Temporal and spatial variability of terrestrial diatoms at the catchment scale: controls on communities

Study area and weather conditions

Diatom sampling was conducted in the Attert River basin, which covers an area of approximately 249 km² and is located for the most part in Luxembourg with the exception of the western part, which is located in Belgium (49°46′13.0″ N, 5°59′9.2″ E). The topography in the north of the basin is typified by a sub-horizontal plateau with elevations ranging between 450 m and 500 m and steep slopes, whereas the southern part has gently sloping lowlands at 300 m a.s.l. (Pfister et al., 2015). Moreover, the area has a complex geological setting. Keuper sandy marls and Jurassic sandstone dominate the southern part of the basin, while the northern part consists of Devonian schists and Luxembourg red sandstone (“Buntsandstein”). Recent alluvial depositions are present in the vicinity of streams. Soil types generally follow the bedrock geology. Cambisols dominate the schists area, whereas Stagnosols and Planosols are the dominant types on marls. Leptosols, Arenosols and Podzols characterize the Luxembourg red sandstone (Cammeraat et al., 2018). Land cover in the basin consists of farmland, grasslands and forested areas (Fig. 1). Farmland is mainly used for growing corn, wheat and rapeseed, while grasslands are predominantly used for cattle grazing. European beech (Fagus sylvatica L.) and spruce (Picea abies L.) dominate the forested areas. The climate of the basin is semi-oceanic.

The sampling campaign was performed monthly from October 2017 to November 2018. The summer prior to sampling was relatively wet (on average 88 mm/month), whereas the average monthly precipitation during the sampling period was 56.4 ± 33 mm. The maximum monthly precipitation was observed in December 2017 and January 2018 (respectively 111.9 mm and 123.5 mm). Monthly precipitation reached a minimum in October 2018 (18 mm). July and August were the warmest months, reaching average temperatures of 20.5 °C and 18.7 °C, while most freezing days were encountered during February and March (28 and 12) (Data retrieved from Administration des Services Techniques de l’Agriculture (ASTA)). Overall, 2018 was characterized by a cold winter and an exceptionally dry and warm summer period.

Sample collection, preparation and diatom identification

In order to capture the intra-annual and spatial variability of diatom communities, soil diatom samples were taken at the soil surface in 16 sites around the catchment every 3–5 weeks for a period of 14 months (n = 224 samples). In order to systematically cover the whole area, a grid of 5 × 5 km was mapped on the study area (Fig. 1). Next, at least one fixed sampling point per cell was chosen considering the differences in geology, soil and land use features, and previous diatom research in the area (Antonelli et al., 2017; Barragán, Wetzel & Ector, 2018). At each sampling campaign, a description of the site was made in order to keep track of the environmental changes in between the sampling periods. Characteristics such as type of land use, disturbances, height of the lower vegetation (i.e., indicator for mowing/grazing) were noted. Also, one sample per site was taken for diatom analysis. According to Barragán, Wetzel & Ector (2018), this should be sufficient to have a diatom community representative for a homogeneous area of at least 75 m².

We used metal rings (diameter: 5.6 cm) to collect small soil cores. Upon arrival in the lab, diatoms were extracted by rinsing the soil surface with MQ water until a 50 mL falcon tube was filled (see Barragán, Wetzel & Ector, 2018). The samples, consisting of soil particles and water, were then fixed with 70% ethanol. Diatom slides were prepared for microscopic counts following the CEN 13946 procedure (European Committee for Standardization, 2003). A minimum of 200 valves were counted and identified on each slide along random transects using a Leica DMR light microscope with a × 100 oil immersion objective and a magnification of × 1,000 (Bate & Newall, 2002).

Diatom identifications were based on following taxonomic references: Ettl & Gärtner (2014); Hustedt (1941); Krammer (2000); Lange-Bertalot (2001); Lange-Bertalot et al. (2017); Levkov, Metzeltin & Pavlov (2013); Wetzel et al. (2015, 2013).

For scanning electron microscope (SEM) analysis, parts of the oxidized suspensions were filtered and rinsed with additional deionized water through a three µm Isopore^™ polycarbonate membrane filter (Merck Millipore). Filters were mounted on aluminum stubs and coated with platinum (30 nm) using a BAL-TEC MED 020 Modular High Vacuum Coating System for 30 s at 100 mA. A Hitachi SU-70 ultra-high-resolution analytical field emission SEM (Hitachi High-Technologies Corporation, Tokyo, Japan), operated at 5 kV and with a 10 mm working distance, was used for the analysis. SEM images were taken using the upper (SE-U) detector signal. Photomicrographs were digitally manipulated and plates containing light and scanning electron micrographs were created using CorelDraw X8. For a more detailed and complete description of the methodology, we refer to Barragán, Wetzel & Ector (2018).

