The Legacy of Mercury Contamination from a Past Leather Manufacturer and Health Risk Assessment in an Urban Area (Pisa Municipality, Italy)

: An abandoned open green space in the urban setting of the Municipality of Pisa (Tuscany, Italy) has been designed for renewal to foster the development of recreational activities and improve the lives of the surrounding communities. However, the geochemical site characterization revealed Pb, Cu, Zn and Hg concentrations in the soil exceeding the thresholds imposed by Italian regulations for residential use. Pb, Cu and Zn contents likely reﬂect the effects of urban vehicle trafﬁc, while Hg contamination represents the legacy of a past artisanal tannery that used Hg(II)-chloride in leather processing in the mid-1900s. Mercury is widely distributed in the area, with the highest concentration in the uppermost soil layer, and reaching about 170 mg/kg in the common dandelion rhizosphere. Chemical extractions and thermal desorption experiments have indicated that most Hg is in the elemental free and matrix-bound fraction, with a possible minor amount (less than 4 wt%) of HgS and negligible methylated forms (0.1 wt%). The data suggest that soil processes could reduce Hg 2+ to volatile Hg 0 . Mercury in groundwater, hosted in a shallow aquitard in the area, was below 0.2 µ g/L. However, the presence of chloride in groundwater might result in the formation of Hg stable aqueous complexes, increasing Hg release from solids. Future water quality monitoring is hence recommended. The risk assessment highlighted that mercury in soil carries a risk of non-cancerous effects, in particular for children, posing the basis for management planning.


Introduction
Mercury is highly toxic for humans, ecosystems, and wildlife (e.g., [1][2][3][4]) and has been included among the ten chemicals of major health concern by the World Health Organization [5]. In particular, organic mercury is a potent neurotoxin [6] which accumulates in the biota and becomes biomagnified in the food chain [7,8], representing an important route for human exposure. Inorganic mercury is generally found in the environment in different oxidation states as elemental Hg 0 that can be converted to vapor at ambient temperature due to its low latent heat of evaporation: mercuric ion Hg 2+ and mercurous ion Hg 2 2+ . The latter is metastable and rarely investigated in natural settings [9], since it is expected to be rapidly disproportionate to elemental mercury and Hg 2+ complexes. The organometallic compounds of Hg 2 2+ are also scarce, whereas mercuric ions in water and soil may be readily converted to organic methylmercury (MeHg, CH 3 Hg + ) and re-methylated to form dimethylmercury (DMeHg, CH 3 HgCH 3 ) across a great diversity of bacterial species [10]. In particular, methylation under reducing conditions primarily occurs through the role of sulfate-reducing bacteria [11][12][13]. Indeed, methylation is a complex process [14][15][16][17][18][19][20], and methylation rates depend on a large number of physico-chemical and biological factors, including the total amount of Hg 2+ [21] and the occurrence of specific molecules (e.g., [22]). Demethylation may occur through abiotic and biotic processes [23], and the extent of MeHg in the ecosystem is the result of both methylation and demethylation processes. Hg 2+ in the water-soil system may form a variety of complexes and matrix-bound compounds, and it is mostly sorbed by natural organic matter [24][25][26][27], Mn-Fe oxy-hydroxides [28,29], and clays. Hg 2+ can be reduced by Fe 2+ -bearing minerals through surface reactions and by humic substances [30][31][32][33][34][35] yielding the formation of gaseous Hg 0 . Mercury vapor may also be sorbed by soil, and transformation to mercuric form may occur. Under favorable conditions, Hg 2+ reacts with sulfur sources leading to the formation of the sparingly soluble cinnabar (α-HgS) and metacinnabar (β-HgS). This process inhibits Hg 2+ methylation and immobilizes mercury in soils and sediments, even though some studies report that HgS in nanoparticles may be a source of MeHg [36]. Since mercury is not degraded by the ecosystem, many cases of present-day Hg contamination are a legacy from the past, adding further complexity in the assessment of contaminated sites. Major anthropogenic sources for mercury include chloralkali plants [37], cement production [38], iron and steel production [39], coal burning [40], mining [41][42][43], and oil refinement [44]. Mercury compounds have been widely used in the past as biocides and preservatives in a variety of applications, before their use became highly restricted or forbidden. In particular, mercury chlorophenol (C 6 H 5 ClHgO 3 ) and mercuric chloride (HgCl 2 ) were extensively applied in cereal and leather tanneries, leaving behind a large risk to the soil/aquifer system. Indeed, traditional tanning processes produced putrefied solid wastes and wastewaters with high concentrations of inorganic salts, heavy metals, and organics, although cleaner approaches to leather processing have been more recently developed [45]. In Italy, tanneries are generally characterized as small-medium size enterprises, often concentrated within municipalities, as in the tanning district of the Tuscany region [46].
Quantifying the total Hg content in soil is not sufficient for assessing the risk of an exposed population; it is necessary to determine the bioaccessible, soluble, and soil Hg vapor fractions. In the literature, several procedures have been proposed to quantify mercury fractionation and speciation [47], including thermal desorption and chemical sequential extraction procedures, even if inconsistent results have been obtained in some cases [48]. It is worth noting that consensual protocols for mercury speciation and certified materials for some Hg species are still lacking.
In the present study, the mercury environmental issues related to the legacy of a small bovine leather tanning manufacturer in an abandoned area within the Municipality of Pisa (Tuscany, Italy) have been investigated. In this context, the aim of this work was to evaluate the impact of Hg contamination in an urban area by determining concentration patterns, as well as the geochemical factors that contribute to mercury's fate. Specifically, the major objectives were: (i) to determine the degree of Hg distribution in soil and groundwater; (ii) to investigate the Hg speciation in soil using thermal desorption and sequential extraction methods; and (iii) to assess the health risk of exposure to mercury within a requalification project for the urban area.

