Light-dominated selection shaping filamentous cyanobacterial assemblages drives odor problem in a drinking water reservoir

Supplementary Material

Authors
Affiliations

Ming Su

Key Laboratory of Drinking Water Science and Technology, Research Center for Eco-Environmental Sciences, Chinese Academy of Sciences

University of Chinese Academy of Sciences

Yiping Zhu

Shanghai Chengtou Raw Water Co. Ltd

Tom Andersen

Department of Biosciences, University of Oslo

Xianyun Wang

National Engineering Research Center of China (South) for Urban Water

Zhiyong Yu

Key Laboratory of Drinking Water Science and Technology, Research Center for Eco-Environmental Sciences, Chinese Academy of Sciences

University of Chinese Academy of Sciences

Jinping Lu

Key Laboratory of Drinking Water Science and Technology, Research Center for Eco-Environmental Sciences, Chinese Academy of Sciences

University of Chinese Academy of Sciences

Yichao Song

Shanghai Chengtou Raw Water Co. Ltd

Tengxin Cao

Key Laboratory of Drinking Water Science and Technology, Research Center for Eco-Environmental Sciences, Chinese Academy of Sciences

University of Chinese Academy of Sciences

Jianwei Yu

Key Laboratory of Drinking Water Science and Technology, Research Center for Eco-Environmental Sciences, Chinese Academy of Sciences

University of Chinese Academy of Sciences

Yu Zhang

Key Laboratory of Drinking Water Science and Technology, Research Center for Eco-Environmental Sciences, Chinese Academy of Sciences

University of Chinese Academy of Sciences

Min Yang

Key Laboratory of Drinking Water Science and Technology, Research Center for Eco-Environmental Sciences, Chinese Academy of Sciences

University of Chinese Academy of Sciences

Published

Thursday, January 9, 2025

1 Key Laboratory of Drinking Water Science and Technology, Research Center for Eco-Environmental Sciences, Chinese Academy of Sciences
2 Shanghai Chengtou Raw Water Co. Ltd
3 Department of Biosciences, University of Oslo
4 National Engineering Research Center of China (South) for Urban Water
5 University of Chinese Academy of Sciences

✉ Correspondence: Min Yang <yangmin@rcees.ac.cn>

Supplementary Material

All data analysis and illustration were performed using R 4.01. Data pretreatment and summary were performed using the dplyr2 and base packages in R, regression analysis including linear and generalized linear models were performed using the stats package1, generalized additive modelling was performed using the mgcv package3,4 quantile regression analysis was performed using the quantreg package5; contour figures were created by the graphics package1, other figures were prepared using the ggplot2 package6.

Fig. S1 Seasonal dynamics of MIB of Yangtze River water (inlet) from 2011 to 2015”

Fig. S2 Composition of filamentous cyanobacteria and Microcystis in QCS Reservoir

Fig. S3 Seasonal dynamics and long-term trends of Planktothrix, Pseudanabaena, Lyngbya, Phormidium and Microcystis in QCS Reservoir

Fig. S4 Seasonal dynamics and long-term trends of nutrients (total nitrogen, nitrate, ammonia and total phosphorus) in QCS Reservoir

