Accumulation of the deep reactive layer
The profiles of many chemical components in the sediments at W2 (Fig. 3) strongly resemble the measured trajectory of electrical conductivity in Enonselkä (Fig. S1), with a rising trend towards a maximum in 1975 and a decline thereafter. The period during which the deep reactive layer was deposited was characterized by high nutrient inputs from municipal wastewater, leading to enhanced production and sedimentation of autochthonous organic matter (Keto, 1982). Urban development also led to elevated Zn concentrations in municipal and industrial wastewaters (Lahti Environment Services, personal communication), leading to elevated Zn contents in the sediments of Enonselkä. Simultaneous high water-column sulfate concentrations, related to both regional atmospheric deposition trends (Monteith et al., 2007) and local industrial sources such as pulp mills (e.g., Chen & Horan, 1998) led to high rates of sulfate reduction and sulfide mineral precipitation in the sediment column, as observed in other boreal systems (Couture et al., 2016). Finally, the enrichments of Fe, Mn, and P likely reflect enhanced redox shuttling of these elements from shallow to deep areas of the lake during intervals of widespread low oxygen conditions (Schaller & Wehrli 1996; Jilbert & Slomp, 2013; Lenz et al., 2015).
Role of the deep reactive layer in regulating modern P cycling
Although water quality in Enonselkä has improved considerably since 1975, the surface sediment enrichment of P is still far in excess of the pre-industrial background (Fig. 3), implying that full recovery of the system from the anthropogenic nutrient loading of the twentieth century remains incomplete. In the following sections, we investigate the modern diagenetic context of the sediments, in order to determine the mechanisms of P burial and regeneration before attempting to quantify their impact upon water-column P cycling in Enonselkä today.
Organic matter remineralization in the deep reactive layer
The convex profiles of ammonium (NH4+) and DIC at W2 (Fig. 4) confirm that remineralization of organic matter is presently taking place in the deep reactive layer (e.g., Burdige, 2006). At 20–60 cm depth, the layer is situated below the depth of significant bioirrigation (Kornijow & Pawlikowski, 2015) and hence microbial remineralization of organic matter is expected to proceed exclusively by anaerobic pathways. Similar convex profiles are observed in porewater Fe2+ and Mn2+, and in porewater CH4, suggesting both dissimilatory reduction of metal oxides and methanogenesis to be occurring simultaneously (Eqs. 4, 5a, 5b). Diagenetic reaction formulae used in the following sections are taken from Reed et al. (2011) and Katsev et al. (2006), with the primary redox reactions represented in the simplified forms given in Canfield (1993):
Metal oxide reduction, e.g., Fe:
$${\text{CH}}_{2} {\text{O}}_{{\left( {\text{s}} \right)}} + 7{\text{CO}}_{{2\left( {\text{aq}} \right)}} + 4{\text{Fe}}\left( {\text{OH}} \right)_{{3\left( {\text{s}} \right)}} \to 4{\text{Fe}}_{{\left( {\text{aq}} \right)}}^{2 + } + 8{\text{HCO}}_{{3\left( {\text{aq}} \right)}}^{ - } + 3{\text{H}}_{2} {\text{O}}_{{\left( {\text{l}} \right)}}$$
(4)
Methanogenesis:
$$2{\text{CH}}_{2} {\text{O}}_{{\left( {\text{s}} \right)}} + 2{\text{H}}_{2} {\text{O}}_{{\left( {\text{l}} \right)}} \to 2{\text{CO}}_{{2\left( {\text{aq}} \right)}} + 4{\text{H}}_{{2\left( {\text{aq}} \right)}}$$
(5a)
$$4{\text{H}}_{{2({\text{aq}})}} + {\text{CO}}_{{2({\text{aq}})}} \to {\text{CH}}_{{4({\text{aq}})}} + 2{\text{H}}_{2} {\text{O}}_{{({\text{l}})}}$$
