You are on page 1of 17

Ecological Indicators 155 (2023) 110952

Contents lists available at ScienceDirect

Ecological Indicators
journal homepage: www.elsevier.com/locate/ecolind

Spatial distribution of surface-sediment diatom assemblages from 45


Tibetan Plateau lakes and the development of a salinity transfer function
Siwei Yu a, b, c, Junbo Wang a, *, Kathleen M. Rühland b, Liping Zhu a, c, Jianting Ju a, Chong Liu a, d,
Qingfeng Ma a, John P. Smol b
a
State Key Laboratory of Tibetan Plateau Earth System, Environment and Resources (TPESER), Nam Co Observation and Research Station (NAMORS), Institute of
Tibetan Plateau Research, Chinese Academy of Sciences, Beijing 100101, China
b
Paleoecological Environmental Assessment and Research Laboratory (PEARL), Department of Biology, Queen’s University, Kingston, Ontario K7L3N6, Canada
c
University of Chinese Academy of Sciences, Beijing 100049, China
d
Piesat Information Technology Co., Ltd, Beijing 100195, China

A R T I C L E I N F O A B S T R A C T

Keywords: The Tibetan Plateau (TP) is often referred to as the “Asian Water Tower”, as it plays a crucial role in supplying
Diatoms water for ~1.9 billion people. Lakes in this arid mountain region have a wide salinity gradient and are sensitive
Tibetan Plateau to the complex interactions between climate and the water cycle, affecting aquatic ecosystems and species
Spatial distribution
distributions. Here, we analyzed relationships between measured environmental variables and the spatial dis­
Salinity
Transfer function
tribution of surface sediment diatom assemblages from 45 lakes located throughout the TP with the intention of
developing a diatom-inference model for salinity. Diatom assemblage composition and diversity differed among
lakes from the western (W), southern (S), northern (N), and southeastern (E) regions of the TP. The most common
taxa observed in the lake set, Pantocsekiella ocellata, Nitzschia palea, and Staurosira venter, tended to have
widespread distributions across the TP. In general, diatom diversity and richness tended to be highest in lakes
with salinity concentrations between 0.1 and 10 g/L and were lowest in high salinity lakes (>20 g/L) of the N
region. Canonical correspondence analysis (CCA) identified lake water salinity and maximum lake depth as
explaining a statistically significant portion of diatom assemblage variation. Weighted-averaging (WA) was used
to develop a diatom-based salinity inference model (R2jack = 0.71, RMSEPjack = 0.28) based on diatom optima
from freshwater to meso-hyposaline lakes of the TP. Diatom-salinity reconstructions and changes in diatom
community composition can potentially be used to track long-term patterns in climate and environmental
conditions across the highly complex mountain landscape of the TP.

1. Introduction the numerous hydrologically closed basin lakes that are common to the
region. These characteristics play a significant role in lake ecosystem
The Tibetan Plateau (TP), also known as “the Third Pole” (Qiu, 2008) structure, diversity, productivity (Wischnewski et al., 2011; Anslan
and “the Asian Water Tower” (Immerzeel et al., 2010), plays an et al., 2020) and in aquatic community composition including diatom
important role in the Earth’s climate system and supplies a critical water species distributions (Wang et al., 2011).
resource for ~1.9 billion people (Immerzeel et al., 2020; Yao et al., Understanding the responses of aquatic organisms to multiple
2012). Because of its location, the TP is highly vulnerable to water stress, stressors and identifying their relationship to key environmental vari­
governance, hydropolitical tensions, and to future climatic and socio­ ables is vital to advancing global freshwater biodiversity research in the
economic changes (Immerzeel et al., 2020). This arid to semi-arid and context of environmental change (Birk et al., 2020; Maasri et al., 2022).
topographically complex mountain region is situated far inland resulting In the TP and many other remote alpine regions, the complexities of
in high evaporation rates and a wide gradient of environmental vari­ aquatic community structure and diversity, as well as biogeographic
ables across lakes (Liao et al., 2020), including salinity. Salinity is gradients, are poorly known (Kärnä et al., 2018). Examining past
important for characterizing the limnology of TP lakes, particularly in (baseline) environmental conditions will provide context to better

* Corresponding author.
E-mail address: wangjb@itpcas.ac.cn (J. Wang).

https://doi.org/10.1016/j.ecolind.2023.110952
Received 30 May 2023; Received in revised form 9 September 2023; Accepted 12 September 2023
Available online 19 September 2023
1470-160X/Crown Copyright © 2023 Published by Elsevier Ltd. This is an open access article under the CC BY license
(http://creativecommons.org/licenses/by/4.0/).
S. Yu et al. Ecological Indicators 155 (2023) 110952

understand contemporary ecological patterns and species distributions mean annual temperature ranging from − 15 ◦ C to 5 ◦ C, relatively short
(Seddon et al., 2014). For example, paleolimnological techniques can be growing seasons and long periods of frost. Most areas of the TP are
used to establish ecological and chemical baseline conditions, assess characterized by an arid/semiarid climate, with annual precipitation
how much the environment has changed relative to the baseline state, ranging from tens to hundreds of millimeters, mostly occurring as rain
and have the potential to qualitatively and quantitatively reconstruct from May to September (Gao and Liu, 2013). The southern and south­
past environmental conditions (Bennion et al., 2004; Battarbee et al., eastern TP has comparatively high rainfall, with mean annual precipi­
2005; Smol, 2019). Here, we explore the modern distributions of diatom tation ranging from 370 mm to over 1800 mm (Lu et al., 2011).
assemblages using a calibration set of lakes in the TP to better under­ Lakes on the TP are relatively isolated with minimal human activ­
stand spatial patterns of environmental variables and diatom species ities. Many TP lakes are hydrologically closed systems, a characteristic
composition in this complex alpine region. We analyse the relationships that is known to have important implications on water quality and
between diatom assemblages and key environmental variables to aquatic biota. With the exception of the southeastern region, the vege­
explore the development of a diatom-based salinity transfer function for tation of the TP is dominated by meadow, steppe, and shrub commu­
the TP. nities. TP vegetation is influenced by climate and altitude and is
Quantitative reconstructions of environmental variables are based distributed along a gradient of decreasing precipitation, changing from
on the relationships between the modern distributions of biotic proxies forest to meadow/shrub, to steppe, then to desert with a zonal pattern
and measured environmental variables that can then be used to estimate from southeast to northwest (Institute of Geography, 1990; Hou, 2001).
past lake-water conditions using subfossil assemblages preserved in lake Vegetation in the eastern and southeastern TP is dominated by montane
sediment cores (Imbrie and Kipp, 1971; Birks et al., 2010). One of the coniferous and broad-leaved forests, whereas in the southern and east-
most commonly used and powerful indicators of environmental change central TP, vegetation is dominated by shrubs (Institute of Geography,
are the siliceous algae, diatoms (Bacillariophyceae). They are successful 1990; Ma et al., 2019). Alpine meadows are mainly distributed in the
paleo-indicators because their cell walls typically preserve well in lake eastern TP above 4000 m a.s.l. and dominated by sedge and grass
sediments and they respond quickly and sensitively to environmental communities. The vegetation of the central and southern TP is domi­
change (Smol 2009; Smol and Stoermer, 2010). Over the past few years, nated by alpine steppe between 4400 and 5400 m a.s.l., and is mainly
several studies from the central and eastern TP have examined the comprised of Poaceae (e.g., Stipa purpurea and S. basiplumosa) and
geographic distribution of diatoms in response to recent climate change Asteraceae (e.g., Artemisia) (Institute of Geography, 1990). Alpine
(Yu et al., 2019; Wang et al., 2020a, Wang et al., 2020b). However, there shrub, one of the dominant vegetation types on the TP, are distributed
is a lack of diatom studies in the western, southern and northern parts of throughout and in large tracts across the alpine forest and meadow
the TP. The scarcity of data is understandable given that these high- transition zones (Zhao et al., 2006).
altitude regions are remote and weather conditions are often harsh,
making fieldwork logistically challenging (Qiu et al., 2008). Examining 3. Methods
the spatial distributions and the environmental optima and tolerances of
diatom taxa in understudied aquatic systems from remote regions will 3.1. Sample collection
help to better understand and interpret environmental changes, partic­
ularly in the context of climate change (Benito et al., 2018; Rivera- The 45 lakes were sampled between 2009 and 2019 and are mainly
Rondón and Catalan, 2020). To date, the majority of diatom-based located in the western, southern, northern, and southeastern TP (Fig. 1).
paleolimnological studies of the TP have focused on qualitative in­ The lakes span a latitudinal gradient of 28.96 ◦ N to 35.9 ◦ N, a longitu­
terpretations of environmental change during the past few centuries (e. dinal gradient of 79.36 ◦ E to 98.85 ◦ E and an altitudinal gradient of
g., Wang et al., 2011; Wischnewski et al., 2011a) to millennia (Li et al., 4240 m a.s.l. to 5220 m a.s.l. These lakes were strategically selected to
2015; Wang et al., 2018; Liao et al., 2020; Su et al., 2022). cover a large geographic area with the aim of acquiring a large gradient
Diatom assemblages show strong relationships to salinity, and of salinity, and to include some of the larger and deeper lakes that are
diatom-based models have been developed to quantitatively track past found in the western and central parts of TP. Surface sediment samples
variations in lake water salinity (reviewed in Fritz et al. 2010). Recon­ were collected from the deepest part (center) of each lake (Fig. 1) using a
structing salinity is particularly appropriate in semi-arid to arid regions Hydro-BIOS Bottom Sampler acc. to Ekman-Birge. The top 2 cm of
of the world where lakes are typically subjected to relatively high sediment was carefully sub-sampled and was used to represent modern
evaporation rates, making them highly susceptible to climate change environments. Based on the few published 210Pb-dated sediment cores
(Cumming and Smol 1993; Wilson et al. 1994; Cumming et al., 1995; from the TP region, including 39 lakes from Wang et al. (2020) and six
Bloom et al. 2003). However, only a few diatom-based salinity models lakes from Yang and Turner (2013), our surface sediment samples are
have been developed from lakes in the eastern Tibetan Plateau (Wang estimated to represent a few years to twenty years. At each lake, water
et al. 2011), and therefore more calibration studies are required to better samples were collected with a polyethylene bottle from a depth of 1-m at
represent this large and highly complex mountain region represented by the same time and location that the surface sediments were sampled. In
a large gradient of salinity. Here, we examined the distribution of the field, water samples were stored in a cooler until sent to the Institute
modern diatom assemblages from the surface sediments of 45 lakes of Tibetan Plateau Research, Chinese Academy of Sciences for analyses.
throughout the TP that cross gradients of altitude, lake depth, pH and Major cations (Na+, K+, Mg2+, Ca2+) and anions (Cl− , SO2− 4 ) were
salinity. These data were then used to develop a diatom-based inference measured using a Dionex-ICS 900 Ion Chromatography System. HCO–3
model that can potentially reconstruct past salinity trends across the TP. concentrations were estimated from the ion balance of major cations and
anions (Murakami et al., 2007). Water temperature, maximum lake
2. Study region depth, pH, turbidity, chlorophyll a, and salinity were measured in the
field using a Yellow Springs Instrument (YSI) EXO2 multiparameter
The TP is the highest and most extensive plateau in the world and is sonde (YSI Incorporated, Yellow Springs, OH, USA).
characterized by a highly complex and heterogeneous landscape that is For the purposes of this study, the TP region was divided into four
sensitive to climate change and other environmental stressors (Liao geographic categories (Fig. 1) based on atmospheric circulation and
et al., 2020). It spans relatively large longitudinal (26.00◦ N-39.78◦ N) moisture conditions (Chen et al., 2010), permafrost development (Wang
and latitudinal (73.32◦ E-104.78◦ E) gradients, with the majority of the et al., 2019), and lake salinity. The western (W) TP is dominated by
region located at elevations above 4500 m a.s.l., covering an area of westerly winds (Westerlies) (An et al., 2012), and the eastern (E) TP is
approximately 2.5 × 106 km2 (Yao et al., 2012). The TP is characterized dominated by the Asian Summer Monsoon. The boundary between the
by cold-dry winters and cool-humid summers (Lu et al., 2011) with Northern (N) TP and Southern (S) TP was based on the modern position

