Technical ReportPDF Available

Abstract and Figures

BRAT was developed to provide more realistic expectation management for beaver assisted restoration and serves as a planning tool intended to help resource managers, restoration practitioners, wildlife biologists and researchers better manage expectations about where beaver might be useful. Specifically, BRAT is a spatially explicit network tool that predicts where along streams and rivers beaver may be useful as a restoration tool and where they may be a nuisance, in which case their impacts can be mitigated or the nuisance beaver can serve as a source population for live-trapping and relocation to areas where they can help achieve restoration and conservation objectives.
Content may be subject to copyright.
IDAHO
BEAVER RESTORATION
ASSESSMENT TOOL
Building Realistic Expectations for Partnering with Beaver
in Conservation and Restoration
Prepared by:
Wally Macfarlane, Maggie Hallerud, Sara Bangen, Chalese Hafen, Micael Albonico, Chad Garlick, Tyler Gibby, Tyler
Hatch, Cashe Rasmussen and Joe Wheaton
Department of Watershed Sciences, 5210 Old Main Hill, Logan, UT 84322-5210
Prepared for:
Idaho Fish and Game, 600 S. Walnut St., Boise, ID 83707
June 2019
Idaho BRAT
2 of 55
Recommended Citation:
Macfarlane WW, Hallerud, MA, Bangen S, Hafen C, Albonico MT, Garlick C, Gibby T, Hatch T, Rasmussen C, and Wheaton JM.
2019. Idaho Beaver Restoration Assessment Tool: Building Realistic Expectations for Partnering with Beaver in Conservation and
Restoration. Prepared Idaho Fish and Game. Logan, UT. 55 pages.
This work is licensed under a Creative Commons Attribution 4.0 International License.
Idaho BRAT
3 of 55
CONTENTS
List of figures .................................................................................................................................................................. 4
Executive Summary ....................................................................................................................................................... 6
Introduction .................................................................................................................................................................... 8
Methods ......................................................................................................................................................................... 9
BRAT - Beaver dam capacity model .......................................................................................................................... 9
BRAT dam building beaver conservation and restoration model ............................................................................. 14
Risk assessment using proximity to human infrastructure................................................................................... 14
Dam building beaver management layers ........................................................................................................... 15
Model calibration ................................................................................................................................................. 16
Model validation ............................................................................................................................................. 16
Results & interpretations .............................................................................................................................................. 18
Model outputs .......................................................................................................................................................... 18
Beaver dam capacity model ................................................................................................................................ 18
Management layers ............................................................................................................................................. 24
Model intermediates ................................................................................................................................................ 27
Model calibration...................................................................................................................................................... 34
Model validation ....................................................................................................................................................... 34
How do dam densities track between predicted and actual?............................................................................... 34
Do the electivity indices increase appreciably from the none to the pervasive class?......................................... 44
Deliverables ................................................................................................................................................................. 47
GIS data layers .................................................................................................................................................... 47
Discussion & recommendations ................................................................................................................................... 48
Caveats ................................................................................................................................................................... 48
Future versions of BRAT ......................................................................................................................................... 49
Riverscapes network models to inform management .............................................................................................. 50
Beaver management recommendations .................................................................................................................. 50
Beaver Dam Analogues ...................................................................................................................................... 50
‘Living with beaver’ strategies ............................................................................................................................. 50
Conclusions ................................................................................................................................................................. 52
Acknowledgments ........................................................................................................................................................ 53
References ................................................................................................................................................................... 54
Idaho BRAT
4 of 55
LIST OF FIGURES
Figure 1: Reach scale illustration of derivation of streamside vs. riparian vegetation scores from 30 vs. 100 m stream
network buffers. A shows the 30 m and 100 m buffers, which we used to summarize intersecting pixels from 30 m
resolution classified LANDFIRE raster in B. Dam building suitability are shown in B and range from 0 (unsuitable; grey)
to 4 (optimal; blue) with red for 1, yellow for 2, and green for 3. C & D contrast the buffer averaged values for the 30 m
buffer (C) and the 100 m buffer (D). ............................................................................................................................. 12
Figure 2: Methodological illustration of inputs (1-5) and output for the combined capacity model of riverscapes capacity
to support beaver dam-building activity. Model output is expressed as dam density (dams/km). ................................ 13
Figure 3: Combined capacity Fuzzy Inference System for capacity of riverscapes to support dam building beaver
activity. This shows the specification of fuzzy membership functions with overlapping values for categorical descriptors
in inputs and the output. ............................................................................................................................................... 14
Figure 4: Modeled beaver dam capacity for existing conditions for the Southwest Region of Idaho............................ 19
Figure 5: Modeled beaver dam capacity for historic conditions for the Southwest Region of Idaho. ........................... 20
Figure 6. Existing vs. historic dam capacity by category for Southwest Region of Idaho. ............................................ 21
Figure 7: Modeled existing dam complex size for the Southwest Region of Idaho. ..................................................... 22
Figure 8: Modeled historic dam complex size for the Southwest Region of Idaho. ...................................................... 23
Figure 9: Modeled conservation and restoration opportunities for the Southwest Region of Idaho. ............................ 24
Figure 10: Modeled areas beavers can build dams but could be undesirable for the Southwest Region of Idaho. ..... 25
Figure 11: Modeled unsuitable/limited dam building opportunities for the Southwest Region of Idaho........................ 26
Figure 12: Little Wood watershed context layers used in the Idaho BRAT model. ....................................................... 27
Figure 13: Little Wood watershed Euclidean distance to roads layer used in the Idaho BRAT model. ........................ 28
Figure 14: Little Wood watershed Euclidean distance to roads within the valley bottom layer used in the Idaho BRAT
model. .......................................................................................................................................................................... 29
Figure 15: Little Wood watershed Euclidean distance to road crossings layer used in the Idaho BRAT model........... 30
Figure 16: Little Wood watershed Euclidean distance to canals layer used in the Idaho BRAT model........................ 31
Figure 17: Little Wood watershed Euclidean distance to human infrastructure layer used in the Idaho BRAT model. 32
Figure 18: Little Wood watershed land use intensity layer used in the Idaho BRAT model. ........................................ 33
Figure 19: Actual beaver dam (based on 2013 0.5 meter NAIP imagery) identified throughout Idaho from desktop dam
censusing. Of the 9,048 dams 8,060 were within 60 m of the NHD stream lines and therefore, were used in the model
validation. ..................................................................................................................................................................... 35
Figure 20: Predicted vs. observed dam counts (per reach) for Idaho. The blue line is line of perfect agreement (1:1
relationship), dashed line is the 50th percentile (median) regression, dotted line is the 75th percentile regression, and
dot-dashed line is the 90th percentile regression (not visible due to direct correspondence with line of perfect agreement
in this graphic). The nearer the higher percentile regression is to the 1:1 ratio, the better the model is performing. ... 36
Figure 21: Predicted vs. observed dam counts (per reach) for Clearwater Region. The blue line is line of perfect
agreement (1:1 relationship), dashed line is the 50th percentile (median) regression, dotted line is the 75th percentile
regression, and dot-dashed line is the 90th percentile regression. The nearer the higher percentile regression is to the
1:1 ratio, the better the model is performing. ............................................................................................................... 37
Figure 22: Predicted vs. observed dam counts (per reach) for the Magic Valley Region. The blue line is line of perfect
agreement (1:1 relationship), dashed line is the 50th percentile (median) regression, dotted line is the 75th percentile
regression, and dot-dashed line is the 90th percentile regression. The nearer the higher percentile regression is to the
1:1 ratio, the better the model is performing. ............................................................................................................... 38
Figure 23: Predicted vs. observed dam counts (per reach) for the Panhandle Region. The blue line is line of perfect
agreement (1:1 relationship), dashed line is the 50th percentile (median) regression, dotted line is the 75th percentile
regression, and dot-dashed line is the 90th percentile regression. The nearer the higher percentile regression is to the
1:1 ratio, the better the model is performing. ............................................................................................................... 39
Idaho BRAT
5 of 55
Figure 24: Predicted vs. observed dam counts (per reach) for the Salmon Region. The blue line is line of perfect
agreement (1:1 relationship), dashed line is the 50th percentile (median) regression, dotted line is the 75th percentile
regression, and dot-dashed line is the 90th percentile regression. The nearer the higher percentile regression is to the
1:1 ratio, the better the model is performing. ............................................................................................................... 40
Figure 25: Predicted vs. observed dam counts (per reach) for the Southeast Region. The blue line is line of perfect
agreement (1:1 relationship), dashed line is the 50th percentile (median) regression, dotted line is the 75th percentile
regression, and dot-dashed line is the 90th percentile regression. The nearer the higher percentile regression is to the
1:1 ratio, the better the model is performing. ............................................................................................................... 41
Figure 26: Predicted vs. observed dam counts (per reach) for the Southwest Region. The blue line is line of perfect
agreement (1:1 relationship), dashed line is the 50th percentile (median) regression, dotted line is the 75th percentile
regression, and dot-dashed line is the 90th percentile regression. The nearer the higher percentile regression is to the
1:1 ratio, the better the model is performing. ............................................................................................................... 42
Figure 27: Predicted vs. observed dam counts (per reach) for the Upper Snake Region. The blue line is line of perfect
agreement (1:1 relationship), dashed line is the 50th percentile (median) regression, dotted line is the 75th percentile
regression, and dot-dashed line is the 90th percentile regression. The nearer the higher percentile regression is to the
1:1 ratio, the better the model is performing. ............................................................................................................... 43
Figure 28: Schematic of a beaver deterrent used to control pond height. .................................................................... 51
Figure 29: Pond leveler installation. From left: securing flexible pipe in cage to protect inflow from being clogged; placing
pipe into beaver pond; rebuilding beaver dam after setting pipe into notched dam at desired water height. ............... 51
Idaho BRAT
6 of 55
EXECUTIVE SUMMARY
This report presents an application of the Beaver Restoration Assessment Tool 3.0.20 (BRAT;
http://brat.riverscapes.xyz/) a tool for building realistic expectations for partnering with beaver in conservation and
restoration (Macfarlane et al., 2017). In this application, we analyzed all the perennial rivers and streams within the
state of Idaho.
