of batch tests were recalculated into P-removal ratio R (%) according to the equation:

$$R=\frac{C_0-C_e}{C_e}·100\%,$$

(16)

where, C₀ and C_e are the initial and equilibrium P-PO₄ concentrations (mg/dm³).

Then, the obtained results of P-removal ratio vs. time, were approximated to the best described trend line by MS Excel. The P-removal ratio value for infiltration time, T, was read from the obtained curves.

3. Results and Discussion

3.1. Sorption of Kinetic Tests

The kinetic tests were performed with surface water to better imitate natural sorbing conditions. The main document regarding water in the EU is Directive 2000/60/EC [58], which gives, for example, descriptions of the different water quality states, the frequency of monitoring, and the obligations of the EU members to maintain good water quality status, but does not provide specific substances or concentrations corresponding to each water quality status. Each EU member has to determine them themselves. According to [59], the surface water (Table 2) is classified as below good condition due to its physicochemical parameters. The main exceedances concern P-PO₄ concentration, which should not exceed 0.101 mg/L for good status and TDS (≤404 mg/L). Because of BOD, the surface water was stated as a II class (good status). EC and pH were classified as a I class (very good status).

Removal of P-PO₄ onto tested materials was strongly affected by the contact time of the materials with surface water and increased with prolonging the time (Figure 2). The retention time was previously widely examined by kinetic studies and is regarded as one of the most influencing sorption agents [39,40,41] and this is also observed in recent research. The initial concentration of P-PO₄ in each sample was 186.59 µg P-PO₄/L. During the first 5 min of contact time, there is observed desorption of P-PO₄ in the case of all tested RMs. After this time, most of the materials began to sorb P-PO₄, except for zeolite (desorption lasted 15 min) and Leca^®, when the desorption process lasted for 30 min of contact time. Releasing P-PO₄ during the first minutes of RMs contact time with surface water may be caused by decreasing equilibrium concentrations of Ca²⁺, P-PO₄ species, and/or OH⁻ ions in solution, which leads to calcium phosphates dissolving [40]. Karczmarczyk et al. [60] tested 29 materials of natural or artificial origin for P-PO₄ release. P-PO₄ was detected in most of the extracts of materials. Most of them are popular building materials used in infiltration facilities (sand, gravel) or as P RMs (limestone, lightweight aggregates, crushed red brick). In field conditions, in the case of Ca-rich materials, the release of P is likely to occur when the solution pH decreases, for example, due to reduced delivery of hydroxide ions from the material (in aged filters) or during increased water flows [40]. For Fe/Al hydroxide, the P sorption is partly reversible when a P group replaces a hydroxyl group on the metal hydroxide surface (i.e., a monodentate complex) [40,61].

Concentrations obtained during the study were recalculated by Equation (1) to sorption capacity. After 72 h, the best unit sorption properties represented AAC (15.482 µg P-PO₄/g) and opoka (15.094 µg P-PO₄/g). In the case of other tested RMs, the sorption were: 10.941; 9.959; 6.995, and 6.974 µg P-PO₄/g for Filtralite^® Nature P; limestone, Leca^®, and zeolite, respectively. The sorption corresponded to the removal rate of P-PO₄ and equaled: 83; 81; 59; 53, 37, and 36% for AAC, opoka, Filtralite^® Nature P; limestone, Leca^®, and zeolite, respectively. Comparing these results with the maximum sorption capacity (S_max) calculated for tested RMs in previous studies [62,63] gave different results. S_max was calculated for RMs according to Langmuir isotherm with an artificial solution of P-PO₄ concentrations ranging from 1 to 1000 mg P-PO₄/L and a constant time of 24 h. The sorption capacity was obtained as follows: Filtralite^® Nature P (116.28 mg P-PO₄/g) > Opoka (100.26 mg P-PO₄/g) > AAC (71.27 mg P-PO₄/g) > zeolite (5.91 mg P-PO₄/g) > Leca^® (5.51 mg P-PO₄/g) > limestone (4.95 mg P-PO₄/g). The differences may arise from using a synthetic solution instead of surface water in previous studies and the modeling character of obtained data.