Site Location and Geomorphological Outlines
The study area was located in the center of the Municipality of Pisa (Tuscany, Italy) ( Figure 1), approximately 0.5 km south of the Leaning Tower and Cathedral Place (the socalled "Square of Miracles"), a UNESCO World Heritage Site. The city of Pisa is located in an alluvial plain, a flat low-lying area with a mean elevation of about 2.5 m a.s.l., bounded by the Ligurian Sea to the west, the Serchio River to the north, the Mt. Pisani to the east, and crossed by the Arno River (Figure 1a). An abrupt change in elevation is recorded in the proximity of the study area; indeed, the area east of the medieval city walls has an elevation of about 5 m a.s.l., whereas the area to the west is located at about 2.5 m a.s.l. (Figure 1b). The reason for this change in elevation is not totally clear [49]. Since protohistoric times, this area has been characterized by a dense hydrographic network related to the combined fluvial activity of the Arno and Serchio rivers [50]. These highly sinuous and low-gradient rivers have frequently changed their course over time, as documented by several geomorphological studies. In particular, a north-south palaeochannel located between 9 and 13 m below ground level was identified nearby the study area [51]. During the last glacial-interglacial transition (MIS 2-MIS 1), the Arno River played a dominant role in the sedimentary and geomorphological evolution of the Pisa plain, forming a prominent incised valley, 40-45 m deep and 5-8 km 2 wide [52]. This palaeovalley was developed in response to the last phase of sea-level fall (culminating in the Last Glacial Maximum) and was progressively filled with estuarine and paludal deposits during the Late Glacial-Early Holocene transgression (about 13,000-8000 cal yr BP). The valley-fill succession is overlain by a laterally extensive, 3-15 m thick lagoonal unit composed of soft grey clays, accumulated approximately between 8000 and 5000 cal yr BP at the time of maximum marine ingression and during the early phases of sea-level highstand. Upwards is a 10-15 m thick deltaic alluvial sedimentary wedge originating from approximately 5000 yr BP. Eneolithic swamp clays, 1-3 m thick, form the lower portion of the late highstand fluvial succession and exhibit upward transition to poorly drained and then well-drained floodplain fine-grained deposits of protohistoric and historic age, respectively. Isolated to locally amalgamated fluvial-channel sand bodies, crevasse/levee sands and sand-silt alternations occur at different stratigraphic profiles. This progradational trend was temporarily interrupted by widespread backswamp development in Pisa's city center at the transition from the Iron age to the early Etruscan age. Stratigraphic profiles and hydrogeological context revealed the occurrence of a silty horizon, about 8 m thick, of alluvial origin with interbedded lenses of sandy deposits forming an aquitard. This succession is bordered below by a continuous sandy level representing the main semi-confined aquifer.
The study site consisted of an abandoned green space (about 1500 m 2 ) adjacent to the medieval defensive walls surrounding the city of Pisa, and was being included in an urban renewal project to become an important public playground and recreational site for inhabitants and tourists. A reconstructed history of the activities on the site revealed that a slaughterhouse operating from the middle of the 19th century to the 1960s was located Since protohistoric times, this area has been characterized by a dense hydrographic network related to the combined fluvial activity of the Arno and Serchio rivers [50]. These highly sinuous and low-gradient rivers have frequently changed their course over time, as documented by several geomorphological studies. In particular, a north-south palaeochannel located between 9 and 13 m below ground level was identified nearby the study area [51]. During the last glacial-interglacial transition (MIS 2-MIS 1), the Arno River played a dominant role in the sedimentary and geomorphological evolution of the Pisa plain, forming a prominent incised valley, 40-45 m deep and 5-8 km 2 wide [52]. This palaeovalley was developed in response to the last phase of sea-level fall (culminating in the Last Glacial Maximum) and was progressively filled with estuarine and paludal deposits during the Late Glacial-Early Holocene transgression (about 13,000-8000 cal yr BP). The valley-fill succession is overlain by a laterally extensive, 3-15 m thick lagoonal unit composed of soft grey clays, accumulated approximately between 8000 and 5000 cal yr BP at the time of maximum marine ingression and during the early phases of sea-level highstand. Upwards is a 10-15 m thick deltaic alluvial sedimentary wedge originating from approximately 5000 yr BP. Eneolithic swamp clays, 1-3 m thick, form the lower portion of the late highstand fluvial succession and exhibit upward transition to poorly drained and then well-drained floodplain fine-grained deposits of protohistoric and historic age, respectively. Isolated to locally amalgamated fluvial-channel sand bodies, crevasse/levee sands and sand-silt alternations occur at different stratigraphic profiles. This progradational trend was temporarily interrupted by widespread backswamp development in Pisa's city center at the transition from the Iron age to the early Etruscan age. Stratigraphic profiles and hydrogeological context revealed the occurrence of a silty horizon, about 8 m thick, of alluvial origin with interbedded lenses of sandy deposits forming an aquitard. This succession is bordered below by a continuous sandy level representing the main semi-confined aquifer.
The study site consisted of an abandoned green space (about 1500 m 2 ) adjacent to the medieval defensive walls surrounding the city of Pisa, and was being included in an urban renewal project to become an important public playground and recreational site for inhabitants and tourists. A reconstructed history of the activities on the site revealed that a slaughterhouse operating from the middle of the 19th century to the 1960s was located about 0.1 km south of the study area. Bovine hides were transferred from the slaughterhouse to a small artisanal tannery, located in the area under study, to be processed into leather. The tannery plants were built without any treatment facilities and environment management systems; as a result, hazardous metals including mercury (in the form of HgCl 2 powder) were discharged directly into the soil. After the tanning activity ceased, the derelict infrastructures were removed to improve the urban landscape without performing any appropriate environmental reclamation work.

