A week of automated continuous measurements of temperature, pH and specific conductance in a mining-polluted river in Sardinia (Italy): The effect of rainfall as a case of serendipity

Figure 4 shows the comparative patterns of temperature (T) and specific conductance (SC) at the two sites RIV3 and RIV10 throughout the period of measurements with the multi-parameter sondes. In the time span before the first rainfall, at the RIV3 site, a constant SC can be observed, while T follows the typical sinusoidal pattern related to the alternation between day and night (Var = 9°C; D% = 58). As SC is a conductivity measurement corrected to 25°C, its invariance between day and night is expected, unless there are external water inputs. With the beginning of the first rainy period, it is interesting to note that the day-night thermal effect on T is cancelled out, while SC decreases abruptly three times due to dilution with rainwater but always returns to the initial value, first very quickly and then more gradually (about 30 h). The rising speed of SC is obviously influenced by the amount of rain falling in the unit of time, always taking into account that the Rio Irvi, for most of the year, is mainly fed by groundwater coming out of the Casargiu gallery (CAS1). The rain on 9–10 May almost halves SC, which takes about 24 h to return to its initial value. In this case, it is interesting to note that the daily variation of T is less influenced by the rain, maintaining a general sinusoidal pattern. Similar patterns are observed at the RIV10 site, although here the SC measurements are more unstable, and the curve is strongly jagged. The main differences with RIV3 are that (i) the range of T is much wider (Var = 18°C; D% = 165); (ii) during the first rainy period, SC decreases progressively instead of falling and rising; (iii) the minimum SC is reached about 12 h after the 9–10 May rainfall. All this can be explained by the fact that the RIV10 site is several kilometres downstream of RIV3, it is located in the coastal plain quite close to the sea, it is less readily affected by groundwater input from CAS1, and it collects a larger volume of runoff during rainfall.

Figure 4

Comparative patterns of temperature and specific conductance at the two sites RIV3 and RIV10 throughout the period of measurements with the multi-parameter sondes.

Figure 5 shows the comparative patterns of temperature (T) and pH at the two sites RIV3 and RIV10 throughout the period of measurements with the multi-parameter sondes. At the RIV3 site, the pH is mildly acidic, ranging from 5.75 to 6.45 (Var = 0.70; D% = 12), and its pattern is highly symmetrical to the T pattern. A discordant T–pH pattern is also present at the RIV10 site, where the pH is acidic, ranging from 3.76 to 4.34 (Var = 0.58; D% = 15). This discordant relation between T and pH has already been shown in other studies but is still rather uncommon since the literature generally reports a concordant T–pH relation linked to the photosynthesis-respiration cycles of aquatic life (at least in rivers with neutral or slightly alkaline pH) according to the following reactions where (org) and (aq) stand for organic and aqueous, respectively (Gammons et al., 2015):

Figure 5

Comparative patterns of temperature and pH at the two sites RIV3 and RIV10, throughout the period of measurements with the multi-parameter sondes.

Daytime T increase (3) photosynthesis : CO 2 + H 2 O + sunlight → CH 2 O (org) + O 2 \text{photosynthesis}:\hspace{.5em}{\text{CO}}_{2}+{H}_{2}O+\text{sunlight}\to {\text{CH}}_{\text{2}}{\text{O}}_{\text{(org)}}+{\text{O}}_{\text{2}} (4) pH increase: HCO 3 − + H + → CO 2(aq) + H 2 O {\text{pH increase: HCO}}_{3}^{-}+{\text{H}}^{\text{+}}\to {\text{CO}}_{\text{2(aq)}}+{\text{H}}_{\text{2}}\text{O}

