Diatom cell-size composition as a novel tool for quantitative estimates of the water table in peatlands

Abstract

Diatom cell-size composition is an indicator of aquatic environmental changes but has been rarely investigated, especially in semi-terrestrial peatlands. In this study, both taxonomic composition and cell-size composition of diatoms were analysed in 41 samples from two montane peatlands, northeastern China. Redundancy analyses revealed that diatom taxonomic composition was significantly related to the depth to the water table (DWT) and Ca²⁺, while cell-size composition was significantly associated with DWT and Si. DWT was the most important factor and its sole effect explained 26.2% and 17.9% of the total variance in taxonomic composition and cell-size composition, respectively. Accordingly, diatom-based water-table transfer functions were developed based on taxonomic composition and cell-size composition, respectively. The maximum-likelihood (ML) model based on diatom taxonomic composition had the best performance, with a correlation coefficient value (R²) of 0.78 and the root mean squared error of prediction (RMSEP) of 6.66 cm. The ML model based on cell-size composition had similar performance, with an R² of 0.78 and the RMSEP of 6.87 cm, suggesting that diatom cell-size composition can be a new quantitative means to track past water-table changes. This method requires further appraisal with palaeoecological data but offers a new option that deserves exploration.

1. Introduction

Diatoms are single-celled, eukaryotic organisms that are widely distributed in almost any aquatic ecosystems such as rivers and lakes, and even semi-terrestrial environments such as peatlands [1]. Due to their narrow optima and tolerances for many environmental variables, diatom taxonomic composition has been widely used for quantitative environmental reconstructions, which provide essential clues for tracking past climate change, ecosystem resilience, lake ontogeny, and human activities [2,3]. For example, diatom taxonomic composition-based water-table transfer functions have been developed to quantitatively infer past water-table changes in the peatlands of northeast China [4,5].

Besides alterations in the taxonomic composition, the cell-size composition of diatom communities does not remain constant and is closely linked to changes in environmental variables [6]. For example, climate warming has intensified thermal stratification in the water column, hence favouring small-sized diatoms with slow sinking velocities in Lake Tahoe [7]. In southeastern Ontario lakes, there is a bloom of small planktonic diatoms under a scenario of high nutrient concentrations and decreased water column transparency [8]. Similarly, diatom demographic and taxonomic shifts toward small cell sizes have led to an overall decrease in the mean cell size in the Great Lakes during the twentieth century [9]. Therefore, cell-size composition of diatom communities is a useful indicator of aquatic environmental changes, including nutrient and light availability and water column stratification [6,8].

Different from lakes and rivers, peatlands are semi-terrestrial ecosystems, where the water table is usually at or near the surface or the land [10]. Diverse microhabitats and spatial heterogeneity due to the discontinuous water availability host a high species diversity of diatoms [11]. The water table is one of the most important variables influencing peatland ecology, development, functions and processes [10]. Previous studies revealed that valve lengths of some dominant diatoms were negatively correlated with the depth to the water table (DWT) in peatlands. For example, Pinnularia borealis [12], and Eunotia paludosa [13–15] have been found to decrease in valve lengths from wet hollows to dry hummocks. Given that valve lengths of dominant diatom species vary with water-table changes, we hypothesize that diatom cell-size composition shifts from large-sized to small-sized valves with an increase in DWT.

To verify our hypothesis, we quantified valve lengths of diatoms in surface samples of peatlands, and then developed a quantitative relationship between cell-size composition and the water table using different transfer function models [2]. Meanwhile, taxonomic composition-based water-table transfer functions were also developed based on the same samples. This study aims to determine the robustness of diatom cell-size composition as a quantitative tool to estimate water-table changes, through comparing the performance of transfer functions developed based on diatom taxonomic composition and cell-size composition, respectively.

2. Materials and methods

The two study sites, i.e. Jinchuan and Hani, are both Sphagnum-dominated peatlands located in the Changbai Mountains of northeastern China (electronic supplementary material, figure S1). The regional climate features long, cold winters with mean annual temperatures of 2–5°C, and mean annual precipitation of 600–1000 mm [16]. Hani Peatland has an average elevation of 899 m, while Jinchuan Peatland is slightly lower with an elevation of 623 m. Vegetation communities in the two peatlands are characterized by Sphagnum mosses and shrubs [16]. A total of 41 Sphagnum samples were collected along the water-table gradient from Hani Peatland (n = 32) and Jinchuan Peatland (n = 9) in August 2016 (electronic supplementary material, table S1). Surface Sphagnum tufts (the upper 2 cm) were cut with scissors at each sampling point, and packed in polyethylene bags. All samples were transported to Hubei Key Laboratory of Regional Ecology and Environmental Change, and were stored in a refrigerator at 4°C until analysis. DWT was measured using a graduated ruler in an approximately 5 cm diameter hole. Conductivity (Cond), pH, and oxidation-reduction potential (ORP) were measured in the field using a calibrated portable instrument (HACH HQ40D®). Meanwhile, a water sample was collected from each hole and brought to the laboratory for further analysis. Water samples were filtered through a 0.45-µm pore size mixed cellulose ester filter membrane for the subsequent analysis of physico-chemical factors. Concentrations of PO³⁺₄, NO⁻₃, Ca²⁺, K⁺, Mg²⁺, Na⁺, Si and dissolved organic carbon (DOC) were measured using standard methods. Detailed procedures are provided in our previous study [5].