Using surface water instead of artificial P solution also affects the sorption. Within many different kinetic and isotherm studies, most of them are obtained with synthetic P solution as adsorbates [24,51,52,61,63,64,65,66,67,68,69]. Such studies are often characterized by better sorption results than those carried out with surface water or wastewater [43,62]. Using artificial P solutions for batch studies creates an ideal condition in which P is isolated from other compounds affecting adsorption [41]. The presence of such compounds, in surface water and wastewater, as nitrogen, P, heavy metals (Ba, Cr, Cu, Ni, Zn), organic matters (citric acid, humic acid, and oxalic acid), anions (sodium nitrate, sodium chloride, sodium carbonate, sodium sulfate), and pharmaceuticals etc., may decrease P adsorption process, and hence the sorption properties of RMs, or even lead to P desorption [64]. The sum of cations and anions in the water is defined as a TDS. The TDS concentration in the examined surface water (Table 2) was below the good state [59], indicating high water pollution. The influence of coexisting anions in water and their impact on reducing sorption capacity by competing with P-PO₄ for the active site was examined and proved [66,67,69,70]. Afridi et al. [64] and Kang et al. [67] examined the influence of P-PO₄ concentration, anions, and heavy metals. They noted that coexisting anions exhibited a notable inhibitory effect on the P removal efficiency. Furthermore, the presence of organic matter reduces P sorption, which Afridi et al. [64] studied. They demonstrate that organic matter significantly inhibits P-PO₄ removal from 32 to 50%. Even if the concentration of BOD in the surface water used for testing corresponds to good ecological status, the presence of organic matter reduced sorption in current research. Furthermore, the compounds present in surface water also reduce the value of K_II in the PSO model, confirmed in previous research [43,71] and the difference between sorption obtained in artificial P solution and surface water may even equal around 38–44% [43].

One of the most critical factors influencing the P sorption, which is also widely examined, is pH [11,41,44,63]. In the case of Ca-RMs, the high pH results from the release of Ca²⁺ that occurs in materials [41]. For most tested materials (AAC, Filtralite^® Nature P; limestone, opoka, and zeolite), the relationship between S_max obtained using the Langmuir isotherm model and the pH was previously tested [63]. The value of R² = 0.994 confirms the statement that P sorption is pH-dependent.

The P-PO₄ sorption also depends on the initial P-PO₄ concentration. In the current research, the initial concentration was lower than in previous studies. Castellar et al. [48] tested Crushed Autoclaved Aerated Concrete for different initial concentrations ranging from 5 to 25 mg/L and stated higher sorption with increasing P-PO₄ concentrations. On the other hand, Łożyńska et al. [44] tested lower concentrations ranging from 1 to 2 mg ${\mathrm{PO}}_4^{3-}$/L. They examined, among others, LECA and limestone with artificial P solution and noted an increased P-PO₄ removal of 14–46% and 11–40% for LECA and limestone, respectively. Deng and Wheatley [70] have suggested that at higher P concentrations, more processes than adsorption can occur, such as complexation and precipitation. The existence of the surface precipitation at higher concentrations and the dominant adsorption process in lower concentrations also were stated by Xu et al. [72]. Cucarella and Renman [41] also noted the same observation.

3.2. Statistic Anaysiss

A lack of normality was detected for the zeolite sorption result (p = 0.0053; p < α). The rest of the sorption results for tested RMs were characterized by p > α. For this reason, the nonparametric test was used.

The Kruskal–Wallis test showed statistically significant differences between the individual group medians (p = 0.00). The detected differences are set in

Table 4. Results of the Kruskal–Wallis test showing the differences between the initial P-PO₄ and removed loads of tested RMs Kruskal–Wallis test: H (6, n = 70) = 33.25802 p = 0.0000. * indicate statistical differences at p < 0.05.

Reactive Material	AAC	Filtralite® Nature P	Leca®	Limestone	Opoka	Zeolite	Initial Load
AAC		1.000000	0.468086	1.000000	1.000000	1.000000	0.043985 *
Filtralite® Nature P	1.000000		0.555554	1.000000	1.000000	1.000000	0.035188 *
Leca®	0.468086	0.555554		1.000000	0.773478	1.000000	0.000002 *
Limestone	1.000000	1.000000	1.000000		1.000000	1.000000	0.001679 *
Opoka	1.000000	1.000000	0.773478	1.000000		1.000000	0.022247 *
Zeolite	1.000000	1.000000	1.000000	1.000000	1.000000		0.000403 *
Initial load	0.043985 *	0.035188 *	0.000002 *	0.001679 *	0.022247 *	0.000403 *

. The sorption results were statistically different than the initial load of P-PO₄. This supports the hypothesis that RMs remove P from surface water.