Sampling
In order to assess the current state of mercury pollution, soil, groundwater and plant environmental media were considered. An illustration of the spatial distribution of sampling stations is given in Figure 2. Two boreholes (PZ1 and PZ2) were drilled in July 2019 by non-destructive techniques, reaching the depth of 9 m below ground level. The extruded cores were collected and carefully stored in plastic core boxes. The cores were examined and cut into sections using a stainless-steel cutter; the outer layer in contact with the core tube was removed to eliminate possible contaminations. Two samples were taken from each drilling at depths between 0.1 m and 1.0 m (samples PZ1-A and PZ2-A), and between 1.2 m and 1.6 m (samples PZ1-B and PZ2-B) in accordance with Italian regulations for risk analysis. Each sample was mixed thoroughly to give a 1 kg composite. An additional borehole (PZ3) was drilled, reaching the depth of 3.0 m below ground level in order to investigate the soil and subsoil layer sequence in more detail. In PZ3 samples were collected from the core at nine different depths from ground level: PZ3-1 (0. . Soil was also collected throughout a trench (2 m wide, 1 m deep) excavated by a ditching machine. A 1 kg composite sample was collected along the full length of the exposed face of the trench, obtaining a composite sample (sample T1-A, 0.1-0.5 m), and one sample was collected from the bottom (sample T1-B, 1 m). During sampling, larger (>20 mm) fragments were removed in the field. Spatially distributed rhizosphere soils surrounding naturally growing dandelion (Taraxacum officinale) plant roots were collected in November 2018 (samples R1 to R12, Figure 2). In order to study the Hg migration from soil to roots and leaves, and to evaluate the exposure to gaseous Hg, the roots and leaves of dandelion were collected in correspondence with rhizosphere stations. One plant of dandelion taken from outside the contaminated area was transferred in a vase in order to prevent Hg uptake from soil and left in the site nearby station R4 from April 14 to May 27, 2019.
All soil samples were air-dried for 24 h to reduce Hg 0 losses [53]. Soils from boreholes were sieved to 2 mm in order to compare contaminant concentration against guideline values.
Dandelion roots were cleaned to carefully remove soil particles; roots and leaves were washed with Milli-Q water, dried at ambient temperature and ground in a mortar.
Boreholes PZ1 and PZ2 were equipped with PVC piezometers with a screen interval between 3 and 6 m below ground level in order to collect groundwater from the aquitard, as the water body most prone to contamination. Groundwater was collected from the PZ1 and PZ2 piezometers in two surveys in July 2019 (groundwater low-stage condition, samples PZ1_W1 and PZ2_W1) and January 2020 (groundwater high-stage condition, samples PZ1_W2 and PZ2_W2) using a low-flow (0.5 L/min) plastic 12 V pump, after purging. Water samples were filtered in the field through 0.45 µm nylon filters and refrigerated in pre-cleaned polyethylene bottles. Ultrapure nitric acid was used as a preservative for major cation and trace element analysis. All soil samples were air-dried for 24 h to reduce Hg 0 losses [53]. Soils from boreholes were sieved to 2 mm in order to compare contaminant concentration against guideline values.
Dandelion roots were cleaned to carefully remove soil particles; roots and leaves were washed with Milli-Q water, dried at ambient temperature and ground in a mortar.
Boreholes PZ1 and PZ2 were equipped with PVC piezometers with a screen interval between 3 and 6 m below ground level in order to collect groundwater from the aquitard, as the water body most prone to contamination. Groundwater was collected from the PZ1 and PZ2 piezometers in two surveys in July 2019 (groundwater low-stage condition, samples PZ1_W1 and PZ2_W1) and January 2020 (groundwater high-stage condition, samples PZ1_W2 and PZ2_W2) using a low-flow (0.5 L/min) plastic 12 V pump, after purging. Water samples were filtered in the field through 0.45 µm nylon filters and refrigerated in pre-cleaned polyethylene bottles. Ultrapure nitric acid was used as a preservative for major cation and trace element analysis.

