Human Activities Shape Global Decomposition Rates In Rivers (2024) - Science PDF

RES EARCH ◥ world. Global scale predictions also contribute RESEARCH ARTICLE t...

RES EARCH ◥ world. Global scale predictions also contribute RESEARCH ARTICLE to a finer scale understanding of decomposition and support efforts to model planetary carbon STREAM ECOLOGY dynamics. Models that can accurately predict current in situ decomposition rates across space Human activities shape global patterns are particularly valuable, enabling manipu- lation of environmental drivers in silico to of decomposition rates in rivers predict impacts under scenarios of future global environmental change. S. D. Tiegs1*†, K. A. Capps2,3*†, D. M. Costello4*†, J. P. Schmidt2*, C. J. Patrick5*, We present a predictive model fitted with J. J. Follstad Shah6, C. J. LeRoy7, the CELLDEX Consortium‡ global data from the Cellulose Decomposition Experiment (CELLDEX), a coordinated, distrib- Rivers and streams contribute to global carbon cycling by decomposing immense quantities of terrestrial uted experiment on cellulose decomposition plant matter. However, decomposition rates are highly variable and large-scale patterns and drivers of in rivers designed to reveal previously un- this process remain poorly understood. Using a cellulose-based assay to reflect the primary constituent documented patterns in decomposition rates of plant detritus, we generated a predictive model (81% variance explained) for cellulose decomposition and the key factors driving this fundamental rates across 514 globally distributed streams. A large number of variables were important for predicting ecosystem-level process. Decomposition of Downloaded from https://www.science.org at Coastal Carolina University on August 27, 2024 decomposition, highlighting the complexity of this process at the global scale. Predicted cellulose cellulose—the most abundant organic polymer decomposition rates, when combined with genus-level litter quality attributes, explain published leaf on the planet and a main constituent of plant litter decomposition rates with high accuracy (70% variance explained). Our global map provides litter—was quantified by more than 150 inves- estimates of rates across vast understudied areas of Earth and reveals rapid decomposition across tigators using a common and well-established continental-scale areas dominated by human activities. cellulose decomposition assay (24). The “cot- ton strip assay” is a standardized approach E for measuring decomposition by using a read- arth’s terrestrial ecosystems produce over regions with contrasting climates (9, 10), conduct- ily available woven cotton fabric (artist’s canvas), 100 billion tons of plant detritus annually ing literature reviews of local field studies (11), comprised of 95% cellulose. The loss of tensile (1, 2) and the fates of this organic matter— developing conceptual models (12, 13), and strength of the fabric is measured, a process for example, long term storage, miner- performing meta analyses (14, 15). Coordi- that is strongly correlated with the microbial alization to greenhouse gasses, or incor- nated, distributed experiments (16–20) have catabolism of cellulose (25). We performed the poration into stream food webs—depend on been particularly insightful as they generate assay in 514 flowing water ecosystems at geo- the rate at which it is decomposed. River eco- directly comparable data across broad geo- referenced field sites on all seven continents, systems are carbon-processing hotspots (3, 4), graphic areas and identify coarse resolution spanning 135° of latitude and each of Earth’s receiving 0.72 billion tons of terrestrial carbon explanatory variables of decomposition rates major terrestrial biomes (19, 20). We used high per year (2), an amount that is disproportion- in rivers, including differences in decomposer resolution (15 arcsecond) climate, soil, geology, ately important relative to the small fraction of communities and biomes. Still, we lack a com- vegetation, and physicochemical data (101 ex- nonglaciated land area (0.58%) that rivers oc- prehensive understanding of how drivers such planatory variables total) in a boosted regres- cupy (5). Rivers connect terrestrial ecosystems as