top of page
Landscape photograph of Boston showing city buildings, greenspace, and Charles River


Ecosystem services are the goods and services that nature provides to people; considering the value of these important goods and services when managing landscapes can help us conserve them. Hydrologic ecosystem services are those services that terrestrial systems provide by altering the flow and quality of water. These services can include increased dry-season flows, reduced flooding, or improved water quality for purposes such as drinking. Hydrologic ecosystem services (or "watershed services") have driven large investments, with more than $24 billion spent in 2015 to conserve watersheds for such services. However, the science linking watershed managment and hydrologic ecosystem services has often lacked rigor and careful consideration of nature's role in these processes. The WRElab uses watershed simulation to critically  investigate how ecosystems provide value to people around water.

Informing ecosystem service investments 

"Water Funds" are land-cover-management approaches that restore upstream ecosystems to improve downstream water resources. Typically, the people downstream benefit from increased hydrologic ecosystem services and provide payments and incentives to upstream communities who steward and improve the ecosystems providing those services. The water fund approach has grown rapidly across the world, and investors adopted it to the  tune of $650 million in 2015. 


In many cases, water fund investments are being made with the desire for increased annual water yield and dry-season flow downstream. We set out to investigate where and when such investments were likely to provide these desired services in consultation with the Natural Capital Project. We applied water fund-style management changes in simulations of 29 watersheds stratified by climate zones across six continents. Our approach was a unique combination of management-scale hydrologic simulations and global breadth. We found that that water-fund-style land-cover-management approaches are unlikely to improve water availability, particularly annual water yield, in most sites. This work provides crucial information for practitioners making hydrologic ecosystem service interventions, and serves as a clarion call for critical science on ecosystem service programs given the reputational risks if promised benefits are not achieved. 

Research Results -- Changes in low from from different scenarios of land-cover management in sites across the globe
Research Results -- Changes in low from from different scenarios of land-cover management in sites across the globe
Research Results -- Changes in low from from different scenarios of land-cover management in sites across the globe

Watershed simulation in 29 sites across the globe show that, in most sites, restoration approaches  (top pie wedge) decreased low flow slightly. Investments in restoring natural vegetation are unlikely to provide desired increases in dry-season flow in most locations. 

We used the same watershed simulations to show that soil characteristics are the primary drivers of dry-season streamflow response to natural land-cover restoration. We developed a graph-connected approach to determine which watersheds responded most similarly to or differently from each other in response to such land-cover change. We then applied statistical tests to determine which watershed characteristics differed most across the clusters. We found that only soil depth and soil water content had statistically significant differences across clusters, suggesting that soil characteristics drive the distinct responses of dry-season flow to land-cover change. The clusters did not have statistically significant differences in their watersheds' climates, locations, sizes or shapes, elevations, or pre-existing land-cover. We further found that natural land-cover restoration tended to increase dry-season flow most beneficially in particularly "flashy" watersheds in which streamflow occurs mostly in short periods of peak flow with relatively little low flow. ​

Research results -- Clusters of similar responses to land-cover restoration in sites across the globe

A graph-based clustering approach showed which watersheds respond most similarly to or differently from each other to natural land-cover restoration. 

Papers & presenations

Dennedy-Frank, P.J. and S. M. Gorelick (2020), Insights on Expected Streamflow Response to Land-Cover Restoration. Journal of Hydrology, 589,

Dennedy-Frank, P. J. and S. M. Gorelick (2019), Insights from watershed simulations around the world: Watershed service-based restoration does not significantly enhance streamflow. Global Environmental Change, 58,

Dennedy-Frank, P.J. and S. M. Gorelick (2018), Spatial Distribution of Hydrologic Ecosystem Service Estimates: Comparing Two Models, Abstract GC13D-0673 presented at 2018 Fall Meeting, AGU, Washington, DC, 10-14 Dec.

Dennedy-Frank, P.J. (2018), Effects of land-cover change on streamflow: Analysis of watershed simulations from around the world. National Center for Environmental Analysis and Synthesis, Santa Barbara, CA, 6 June.

Planning the Camboriú Water Fund

Camboriú is a beach destination city in Brazil's Atlantic Forest biome, which is known for high biodiversity and endemism. The Camború metro area swells from 200,000 to 800,000 people during the high season. That influx of people strains the city's water supply; this is exacerbated by the lack of storage in the watershed and the use of water in the treatment process for high sediment loads. The water company for Camboriú decided to reduce sediment loading by protecting and restoring natural forests in areas with the potential to produce high sediment loads with the hope it would be an economical option to help secure the future water supply. 

Research results -- Workflow for evaluating Camboriú Water Fund return-on-investment

Properly assessing the value of ecosystem services requires a complex process that includes biophysical modeling, spatial targeting and budgeting, and economic analysis as shown in the process used to assess the return-on-investment for the Camboriú Water Fund.