Diatom sample preparation was performed according to standard methods [17]. For diatom taxonomic composition, a minimum of 300 diatom valves were counted from each sample. Data of diatom taxonomic composition and environmental variables are sourced from our previous study [5]. In this study, lengths of more than 300 diatom valves were measured in each sample using an Olympus BX53 microscope with 100 Plan N equipped with an Olympus DP27 digital camera and the cellSens software. The precision of the measurement is 0.1 µm. In order to establish diatom cell-size composition, all diatom valves are grouped in 18 size classes at 2-µm intervals (i.e. 7–9 µm, 9–11 µm, 11–13 µm, etc.). Large-sized diatoms with valve lengths greater than 41 µm are rare and hence classified into one group. The species dataset comprises 32 main species (taxa with an abundance of at least 1% in at least one sample). Diatom data and environmental factors (excluding pH) were square root transformed and log-normalized prior to ordination analyses, respectively. An initial detrended correspondence analysis (DCA) revealed that the first axis gradient length of both taxonomic composition and cell-size composition was less than 3 s.d., suggesting that a linear ordination model (redundancy analysis, RDA) can be used to explore the relationship between diatom data and environmental factors. Forward selection, with the false discovery rate correction, and the Monte Carlo tests (p < 0.05, n = 999 unrestricted permutations) were used to find the significant explanatory variables. Further partial RDAs were conducted to investigate the sole effect of each significant variable on diatom taxonomic composition and cell-size composition, respectively. All ordination analyses were conducted using the CANOCO 5 software [18].

The transfer function was developed using different models, including weighted averaging (WA), maximum-likelihood (ML) and weighted average partial least squares (WAPLS). Firstly, the training set containing all samples was used to develop the transfer function. The performance of different models was assessed using three metrics, including the regression value R², the root mean squared error of prediction (RMSEP), and the maximum bias (MAX.bias) [2]. Secondly, to enhance the efficiency of the transfer functions, the samples with residuals exceeding 20% of the DWT gradient (12.8 cm in this study) were excluded [19], and the transfer functions were subsequently re-established. To evaluate the predictive accuracy of the transfer functions, the leave-one-out cross-validation method was employed. The transfer functions were developed through the rojia package [20] in RStudio [21].

3. Results

The training set includes diatom data and 13 environmental variables in 41 surface samples (electronic supplementary material, table S1 and electronic supplementary material, figure S2). DWT ranged from 2 to 66 cm, representing the gradient from wet hollows to dry hummocks. Sample codes were assigned to portray the DWT gradient after a post-analysis. Cell lengths of 20 300 diatom valves were measured in 41 samples, which were classified into 18 cell-size classes at 2-µm intervals (electronic supplementary material, figure S3). In this dataset, 74 diatom species belonging to 25 genera were identified. From the hollows to the hummocks, the diatoms gradually transitioned from wetness-adapted species (e.g. Eunotia nymanniana, Kobayasiella parasubtilissima) to drought-tolerant taxa such as Chamaepinnularia hassiaca, Eunotia paludosa medium type, and E. paludosa small type (figure 1 and electronic supplementary material, figure S4b). According to diatom cell-size composition, large-sized diatoms were gradually replaced by small-sized diatom from wet hollows to dry hummocks (figure 1). The mean, range and median of cell lengths measured for each sample can be found in electronic supplementary material, figure S3.

Figure 1.

Diatom diagram showing percentages of major cell-size classes (green bars) and diatom species (blue bars) in 41 surface samples, which are displayed along the gradient of the depth to the water table (DWT). Major cell-size class and diatom species with a relative percentage of more than 10% in at least one sample are shown. Sample codes 1–41 correspond to electronic supplementary material, table S1.

