Research ArticleECOLOGY

Terrestrial support of lake food webs: Synthesis reveals controls over cross-ecosystem resource use

See allHide authors and affiliations

Science Advances  22 Mar 2017:
Vol. 3, no. 3, e1601765
DOI: 10.1126/sciadv.1601765
  • Fig. 1 Model recovers known parameters across 100 simulated data sets that replicate our empirical observations.

    Mean posterior distributions of the effects of (A) DOC (gray), normalized difference vegetation index (NDVI) (pink), ratio of lake perimeter to area (blue), and area of woody vegetation per meter shoreline (green) on availability of allochthonous resources and (B) allochthonous resources (purple), lake chlorophyll a (red), and an allochthonous resources–chlorophyll a interaction (orange) on terrestrial resource use (ϕT); dashed lines are known prior distributions. (C) Mean predicted ϕT versus observed (that is, known) ϕT for 559 consumer observations in each of the 100 simulations. Warmer colors indicate greater concentration of points (total n = 55,900). (D) Percent bias in mean predicted ϕT values. Darker shading indicates greater concentration of points. Lines are splines fitted through observations on one (δ2H only; pink), two (δ13C-δ15N; green), or three (δ13C-δ15N-δ2H; blue) isotopes.

  • Fig. 2 Terrestrial resource (ϕT) use by lake zooplankton.

    (A) Mean posterior estimates of ϕT for each of the 559 consumer observations. (B) Scaled distributions of key catchment characteristics and unscaled means and SDs. (C) Focal lake regions (n = 14) superimposed on water bodies at a resolution of 1 km and a proxy of vegetation density (NDVI) at a resolution of 0.1° in September 2015 (NASA Earth Observations data repository, http://neo.sci.gsfc.nasa.gov/).

  • Fig. 3 Modeled network of factors influencing terrestrial resource use (ϕT) by aquatic consumers across 147 lakes.

    Arrows point at modeled variables, with mean effects of one variable on another proportional to standardized effect size (see legend). Lines ending in circles are interactions. The asterisk symbol (*) indicates random variation among consumers, with colors showing direction of significant effects. Black lines are intercepts with no “effect direction,” ellipses are unobserved (that is, latent) variables, and gray boxes are covariates included to explain the connections between modeled variables and predictors of interest better. Five mechanisms explaining variation in ϕT are associated with broken boxes. NDVI, vegetation density; temp, mean monthly temperature of warmest quarter; woody, area of woody vegetation in catchment per meter shoreline; LP/LA, ratio of lake perimeter to lake area; group ID, research group that collected the data (such as to account for variation in sampling). Bayesian R2 for consumers with one (δ2H only), two (δ13C-δ15N), or three (δ13C-δ15N-δ2H) observed isotopes were 0.64, 0.98, and 0.99, respectively (fig. S5).

  • Fig. 4 ϕT increases with t-OM.

    Specifically, ϕT increased with the estimated availability of DOC (A), POC (B), and their summed contribution toward allochthonous (alloc) resources (C). Points are mean estimated ϕT values for each of the 409 consumer observations with corresponding water chemistry measurements. The solid line denotes the mean increase across all consumers at mean levels of all other water chemistry variables, with the shaded polygon denoting 95% CI and dotted lines denoting consumer-specific responses.

  • Fig. 5 Consumer-specific variation in ϕT.

    Means ± 95% CIs plotted for the effect of allochthonous (alloc) resources on ϕT (A), the change in effect of allochthonous resources on ϕT with increasing lake water chlorophyll a (B), and ϕT at mean water chemistry levels across sites (C). zoopl, zooplankton.