2
S. Yu et al. Ecological Indicators 155 (2023) 110952

Fig. 1. Map showing the location of the 45 study lakes in the Tibetan Plateau (TP) relative to the Westerlies and Indian Summer Monsoon (ISM). The study lakes are
categorized into four geographic regions within the TP (W, N, S, E) based on differences in atmospheric circulation and permafrost development (from Chen et al.,
2010; Wang et al., 2019) and are delineated by the blue dashed lines. (For interpretation of the references to colour in this figure legend, the reader is referred to the
web version of this article.)

of the Asian Summer Monsoon limit (Chen et al., 2010), the spatial diatom valves were counted using a Leica DMR HC light microscope
distribution of permafrost development (Wang et al., 2019), and lake using oil immersion at a magnification of 1000 X and fitted with dif­
salinity concentrations (the salinity range of lakes in the S region ferential interference contrast optics. Diatom identifications were made
(average salinity: 2.4 g/L) is generally smaller than that of lakes in the N to the lowest taxonomic level possible (often variety) using a selection of
region (average salinity: 23.5 g/L) (Fig. 2). taxonomic sources, mainly Krammer and Lange-Bertalot (1986, 1988,
1991a, 1991b), Algaebase (Guiry and Guiry, 2023: algaebase.org) and
the Diatoms of North America online source (diatoms.org). Diatom
3.2. Diatom analysis
counts were expressed as percent relative abundances.
Diatom samples were prepared using standard preparation methods
at the Paleoecological Environmental Research Laboratory (PEARL), 3.3. Numerical analyses
Queen’s University (Canada) that closely followed Pienitz and Smol
(1993). Briefly, approximately 0.05 g of freeze-dried sediment from each Prior to numerical analyses (with the exception of diatom richness
interval was placed into glass scintillation vials and 15 mL of a 50:50 M and diversity), diatom species relative abundances were square-root
ratio of concentrated sulfuric (H2SO4) and nitric (HNO3) acid was added transformed to downweight the influence of the dominant taxa (Leg­
to digest the organic material. The samples were placed into a hot water endre and Gallagher, 2001). The physical, chemical and geographical
bath and heated to ~ 80 ◦ C for at least 2 h and then left to settle for environmental variables were tested for normality using the Shapiro-
approximately 24 h. The supernatant was removed, and the slurries Wilk test (Shapiro and Wilk, 1965; R package ‘vegan’ v2.5–7, Oksanen
were then rinsed with distilled water. This process was repeated until a et al., 2020) (except for pH as this was already on a logarithmic scale)
litmus test indicated that the samples were no longer acidic. The diatom and, where needed, were log10(x + 1) transformed to correct skewness.
slurries were strewn onto coverslips at four different dilutions and To examine the relationships among environmental variables, a Pearson
allowed to evaporate before being mounted onto glass microscope slides correlation matrix was examined and significantly correlated variables
with Naphrax (refractive index of 1.74). For each sample, at least 300 (p < 0.05, p < 0.01, and p < 0.001) were identified using the R software

3
S. Yu et al. Ecological Indicators 155 (2023) 110952

Fig. 2. Comparison of frequency distributions (box and whiskers plots) of key measured variables in the 45 Tibetan Plateau study lakes including: a) altitude, b)
maximum lake depth, c) lake temperature, d) pH, e) chlorophyll a concentration, and f) salinity across the four geographical regions. In each box plot, the vertical
lines represent the range in measured values and the horizontal lines depict the median values.

(R Core Team, 2018) and plotted using corrplot. Spatial variations of the the CCA (Ter Braak and Prentice, 2004). To determine whether any of
diatom diversity measurements (richness) made from the different re­ the 45 TP lakes were outliers in terms of both environmental variables
gions in our lakeset (i.e., S, N, W and E regions of the TP) were explored and diatom species composition, PCA axes 1 and 2 samples scores
using the Shannon Index and implemented using Past software (Hammer (environmental variables) and DCA axes 1 and 2 (species scores) were
et al., 2001). examined. If the sample scores for both axes of a given lake were outside
All ordination analyses including principal component analysis one standard deviation of the mean sample scores of the lakeset in both
(PCA), detrended correspondence analysis (DCA), detrended canonical PCA and DCA (Fig. S1), that lake would be considered an outlier (Hall
correspondence analysis (DCCA), and canonical correspondence anal­ and Smol, 1992). Based on these tests, Mang Co (located far from the
ysis (CCA) were performed with CANOCO version 5.0 (Ter Braak and other lakes) was considered a statistical outlier, and therefore model
Šmilauer, 2012) and were based on diatom taxa that had a relative development was based on the remaining 44 TP lakes.
abundance >1% and occurred in at least two lakes (total of 125 diatom A CCA with forward selection (Ter Braak and Šmilauer, 2012) was
taxa). PCA was used to explore the relationships between environmental performed, together with a Monte Carlo permutation test, to identify
variables and to identify important environmental gradients in the TP which of the remaining variables explained a statistically significant
lakes. DCA using the options of detrending by segments, nonlinear amount of variation in the diatom data. To explore relationships among
rescaling, and no downweighting of rare taxa was used to determine environmental variables and diatom taxa, a final CCA was performed
whether linear (RDA) or unimodal (CCA) direct ordination analysis with the variables that were identified as being significant run actively,
should be used. A DCA gradient length >4 SD units suggests a unimodal and additional variables of interest that were not deemed significant,
model is most appropriate (Birks, 1995; Ter Braak and Šmilauer, 2012) run passively in the ordination.
and because the gradient length for the TP lakes was 5.6 SD units, a CCA To determine the species gradient length for the forward selected
was used to explore the relationships among diatom taxa and measured variable of interest (e.g., salinity), a detrended CCA with salinity was
environmental variables. used to determine its ability to maximize the dispersion of diatom scores
To identify variables that displayed multiple correlations with other (Ter Braak and Šmilauer, 2012). To assess whether the forward selected
variables, CCA was used to examine high variance inflation factors variables were statistically significant, a CCA constrained to that vari­
(VIFs): environmental variables with VIFs >10 were eliminated in able (i.e., sole explanatory variable) together with a Monte Carlo per­
sequence, starting with the variable having the highest inflation factor in mutation (999 permutations, p > 0.05) was examined (Ter Braak and

4
S. Yu et al. Ecological Indicators 155 (2023) 110952

Šmilauer, 2012). A variety of inference models were explored for diatom assemblage composition was assessed by examining the ratio
reconstructing salinity based on modern diatom distributions across the (λ1/λ2) of the constrained eigenvalue (λ1) to the first unconstrained
measured salinity gradient including Weighted Averaging (WA), Locally eigenvalue (λ2) of a DCCA with salinity included as the sole predictive
Weighted Weighted Averaging (LWWA), Maximum Likelihood (ML), variable (Ter Braak, 1992). Estimated salinity optima were shown only
Partial Least Squares (PLS) and Weighted Averaging Partial Least for diatom taxa that had an effective number of occurrences (Hill’s N2)
Squares (WAPLS). Jackknifed cross-validation procedures were used to > 4 because taxa that are not well represented in the calibration set will
estimate prediction errors and test the predictive ability of different not have reliable optima estimates (Hill, 1973; Cumming and Smol,
transfer functions (Birks, 1995). For WAPLS model options, each new 1993). Optima and tolerances of diatom taxa and transfer functions were
component was tested and retained only if it resulted in at least a 5% determined using the software program C2 version 1.5 (Juggins, 2007).
reduction in the prediction error (in cross-validation) of the root mean
square error of prediction (RMSEP) (Birks et al., 1998). WA and WAPLS
regression were used to estimate the relationship of each taxon in the
calibration set to measured salinity providing estimates of species op­
tima. The reliability of the salinity inference model as a predictor of

Fig. 3. Piper diagram showing cation and anion concentrations of the 45 TP lakes, categorized by geographic region (colors of circles). The lower left ternary
diagram includes the concentrations of sodium, calcium, and magnesium, and the lower right ternary diagram shows cation concentrations of carbonate sulfate,
bicarbonate, and chloride. The diamond plot illustrates lake types based on the combined cation and anion concentrations.

5
S. Yu et al. Ecological Indicators 155 (2023) 110952

4. Results (Table 1).


PCA analyses revealed that the PCA axis 1 (λ1 = 0.48) and PCA axis 2
4.1. Trends in environmental variables (λ2 = 0.13) captured 61% of the variation in the environmental data
(Fig. 4). The strongest direction of variation (PCA axis 1) was a gradient
The 45 study lakes encompass a broad range in physical character­ of major ions (Na+, K+, Mg2+, Cl− , and SO2− 4 ) and the second environ­
istics including maximum lake depth (0.5 m to 109.6 m, median = 13 m) mental gradient (PCA axis 2) was characterized by positive correlations
(Fig. 2b) and lake surface area (LSA) (3.1 km2 to 2420 km2, median = with latitude and altitude and by negative correlations with HCO–3, pH
44 km2) (Table S1). There were 19 lakes (<10 m, median = 1 m), 21 and maximum lake depth (Fig. 4). Lakes located in the N region (with
lakes between 10 m and 50 m (median = 25 m), and 5 lakes that were the exception of sites 33 and 44) tended to group together on the left side
deeper than 50 m (median = 83.5 m) (Table S1). The relatively large of axis 1 indicating higher concentrations of salinity, Na+, K+, Mg2+,
altitudinal gradient of this study (4200 m a.s.l. to 5200 m a.s.l.: Fig. 2a) Cl− , and SO2-
4 (Fig. 4), in contrast to the majority of lower salinity lakes
included lakes situated both above and below the timberline. Lakes in in the W region that grouped on the right side of PCA axis 1 (Fig. 4).
the W region spanned the entire altitudinal range sampled in the study
with the highest median elevation (4900 m a.s.l: Fig. 2a), whereas lakes Diatom trends and relationships with environmental variables and
in the S and N regions were generally located at lower and higher alti­ geographic regions
tudes (median elevation of 4636 m a.s.l, and 4900 m a.s.l., respectively)
(Fig. 2a). In general, the TP lakes were alkaline (median pH = 9.1, range A total of 142 diatom taxa representing 48 genera were identified in
= 7 to 10.5) with a similar median value in the W, S and N regions the 45 TP lakes of which 125 taxa (relative abundance of at least 1% in 2
(Fig. 2d). Water temperature was highly variable ranging from 0 ◦ C to lakes) were retained for ordination analysis and salinity transfer func­
20.1 ◦ C, with the greatest range in the N region (Fig. 2c). Measured tion development. Taxonomic trends across geographic regions were
chlorophyll a concentrations were similar in lakes across geographic examined for 37 of the most common diatom taxa that had relative
regions with a few exceptions that were higher, ranging from 0.3 µg/L to abundances > 15% and each of the three most common taxa had a high
10.5 µg/L (median = 2 µg/L; Fig. 2e). optimum in different lake types (Fig. 5). Spatial patterns of the relative
One of the goals of this study was to examine salinity across the TP abundances of the three most commonly occurring diatom taxa observed
and therefore our sampling strategy aimed to target lakes along a in our lakes (Pantocsekiella.
gradient of salinity concentrations. The study lakes spanned a range in ocellata, Staurosira venter, and Nitzschia palea) were examined across
salinity from 0.1 g/L to 85.3 g/L, although most lakes were below 40 measured gradients of altitude, maximum lake depth, water tempera­
mg/L (median = 15 mg/L) (Fig. 2f). Lakes in the N region had the widest ture, pH, and chlorophyll a (Fig. 6). The WA estimates of salinity optima
range in salinity with most lakes having concentrations>3 g/L. In were only reported for 35 diatom taxa that had a Hill’s N2 > 4 (Table 6).
contrast, the majority of lakes in the W and S regions had salinity con­ The majority of diatom taxa in the surface sediments of the 45 TP
centrations lower than 2 g/L with notably narrower ranges than lakes in lakes were benthic, including Navicula sensu lato (s.l.) spp., Nitzschia spp.
the W region (Fig. 2f). Based on commonly applied salinity categories (e.g., N. fonticola, N. frustulum, and N. palea), Tryblionella spp.
(Hammer, 1986; Meybeck, 1995), the 45 lakes in this study would be (T. hungarica, T. angustata, and T. constricta), Staurosira spp. (Staurosira.
classified as follows: 21 lakes were freshwater (<1 g/L), 6 lakes were construens, Staurosira venter), and a variety of epiphytic taxa (e.g., Coc­
subsaline (1–3 g/L), 13 lakes were hyposaline (3–20 g/L), and 5 lakes coneis, Amphora, Epithemia, Surirella, Achnanthidium (s.l.) spp., Masto­
were meso-hypersaline (>20 g/L). The majority of study lakes in the W gloia smithii and Karayevia clevei) (Fig. 5). Planktonic diatoms were
and S geographic regions, as well as the one lake in the E region, were mainly represented by Pantocsekiella ocellata, Cyclotella distinguenda,
subsaline to hyposaline (Fig. 2f). In contrast, the lakes in the N region Lindavia affinis, Pantocsekiella tripartita, Lindavia bodanica and Stepha­
(with the exception of Taiyang Lake) are closed basin systems and were nodiscus medius. Lakes in the W region had diatom assemblages that were
meso-hypersaline with higher concentrations of salinity (Fig. 2f) and more commonly comprised of Cocconeis lineata, Gyrosigma acuminatum,
major ions (Fig. 3). Stephanodiscus medius and Denticula kuetzingii, often in high abundances
The concentrations of individual ions measured in the 45 TP lakes in lakes with salinities below 1 g/L (Fig. 5). Diatom assemblages from
(Table S2), indicated that sodium, chloride, and sulfate were prevalent lakes in the S region, where salinities were below 6 g/L, were commonly
anions in samples from the majority of sampling sites. As shown in the comprised of Amphora (s.l.) spp. (e.g., A. ovalis, A. pediculus, A. libyca)
Piper diagram (Fig. 3), the 45 lakes can be divided into lakes charac­ and a variety of Lindavia spp. (e.g., L. affinis, L. radiosa, and L. bodanica)
terized by: (Type A) Ca2+-Mg2+-SO2− 2+
4 ; (Type B) Ca –HCO3 ; (Type C),