3.3. Kinetic Models

In the kinetic studies of P-PO₄, four of the most popular kinetic models (pseudo-first order, pseudo-second-order, Elovich, and intra-particle diffusion model) were used to know the sorption mechanisms better. The linearized form of all model equations was used to determine the constants based on slopes and intercepts according to the Equations (2)–(5). The results of constants and coefficients of determinations are set in Table 2. In the case of the pseudo-first kinetic model, this one does not adequately describe the adsorption process that confirms R² ranged from 0.3388 to 0.7748. The research showed that the pseudo-second kinetic model is more suitable than the other tested models. Most of the tested materials (AAC, Filtralite^® Nature P, limestone, opoka, and zeolite) were characterized by good agreement to pseudo-second kinetic model with R² ranging from 0.9653 to 0.9999, which is widely used in liquid-phase adsorption systems and is worth of investigating because of feasibility in engineering applications [73]. Higher exhibited correlation values to the pseudo-second kinetic model suggested that P-PO₄ adsorption is in agreement with chemisorption, which is also confirmed by previous studies [66].

In the case of sorption properties of tested materials, these were also confirmed by q_e values representing an excellent fit to unit sorption of tested materials (R² = 0.54). Only in the case of Leca^® was the highest fitting observed for the Elovich model (R² = 0.9802) that described a chemical adsorption process and is suitable for systems with heterogenous adsorption surfaces [73]. Essentially, low fitting is observed in the case of the intra-particle diffusion model (R² ranged from 0.4659 to 0.9036), indicating the adsorption is not responsive to the diffusion mechanism [68].

3.4. Reservoir P-Reactive Layer Dimension and Economic Analysis

Assuming the reservoir dimensions of the bottom sides a = 10.5 m, b = 11.0 m and the depth of the reservoir h = 1.1 m with a slope of the sides m = 2 [-], the calculations according Equations (10)–(15) were made.

For a reservoir with a rectangular base area, the minimum infiltration area A_min = 115.50 m² and the maximum infiltration area A_sides = 127.41 m². The calculated volume of the reservoir V_cal = 183.40 m³ and the real volume of the reservoir V_real = 186.18 m³. Hence, the assumption $V_{cal}\le V_{real}$ was fulfilled.

The data of equilibrium concentrations (Figure 2) were recalculated by Equation (16) to the P-removal ratio. The results were fitted to the logarithmic function with R² ranging from 0.8096 to 0.9796 (Figure 3). The infiltration time T, calculated according to Equation (14), is around 14 h. Reading the time T from Figure 3, we are able to calculate the P-PO₄ removal by tested RMs during the infiltration time is: 55, 47, 22, 38, 54, and 35% for AAC, Filtralite^® Nature P, Leca^®, limestone, opoka and zeolite, respectively. These correspond to the decrease of P-PO₄ concentration to: AAC (84 µg/L) > opoka (86 µg/L) > Filtralite^® Nature P (99 µg/L) > limestone (116 µg/L) > zeolite (121 µg/L) >, Leca^® (146 µg/L). The limit concentration above which water is exposed to eutrophication equals 0.1 mg P-PO₄/L [14] or good ecological status (0.101 mg P-PO₄/L [59]) is completed only for AAC, opoka, and Filtralite^® Nature P.

In the infiltration reservoir, water infiltrates through the ground by the bottom and sides of the reservoir. The possible implementation of RMs as a filtration layer is on the bottom or bottom and sides of the reservoir. The P-reactive layer was compared to other RMs applications implemented at bio-/infiltrations facilities, such as bioswales and bioretention systems [35,37,74], infiltration rain gardens [35,75] or reactive layer at green roofs [46,47]. In the case of the bioswales, Jiang et al. [35] tested the possibility of reducing TP by a 0.2 m filtration layer made of slag, and 0.5 m of gravel. During two years of observation, they noted an average TP removal of 61 and 78% for bioswale A and B, respectively. Such a statement is also confirmed by Ekka et al. [37], who stated that bioswale is likely the best practice for P-PO₄ removal. The bioretention system with the addition of 1% (w/w) of Fe-Osorb improves the treatment performance of P by more than 40% compared to the standard bioretention system [76]. Weiss et al. [75] compared two types of rain garden with and without a reactive layer of sand mixed with iron shavings at 5% w/w and obtained the removal of over 90% of P-PO₄ from stormwater runoff in the case of the rain garden enhanced by reactive layer. For the green roof, a 0.02 m layer of Polonite^® underlying the substrate is enough for a P-PO₄ removal of 96% [46].