(5b)
Although these reactions are typically considered to be competitive (e.g., Karvinen et al., 2015), with dissimilatory reduction of metal oxides the more energetically favorable pathway and hence observed at a shallower depth than methanogenesis (e.g., Arndt et al., 2013), vertical overlap of diagenetic zones is common in non-steady-state systems influenced by recent eutrophication (e.g., Jilbert et al., 2018; Sawicka & Bruchert, 2017). Additionally, Fe and Mn may be released into porewaters in the deep reactive layer via oxide-mediated anaerobic oxidation of methane (AOM) (Beal et al., 2009; Sivan et al., 2011; Egger et al., 2015b):
$${\text{CH}}_{{4({\text{aq}})}} + 8{\text{Fe}}({\text{OH}})_{{3({\text{s}})}} + 15{\text{H}}_{{({\text{aq}})}}^{ + } \to {\text{HCO}}_{{3({\text{aq}})}}^{ - } + 21{\text{H}}_{2} {\text{O}}_{{({\text{l}})}} + 8{\text{Fe}}_{{({\text{aq}})}}^{2 + } .$$
(6)
Diagenetic processes in the upper sediments
The porewater zonation of the upper sediments at W2 is complex and likely influenced by a combination of microbial and macrofaunal processes. The uppermost 10 cm are characterized by signals of organic matter remineralization by anaerobic pathways as described in Eqs. 4, 5a and 5b, including production of NH4+, CH4, and DIC (see zoomed profiles in Fig. 4b). The concave profile of porewater S also indicates organoclastic sulfate reduction in this layer:
$$2{\text{CH}}_{2} {\text{O}}_{{({\text{s}})}} + {\text{SO}}_{{4({\text{aq}})}}^{2 - } \to {\text{H}}_{2} {\text{S}}_{{({\text{aq}})}} + 2{\text{HCO}}_{{3({\text{aq}})}}^{ - }$$
(7)
The interval 10–20 cm is characterized by net removal of the reduced species NH4+, CH4, Fe2+, and Mn2+, as shown by concave profiles of these parameters (Fig. 4b). The most plausible explanation for these profiles is that advection of O2 by burrowing macrofauna outweighs the microbial remineralization of organic matter in this layer, leading to net consumption of reduced species in simple aerobic oxidation reactions (Eqs. 8–11). Hence we refer to this layer as the zone of net oxidation (Fig. 4b):
$${\text{NH}}_{{4({\text{aq}})}}^{ + } + 2{\text{O}}_{{2({\text{aq}})}} + 2{\text{HCO}}_{{3({\text{aq}})}}^{ - } \to {\text{NO}}_{{3({\text{aq}})}}^{ - } + 3{\text{H}}_{2} {\text{O}}_{{({\text{l}})}} + 2{\text{CO}}_{{2({\text{aq}})}}$$
(8)
$${\text{CH}}_{{4({\text{aq}})}} + 2{\text{O}}_{{2({\text{aq}})}} \to {\text{CO}}_{{2({\text{aq}})}} + 2{\text{H}}_{2} {\text{O}}_{{({\text{l}})}}$$
(9)
$$4{\text{Fe}}_{{({\text{aq}})}}^{2 + } + {\text{O}}_{{2({\text{aq}})}} + 8{\text{HCO}}_{{3({\text{aq}})}}^{ - } + {\text{H}}_{2} {\text{O}}_{{({\text{l}})}} \to 4{\text{Fe}}({\text{OH}})_{{3({\text{s}})}} + 8{\text{CO}}_{{2({\text{aq}})}}$$
(10)
$$2{\text{Mn}}_{{({\text{aq}})}}^{2 + } + {\text{O}}_{{2({\text{aq}})}} + 4{\text{HCO}}_{{3({\text{aq}})}}^{ - } \to 2{\text{MnO}}_{{2({\text{s}})}} + 2{\text{H}}_{2} {\text{O}}_{{({\text{l}})}} + 4{\text{CO}}_{{2({\text{aq}})}} .$$
(11)
The benthic macrofaunal community of Enonselkä is dominated by Chironomus and Potamothrix species (Iso-Tuisku, 2017). These organisms are capable of burrowing throughout the uppermost 20 cm of the sediment column (Hökkä, 1990) and hence advecting bottom-water oxygen into the 10–20 cm interval, as well as diluting the porewaters of this interval with bottom water. A similar influence of burrowing chironomid larvae on porewater NH4+ and Fe2+ concentrations was demonstrated by Lewandowski et al. (2007), and on porewater CH4 by Clayer et al. (2016).