Analytical Procedures
Soil samples from boreholes PZ1 and PZ2 and trench T1 were digested by using a Milestone Ethos Easy microwave platform (US EPA Method 3051A). The concentration of a set of elements was determined by ICP-MS using a Perkin Elmer NexION 300X. The analytical uncertainty was evaluated by the analysis of the soil reference material NIST SRM 2711a. In general, the accuracy was better than 10%. Precision values, as relative standard deviation, were better than 5% for Li, Be, Mn, Ni, Ag, Sn, Cd, Tl, Pb, Fe and As, and within 10% for Co, Cu, Zn, Sr, Sb, Ba, Th, U, V, and Cr. Hexavalent Cr was measured following the US EPA Method 3060A. Total organic carbon (TOC) was determined according to EN 15936:2012. Additionally, soil pH was measured [54].
In order to quantify the Hg species in PZ1-A and PZ2-A soils, mercury speciation was investigated by sequential chemical extraction according to [55]; the average value for Hg recovery was 86 ± 5%. Methyl-Hg analyses were performed using the extraction

Analytical Procedures
Soil samples from boreholes PZ1 and PZ2 and trench T1 were digested by using a Milestone Ethos Easy microwave platform (US EPA Method 3051A). The concentration of a set of elements was determined by ICP-MS using a Perkin Elmer NexION 300X. The analytical uncertainty was evaluated by the analysis of the soil reference material NIST SRM 2711a. In general, the accuracy was better than 10%. Precision values, as relative standard deviation, were better than 5% for Li, Be, Mn, Ni, Ag, Sn, Cd, Tl, Pb, Fe and As, and within 10% for Co, Cu, Zn, Sr, Sb, Ba, Th, U, V, and Cr. Hexavalent Cr was measured following the US EPA Method 3060A. Total organic carbon (TOC) was determined according to EN 15936:2012. Additionally, soil pH was measured [54].
In order to quantify the Hg species in PZ1-A and PZ2-A soils, mercury speciation was investigated by sequential chemical extraction according to [55]; the average value for Hg recovery was 86 ± 5%. Methyl-Hg analyses were performed using the extraction scheme proposed by [56] and approved by the Tuscany Region Environmental Protection Agency. The elemental mercury fraction in soil was estimated by thermal desorption techniques: continuous heating experiments from ambient temperature to 250 • C were performed by weighting about 50 mg of soil samples into a 250 mL borosilicate glass flask in a heating furnace. Temperature was digitally controlled to ±1 • C; the applied heating rate ranged from 8 to 35 • C/min. Hg-release curves during heating were recorded by measuring the vapor emission by online Lumex RA-915M, using air flow as a carrier.
On-site measurements were conducted to determine the Hg gas release of the soil using a probe assembly inserted at a depth of 50 cm into the ground nearby the PZ1 station. Mercury soil gas was extracted by connecting a syringe to the probe and collected (after purging) through Carulite sorbent traps and analysed by Milestone DMA-80.

Risk Assessment
In the present study, the mercury risk assessment was performed using a site-specific approach [57,58] and the Risk-net software (version 3.1.1, September 2019, available at http://www.reconnet.net/Software.htm (accessed on 21 February 2022)). The exposure routes considered were shallow soil ingestion, dermal contact, particulate inhalation, and outdoor vapour. Human receptors were identified for both adults and children.
The Hazard Quozient (HQ) for non-carcinogenic chronic effects in humans was calculated for each exposure pathway. In case of direct ingestion and dermal contact, HQ (i.e., HQ ingestion and HQ dermal ) was calculated by dividing the chronic daily intake (CDI, mg/kg/day) by the corresponding reference dose (RfD, mg/kg/day) [59][60][61], defined as the maximum daily exposure to a toxic agent that would not produce any appreciable deleterious effects on human health: where CDI represents the exposure to a toxic agent, averaged over a long period of time, through ingestion (CDI ing ) or dermal contact (CDI derm ): where C POE is the exposure point concentration of mercury in soil (mg/kg); EF is the exposure frequency (150 day/year, a value suitable for a non-residential use). For the remaining parameters, recommend values [62][63][64] were used: R ing is the ingestion rate (100 mg/day for adult, 200 mg/day for children); ED is the exposure duration (24 years for adults, 6 years for children); SA is the exposed skin area (5700 cm 2 for adults, 2800 cm 2 for children); SAF is the skin adherence factor (0.07 mg/cm 2 day for adults, 0.2 mg/cm 2 day for children); ABS is the dermal absorption factor (0.01 unitless); BW is the average body weight (70 kg for adults,15 kg for children); AT is the average time of exposure to non-carcinogens (ED × 365 day/year). In case of inhalation pathways (vapor and dust), HQ (i.e., HQ inhalation ) was calculated by dividing the exposure concentration (EC, mg/m 3 ) to the reference toxicity concentration value (RfC, mg/m 3 ), rather than on the inhalation rate and average body weight: where RfC was set to 3.0 × 10 −4 mg/m 3 [61] and EC was estimated as starting from the predicted concentration in air (mg/m 3 ), according to: ET is the exposure time (3 h/day, suitable for a recreational area), AT is given by ED × 365 day/year × 24 h/day and C POE is the concentration in air at the exposure point (mg/m 3 ) given by: FT is the estimated transport factor and C S is the concentration at the source of emission (<C POE ). FTs are obtained by mathematical models considering direction, rate, and extent of chemical migration, if the point of exposure for a receptor is not at the source. According to the ASTM standard [57], the transport model related to dust inhalation pathways is identified with the particle emission factor (V Fp ), while in the case vaporization of a compound from surface soil to outdoor air, the FT is the volatilization factor (VF SS ). Both V Fp and VF SS , predicting the attenuation of chemical concern away from the soil source, were calculated by using both the site-specific data (source geometry, vadose zone soil feature and metereological parameters) and physico-chemical properties obtained during field investigations, in addition to the conservative default values suggested by the ASTM standard: 7.42 × 10 −7 and 4.60 × 10 −12 (mg/m 3 -air)/(mg/kg-soil), respectively.
The cumulative risk from simultaneous exposure to several non-carcinogens is defined by the screening level hazard index (HI) as: HI = HQ ingestion + HQ dermal + HQ inhalation An HI value less than unity indicates that the risk is acceptable; otherwise, if HI exceeds one, non-carcinogenic effects are in danger of occurring, with probability increasing as the value of the HI increases.