Fig. S5 Correlation between factors and filamentous abundances
Year Season Min Max Median IQR Mean SD SE CI
2013 Spring 0.037 0.078 0.055 0.013 0.055 0.010 0.002 0.004
2013 Summer 0.033 0.139 0.058 0.018 0.063 0.023 0.003 0.006
2013 Autumn 0.035 0.127 0.067 0.027 0.065 0.021 0.004 0.007
2013 Winter 0.028 0.201 0.053 0.089 0.088 0.064 0.021 0.050
2014 Spring 0.045 0.107 0.071 0.019 0.074 0.016 0.002 0.005
2014 Summer 0.032 0.093 0.059 0.016 0.061 0.013 0.001 0.003
2014 Autumn 0.024 0.105 0.044 0.027 0.051 0.019 0.003 0.007
2014 Winter 0.034 0.183 0.067 0.024 0.078 0.034 0.007 0.015
2015 Spring 0.034 0.104 0.069 0.019 0.068 0.016 0.003 0.006
2015 Summer 0.033 0.109 0.058 0.022 0.062 0.016 0.002 0.005
2015 Autumn 0.039 0.077 0.054 0.008 0.056 0.011 0.003 0.007
2015 Winter 0.022 0.081 0.053 0.028 0.052 0.022 0.009 0.023
Table S1 Descriptive statistics of ammonia nitrogen in QCS Reservoir (unit: mg L-1)
Year Season Min Max Median IQR Mean SD SE CI
2013 Spring 1.213 1.702 1.415 0.179 1.431 0.118 0.024 0.05
2013 Summer 0.759 1.665 1.110 0.269 1.137 0.194 0.025 0.05
2013 Autumn 0.843 1.71 1.029 0.445 1.140 0.281 0.047 0.095
2013 Winter 1.390 1.828 1.557 0.167 1.600 0.136 0.045 0.105
2014 Spring 1.578 2.162 1.796 0.121 1.800 0.130 0.019 0.038
2014 Summer 0.944 1.737 1.293 0.176 1.306 0.163 0.019 0.038
2014 Autumn 1.159 1.707 1.309 0.119 1.352 0.129 0.023 0.047
2014 Winter 1.561 2.362 2.008 0.196 2.008 0.189 0.039 0.082
2015 Spring 1.157 1.793 1.485 0.271 1.474 0.186 0.033 0.068
2015 Summer 0.895 1.867 1.411 0.414 1.333 0.264 0.037 0.074
2015 Autumn 0.729 1.365 1.145 0.234 1.092 0.189 0.052 0.114
2015 Winter 1.595 1.774 1.708 0.131 1.694 0.078 0.032 0.082
Table S2 Descriptive statistics of nitrate nitrogen in QCS Reservoir (unit: mg L-1)
Year Season Min Max Median IQR Mean SD SE CI
2013 Spring 1.008 1.469 1.326 0.156 1.313 0.113 0.023 0.048
2013 Summer 0.683 1.411 1.126 0.150 1.094 0.153 0.020 0.040
2013 Autumn 0.319 1.592 1.006 0.583 1.052 0.351 0.059 0.119
2013 Winter 1.367 1.735 1.543 0.161 1.542 0.132 0.044 0.101
2014 Spring 1.478 2.265 1.808 0.217 1.795 0.164 0.024 0.048
2014 Summer 0.808 1.705 1.237 0.192 1.228 0.195 0.023 0.045
2014 Autumn 0.959 1.601 1.236 0.125 1.250 0.142 0.025 0.052
2014 Winter 1.361 2.429 2.073 0.337 1.974 0.259 0.054 0.112
2015 Spring 1.234 1.875 1.419 0.191 1.450 0.154 0.028 0.056
2015 Summer 0.718 1.654 1.317 0.355 1.292 0.223 0.031 0.063
2015 Autumn 0.726 1.348 1.013 0.135 1.017 0.156 0.043 0.094
2015 Winter 1.348 1.75 1.59 0.279 1.581 0.177 0.072 0.186
Table S3 Descriptive statistics of total nitrogen (TN) in QCS Reservoir (unit: mg L-1)
Year Season Min Max Median IQR Mean SD SE CI
2013 Spring 0.053 0.095 0.077 0.016 0.078 0.012 0.003 0.005
2013 Summer 0.033 0.25 0.064 0.023 0.07 0.036 0.005 0.009
2013 Autumn 0.046 0.121 0.070 0.017 0.072 0.015 0.003 0.005
2013 Winter 0.055 0.095 0.060 0.018 0.067 0.013 0.004 0.010
2014 Spring 0.050 0.133 0.089 0.022 0.089 0.019 0.003 0.005
2014 Summer 0.047 0.118 0.081 0.024 0.079 0.017 0.002 0.004
2014 Autumn 0.056 0.126 0.084 0.011 0.088 0.013 0.002 0.005
2014 Winter 0.058 0.133 0.092 0.037 0.091 0.021 0.004 0.009
2015 Spring 0.042 0.115 0.073 0.023 0.073 0.019 0.003 0.007
2015 Summer 0.046 0.222 0.077 0.022 0.083 0.027 0.004 0.007
2015 Autumn 0.065 0.092 0.072 0.012 0.075 0.008 0.002 0.005
2015 Winter 0.069 0.096 0.075 0.006 0.078 0.010 0.004 0.010
Table S4 Descriptive statistics of total phosphrus (TP) in QCS Reservoir (unit: mg L-1)
Year Season Min Max Median IQR Mean SD SE CI
2013 Spring 11.208 24.642 19.468 5.594 19.475 4.088 0.834 1.726
2013 Summer 24.563 30.343 28.701 1.977 28.348 1.609 0.208 0.416
2013 Autumn 11.564 27.028 21.502 7.032 20.788 5.019 0.836 1.698
2013 Winter 3.927 9.514 5.846 4.501 6.853 2.403 0.801 1.847
2014 Spring 12.100 25.821 20.99 6.767 20.322 4.091 0.591 1.188
2014 Summer 25.688 28.549 27.386 0.832 27.367 0.691 0.08 0.159
2014 Autumn 12.874 26.943 21.812 7.086 21.447 4.271 0.767 1.567
2014 Winter 7.19 11.001 8.403 1.759 8.637 1.168 0.244 0.505
2015 Spring 11.17 25.049 22.13 3.92 20.856 4.131 0.742 1.515
2015 Summer 24.69 31.491 28.798 2.92 28.475 1.953 0.273 0.549
2015 Autumn 16.102 28.22 26.413 5.503 23.534 4.587 1.272 2.772
2015 Winter 7.861 8.981 8.092 0.554 8.278 0.447 0.182 0.469
Table S5 Descriptive statistics of water temperature in QCS Reservoir (unit: °C)
Year Season Min Max Median IQR Mean SD SE CI
2013 Spring 0.815 1.879 1.273 0.410 1.322 0.314 0.064 0.133
2013 Summer 0.901 2.261 1.638 0.810 1.621 0.416 0.054 0.107
2013 Autumn 0.701 1.468 1.152 0.409 1.110 0.247 0.041 0.084
2013 Winter 0.483 1.311 0.798 0.139 0.873 0.298 0.099 0.229
2014 Spring 0.854 1.977 1.552 0.395 1.528 0.284 0.041 0.083
2014 Summer 0.666 1.840 1.226 0.453 1.235 0.316 0.036 0.073
2014 Autumn 0.643 1.487 0.957 0.368 0.999 0.254 0.046 0.093
2014 Winter 0.461 1.254 0.938 0.241 0.857 0.248 0.052 0.107
2015 Spring 0.870 2.033 1.462 0.572 1.418 0.355 0.064 0.130
2015 Summer 0.734 2.138 1.235 0.482 1.293 0.352 0.049 0.099
2015 Autumn 0.638 1.460 1.253 0.185 1.195 0.329 0.091 0.199
2015 Winter 0.415 0.994 0.598 0.297 0.700 0.239 0.097 0.251
Table S6 Descriptive statistics of pre-week PAR in QCS Reservoir (unit: mol m-2 d-1)
Term EDF Ref.DF Statistic P.Value
s(nweek) 5.293509 8.000000 21.395662 0.000000
s(ndate) 1.757591 1.757591 3.755324 0.091522
Table S7 Summary of time series analysis for Planktothrix; s(nweek) denotes the seasonal pattern, and s(ndate) denotes the long-term pattern