Statistical analysis

Before carrying out the statistical analysis, the species dataset was reduced by only keeping the taxa with a relative abundance of at least 2% in at least two samples or 4% in one sample. Samples which contained less than 100 valves were removed from the analysis (18 of the 224 samples). Of the 18 samples, 12 were taken during the dry summer period and in the aftermath of it (July: 1, August: 3, September: 5, October: 3). A Detrended Correspondence Analysis (DCA) was performed to see how communities relate to each other and to estimate the length of the gradient (Hill & Gauch, 1980). The analysis revealed a gradient longer than 5.0 SD (Standard deviation of species turnover). Also, the non-parametric statistical test Analysis of Similarities (ANOSIM) using Bray–Curtis as a dissimilarity metric, was used to test if communities significantly differ between land use types (Clarke, 1993). Temporal variation was assessed using the same analysis performed for each type of land use. Significance of both analyses were tested with the default Monte Carlo permutation test (perm-999). The effect of geographical distance on the species composition of the communities was assessed with a Mantel test. Species data were transformed (Bray–Curtis) to a dissimilarity matrix and tested if there was significant correlation with the geographical distance matrix. The latter was obtained using the rdist.earth function from the R-package fields (Nychka et al., 2017).

The characteristic species and their indicator value for the different land uses (i.e., agricultural field, grazed grassland, agricultural grassland, undisturbed grassland and forest) were assessed using the option IndVal.g, which is implemented in the multipatt function included in the indicspecies package. This function takes into account species presences/absences and relative abundances per type of land use (De Cáceres, Legendre & Moretti, 2010). Also, the Shannon–Wiener index was computed for each sample with the function diversity. Additionally, IPS values were calculated (Cemagref, 1982) for the five different land uses to check if differences in communities could be expressed by a single value. This index takes into account the abundance of each species in the sample, their sensitivity value (i.e., sensitivity to pollution) and their indicator value (i.e., relative probability of each species to occur in one of five saprobity classes). Calculations for both, Shannon–Wiener and IPS index, were done on the whole dataset of soil diatom species. The IPS indices were computed with the software Omnidia (version 6.0.8, 2018) (Lecointe, Coste & Prygiel, 1993) and compared with ANOVA (analysis of variance). IPS was chosen for this study, since it is the index with the higher number of taxa included and the most updated. Furthermore, a CONISS cluster analysis was done for sites 9, 10 and 13 to see when diatom community composition changed following land management practices. This was conducted with the function strat.plot in the Rioja package (Juggins, 2017). All aforementioned statistical analyses were performed using the R statistical program (R v. 3.5.0.; http://www.r-project.org/) and additional functions from the R-package vegan version 2.5–4 (if not mentioned differently).

Seasonal variation

Our research showed that seasonal variation in soil diatom communities was rather limited or almost absent when looking at the communities in the forest sites. A reason behind this stationarity could be that forests create a relatively stable microhabitat for diatoms. They exhibit generally a higher and more constant moisture than grasslands (James et al., 2003). The temperature fluctuations are usually smaller and maximum temperatures are lower. Also, the forest canopy provides a good protection against direct UV-radiation, which is considered to be harmful to algae (Ress, 2012), and ensures that the topsoil layer dries out less quickly. Although, forest soils are receiving high amounts of leaf litter and seasonal changes resulting from that (i.e., decomposition and mineralization processes), but this does not seem to have any visible effect on soil diatoms. Similar to forests, several spring types have shown to create a stable environment for aquatic diatoms (i.e., fairly constant pH, temperature and trophy level), leading to limited seasonal changes in the communities (Cantonati, 1998). Cantonati, Gerecke & Bertuzzi (2006) therefore concluded that this actually facilitates the use of diatoms as ecological indicators since fewer numbers of surveys are required. The same principle could be applied to forest soils, meaning that sampling once a year should be sufficient.

In addition to the minor seasonal influences in forests, we did not observe any significant temporal variation in the other habitat types, except for communities living on agricultural fields. There, we observed a general change from Mayamaea to Hantzschia dominated communities. However, this variation could be explained by farming practices such as plowing and manuring, since the compositional changes in the communities coincides with the beginning (April and May) and ending (October and November) of the “farming season.” Nevertheless, we noticed some small signs of seasonality, especially by some species who are strongly dependent on soil moisture content (e.g., N. pusilla (J. Foets, 2018, unpublished data)). Also, the fact that we experienced an exceptionally dry and warm summer period during the sampling campaign, did not help to observe potential seasonality since sufficient moisture availability is essential for diatom survival and reproduction (Van de Vijver, Ledeganck & Beyens, 2002). So, seasonality in diatom community composition could occur in (disturbed) grassland soils, but it is rather unlikely that it will play a significant role in the composition of soil diatom communities.