Soil and Plants
TOC, pH, trace element data and total Hg concentration in borehole samples, rhizosphere, and dandelion leaves are reported in Tables 1 and 2. The total organic carbon content ranged between 0.42 and 1.92 wt%. Samples from both boreholes and trenches were characterized by relatively high Fe, Mn, and Cr content (from 22 to 35 g/kg; from 495 to 1143 mg/kg; and from 76 to 129 mg/kg, respectively). Lead concentration in PZ1-A, PZ1-B, PZ2-A, T1-A, and T1-B exceeded the maximum concentration level (MCL) imposed by Italian regulations for residential and recreational areas (100 mg/kg). Cu and Zn exceeded the MCL (120 mg/kg and 150 mg/kg, respectively) in PZ1-A, T1-B, PZ2-A, and T1-B. It has to be noted that the Cr, Ni, and V contents overlapped with the concentration patterns that characterized the natural background of the Pisa plain sediments [65], interpreted as having originated from erosion and the weathering of ophiolitic rocks in the Serchio and Arno River Apennine catchments. Pb, Cu and Zn enrichments were not attributable to the Olocene sediments of the Pisa floodplain, and reflected inputs from anthropogenic sources such as the effects of urban vehicles [66]. Total mercury was above the MCL (1 mg/kg) in the borehole and trench samples (except PZ2-B), reaching 54 mg/kg; the Hg T concentration in the PZ3 profile showed that Hg contamination was confined to the shallower soil layers. Mercury in the dandelion rhizosphere soil was widespread at concentrations exceeding MCL for almost the entire area, even if not homogeneously distributed, reaching 172 mg/kg. The Hg in dandelion leaves ranged between 0.42 and 1.50 mg/kg; no correlation with the Hg content in the rhizosphere was observed. The Hg content in the roots was between 0.01 and 0.03 mg/kg, indicating that the level of Hg uptake by plants from soil was minimal, or mercury does not reside in the roots. It is worth noting that leaves from dandelion plants growing in vases had comparable Hg content to the leaves of plants growing in the field. This suggests that the Hg uptake in leaves mainly comes from atmospheric mercury (e.g., [67]), most likely Hg 0 . It is worth noting that the mercury vapor concentration during soil gas sampling obtained by suck-up techniques [68] reached 1300 ng/m 3 , confirming that gaseous elemental mercury in soil has been spilled outside.

Hg Speciation in Soil
In order to quantify the Hg forms in soils, the PZ1-A and PZ2-A samples were chosen as the best representative of the area in terms of soil mercury contamination ( Figure 2, Table 1).

Hg Determined by Chemical Extraction
The results of sequential extraction for PZ1-A and PZ2-A soils are graphically shown in Figure 3.

Hg Speciation in Soil
In order to quantify the Hg forms in soils, the PZ1-A and PZ2-A samples were chosen as the best representative of the area in terms of soil mercury contamination ( Figure 2, Table 1).