Partial RDA revealed that the λ₁/λ₂ ratio of DWT was 0.87 based on diatom size composition and 1.25 based on taxonomic composition (electronic supplementary material, table S5). Ca²⁺ and Si had relatively lower λ₁/λ₂ ratios (less than 0.4) in comparison with DWT. In addition, the unique effect of DWT explained 17.9% and 26.2% of the total variance in cell size and taxonomic composition, respectively. Therefore, DWT was the most important variable for explaining variations in both cell-size composition and taxonomic composition, and it can be used to develop transfer functions.

For both cell-size and taxonomic composition, the ML model has the highest correlation coefficient value R² and the lowest RMSEP, and hence the water-table transfer function can be developed using the ML model. The final ML models with leave-one-out cross-validation based on taxonomic composition and cell-size composition had relatively high R² values (0.78 versus 0.78) but low RMSEP values (6.66 versus 6.87 cm) (electronic supplementary material, tables S3 and S4). The two models had similar R² and RMSEP values, while the final ML model based on cell-size composition had a much lower maximum bias (figure 2).

Figure 2.

Predication performance of ML model based on diatom cell-size composition (a,b) and taxonomic composition (c,d). Red crosses represent samples with greater than 20% residuals.

4. Discussion

Hummocks and hollows are commonly formed in peatlands due to the varying rates of peat accumulation and decomposition in different microhabitats [22]. Hummocks range in height from a few centimetres up to a metre, and the hummock tops are generally dry habitats with low availability of water, nutrients and major cations in comparison with wet hollows [10]. In this study, concentrations of Mg²⁺, Na⁺ and Si were negatively correlated with DWT (electronic supplementary material, table S2), suggesting depletion of inorganic elements on dry hummocks. On hummock tops, water and inorganic elements mainly originate from atmospheric rainfall and deposition, as well as upward delivery from underground water by the capillary action of Sphagnum mosses [23]. In these extreme microhabitats, both diatom taxonomic composition and cell-size composition were significantly correlated with the inorganic elements Ca²⁺ and Si, respectively, and DWT (electronic supplementary material, figure S5). Significant correlations between diatom taxonomic composition and the water table have been observed in peatlands of the Western Carpathian flysch zone [17], northern Canada [15], and China [4,24]. Consistent with previous studies, DWT is the primary environmental factor that influences diatom taxonomic composition in peatlands (electronic supplementary material, figure S5). Generally, aerophytic taxa, such as E. paludosa and C. hassiaca thrive on dry hummocks [13], while wetness-preferring species such as K. parasubtilissima and E. nymanniana are mainly found in wet hollows [25]. In addition, diatom cell-size composition is significantly correlated with Si, which is crucial for building diatom frustules [26,27]. Silica may be more limiting than nitrogen or phosphorus in oligotrophic peatlands [27]. Large-sized (≥ 19 µm in valve length) diatoms are positively correlated with Si concentration in this study (electronic supplementary material, figure S5). Calcium concentration represents an important chemical gradient in peaty environments and has been found to be a determinant of diatom species- and genus-level composition in montane peatlands of central China [28]. Besides diatom taxonomic composition, diatom cell size in peatlands has been found to be responsive to water-table changes [12–14]. Small-sized (less than 19 µm in valve length) diatoms are positively correlated with DWT (electronic supplementary material, figure S5), suggesting that small-sized diatoms enjoy a considerable competitive advantage over large-sized counterparts in dry habitats. For example, small-sized E. paludosa (6–15 µm in valve length) and C. hassiaca (10–13 µm in valve length) have been found to thrive on dry hummocks in the two peatlands.

There are several potential explanations for the relationship between cell-size composition and the water table. Firstly, small-sized diatoms have high surface area to volume ratios, and hence small diffusion boundary layers can make the acquisition of water and nutrients more efficient in the extremely dry habitats of hummock tops [29]. In this study, negative correlations between Mg²⁺, Na⁺, Si and DWT were suggestive of the depletion of cations and silica on hummock tops. This means that limiting supplies of water, cations and silica on hummocks select for small-sized diatoms on dry hummocks. In these extreme environments, diatoms live in close vicinity to Sphagnum cell walls where water and cation exchanges take place [17]. Sphagnum growth is affected by DWT due to its lack of vascular root access to water sources [10], and hence water-table fluctuations can influence diatom cell-size composition indirectly through mediating Sphagnum growth. Secondly, diatoms can respond to environmental changes through morphological alterations by size reduction [30]. The less than 19 µm cell-size classes have their DWT optima above 30 cm, while the ≥ 19 µm cell-size classes have their DWT optima below 30 cm (electronic supplementary material, figure S6). For example, the 7–9 µm cell-size class has a DWT optimum of 44 cm, while the ≥ 41 µm cell-size class has a DWT optimum of 7 cm (electronic supplementary material, figure S6).