Burial phases of reactive P
The observation that total PBE is consistently 25–30% lower than Total PTAD in the sediments of W2 (Fig. 3a) concurs with Hartikainen et al. (1996), who suggested that 25–30% of total P in sediments of the Enonselkä basin is bound in stable forms that cannot be extracted by most common methods. Apparently, only total sediment digestion approaches, such as our triple-acid digestion or the NaCO3 smelting method used by Hartikainen et al. (1996), are capable of extracting 100% of sedimentary P in this system. The nature of the most refractory P phase remains unclear, but it is unlikely that it participates in biogeochemical cycles due to its inaccessibility.
The large component of NaOH-soluble P (Fig. 3b) indicates that the majority of P accumulated in the upper sediments during the late twentieth century is present in reactive forms. We interpret NaOH-soluble P to represent a combination of P bound to oxides (primarily Fe- and Mn-, rather than Al-bound, as shown by Koski-Vähälä et al., 2001), labile organic P (Hartikainen et al., 1996), as well as authigenic manganous vivianite ((Fe, Mn)3(PO4)2·8H2O)). Indeed, Rothe et al. (2015) showed that vivianite is soluble in NaOH and is therefore indistinguishable from oxide-bound P in the Hieltjes & Lijklema (1980) sequential extraction (Table 2).
The Micro-XRF data allow more detailed investigation of the distribution of P in the sediments and co-location of P with other elements, and therefore can be used to augment the sequential extraction data. This approach is particularly useful for identifying heterogeneously distributed enrichments linked to authigenic or biogenic mineral formation (e.g., Jilbert & Slomp, 2013). The conspicuous clusters of P, Fe, and Mn enrichments visible in 2D Micro-XRF maps (Fig. 5a) are suggestive of an important burial phase of P at W2. To investigate how the occurrence of this phase varies with depth in the sediments, we first isolated 800 P-rich pixels from a line scan across several prominent clusters in the deep reactive layer. We then determined the P/Mn count ratio distribution of these pixels for use as a tracer of cluster occurrence throughout the upper 100 cm (Fig. 5b). P and Mn in the isolated pixels were significantly correlated to the 99% confidence level, with a least-squares regression yielding a P/Mn of 1.37 (R2 = 0.54, p = 0.88, Fig. 5b). Similarly, the median value of P/Mn in the 800 pixels was 1.33.
We subsequently plotted the frequency distribution of pixel P/Mn ratios in seven 2.5 cm scans (each representing 1000 pixels) of Micro-XRF data throughout the uppermost 100 cm of the sediments. The results show that pixel P/Mn ratios at 80–100 cm depth (scans 6 and 7 in Fig. 5c, d) are tightly distributed in the range 0–0.25, implying a relatively homogeneous distribution of P in the sediments, with a low abundance of P-Fe–Mn clusters. In the shallower layers (0–60 cm), the frequency distributions are much wider, with higher median values of P/Mn (0.4–0.7), implying an elevated contribution of P-Fe–Mn clusters to total P counts. Scan 5 (approx. 70 cm depth) is transitional between the two conditions. Hence, the appearance of the clusters roughly corresponds to the base of the deep reactive layer, and their occurrence persists upwards to the surface sediments.
The clustered co-enrichments of P, Fe, and Mn are strongly reminiscent of previous reports of authigenic manganous vivianite in freshwater and coastal marine sediments (e.g., Fagel et al., 2005, Egger et al., 2015a). The presence of vivianite is supported by two further lines of evidence. First, spherical blue nodules of 50–150 µm diameter could be isolated from the > 50 µm heavy fraction from the depths corresponding to Micro-XRF scans 1–5 in the surface sediments and deep reactive layer (Fig. 5), but could not be found in samples 6–7, which were dominated by mica and other silicates. These blue nodules bear a morphological resemblance to those presented in Rothe et al. (2014) and display SEM–EDS spectra consistent with manganous vivianite (Fig. 6a). Second, PHREEQC analysis shows that the porewater solution is supersaturated with respect to vivianite at all investigated depths (Fig. 6b). Although porewater supersaturation does not guarantee authigenic vivianite formation (Rothe et al., 2014), it is a pre-requisite for the process to occur spontaneously according to the following equilibrium reaction:
$$3{\text{Fe}}_{{({\text{aq}})}}^{2 + } + 2{\text{H}}_{2} {\text{PO}}_{{4({\text{aq}})}}^{ - } \to {\text{Fe}}_{3} ({\text{PO}}_{4} )_{{2({\text{s}})}} + 4{\text{H}}_{{({\text{aq}})}}^{ + } .$$
(12)
Hupfer et al. (2019) recently described how Fe oxide-bound P in the linings of abandoned chironomid burrows can transform in situ to vivianite in sediments of high Fe:S ratio, due to the excess of dissolved Fe over H2S during the transition to reducing conditions. This mechanism is plausible for the sediments of Enonselkä, where macrofaunal burrows exert a significant influence over the diagenetic zonation, and Fe content is strongly in excess of S (Fig. 3).