(Fig. 5). Lakes in the N region where the highest salinity concentrations
Na+-K+- Cl− ; and (Type D) Na+–HCO−3 . The majority of our TP sites can were measured (median = 12.5 g/L, maximum = 85.3 g/L) had higher
be characterized as Type C (Na+-K+- Cl− ), and Type D (Na+–HCO−3 ) relative abundances of Anomoeoneis costata, Nitzschia palea, Tryblionella
lakes (Fig. 3), indicating that most lakes are rich in carbonate and hungarica, Cyclotella distinguenda and S. venter (Fig. 5). Cyclotella dis­
chloride. The hydrochemical characteristics of the study lakes can be tinguenda was observed in only two lakes in the N region (NM4: salinity
categorized along a general trend from freshwater to saltwater to dry = 85.3 g/L and Longwe Co: salinity = 37 g/L), whereas Cyclostephanos
salt-lake conditions, following the sequence from carbonate to sulfate, dubius was only observed in Mang Co, the one lake representing the E
magnesium, and sulfate, and finally to chloride (Wang et al., 2013). region (salinity = 0.2 g/L), where it dominated (>88% relative abun­
Lakes in the N region were characterized by Na+-K+- Cl− and lakes in the dance) the assemblage (Fig. 5).
S region by Na+–HCO−3 (Fig. 3). Lakes in the W region were character­ The three most common diatom taxa observed in the 45 TP lakes,
ized by Ca2+–HCO−3 or Na+–HCO−3 . The one lake representing the E P. ocellata, N. palea, and S. venter, often showed distinct trends across
region (Lake Mang Co) was dominated by Ca2+-Mg2+-SO2− 4 (Fig. 3). environmental gradients and across geographic regions (Fig. 6).
A Pearson correlation matrix identified several significant correla­ P. ocellata was the most common and abundant taxon in this lake-set and
tions among the chemical, physical, and geographical variables was observed in 16 of the 45 lakes (Fig. 6a). This planktonic taxon was
measured in this study (Table 1). As expected, strong positive relation­ observed in high relative abundance in the N, S, and W geographic re­
ships (p < 0.001) were observed between total salinity and the dominant gions across a broad gradient of altitude (Fig. 6ai), water temperature
ions (Na+, K+, Mg2+, Cl− , and SO2− 4 , Table 1). Latitude and altitude had (Fig. 6aiii) and pH (Fig. 6aiv). Although this taxon occurred in both
a strong, positive correlation (r = 0.554, p < 0.001), altitude was shallow and deep lakes, it had distinctly higher relative abundances at
negatively associated with temperature (r = -0.513, p < 0.001) and maximum lake depths greater than ~ 10 m, particularly in the deeper
water depth (r = -0.328, p < 0.05) (Table 1), whereas water depth had a lakes of the S region (Fig. 6aii). In addition, P. ocellata occurred in higher
strong positive relationship with lake area (r = 0.614, p < 0.001) relative abundances in lakes with chlorophyll a below 3.5 µg/L

6
S. Yu et al.
Table 1
Pearson’s correlation matrix of measured environmental variables in the 45 TP lakes: latitude (o), longitude (o), altitude (m), lake area (km2), salinity (g/L), water depth (m), lake temperature (oC), pH (unit), chlorophyll a
(μg/L), Na+ (mg/L), K+ (mg/L), Mg2+ (mg/L), Ca2+ (mg/L), Cl− (mg/L), SO2− 4 (mg/L) and HCO3 (mg/L). One asterisk, two asterisks, and three asterisks indicate 0.05, 0.01, and 0.001 significance levels, respectively.

Latitude Longitude Altitude Area Salinity Depth Temperature pH Chlorophyll a Na+ K+ Mg2+ Ca2+ Cl− SO2−
4 HCO–3

Latitude 1 − 0.45*** 0.55 − 0.24 0.05 − 0.34 * − 0.42 − 0.33 0.03 0.01 − 0.03 − 0.02 − 0.05 0.15 − 0.15 − 0.23
*** ** *
Longitude 1 − 0.25 0.34 0.15 0.31 − 0.11 − 0.26 − 0.20 − 0.05 − 0.03 0.02 − 0.02 − 0.02 0.01 − 0.02
*
Altitude 1 − 0.25 − 0.01 − 0.33 − 0.51*** − 0.13 0.02 0.08 0.08 − 0.04 0.16 0.09 0.04 − 0.06
*
Area 1 0.18 0.61 0.17 − 0.02 − 0.17 0.02 0.1 0.03 0.15 0.07 0.09 0.05
***
Salinity 1 0.04 − 0.07 0.21 0.08 0.88 0.81 0.70 − 0.01 0.82 0.71 0.55
*** *** *** *** ***
Depth 1 0.02 − 0.16 − 0.18 − 0.09 − 0.07 − 0.03 − 0.01 − 0.02 0.01 0.08
Temperature 1 0.20 − 0.06 − 0.07 − 0.03 − 0.11 − 0.08 − 0.12 0.03 − 0.08
7

pH 1 0.32 0.36* 0.38 0.27 − 0.12 0.27 0.42 0.43


* * ** **
Chlorophyll a 1 0.16 0.11 0.03 0.28 0.11 0.08 0.20
Na+ 1 0.93 0.75 0.16 0.83 0.84 0.56
*** *** *** *** ***
K+ 1 0.79 0.13 0.79 0.82 0.55
*** *** *** ***
Mg2+ 1 0.12 0.81 0.62 0.32
*** ***
2+
Ca 1 0.08 0.14 − 0.1
Cl− 1 0.71 0.28
***
SO42− 1 0.43
**
HCO–3 1

Ecological Indicators 155 (2023) 110952


S. Yu et al. Ecological Indicators 155 (2023) 110952

Fig. 4. Principal component analysis (PCA) of physical, chemical, and geographical variables from the 45 TP study lakes, categorized by geographic region (colored
circles). The corresponding lake name for the numbers beside each circle in the figure can be found in Table S2 in the Supplemental material.

(Fig. 6av) and with salinity concentrations below 5 g/L (Fig. 6avi). water temperature and pH gradients but was in highest abundances in
Benthic S. venter occurred at highest relative abundances in S and N warmer and alkaline S region lakes (Fig. 6aii and 6aiv). S. venter was
region lakes located at altitudes between ~ 4600 m a.sl. to 4900 m a.s.l. most prevalent in S region lakes with chlorophyll a below ~ 3.5 µg/L
(Fig. 6bi) and in shallower N, S, and W region lakes below ~ 3 m (Fig. 6bv) and salinity below 2.0 g/L (Fig. 6bvi). Nitzschia palea occurred
maximum depth (Fig. 6bii). This taxon was observed across the full in high relative abundances (>10%) over a broad range of altitudes in

8
S. Yu et al. Ecological Indicators 155 (2023) 110952

Fig. 5. Relative abundances of 37 dominant (maximum relative abundance > 15% in at least 2 lakes, and their salinity optima were in the different lake types.)
diatom taxa in the 45 TP lake calibration set, with different colors depicting geographic categorizations. The lakes were organized by region, and lakes within each
region were ranked from low to high salinity. The relative abundances of diatoms in region W, S, N, E are shown with green, pink, orange and red bars, respectively.
(For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

the N, S, and W regions (Fig. 6ci) and over a range of lake depths from cumulative variance of diatom species (Table 3), with axis 1 repre­
relatively deep to shallow lakes (3.4 to 40 m) (Fig. 6cii). This taxon was senting a gradient of salinity (and related variables) and axis 2, a
observed in highest relative abundances (>20%) in warmer waters from gradient of maximum lake depth (Fig. 8). The CCA ordination plots show
both N and S region lakes (Fig. 6ciii) and in N, S, and W lakes across a a clear separation of the more saline, N region lakes (except for lake 36)
broad range of pH (Fig. 6civ). It occurred in highest relative abundances that plotted in the right, lower quadrant of the ordination. Lakes in the
(>10%) across a relatively wide gradient of chlorophyll a in both N and W and S regions were lower in salinity and clustered on the left of axis 1,
S region lakes (Fig. 6cv) and across a wide gradient of salinity from and with a gradient of lake depth along axis 2 (Fig. 8A and 8C). Altitude
freshwater to 85.3 g/L in the N, S and W region (Fig. 6cvi). (run passively) was positively related to salinity, suggesting that higher
Diatom richness and diversity showed distinct spatial patterns altitude lakes were generally higher in salinity. Salinity and correlated
among the lakes across the W, N, and S regions (Fig. 7). In general, lakes variables, Mg2+, Cl− and SO2− 4 (run passively) were negatively related to
with salinity concentrations between 0.1 and 10 g/L had higher diatom maximum water depth and plotted on the opposite sides of the CCA
diversity and richness than lakes with higher salinities (Fig. 7a,b). Lakes ordination (Fig. 8C). Lake Mang Co (45), the only lake from the E region,
with salinities >20 g/L were mainly located in the N region and were displayed a strong association with longitude and Ca2+ (Fig. 8A and 8C).
characterized by relatively low biodiversity and richness compared with Nitzschia spp. (species code: 89, 91, 96, 99, 100), Fallacia pygmaea
other regions (Fig. 7). The S region had the largest range of diatom di­ (49), C. distinguenda (22), and Denticula subtilis (36) plot in the right
versity and richness. portion of the CCA plot (Fig. 8B), showing that these taxa commonly
occur in lakes with higher concentrations of salinity and major ions
(Fig. 8B). Gomphonema spp. (54, 58), Lindavia radiosa (63), P. ocellata
4.2. Ordination (24), and Pinnularia rhombarea (104) plot in the left side of the CCA
ordination space, indicating that they tend to occur in deeper lakes
Ordination analyses were performed on 125 diatom taxa (1% in 2 (Fig. 8B). C. dubius (26), Eolimna sp. (42), and Gomphonema spp. (59, 60)
lakes) from the TP calibration set. Further data screening included an showed positive associations with deeper TP lakes with higher Ca2+
examination of variance inflation factors (VIFs) through a series of CCAs concentrations and located at higher longitudes. The relative abundance
that identified Na+ and K+ as having multiple correlations (VIFs > 10) of Gomphonema spp., Stauroneis livingstonii (110) and Stephanodiscus
and were therefore eliminated from further analyses. Forward selection spp., which characterized diatom assemblages in the S and W regions,
in CCA was performed on the remaining variables from this lake set, increased with increasing pH and bicarbonate (both run passively)
which identified salinity and maximum lake depth as being significant (Fig. 8).
(Monte Carlo permutation test, p < 0.001) and explaining the highest
variation in diatom assemblage composition (4.2% and 3.7%, respec­
tively) (Table 2). 4.3. Diatom-based salinity transfer function
The relationships between diatom species and environmental vari­
ables were explored by running a CCA with salinity and maximum water Salinity model development was based on the modern diatom dis­
depth as active variables, and a suite of variables of interest that were tributions of 125 taxa (1% in 2 lakes) across the measured salinity
not found to be significant as passive variables in the ordination (i.e. pH, gradient of 44 lakes (excluding lake Mang Co). The results of a PCA of
chlorophyll a, longitude, altitude, lake area, HCO–3, SO2− 2+
4 , Mg , Ca ,
2+
the environmental variables and a DCA of the species data identified
Cl ) (Fig. 8). The first four axes of the CCA explained 17.09% of the

Mang Co as an outlier (Fig. S1) and therefore was not included in model

9
S. Yu et al. Ecological Indicators 155 (2023) 110952

Fig. 6. Responses of selected dominant diatom taxa (Pantocsekiella ocellata, Staurosira venter and Nitzschia palea) with the highest occurrence in 45 lakes with the
different regions related to altitude, lake depth, temperature, pH, chlorophyll a and lakewater salinity. The various categories from freshwater to hypersaline are
delineated by the dashed vertical lines.