Fig. S6 Time series analysis of Planktothrix based on seasonal and long-term trend smooth functions

Seasonal and long-term trends of Planktothrix and Pseudanabaena were evaluated using time series analysis, as illustrated in Table S7, Fig. S6, Table S8 and Fig. S7.

Term EDF Ref.DF Statistic P.Value
s(nweek) 6.246786 8.000000 11.302972 0.000000
s(ndate) 1.943873 1.943873 82.456653 0.000000
Table S8 Summary of time series analysis for Pseudanabaena; s(nweek) denotes the seasonal pattern, and s(ndate) denotes the long-term pattern

Fig. S7 Time series analysis for Pseudanabaena based on seasonal and long-term trend smooth functions

We select water temperature (temp), pre-week photosynthetically active radiation (weekPAR), total nitrogen (TN), nitrate (NO3), total phosphate (TP), ammonia (NH4), wind speed (wind) and maximum daily air temperature (maxtemp) as the potential predictors for abundances of Planktothrix (Table S9) and Pseudanabaena (Table S10). Linear model (\(Y = \sum{\beta_iX_i} + \beta_0 + \epsilon\), \(\epsilon \in N(0, 1), i = 1, 2, ...\)) between these predictors (\(X_i\), i = 1, 2, …) and logarithm transformed abundances (\(Y = log_{10}(1 + N)\)).