climate, geology, vegetation, water quality, sion tree algorithm to develop the first global, with aquatic storage compartments includ- and soils interact to govern organic matter high-resolution predictive model of organic ing floodplains, lakes, and oceans, playing vital decomposition at large scales. Such knowl- matter decomposition in rivers. We then tested roles in the global carbon cycle and function- edge gaps are particularly evident across the the utility of the cellulose model by using pre- ing both as organic matter conduits and reactors. tropics and in lower income economies— dicted cellulose decomposition rates and genus- Despite the widely recognized importance of ecologically important areas where rivers are level leaf litter chemistry traits to explain 895 flowing water in global carbon cycling (6–8), grossly understudied relative to those in north- leaf litter decomposition estimates from studies our understanding of variation in organic ern temperate zones. Quantifying patterns and conducted at 559 locations across the globe. We matter decomposition rates and their driv- controls of decomposition in these areas is found that cellulose decomposition rates are an ers at large spatial scales is still limited (2). critical, however, as much of Earth’s terres- excellent proxy for litter decomposition rates. Large-scale spatial variation in organic mat- trial plant matter is annually produced in Further, our models indicate the physicoche- ter decomposition in rivers and streams has tropical forests (net primary production 16.0 mical factors at river and watershed scales in- been estimated by comparing leaf litter de- to 23.1 billion tons of carbon) (21, 22), and teract with characteristics of the organic matter composition rates from studies conducted in tropical rivers deliver 48 to 64% of the carbon being decomposed (e.g., leaf litter chemistry) to moving from rivers to the ocean (23). create heterogeneous spatial patterns in river- 1 Department of Biological Sciences, Oakland University, Effectively modeling carbon dynamics at the ine decomposition across the planet. Rochester, MI 48309, USA. 2Odum School of Ecology, University of Georgia, Athens, GA 30602, USA. 3Savannah global scale—including areas where field data River Ecology Laboratory, University of Georgia, Aiken, SC are scarce—requires a more mechanistic and Climate, geology, soils, and water quality 29802, USA. 4Department of Biological Sciences, Kent State process-based understanding of the many en- explain cellulose decomposition rates University, Kent, OH 44242, USA. 5Virginia Institute of Marine Science, Coastal Ocean Processes Section, William & vironmental and biotic factors that drive or- Climate, geology, soil, and water quality vari- Mary, Gloucester Point, VA 23062, USA. 6School of the ganic matter decomposition. Accurate estimates ables explain 81% of variance in field measure- Environment, Society, and Sustainability, University of Utah, generated by combining existing empirical ments of cellulose decomposition. Because Salt Lake City, UT 84112, USA. 7Environmental Studies Program, The Evergreen State College, Olympia, WA 98505, USA. measurements with fine-scale geospatial and a standardized cellulose substrate was used at *Corresponding author. Email: [email protected] (S.D.T); environmental data can provide multiple ben- all field sites, observed variation in decom- [email protected] (K.A.C.); [email protected] (D.M.C.); efits. They can reduce the need for data collec- position rates can be attributed unequivocally [email protected] (J.P.S.); [email protected] (C.J.P.) tion from remote or difficult to access regions, to the activity of microbial communities and †These authors contributed equally to this work. ‡CELLDEX Consortium authors and affiliations are listed in the subsequently generating baseline estimates environmental drivers. Prior efforts have ex- supplementary materials. for decomposition in understudied areas of the plained broad variation in decomposition rates Tiegs et al., Science 384, 1191–1195 (2024) 14 June 2024 1 of 5 RES EARCH | R E S E A R C H A R T I C L E across riverine ecosystems as a function of ature (IV = 2.5; Fig. 1F). Our data and ap- tent (IV = 2.0), population count (IV = 