A completely different way of applying RMs to decrease P-PO₄ concentration in urban areas is presented by Waajen et al. [36]. They monitored two urban ponds after the implementation of powder fraction of lanthanum-modified bentonite and biomanipulation (LB) and flocculant polyaluminium chloride with biomanipulation (LPB). They observed that P-PO₄ reduction ranged from 10 to 68% for the LB variant, and from 31 to 70% for LPB, respectively. Moreover, they stated that using lanthanum-modified bentonite in both configurations gave comparable results to dredging in P-PO₄ reduction.

Liu et al. stated that P-PO₄ retention-release processes often occur at a shallow depth of infiltration facilities, even less than 0.1 m [74]. For that reason, the designed depth of the P-reactive layer is 0.1 m. The RMs with the highest sorption were selected for further analysis (AAC, opoka and Filtralite^® Nature P). Assuming the bottom area of the tank (115.5 m²) and the thickness of the P-reactive layer (0.1 m), the mass of the material is 3465, 4274, and 9009 kg of AAC, Filtralite Nature P, and opoka, respectively. Covering the sides of the tank (243 m²) with a P-reactive layer, the mass of the material will increase to 7287, 8988, and 18,947 kg (

Table 5. Results of kinetic analysis of P-PO₄ adsorption on tested RMs by adsorption models.

	AAC	Filtralite® Nature P	Leca®	Limestone	Opoka	Zeolite
Pseudo-first order model
KI (min−1)	1.7∙10−5	9.0∙10−5	170∙10−5	9.0∙10−5	17∙10−5	9.0∙10−5
qe (µg/g)	2.479	13.813	0.9340	15.669	1.956	1.949
R2 (-)	0.6081	0.5809	0.7495	0.7748	0.5185	0.3388
Pseudo-second order model
KII (g/µg min)	28∙10−5	88∙10−5	35∙10−5	54∙10−5	27∙10−5	12∙10−5
qe (µg/g)	15.221	10.427	7.199	9.990	15.015	0.083
R2 (-)	0.9653	0.9843	0.9277	0.9864	0.9780	0.9999
Elovich model
A	0.0010	0.0664	0.000001	0.0026	0.0013	0.0014
Β	0.43	0.64	0.60	0.63	0.44	0.71
R2	0.9103	0.8024	0.9802	0.8135	0.9740	0.8084
Intra-particle diffusion model
Kd (µg/g min0.5)	0.2298	0.1365	0.1627	0.1542	0.2290	0.1126
C (µg/g)	1.6988	3.059	2.1121	1.2680	1.5430	1.3406
R2 (-)	0.7932	0.5511	0.8509	0.8385	0.9036	0.4659

).

The infiltration reservoirs may also work in connection with water pre-treatment facilities, with a separate sedimentation zone planted with aquatic vegetation acting as biofilters periodically filled with water [77].

Besides the excavation volumes, the infiltration P-reactive layer is considered as the main cost factor for the infiltration facilities [37,78]. The cost of building reservoirs is around 215 EUR/m² [79]. The cost statement for the considered two variants (bottom and bottom with sides of reservoirs) of the P-reactive layer with the potential removal loads of P-PO₄ are set in

Table 6. Cost calculation for P-reactive infiltration layer.