Hg Determined by Chemical Extraction
The results of sequential extraction for PZ1-A and PZ2-A soils are graphically shown in Figure 3. The soil samples had similar distributions of Hg species, with the F4 fraction (12 M HNO3 extractant) being the most abundant (79 ± 9%). This fraction was intended as a representative of both elemental mercury and mercury bound to Fe and Mn oxides. The F5 fraction (aqua regia extractant) revealed the possible occurrence of a minor amount of mercury sulfides (3.7 ± 2.6%) suggesting that, in general, conditions were not favorable to the spontaneous stabilization of mercury by conversion to cinnabar or metacinnabar. The F3 fraction (1 M KOH extractant), representing organo-chelated Hg, was 2.6 ± 1.1%, sug- The soil samples had similar distributions of Hg species, with the F4 fraction (12 M HNO 3 extractant) being the most abundant (79 ± 9%). This fraction was intended as a representative of both elemental mercury and mercury bound to Fe and Mn oxides. The F5 fraction (aqua regia extractant) revealed the possible occurrence of a minor amount of mercury sulfides (3.7 ± 2.6%) suggesting that, in general, conditions were not favorable to the spontaneous stabilization of mercury by conversion to cinnabar or metacinnabar. The F3 fraction (1 M KOH extractant), representing organo-chelated Hg, was 2.6 ± 1.1%, suggesting a minor fraction of Hg associated with soil organic matter. The Hg sequential extraction experiments also revealed that the F1 fraction (distilled water) and F2 fraction (0.01 M HCl + 0.1 M CH 3 COOH) extracted a negligible amount of mercury from the soil (0.4 ± 0.3% and 0.2 ± 0.2%, respectively).
Methylmercury from ingestion and inhalation pathways easily enter the bloodstream and rapidly distribute to all tissues, accumulating in the brain. The presence of methylmercury in soil is hence of particular importance in characterizing mercury hazard. The extraction results indicated very low values of methylmercury species in the studied soil. In particular, the measured MeHg concentrations of the PZ1-A and PZ2-A samples corresponded to 0.1%, respectively, of total Hg content.

Hg Thermal Desorption Analysis
Considering the comparable results obtained during the sequential chemical extraction for PZ1-A and PZ2-A soils, thermo-desorption experiments at different heating rates were applied only to the PZ1-A sample. The obtained thermograms had the same shape and were fitted by two peaks ranging between 110-130 • C (peak #1) and 160-200 • C (peak #2) in experiments at variable heating rates. An example is shown in Figure 4.

Hg Thermal Desorption Analysis
Considering the comparable results obtained during the sequential chemical extraction for PZ1-A and PZ2-A soils, thermo-desorption experiments at different heating rates were applied only to the PZ1-A sample. The obtained thermograms had the same shape and were fitted by two peaks ranging between 110-130 °C (peak #1) and 160-200 °C (peak #2) in experiments at variable heating rates. An example is shown in Figure 4. Peak #1 is attributable to the desorption of Hg 0 free in the soil matrix ( [69]]. In order to assign the peak #2 to Hg species, activation energy was obtained by thermo-release curves measured at the heating rates (β) of 8, 10, 12, 20, 35 °C/min. As expected, the peak maximum (Tpeak) shifted to a higher temperature with increasing β [70,71]. Tpeak and β can be used to determine the first-order activation energy (Ea) of Hg vapor emission from soil according to:

= +
where A represents the Arrhenius pre-exponential factor and R the Boltzmann gas constant [71]. In a series of heating experiments with different β a linear correlation is hence expected between T 2 peak/β and 1/Tpeak, whose slope is given by Ea/R. The results obtained for PZ1-A soil are graphically shown in Figure 5. Peak #1 is attributable to the desorption of Hg 0 free in the soil matrix [69]. In order to assign the peak #2 to Hg species, activation energy was obtained by thermo-release curves measured at the heating rates (β) of 8, 10, 12, 20, 35 • C/min. As expected, the peak maximum (T peak ) shifted to a higher temperature with increasing β [70,71]. T peak and β can be used to determine the first-order activation energy (E a ) of Hg vapor emission from soil according to: where A represents the Arrhenius pre-exponential factor and R the Boltzmann gas constant [71]. In a series of heating experiments with different β a linear correlation is hence expected between T 2 peak /β and 1/T peak , whose slope is given by E a /R. The results obtained for PZ1-A soil are graphically shown in Figure 5. The slope of the regression (6.60) gives an activation energy of 13.2 Kcal/mol. A range of desorption activation energies have been reported for Hg gas emission from soils, mostly depending on the physico-chemical properties of the soil and sorption processes [72][73][74][75]. The E a obtained in this study was close to the enthalpy of the Hg 0 liq → Hg 0 gas reaction (about 14 Kcal/mol), and might reflect the vaporization of elemental Hg more deeply retained within the soil structure, compared with that represented by peak #1. Indeed, a double peak for Hg 0 release with a maximum at approximately 100 • C and 200 • C has been reported in the literature (e.g., [76]) and ascribed to Hg 0 as variably bound to Fe oxydydroxides, while a higher desorption temperature (200-300 • C) is attributed to matrix bound Hg 2+ [69,77,78].
There is no clear consensus about this matter; however, considering the obtained E a we can assign peak #2 to matrix-bound Hg 0 , where the matrix is a mineral phase. Most of the Hg 2+ released during the pristine contamination would hence have been reduced to Hg 0 . Indeed, in natural settings, divalent mercury can be reduced to elemental mercury by specific inorganic reducing agents, such as bacteria and sunlight; the process of the biotic and abiotic production of elemental mercury in the soil-water system via Hg 2+ reduction represents a pathway in the biogeochemical cycle of mercury. Furthermore, experimental studies have demonstrated that the ferrous iron associated with iron oxyhydroxides, such as magnetite grains commonly found in the Pisa plain sediments [52] and other minerals (e.g., siderite), is a reductant of Hg 2+ , with the formation of elemental Hg 0 by surface reactions (e.g., [79]).
The deconvolution of the desorption curve, scaled to the % recovery in the F4 fraction obtained by chemical sequential extraction, yielded that about 25% of total mercury in soil was free Hg 0 (peak #1) and 55% was matrix-bound Hg 0 (peak #2). It is worth mentioning that cinnabar and/or metacinnabar, released at higher temperature and highlighted by sequential extraction, were not detected by thermograms, suggesting that their amount was below the peak resolution.