The backbone to BRAT is a capacity model developed to assess the upper limits of riverscapes to support beaver dam-
building activities. It outputs an estimated density of dams (i.e. dams per length of stream) and a rough count of an
upper limit (i.e. capacity) of how many dams the conditions in and surrounding a reach could support. Both existing
and historic capacity were estimated using readily available spatial datasets to evaluate seven lines of evidence: (1) a
reliable water source; (2) stream bank vegetation conducive to foraging and dam building; (3) vegetation within 100 m
of edge of stream to support expansion of dam complexes and maintain large beaver colonies; (4) likelihood that dams
could be built across the channel during low flows; (5) the likelihood that a beaver dam on a river or stream is capable
of withstanding typical floods; (6) evidence of suitable stream gradient; and (7) evidence that river is too large to allow
dams to be built and to persist. Fuzzy inference systems were used to combine these lines of evidence while accounting
for categorical ambiguity and uncertainty in the continuous inputs driving the models. The existing model estimate of
capacity was driven with LANDFIRE 30 m resolution vegetation data from 2016, whereas the ‘historic’ estimate
represents a pre-European settlement model of vegetation, also from LANDFIRE.
The estimated existing Idaho statewide capacity is 994,299 dams or roughly 8 dams/km. By contrast, the same model
driven with estimates of historic vegetation types estimated the statewide capacity at 1,743,459 dams or roughly 15
dams/km reflecting a 43% loss compared to historic capacity. Comparison of the Idaho Fish and Game administrative
boundaries of Idaho are illustrated in Table 1.
Nearly all of the capacity loss from historic conditions can be explained in terms of riparian vegetation loss, vegetation
conversion and degradation associated with high intensity land use including: 1) conversion of valley bottoms to urban
and agricultural land uses, 2) overgrazing in riparian and upland areas, 3) conifer encroachment of wet meadow areas.
Despite the losses in beaver dam capacity, Idaho’s waterways are still capable of supporting and sustaining a
substantial amount of beaver dam-building activity (994,299 dams).
Table 1: Estimated existing and historic capacity by the administrative region of Idaho
Idaho BRAT
7 of 55
Identifying these losses in beaver dam capacity incentivizes plans for restoration and conservation opportunities to be
considered. To aid groups in their decisions and what possible risks may arise the BRAT model supplies the following
management outputs: 1) potential risk areas, 2) unsuitable or limited dam building opportunities, and 3) conservation
and restoration opportunities. As such, the BRAT model identifies where streams are relative to human infrastructure
and high intensity land use, and conservatively shows how that aligns with where beaver could build dams.
The existing capacity model was verified statewide, using 8,060 actual dam locations. Results indicate that the model
effectively segregates the factors controlling beaver dam occurrence and density 89.97% of the time. The model
verification results also indicate that beavers preferentially dam in reaches with higher modeled dam capacity while
avoiding those with lower capacity.
The spatially explicit outputs from this application of BRAT provides Idaho Fish and Game (IDF&G) staff with the
information needed to understand patterns of beaver dam capacity, potential risks to human infrastructure, as well as
constraints and opportunities for using dam building beaver in restoration and conservation. This information helps with
both statewide planning efforts and individual watershed scale planning as well as design and on-the-ground
implementation of conservation and restoration activities.
Idaho BRAT
8 of 55
INTRODUCTION
The current scope of stream and riparian degradation in the western U.S. is enormous (Paulsen et al., 2006). Even
with more than $10 billion spent annually, these traditional restoration efforts are barely scratching the surface of what
could be restored. Moreover, a disproportionate amount of funds are spent on too few miles of streams and rivers
leaving millions of miles of degraded streams neglected. To fill this gap, restoration practitioners are increasingly trying
restoration techniques that are more costeffective, less intensive, and can more practically scale up to the enormous
scope of degradation (Silverman et al., 2018).
Watershed conservation and restoration involving beaver is not new. For example, in fall 1948, 76 live nuisance beaver
were parachuted into the Frank Church Wilderness of Idaho with only one fatality (Heter, 1950). Beaver assisted
restoration, without the use of parachutes, is now broadly appreciated as an effective low-tech stream, riparian and
wet meadow restoration technique.
This appreciation of beaver as a restoration agent stems from beavers long recognized status asecosystem engineers’
(Wright et al., 2002) with the capability of restoring streams, rivers and wetlands to the benefit of numerous upland,
riparian and aquatic flora and fauna (Naiman et al., 1986). Beaver dam-building activities lead to a cascade of
hydrologic, geomorphic, and ecological feedbacks that increase stream complexity and benefit aquatic and terrestrial
biota (Rosell et al., 2005). Dam complexes increase system roughness and resilience, increase groundwater recharge
and elevate water tables, create ponds, wetlands and critical habitat for fish, amphibians, small mammals, and
vegetation. Beaver dams expand riparian areas, change timing, delivery and storage or water, sediment and nutrients
(Bird et al., 2011). As a result, beaver are increasingly being used as a key component of stream restoration strategies
(e.g. Curran and Cannatelli, 2014). Using beaver dam building as a restoration agent actually produced a population
level increase in density, survival and production of ESA listed salmon (Bouwes et al., 2016); (Pollock et al., 2014).
Nevertheless, until the development of the Beaver Restoration Assessment Tool (BRAT) (Macfarlane et al., 2017),
predictive spatial models resolving where beaver dams within a drainage network can be built and sustained have
been lacking. In summary, dam building beaver need water and wood. The type and extent of wood/vegetation matters
most to predict dam building capacity, whereas the flow regime acts to potentially limit dam capacity. Existing dam
capacity estimates where and to what extent beaver can build dams now. Or in other words, how many beaver dams
could a particular stream reach support? Dam building capacity is reported as number of dams per mile (or km). Historic
capacity estimates where and to what extent beaver could build dams historically. Also lacking until the development
of BRAT was a model that identifies places where beaver might build dams that could be in direct conflict with land use
and management priorities (e.g., damming of culverts or irrigation canals and flooding of roads or railroads).