10
S. Yu et al. Ecological Indicators 155 (2023) 110952

models showed the strongest predictive abilities. The performance sta­


tistics of WA-PLS with one component (WAPLS-1) were similar to WA-
PLS with additional components (Table 5) and therefore did not show
a notable improvement over simple WA inference models. A WA model
with no tolerance down-weighting for both inverse and classical
deshrinking (R2jack = 0.71) outperformed WA models with tolerance
downweighting (R2jack = 0.49 and 0.50) (Table 5). Based on these com­
parisons of model performance, we chose a simple WA model with no
tolerance downweighting and with classical deshrinking (the mean and
maximum bias and RMSEP were slightly lower than the inverse
deshrinking model: Table 5). The salinity transfer function developed
from our TP lakes had a relatively high jackknifed coefficient of deter­
mination (R2jack = 0.71), and a relatively low RMSEPlog (0.28) (Fig. 9).
Fig. 7. Distribution of diatom (a) taxon richness and (b) Shannon’s diversity in
the 45 TP lakes. 4.4. Species ecological optima and tolerances for salinity

The estimated salinity optima that form the basis of the WA-Cla
Table 2
model are included for the 35 most frequently occurring diatom taxa
Total variance explained and p-values for forward selected variables (salinity
in the TP calibration set (i.e., taxa that occurred at>1% relative
and maximum lake depth), that were actively included in the canonical corre­
spondence analysis (CCA) and that significantly and independently explained
diatom distributions in the 45-lake TP calibration set. Table 3
Eigenvalues, explained variation, species-environmental correlations, explained
Forward-selected environmental variable Variance explained (%) P-value
fitted variation for the first four canonical correspondence analysis (CCA) axes of
Salinity 4.2 0.001 the TP lakes calibration set, showing the total inertia and the proportion of the
Depth 3.7 0.001 total variation explained.
Statistic Axis 1 Axis 2 Axis 3 Axis 4 Value
development. The ratio of the eigenvalue for the first constrained DCCA Eigenvalues 0.56 0.42 0.40 0.37
axis to the eigenvalue for the second unconstrained DCCA axis for Explained variation (cumulative) 5.46 9.56 13.49 17.09
salinity (λ1/λ2 = 0.76; Table 4) suggests this variable has moderate Species-environmental 0.97 0.93 0.94 0.94
correlation
explanatory power for diatom distributions (Ter Braak, 1988) and is
Explained fitted variation 14.17 24.80 34.98 44.32
comparable to the value reported for salinity by Bloom et al. (2003) (λ1/ (cumulative)
λ2 = 0.67). Although a ratio below 1.0 suggests it is not the only envi­ Total Inertia CCA= 10.22
ronmental gradient that the diatom taxa are responding to (Juggins, Explanatory variables 38.6%
2013), it accounted for the highest variation in the diatom assemblage
composition in the TP calibration lakes (4.2 %; p < 0.05) and was
therefore explored for the development of a transfer function. Table 4
DCCA with the first axis constrained to salinity produced a gradient Results of the detrended canonical correspondence analysis
length of 5.9 SD, indicating that a unimodal inference model was (DCCA) and canonical correspondence analysis (CCA) con­
appropriate (Birks, 1995). A comparison of the performance statistics of strained to salinity for diatom taxa of the 45 TP lakes.
a variety of salinity inference models developed from our TP calibration Environmental variable Salinity
set indicated that simple weighted average (WA) models or models Gradient length (SD units) 5.9
based on WA partial least squares (WA-PLS) outperformed more com­ DCCA axis 1 λ1 0.44
plex models including locally weighted WA (LWWA), maximum likeli­ λ1/λ2 0.76
hood (ML), and PLS models (Table 5). PLS and ML models had weaker P-value 0.002
Variance explained (%) 10.22
jackknifed correlations between the observed diatom-inferred salinity
CCA axis 1 λ1 0.43
measurements and showed larger maximum bias, whereas the LWWA P-value 0.001
model had a larger average bias (Table 5). The WA and WA-PLS salinity

Fig. 8. Canonical correspondence analysis (CCA) biplots of (A) sampling sites categorized by geographic region (colored circles for W, S, N and E), (B) diatom taxa,
and (C) environmental variables in the 45 TP lakes. Table S1 provides corresponding lake names to the numbers in (A) and Table S3 provides corresponding diatom
taxonomic names to the numbers in (B). Solid lines with arrows represent forward-selected environmental variables, and dashed lines with arrows represent
environmental variables that were plotted passively in the ordination.

11
S. Yu et al. Ecological Indicators 155 (2023) 110952

Table 5
Performance statistics (R2, jackknifed R2 (R2jack) and root-mean-squared-error (RMSE) and RMSE of prediction (RMSEP) for weighted averaging (WA), locally weighted
weighted averaging (LWWA), maximum likelihood (ML), partial least squares (PLS) and weighted average partial least squares (WAPLS)-based diatom-salinity
inference models for 44 lakes on the Tibetan Plateau. The cross-validation of leave-one-out (jackknifing) was applied.
Model R2 RMSE (log10(Salinity + 1)) R2jack Average Bias Max Bias RMSEP (log10(Salinity + 1))

WA_Inv 0.90 0.16 0.71 0.03 0.33 0.29


WA_Cla 0.90 0.17 0.71 0.03 0.32 0.28
WATOL_Inv 0.87 0.19 0.49 0.1 1.05 0.39
WATOL_Cla 0.87 0.20 0.50 0.11 1.06 0.39
WAPLS-1 0.90 0.16 0.72 0.02 0.33 0.29
WAPLS-2 0.97 0.09 0.72 0.02 0.43 0.28
WAPLS-3 0.99 0.06 0.72 0.03 0.42 0.28
WAPLS-4 0.99 0.04 0.72 0.03 0.43 0.28
WAPLS-5 1 0.03 0.71 0.03 0.39 0.28
PLS-1 0.71 0.28 0.48 0.0006 0.70 0.38
PLS-2 0.89 0.17 0.61 − 0.008 0.63 0.33
PLS-3 0.94 0.13 0.64 0.002 0.64 0.31
PLS-4 0.97 0.09 0.63 0.002 0.49 0.32
PLS-5 0.99 0.06 0.6 0.006 0.42 0.33
ML 0.83 0.23 0.48 0.01 0.60 0.31
LWWA_Inv 0.92 0.15 0.7 0.27 0.63 0.30
LWWA_Cla 0.92 0.15 0.7 0.18 0.57 0.29

WA_Inv: WA (inverse deshrinking); WA_Cla: WA (classical deshrinking); WA_Inv: WA (tol. Downweighted inverse deshrinking); WATOL_Cla: WA (tol. Downweighted
classical deshrinking); LWWA_Inv: LWWA (inverse deshrinking); LWWA_Cla: LWWA (classical deshrinking).

Table 6
Tibetan Plateau diatom taxa that occurred with relative abundances>1% in at least two lakes in the 44 calibration lakes, species code, effective number of occurrences
(Hill’s N2 > 4), jackknifed WA_Cla (Weighted average with classical deshrinking) salinity optima (Jack_Opt, back-transformed salinity concentrations (g/L)), and
tolerances (Jack_Tol, back-transformed salinity concentrations (g/L)), max abundant (Max_Abund), mean abundant (Mean_Abund.), salinity category (Sa_categ) (F =
Freshwater lake; Sub = Subsaline lake; Hypo = Hyposaline lake; Meso-hype = Meso-hypersaline lake). The diatom taxa are ordered from low to high salinity optima.
NO. Diatom taxa Jack_Opt Jack_Tol Hill’s N2 Max_Abund. (%) Mean_Abund.(%) S_categ.

1 Navicula radiosa 0.44 0.73 5.89 14.49 0.77 F


2 Pinnularia viridiformis var. minor 0.50 0.81 6.85 8.24 0.6 F
3 Navicula cari 0.57 1.26 5.73 2.51 0.06 F
4 Amphora ovalis 0.68 0.71 8.41 29.34 1.61 F
5 Cymbella proxima 0.69 0.59 4.18 4.35 0.22 F
6 Staurosira venter 0.74 1.40 8.66 38.46 2.23 F
7 Cocconeis lineata 0.83 1.07 10.36 85.50 4.04 F
8 Mastogloia smithii 0.92 1.17 5.30 27.51 1.36 F
9 Epithemia sorex 0.95 0.79 5.63 4.76 0.3 F
10 Halamphora veneta 0.99 1.10 6.82 8.90 0.61 F
11 Gyrosigma acuminatum 1.00 1.46 10.12 46.01 1.95 F
12 Cymbella hantzschiana 1.07 2.44 5.11 9.49 0.47 Sub
13 Pantocsekiella ocellata 1.20 0.96 14.69 79.46 13.54 Sub
14 Encyonopsis microcephala 1.20 1.65 6.90 11.24 0.65 Sub
15 Navicula recens 1.26 1.56 4.19 11.49 0.51 Sub
16 Amphora pediculus 1.30 1.58 14.04 39.04 2.44 Sub
17 Pseudostaurosira brevistriata 1.46 0.82 4.79 2.97 0.23 Sub
18 Denticula kuetzingii 1.47 1.84 7.33 29.80 2.31 Sub
19 Encyonema silesiacum 1.48 1.21 5.92 11.96 0.48 Sub
20 Karayevia clevei 1.78 1.50 6.44 33.60 1.34 Sub
21 Caloneis bacillum 1.96 1.05 5.71 43.05 1.74 Sub
22 Staurosirella pinnata 1.97 1.69 9.80 10.14 0.82 Sub
23 Ulnaria ulna 2.00 1.78 10.35 13.90 0.98 Sub
24 Diploneis elliptica 2.00 1.38 7.84 9.24 0.65 Sub
25 Staurosira construens 2.12 1.14 9.11 8.46 0.91 Sub
26 Epithemia argus 2.51 1.48 4.38 60.26 1.65 Sub
27 Surirella crumena 2.68 0.96 7.01 21.43 1.09 Sub
28 Anomoeoneis costata 3.39 1.38 5.89 86.00 4.12 Hypo
29 Nitzschia fonticola 3.42 3.36 14.87 62.31 4.31 Hypo
30 Surirella ovalis 3.47 0.92 4.41 11.00 0.47 Hypo
31 Pantocsekiella tripartita 3.96 1.53 4.08 89.51 3.33 Hypo
32 Tryblionella hungarica 12.09 3.18 5.03 78.99 2.74 Hypo
33 Neidium ampliatum 13.61 6.44 4.52 8.88 0.43 Hypo
34 Nitzschia palea 13.69 3.19 8.48 79.47 6.28 Hypo
35 Navicula veneta 20.79 4.51 4.49 26.63 0.99 Meso-hype

abundance in at least two samples and that had a Hill’s N2 > 4) and are with optima in the hyposaline category (3.0 to 20.0 g/L) (Table 6). Only
presented in Table 6. The salinity optima for these common diatom taxa one of the 35 dominant taxa, Navicula veneta, had a salinity optimum in
ranged from 0.4 − 20.8 g/L, spanning a gradient from freshwater to the meso-hypersaline range (>20.0 g/L) (Table 6). Three of the most
hyposaline conditions (Table 6). Of the 35 dominant diatoms, there were common and abundant taxa included Nitzschia palea (hyposaline opti­
eleven taxa with optima in the freshwater category (0.44 to 1.0 g/L), mum of 13.7 g/L, mean relative abundance 6.3%, maximum relative
sixteen with optima in the subsaline category (1.0 to 3.0 g/L), and seven abundance 79.5%), P. ocellata (subsaline optimum of 1.2 g/L, mean