The combined evidence suggests that vivianite formation occurs in the modern sediments of Enonselkä, and therefore constitutes an important permanent burial pathway for P. Yet, part of the NaOH-soluble P fraction in the surface sediments and deep reactive layer (Fig. 3b) is very likely contributed by oxide-bound and organic P, which remains available for diagenetic reactions.
Vertical mobility of P from the deep reactive layer to surface sediments
The porewater data from W2 confirm that P is remobilized as phosphate in the upper part of the deep reactive layer (20–40 cm depth, Fig. 4). Very similar profiles were observed at site W1 (Fig. S2), implying that such mobilization from the deep reactive layer is common throughout the deep areas of the lake (> 10 m water depth). Furthermore the porewater profiles at W1 are very similar in both summer and winter (Fig. S5) indicating a continuous upwards diffusive flux from this layer throughout the year. The release of P into porewaters is expected to reflect the combined effect of direct remineralization from labile organic matter via reactions 4, 5a and 5b, as well as additional release of P associated with Fe and Mn oxides via reactions 4 and 6. This additional contribution from oxide-bound P may be significant. We calculated the theoretical porewater P profile assuming P release only from organic matter remineralization, using principles outlined in Burdige & Komada (2011):
$$\frac{{d\Delta \left[ {\text{DIC}} \right]}}{{d\Delta \left[ {{\text{H}}_{2} {\text{PO}}_{4}^{ - } } \right]}} = - r_{{{\text{C}}:{\text{P}}}} \frac{{D_{{{\text{H}}_{2} {\text{PO}}_{4}^{ - } }} }}{{D_{\text{DIC}} }},$$
(13)
where dΔ[species] = change in porewater concentration gradient over depth, rC:P = expected molar ratio of C:P in remineralized organic matter (assumed to be 106), and Dspecies = diffusion coefficients of H2PO4− and HCO3− at in situ temperature (Li & Gregory, 1974). The bottom-water P value in the calculation was fixed at the measured concentration of 15 µmol/L. This exercise shows that measured porewater P concentrations are significantly elevated with respect to the value estimated from remineralization of organic matter alone (Fig. 7), confirming that a large proportion of the P remobilized in the deep reactive layer is derived from oxide reduction.
The remobilization of P in the deep reactive layer generates a porewater concentration gradient (e.g., 15.02 × 10−3 µmol/cm4 at W2, Fig. 7), leading to a vertical upwards diffusive flux of P towards the zone of net oxidation at 10–20 cm depth. In the zone of net oxidation, in turn, the concave porewater P profile indicates net consumption of P, most likely due to co-precipitation with Fe and Mn oxides in burrows as described by Lewandowski et al. (2007). However, this immobilization of P into the solid phase does not imply that there is zero further transport towards the sediment surface. Bioturbation by Chironomus and Potamothrix is expected to mix the upper sediments, leading to biodiffusion of solid-phase constituents throughout the zone inhabited by benthic fauna (e.g., Boudreau, 1986). Part of the newly formed oxide-bound P will hence be transported vertically upwards, and taken up into repeated cycles of reduction and oxidation in the upper sediments (Hupfer & Lewandowski, 2008).