Currently there is massive enthusiasm to use beaver as a restoration strategy as well as a growing need to
constructively temper this enthusiasm. As such, BRAT was developed to provide more realistic expectation
management for beaver assisted restoration and serves as a planning tool intended to help resource managers,
restoration practitioners, wildlife biologists and researchers better manage expectations about where beaver might be
useful. Specifically, BRAT is a spatially explicit network tool that predicts where along streams and rivers beaver may
be useful as a restoration tool and where they may be a nuisance, in which case their impacts can be mitigated or the
nuisance beaver can serve as a source population for live-trapping and relocation to areas where they can help achieve
restoration and conservation objectives. The model has been run for the entire state of Utah (Macfarlane et al., 2017)
and is currently used by the Utah Division of Wildlife Resources to help manage their beaver populations, carry out
habitat restoration and implement their beaver management plan. BRAT has also been run for large portions of
Wyoming, Oregon and California along with smaller portions of New York, Washington and the United Kingdom.
BRAT is a tool for imaging what is possible. As (Goldfarb, 2018; page 237) elegantly states, Although BRAT’s primary
value is technical, it is, too, an achievement of the imagination, a method for visualizing the magnificently ponded world
Idaho BRAT
9 of 55
that predated European trappinga time machine to the Castorocene”. The BRAT capacity model outputs ‘paint’ a
vision of what watersheds once were, and what they could be if beaver were allowed to return. BRAT provides a vision
for what’s possible and how partnering with beaver can help increase the resilience of watersheds to drought, fire, and
climate change (Silverman et al., 2018). BRAT can be used for planning and outreach, expectation management, and
conservation and restoration prioritization.
This project focused geographically on the entire state of Idaho and project deliverables will be used as a tool for
managing dam-building beaver. The three main objectives of the project were to:
1. Run BRAT for Idaho;
2. Validate BRAT statewide using field and Google Earth reconnaissance; and
3. Synthesize findings from BRAT into recommendations for beaver management.
This report’s primary purpose is to document the fulfillment of these three objectives and explain how the analyses and
tools presented can assist in the management of dam-building beaver populations across Idaho.
METHODS
BRAT - Beaver dam capacity model
Beaver dams, not beaver themselves, provide the impacts to the geomorphology of the streams that we seek. As such,
the BRAT model estimates beaver dam capacity not beaver habitat. While beaver can survive in wide range of
conditions, where they build dams is more limited. Dam building activity varies dramatically according to flow regime
and availability of dam building materials. Thus, BRAT’s backbone is a capacity model developed to assess the upper
limits of riverscapes to support beaver dam-building. Our estimates of beaver dam capacity come from seven lines of
evidence: (1) a reliable water source; (2) stream bank vegetation conducive to foraging and dam building; (3) vegetation
within 100 m of edge of stream to support expansion of dam complexes and maintain large beaver colonies; (4)
likelihood that dams could be built across the channel during low flows; (5) the likelihood that a beaver dam on a river
or stream is capable of withstanding typical floods; (6) evidence of suitable stream gradient; and (7) evidence that river
is too large to allow dams to be built and to persist.
The four primary questions that the BRAT capacity model asks:
1. Is there enough water present to maintain a pond?
2. Are enough and the right types of woody resources present to support dam building?
3. Can beaver build a dam at base flows?
4. Can dams withstand typical floods?
With the BRAT model, we approximate quantitative answers to these four questions with GIS data. For this application,
we used the following publicly available datasets of national extent (Table 1) that provide direct approximations for
these lines of evidence based on remotely sensed imagery and regionally derived empirical relationships. While we
recognize that higher resolution inputs with greater accuracy and exactness could result in more precise model outputs,
such higher resolution data are not freely available and as such we use freely and broadly available datasets of coarser
resolution with the belief that these datasets will provide useful estimates.
Idaho BRAT
10 of 55
Table 2: Input data used to represent the lines of evidence of Idaho BRAT beaver dam capacity model.
Input Data Criteria Source
Streams and
rivers Perennial water USGS National Hydrography Dataset
http://nhd.usgs.gov/
LANDFIRE 2016
& LANDFIRE
2014 (EVT and
BPS)
Riparian vegetation
LANDFIRE land cover data
http://www.landfire.gov/
USGS baseflow
equations Dam could be built
Wood et al. 2009
Hortness and Berenbrock 2001
Wilkowske et al. 2008
USGS 2-year
peak flow
equations
Dam could withstand floods Wood et al. 2016
Berenbrock 2002
10 m DEM Evidence of stream gradient USDA NRCS Geospatial Data Gateway
http://datagateway.nrcs.usda.gov/
Idaho BRAT
11 of 55
The beaver dam capacity model is described thoroughly in Macfarlane et al. (2017), and detailed online documentation
describing how to run the model is available at http://brat.riverscapes.xyz/. Therefore, in this report, we only briefly
describe the capacity model. Our capacity model estimates the capacity of riverscapes to support dam-building activity
by approximating the maximum number of dams that can be sustained, based on vegetation resources and typical
stream flows. Model outputs are calibrated to a range of dam densities found in nature and reported in the literature,
which locally can be as high as 40 dams per km, or roughly one dam every 25 m. These high densities are only found
where multiple colonies maintain large dam complexes, which vary from 3 to 15 dams each (Gurnell, 1998). We express
the model output in dams per length (miles) because a) it is directly comparable to densities that can be calculated in
GIS from field GPS measurements, b) densities can also be approximated with aerial imagery and/or overflights, and
c) linear dam density is commonly reported in the literature so there are valid estimates for direct comparison. The
output categories are as follows:
None 0 dams: segments deemed not capable of supporting dam building activity
Rare > 0-1 dam/km: segments barely capable of supporting dam building activity; likely used by dispersing
beaver
Occasional > 1-5 dams/km: segments that are not ideal, but can support an occasional dam or small colony
Frequent > 5-15 dams/km: segments that can support multiple colonies and dam complexes, but may be
slightly resource limited
Pervasive > 15-40 dams/km: segments that can support extensive dam complexes and many colonies.
To assess evidence of a stream within a network being a reliable water source for dam-building beaver we use the
National Hydrography Dataset (NHD) cartographically derived 1:24 000 drainage network. The NHD network
differentiates between perennial, intermittent, and ephemeral watercourses. We use the perennial designation
segmented into 300 m long segments because a) this is a reasonable length over which to approximate reach-averaged
slope from a 10 m DEM, and b) 300 m segments produce a reasonable length along which to sample 30 m LANDFIRE
vegetation data within buffers and get a representative sample.
To assess beaver forage and building material preferences, we classify LANDFIRE EVT 2016 (first made available in
2019), a nationwide 30 m Landsat satellite imagery-based landcover classification (LANDFIRE, 2016), into beaver
dam-building material preference categories. Where 2016 LANDFIRE data was not available during modeling,
LANDFIRE EVT 2014 data was used. Based on these preferences, we assign a single numeric suitability value from
0-4 to each of the land cover classes, with zero representing unsuitable food/building material and four representing
preferred food and building material. The result is a look-up table of LANDFIRE land cover classes and associated
beaver preference values that is applied to raster data on a cell-by-cell basis.
Riverscapes with narrow riparian corridors limit beaver dam construction opportunities relative to those with expansive
riparian areas and/or adjacent deciduous forests with preferred woody browse (e.g. aspen). To represent this important
distinction, we generate two buffers along the drainage network in which we assessed beaver dam-building preference
values:
A 30 m buffer representing the streamside vegetation; and
A 100 m buffer representing the maximum harvest distance (Figure 1).