Water
Physico-chemical parameters, major ions, and trace element content of groundwater are reported in Table 3. In PZ1_W and PZ2_W temperature was quite constant in both the summer and winter surveys (between 15.4 and 17.3 • C), despite large changes in the air temperature. Water had a circum-neutral pH, and electrical conductivity was somewhat lower during the dry season (July 2019; EC = 1955-2230 µS/cm) compared with the wet season (January 2020; EC = 2600-3370 µS/cm). Dissolved oxygen ranged between 2.0 and 3.6 mg/L, indicating that the original oxygen content was significantly modified through chemical and biological processes. In particular, the occurrence of NH 4 + might reflect the decay of buried organic matter; indeed, at the Eh-pH conditions measured in groundwater, NH 4 + oxidized to nitrate-nitrogen, suggesting the occurrence of water bodies in the aquitard of anoxic conditions not directly sampled during the surveys. The major ion chemistry indicated that PZ1_W and PZ2_W belonged to Na-Cl-type water, according to the classification based on the Piper diagram ( Figure 6). The relatively high Na and Cl content indicated the contribution of a marine component. However, the SO 4 2− content was lower compared with what was expected on the basis of the Cl − /SO 4 2− ratio in seawater, pointing to sulfate reduction by organic matter. This hypothesis was supported by the H 2 S "rotten egg" smell and the relatively high HCO 3 − content measured. The seasonal reducing environment, occurring in soil as a result of the different hydrogeochemical conditions and sulfate reduction by sulfate-reducing microbes, might locally facilitate the formation of minor amounts HgS from elemental mercury [80].
Groundwater was characterized by high Mn, relatively high Fe content, and in some cases Ni above MCL (50,200 and 20 µg/L for Mn, Fe and Ni, respectively), as expected on the basis of the lithology of the aquitard originating from the erosion of ophiolitic rock outcropping in the Apennine catchments, and commonly found in groundwater in the coastal plain of Pisa. Arsenic in PZ1_W1 slightly exceeded the MCL (10 µg/L), possibly reflecting As hosted in sulfide veins related to the hydrothermal activity across the northern Apennine ophiolites. With the exception of Mn, Ni, and As in PZ1_W1, the concentrations of potentially toxic elements were below the corresponding MCL imposed by Italian regulations. It is worth noting that Hg was below the detection limit of 0.2 µg/L in all water samples.

Risk Analysis
Non-carcinogenic health risk assessment was performed by calculating HQ and HI. The data obtained using the highest mercury concentration measured on site (172 mg/kg) and speciation results (percent yields of 3.7, 25, 55 and 0.1 for mercury sulfide, free Hg 0 , matrix-bound Hg 0 and methylmercury, respectively; see Section 4.2) are shown in Table 4.  Groundwater was characterized by high Mn, relatively high Fe content, and in some cases Ni above MCL (50,200 and 20 µg/L for Mn, Fe and Ni, respectively), as expected on the basis of the lithology of the aquitard originating from the erosion of ophiolitic rock outcropping in the Apennine catchments, and commonly found in groundwater in the coastal plain of Pisa. Arsenic in PZ1_W1 slightly exceeded the MCL (10 µg/L), possibly reflecting As hosted in sulfide veins related to the hydrothermal activity across the northern Apennine ophiolites. With the exception of Mn, Ni, and As in PZ1_W1, the concentrations of potentially toxic elements were below the corresponding MCL imposed by Italian regulations. It is worth noting that Hg was below the detection limit of 0.2 µg/L in all water samples.

Risk Analysis
Non-carcinogenic health risk assessment was performed by calculating HQ and HI. The data obtained using the highest mercury concentration measured on site (172 mg/kg) and speciation results (percent yields of 3.7, 25, 55 and 0.1 for mercury sulfide, free Hg 0 , matrix-bound Hg 0 and methylmercury, respectively; see Section 4.2) are shown in Table  4.  It was observed that the risk was associated with soil ingestion by children, while dermal contact and inhalation pathways did not produce a significant health risk. In particular, an HQ ingestion exceeding unit was obtained for children for the cases of free and matrix-bound Hg 0 (HQ of 1.47 and 3.24, respectively), while for sulfide-bound Hg and methylmercury the HQ value is negligible.
The hazard index calculated considering all the exposure routes active on-site exceeded unity (HI = 4.99), and therefore posed a (non-cancerous) risk to children, who represent the major target users of the playground space, and the most vulnerable group of city residents.
It is worth mentioning that the speciation data obtained in this study allowed us to perform an actual risk assessment instead of using the total Hg concentration for each exposure pathway, with the latter leading to an overestimation of the risk.

