East Branch Penobscot River Salmon Habitat Rehabilitation Through Evaluation of Flow Regimes

Maine’s Penobscot River contains some of the last remaining spawning habitat in the United States for the endangered Atlantic salmon. The WPES team is collaborating on a project led by the Penobscot Nation and sponsored by Maine Sea Grant to evaluate flow regimes affecting Atlantic salmon habitat in the East Branch of the Penobscot River, with a major focus on examining scenarios related to watershed conditions, climate, civil infrastructure, and human generated disturbances to help guide river management decision-making. WPES researchers are using historical information and hydrologic and spatial data to parameterize and calibrate a hydrologic model of the East Branch watershed to advance our understanding of the many factors affecting watershed and river conditions.

Project Context

The Penobscot River region, including the East Branch watershed, has been the home of the Penobscot people for at least 11,000 years following deglaciation of Maine at the end of the last ice age. The Penobscot practiced subsistence hunting and fishing and used the river for transportation when moving seasonally between headwater forests and the Gulf of Maine coast.  Atlantic salmon made up as much as a third of the Penobscot diet prior to the arrival of European settlers. To learn more about the Penobscot Nation perspective on the importance of Atlantic salmon restoration and see how the WPES team’s evaluation of flow regimes in the East Branch contributes, see the Maine Public Film Series short documentary, .

The overarching salmon restoration project incorporates elements of ecohydrology (the study of interactions between hydrology and ecosystem functioning) and the land use history of the watershed. Below, we’ve compiled some of what other researchers can tell us about these topics to help readers contextualize our evaluation of East Branch flow regimes.

Simple map of the Gulf of Maine and state of Maine. The Penobscot River watershed, stretching across the middle of the state, is highlighted. Within the Penobscot River watershed area, the smaller East Branch Penobscot River sub-watershed at its northern end is also highlighted.

Atlantic Salmon Ecohydrology

While Atlantic salmon were once abundant in Maine’s rivers, they are now listed as an endangered species, due in part to habitat degradation and loss of access to spawning habitat in river headwaters [Zydlewski et al., 2023].  

Biologists observe that salmon thrive in very cold river and stream temperatures, with clean, flowing water and coarse gravel bedding in which to deposit their eggs, and that they are sensitive to changing river conditions [Louhi, 2008; Harrison, 2019].  The river flows that lead to suitable habitat conditions are controlled by a combination of natural watershed processes and influenced by civil infrastructure such as dams within the river network. Ecohydrology is useful in determining the quantitative connections between watershed runoff and aquatic organisms; ecohydraulics can ascertain where river flows will impact lifeforms and water quality and how engineering solutions can manage the passage of water effectively [Kattel, 2024]. Dam operations within a watershed affect the magnitude and timing of freshwater flows through rivers: during spring freshets, dams can store water from high magnitude runoff events, which can decrease downstream flooding; during late summer and fall dry periods, dams are used to augment baseflows by releasing stored lake water [Boulange et al., 2021]. Other human activities in the landscape can also affect river conditions: clearing of trees can lead to shorebed deforestation, which causes soil erosion and channel narrowing. Changes in river geomorphology also affect runoff patterns and water temperature [Sweeney et al., 2004].

The WPES team is working to provide the types of hydrologic and hydraulic information and scenarios for the East Branch that fish biologists can use to assess how flow regimes controlled by precipitation and dam operations influence suitability of potential salmon spawning habitat downstream of Matagamon Dam.

Historical Changes to Penobscot River Watershed

Logging across the state that began with the harvesting of white pine for ships’ masts by European colonists in the 17th century intensified into industrial-scale clear cutting operations by the 19th century and brought significant changes to the East Branch watershed. 

Loggers intensively modified the natural channels to efficiently float the millions of board-feet per year of timber through the river network to the major lumber port of Bangor. Sections of stream were dredged or straightened to prevent logs from snagging on rocks and other obstructions. Temporary wooden dams were constructed on some streams to hold back flow until water levels were high then blown with dynamite to send logs downstream on the flood wave, causing additional channel erosion and widening. More permanent dams were constructed to create large reservoirs and regulate river discharge [Anderson, 1979].  Collectively, these activities degraded or destroyed natural salmon spawning habitat, impeded or blocked salmon access to the Penobscot headwaters, and changed the timing and magnitude of freshwater flows through the East Branch. Log drives on the Penobscot River continued until the early 1970s before being fully replaced by truck transport along logging roads. 