Supplementary Materials

  • Supplementary material for this article is available at http://advances.sciencemag.org/cgi/content/full/3/3/e1601765/DC1

    method S1. Additional details for geospatial analyses.

    method S2. Additional details for statistical analysis.

    method S3. Validation and sensitivity of the Bayesian mixing model.

    fig. S1. End members used in mixing model and corresponding with each of the 559 consumer observations.

    fig. S2. Sensitivity of Bayesian mixing model to changes in 7 SDs.

    fig. S3. Sensitivity of Bayesian mixing model to misinformed dietary priors.

    fig. S4. Model recovers known parameters despite not accounting for data sets with consumer use of MOB.

    fig. S5. Predicted isotope ratios versus observed isotope ratios for 559 consumer observations.

    fig. S6. Prior (light gray curves) and posterior (dark gray curves) of ϕT for each of the 559 observations organized by consumer type.

    fig. S7. Lake area distributions globally (black lines) and within our data set (blue lines).

    fig. S8. DOC distributions from 7514 worldwide lakes.

    fig. S9. Chlorophyll a distribution from 80,012 worldwide lakes.

    fig. S10. Model recovers known parameters across 100 simulated data sets that span the range of ϕT (that is, 0 to 1).

    fig. S11. Catchment area estimated for 147 lakes in our isotope data set.

    fig. S12. Proportion of each catchment covered with one of four woody vegetation types.

    fig. S13. Vegetation, geomorphology, and soil characteristics.

    fig. S14. Catchment area for 46 lakes.

    fig. S15. Percent overlap in catchments of each of the 46 lakes delineated with three different approaches.

    fig. S16. Model recovers known parameters despite random noise around the mean effects of covariates predicting the availability of allochthonous resources ξkl.

    fig. S17. Alternate ways of modeling t-OM deposition.

    table S1. Mean and 95% CIs for model parameter estimates associated with eqs. S1 to S11.

    table S2. Key symbols and abbreviations used in the text and the Supplementary Materials and Methods.

    table S3. Reclassification of 2005 North America Land Cover.

    table S4. Reclassification of 2006 European Land Cover.

    table S5. Consumer-specific dietary parameters.

    data file S1. Site-level summary of water quality and catchment characteristics for 147 lakes.

    data file S2. R code for stable isotope mixing model.

    References (6594)

  • Supplementary Materials

    This PDF file includes:

    • method S1. Additional details for geospatial analyses.
    • method S2. Additional details for statistical analysis.
    • method S3. Validation and sensitivity of the Bayesian mixing model.
    • fig. S1. End members used in mixing model and corresponding with each of the 559 consumer observations.
    • fig. S2. Sensitivity of Bayesian mixing model to changes in 7 SDs.
    • fig. S3. Sensitivity of Bayesian mixing model to misinformed dietary priors.
    • fig. S4. Model recovers known parameters despite not accounting for data sets with consumer use of MOB.
    • fig. S5. Predicted isotope ratios versus observed isotope ratios for 559 consumer observations.
    • fig. S6. Prior (light gray curves) and posterior (dark gray curves) of ϕT for each of the 559 observations organized by consumer type.
    • fig. S7. Lake area distributions globally (black lines) and within our data set (blue lines).
    • fig. S8. DOC distributions from 7514 worldwide lakes.
    • fig. S9. Chlorophyll a distribution from 80,012 worldwide lakes.
    • fig. S10. Model recovers known parameters across 100 simulated data sets that span the range of ϕT (that is, 0 to 1).
    • fig. S11. Catchment area estimated for 147 lakes in our isotope data set.
    • fig. S12. Proportion of each catchment covered with one of four woody vegetation types.
    • fig. S13. Vegetation, geomorphology, and soil characteristics.
    • fig. S14. Catchment area for 46 lakes.
    • fig. S15. Percent overlap in catchments of each of the 46 lakes delineated with three different approaches.
    • fig. S16. Model recovers known parameters despite random noise around the mean effects of covariates predicting the availability of allochthonous resources ξkl.
    • fig. S17. Alternate ways of modeling t-OM deposition.
    • table S1. Mean and 95% CIs for model parameter estimates associated with eqs. S1 to S11.
    • table S2. Key symbols and abbreviations used in the text and the Supplementary Materials and Methods.
    • table S3. Reclassification of 2005 North America Land Cover.
    • table S4. Reclassification of 2006 European Land Cover.
    • table S5. Consumer-specific dietary parameters.
    • References (65–94)

    Download PDF

    Other Supplementary Material for this manuscript includes the following:

    • data file S1 (.csv). Site-level summary of water quality and catchment characteristics for 147 lakes.
    • data file S2 (.txt). R code for stable isotope mixing model.

    Files in this Data Supplement:

Navigate This Article