In the uppermost 10 cm, the convex porewater P profile indicates net release due to high rates of remineralization reactions. In this interval, a direct diffusive flux of P towards the sediment surface is indicated by the porewater profile (Fig. 4). When oxic conditions prevail at the sediment–water interface (in practice, for the majority of the annual cycle), the upwards diffusive flux of P in the uppermost 10 cm likely does not escape directly to the bottom water. Rather, precipitation of oxide minerals at the sediment surface traps this P in the uppermost millimeters of the sediment column, via reactions 10 and 11 (e.g., Reed et al., 2011). The actual transfer of P from the sediment to the water column in the deeper areas of the lake occurs primarily during summer hypoxic intervals, upon the reduction of oxides in the surface sediments and consequent diffusive efflux of P (Tammeorg et al. 2017). Although physical resuspension of sediments and porewaters accounts for the majority of the total flux of P from the sediments in Enonselkä (Niemistö et al., 2012), this process is primarily restricted to shallower areas (0–10 m water depth).
Basinwide estimates of P burial and regeneration from the deep reactive layer
Sedimentary P burial is the key process removing reactive P from aquatic biogeochemical cycles (e.g., Ruttenberg, 2003). The rate and distribution of P burial thus exerts a strong influence over the recovery trajectory in eutrophied aquatic systems (Jilbert et al., 2015). A typical approach to estimate permanent P burial in lake sediments under steady-state conditions is to calculate the mass accumulation rate of P at the depth at which solid-phase P contents stabilize below the active surface layer (Håkanson, 2003), often taken to be 10–25 cm (Søndergaard et al., 2003). In the deep basins of Enonselkä, this approach must be modified because of the influence of the deep reactive layer. At sites where the deep reactive layer is evident in the sediment data (in practice all sites > 10 m water depth, labeled C and W in Fig. 1/Table 1), we applied the principle that the decline in porewater P at the base of the deep reactive layer represents the ultimate lock-in depth for P in these sediments, since the diffusive gradient is reversed and the layer lies below the depth of bioturbation. Hence, solid-phase P passing through this horizon is assumed to be permanently buried. The modern burial rate of P at these sites is thus calculated by
$${\text{MAR}}_{\text{P}} = {\text{MAR}}_{{{\text{tot}}.}} \times P_{\text{tot - LID}} ,$$
(14)
where MARP = Mass accumulation rate (burial rate) of P in mg/cm2/yr, MARtot. = total mass accumulation rate of solid material in g/cm2/yr, and Ptot-LID = total PTAD content at the lock-in depth in mg/g. MARtot., in turn, was calculated from linear sedimentation rates (LSR, cm/yr) assuming volumetric porosity of 0.9 and solid-phase density of 2.65 g/cm3:
$${\text{MAR}}_{{{\text{tot}}.}} = {\text{LSR}}_{{{\text{tot}} .}} \times 0.265.$$
(15)
Burial rates of P were subsequently converted to units of mg/m2/d for comparison with earlier studies. For most sites, lock-in depth was estimated on the basis of the solid-phase profiles due to the absence of porewater data (Fig. S3). Here, the lock-in depth was assumed to be equivalent to the base of the deep reactive layer, as observed at W2.
Sites of water depth 0–10 m display more classical profiles of sedimentary P in eutrophic lakes (Fig. S4) as described in Carey & Rydin (2011). The absence of a P enrichment corresponding to the deep reactive layer at these sites implies either (1) that resuspension and focusing of sediments in these areas prevented the formation of such a layer entirely, or (2) that diagenetic processes have effectively recycled the reactive P back the sediment surface over recent decades. For these sites, we used the stable background value of Ptot. to estimate burial rates of P:
$${\text{MAR}}_{\text{P}} = {\text{MAR}}_{{{\text{tot}}.}} \times P_{{{\text{back}}.}} ,$$
(16)
where Pback. is the stable background value at the base of the core (Fig. S4).