1.3), and exogenous factors such as temperature (14, 19) proach also highlight watershed-level charac- river regulation (IV = 1.3) still emerge as im- and concentrations of dissolved nutrients teristics that have been given little attention portant (table S1). Notably, relationships be- (17, 20, 26), as well as litter traits (15, 27, 28). previously, such as sub-watershed lake area tween explanatory variables and decomposition Our model supports those findings and shows (limnicity) (IV = 6.9; Fig. 1B), actual evapo- rates are frequently nonlinear, revealing thresh- that climatic and water quality parameters are transpiration in the watershed (IV = 4.4; Fig. 1E), olds beyond which there are abrupt changes in among the most important explanatory vari- and the chemical and physical properties of decomposition rates (e.g., Fig. 1, B, D, and E). ables of decomposition rates (Fig. 1). However, a soil (table S1). Subwatershed lake area was a Water temperature has a strong positive effect relatively large number of explanatory vari- high ranking variable and its negative rela- on cellulose decomposition (Fig. 1A) and there ables (n = 26) have importance values greater tionship with decomposition rates may be is an optimal range (5 to 13°C) of annual air than 1.0 (table S1), and no single variable con- explained by the disproportionately greater temperature with estimated lower rates in both tributes >15% to the explanatory power of the abundance of lakes at high northern latitudes cooler and warmer watersheds (Fig. 1F). model (table S1). This result reveals the com- where water temperatures are low (Fig. 1B). plexity of the many drivers that influence or- Alternatively, lower nutrient concentrations and Extrapolating to global patterns of ganic matter decomposition at the global scale. suppressed hydrological variability may have decomposition rates Top explanatory variables of cellulose de- also contributed to the negative influence of Our model and map of riverine cellulose de- composition include expected attributes like limnicity on decomposition. Although our study composition reveals pronounced, large-scale Downloaded from https://www.science.org at Coastal Carolina University on August 27, 2024 mean daily water temperature [importance value sites were selected to have minimal human im- spatial patterns of organic matter processing (IV) = 14.0; Fig. 1A], nitrogen and phospho- pacts relative to their region of study (19), variables (Fig. 2). Rates generally increase with decreasing rus availability (IV = 6.7 and 4.9, respectively; associated with anthropogenic development latitude, with rapid rates in tropical regions (e.g., Fig. 1, C and D), and mean annual air temper- such as dissolved nutrient yields, cropland ex- Central America, Amazon basin, Western Africa, 0.025 A 0.025 B 0.020 0.020 0.015 0.015 0.010 0.010 0 10 20 30 1 10 100 Mean daily stream temp during deployment (°C) Subwatershed lake area+1 (%) C D Cellulose decomposition rate (Kd d –1) 0.025 0.025 0.020 0.020 0.015 0.015 0.010 0.010 0.01 0.1 1 10 0.001 0.01 0.1 Stream NO3+NO2 yield (kg ha−1 yr−1) Stream DRP yield (kg ha−1 yr−1) 0.025 E 0.025 F 0.020 0.020 0.015 0.015 0.010 0.010 0 50 100 150 −10 0 10 20 30 AET during month of deployment (mm) Mean annual air temperature (°C) Fig. 1. Partial dependence plots (black lines) of the top variables that explain and predict cellulose decomposition rates (Kd). (A to F) Background maps show global distributions of explanatory variables in a Mollweide projection. The boosted regression tree model explains 81% of the variance in decomposition rates across the 514 streams used in our study. Most top variables relate to climate and water quality and effects exhibit nonlinear threshold responses. Black ticks above the x-axis indicate decile breaks. Tiegs et al., Science 384, 1191–1195 (2024) 14 June 2024 2 of 5 RES EARCH | R E S E A R C H A R T I C L E Indo Pacific) and areas characterized by volcanic as cotton fabric) lacks the chemical complex- substrate and natural litter (Fig. 3A and activity and young soils, an effect previously ity of organic matter that naturally enters table S2). These results provide strong support documented only at more local scales (29). running waters, we also tested