Working with the Nature Conservancy, we helped assess the return on investment of this hydrologic ecosystem service approach to securing a sustainable water supply. This process involved 5 steps: 1) modeling land-use/land-cover change to project future changes based on historical land-cover trends determined from high-resolution satellite imagery; 2) modeling watershed hydrology and sediment export both with and without forest conservation and restoration in the areas determined from that land-use/land-cover modeling; 3) targeting interventions to maximally reduce sediment loads within the available budget based on the watershed sediment export and land-use/land-cover modeling; 4) estimating the full costs of the program, including administrative overhead and maintenance; and 5) determining the benefits of the of program and thus the return-on-investment (ROI).

We found that the ROI exceeded 1 for the water supply system after 44 years. This long investment return time suggests that moderate co-benefits could make such restoration much more attractive to investors. The business case for watershed restoration in these types of conditions is likely to be sensitive both to the local socio-economic conditions and to institutional arrangements that can reduce overhead costs and ensure a clear flow of investments and benefits.  

Papers & presenations

Kroeger, T., C. Klemz, T. Boucher, J.R.B. Fisher, E. Acosta, A.T. Cavassani, P.J. Dennedy-Frank, L. Garbossa, E. Blainski, R.C. Santos, S. Giberti, P. Petry, D. Shemie, and K.C. Dacol (2019), Returns on investment in watershed conservation: Application of a best practices analytical framework to the Rio Camboriú Water Producer program, Santa Catarina, Brazil. Science of the Total Environment, 657, 1368-1381,

Fisher, J.R.B., E. Acosta, P.J. Dennedy-Frank, T. Kroeger, & T. M. Boucher (2017), Impact of satellite imagery spatial resolution on land use classification accuracy and modeled water quality. Remote Sensing in Ecology and Conservation,

Kroeger T., C. Klemz, D. Shemie, T. Boucher, J.R.B. Fisher, E. Acosta, P.J. Dennedy-Frank, A. Targa Cavassani, L. Garbossa, E. Blainski, R. Comparim Santos, P. Petry, S. Giberti and K. Dacol (2017). Assessing the Return on Investment in Watershed Conservation: Best Practices Approach and Case Study for the Rio Camboriú PWS Program, Santa Catarina, Brazil. The Nature Conservancy, Arlington, VA.

Kroeger, T., C. Klemz, P. Petry, D. Shemie, E. Blainski, T. Boucher, J. Fisher, J. Guimaraes, P.J. Dennedy-Frank and K. Dacol (2014), Estimating the ‘return on investment’ in natural infrastructure: Rio Camboriù Watershed, Santa Catarina State, Brazil. ACES Conference, Washington, DC, 08 – 12 Dec.

Testing models for ecosystem service analysis

Numerous groups have developed models to estimate the production of ecosystem services and the value those services provide to people. However, these models sometimes trade off appropriate rigor for ease of use by non-experts. This tradeoff is not necessarily bad but it needs to be acknowledged, particularly in a field like hydrology where a long history of water management and research could be integrated when considering the production of hydrologic ecosystem services.


We have used existing models and datasets to test newer models' estimates of hydrologic ecosystem services. In one case, we compared results from the Natural Capital Project's InVEST Water Yield model and the Soil and Water Assessment Tool (SWAT), a widely-used hydrologic model. We showed that InVEST and SWAT provide similar estimates of water yield in some locations but not in others. In related work, we helped benchmark InVEST's new Sediment Delivery Ratio model. We also contributed careful thinking to analyses that sought to quantify the potential range of flow regulation services by using reservoirs of different sizes as a clear set of comparisons. 

Research results -- Model comparisons between SWAT and InVEST

We compared estimates of annual water yield from InVEST, an ecosystem service-targeted model and SWAT, a widely used hydrologic model. We showed that while the two may provide very similar estimates of annual water yield distribution in some locations like the agriculture-dominated Wildcat Creek Watershed they provide quite different estimates in a more forest-dominated watershed like the Upper Upatoi Creek Watershed. 

Papers & presenations

Dennedy-Frank, P.J., R.L. Muenich, I. Chaubey and G. Ziv (2016), Comparing two tools for ecosystem service assessments regarding water resources decisions, Journal of Environmental Management, 177, 331-340,

Dennedy-Frank, P.J., Y. Ghile, S. M. Gorelick, R. A. Logsdon, I. Chaubey and G. Ziv (2014), Spatial Distribution of Hydrologic Ecosystem Service Estimates: Comparing Two Models, Abstract GC13D-0673 presented at 2014 Fall Meeting, AGU, San Francisco, CA, 15-19 Dec.

Guswa, A.J., P. Hamel, P.J. Dennedy-Frank (2017), Potential effects of landscape change on water supplies in the presence of reservoir storage. Water Resources Research, 53, 2679-2692,

Hamel, P., K. Falinski, R. Sharp, D.A. Auerbach, M. Sanchez-Canales, and P.J. Dennedy-Frank (2016), Sediment delivery modeling in practice: Comparing the effects of watershed characteristics and data resolution across hydroclimatic regions. Science of the Total Environment, 580, 1381-1388,

bottom of page