12
S. Yu et al. Ecological Indicators 155 (2023) 110952

Fig. 9. Plots of observed versus inferred log-transformed salinity (left) and the residuals (right) using the weighted averaging with classical deshrinking inference
model (WA-Cla).

relative abundance 13.5%, maximum relative abundance 79.5%), and from the TP (Liu et al., 2021).
Amphora pediculus (subsaline optimum of 1.3 g/L, mean relative abun­ Lakes in the southern TP were generally warmer than lakes in the
dance 2.5%, maximum relative abundance 39.0%). northern TP, likely because they experience a shorter period of ice cover
(Pi et al., 2021). Chlorophyll a concentrations were generally similar
5. Discussion across the four regions, with a few exceptions within the S region where
concentrations in the two closed basin lakes (Angrenjin Co, Angzi Co)
The climate of the Tibetan Plateau in central Asia is largely were higher. Higher chlorophyll a concentrations in TP lakes were found
controlled by Asian summer monsoons that produce strong gradients in to be related to warmer conditions and shorter ice cover duration (Pi
precipitation and moisture, resulting in lakes with a large range in et al., 2021). A longer and warmer growing season for algae and aquatic
salinity (major ion) concentrations (Yang et al., 2003; Li et al., 2016). plants would lead to higher primary production and chlorophyll a
Because diatoms respond sensitively to changes in salinity (Wilson et al., concentrations. In addition to climate changes, higher chlorophyll a
1996; Fritz et al., 1991), diatom-based salinity inference models have concentrations in Angrenjin Co can also be explained by its proximity to
the potential to be used to track past climate scenarios (Wilson et al., human activities in the northern part of Angren County.
1996; Saunders, 2011; Wang et al., 2011). This is of particular interest in
the remote, high mountain lakes of the TP that are highly sensitive to 5.2. Spatial distributions of diatom taxa and its environmental indication
climate change.
The distribution of the main diatom taxa across the TP showed
5.1. General trends in environmental variables across Tibetan Plateau distinct separation among the different geographic regions, particularly
lakes between the W and N regions (Fig. 5). This clearly highlights the distinct
limnological differences among lakes in this heterogeneous mountain
Water chemistry and environmental variables examined across our landscape. The three most common and dominant diatom taxa in our
TP lakes generally reflect the large limnological diversity of lakes in this study (P. ocellata, S. venter, and N. palea) often showed clear preferences
high-altitude, heterogeneous region. The majority of our 45 TP lakes had for specific limnological conditions (Fig. 6). For example, P. ocellata
major ion concentrations that were substantially higher (by as much as clearly dominated lakes in the S region and preferred low salinity, low
ten times) than what has been commonly reported in other regions of productivity, and relatively high pH conditions, but it also occurs in high
China (Table S3). One of the clearest trends across our 45 lakes were the abundance across a range of elevations and water depths (Fig. 6a). This
large spatial differences in lake-water salinity, pH and major ion con­ planktonic diatom is one of the most commonly reported taxa in pale­
centrations that generally increased with altitude (Tables S1 and S2). In olimnological studies of the TP (e.g., Chen et al., 2021; Zhang et al.,
particular, the closed-basin lakes, which are most common to the N re­ 2023) and often dominated assemblages for the entire record (Wisch­
gion (e.g., Tangra Yum Co, Nam Co), were chemically distinct and were newski et al., 2014; Liao et al., 2020; Wang et al., 2020; He et al., 2022).
characterized by Na+-K+- Cl− (Fig. 3). The lack of inflow and outflow in Although our study did not include lake-water nutrient measurements, it
these hydrologically closed lakes make them susceptible to increased is well-documented that P. ocellata prefers oligotrophic, non-acidified
evaporation and crystallization, greatly affecting their chemical conditions (Yang et al., 2003; Wang et al., 2014; Liao et al., 2020) and
composition (Qiao et al., 2017; Wang et al., 2010). Therefore, closed based on our results, prefers to live in lakes with salinity concentrations
lakes in the N region are likely most susceptible to climate-related lake- lower than 6.1 g/L (Fig. 6avi). Not surprisingly, benthic S. venter and
level changes. In contrast, the majority of lakes in the S and W regions N. palea dominated the shallower lakes of the calibration set (Fig. 6bii,
are hydrologically open with major ions dominated by Na+–HCO−3 and 6cii). Based on our results, N. palea had a clear preference for lakes with
Ca2+–HCO−3 , respectively. Water chemistry was distinct across the a salinity higher than 8 g/L (Fig. 6cvi), whereas S. venter was usually
various regions largely owing to differences in evaporation that would found in freshwater lakes (Fig. 6bvi).
affect ion concentrations (i.e. Na+, Cl− , Ca2+, Mg2+) and pH. For The majority of surface sediment diatom assemblages in the 45 TP
example, the high evaporation rates of the W region resulted in high lakes were dominated by a variety of benthic taxa, particularly in lakes
Ca2+ concentrations precipitated from calcium carbonate. Differences in in the N and W regions, where lakes were generally shallower than lakes
pH among the lakes from the four different regions tended to increase in the S and E regions (Fig. 2b). In shallow lakes, light is sufficient to
with latitude and is in agreement with previous limnological studies allow aquatic plants and epiphytic diatoms to proliferate. Therefore,

13
S. Yu et al. Ecological Indicators 155 (2023) 110952

benthic taxa in the shallow W region lakes (i.e., Bangong Lake, Sumxi differences in lake depth, the only other measured variable found to
Co, Hongshan Hu, Zongxiong Co) and N region lakes (i.e., Serling Co, significantly explain diatom distributions, although not correlated with
Longwe Co, Dong Co) were largely represented by epiphytic taxa such as salinity (Fig. 8). In addition to the variables available for our study, it is
C. placentula var. lineata and G. acuminatum. Most of high salinity lakes likely that some of the diatom variation can be explained by factors that
in the N region were dominated by benthic taxa, which can be explained were not measured, such as nutrients (e.g., total phosphorus) and
by high evaporation rates relative to low precipitation in this remote climate-related changes to lake water properties. Other variables that
inland region of the TP (Li et al., 2016). were not included in the ordination analysis such as hydrology (e.g.,
Lakes located in the arid and semi-arid western and central TP closed versus open-basin lakes) would also have an effect on salinity
(Longwe Co, Dong Co and Serling Co) had distinct diatom assemblages where closed-basin systems tend to have higher salinity concentrations,
that appear to be mainly driven by salinity, major ions and water depth particularly when evaporation is greater than precipitation. Acknowl­
(Fig. 8). These lakes tend to have higher salinity concentrations as they edging that diatom assemblages are likely also responding to other
are located far inland where the Himalayan mountains provide a barrier variables in TP lakes, the development of a salinity transfer function is of
from the ocean resulting in low precipitation (Liu et al., 2015). For interest given the importance of this variable in characterizing these
example, diatom assemblages in these higher salinity lakes were lakes.
comprised of a variety of brackish water taxa such as Anomoeoneis cos­ Simple weighted-averaging with classical deshrinking (WA-Cla) was
tata and Nitzschia palea. In meso-hypersaline lakes (e.g., Longwe Co, deemed to be the most robust salinity inference model for our TP lakes
Dong Co and Yagen Co), Nitzschia spp. and A. costata dominated the (Table 5). As was found in our examination of performance statistics,
assemblages. Our findings are in agreement with a previous study of WA models often outperform other more complex calibration ap­
Serling Co that reported A. costata relative abundances were highest in proaches and are therefore widely used to examine the relationship
shallow lakes that are relatively high in specific conductance and between diatom taxa and the environmental variable of interest (Birks
salinity (Yu et al., 2019). Therefore, specific diatom species are available et al., 1998; Birks and Simpson, 2013; Cumming et al., 2015). The choice
to assess the level of salinization (Stenger-Kovács et al., 2023). of simple WA regression for developing a salinity model is consistent
Planktonic Cyclostephanos dubius was only observed in one of the 45 with previous calibration studies from the Interior Plateau of British
TP lakes (Mang Co, the only lake representing the E region) where it Columbia (Wilson et al. 1994), and in Arctic (Moser et al. 2000) and
dominated the assemblages. Cyclostephanos taxa are well-documented Antarctic (Verleyen et al. 2003) regions. However, our TP calibration set
indicators of eutrophic conditions (Bradshaw and Anderson, 2003; tended to have more lakes in the higher end of the salinity gradient (18
Hausmann and Kienast 2006; Laperriere et al., 2008; Thienpont et al., lakes>3 g/L) whereas in many diatom-salinity studies from other re­
2013) and, although lake-water nutrient concentrations were not gions, fewer lakes had salinity concentrations>3 g/L. For example,
measured, the high relative abundances of C. dubius in Mang Co suggests among the 57 lakes in the Sierra Nevada (California, USA), there were
this lake is high in nutrients. only 8 lakes with salinity concentrations>3 g/L (Bloom et al., 2003),
In general, the salinity of the southern lakes is significantly lower and among the 102 lakes in British Columbia (Canada), only 36 lakes
than that of the northern lakes, and the southern lakes have a larger had salinity concentrations>3 g/L (Wilson et al., 1994).
gradient of diatom richness and diversity (Fig. 7a,b). This result suggests Although the inferred versus observed salinity values based on
that diatoms in the central TP lakes may have a latitudinal diversity diatom composition for the WA-Cla model exhibits a relatively strong
gradient (LDG). This observation is similar to that made by Hillebrand relationship, the lakes in the TP dataset were not evenly distributed
and Azovski (2001) using 148 local and 25 regional freshwater diatom across the salinity gradient (Fig. 9A). The majority of lakes had salinity
data sets, which also showed that there was a significant diatom LDG. values at the lower end of the gradient and lakes with moderate salinity
concentrations were not well represented in the dataset as they were
5.3. Diatom-based salinity transfer function rare and, therefore, not sampled. A comparison of the observed versus
predicted salinity values indicated that our WA-Cla model tends to over-
Given the strong control that the East Asian and Indian summer predict salinity in freshwater lakes (salinity<1.0 g/L (log-salinity <
monsoons have on precipitation and moisture in the TP (An et al., 2012), 0.3)) and under-predict salinity in lakes with concentrations>1.0 g/L
this region typically encompasses a large range in salinity from fresh­ (log-salinity < 0.3). This may be because the 44 calibration lakes were
water to hypersaline lakes (Liu et al., 2021; Wang et al., 2011). Diatom- not evenly distributed across the salinity gradient, with fewer lakes
salinity transfer function studies typically aim to track changes in representing the meso-hyposaline (9–30 g/L) part of the gradient
climate over time, with the assumption that changes in precipitation and (Fig. 9). The performance results of our WA-Cla diatom salinity model
effective moisture will be large enough to clearly alter lake water (R2 = 0.9, R2jack = 0.71, RMSEPjack = 0.28, N = 44) were similar to a
salinity (Fritz et al., 2013) and would be registered in sedimentary diatom-salinity transfer function developed with a WAPLS model from
diatom records. In the arid TP region, where evaporative forcing is lakes in the eastern TP by Wang et al. (2011) (R2jack = 0.738, RMSEPjack
strong, these conditions are easily met (Sun et al., 1990). In particular, = 0.38, N = 90). The range in salinity measured from lakes in our study
when evaporation is greater than precipitation, the numerous hydro­ (0.1 to 85.3 g/L; 18 of 44 lakes (41%) with salinity > 3 g/L) were similar
logically closed lakes in the TP have great potential to concentrate dis­ to lakes from the eastern TP (0.1 to 123 g/L; 38 of 90 lakes (42%) with
solved salts (China salty lakes resources and environmental database, salinity > 3 g/L) but the majority of our study lakes were located at
2019). higher elevations (30 of 44 lakes (68%) above 4600 m a.s.l.) than the
Salinity was found to be important for explaining the distribution eastern TP lakes (27 of 90 lakes (30%)) were above 4600 m a.s.l.). The
and main variation in diatom assemblage composition in the TP cali­ lower number of high elevation sites is not surprising, given the diffi­
bration lakes, a trend that has been well-documented for this high culty in sampling these altitudes in the western and central TP. In gen­
mountain, monsoon-dominated region (Yang et al., 2003; Wang et al., eral, the diatom-salinity transfer function obtained in this study would
2011; Li et al., 2021). Although salinity explained the greatest variance be most applicable to the semiarid and arid regions of the TP. Increasing
in diatom assemblage composition in our TP lakes, the relative explan­ the number of lakes in the meso-hyposaline range would be a step to­
atory power of this variable to predict diatom assemblage composition wards improving diatom-based salinity inferences in the TP. Given that
(λ1/λ2 of DCCA = 0.76) was moderate. Ter Braak (1988) suggested that sampling these lakes may prove logistically difficult, future studies
λ1/λ2 values should be at least 1.0 for reconstructions to be useful, and might consider merging existing TP calibration sets.
following this rule, salinity was not the only variable determining Salinity optima
diatom distributions in our TP calibration lakes (Juggins, 2013). In Paleolimnological studies provide key information that enable as­
particular, diatom variation in our TP lakes was also influenced by sessments on the extent of environmental change in the context of