Habitat variation

We found that the type of land use was a key factor in defining the species composition. This result is consistent with previous works that showed different communities with different levels of disturbance (Antonelli et al., 2017; Stanek-Tarkowska et al., 2018; Stanek-Tarkowska & Noga, 2012; Vacht et al., 2014). In contrast to these works, we compared two to three habitat types more at the same time. Although our analysis could not split the communities sampled on agricultural grasslands from the ones taken on undisturbed grasslands, our results still show that species are perhaps more sensitive and indicative for anthropic disturbances than we previously thought. Indeed, we were able to define several indicator species for each type of land use and when comparing those species with earlier works, we noticed many similarities. For instance, Barragán, Wetzel & Ector (2018) also reported S. lundii from a meadow and P. perirrorata and Eunotia spp. from a forest site, whereas Stanek-Tarkowska et al. (2013); Stanek-Tarkowska & Noga (2012) observed S. borichii in high abundances on farmland. This habitat specificity of soil diatoms is an interesting outcome and should be explored more often in the future, certainly in the light of using soil diatoms as environmental markers.

Another interesting outcome of this study was the follow-up of the communities after a drastic change in the landscape following land management practices. Two of our 16 sites (9 and 10) went from an agricultural grassland to an agricultural field and one site (13) was transformed from an undisturbed to a grazed grassland. Generally, species colonizing the early stages of a secondary succession have fast immigration rates rather than high reproduction and growth rates (Stevenson, 1986; Stevenson et al., 1991). This is also what we observed in the three cases with Pinnularia borealis Ehrenberg, H. amphioxys and H. abundans often replacing N. pusilla and Mayamaea spp. The former three taxa are very common, large terrestrial diatoms (±75 µm) and are known to be well-adapted to living in extreme environments (Dodd & Stoermer, 1962; Pipe & Cullimore, 1984; Stock et al., 2018; Van Dam, Mertens & Sinkeldam, 1994). These diatoms are considered to be slow growers, since they spend a large part of their energy in developing complex structures to withstand (very) harsh conditions (Pinseel et al., 2019; Ress, 2012). For example, Hantzschia spp. create internal valves to better retain moisture and decrease water loss (Dodd & Stoermer, 1962; Round, Crawford & Mann, 1990). On the other hand, N. pusilla and Mayamaea spp. are rather small (10–15 µm) and more fragile species (i.e., less silicified cells). We therefore expect them to grow and reproduce faster, but they need preferably a more stabilized habitat. Consequently, changes in land use could have large, similar effects on the community, even in (semi-) disturbed areas.

Besides the identification of important pioneer species, we also noticed that soil diatom communities need between 1 and 2 months to colonize a new, disturbed habitat depending on the prevailing environmental conditions. This recolonization period is longer compared to freshwater communities. They generally need 14 days to 1 month to establish a new, stable community (Blinn, Fredericksen & Korte, 1980; Oemke & Burton, 1986). We stipulate that this difference is mainly due to three reasons: (1) the absolute abundance of soil diatoms is much lower compared to freshwater diatoms (Loescher, 1972) (2) water is a more efficient dispersal mechanism than wind (Kristiansen, 1996) and (3) nutrient exchange via osmosis and reproduction is easier in aquatic conditions. However, it is to be noted that some shifts happened during exceptionally warm, dry weather conditions and therefore recolonization could be shorter in more optimal (moist) conditions. Unfortunately, we were not able to monitor a second succession stage to see if communities go back to their previous composition or to a different one. So, the reestablishment of a stable community on disturbed terrestrial habitats generally happens after 1–2 months with the more tolerant species replacing the rather sensitive ones.

IPS and future modifications

Despite that the IPS index was developed for water quality assessment, it works very well in separating the different land use types as it is shown here and previously by Antonelli et al. (2017) and Barragán, Wetzel & Ector (2018). Nevertheless, this index should be adapted to terrestrial environments if it is to be used further. In the first place, we should be able now (i.e., with the recent results) to fill in the missing indicator and pollution values for, at least, the most abundant and common terrestrial diatom taxa. The IPS index was chosen, because it is species-specific and is supported by a large and updated database. However, some important terrestrial diatoms are not incorporated in it. For instance, Antonelli et al. (2017) could not include certain abundant species in their analysis, while some of our IPS values were calculated using only 60% of the total valves counted. Secondly, there is a need to update and reevaluate the indicator and sensitivity values, since changes in diatom taxonomy happen very fast leading to new names and/or species descriptions. This in turn leads then to different results in quality indices calculations (Kahlert et al., 2016). Also, both values are inferred from aquatic surveys and many of them are probably not reflecting the behavior of soil diatoms in terrestrial habitats (Antonelli et al., 2017). Finally, we should look at incorporating an additional factor to make the index more specific for terrestrial habitats and eventually decrease the intra-variability and increase the inter-variability between the land use types. For instance, by multiplying that factor it will give more weight to the very specific indicator taxa. Hopefully, with those modifications we will be able to predict anthropic disturbances more precisely in the future.