Discussion
The measured metal concentrations revealed that soil samples in the studied urban green space-planned to be converted into a recreational area in the town of Pisacomprised potentially toxic elements. The major threat was related to the occurrence of mercury, largely exceeding the guideline concentration values for residential use imposed by Italian regulations. Mercury in the soil was not attributable to lithogenic sources, representing the legacy of HgCl 2 applied in the past artisanal tannery activities that characterized the site. The data showed that mercury contamination was widespread, with the highest concentration, reaching 172 mg/kg, measured in the rhizosphere.
The analytical results from soil gas sampling revealed that mercury degassing is an active process on the site, in agreement with the results obtained by the thermal desorption laboratory experiments demonstrating the occurrence of free Hg 0 and Hg 0 retained in the soil, produced by reducible matrix-bound Hg 2+ . Furthermore, the sequential chemical extraction indicated that the main Hg species in soil were extracted by the 12 M HNO 3 fraction, representative of both elemental and matrix-bound Hg. We can suppose that the matrix-bound Hg mostly comprised Hg adsorbed in the mineral surface, in particular onto Fe-Mn oxyhydroxides commonly found in the Pisa alluvial plain. The relatively high pH of the soil may have enhanced the Fe-Mn oxyhydroxides' sorptive capacity [81].
The chemical sequential extraction also indicated that the organo-chelated Hg fraction was below 3 wt%. A minor amount of mercury (below 4 wt% from the sequential extraction results) had possibly formed sulfides. Indeed, in anaerobic Hg-contaminated soils, the microbial reduction of sulfate can lead to the in situ formation of mercuric sulfide (HgS), occurring as cinnabar or metacinnabar.
The obtained results hence highlight that Hg 2+ from the original industrial input of mercuric chloride has mostly undergone reduction-therefore being less toxic to the environment and comprising less reactive Hg 0 -but is still highly toxic to humans when inhaled [82], heterogeneously distributed, and variably bound to the soil matrix. The microbial reduction of Hg 2+ to Hg 0 has been documented in periodically flooded anaerobic soils, such as the site under study during heavy raining events [83]. Reduction reactions in surface soils may also be enhanced photochemically [84].
Mercury contamination has not been detected in groundwater, indicating that Hg 2+ reduction does not lead to the release of a significant amount of Hg in solution. This is interpreted as reflecting soil speciation, revealing the occurrence of insoluble HgS in soil, stable organo-chelated Hg complexes, and Hg 0 having relatively low solubility (about 6.0 × 10 −5 g/L at 25 • C). This was also confirmed by water leachates during the sequential extraction experiments (F1 fraction), which contained negligible amounts of mercury. It should also be noted that the water table level in the study area was generally not sufficiently high enough to be in contact with the contaminated soil layer, preventing the direct scavenging of Hg from groundwater. In addition, in the case of Hg release to groundwater, lateral fluxes in the aquitard would be limited due to the low hydraulic conductivity, reducing the possibility of soluble Hg migration pathways. Even though, on the basis of the available data, there is no evidence that Hg enters the groundwater, the observed increase in the Cl − concentration might result in forming stable aqueous complexes with Hg, enhancing mercury leaching during the interaction between groundwater and the mercury mineral phases occurring at depth [85]. Indeed, the redox profile in the aquifer and the dynamic between infiltrating rainwater and deeper saltwater lenses may favor Hg-Cl complexes, promoting the possible dissolution of Hg-bearing sulfides. This requires the monitoring of groundwater to be continued.

Conclusions
The re-use of dismissed areas by urban greening policies represents an effective way of improving the quality of life for citizens, and therefore has an important social function. In the present study, the soil and groundwater quality of an abandoned site in the city center of Pisa, planned to become a new public green space, was investigated by geochemical analysis. The obtained results highlighted that mercury contamination in the soil is the major threat. According to historical records, this mercury contamination represents the legacy of a past artisanal tannery that used HgCl 2 in leather processing. The activity ceased without a reclamation program, and specific monitoring and management were not carried out. The data indicate that Hg contamination is widespread, but confined to the uppermost soil layers, and that this contaminant is not mobilized to groundwater. Soil chemical sequential extraction, thermal desorption experiments, and field measurements indicated that redox conversion of Hg 2+ into Hg 0 occurred in the soil, and that Hg 0 was heterogeneously distributed and variably bound to the soil matrix. A small fraction of mercury was likely fixed in soil as HgS. Even if elemental mercury was less reactive and less toxic to the environment with respect to Hg 2+ and organomercurials, the risk assessment based on the speciation data indicates an unacceptable risk (HI > 1) due to the possible direct soil ingestion of children. This study demonstrates the usefulness of a site-specific investigation approach providing geochemical data for modelling the risks associated with human exposure to hazardous elements in an urban environment-in this instance mercury. Furthermore, risk assessments should be the main basis for remedial action regarding human health protection.