Perhaps the most significant alteration to the East Branch watershed was the construction of the Telos Cut canal in 1841.  A natural drainage divide separating the East Branch and Allagash River watersheds existed to the northwest of Mt. Katahdin between Webster Pond and Telos Lake.  Runoff entering Telos and Chamberlain Lakes naturally flowed north from Chamberlain into Eagle and Churchill Lakes and ultimately via the Allagash and Saint John Rivers to the coastal city of Saint John, New Brunswick. Landowners who wanted to float logs south to Bangor rather than to Canada had a ~1 mile long canal dug between Webster Pond and Telos Lake to allow the latter to drain eastward into the East Branch.  A series of dams were constructed at the Telos Cut, Chamberlain Lake, Eagle Lake, and Churchill Lake to act as a lock system to allow logs from as far north as Churchill Lake to be floated upstream to Telos Lake and through the canal on their journey to Bangor [Anderson, 1979]. 

Today, three actively-managed dams remain in the East Branch watershed. Lock Dam, on the north side of Chamberlain Lake at the site of one of the original log transport locks, prevents Chamberlain Lake waters from draining naturally to the north into Eagle Lake, forming a human-imposed drainage divide between the East Branch and Allagash River watersheds and diverting runoff from a ~250 square mile area into the Penobscot River. Telos Dam controls water levels in the hydrologically-connected Telos and Chamberlain Lakes and regulates flows into the natural East Branch watershed through Telos Cut. Farther downstream, Matagamon Dam controls water levels in Grand Lake Matagamon and regulates flows into the East Branch Penobscot mainstem.

Anderson, Hayden L.. “Penobscot Waterways Canals and Waterway Improvements on the Penobscot River, 1816-1921.” Maine History 19, 1 (1979): 21-46.

Boulange, J., Hanasaki, N., Yamazaki, D., and Pokhrel, Y., 2021, Role of dams in reducing global flood exposure under climate change: Nature Communications, v. 12, p. 417, doi:.

Harrison, L.R., Bray, E., Overstreet, B., Legleiter, C.J., Brown, R.A., Merz, J.E., Bond, R.M., Nicol, C.L., and Dunne, T., 2019, Physical Controls on Salmon Redd Site Selection in Restored Reaches of a Regulated, Gravel-Bed River: Water Resources Research, v. 55, p. 8942–8966, doi:.

Kattel, G., 2024, Integrating ecohydrology and ecohydraulics: Ecohydrology, v. 17, doi:.

Louhi, P., Mäki-Petäys, A., and Erkinaro, J., 2008, Spawning habitat of Atlantic salmon and brown trout: general criteria and intragravel factors: River Research and Applications, v. 24, p. 330–339, doi:.

Sweeney, B.W., Bott, T.L., Jackson, J.K., Kaplan, L.A., Newbold, J.D., Standley, L.J., Hession, W.C., and Horwitz, R.J., 2004, Riparian deforestation, stream narrowing, and loss of stream ecosystem services: Proceedings of the National Academy of Sciences, v. 101, p. 14132–14137, doi:.

Zydlewski, J. et al., 2023, Seven dam challenges for migratory fish: insights from the Penobscot River: Frontiers in Ecology and Evolution, v. 11, doi:.

Hydrologic Modeling

To test scenarios and make predictions about river flow regimes under different conditions affected by both natural processes (e.g., precipitation patterns) and human interventions (e.g., dam operations), WPES has created a hydrologic model of the East Branch Penobscot River watershed using the US Army Corps of Engineers Hydrologic Engineering Center Hydrologic Modeling System (). This type of model is designed to represent hydrologic and hydraulic processes at the scale of watershed subbasins and river reaches over time scales that can range from single storm events to years or decades.  In general, the model makes hydrologic predictions by generating runoff from precipitation and routing it into and through the river network and to the river outlet. The following subsections describe the hydrologic processes the model must account for, the spatial and civil infrastructure data used to develop the basin model describing our watershed, and the climatic and meteorological time series that act upon the basin model to generate predictive output.

Terrain map of East Branch Penobscot River watershed above USGS gage at Grindstone, broken up in to sub-watersheds with major lakes and river network segments. Dams and USGS gages are labeled.
Map view of the hydrologic model schematic for the East Branch Penobscot River, including subbasins, river reaches, and lakes.