The exponential increase in sedimentation rate observed with increasing water depth (Fig. 3d) is consistent with earlier studies reporting intensive sediment focusing into the deeps of Enonselkä (e.g., Nykänen et al., 2010; Niemistö et al., 2012; Tammeorg et al., 2018). This phenomenon strongly influences the spatial distribution of P burial (Fig. 8). To estimate basin-wide P burial rates, we first drew a logarithmic trendline through the P burial data for all sites, plotted against water depth (Fig. 8b). We then divided the basin into four water-depth classes as defined by the available bathymetric data, and used the trendline to estimate mean burial rates in each depth class. Subsequently, we multiplied these values by the areal coverage of each depth class as estimated from Fig. 1, to yield total P burial rates by depth class, as well as total basin-wide P burial. This exercise shows that the most important discrete depth class for P burial in Enonselkä is 10–20 m water depth (Fig. 8c), due to a combination of relatively high P burial rates and significant areal coverage. The total rate of P burial in Enonselkä as estimated by this method is 37,200 kg/yr, or 3.91 mg/m2/d for the entire basin on an area-normalized basis.
To estimate the basin-wide upwards vertical flux of P from the deep reactive layer to the zone of net oxidation, we assumed all sediment areas below 10 m water depth (in total 5.85 km2, Figs. 1, 8a) to display similar porewater zonation to that observed at W1 and W2. Based on the porewater P profiles, the upwards flux at W1 is 0.40 mg/m2/d, and at W2, 1.34 mg/m2/d (Fig. S7). Using an average of these values (0.87 mg/m2/d) and assuming 22% of Enonselkä sediments fall in the depth range > 10 m water depth (Fig. 8a), we calculate a total upwards diffusive flux in Enonselkä of 1860 kg/yr, or 0.19 mg/m2/d for the entire basin on an area-normalized basis.
Enonselkä P budget and the relevance of the deep reactive layer for the recovery from eutrophication
We constructed a new contemporary P budget for Enonselkä basin, taking into account sediment and porewater fluxes from this study, as well as previous estimates of other components of the P cycle (Fig. 9). The budget is based on the conceptual premise that the surface sediment layer (0–20 cm) is the primary reservoir of P in Enonselkä, and that the balance of vertical fluxes in and out of this reservoir determines the trajectory and speed of trophic change in the ecosystem as a whole. We divide the surface sediment box into shallow-water areas, in which only vertical downward (i.e., burial) fluxes from the base of the surface sediment layer are permitted; and deep-water areas, in which bidirectional exchange at the base of the surface sediment layer is permitted, due to regeneration of P from the deep reactive layer. In deep areas, permanent burial occurs at the base of the deep reactive layer. Release of P from the sediments to the water column proceeds by resuspension in shallow areas, as shown by Niemistö et al. (2012), and diffusive efflux in deep areas susceptible to low oxygen conditions, as shown by Tammeorg et al. (2017). Gross sedimentation of newly produced organic P, and settling resuspended material, deposits P to the sediments throughout the lake.
The budget reveals that annual cycling of P between the surface sediments and water column is intense, relative to the inputs and outputs from the system as a whole. Resuspension and gross sedimentation are approximately in balance, and these fluxes cycle the equivalent of 6% of the surface sediment P pool annually. The large magnitude of these fluxes is maintained by high P contents in the surface sediments, as well as the generally shallow bathymetry of Enonselkä, which makes the system vulnerable to wind-driven mobilization of the surface sediments.
Two key further observations may be drawn from Fig. 9. Firstly, that the budget as a whole is not in steady state. There is a net annual loss of 31,150 kg P, principally through sedimentary P burial. This value is equivalent to 1.2% of the surface sediment reservoir annually, indicating that the reservoir is in steady decline. We suggest an error margin of ± 0.2% (of the surface sediment reservoir) for this value, although in practice this margin is difficult to determine without a denser network of sediment cores to better constrain the relationship in Fig. 8b. This steady loss of recyclable P concurs with the long-term trajectory towards improved water quality in Enonselkä (Fig. S1). The second observation is that the upwards flux of P from the deep reactive layer to the surface sediments is significant (1860 kg/yr), and equivalent to > 40% of the present-day value of external P loading to Enonselkä. Therefore, the deep reactive layer still plays a role in maintaining high P contents in the surface sediments today, and may be partly responsible for the slow recovery trajectory over the period 1975–present. Moreover, the present-day value for the upwards diffusive flux of 1860 kg/yr is likely to be in decline, since the porewater concentration gradient is becoming shallower as the deep reactive layer is buried. Hence, the significance of the flux may have been greater in previous decades.