how accurately for the critical influence that environmental Notably, fluvial ecosystems in these regions are our modeling approach could explain varia- drivers have in regulating riverine litter de- among the least studied on the planet (Fig. 2, tion in the decomposition rates of terrestrial composition, including those affected by an- inset) despite having high rates of terrestrial leaf litter in rivers reported by ecologists thropogenic activities. primary production (22) and carbon export to worldwide. To this end we independently Prior research at large scales has stressed the ocean (23). Vast areas in middle latitudes validated model forecasts using 895 litter the importance of litter quality as the predomi- with ubiquitous human impacts—central Europe, decomposition rates from 559 locations and nant control of decomposition rates in rivers eastern China, central North America, south- representing 35 genera of terrestrial plants (27). (15). Our results demonstrate that in addition eastern South America, and Japan—also sup- We also used leaf and litter trait data at the to leaf litter traits, environmental factors such port elevated decomposition rates, strongly genus level (30, 31) and experimental condi- as temperature and nutrient availability are suggesting continental-scale human impacts tions (14, 27) as explanatory variables to critically important in regulating decomposition on carbon cycling in rivers. By contrast, areas account for variation among decomposition rates at larger spatial scales. Our validation of boreal forests—characterized by short grow- estimates resulting from differences in leaf model also reveals that invertebrate access to ing seasons, low temperatures, and peaty, acidic, litter quality (e.g., lignin, hemicellulose, tannin, leaves, as assessed by experimentally manipu- water logged soils—exhibit slower rates of nutrient content) and the feeding activity of lating litter bag mesh size, greatly increases Downloaded from https://www.science.org at Coastal Carolina University on August 27, 2024 organic matter decomposition, especially in invertebrates (Fig. 3A and table S2). Our cellu- the rate of decomposition in all but the fastest northern Asia, eastern Scandinavia, and north- lose decomposition model predictions coupled decomposing leaves (Fig. 3A). Finally, litter eastern Canada. with litter traits account for 70% of the varia- chemistry contributes to the explanatory power tion in leaf litter decomposition. Notably, the of the model in expected ways, with plant Validating predicted cellulose decomposition explanatory power of this model is over- genera characterized by high lignin content rates with leaf litter decomposition rates whelmingly driven by predicted rates of cellu- (IV = 11.9; Fig. 3B) and low litter nitrogen Recognizing that the substrate used in our lose decomposition (IV = 39.5), despite the stark content (C:N, IV = 5.45 and N, IV = 5.23; Fig. 3, standardized decomposition assay (cellulose differences in quality between the cellulose C and D), exhibiting slower decomposition. Cellulose Kd (d –1) 0.08 0.05 0.03 0.02 0.01 0.005 NA Fig. 2. Predicted mean annual cellulose decomposition rates (Kd) revealing broad spatial patterns in decomposition rates. We did not predict Kd for sub watersheds with ≤10 ha of sub basin area, nor for Antarctica, for which we did not have values for most predictor variables. Inset shows study sites for cellulose (light circles) and leaf litter (dark circles) decomposition measurements. Map and insert are Mollweide projection. Tiegs et al., Science 384, 1191–1195 (2024) 14 June 2024 3 of 5 RES EARCH | R E S E A R C H A R T I C L E 0.025 A decomposition may be particularly susceptible Detritivore+Microbe or resistant to global change, thereby inform- 0.020 Microbe ing freshwater conservation efforts. As proof 0.015 of concept, we examined potential changes in predicted litter decomposition rates associ- 0.010 ated with changes in pine oak forest com- 0.005 position in Mexican watersheds invaded by pine bark beetle (Dendroctonus mexicanus) 0.002 0.005 0.01 0.02 0.05 0.1 Cellulose decomp rate (Kd d –1) (35). This invasion is expected to be partic- ularly severe in the watershed of the Rio 0.020 B Grande de Santiago, a