We based these buffer distances on documented distances from water that beaver typically travel to harvest woody
stems for dam and lodge construction, and winter food caches. Many studies indicate that most of the woody species
utilized by beaver occur within 30 m of the edge of water and that a majority of foraging occurs within 100 m.
Idaho BRAT
12 of 55
Figure 1: Reach scale illustration of derivation of streamside vs. riparian vegetation scores from 30 vs. 100 m stream network buffers. A shows
the 30 m and 100 m buffers, which we used to summarize intersecting pixels from 30 m resolution classified LANDFIRE raster in B. Dam
building suitability are shown in B and range from 0 (unsuitable; grey) to 4 (optimal; blue) with red for 1, yellow for 2, and green for 3. C & D
contrast the buffer averaged values for the 30 m buffer (C) and the 100 m buffer (D).
To infer whether it is likely that beaver could physically build a dam during low-flow conditions, we calculate stream
power (Ω= ρgQS) at baseflow.
Where Ω is the stream power (in Watts), ρ is the density of water (1000 kg/m3), g is acceleration due to gravity (9.8
m/s2), Q is discharge (m3/s), and S is the channel slope.
To infer the likelihood that a beaver dam will persist once built, the two-year recurrence interval peak flood (Q2) stream
power is calculated for each reach based on drainage area and USGS regional curves. To calculate reach slope we
use the NHD network segmented into 300 m long reaches and extract elevations at top and bottom of each reach
based on the DEM and divide by reach length. The two slope values that matter for the BRAT capacity model are <
0.5 percent slope because dam density goes down in very flat areas and > 23% slope because dams cannot be built
and sustained in very steep reaches. All seven lines of evidence (described above) are combined within a fuzzy
inference system (FIS) to estimate the maximum beaver dam density (dams/km) of riverscapes (Figure 2 and Figure
3).
Idaho BRAT
13 of 55
Figure 2: Methodological illustration of inputs (1-5) and output for the combined capacity model of riverscapes capacity to support beaver dam-
building activity. Model output is expressed as dam density (dams/km).
Idaho BRAT
14 of 55
Figure 3: Combined capacity Fuzzy Inference System for capacity of riverscapes to support dam building beaver activity. This shows the
specification of fuzzy membership functions with overlapping values for categorical descriptors in inputs and the output.
BRAT dam building beaver conservation and restoration model
The BRAT capacity model can help explain dam density patterns and to explore appropriate locations for Beaver Dam
Analogs (BDA) and restoration using beaver. However, a beaver dam capacity model alone is not enough to effectively
plan for large scale management and restoration of dam-building beaver. In this section, we describe how the BRAT
model assess risk and opportunity for using beaver in conservation and restoration and we argue that suitable dam-
building habitat for beaver can be planned for at statewide and watershed levels using the outputs of the capacity and
management models. The different components of the conservation and restoration model are described below and
together makeup a robust model for assessing risk based on where streams with beaver dam capacity are relative to
human infrastructure and high intensity land use.
Risk assessment using proximity to human infrastructure
There is no doubt that beaver can be a destructive nuisance in the built environment and anywhere human
infrastructure exists (Bhat et al., 1993, Hill, 1976, McKinstry and Anderson, 1999). Beaver can clog culverts, interfere
with diversions, flood public and private infrastructure, and harvest trees in undesirable locations. In an attempt to
identify potential conflicts in the built environment, the BRAT model identifies where streams are relative to
infrastructure and high intensity land use, and conservatively shows where that aligns with streams where beavers
could potentially build dams.
Idaho BRAT
15 of 55
For example, beaver are not welcome or useful additions to diversion structures or irrigation canals. Thus, distance
from canals and points of diversion are cast in terms of ‘beaver distances’ (immediately adjacent, within normal forage
range, within plausible forage range, or far enough to be outside the range of concern). The model casts the following
human infrastructure in terms of ‘beaver distances’:
1. distance to roads;
2. distance to road crossings;
3. distance to railroads;
4. distance to canals;
5. distance to points of diversion;
6. average land use intensity and;
7. distance to nearest infrastructure
However, these calculations are unrealistically restrictive because the analysis only shows how close streams are to
the above listed infrastructure, regardless of whether the infrastructure is ‘cloggable’ or ‘floodable’. To account for
‘cloggable’ or ‘floodable’ infrastructure the model defines the areas beaver dams could flood the valley bottoms. For
example, while roads can be examined in terms of distance from the channel (as described above), we should only be
interested in roads within the valley bottom (stream channels and active floodplain) if we are concerned about road
flooding problems. Thus, we calculate proximity to the following ‘floodable’ or ‘cloggable’ infrastructure:
1. roads within valley bottoms;
2. railroads within valley bottoms;
3. road crossings;
4. canals and;
5. points of diversion
Once again calculating proximity to ‘floodable’ or ‘cloggable’ infrastructure is limited because for example, a ‘road
crossing’ could be a large bridge with plenty of clearance which is not necessarily a problem, or it could be a small
culvert which is much more likely to have a clogging issue. Also, this ‘floodable’ or ‘cloggable’ infrastructure analysis is
unrealistically restrictive because this analysis only shows how close the stream is to these ‘floodable’ or ‘cloggable’
infrastructure not how close the beaver might build a dam to that infrastructure.
The dam building conservation and restoration model accounts for risk by identifying ‘floodable’ or ‘cloggable’
infrastructure that is in close proximity to the drainage network, where land use intensity is high, and where beaver are
likely to build dams along that watercourse. In other words, the model identifies reaches as having risk of human-
beaver conflict given that beaver can build dams on that reach and that those dams could potentially have a negative
effect on nearby infrastructure. The quantification of risk for each reach is based on the number of beaver dams that
could be built, how close infrastructure especially ‘floodable’ or ‘cloggable’ infrastructure is, and how high land use
intensity is. This assessment is an over-prediction of potential risk and provides a conservative prediction of potential
risk areas.
Dam building beaver management layers
We have developed dam building beaver management layers and associated maps that combine capacity estimates
and risk analysis (described above) for use in beaver related conservation and restoration. The management outputs
identify streams that are: low-risk with restoration and conservation opportunities, moderate or high risk where beavers
can build dams and; natural or anthropogenically limited areas where beaver cannot build dams now. Specifically, the
Idaho BRAT
16 of 55
three beaver management layers/maps are: 1) unsuitable or limited beaver dam building opportunities, 2) areas
beavers can build dams, but could be undesirable, and 3) ‘possible beaver dam conservation or restoration
opportunities’. An additional layer based on the surveyed dams throughout Idaho was produced representing ‘current
beaver dam management strategies’. Further documentation and discussion/development of these layers/maps can
be found here.
Model calibration
IDF&G staff selected three watersheds, Lower Henry’s Fork, Little Wood, and Jordan for BRAT model calibration.
These watersheds were selected because either ongoing beaver related conservation and restoration efforts were
taking place and/or future BDA and or translocation efforts were planned. Provisional BRAT model runs including
inputs, intermediates and outputs were provided to IDF&G staff familiar with these watersheds. Instructions on how to
effectively interrogate these datasets along with a standard list of questions to evaluate the accuracy and utility of the
model was also provided to the staff members.
Model validation
A capacity model is difficult to ‘validate’ because rarely, if ever, would the entire riverscape be at ‘capacity’. However,
since the BRAT model output is dam density, direct comparison to actual dam densities is a useful form of model
validation. We validated our model in three different ways. First, model outputs were ground truthed to confirm whether
or not the predictions seemed reasonable (e.g. places with no evidence of beaver dams are modeled with a capacity
equal to 0 dams/km). Second, actual beaver dam locations were used to calculate densities and compare surveyed
dam densities to modeled dam capacity estimates. Finally, an electivity index was used to show whether higher
preference was exhibited for beaver dam construction in reaches that predicted higher capacities.