14
S. Yu et al. Ecological Indicators 155 (2023) 110952

natural, background conditions (Smol and Cumming, 2000). To more represented in the lake set. The spatial distribution of diatoms in lakes
accurately interpret changes in diatom assemblage composition over across the TP together with the diatom-salinity transfer function could
time, it is necessary to define the environmental optima of taxa, be used to track past changes in salinity and environmental conditions
particularly in understudied regions such as the TP where diatom-based that may reveal long-term patterns in climate in this complex, high
salinity studies are scarce. The estimated salinity optima of the 35 most mountain region.
common diatom taxa in the TP lakes (Table 6) form the basis of the
simple WA_Cla model developed in this study. Categorizing salinity
preferences (based on optima) for these taxa may be useful for future Declaration of Competing Interest
researchers. Although a few diatom-based salinity models have been
developed for the TP (Yang et al., 2004; Wang et al., 2011), this is a vast, The authors declare that they have no known competing financial
highly heterogeneous and complex mountain region that would require interests or personal relationships that could have appeared to influence
additional calibration sets from various locations to provide a more the work reported in this paper.
complete distribution of lakes across the salinity gradient. Information
on diatom distributions and autecological information will undoubtedly Data availability
be refined as more lakes are examined in this understudied high
mountain region. Data will be made available on request.
The estimated salinity optima of the most common diatom taxa in
this study were compared to the salinity optima estimated from other Acknowledgements
diatom-based studies (including outside of the TP region). However,
diatom taxa from the TP lakes were often particular to the region and This study was jointly funded by the Strategic Priority Research
therefore not all taxa in Table 6 have been reported in the available Program of the Chinese Academy of Sciences (Grant No.
literature on diatom salinity model construction. In general, many of our XDA2007010102), the Second Tibetan Plateau Scientific Expedition and
taxa had estimated salinity optima that were comparable to optima re­ Research (STEP) (2019QZKK0202), the National Natural Sciences
ported by other diatom-based salinity studies. For example, the most Foundation of China (Grant No. 41877168), the Science and Technology
common planktonic taxon in our calibration set (P. ocellata) had a Department of Tibet (XZ202001YD0013C) and China Scholarship
salinity optimum of 1.2 g/L (subsaline waters) that was slightly higher Council (Grant No. 20204910715). We thank Qiangqiang Kou and Teng
than the estimated optimum of 0.9 g/L (freshwater) from an eastern TP Xu, Cunlin Li for helping with the sampling in the fieldwork. We also
calibration set (Wang et al., 2011). Ulnaria ulna had an optimum (2 g/L) thank Qingmin You and Yameng Li for helping with the diatom
in the subsaline category that was comparable (although slightly higher) identification.
to this taxon’s optimum estimated from other datasets from south-west
of Australia (1.5 g/L; Taukulis and John, 2009) and in the northern
Great Plains, USA (1.6 g/L; Fritz et al., 1993). Based on the estimated Appendix A. Supplementary data
optimum from our TP calibration set, Tryblionella hungarica, Nitzschia
palea, Neidium ampliatum, and A. costata prefer hyposaline lake condi­ Supplementary data to this article can be found online at https://doi.
org/10.1016/j.ecolind.2023.110952.
tions (Table 6), which is consistent with other studies that likewise
found these taxa to occur in high salinity conditions (Martínez et al.,
2003). Navicula veneta, a species considered to be a salt generalist (Veres References
et al., 1995) and shown to withstand meso-hyposaline conditions
An, Z., Colman, S.M., Zhou, W., Li, X., Brown, E.T., Jull, A.J., Cai, Y., Huang, Y., Lu, X.,
(Blanco et al., 2013), had a salinity optimum of 20.8 g/L in the western
Chang, H., Song, Y., Sun, Y., Xu, H., Liu, W., Jin, Z., Liu, X., Cheng, P., Liu, Y., Ai, L.,
and central TP (Table 6). This taxon is usually associated with high Li, X., Liu, X., Yan, L., Shi, Z., Wang, X., Wu, F., Qiang, X., Dong, J., Lu, F., Xu, X.,
conductivity waters and has also been shown to dominate mesosaline 2012. Interplay between the Westerlies and Asian monsoon recorded in Lake
Qinghai sediments since 32 ka. Sci. Rep. 2, 619. https://doi.org/10.1038/
ponds (specific conductivity: 10.9–17.3 mS/cm) in the southern Iberian
srep00619.
Peninsula (Fernández-Moreno et al., 2022). Anslan, S., Azizi Rad, M., Buckel, J., Echeverria Galindo, P., Kai, J., Kang, W., Keys, L.,
Maurischat, P., Nieberding, F., Reinosch, E., Tang, H., Tran, T.V., Wang, Y.,
Schwalb, A., 2020. Reviews and syntheses: How do abiotic and biotic processes
6. Conclusions
respond to climatic variations in the Nam Co catchment (Tibetan Plateau)?
Biogeosciences 17 (5), 1261–1279.
The 45-lake calibration set from the Tibetan Plateau provides Battarbee, R.W., Anderson, N.J., Jeppesen, E., Leavitt, P.R., 2005. Combining
important information on spatial distributions and diatom- palaeolimnological and limnological approaches in assessing lake ecosystem
response to nutrient reduction. Freshw. Biol. 50, 1365–2427. https://doi.org/
environmental relationships in this climatically sensitive high moun­ 10.1111/j.1365-2427.2005.01427.x.
tain region. Diatom assemblage distributions in our TP lakes were Benito, X., Fritz, S.C., Steinitz-Kannan, M., Tapia, P.M., Kelly, M.A., Lowell, T.V., Gill, J.,
strongly influenced by lakewater salinity. In particular, the N region 2018. Geo-climatic factors drive diatom community distribution in tropical South
American freshwaters. J. Ecol. 106 (4), 1660–1672.
included lakes with relatively high salinity concentrations and had Bennion, H., Fluin, J., Simpson, G.L., 2004. Assessing eutrophication and reference
distinct diatom assemblage composition. Of the most common diatom conditions for Scottish freshwater lochs using subfossil diatoms. J. Appl. Ecol. 41,
taxa observed in our lakes, P. ocellata and S. venter were found in 124–138. https://doi.org/10.1111/j.1365-2664.2004.00874.x.
Birk, S., Chapman, D., Carvalho, L., Spears, B.M., Andersen, H.E., Argillier, C., Auer, S.,
freshwater lakes, whereas N. palea had highest abundances in hyposa­ Baattrup-Pedersen, A., Banin, L., Beklioğlu, M., Bondar-Kunze, E., Borja, A.,
line lakes. Although salinity had a strong influence on diatom assem­ Branco, P., Bucak, T., Buijse, A.D., Cardoso, A.C., Couture, R.M., Cremona, F., de
blage composition, other variables including water depth also played a Zwart, D., Feld, C.K., Ferreira, M.T., Feuchtmayr, H., Gessner, M.O., Gieswein, A.,
Globevnik, L., Graeber, D., Graf, W., Gutiérrez-Cánovas, C., Hanganu, J., Işkın, U.,
role in structuring the diatom assemblages. A simple WA model resulted Järvinen, M., Jeppesen, E., Kotamäki, N., Kuijper, M., Lemm, J.U., Lu, S., Solheim, A.
in a relatively robust salinity transfer function for TP lakes based on L., Mischke, U., Moe, S.J., Nõges, P., Nõges, T., Ormerod, S.J., Panagopoulos, Y.,
estimated diatom salinity optima that cover a large gradient including Phillips, G., Posthuma, L., Pouso, S., Prudhomme, C., Rankinen, K., Rasmussen, J.J.,
Richardson, J., Sagouis, A., Santos, J.M., Schäfer, R.B., Schinegger, R., Schmutz, S.,
freshwater (0.1–1.0 g/L), subsaline (1.0–3.0 g/L), hyposaline (3.0–20.0
Schneider, S.C., Schülting, L., Segurado, P., Stefanidis, K., Sures, B., Thackeray, S.J.,
g/L), and meso-hyposaline water (>20.0 g/L) lakes. This relatively large Turunen, J., Uyarra, M.C., Venohr, M., von der Ohe, P.C., Willby, N., Hering, D.,
gradient in salinity across the TP provides an opportunity to explore past 2020. Impacts of multiple stressors on freshwater biota across spatial scales and
changes in lake levels using paleo-salinity reconstructions. However, the ecosystems. Nat. Ecol. Evol. 4, 1060–1068. https://doi.org/10.1038/s41559-020-
1216-4.
salinity inference model will likely be improved by including a greater Birks, H.J.B., Simpson, G.L., 2013. ‘Diatoms and pH reconstruction’ (1990) revisited.
number of lakes in the meso-hyposaline range as these were under- J. Paleolimnol. 49, 363–371. https://doi.org/10.1007/s10933-013-9697-7.