major conduit of or- ganic matter to the Pacific Ocean in Mexico 0.015 (Fig. 4). Our forecasts predict that insect- induced canopy replacement from pine to oak Litter decomposition rate (Kd d –1) 0.010 would cause decomposition rates to increase and become more variable (2.5- to 3.8-fold increase), with larger increases in decom- 10 20 30 Downloaded from https://www.science.org at Coastal Carolina University on August 27, 2024 position associated with watersheds with Litter lignin (% dm) greater evapotranspiration and drier soils 0.020 C (fig. S1). To promote the use of our models for forecasting we created an easy-to-use, open-source online application where users 0.015 can estimate both cotton strip and leaf litter decomposition rates for any river across the 0.010 globe (https://shiny-bsci.kent.edu/CELLDEX/). 20 40 60 80 100 120 140 Conclusions and implications Litter C:N (molar) By pairing a distributed field experiment with publicly available environmental data, we 0.020 D created the first high-resolution map and predictions of organic matter decomposition 0.015 rates in flowing waters worldwide. Our model demonstrates that cellulose decomposition 0.010 results from diverse, interacting, and non- linear environmental forcings that can best be described with complex, data-rich models. Al- 0.5 1.0 1.5 2.0 2.5 Litter N (% dm) though the standard cotton fabric used lacks the biochemical complexity of leaf litter, our Fig. 3. Partial dependence plots of the top variables that explain leaf litter decomposition rates (Kd). relatively simple organic matter substrate is The boosted regression tree model explains 70% of the variance in rates across 895 published values of an excellent proxy for leaf litter in decom- leaf litter decomposition (27). Top explanatory variables included our modeled cellulose decomposition rates, position studies, as demonstrated by our invertebrate access to the leaf material, and attributes related to litter quality at the genus level. Smooth fits model predictions. Simplification of the leaf (GAM) show the relationship between cellulose decomposition rate and litter decomposition for the litter bag assay allowed us to both achieve two different common litter bag mesh sizes that allow or exclude invertebrates (A). The smooth fits capture standardized results and fill extensive geo- the general environmental effects on decomposition, whereas the partial dependency plots (thin lines) are graphic gaps in remote and low resourced noisier due to covariation in leaf quality and environmental conditions (i.e., certain leaf types are used in areas, demonstrating the power of coordi- certain regions). Black ticks above x axis indicate decile breaks. Note the change in y axis between nated, distributed experiments (36). Although (A), (B), (C), and (D). our datasets were large when compared with other studies of organic matter decomposition, the field data used were relatively limited in Other litter traits (e.g., P content, cellulose) available data we show that cellulose decom- both space and time, which makes our strong provide little additional explanatory power and position can be an excellent proxy for litter explanatory power all the more valuable. Thus, these leaf traits explain no more variation than decomposition, and our composite model of this work also underscores the power of ma- expected by chance (table S2). It is well- environmental drivers makes reliable estimates chine learning algorithms and large geographic recognized that leaf litter chemistry can vary of litter decomposition at a global scale. databases of environmental data (e.g., Hydro- among individuals within a species (32, 33) BASINS) (37, 38) plus the critical value of and even individual leaves from a single tree Forecasting decomposition under global temporally and geographically extensive data (34); thus our model may underestimate the environmental change from simple but standardized coordinated ex- importance of individual-level variation in leaf The high explanatory power of our cellulose periments (e.g., CELLDEX). and litter chemistry in driving decomposition. and leaf litter decomposition models enables Given the pressing need for measuring eco- Greater measurement and reporting of litter