Term Estimate Std.Error Statistic P.Value
(Intercept) 0.055381 0.204114 0.271325 0.786326
temp 0.020686 0.006190 3.342053 0.000936
weekPAR -0.043151 0.062931 -0.685693 0.493431
TN -0.125703 0.134448 -0.934960 0.350554
NO3 -0.110292 0.143055 -0.770975 0.441323
TP 0.915797 1.105912 0.828092 0.408270
NH4 -0.721135 0.911296 -0.791330 0.429371
wind 0.085435 0.042653 2.003032 0.046064
maxtemp -0.003568 0.005304 -0.672707 0.501646
Table S9 Correlation analysis between Planktothrix and environmental factors (named as LM1)
Term Estimate Std.Error Statistic P.Value
(Intercept) -0.249268 0.208706 -1.194353 0.233274
temp 0.041253 0.006329 6.518053 2.97e-10
weekPAR -0.119826 0.064347 -1.862196 0.063543
TN 0.189667 0.137473 1.379667 0.168707
NO3 -0.068600 0.146274 -0.468984 0.639418
TP 0.202906 1.130792 0.179437 0.857715
NH4 -0.737178 0.931798 -0.791136 0.429484
wind -0.034087 0.043612 -0.781579 0.435072
maxtemp -0.012824 0.005423 -2.364817 0.018669
Table S10 Correlation analysis between Pseudanabaena and environmental factors (named as LM2)

We performed Backward Stepwise Regressions to identify the significant variables responsible for the abundances of Planktothrix and Pseudanabaena. The regression started with a model that contains all variables, and then removing the least significant variables one by one, until a pre-specified stopping rule (here we use AIC rule) is reached.

Below is the Backward Stepwise Regressions of linear model for Planktothrix, the water temperature (temp), pre-week PAR (weekPAR), total nitrogen (TN) and wind speed (wind) were considered as effect predictors for Planktothrix.

Term Estimate Std.Error Statistic P.Value
(Intercept) 0.029206 0.183201 0.159421 0.873442
temp 0.018787 0.003960 4.744185 3.21e-06
weekPAR -0.066606 0.057760 -1.153142 0.249749
TN -0.194775 0.073604 -2.646272 0.008558
wind 0.075903 0.040487 1.874726 0.061780
Table S11 Backward Stepwise Regression of linear models for Planktothrix

Below is the Backward Stepwise Regressions of linear model for Pseudanabaena, the water temperature (temp), pre-week PAR (weekPAR), total nitrogen (TN), ammonia (NH4) and maximum daily air temperature (maxtemp) were considered as effect predictors for Pseudanabaena.

Term Estimate Std.Error Statistic P.Value
(Intercept) -0.351 0.162542 -2.159404 0.031596
temp 0.043007 0.005947 7.231744 3.84e-12
weekPAR -0.118911 0.063542 -1.871372 0.062248
TN 0.142841 0.078959 1.809052 0.071424
NH4 -0.888384 0.900174 -0.986903 0.324470
maxtemp -0.013610 0.005303 -2.566502 0.010749
Table S12 Backward Stepwise Regression of linear models for Pseudanabaena

Backward Stepwise Regression approaches were performed for the two linear models to find out the optimum selections of predictors, as summarize in Table S11 (Planktothrix) and Table S12 (Pseudanabaena), respectively. Water temperature (temp), pre-week PAR (weekPAR), total nitrogen (TN), wind speed (wind) and maximum daily air temperature (maxtemp) were sorted out.

Parameter VIF
temp 2.025278
weekPAR 1.449414
TN 1.734340
TP 1.148326
NH4 1.179282
wind 1.119961
Table S13 VIF analysis of selected predictors of Planktothrix and Pseudanabaena models

According to the Backward Stepwise Regressions, water temperature (temp), pre-week PAR (weekPAR), total nitrogen (TN), total phosphate (TP), ammonia (NH4) and wind speed (wind) were selected as the valid predictors and used for following general additive models (GAMs). Variance inflation factors (VIF) were calculated and validated the rationality (<5) for these predictors , as summarized in Table S13.