15
S. Yu et al. Ecological Indicators 155 (2023) 110952

Birks, H.J.B., Frey, D.G., Deevey, E.S., 1998. Numerical tools in palaeolimnology – Imbrie, J., Kipp, N.G., 1971. A new micropaleontological method for quantitative
Progress, potentialities, and problems. J. Paleolimnol. 20, 307–332. https://doi.org/ paleoclimatology: Application to a Late Pleistocene Caribbean core, The Late
10.1023/A:1008038808690. Cenozoic Glacial Ages K. Yale Univ. Press, New Haven, Conn. Turekian, 71–181,.
Birks, H.J.B., Heiri, O., Seppä, H., Bjune, A.E., 2010. Strengths and weaknesses of Immerzeel, W.W., Van Beek, L.P.H., Bierkens, M.F.P., 2010. Climate change will affect
quantitative climate reconstructions based on Late-Quaternary. The Open Ecology the Asian water towers. Science 328, 1382–1385. https://doi.org/10.1126/
Journal 3, 68–110. https://doi.org/10.2174/1874213001003020068. science.1183188.
Birks, H.J.B., 1995. Quantitative palaeoenvironmental reconstructions, in: Statistical Immerzeel, W.W., Lutz, A.F., Andrade, M., Bahl, A., Biemans, H., Bolch, T., Hyde, S.,
Modelling of Quaternary Science Data, Technical Guide 5, edited by: Maddy, D., and Brumby, S., Davies, B.J., Elmore, A.C., Emmer, A., Feng, M., Fernández, A.,
Brew, J. S., Quaternary Research Association, Cambridge, 271 pp. Haritashya, U., Kargel, J.S., Koppes, M., Kraaijenbrink, P.D.A., Kulkarni, A.V.,
Blanco, S., Álvarez-Blanco, I., Cejudo-Figueiras, C., De Godos, I., Bécares, E., Muñoz, R., Mayewski, P.A., Nepal, S., Pacheco, P., Painter, T.H., Pellicciotti, F., Rajaram, H.,
Guzman, H.O., Vargas, V.A., Soto, R., 2013. New diatom taxa from high-altitude Rupper, S., Sinisalo, A., Shrestha, A.B., Viviroli, D., Wada, Y., Xiao, C., Yao, T.,
Andean saline lakes. Diatom Res. 28 (1), 13–27. Baillie, J.E.M., 2020. Importance and vulnerability of the world’s water towers.
Bloom, A.M., Moser, K.A., Porinchu, D.F., MacDonald, G.M., 2003. Diatom-inference Nature 577, 364. https://doi.org/10.1038/s41586-019-1822-y.
models for surface-water temperature and salinity developed from a 57-lake Institute of Geography, 1990. Map of the Qinghai-Tibetan Plateau. Science Press, Beijing,
calibration set from the Sierra Nevada, California, USA. J. Paleolimnol. 29, 235–255. pp. 102–112 in Chinese.
https://doi.org/10.1023/A:1023297407233. Juggins, S., 2007. C2, Software for ecological and palaeoecological data analysis and
Bradshaw, E.G., Anderson, N.J., 2003. Environmental factors that control the abundance visualisation, User guide Version 1.5. School of Geography. Politics & Sociology,
of Cyclostephanos dubius (Bacillariophyceae) in Danish lakes, from seasonal to Newcastle University, Newcastle.
century scale. Eur. J. Phycol. 38, 265–276. https://doi.org/10.1080/ Juggins, S., 2013. Quantitative reconstructions in palaeolimnology: new paradigm or
0967026031000136349. sick science? Quat. Sci. Rev. 64, 20–32. https://doi.org/10.1016/j.
Chen, F.H., Chen, J.H., Holmes, J., Boomer, I., Austin, P., Gates, J.B., Wang, N.L., quascirev.2012.12.014.
Brooks, S.J., Zhang, J.W., 2010. Moisture changes over the last millennium in arid Kärnä, O.M., Heino, J., Grönroos, M., Hjort, J., 2018. The added value of geodiversity
central Asia: a review, synthesis and comparison with monsoon region. Quat. Sci. indices in explaining variation of stream macroinvertebrate diversity. Ecol. Ind. 94,
Rev. 29, 1055–1068. https://doi.org/10.1016/j.quascirev.2010.01.005. 420–429. https://doi.org/10.1016/j.ecolind.2018.06.034.
Chen, J., Liu, J.B., Rühland, K.M., Smol, J.P., Zhang, X.S., Zhang, Z.P., Zhou, A.F., Krammer, K., Lange-Bertalot, H., 1986,1988,1991a,1991b. Bacillariophyceae.
Shen, Z.W., Chen, F.H., 2021. Aquatic ecosystem responses to environmental and Süßwasserflora von Mitteleuropa. Gustav Fischer Verlag, Jena.
climatic changes in NE China since the last deglaciation (~17, 500 cal yr BP) tracked Laperriere, L., Fallu, M.A., Hausmann, S., Pienitz, R., Muir, D., 2008. Paleolimnological
by diatom assemblages from Lake Moon. Quat. Sci. Rev. 272, 107218 https://doi. evidence of mining and demographic impacts on Lac Dauriat, Schefferville (subarctic
org/10.1016/j.quascirev.2021.107218. Québec, Canada). J. Paleolimnol. 40, 309–324. https://doi.org/10.1007/s10933-
China salty lakes resources and environment database, 2019. Qinghai Institute of Salty 007-9162-6.
Lakes. Chinese Academy of Sciences, Xining, China. Legendre, P., Gallagher, E.D., 2001. Ecologically meaningful transformations for
Cumming, B.F., Smol, J.P., 1993. Development of diatom-based salinity models for ordination of species data. Oecologia 129 (2), 271–280. https://doi.org/10.1007/
paleoclimatic research from lakes in British Columbia (Canada). In: van Dam, H. s004420100716.
(Ed.), Twelfth International Diatom Symposium. Springer Netherlands, Dordrecht, Li, Z.X., Gao, Y., Wang, S.Y., Lu, Y., Sun, K., Jia, J.J., Wang, Y.F., 2021. Phytoplankton
pp. 179–196. community response to nutrients along lake salinity and altitude gradients on the
Cumming, B.F., Wilson, S.E., Hall, R.I., Smol, J.P., 1995. Diatoms from British Columbia Qinghai-Tibet Plateau. Ecol. Ind. 128, 107848 https://doi.org/10.1016/j.
(Canada) lakes and their relationship to salinity, nutrients and other limnological ecolind.2021.107848.
variables. In: Lange-Bertalot, H. (Ed.), Bibliotheca Diatomologica, Band 31. J. Li, C.D., Kang, S.C., Liu, Y.Q., Hou, J.Z., Guo, J.M., Liu, X.B., Cong, Z.Y., Zhang, Q.G.,
Cramer, Stuttgart, p. 207. 2016. Distribution of major ions in waters and their response to regional climatic
Fernández-Moreno, D., Sánchez-Castillo, P.M., Delgado, C., Almeida, S.F.P., 2022. change in Tibetan lakes. Journal of Lake Sciences 28 (4), 743–754. In Chinese with
Diatom species that characterize saline ponds (Southern Spain) with the description English abstract) DOI: 10.18307/2016.0407.
of a new Navicula species. Wetlands 42, 14. https://doi.org/10.1007/s13157-021- Li, Y.L., Liu, E.F., Xiao, X.Y., Zhang, E.L., Ji, M., 2015. Diatom response to Asian
01529-z. monsoon variability during the Holocene in a deep lake at the southeastern margin
Fritz, S.C., Juggins, S., Battarbee, R.W., Engstrom, D.R., 1991. Reconstruction of past of the Tibetan Plateau. Boreas 44, 785–793. https://doi.org/10.1111/bor.12128.
changes in salinity and climate using a diatom-based transfer function. Nature 352, Liao, M.N., Herzschuh, U., Wang, Y.B., Liu, X.Q., Ni, J., Li, K., 2020. Lake diatom
706–708. https://doi.org/10.1038/352706a0. response to climate change and sedimentary events on the southeastern Tibetan
Fritz, S.C., Juggins, S., Battarbee, R.W., 1993. Diatom assemblages and ionic Plateau during the last millennium. Quat. Sci. Rev. 241, 106409 https://doi.org/
characterization of lakes of the northern Great Plains, North America: a tool for 10.1016/j.quascirev.2020.106409.
reconstructing past salinity and climate fluctuations. Can. J. Fish. Aquat. Sci. 50 (9), Liu, X.D., Guo, Q.C., Guo, Z.T., Yin, Z.Y., Dong, B.W., Smith, R., 2015. Where were the
1844–1856. https://doi.org/10.1139/f93-20. monsoon regions and arid zones in Asia prior to the Tibetan Plateau uplift? Natl. Sci.
Fritz, S.C., Cumming, B.F., Gasse, F., Laird, K., 2010. Diatoms as indicators of hydrologic Rev. 2, 403–416. https://doi.org/10.1093/nsr/nwv068.
and climate change in saline lakes. In: Smol, J.P., Stoermer, E.F. (Eds.), the Diatoms: Liu, C., Zhu, L., Wang, J., Ju, J., Ma, Q., Qiao, B., Wang, Y., Xu, T., Chen, H., Kou, Q.,
Applications for the Environmental and Earth Sciences, 2nd ed. Cambridge Zhang, R., Kai, J., 2021. In-situ water quality investigation of the lakes on the
University Press, Cambridge, pp. 186–208. Tibetan Plateau. Science Bulletin 66 (17), 1727–1730.
Fritz, S.C., 2013. Salinity and climate reconstructions from continental lakes. Lu, H.Y., Wu, N.Q., Liu, K.B., Zhu, L.P., Yang, X.D., Yao, T.D., Wang, L., Li, Q., Liu, X.Q.,
Encyclopedia of Quaternary Science, 2nd ed. (Scott A. Elias and Cary J. Mock, Shen, C.M., Li, X.Q., Tong, G.B., Jiang, H., 2011. Modern pollen distributions in
editors) pp. 507-515. Qinghai-Tibetan Plateau and the development of transfer functions for
Gao, Y.C., Liu, M.F., 2013. Evaluation of high-resolution satellite precipitation products reconstructing Holocene environmental changes. Quat. Sci. Rev. 30, 947–966.
using rain gauge observations over the Tibetan Plateau. Hydrol. Earth Syst. Sci. 17, https://doi.org/10.1016/j.quascirev.2011.01.008.
837–849. https://doi.org/10.5194/hess-17-837-2013. Ma, Q.F., Zhu, L.P., Lü, X.M., Wang, J.B., Ju, J.T., Kasper, T., Daut, G., Haberzettl, T.,
Guiry, M.D., Guiry, G.M., 2023. AlgaeBase. World-wide electronic publication, National 2019. Late glacial and Holocene vegetation and climate variations at Lake Tangra
University of Ireland, Galway. Yumco, central Tibetan Plateau. Global Planet. Change 174, 16–25. https://doi.org/
Hall, R.I., Smol, J.P., 1992. A weighted-averaging regression and calibration model for 10.1016/j.gloplacha.2019.01.004.
inferring total phosphorus concentration from diatoms in British Columbia (Canada) Maasri, A., Jähnig, S.C., Adamescu, M.C., Adrian, R., Baigun, C., Baird, D.J., Batista-
lakes. Freshw. Biol. 27, 417–434. https://doi.org/10.1111/j.1365-2427.1992. Morales, A., Bonada, N., Brown, L.E., Cai, Q., Campos-Silva, J.V., Clausnitzer, V.,
tb00551.x. Contreras-MacBeath, T., Cooke, S.J., Datry, T., Delacámara, G., De Meester, L.,
Hammer, U.T., 1986. Saline lake ecosystems of the world, Vol. 59. Springer Science & Dijkstra, K.-D., Do, V.T., Domisch, S., Dudgeon, D., Erös, T., Freitag, H., Freyhof, J.,
Business Media. Friedrich, J., Friedrichs-Manthey, M., Geist, J., Gessner, M.O., Goethals, P.,
Hammer, Ø., Harper, D.A., Ryan, P.D., 2001. PAST: Paleontological statistics software Gollock, M., Gordon, C., Grossart, H.-P., Gulemvuga, G., Gutiérrez-Fonseca, P.E.,
package for education and data analysis. Palaeontol. Electron. 4, 9. http://palaeo-el Haase, P., Hering, D., Hahn, H.J., Hawkins, C.P., He, F., Heino, J., Hermoso, V.,
ectronica.org. Hogan, Z., Hölker, F., Jeschke, J.M., Jiang, M., Johnson, R.K., Kalinkat, G.,
Hausmann, S., Kienast, F., 2006. A diatom-inference model for nutrients screened to Karimov, B.K., Kasangaki, A., Kimirei, I.A., Kohlmann, B., Kuemmerlen, M.,
reduce the influence of background variables: application to varved sediments of Kuiper, J.J., Kupilas, B., Langhans, S.D., Lansdown, R., Leese, F., Magbanua, F.S.,
Greifensee and evaluation with measured data. Palaeogeogr. Palaeoclimatol. Matsuzaki, S.-I., Monaghan, M.T., Mumladze, L., Muzon, J., Mvogo Ndongo, P.A.,
Palaeoecol. 233 (1–2), 96–112. https://doi.org/10.1016/j.palaeo.2005.09.019. Nejstgaard, J.C., Nikitina, O., Ochs, C., Odume, O., Opperman, J.J., Patricio, H.,
He, J.N., Liu, J.B., Rühland, K.M., Zhang, J.F., Chen, Z.T., Dong, H.R., Smol, J.P., 2022. Pauls, S., Raghavan, R., Ramírez, A., Rashni, B., Ross-Gillespie, V., Samways, M.J.,
Response of lake diatoms to rapid 21st century warming on the southeastern Tibetan Schäfer, R.B., Schmidt-Kloiber, A., Seehausen, O., Shah, D.N., Sharma, S.,
Plateau. Anthropocene 39, 100345. https://doi.org/10.1016/j.ancene.2022.100345. Soininen, J., Sommerwerk, N., Stockwell, J.D., Suhling, F., Tachamo Shah, R.D.,
Hill, M.O., 1973. Diversity and evenness: a unifying notation and its consequences. Tharme, R.E., Thorp, J.H., Tickner, D., Tockner, K., Tonkin, J.D., Valle, M., Vitule, J.,
Ecology 54, 427–432. https://doi.org/10.2307/1934352. Volk, M., Wang, D., Wolter, C., Worischka, S., Thrall, P., 2022. A global agenda for
Hillebrand, H., Azovski, A.I., 2001. Body size determines the strength of the latitudinal advancing freshwater biodiversity research. Ecol. Lett. 25 (2), 255–263.
diversity gradient. Ecography 24. https://doi.org/10.1034/j1600- Meybeck, M., 1995. Physics and chemistry of lakes (eds Lerman A., Imboden D.M., Gat
0587.2001.240302.x. J.), Springer-Verlag, 36.
Hou, X. (Ed.), 2001. Vegetation Atlas of China. Science Press, Beijing in Chinese. Moser, K.A., Korhola, A., Weckstrröm, T.B., Pienitz, R., Smol, J.P., Douglas, M.S.V.,
Hay, M.B., 2000. Paleohydrology inferred from diatoms in northern latitude regions.
J. Paleolimnol. 24, 93–107. https://doi.org/10.1023/A:1008173901591.