Nighttime T decrease (5) respiration: CH 2 O (org) + O 2 → CO 2 + H 2 O {\text{respiration: CH}}_{\text{2}}{\text{O}}_{\text{(org)}}+{\text{O}}_{\text{2}}\to {\text{CO}}_{\text{2}}+{\text{H}}_{\text{2}}\text{O} (6) pH decrease: CO 2(aq) + H 2 O → HCO 3 − + H + {\text{pH decrease: CO}}_{\text{2(aq)}}+{\text{H}}_{\text{2}}O\to {HCO}_{3}^{-}+{\text{H}}^{\text{+}} Even in metal-polluted rivers where there may be very little aquatic life, the effect of temperature on CO₂ solubility would lead to the result described above in each case. However, if we focus on the production–consumption of O₂ related to the photosynthesis–respiration cycles, other oxidation–reduction reactions can be influenced, such as those involving the Fe species that, together with SO₄, Zn, Mg and Ca, dominate the chemical composition of the Rio Irvi water (Frau et al., 2015). In particular, when the following reactions are considered: (7) Fe 2+ + 0 .25O 2 + H + ↔ Fe 3+ + 0 .5H 2 O {\text{Fe}}^{\text{2+}}\text{ + 0}{\text{.25O}}_{\text{2}}\hspace{.25em}{\text{+ H}}^{\text{+}}\leftrightarrow {\text{Fe}}^{\text{3+}}\text{ + 0}{\text{.5H}}_{\text{2}}\text{O} (8) Fe 3+ + 3H 2 O ↔ Fe ( OH ) 3(am) + 3H + {\text{Fe}}^{\text{3+}}+{\text{3H}}_{\text{2}}O\leftrightarrow \text{Fe}{(\text{OH})}_{\text{3(am)}}+{\text{3H}}^{\text{+}} (9) Fe 2+ + 0 .25O 2 + 2 .5H 2 O ↔ Fe ( OH ) 3(am) + 2H + {\text{Fe}}^{\text{2+}}\text{ + 0}{\text{.25O}}_{\text{2}}\hspace{.25em}\text{+ 2}{\text{.5H}}_{\text{2}}O\leftrightarrow \text{Fe}{(\text{OH})}_{\text{3(am)}}+{\text{2H}}^{\text{+}} where (am) stands for amorphous and reaction (9) is the combination of reactions (7) and (8), it is apparent that daytime O₂ production from photosynthesis shifts the reaction (9) to the right, generating acidity, while nighttime O₂ consumption from respiration shifts the reaction (9) to the left or simply slows it down, consuming or not-producing acidity and partially redissolving the previously precipitated Fe(III)-hydroxide. As described in a previous study (Frau et al., 2015), the reaction (9) predominates at the RIV10 site, while the RIV3 site is characterized by the precipitation of GR (sulphate green rust) according to the following reaction: (10) 6Fe 2+ + SO 4 2 − + 0 .5O 2 + 19H 2 O ↔ Fe 4 II Fe 2 III ( OH ) 12 ( SO 4 ) ⋅ 8H 2 O + 10H + {\text{6Fe}}^{\text{2+}}+{\text{SO}}_{4}^{2-}+\text{0}{\text{.5O}}_{\text{2}}+{\text{19H}}_{\text{2}}O\leftrightarrow {\text{Fe}}_{4}^{\text{II}}{\text{Fe}}_{2}^{\text{III}}{(\text{OH})}_{\text{12}}({\text{SO}}_{\text{4}})\cdot {\text{8H}}_{\text{2}}{\text{O + 10H}}^{\text{+}} which is analogous to reaction (9) in determining the pH pattern opposite to the T pattern. In both cases, the effect of temperature on O₂ solubility should not be so important since O₂ is much less soluble in water than CO₂.

An opposite relation between T and pH could also be due to photoreduction of aqueous Fe(III) according to such a reaction: (11) Fe 3+ + 0 .5H 2 O + sunlight → Fe 2+ + 0 .25O 2 + H + {\text{Fe}}^{\text{3+}}+\text{0}{\text{.5H}}_{\text{2}}\text{O + sunlight}\to {\text{Fe}}^{\text{2+}}+\text{0}{\text{.25O}}_{\text{2}}+{\text{H}}^{\text{+}} Gammons et al. (2005a) also suggested the importance of aqueous Fe³⁺ speciation in determining a diel cycle of pH analogous to that described in this article, according to the following reaction: (12) Fe ( SO 4 ) + + 0 .5H 2 O ↔ Fe 2+ + SO 4 2 − + 0 .25O 2 + H + \text{Fe}{({\text{SO}}_{\text{4}})}^{\text{+}}+{\text{0}\text{.5H}}_{\text{2}}O\hspace{1em}\leftrightarrow {\text{Fe}}^{\text{2+}}+{\text{SO}}_{4}^{2-}+\text{0}{\text{.25O}}_{\text{2}}+{\text{H}}^{\text{+}} However, Fe speciation at both RIV3 and RIV10 sites is by far dominated by the free Fe²⁺ ion, while the dominant Fe(III) species are Fe(OH) 2 + {\text{Fe(OH)}}_{2}^{+} and Fe(SO₄)⁺, respectively, but with concentrations three–four orders of magnitude lower than Fe²⁺. Therefore, it seems very unlikely that changes in pH during 24 h are determined by the formation/dissociation of the aqueous Fe(SO₄)⁺ complex.

There is no doubt that the complex biological/microbial activity present in the Rio Irvi, which is still at an early stage of study and understanding, has a decisive influence on the Fe (and other metals) cycle in solution and in precipitates, and consequently also on pH (Gammons et al., 2015; Lueder et al., 2020; Morris et al., 2005; Paganin et al., 2021).