Fig. S8 Correlation between environmental factors

Correlation coefficients among these predictors were computed by Pearson Method, as illustrated in Fig. S8. The results were used for the optimization of non-parameter smooth functions (smoother) in following GAMs for Planktothrix and Pseudanabaena. Specifically, high correlation coefficient between water temperature (temp) and pre-week PAR (PAR) suggests they are interacting predictors; similarly, total nitrogen (TN) and ammonia (NH4) are interacting predictors.

Term edf ref.df Statistic p-value
t2(temp, weekPAR) 2.999974 2.999974 8.441657 2.12287e-05
t2(TN, NH4) 2.999898 2.999898 2.800880 0.04014
s(TP) 1.000008 1.000008 0.759490 0.38418
s(wind) 1.000000 1.000000 3.944025 0.04794
Table S14 Summary of GAM model with 6 predictors for Planktothrix (named as GAM1)

Iterations between water temperature and pre-week PAR (t2(temp, weekPAR)), and between total nitrogen and ammonia (t2(TN, NH4)) were both evaluated by the tensor products (\(f_1(x_1) \otimes f_2(x_2)\)).

The results of “GAM1” suggest water temperature (temp), pre-week PAR (weekPAR), total nitrogen (TN) and ammonia (NH4) are most important predictors for Planktothrix abundance.

Term edf ref.df Statistic p-value
t2(temp, weekPAR) 5.733091 5.733091 9.604014 0
t2(TN, NH4) 2.999999 2.999999 0.964346 0.41001
s(TP) 2.236544 2.236544 2.650483 0.1095
s(wind) 4.787511 4.787511 3.886710 0.00265
Table S15 Summary of GAM model with 6 predictors for Pseudanabaena (named as GAM2)

The results of “GAM2” suggest water temperature (temp), pre-week PAR (weekPAR), total phosphrus (TP) and wind speed (wind) are most important predictors for Pseudanabaena abundance.

Term edf ref.df Statistic p-value
t2(temp, weekPAR) 2.999993 2.999993 8.439392 1.912541e-05
t2(TN, NH4) 2.999870 2.999870 4.088651 0.00689
Table S16 Summary of GAM model with 4 predictors for Planktothrix (named as GAM3)

According to “GAM1”, the predictors of Planktothrix abundances were further optimized to 4 factors, which are water temperature (temp), pre-week PAR (weekPAR), total nitrogen (TN) and ammonia (NH4). The summary of optimized Planktothrix GAM (GAM3) results is shown in Table S16.

Term edf ref.df Statistic p-value
t2(temp, weekPAR) 5.870002 5.870002 9.200651 0
s(TP) 2.222784 2.222784 3.073369 0.06404
s(wind) 4.665106 4.665106 3.782337 0.00378
Table S17 Summary of GAM model with 4 predictors for Pseudanabaena (named as GAM4)

According to “GAM2”, the predictors of Pseudanabaena abundances were further optimized to 4 factors, which are water temperature (temp), pre-week PAR (weekPAR), total phosphorus (TP) and wind speed (wind). The summary of optimized Pseudanabaena GAM (GAM4) results is shown in Table S17.

R Language demonstration code

# demo data frame
require(lubridate)
require(mgcv)
modeldf <- data.frame(date = ymd("2020-01-01") + 0:365,
                      y = rnorm(366))
# seasonal pattern: week number
modeldf$x1 <- week(modeldf$date)
# long-term pattern: decimal number
modeldf$x2 <- year(modeldf$date) + yday(modeldf$date) / 366
# demo gam model
m <- gamm(y ~ s(x1, bs = "cc") + s(x2), data = modeldf)

References

1.
R Core Team. R: A language and environment for statistical computing. (R Foundation for Statistical Computing, 2020).
2.
Wickham, H., François, R., Henry, L. & Müller, K. dplyr: A grammar of data manipulation. (2018).
3.
Wood, S. N. Stable and efficient multiple smoothing parameter estimation for generalized additive models. Journal of the American Statistical Association 99, 673–686 (2004).
4.
Wood, S. N. Fast stable restricted maximum likelihood and marginal likelihood estimation of semiparametric generalized linear models. Journal of the Royal Statistical Society (b) 73, 3–36 (2011).
5.
Koenker, R. Quantreg: Quantile regression. (2017).
6.
Wickham, H. ggplot2: Elegant graphics for data analysis. (Springer-verlag New York, 2016).