To facilitate model validation, actual dam counts were collected using virtual reconnaissance in a web map with 2013
0.5-meter NAIP Imagery. A trained technician used the web map to examine the entire stream network within watershed
for beaver dams. The technician navigated up and down every stream in the drainage network at an ‘eye altitude’ of
roughly 500-600 m above ground (~scale 1:2000) and when potential dams were identified the technician zoomed in
and assessed other lines of visual evidence (e.g. pond shape, evidence of dam, evidence of riparian harvest, evidence
of skid trails, etc.). When likely beaver dams were identified, locations were recorded. The resulting dam location data
were used for model validation. We refer to this exercise as ‘censusing’ because the entire perennial network is
sampled. However, it is important to point out that this is not a complete census, it is a sample from a snap shot in time
of imagery. Moreover, these methods generally under-sample total dams, especially in forested ecosystems where
shadows obstruct ground imagery.
Dam counts were plotted against predicted existing capacity counts and a quantile regression was performed on the
50th, 75th, and 90th percentiles of these data. Quantile regression using upper percentiles (i.e., 75th, 90th) is used to
evaluate habitat models because many of these systems have low surveyed dam densities. Upper percentiles
represent the highest levels of surveyed dam density which correspond most closely with true capacity (Cade and
Noon, 2003), which is what we are modeling. Dam counts were also compared to predicted capacity counts spatially
in the “predicted dam density vs. surveyed dam density” layer.
Finally, to assess whether or not beaver dam-building was preferentially taking place in reaches with higher capacity
estimates, an electivity index (EI) was calculated. This logic, follows conceptually from the ‘ideal free distribution’
(Fretwell and Lucas, 1970), such that the distribution of beaver dams (in this case) should match the distribution of
resources to support such construction and maintenance activities. Following Pasternack (2011) an EI, was calculated
for each segment type (i):
=(
)
(
)
Idaho BRAT
17 of 55
where  is the number of beaver dams surveyed in segment type i and  is the length of that segment type. The EI
essentially normalizes utilization by availability such that i) an EI value of one indicates utilization of available habitat
without preference or avoidance, ii) an EI value less than one indicates avoidance of a particular habitat, whereas iii)
an EI value greater than one indicates preference for a habitat. The segment types (i) are a classification that
corresponds to the linguistic categories used in the FIS. If the capacity model is effectively segregating actual dam
densities, we would expect an EI close to zero for the none and rare classes, less than one for the occasional class,
greater than one for the frequent class, and much greater than one for the pervasive class.
Idaho BRAT
18 of 55
RESULTS & INTERPRETATIONS
Model outputs
The Idaho BRAT outputs consist of seven stream network classifications: 1) existing beaver dam capacity (Figure 4),
2) historic beaver dam capacity (Figure 5), 3) existing dam complex size (Figure 6), 4) historic dam complex size
(Figure 7), 5) restoration or conservation opportunities (Figure 8), 6) areas beavers can build dams, but could be
undesirable (Figure 9), and 7) unsuitable or limited beaver dam opportunities areas beavers can build dams (Figure
10). These figures represent results from the Southwest Region. Additional statewide, regional and watershed level
data and summary products can be found here.
Beaver dam capacity model
Idaho statewide & by administrative regions
The BRAT beaver dam capacity model was run for existing (based primarily on 2016 imagery) and an estimate of
historic conditions based on LANDFIRE BPS data. The estimated existing Idaho statewide capacity is 994,299 dams
or 8 dams/km (Table 3). By contrast, the same model driven with estimates of historic vegetation types estimated the
statewide capacity at 1,743,459 dams or 15 dams/km reflecting a 43% loss compared to historic capacity (Table 3).
Comparison of the IDF&G administrative regions of Idaho are also illustrated in (Table 3).
Nearly all of the capacity loss from historic conditions can be explained in terms of riparian vegetation loss, vegetation
conversion and degradation associated with high intensity land use including: 1) conversion of valley bottoms to urban
and agricultural land uses, 2) overgrazing in riparian and upland areas, 3) conifer encroachment of wet meadow areas.
Despite the losses in beaver dam capacity, Idaho’s waterways are still capable of supporting and sustaining a
substantial amount of beaver dam-building activity (994,299 dams).
Table 3: Gross summary of contrast between existing and historic beaver dam capacity estimates for Idaho by administrative region.
Idaho BRAT
19 of 55
Figure 4: Modeled beaver dam capacity for existing conditions for the Southwest Region of Idaho.
Idaho BRAT
20 of 55
Figure 5: Modeled beaver dam capacity for historic conditions for the Southwest Region of Idaho.
Idaho BRAT
21 of 55
Figure 6 shows the comparison between existing and historic capacity for the Southwest Region of Idaho.
Figure 6. Existing vs. historic dam capacity by category for Southwest Region of Idaho.
Dam building capacity can be re-cast in terms of what size dam complex (single dam, small complex (1-3 dams),
medium complex (3-5 dams), or large complex (greater than 5 dams) can fit in a reach. Existing dam complex size for
southwest region of Idaho is mapped in Figure 7. Historic dam complex size for southwest region of Idaho is mapped
in Figure 8.
Idaho BRAT
22 of 55
Figure 7: Modeled existing dam complex size for the Southwest Region of Idaho.
Idaho BRAT
23 of 55
Figure 8: Modeled historic dam complex size for the Southwest Region of Idaho.
Idaho BRAT
24 of 55
Management layers
Conservation and restoration opportunities
The BRAT model identifies opportunities where low-risk restoration and conservation opportunities exist for using
beaver in stream conservation and restoration (Figure 9). The layer/map represents areas with limited ‘risk’ of human-
beaver conflict and some level of existing capacity. The map consists of the following categories: i) ‘easiest - low
hanging fruit’ has capacity, just needs beaver, ii) ‘straight forward - quick return’ is currently occasional capacity but
historically was higher capacity, iii) ‘strategic’ is currently degraded but historically was higher capacity. These areas
typically need long-term riparian recovery first (e.g. grazing management), and 4) other’. The othercategory is based
on higher ‘risk’ of human-beaver conflict and lower existing dam building capacity (i.e., reaches that are likely not worth
beaver dam related conservation and restoration actions). 27% of Idaho is categorized as ‘easiest’ conservation and
restoration opportunities, suggesting there are many low-hanging fruit beaver-related restoration opportunities
available.
Figure 9: Modeled conservation and restoration opportunities for the Southwest Region of Idaho.
Idaho BRAT
25 of 55
Potential risk areas
The BRAT model identifies potential risks areas -- streams that are close to human infrastructure or high land use
intensity and where the capacity model estimates that beavers can build dams. The layer/map is called ‘areas beavers
can build dams, but could be undesirable(Figure 10). 59% of Idaho is categorized as ‘negligible’ risk, suggesting there
are many low risk beaver-related restoration opportunities available.
Figure 10: Modeled areas beavers can build dams but could be undesirable for the Southwest Region of Idaho.
Idaho BRAT
26 of 55
Unsuitable or limited dam building opportunities
The BRAT model also identifies areas where beaver cannot build dams now, and differentiates these into
anthropogenically and naturally limiting areas. The layer/map is called unsuitable or limited dam building opportunities
(Figure 11). Only 18% of all perennial streams in Idaho were categorized as unsuitable for beaver dam building.
Figure 11: Modeled unsuitable/limited dam building opportunities for the Southwest Region of Idaho.
Idaho BRAT
27 of 55
Model intermediates
Anthropogenic constraints layers
The BRAT model identifies anthropogenic constraints based on proximity to human infrastructure (Figure 12) and
includes distance to roads (Figure 13), distance to roads within valley bottoms (Figure 14), distance to road crossings
(Figure 15), distance to railroads, distance to railroads within valley bottoms, distance to canals (Figure 16), distance
to points of diversion, and distance to nearest infrastructure (Figure 17). The BRAT model identifies also identifies land
use intensity (Figure 18). This proximity analysis highlights possible risks (i.e. flooding and clogging) that might occur
to this infrastructure. These figures are results from the Little Wood watershed. Additional watershed level intermediate
data can be found here.