forecasting of decomposition rates under al- system functions for biomonitoring and bio- chemistry, especially nitrogen and lignin con- tered climate, land cover, soil conditions, and assessment (39, 40), our globally distributed tent, will improve understanding of endogenous nutrient loading scenarios. These predictions experiment provides a template for matching controls at global scales. Despite limitations in can identify locations across the globe where observational data with model predictions. This Tiegs et al., Science 384, 1191–1195 (2024) 14 June 2024 4 of 5 RES EARCH | R E S E A R C H A R T I C L E 22. C. B. Field, M. J. Behrenfeld, J. T. Randerson, P. Falkowski, Science 281, 237–240 (1998). 23. T. H. Huang, Y. H. Fu, P. Y. Pan, C. T. A. Chen, Curr. Opin. Environ. Sustain. 4, 162–169 (2012). 24. S. D. Tiegs, J. E. Clapcott, N. A. Griffiths, A. J. Boulton, Ecol. Indic. 32, 131–139 (2013). 25. J. Mancuso, J. L. Tank, U. H. Mahl, A. Vincent, S. D. Tiegs, Aquat. Sci. 85, 83 (2023). 26. M. Ardón et al., Biol. Rev. Camb. Philos. Soc. 96, 692–715 (2020). 27. C. J. LeRoy et al., J. Ecol. 108, 17–35 (2020). 28. K. Yue et al., Biol. Rev. Camb. Philos. Soc. 97, 2023–2038 (2022). 29. A. D. Rosemond, C. M. Pringle, A. Ramírez, M. J. Paul, J. L. Meyer, Limnol. Oceanogr. 47, 278–289 (2002). 30. L. Boyero et al., Sci. Rep. 7, 10562 (2017). 31. J. Kattge et al., Glob. Change Biol. 17, 2905–2935 (2011). Pinus 32. C. J. LeRoy, T. G. Whitham, S. C. Wooley, J. C. Marks, J. N. Am. Quercus Relative frequency Benthol. Soc. 26, 426–438 (2007). 33. A. Lecerf, E. Chauvet, Basic Appl. Ecol. 9, 598–605 (2008). 34. T. Sariyildiz, J. M. Anderson, Biol. Fertil. Soils 37, 137–146 (2003). Downloaded from https://www.science.org at Coastal Carolina University on August 27, 2024 35. A. González-Hernández, R. Morales-Villafaña, M. E. Romero-Sánchez, B. Islas-Trejo, R. Pérez-Miranda, J. For. Res. 31, 649–659 (2018). 36. L. H. Fraser et al., Front. Ecol. Environ. 11, 147–155 (2013). 37. B. Lehner, K. Verdin, A. Jarvis, Eos 89, 93–94 (2008). 38. B. Lehner, G. Grill, Hydrol. Processes 27, 2171–2186 0.000 0.005 0.010 0.015 (2011). 1 39. M. O. Gessner, E. Chauvet, Ecol. Appl. 12, 498–510 (2002). Decomposition rate (d ) 40. M. C. Jackson et al., Adv. Ecol. Res. 55, 615–636 (2016). 41. K. A. Wilson et al., PLOS Biol. 14, e1002413 (2016). Fig. 4. Distribution of temperate coniferous forests in Mexico (all points) and locations (orange) 42. A. D. Rosemond et al., Science 347, 1142–1145 (2015). 43. D. Costello et al., Zenodo (2024); https://zenodo.org/records/ where there is a moderate to high risk of pine bark beetle (Dendroctonus mexicanus) invasion 11035638 [adapted from (35)] that drives a shift from coniferous to deciduous forest. Inset shows the density 44. J. Cheng, B. Schloerke, B. Karambelkar, Y. Xie, leaflet: Create Interactive Web Maps with the JavaScript “Leaflet” Library, distribution of predicted litter decomposition rates for streams in areas of moderate to high invasion risk both version 2.2.1, Comprehensive R Archive Network (2023); for pine litter (green solid line) and oak litter (orange dashed line). Our model predicts that full canopy https://cran.r-project.org/web/packages/leaflet/index.html. replacement from pine to oak would increase leaf litter decomposition rates 2.5 to 3.8 fold with a greater increase predicted in watersheds with greater evapotranspiration and drier soils. Base from US Geological AC KNOWLED GME NTS We are grateful for the efforts of the many people who assisted Survey, The National Map, 2023; Web Mercator projection; created in the R package leaflet 2.2.1 (44). with the CELLDEX project in the lab and in the field: We thank J. Mancuso for edits on an earlier version of this manuscript, D. Ethaiya for logistical assistance during the CELLDEX project, and J. Talbott for assistance with the Shiny application. Any use of approach provided baseline data for estimated shorter-term carbon storage (42) and re- trade, firm, or product names is for descriptive purposes only and does not imply endorsement by the US Government. Funding: This decomposition rates across immense, unstudied ductions in carbon transfer to longer-term work was sponsored by an Ecuadorian National Science Foundation areas of