Hydrologic & Hydraulic Processes

Briefly, these are major processes that the model must account for:

  • Runoff generation – conversion of precipitation into surface water runoff.  This is a function of precipitation intensity and depth, losses to interception by vegetation, and losses to infiltration into the soil.
  • Snowpack dynamics – seasonal accumulation of precipitation as snow and ice that can persist on the ground surface throughout the winter months and melts in spring freshet runoffs.
  • Overland flow routing – efficiency and timing of precipitation or snowmelt runoff from the landscape into stream networks, affected by surface slope, flow path lengths, and surface roughness from vegetation or microtopography.
  • Channelized stream flow – timing and hydraulics of flow through stream and river channels, controlled by channel dimensions, bed roughness, water surface slope.
  • Baseflow – efficiency and timing of groundwater flow into stream networks, controlled by soil saturation and hydraulic conductivity in the soil layers.
  • Evapotranspiration – removal of available water from vegetation, land surface, and soil. Transpiration is the vegetation extracting available water from the soil through plant roots. These are functions of vegetation type, temperature, wind, and atmospheric conditions.
  • Storage – temporary detention of runoff in puddles, ponds, and lakes.

Spatial Data

The basin model representing the physical setting within the hydrologic model is constructed using existing spatial data:

  • Elevation data – used for calculations of surface slope, flow routing, overland flow lengths, subbasin delineation, and microtopographic surface storage. The East Branch model uses USGS 1/3 arc second elevation rasters reprojected to 10 meter cells.
  • Stream networks – channel networks derived from USGS National Hydrography Dataset data convey flow from upstream subbasins.
  • Land cover – used for calculations of runoff generation, evapotranspiration, canopy interception, and overland flow routing. The East Branch model uses USGS National Land Cover Database (NLCD) data.
  • Soils – soil characteristics control rates of infiltration (inversely related to runoff generation) and base flow, transpiration, soil percolation losses to deep groundwater, and total available soil storage. The East Branch model uses soil characteristics derived from NRCS SSURGO soils data.
  • Water bodies / bathymetry – lakes and large ponds act as storage reservoirs that slowly release water. Water body surface dimensions from USGS National Hydrography Dataset and storage volumes calculated using bathymetry interpolated from lake soundings.

In total, 61 parameterized subbasins and 15 reservoirs are included in our ~2,800 km2 basin model.

Civil Infrastructure

Three major dams controlling outflow from large reservoirs are currently represented in the hydrologic model:

  • Lock Dam, an earthen dam with two small sluice gates, prevents Chamberlain Lake from naturally draining north into the Allagash River system.  A small amount of outflow is allowed through the gates to provide baseflow for the Allagash; this water leaves the model.
  • Telos Dam, a wood crib dam last replaced in the 1980s, controls outflow from the hydrologically connected Telos and Chamberlain Lakes in the northwest of the East Branch watershed. Discharge through this dam flows east through Telos Cut to Webster Brook and into Grand Lake Matagamon.  Four sluice gates are currently operated; the structure also has two inactive radial gates and an emergency spillway.
  • Matagamon Dam, a concrete dam built in 1941 to replace an 1880s wood structure, controls outflow from Grand Lake Matagamon into the East Branch mainstem.  Four main sluice gates are currently operated; the structure also has an inactive log sluice gate and spillway with removable flashboards.

All three dams were modeled from dam schematics with individually controllable sluice gates, allowing real dam operations to be included in the model as hourly time series.

  • View of Matagamon Dam, an aging concrete structure spanning the river with a series of gates to allow controlled release of water.
    Matagamon Dam, looking upstream from East Branch mainstem
  • Side-on view of Telos Dam, a wooden structure spanning the river. From this angle the gates cannot be seen.
    Telos Dam, with Telos Cut downstream to left
  • Side-on view from atop Lock Dam, a long earthen berm that crests just a few feet above the water level of Chamberlain Lake beside it. In the foreground are controls for the underground sluice gates and a hatch to access the gates.
    Lock Dam, with Chamberlain Lake to left

Weather & Water Cycle Time Series

In our HEC-HMS model, meteorological and climatic inputs are applied to the East Branch watershed basin model over 2000 meter grids that compute an average for each subbasin and then compute hydrologic conditions. For this model, we use gridded time series data produced by the European Commission’s Copernicus program. Copernicus’ data are reanalysis products that interpolate weather station observations and remotely sensed data into a regular grid of values on an hourly time step. Meteorological and climatic data inputs acting on the model include:

  • Precipitation – The source of fresh water for the model simulation. When air temperature in the model is sufficiently cold, precipitation type transitions from rain to snow. 
  • Temperature – Temperature affects evapotranspiration by determining the potential rate of water loss and snowmelt by acting as an energy index for melt rate. 
  • Radiation – Short wave solar radiation is radiant energy produced by the sun and is normally associated with daylight hours. Longwave thermal radiation is emitted from all living and non-living bodies. Both types of radiation are used in model calculations for evapotranspiration and snowmelt processes. 
  • Air Pressure – Barometric air pressure is the force exerted against the surface of the earth by the weight of the air above it. Air pressure is used in model calculations for evapotranspiration and snowmelt. 
  • Windspeed – Wind velocities near the ground surface are used in model calculations for evapotranspiration and snowmelt.
  • Dew Point – The temperature to which air must be cooled to become fully saturated so that water vapor condenses to form dew. Dew point is used in calculations of evapotranspiration and snowmelt. 
  • Initial Snowpack – Snowpack provides storage of precipitation during cold periods and is a source of additional surface runoff during spring freshets and other melt events. It is measured in terms of snow water equivalent (SWE, the volume of liquid water held as frozen snow in the snowpack). The model uses Copernicus ERA5 hourly gridded snow data for SWE for the first time step of a model run, then makes its own snowpack calculations for subsequent time steps.
  • Initial Albedo – Albedo is a measure of how much incoming light or solar radiation energy reflects off ground surfaces and back into space. Albedo is used in model calculations for snowmelt. The model uses Copernicus ERA5 hourly gridded reflective surface data for albedo for the first time step, then makes its own albedo calculations for subsequent time steps.
  • Reservoir Monthly Evaporation – The rate at which water evaporates from the surface of lake or pond reservoirs included in the model. The model uses long-term monthly averages of Copernicus ERA5 gridded evaporation data for the East Branch watershed to calculate hourly evaporation from reservoirs. 

Calibration and Validation

No model can perfectly capture the complications of real hydrologic systems.  In a model, hydrologic processes are simplified, landscapes are generalized into lumped subbasins or gridded cells, and the spatial data used to build the model may also not perfectly represent the real system, particularly for data like soils that are interpolated from observations at a limited set of sampling locations.  Calibration is a process of modifying initial model parameter estimates within physically realistic ranges to obtain simulated results that match known conditions. All models require calibration and validation before they can be trusted for scenario testing. 

Our watershed hydrologic model is calibrated using river discharge data from multiple USGS river gages, including , a recently installed gage , and , a tributary of the East Branch.  Calibration of the snowmelt parameters was performed using historical data from the .  The calibration process on an initialized model is generally performed in three repeating steps: 1) running the simulation using real meteorologic time series data, 2) comparing simulation results with the observed gage discharge data for the same time period, and 3) adjusting basin model parameters (e.g., maximum soil infiltration rate) to attempt to reduce differences between modeled and observed discharge values, then returning to step 1.  This process continues until the simulation results closely match real-world discharge data.  Calibration generally begins by focusing on short time series such as single storm events, then building up to fine-tuning across longer time series.

Model validation is the process of determining if a calibrated model accurately represents the physical process and provides predictive capability beyond the calibration period.  Our model was calibrated using gage data for large storm events in 2017 and 2020, then validated against a full 2015-2024 hourly time series that included relatively wet, relatively dry, and “normal” precipitation and observed discharge conditions.  Validation identified types of conditions where model performance is strongest and where its predictions may have larger error bars, and also helped reveal occasional discrepancies between the ERA5 reanalysis precipitation time series and observations at nearby NOAA weather stations that could explain departures between modeled and observed stream discharge for those periods.

Example of observed and (preliminary) modeled hydrograph data for Seboeis River gage watershed, Dec. 2015 – Dec. 2024

Scenario Testing

The primary purpose of our model development is evaluation of flow regimes in the East Branch Penobscot River.  The validated model can be used to explore multiple scenarios related to flow regimes, including forecasting for different precipitation inputs, investigating the effects of dam operations on modern flow regimes, and hindcasting to estimate pre-dam flow regimes.  We are also using a hydraulic model using the Army Corps’ HEC-RAS (River Analysis System) platform to look more closely at in-channel flow velocities and other hydraulic conditions in the several miles of the East Branch Penobscot River mainstem directly downstream of Matagamon Dam.

Check back for more about our scenario testing as the project continues!