Specification	AAC	Filtralite® Nature P	Opoka
Price, EUR/m3	215	232	332
Volume of P-reactive layer (bottom) = 23.1 m3
Cost of P-filtration layer, EUR	2489	2679	3833
Total mass of P-reactive layer, kg	3465	4274	9009
Potential removed mass of P-PO4 in infiltration time, g	29	42	77
Volume of P-reactive layer (bottom+sides) = 48.6 m3
Cost of P-filtration layer, EUR	5234	5634	8062
Total mass of P-reactive layer, kg	7287	8988	18,947
Potential removed mass of P-PO4 in infiltration time, g	61	89	163
Unite cost of P-PO4 removed in infiltration time, EUR/g	85.53	63.39	49.57

. The best environmental effective in P-PO₄ removal is opoka because of its high bulk density (about twice as high as other materials), on the contrary, it has a higher cost per m³ too. The high bulk density corresponded to the most increased mass of material necessary for constructing the P-reactive layer. However, recalculating the costs into the unit cost of P-PO₄ removal, opoka has the lower value (49.57 EUR/g). The least economically effective material is AAC (85.53 EUR/g), which is influenced by the low bulk density. Kumar et al. [24] calculated the cost of P-PO₄ adsorption using reusable porous metal oxide at USD 100–200/kg P for reducing the P-PO₄ to ultra-low concentrations. For a reservoir of an area of 115 m², the cost of the P-reactive layer covering the bottom ranged from 6 to 9% of the total cost. In the case of implementation of the layer both at the bottom and sides, the price will increase to 12–19%, depending on the material used. However, it should be mentioned that the costs of degradation of the water environment, eutrophication, loss of recreational values, or even the cost of flooding are always higher.

3.5. Limitations of Potential RMs Field Applications

It should be remembered that every interference and implementation of any amendments to the natural environment may pose certain risks. The same applies to the implementation of RMs. Both the acidic and alkaline pH of the leachate from RMs may have a harmful effect on the soil and occurring processes [11]. It should also be expected that in the case of most common construction materials, not only P-PO₄ is leached out, but also heavy metals (Cu, Ni, Cd, and Zn) [60] that are toxic to the environment. Ca- and Mg-based RMs seem to be a better choice for the environment than Fe and Al because of their lower toxicity [39]. In the case of RMs, P removal efficiency is strongly affected by initial P concentration, which was discussed before. At low concentrations, RMs may not be effective in P removing. Furthermore, the varying concentrations of different pollutants in treated water can decrease P sorption too. Moreover, in urban areas, the problem may be land availability, which is a common limitation. Design facilities are often land-intensive and not economically feasible in urban areas [49]. In the case of the conducted research, the reservoir covers 0.34% of the considered space, which seems to be an acceptable ratio.

4. Conclusions

On the basis of the conducted research, the sorption obtained during kinetic studies after 72 h of contact time follow the order AAC (15.482 µgP-PO₄/g) > opoka (15.094 µgP-PO₄/g) > Filtralite^® Nature P (10.941 µgP-PO₄/g) > limestone (9.959 µgP-PO₄/g) > Leca^® (6.995 µgP-PO₄/g) > zeolite (6.974 µgP-PO₄/g). Most of the tested RMs follow the pseudo-second kinetic model that suggests chemisorption may be a dominant P-PO₄ removal process. Only Leca^® follows the Elovich model. AAC, opoka and, Filtralite^® Nature P were chosen for the further calculations because of the best sorption properties.

The results confirm the hypothesis put forward in the paper that RMs remove P from the water environment in urban areas. It is also reflected in statistical analysis showing that the initial P-PO₄ load differs statistically significant from the loads removed by all tested materials.

The primary water services which fulfill the reservoir are the storage of around 180 m³ of water, infiltrate water during 14 h, and removal of 29–77 or 61–163 g of P-PO₄, depending on the volume of the layer. The unit cost of removing P-PO₄ by the P-reactive layer ranged from 49.57 EUR/P-PO₄ g for opoka to 85.53 EUR/P-PO₄ g for AAC, and is still lower than the environmental, social, and economic cost of eutrophication. However, it has to be remembered that laboratory-scale experiments, even based on natural surface water instead of an artificial solution, do not give entirely representative results. A pilot-scale investigation should precede the implementation.

On the other hand, urbanization and climate change leading to decreased water quality and eutrophication of water bodies are global problems. A better understanding of the relationship between these issues and associated with the economic impact is crucial for better management and investments of NSBs and sustainable development mitigating the effects of urbanization on climate change. Moreover, we should think globally and act locally to solve the issues because local actions can have tangible results.