the planet and supports the develop- storage compartments, such as reservoirs, PROMETEO award (to S.D.T.). Original data compilation of leaf ment of biomonitoring networks in areas where floodplains, and oceans. litter decomposition rates from primary literature was supported by a working group grant to J.J.F.S. from the US National Science they are most needed (41). To further advance Foundation (Division of Environmental Biology #1545288 and large scale monitoring and assessment, we have RE FERENCES AND NOTES #1929393) through the US Long Term Ecological Research made these modeling approaches accessible Network. A portion of the salary of K.A.C. was supported by the 1. J. Cebrian, Am. Nat. 154, 449–468 (1999). Department of Energy Office of Environmental Management under through an open-source online mapping tool. 2. T. J. Battin et al., Nature 613, 449–459 (2023). Award Number DE-EM0005228 to the University of Georgia Application of the models to current and fu- 3. M. E. McClain et al., Ecosystems 6, 301–312 (2003). Research Foundation. Author contributions: Conceptualization: C.J.P., 4. E. R. Hotchkiss et al., Nat. Geosci. 8, 696–699 (2015). ture environmental threats will enable scien- D.M.C., J.P.S., K.A.C., and S.D.T. Methodology: C.J.P., D.M.C., J.P.S., 5. G. H. Allen, T. M. Pavelsky, Science 361, 585–588 K.A.C., and S.D.T. Investigation: All coauthors. C.L., and J.J.F.S. also tists and natural resource managers to forecast (2018). provided data from a literature review. Visualization: D.M.C. and J.P.S. changes in the functioning of river networks at 6. P. Regnier et al., Nat. Geosci. 6, 597–607 (2013). Funding acquisition: S.D.T. Writing - original draft: K.A.C. and S.D.T. 7. A. Marx et al., Rev. Geophys. 55, 560–585 (2017). a planetary scale. 8. P. A. Raymond et al., Nature 503, 355–359 (2013). Writing - review and editing: B.G.M., C.J.P., D.M.C., F.J.B., J.P.S., K.A.C., M.O.G., N.A.G., S.D.T., and G.W. Competing interests/Disclaimer: Cellulose decomposition is strongly influ- 9. V. Ferreira, A. C. Encalada, M. A. S. Graça, Freshw. Sci. 31, 945–962 (2012). Authors declare that they have no competing interests. Data and enced by multiple interacting environmental 10. A. Bruder, M. H. Schindler, M. S. Moretti, M. O. Gessner, materials availability: All data and code for analyses and figures are drivers that continue to be affected by anthro- Freshw. Biol. 59, 438–449 (2014). available on GitHub (43). License information: Copyright © 2024 pogenic activities. Undoubtedly, climate change, 11. J. C. Marks, Annu. Rev. Ecol. Evol. Syst. 50, 547–568 the authors, some rights reserved; exclusive licensee American (2019). Association for the Advancement of Science. No claim to original US increased nutrient loading, intensified land 12. M. A. S. Graça, Int. Rev. Hydrobiol. 86, 383–393 (2001). government works. https://www.science.org/content/page/science- use modification, and changes in vegetation 13. T. V. Royer, G. W. Minshall, J. N. Am. Benthol. Soc. 22, licenses-journal-article-reuse cover will continue to alter organic matter 352–358 (2003). 14. J. J. Follstad Shah et al., Glob. Change Biol. 23, 3064–3075 processing in rivers and streams. Notably, SUPPLEMENTARY MATERIALS (2017). key human influenced drivers of cellulose 15. M. Zhang et al., Glob. Ecol. Biogeogr. 28, 1469–1486 science.org/doi/10.1126/science.adn1262 decomposition—especially nutrient loading and (2019). CELLDEX Consortium Authors and Affiliations 16. L. Boyero et al., Ecol. Lett. 14, 289–294 (2011). Materials and Methods temperature—are positively related to decompo- 17. G. Woodward et al., Science 336, 1438–1440 (2012). Fig. S1 sition rates. A critical implication is that in the 18. I. T. Handa et al., Nature 509, 218–221 (2014). Tables S1 and S2 presence of continued environmental change, 19. S. D. Tiegs et al., Sci. Adv. 5, eaav0486 (2019). References (45–55) 20. D. M. Costello et al., Global Biogeochemical Cycles 36, organic matter decomposition rates will likely e2021GB007163 (2022). Submitted 4 December 2023; accepted 9 May 2024 increase in rivers, resulting in declines in 21. C. M. Gough, Nature Education Knowledge 3, 28 (2011). 10.1126/science.adn1262 Tiegs et al., Science 384, 1191–1195 (2024) 14 June 2024 5 of 5