Context layers
Figure 12 shows the Little Wood watershed as an example of the context layers: canals, valley bottoms, roads, and
railroads that the model used to calculate overall and individual proximity of streams to human infrastructure.
Figure 12: Little Wood watershed context layers used in the Idaho BRAT model.
Idaho BRAT
28 of 55
Distance to road
Figure 13 shows the Euclidean distance of roads to the perennial stream network and was calculated because beaver
dam building can flood roads. Hence, roads in close proximity to streams where beaver can build dams are considered
a potential risk area.
Figure 13: Little Wood watershed Euclidean distance to roads layer used in the Idaho BRAT model.
.
Idaho BRAT
29 of 55
Distance to roads within valley bottoms
Figure 14 shows the Euclidean distance of roads within valley bottoms to the perennial stream network and was
calculated because beaver dam building can flood roads that are in the floodplain (valley bottom). Hence, roads within
valley bottoms in close proximity to streams where beaver can build dams are considered a high risk area.
Figure 14: Little Wood watershed Euclidean distance to roads within the valley bottom layer used in the Idaho BRAT model.
Idaho BRAT
30 of 55
Distance to road crossing
Figure 15 shows the Euclidean distance of road crossings to the perennial stream network and was calculated because
road crossings can be advantageous locations for beavers to build dams but can cause damage to infrastructure in
these areas. For example, beaver dams can be constructed on the upstream end of culverts causing potential clogging
of issues and flooding of roadways. Hence, road crossings are considered a high risk area.
Figure 15: Little Wood watershed Euclidean distance to road crossings layer used in the Idaho BRAT model.
Idaho BRAT
31 of 55
Distance to canal
Figure 16 shows the Euclidean distance of canals to the perennial stream network and was calculated because beaver
dams in canals can alter the flow of water in canals designed to provide stable and undisturbed water delivery. As
such, canals in close proximity to streams where beaver can build dams are considered a high risk area.
Figure 16: Little Wood watershed Euclidean distance to canals layer used in the Idaho BRAT model.
Idaho BRAT
32 of 55
Distance to infrastructure
Figure 17 shows the minimum distance of streams to human infrastructure, including roads, roads in the valley
bottom, road crossings, and canals. This is a simple visualization of overall risk to man-made structures and shows
reaches where beaver may cause impacts, though risk of an individual reach varies by the susceptibility of
infrastructure to flooding and the likelihood of beaver building dams there.
Figure 17: Little Wood watershed Euclidean distance to human infrastructure layer used in the Idaho BRAT model.
Idaho BRAT
33 of 55
Land use intensity
Figure 18 shows land use intensity for the perennial stream network and was modeled with the rational that the
higher the land use intensity score the higher the potential human-beaver conflict.
Figure 18: Little Wood watershed land use intensity layer used in the Idaho BRAT model.
Idaho BRAT
34 of 55
Model calibration
IDF&G staff provided concrete and detailed written feedback regarding the accuracy, and usability of the BRAT
model in the calibration watersheds. The vast majority of the feedback can be assigned to the following four
categories:
1. Low resolution input data: As anticipated, IDF&G staff provided examples of where input vegetation and land
use data was lacking in detail and correctness. The input data for our provisional run for both landcover and land
use was 2014 LANDFIRE EVT data. It was fortuitous that in May 2019, during this review process, LANDFIRE
made the 2016 existing landcover/land use data available and we were able to rerun the BRAT model primarily
using this updated dataset, though LANDFIRE 2014 had to be used for small portions of southern Idaho.
2. Inadequate canal and irrigation infrastructure: We resolved this issue by identifying points of diversion data,
adding ditches to the dataset, and modifying the conservation and restoration model to identify any canals or
ditches as high risk areas.
3. Lack of land ownership information: We improved the model to include land ownership and created a script
that calculated distance from private land as a way to help identify areas suitable for beaver translocation.
4. Risk model parameters are too restrictive: As directed we adjusted the risk model parameters to make the
model less restrictive.
Model validation
We used three forms of model verification to assess the performance of the capacity model.
1. Are there surveyed dams where the model predicted existing dam capacity as none?
2. How do dam densities track between predicted and actual?
3. Do the electivity indices increase appreciably from the none to the pervasive class?
How do dam densities track between predicted and actual?
Figure 19 shows a total of 9,048 actual beaver dams were identified throughout Idaho based on virtual reconnaissance
using a webmap. Dams were concentrated in reaches where the BRAT capacity model estimated frequent and
pervasive dam densities
Idaho BRAT
35 of 55
Figure 19: Actual beaver dam (based on 2013 0.5 meter NAIP imagery) identified throughout Idaho from desktop dam censusing. Of the 9,048
dams 8,060 were within 60 m of the NHD stream lines and therefore, were used in the model validation.
Idaho BRAT
36 of 55
Idaho Statewide
We verified the performance of the existing capacity model using a total of 8,060 actual beaver dams representing
0.8% of the 118,319 km of perennial streams across the state of Idaho. 70 beaver dams were found where the model
predicted no dams could be supported. Of the total 3,723 stream segments with validation dam counts 382 exceeded
the capacity estimates indicating that the model effectively segregates the factors controlling beaver dam occurrence
and density 89.74% of the time (Figure 20). Thus, the model run with 30 m LANDFIRE data only underestimated
capacity 10.26% of the time (Figure 20). Validation data can be found here.
Figure
20: Predicted vs. observed dam counts (per reach) for Idaho. The blue line is line of perfect agreement (1:1 relationship),
dashed line is the 50th percentile (median) regression, dotted line is the 75
th percentile regression, and dot-dashed line is the 90th
percentile regression (not visible due to direct correspondence with line of perfect agreem
ent in this graphic). The nearer the higher
percentile regression is to the 1:1 ratio, the better the model is performing.
Idaho BRAT
37 of 55
Clearwater Region
For the Clearwater Region we verified the performance of the existing capacity model using 274 dams representing
0.2% of the 22,901km of perennial streams across the region. 8 beaver dams were found where the model predicted
no dams could be supported. Of the total 149 stream segments with validation dam counts 19 exceeded the capacity
estimates indicating that the model effectively segregates the factors controlling beaver dam occurrence and density
87.25% of the time (Figure 21). Thus, the model run with 30 m LANDFIRE data underestimated capacity 12.75% of
the time (Figure 21). Validation data can be found here.
Figure 21: Predicted vs. observed dam counts (per reach) for Clearwater Region. The blue line is line of perfect agreement (1:1
relationship), dashed line is the 50th percentile (median) regression, dotted line is the 75th percentile regression, and dot-dashed line
is the 90th percentile regression. The nearer the higher percentile regression is to the 1:1 ratio, the better the model is performing.
Idaho BRAT
38 of 55
Magic Valley Region
For the Magic Valley Region we verified the performance of the existing capacity model using 864 dams representing
0.7% of the 15,690 km of perennial streams across the region. 11 beaver dams were found where the model predicted
no dams could be supported. Of the total 406 stream segments with validation dam counts 60 exceeded the capacity
estimates indicating that the model effectively segregates the factors controlling beaver dam occurrence and density
85.22% of the time (Figure 22). Thus, the model run with 30 m LANDFIRE data underestimated capacity 14.78% of
the time (Figure 22).
Figure
22: Predicted vs. observed dam counts (per reach) for the Magic Valley Region. The blue line is line of perfect agreement
(1:1 relationship), dashed line is the 50th percentile (median) regression, dotted
line is the 75th percentile regression, and dot-
dashed line is the 90
th percentile regression. The nearer the higher percentile regression is to the 1:1 ratio, the better the model is
performing.