It is also interesting to observe the effect of the first rainfall period in which T and pH remain almost constant at the RIV3 site, while at the RIV10 site, a relative constancy of T is accompanied by a clear increase in pH. This can be explained by the fact that dilution with rainwater, to which a pH of 5.65 (pure rain in equilibrium with atmospheric CO₂) can be attributed, does not substantially change the mildly acidic pH of the river water at the RIV3 site, while it causes the increase of the markedly acidic pH at the RIV10 site. Pure rain in equilibrium with atmospheric CO₂ can be used in subsequent simulations, as the concentrations of solutes (in particular, sulphate and iron) in rainwater are negligible compared to the water of the Rio Irvi, which contains very high concentrations of dissolved ions (Table 1). For example, studies conducted on rainwater in Sardinia (Caboi et al., 1995) have determined dissolved iron concentrations averaging 13 μg/L, compared with concentrations exceeding 100 mg/L at the two sites RIV3 and RIV10, and dissolved sulphate concentrations in the range of 5–11 mg/L, compared with concentrations of around 3,500 mg/L in the Rio Irvi.

A simulation was carried out with PHREEQC to estimate the volume of rain needed to raise the pH at the RIV10 site; Figure 6 shows that, starting from pH of 3.89 in the absence of rain, a progressive mixing of rain up to 70% of the water flowing in the Rio Irvi is needed to achieve pH of 4.30 at the end of the rainfall period from 7 to 8 May. It is evident that the slope of the two curves shown in Figure 6 cannot be the same, as the simulation does not take into account the time as a variable in the mixing of river water and rainwater.

Figure 6

Measured and simulated pH patterns at the RIV10 site during a rainfall period.

The effect of rainfall on river chemistry can potentially influence the precipitation/dissolution of a mineral phase, especially when it is an unstable phase, as in the case of the formation of GR in the first stretch of the Rio Irvi, where Fe(II) still persists in solution together with Fe(III). To highlight this process, a simulation was carried out with PHREEQC. The saturation index (SI) values of GR were calculated by PHREEQC with the standard formula SI = log(IAP) – log(K _sp), where IAP is the ion activity product and K _sp is the solubility product constant of the mineral phase. For further information on the method used to calculate the K _sp of GR, refer to Frau et al. (2015).

Figure 7 shows that, starting from the chemical composition at the RIV3 site reported in Table 1, the initial oversaturation (SI = 0.72) with respect to GR progressively shifts to saturation (SI = −0.02) and then to a marked undersaturation (SI = −1.21) when, respectively, 40 and 70% of the water flowing in the Rio Irvi is represented by rain. This process may be periodically important in contributing to the metal load discharged from the Rio Irvi and other polluted rivers to the sea (Frau et al., 2015). Indeed, a phenomenon of partial removal of Fe precipitates from the Rio Irvi bed has sometimes been observed during heavy rainfall. In that case, there is a probable concomitance of a dissolution process of riverbed precipitates, as a result of undersaturation caused by the dilution of river water with rainwater, with a physical transport of particles of riverbed precipitates, due to the increase in the flow rate of river water.

Figure 7

Simulated variation of the saturation index (SI) of sulphate green rust at the RIV3 site as a consequence of dilution due to rainfall.

It should be noted that rainfall measurements and discharge data were not part of the original monitoring design. The rainfall events discussed here are therefore interpreted primarily in terms of their hydrogeochemical effects on river water chemistry rather than as part of a hydrological rainfall-runoff analysis. In any case, previous studies (De Giudici et al., 2018; Frau et al., 2015) and an examination of the almost identical flow at the two sites, RIV3 and RIV10 (Table 1), indicate that, under dry conditions, the bed of the Rio Irvi is essentially sealed by extremely abundant muddy iron precipitates (amorphous or slightly crystalline Fe(III)-(oxy)hydroxide and sulphate green rust) that line the entire course of the river. Consequently, the inflow of groundwater from beneath the riverbed can be considered negligible, even during rainfall. On the other hand, especially during heavy rainfall, there will obviously be inflows of runoff water and, most likely, groundwater entering from the sides due to the lateral expansion of the stream network, with a consequent dilution effect on river water.

Figure 8a shows the comparative patterns of pH at the two sites, RIV3 and RIV10. The maxima and minima of the curve during the non-rainy periods at the RIV10 site are shifted forward a few hours compared to the curve at the RIV3 site. Similar shifts are also observed for T (Figure 8b) and SC (Figure 8c). As previously explained, this is mainly due to the fact that the RIV3 site is located upstream in a mountainous area, while the RIV10 site is located several kilometres downstream near the Piscinas beach.

Figure 8

Comparative patterns of pH (a), temperature (b) and specific conductance (c) at the two sites RIV3 and RIV10 throughout the period of measurements with the multi-parameter sondes.