16
S. Yu et al. Ecological Indicators 155 (2023) 110952

Murakami, T., Terai, H., Yoshiyama, Y., Tezuka, T., Zhu, L., Matsunaka, T., moisture balance in continental Antarctic lakes. J. Paleolimnol. 30, 195–215.
Nishimura, M., 2007. The second investigation of Lake Puma Yum Co located in the https://doi.org/10.1023/A1025570904093.
Southern Tibetan Plateau, China. Limnology 8, 331–335. https://doi.org/10.1007/ Wang, R., Hu, Z., Wang, Q., Xu, M., Zheng, W., Zhang, K.e., Yang, X., 2020b. Discrepancy
s10201-007-0208-2. in the responses of diatom diversity to indirect and direct human activities in lakes of
Oksanen, A., Savela, N., Latikka, R., Koivula, A., 2020. Trust toward robots and artificial the southeastern Tibetan Plateau, China. Anthropocene 30, 100243.
intelligence: an experimental approach to human–technology interactions. Front. Wang, L., Jiang, W.Y., Jiang, D.B., Zou, Y.F., Liu, Y.Y., Zhang, E.L., Hao, Q.Z., Zhang, D.
Psychol. 11 https://doi.org/10.3389/fpsyg.2020.568256. G., Zhang, D.T., Peng, Z.Y., Xu, B., Yang, X.D., Lu, H.Y., 2018. Prolonged heavy
Pi, X.H., Feng, L., Li, W.F., Liu, J.G., Kuang, X.X., Shi, K., Qi, W., Chen, D.L., Tang, J., snowfall during the Younger Dryas. J. Geophys. Res. Atmos. 123 https://doi.org/
2021. Chlorophyll-a concentrations in 82 large alpine lakes on the Tibetan Plateau 10.1029/2018JD029271.
during 2003–2017: temporal-spatial variations and influencing factors. Int. J. Digital Wang, P., Shang, Y.N., Shen, L.C., Wu, K.Y., Xiao, Q., 2013. Characteristics and evolution
Earth 14, 714–735. https://doi.org/10.1080/17538947.2021.1872722. of hydrochemical compositions of freshwater lake in Tibetan plateau. Huan Jing Ke
Pienitz, R., Smol, J.P., 1993. Diatom assemblages and their relationship to environmental Xue 34, 874–881. In Chinese with English abstract.
variables in lakes from the boreal forest-tundra ecotone near Yellowknife, Northwest Wang, M.D., Tian, Q., Li, X.M., Liang, J., He, Y., Hou, J.Z., 2020a. TEX86 as a potential
Territories, Canada. Hydrobiologia 269, 391–404. https://doi.org/10.1007/ proxy of lake water pH in the Tibetan Plateau. Palaegeography, Palaeoclimatology,
BF00028037. Palaeoecology 538, 109381. https://doi.org/10.1016/j.palaeo.2019.109381.
Qiao, B.J., Wang, J.B., Huang, L., Zhu, L.P., 2017. Characteristics and seasonal variations Wang, T.Y., Wu, T.H., Wang, P., Li, R., Xie, C.W., Zou, D.F., 2019. Spatial distribution
in the hydrochemistry of the Tangra Yumco basin, central Tibetan Plateau, and and changes of permafrost on the Qinghai-Tibet Plateau revealed by statistical
responses to the Indian summer monsoon. Environ. Earth Sci. 76, 162. https://doi. models during the period of 1980 to 2010. Sci. Total Environ. 650, 661–670. https://
org/10.1007/s12665-017-6479-y. doi.org/10.1016/j.scitotenv. 2018.08.398.
Qiu, J., 2008. The third pole. Nature 454, 393–396. https://doi.org/10.1038/454393a. Wang, R., Yang, X.D., Langdon, P., Zhang, E.L., 2011. Limnological responses to warming
R core team,, 2018. R: A language and environment for statistical computing. R on the Xizang Plateau, Tibet, over the past 200 years. J. Paleolimnol. 45, 257–271.
Foundation for Statistical Computing, Vienna, Austria. https://doi.org/10.1007/s10933-011-9496-y.
Rivera-Rondón, C.A., Catalan, J., 2020. Diatoms as indicators of the multivariate Wang, J.B., Zhu, L.P., Wang, Y., Ju, J.T., Xie, M.P., Daut, G., 2010. Comparisons between
environment of mountain lakes. Sci. Total Environ. 703, 135517 https://doi.org/ the chemical compositions of lake water, inflowing river water, and lake sediment in
10.1016/j.scitotenv.2019.135517. Nam Co, central Tibetan Plateau, China and their controlling mechanisms. J. Great
Saunders, K.M., 2011. A diatom dataset and diatom-salinity inference model for Lakes Res. 36, 587–595. https://doi.org/10.1016/j.jglr.2010.06.013.
southeast Australian estuaries and coastal lakes. J. Paleolimnol. 46 (4), 525–542. Wilson, S.E., Cumming, B.F., Smol, J.P., 1994. Diatom-salinity relationships in 111 lakes
Seddon, P.J., Griffiths, C.J., Soorae, P.S., Armstrong, D.P., 2014. Reversing defaunation: from the Interior Plateau of British Columbia, Canada: the development of diatom-
Restoring species in a changing world. Science 345, 406–412. https://doi.org/ based models for paleosalinity reconstructions. J. Paleolimnol. 12 (3), 197–221.
10.1126/science.1251818. Wilson, S.E., Cumming, B.F., Smol, J.P., 1996. Assessing the reliability of salinity
Smol, J.P., 2009. Pollution of lakes and rivers: a paleoenvironmental perspective. John inference models from diatom assemblages: an examination of a 219-lake data set
Wiley & Sons. from western North America. Journal of Fisheries and Aquatic Sciences 53 (7),
Smol, J.P., 2019. Under the radar: long-term perspectives on ecological changes in lakes. 1580–1594. https://doi.org/10.1139/f96-094.
Proc. R. Soc. B 286, 20190834. https://doi.org/10.1098/rspb.2019.0834. Wischnewski, J., Kramer, A., Kong, Z., Mackay, A.W., Simpson, G.L., Mischke, S.,
Smol, J.P., Cumming, B.F., 2000. Tracking long-term changes in climate using algal Herzschuh, U., 2011a. Terrestrial and aquatic responses to climate change and
indicators in lake sediments. J. Phycol. 36, 986–1011. https://doi.org/10.1046/ human impact on the southeastern Tibetan Plateau during the past two centuries.
j.1529-8817.2000.00049.x. Glob. Chang. Biol. 17, 3376–3391. https://doi.org/10.1111/j.1365-
Smol, J.P., Stoermer, E.F., (editors),, 2010. The diatoms: applications for the 2486.2011.02474.x.
environmental and earth sciences. Cambridge University Press. Wischnewski, J., Mischke, S., Wang, Y., Herzschuh, U., 2011b. Reconstructing climate
Stenger-Kovács, C., Béres, V.B., Buczkó, K., Tapolczai, K., Padisák, J., Selmeczy, G.B., variability on the northeastern Tibetan Plateau since the last Lateglacial – a multi-
Lengyel, E., 2023. Diatom community response to inland water salinization: a proxy, dual-site approach comparing terrestrial and aquatic signals. Quat. Sci. Rev.
review. Hydrobiologia. https://doi.org/10.1007/s10750-023-05167-w. 30, 82–97. https://doi.org/10.1016/j.quascirev.2010.10.001.
Su, Y.L., Li, K.Y., Zhang, Y.D., Liu, Z.W., Wang, T.J., Jeppesen, E., Middelburg, J.J., Wischnewski, J., Herzschuh, U., Rühland, K.M., Brauning, A., Mischke, S., Smol, J.P.,
Smol, J.P., 2022. Effects of climate change and nutrients concentrations on carbon Wang, L., 2014. Recent ecological responses to climate variability and human
sources for zooplankton in a Tibetan Plateau lake over the past millennium. impacts in the Nianbaoyeze Mountains (eastern Tibetan Plateau) inferred from
J. Paleolimnol. 68, 249–263. https://doi.org/10.1007/s10933-022-00245-. pollen, diatom and tree-ring data. J. Paleolimnol. 51, 287–302. https://doi.org/
Sun, H., Liao, K., Pan, Y., Wang, J., 1990. Altas of the Qinghai-Tibet plateau. Science 10.1007/s10933-013-9747-1.
Press, Beijing, p. 17.67-69.. Yang, X.D., Kamenik, C., Schmidt, R., Wang, S.M., 2003. Diatom-based conductivity and
Taukulis, F.E., John, J., 2009. Development of a diatom-based transfer function for lakes water level inference models from eastern Tibetan (Qinghai-Xizang) Plateau lakes.
and streams severely impacted by secondary salinity in the south-west region of J. Paleolimnol. 30, 1–19. https://doi.org/10.1023/A:1024703012475.
Western Australia. Hydrobiologia 626 (1), 129–143. https://doi.org/10.1007/ Yang, H.D., Turner, S., 2013. Radiometric dating for recent lake sediments on the Tibetan
s10750-009-9741-9. Plateau. Hydrobiologia 713, 73–86. https://doi.org/10.1007/s10750-013-1493-x.
Ter Braak, C.J.F., 1988. CANOCO – a FORTRAN Program for Canonical Community Yao, T., Thompson, L.G., Mosbrugger, V., Zhang, F., Ma, Y., Luo, T., Xu, B., Yang, X.,
Ordination by [Partial] [Detrended] [Canonical] Correspondence Analysis, Principal Joswiak, D.R., Wang, W., Joswiak, M.E., Devkota, L.P., Tayal, S., Jilani, R.,
Components Analysis and Redundancy Analysis (Version 2.1). Technical Report Fayziev, R., 2012. Third Pole Environment (TPE). Environ. Dev. 3, 52–64. https://
LWA-88-02. Agricultural Mathematics Group, Wageningen. doi.org/10.1016/j.envdev. 2012.04.002.
Ter Braak, C.J.F., 1992. CANOCO a Fortran program for Canonical Community Yu, S.W., Wang, J.B., Li, Y.M., Peng, P., Kai, J.L., Kou, Q.Q., Laug, A., 2019. Spatial
Ordination. Microcomputer Power, New York, p. 351. distribution of diatom assemblages in the surface sediments of Selin Co, central
Ter Braak, C.J.F., Šmilauer, P., 2012. Canoco reference manual and user’s guide: Tibetan Plateau, China, and the controlling factors. J. Great Lakes Res. 45,
software for ordination, version 5.0. Microcomputer Power. 1069–1079. https://doi.org/10.1016/j.jglr.2019.09.006.
Ter Braak, C.J.F., Prentice, I.C., 2004. A theory of gradient analysis. Adv. Ecol. Res. 34, Zhang, X.S., Liu, J.B., Rühland, K.M., Jia, X., Reed, J.M., Li, Y.L., Shen, Z.W., Zhao, J.J.,
235–282. https://doi.org/10.1016/S0065-2504(08)60183-X. Chen, J., Wang, H.P., Wang, X., Smol, J.P., Chen, F.H., 2023. Concurrent mid-
Thienpont, J.R., Ruehland, K.M., Pisaric, M.F., Kokelj, S.V., Kimpe, L.E., Blais, J.M., Holocene East Asian temperature and summer monsoon maxima forced by high- and
Smol, J.P., 2013. Biological responses to permafrost thaw slumping in Canadian low-latitude interplay. Global Planet. Change 220, 104008. https://doi.org/
Arctic lakes. Freshwater Biol. 58, 337–353. https://doi.org/10.1111/fwb.12061. 10.1016/j.gloplacha.2022.104008.
Veres, A.J., Pienitz, R., Smol, J.P., 1995. Lake water salinity and periphytic diatom Zhao, L., Li, Y.N., Xu, S.X., Zhou, H.K., Gu, S., Yu, G.R., Zhao, X.Q., 2006. Diurnal,
succession in three subarctic lakes, Yukon Territory, Canada. Arctic 48, 63–70. seasonal and annual variation in net ecosystem CO2 exchange of an alpine shrubland
http://www.jstor.org/stable/40511618. on Qinghai-Tibetan plateau. Glob. Chang. Biol. 12, 1940–1953. https://doi.org/
Verleyen, E., Hodgson, D.A., Vyverman, W., Roberts, D., McMinn, A., Vanhoutte, K., 10.1111/j.1365-2486.2006.01197.x.
Sabbe, K., 2003. Modelling diatom responses to climatic induced fluctuations in the

17

You might also like