Idaho BRAT
39 of 55
Panhandle Region
For the Panhandle Region we verified the performance of the existing capacity model using 684 dams representing
0.8% of the 11,962 km of perennial streams across the region. 4 beaver dams were found where the model predicted
no dams could be supported. Of the total 357 stream segments with validation dam counts 20 exceeded the capacity
estimates indicating that the model effectively segregates the factors controlling beaver dam occurrence and density
94.40% of the time (Figure 23). Thus, the model run with 30 m LANDFIRE data underestimated capacity 5.60% of the
time (Figure 23).
Figure
23: Predicted vs. observed dam counts (per reach) for the Panhandle Region. The blue line is line of perfect agreement (1:1
relationship), dashed line is
the 50th percentile (median) regression, dotted line is the 75th percentile regression, and dot-
dashed line
is the 90
th percentile regression. The nearer the higher percentile regression is to the 1:1 ratio, the better the model is performing.
Idaho BRAT
40 of 55
Salmon Region
For the Salmon Region we verified the performance of the existing capacity model using 696 dams representing 0.7%
of the 13,867 km of perennial streams across the region. 18 beaver dams were found where the model predicted no
dams could be supported. Of the total 379 stream segments with validation dam counts 38 exceeded the capacity
estimates indicating that the model effectively segregates the factors controlling beaver dam occurrence and density
89.97% of the time (Figure 24). Thus, the model run with 30 m LANDFIRE data underestimated capacity 10.03% of
the time (Figure 24).
Figure
24: Predicted vs. observed dam counts (per reach) for the Salmon Region. The blue line is line of perfect agreement (1:1
relationship), dashed line is
the 50th percentile (median) regression, dotted line is the 75th percentile regression, and dot-
dashed line
is the 90
th percentile regression. The nearer the higher percentile regression is to the 1:1 ratio, the better the model is performing.
Idaho BRAT
41 of 55
Southeast Region
For the Southeast Region we verified the performance of the existing capacity model using 2,248 dams representing
2.1% of the 10,469 km of perennial streams across the region. 2 beaver dams were found where the model predicted
no dams could be supported. Of the total 843 stream segments with validation dam counts 102 exceeded the capacity
estimates indicating that the model effectively segregates the factors controlling beaver dam occurrence and density
87.90% of the time (Figure 25). Thus, the model run with 30 m LANDFIRE data underestimated capacity 12.10% of
the time (Figure 25).
Figure
25: Predicted vs. observed dam counts (per reach) for the Southeast Region. The blue line is line of perfect agreement (1:1
relationship), dashed line is t
he 50th percentile (median) regression, dotted line is the 75th percentile regression, and dot-
dashed line
is the 90
th percentile regression. The nearer the higher percentile regression is to the 1:1 ratio, the better the model is performing.
Idaho BRAT
42 of 55
Southwest Region
For the Southwest Region we verified the performance of the existing capacity model using 1,002 dams representing
0.5% of the 30,770 km of perennial streams across the region. 23 beaver dams were found where the model predicted
no dams could be supported. Of the total 614 stream segments with validation dam counts 66 exceeded the capacity
estimates indicating that the model effectively segregates the factors controlling beaver dam occurrence and density
89.25% of the time (Figure 26). Thus, the model run with 30 m LANDFIRE data underestimated capacity 10.75% of
the time (Figure 26).
Figure
26: Predicted vs. observed dam counts (per reach) for the Southwest Region. The blue line is line of perfect agreement (1:1
relationship), dashed line i
s the 50th percentile (median) regression, dotted line is the 75th percentile regression, and dot-
dashed line
is the 90
th percentile regression. The nearer the higher percentile regression is to the 1:1 ratio, the better the model is performing.
Idaho BRAT
43 of 55
Upper Snake Region
For the Upper Snake Region we verified the performance of the existing capacity model using 2,292 dams representing
2.0% of the 12,492 km of perennial streams across the region. 4 beaver dams were found where the model predicted
no dams could be supported. Of the total 980 stream segments with validation dam counts 78 exceeded the capacity
estimates indicating that the model effectively segregates the factors controlling beaver dam occurrence and density
92.04% of the time (Figure 27). Thus, the model run with 30 m LANDFIRE data underestimated capacity 7.96% of the
time (Figure 27).
Figure
27: Predicted vs. observed dam counts (per reach) for the Upper Snake Region. The blue line is line of perfect agreement
(1:1 relationship), dashed line is the 50th percentile (median) regression, dotted line is the 75
th percentile regression, and dot-
dashed line is the 90
th percentile regression. The nearer the higher percentile regression is to the 1:1 ratio, the better the model is
performing.
Idaho BRAT
44 of 55
Do the electivity indices increase appreciably from the none to the pervasive class?
The Electivity indices results in Table 4, Table 5, Table 6, Table 7, Table 8, Table 9, Table 10 and Table 11 show that
throughout the perennial streams of Idaho beavers preferentially dam in reaches with higher modelled dam capacity
while avoiding those with lower capacity. That is, beaver exhibited avoidance of reaches predicted as supporting none,
rare or occasional densities, and beaver exhibited preference for areas predicted as having frequent or pervasive dam
densities.
Table 4: Existing number of dams and BRAT modeled capacity estimates Idaho statewide.
Table 5: Existing number of dams and BRAT modeled capacity estimates for the Clearwater Region.
Table 6: Existing number of dams and BRAT modeled capacity estimates for the Magic Valley Region.
Idaho BRAT
45 of 55
Table 7: Existing number of dams and BRAT modeled capacity estimates for the Panhandle Region.
Table 8: Existing number of dams and BRAT modeled capacity estimates for the Salmon Region.
Table 9: Existing number of dams and BRAT modeled capacity estimates for the Southeast Region.
Idaho BRAT
46 of 55
Table 10: Existing number of dams and BRAT modeled capacity estimates for the Southwest Region.
Table 11: Existing number of dams and BRAT modeled capacity estimates for the Upper Snake Region.
Idaho BRAT
47 of 55
DELIVERABLES
This project will consist of implementing BRAT for the entire state of Idaho. This included calibration, validation and
refinement of the BRAT models based on feedback for IDF&G staff.
1. Develop and run BRAT on the perennial portion of the 1:24K NHD network (60,484 miles) segmented at 300
m including:
Existing (based on 2016 LANDFIRE data, the most current data available) beaver dam capacity
estimates (dams /km)
Historic beaver dam capacity estimates (based on LANDFIRE BPS data (dams /km)
Potential for human beaver conflict (probability)
o Refine conflict model. This resulted in a model that asses risk and opportunity for using
beaver in conservation and restoration.
Beaver management, restoration and conservation predictions reflecting management priorities
2. Calibration and validation of BRAT performance using harvest and nuisance trapping data.
3. Google Earth-based virtual inventory of existing beaver dams and field based data collection of dam
locations.
Deliverable data products include:
1. KMZ of each of four primary BRAT outputs (existing capacity, historic capacity, risk, limiting factors, and
conservation and restoration opportunities)
2. Shapefiles and layer packages of each of four primary outputs (existing capacity, historic capacity, risk,
limiting factors, and conservation and restoration opportunities)
3. Posting of existing and historic beaver dam capacity output layers to public website
4. Full interpretive report with management recommendations
5. Idaho-wide virtual inventory of existing beaver dams as KMZ
6. Atlas of BRAT outputs
GIS data layers
The GIS data layers that make up the maps are available in KMZ for the perennial network of each HUC 8 watershed,
as an example American Falls data is found here, shapefile here and layer package formats here and enable
visualization and querying in GIS programs. We encourage the use of the layer packages because this format provides
all the inputs, intermediates and outputs symbolized in a standard format which increases their usability. Viewing the
KMZ files in Google Earth or ArcGIS Earth is an effective way to visualize and interrogate these output datasets
because of the 3-D capabilities, image rendering speed and the quality of the base imagery. If you need help using the
GIS data we have developed a series of tutorial videos and other instructions found here. For non-GIS users we have
generated an Esri Story Map of the project that can be viewed here and a map atlas of BRAT outputs which, can be
found here.