Diatom vegetative cell division generally involves a successive diminution in mean cell size as daughter cells are generated by the laying down of daughter thecae back-to-back within the mother cell [3]. The extreme environments of hummocks may not provide ample opportunities for diatoms to undergo sexual reproduction; diatoms undergo successive cell division, leading to the bloom of small-sized diatoms with faster metabolism as a stress-tolerance mechanism [13,15,30]. By contrast, diatom cell size can be restored after sexual reproduction through gamete production, auxosporulation and the production of new silicified initial cells in moist and nutrient-rich hollows [3]. Further laboratory culture experiments are needed to explore the potential mechanisms responsible for diatom cell-size reduction in response to drought.

According to the results of partial RDAs (electronic supplementary material, table S5), DWT was identified as the most significant environmental factor influencing both diatom taxonomic composition and cell-size composition. A λ₁/λ₂ ratio greater than 1.0 indicates that the variable of interest represents an important ecological gradient in the training set [31]. The λ₁/λ₂ ratio of DWT (1.25) for diatom taxonomic composition was greater than the criterion, while the λ₁/λ₂ ratio of DWT (0.87) for cell-size composition was slightly lower than the criterion. This means that more critical assessment should be undertaken including building a transfer function based on diatom cell-size composition [31].

This study used the leave-one-out cross-validation to further verify the transfer function models. After the cross-validation, the ML models had the best performance to infer water-table changes. The final ML models based on taxonomic composition and cell-size composition had the same R², while the former had a lower RMSEP and the later had a smaller maximum bias (electronic supplementary material, tables S3 and S4). This suggests that diatom cell-size composition can be a novel tool to infer water-table changes in peatlands.

Previous studies have revealed that the transfer function generally overestimated DWT at wet sites, but underestimated DWT at dry sites [19], partly resulting from uneven sampling along the water-table gradient, e.g. fewer samples at the upper and lower ends of the water-table gradient. In addition, diatom assemblages are generally dominated by a single aerophytic species (e.g. E. paludosa) on dry hummocks in Sphagnum peatlands [14,15,17]. In an early taxonomic composition-based transfer function, the predominance of E. paludosa with a DWT optimum of 33 cm might lead to the underestimation of DWT at dry sites with DWT > 40 cm [5]. E. paludosa has a relatively large range of valve lengths from 6 to 45 µm [32]. Despite the dominance of the single species, valve lengths of E. paludosa become shorter with an increase in DWT [33]. In the study region, the increase in small-sized valves on dry hummocks was also observed in other dominant species such as E. nymanniana [33] and C. hassiaca (X Chen, S Xu, B Huang 2022, unpublished data). This indicates demographic shifts towards smaller-sized individuals within some dominant species (electronic supplementary material, figure S7). For quantitative estimation of water-table changes, diatom cell-size classes can be an important supplement to taxonomic composition. However, further investigations on valve-length variations within species along the water-table gradient are needed to provide a scientific basis for the routine application of diatom cell-size composition.

In conclusion, this study presents diatom-based water-table transfer function models based on taxonomic composition and cell-size composition, respectively. The ML model had the best performance. The cell-size composition-based ML model had similar performance to the taxonomic composition-based ML model, suggesting that diatom cell-size composition can be a novel indicator of water-table changes in peatlands.

Ethics

This work did not require ethical approval from a human subject or animal welfare committee.

Data accessibility

Data are available from the Dryad Digital Repository: https://doi.org/10.5061/dryad.t1g1jwt9g [34]. https://datadryad.org/stash/share/mhAGkbNuPsRMTgxvz_TnjHfdDJxLFctVMWqEZoJ9Ml8

The data are provided in electronic supplementary material [35].

Declaration of AI use

We have not used AI-assisted technologies in creating this article.

Authors' contributions

S.X.: data curation, formal analysis, investigation, writing—original draft, writing—review and editing; B.H.: data curation, investigation, writing—original draft, writing—review and editing; L.Z.: data curation, investigation, writing—original draft, writing—review and editing; X.H.: data curation, funding acquisition, resources, writing—review and editing; Z.-J.B.: funding acquisition, investigation, resources, writing—review and editing; X.C.: conceptualization, formal analysis, investigation, writing—original draft, writing—review and editing.

All authors gave final approval for publication and agreed to be held accountable for the work performed therein.

Conflict of interest declaration

We declare we have no competing interests.

Funding

This study was supported by the National Natural Science Foundation of China (U23A2003 and U20A2094).