Including the subsurface in ecosystem services

Ecosystem-service assessments often fail to account for groundwater’s role in the ecosystem. Whether groundwater is important for these services depends strongly on the assessment scale and the local context.

Groundwater supports crucial ecosystems around the world, including riparian zones, meadows, lakes, and wetlands1, and represents ~40% of the world’s freshwater resources2,3. However, groundwater’s influence on ecosystem services, the benefits that ecosystems provide to people, has often been ignored or oversimplified in both global assessments4 and local planning studies5. In this issue, Qiu and colleagues6 simulate the potential ecosystem services produced in an agricultural landscape (Fig. 1) both when including groundwater in the model and when excluding it.

Fig. 1: Groundwater’s presence is seen in the local lakes, wetlands, and vegetation patterns of many agricultural regions that produce valuable goods and services.

Chris Hepburn / Getty

Qiu and colleagues investigate how estimates of ecosystem-service production change when groundwater is excluded from models.

The subsurface’s role in the ecosystem is challenging to study for several reasons: nonlinear physical and biogeochemical processes dominate behaviour; heterogeneous soils and rock formations cause further variability; and high-resolution images or samples of these belowground processes and properties are difficult to obtain. What do we miss when we make our analysis more tractable by excluding these subsurface processes?

Qiu and colleagues6 investigate how eight biophysical indicators change with and without groundwater in the model: recharge and overland flow; sediment, nitrogen, and phosphorus fluxes; crop and grass yield; and carbon storage. Importantly, these biophysical indicators signal quantities produced by the ecosystem, not the services (that is, the flow of benefits) provided to people. The authors model an agricultural watershed where previous research and sampling produced a detailed understanding of subsurface behaviour7. The lessons learned here will be applicable in similar watersheds around the world. They demonstrate the effect of omitting groundwater by adopting free drainage, which treats the water table as being so deep that it does not affect the lower boundary of the model, as the boundary condition for the counterfactual ‘without groundwater’ case. Further tests investigate the effects of continuous dry and wet years as well as different land covers and soils.

Their results suggest that whether groundwater is important for producing ecosystem services depends on the scale. At the watershed scale, removing groundwater from the model changes the medians of all but one biophysical indicator by <10%. Subcontinental- to global-scale ecosystem service assessments rarely estimate biophysical indicators with 10% accuracy anyway. Such assessments may choose to ignore groundwater to avoid the increased complexity of models that include it, though the results may then be less useful for specific management decisions. Scientists might focus instead on better estimating climate, especially likely precipitation shifts, since biophysical indicators under wet and dry climates differ by a factor of two or three in Qiu and colleagues’ models. At the pixel scale (4.84 ha), by contrast, the authors find that the biophysical indicators change by more than 100% when groundwater is excluded for all but one service. Local ecosystem service assessments may need to realistically account for groundwater to target interventions for the desired return and to avoid interventions producing minimal service. Models that include groundwater may require more detailed data and analyses than are often available for such local planning. New understanding will be required to develop inexpensive and feasible assessment approaches that can scale ecosystem-service investment8.

Why do these biophysical indicators respond differently at different scales to removing groundwater? There seem to be two main causes, each affecting a subset of metrics. First, for crop and grass yield and carbon storage, groundwater’s impacts offset each other at the larger, watershed scale, creating a small average change. Groundwater provides plants in drier areas with a water subsidy9, but if the water table is too shallow it deprives plant roots of oxygen, reducing yields. Second, for recharge, overland flow, and phosphorus and sediment fluxes, groundwater has a large effect mostly when depth-to-water is <1 m. Outside this small fraction of the watershed, groundwater has little effect on these biophysical indicators, so the watershed-scale response is small. Of course, if groundwater was very shallow in a larger portion of the watershed, removing that groundwater might have a larger effect.

Qiu and colleagues modelled only the biophysical indicators of ecosystem-service production, not the flow of benefits provided to people. These flows may vary nonlinearly for the same ecosystem-service production depending on context. As an example, shallow groundwater reduces recharge — but recharge is less important for crucial drinking water and irrigation services where shallow groundwater itself provides those services. The authors did not directly engage stakeholders to quantitatively evaluate the magnitude or spatiotemporal variability of ecosystem-service demand, so their results may not link strongly to the actual flows of benefits. This step is essential to affect investments or policies. In addition, they did not investigate the covariation of the biophysical indicators of ecosystem-service supply, an important step for assembling portfolios of services likely to drive investment. The study, however, estimates these biophysical indicators particularly well, using state-of-the-science models that include explicit groundwater flow and dynamically responsive plant growth and nutrient interactions.

The work by Qiu et al. demonstrates groundwater’s scale- and context-dependent effects on several biophysical indicators of ecosystem services. It provides crucial understanding of how our modelling choices affect estimates of ecosystem-service production and starts to build physical understanding that may someday usefully simplify ecosystem-service models and clarify critical monitoring needs. It provides an example of a detailed, careful study of ecosystem-service production in one location, and serves as a call for further studies in different contexts and of ecosystem-service demand to link with this detailed understanding of its supply.


  1. 1.

    Orellana, F., Verma, P., Loheide, S. P. II & Daly, E. Rev. Geophys. 50, RG3003 (2012).

    Article  Google Scholar 

  2. 2.

    Gleeson, T., Wada, Y., Bierkens, M. F. P. & van Beek, L. P. H. Nature 488, 197–200 (2012).

    CAS  Article  Google Scholar 

  3. 3.

    Siebert, S. et al. Hydrol. Earth Syst. Sci. 14, 1863–1880 (2010).

    Article  Google Scholar 

  4. 4.

    Kim, H. et al. Geosci. Model Dev. 11, 4537–4562 (2018).

    Article  Google Scholar 

  5. 5.

    Vogl, A. L. et al. J. Environ. Manag. 195, 78–91 (2016).

    Article  Google Scholar 

  6. 6.

    Qiu, J. et al. Nat. Sustain. (2019).

  7. 7.

    Carpenter, S. R. et al. BioScience 57, 323–335 (2007).

    Article  Google Scholar 

  8. 8.

    Daily, G. C. et al. Front. Ecol. Environ. 7, 21–28 (2009).

    Article  Google Scholar 

  9. 9.

    Hwang, T., Band, L. E., Vose, J. M. & Tague, C. Water Resour. Res. 48, W06514 (2012).

    Article  Google Scholar 

Download references

Author information



Corresponding author

Correspondence to P. James Dennedy-Frank.

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Dennedy-Frank, P.J. Including the subsurface in ecosystem services. Nat Sustain 2, 443–444 (2019).

Download citation


Nature Briefing

Sign up for the Nature Briefing newsletter — what matters in science, free to your inbox daily.

Get the most important science stories of the day, free in your inbox. Sign up for Nature Briefing