Idaho BRAT
48 of 55
DISCUSSION & RECOMMENDATIONS
Caveats
Although BRAT was run with freely-available, national data, and produces very accurate results, some caveats should
be kept in mind:
The capacity models are only as good as the inputs. As shown in the results section, the logic of the capacity
model and model performance is robust (i.e. the model gets the right answers for the right reasons and the
wrong answers for the right reasons namely, if the inputs are inaccurate). However, this does not mean that
the model will be correct on each and every reach of stream that was modeled.
The vegetation mapping, digital elevation model and drainage network (stream position) mapping are all
relatively coarse, and have inherent inaccuracies, when examined at scales at the limits of their precision.
The BRAT samples across reasonable extents at reasonable, reach-scale resolutions to make the
significance of those inaccuracies less impactful. As such, model verification shows that the model does a
excellent job at capturing capacity in most cases.
Small streams with narrow bands of riparian vegetation may not have the spatial extent to be resolved in 30
m datasets, like LANDFIRE. This is particularly true for highly incised streams because incised streams are
hydrologically disconnected from their channels and as floodplains and channels are decoupled, riparian plant
performance declines, eliminating the existence of many riparian species. As such, in such settings, higher
resolution vegetation inputs may be more appropriate (e.g., Macfarlane et al., 2016), or an on-the-ground
assessment may be necessary. Fortunately, with minor modifications, the BRAT tool can be run with higher
resolution input data. Even with this know limitation, we still find that 30 m resolution LANDFIRE data is an
appropriate input for watershed-scale evaluations of beaver dam capacity.
While investment in higher resolution data could improve the precision of the model outputs in some localities,
and pixel-by-pixel accuracy, it could cost a lot more without that great of an increase in accuracy or utility of
the outputs. If these high resolution data become freely available in the future, or very affordable, it may make
sense to run an updated version of the BRAT model with new inputs. However, we do not recommend
undertaking expensive data acquisition campaigns for the sole purpose of improving the model outputs.
The capacity model and infrastructure layers are all based on a 2016 snapshot of conditions.
The BRAT model predicts only the maximum number of dams that can be supported, not the expected number
of dams across a given area.
Figure 9 shows where there are low-risk restoration and conservation opportunities using beaver. However,
this map is an under estimation of opportunities (i.e., conservative)
How does one determine whether a reach should be a conservation or restoration reach? The BRAT model
does this by comparing realized dam counts to existing capacity. ‘Low hanging fruit’ reaches with >25% of
existing capacity already occupied by dams are flagged as ‘Immediate - Beaver Conservation reaches where,
for example, a trapping closure could be implemented. ‘Low hanging fruit’ reaches with no realized-capacity
and/or under-utilized-capacity should be target for restoration and/or translocation. These reaches are good
candidates for using BDAs to promote beaver retention. ‘Quick-return’ reaches are also good candidates for
BDAs and better land management to improve conditions for encouraging beaver recovery in the area.
Figure 10 is a conservative first cut of ‘risk’ and just because this layer/map shows there to be some risk does
not mean that beaver impacts will actually be a problem and if beaver impacts occur, they might easily be
mitigated.
The BRAT model can be manually run as a simple inference system in the field using a data capture form
found here. Using the field-based BRAT data form one can override the model outputs that are based on
remotely sensed data. The field-based BRAT data collection, that relies on ocular estimates and expert
opinion can be done in an opportunistic way as one traverses the watershed or could be a more planned data
capture campaign focused on priority restoration areas.
Idaho BRAT
49 of 55
In this medium-sized contract ($74,694), the expectation management and vision provided by the BRAT model
could stand to save millions of dollars. If one considers the cost of current restoration practices, the scope of
areas that could use improvement, and the relatively low-cost of beaver-assisted restoration, dramatic gains
and improvements could be made.
As the existing capacity model and statewide dam census indicates, Idaho is only at .8% of existing capacity
indicating there are numerous opportunities throughout the watershed to increase the amount of beaver dam-
building activities.
Future versions of BRAT
Since 2014, BRAT has undergone continuous development and during the scope of this work has undergone a major
overhaul and is now BRAT 3.0.20. However, the heart of the BRAT model the capacity estimates has not changed
and will not change in subsequent versions of BRAT. BRAT 4 is planned to be released in 2020. BRAT 4 will be open
source and web-based without any Esri dependences. BRAT 4 will also be fully compatible with RAVE
(http://rave.riverscapes.xyz), which allows users to easily add model outputs to map and view metadata from the
project. Unfortunately, BRAT 3.0.20 outputs are not RAVE compliant at this point. Instead, shapefiles and layer
packages (to maintain organization and symbology) have been provided here in this report. If significant changes occur
in the watershed (e.g. through disturbances like fire, land-use practices and restoration) it may make sense to re-run
BRAT capacity model for a future snap-shot in time.
Idaho BRAT
50 of 55
Riverscapes network models to inform management
There are a variety of ‘sister’ network models in the Riverscapes Consortium (http://riverscapes.xyz). For example, the
Riparian Condition Assessment Toolbox (RCAT: http://rcat.riverscapes.xyz), could be quite helpful for examining
riparian conditions to contextualize BRAT results as well as exploring recovery potential for riparian improvement to
expand beaver dam-building capacity. By contrast, the Geomorphic Network Assessment Tool (GNAT-
http://gnat.riverscapes.xyz) could be helpful in terms of building more realistic expectations for defining what is
physically and geomorphically possible at a given site.
Beaver management recommendations
Beaver Dam Analogues
From our experience working in locations such as Birch Creek, Idaho (http://beaver.joewheaton.org/logan-workshop-
materials.html - See Jay Wilde’s story) we have found that building Beaver Dam Analogues (BDAs) at release sites to
provide cover from predation can significantly increase the likelihood of successful translocations. As such, we
recommend that BDAs be built at release sites where appropriate.
We have also learned that BDAs and PALS (post-assisted log structures) can be an affordable addition to streams and
that these structures allow the opportunistic behavior of beaver to be taken advantage of (Wheaton et al., 2019).
Specifically, we have seen beaver switch from bank-lodging to dam building where these structural elements are
available.
‘Living with beaver’ strategies
Traditionally, beaver management has relied on lethal trapping to prevent threats to infrastructure posed by beaver
dam building activity. The increased awareness of the ecosystem benefits provided by beaver activity and their ability
to help achieve a number of restoration goals has spurred the development of approaches capable of mitigating the
negative results of beaver activity in order to retain the benefits such activity produces. Here we summarize a number
of ‘living with beaver’ strategies.
Breach dam
Breaching or partial breaching (i.e., notching) a dam is an effective way to mitigate the risk of flooding due to a specific
dam, if that dam is no longer being actively maintained by beaver. Breaching, rather than full removal, allows managers
to effectively control the water height of the dam while also still retaining the ecosystem services provided by such a
dam. Breaching a dam is not an effective strategy if the dam is being actively maintained, given beavers’ ability to
repair breaches within short periods of time (i.e., hours to days).
Notch dam and install beaver deterrent
In areas where an actively maintained dam is posing a threat of flooding but has not reached a critical level, notching
the dam to reduce the pond height and installing a beaver deterrent may reduce the threat of flooding. A beaver
deterrent is simply a white sheet that is strung between two fence posts and placed just upstream of the notched dam,
such that it can move freely in the wind. The sheet is cut vertically to create strips that can blow in the wind. The
movement of the sheet deters beaver from repairing the notched dam (Figure 28). This approach is very inexpensive
and an excellent first approach to dealing with potentially threatening pond heights.
Idaho BRAT
51 of 55
Figure 28: Schematic of a beaver deterrent used to control pond height.
Install pond leveler to control pond height
Pond levelers are another way managers and land owners can mitigate the risk of flooding due to beaver activity while
allowing beaver to remain in a given area. Pond levelers installation typically requires a half-day of labor for 2-3 people
and materials cost approximately $600 1000 depending on site specific conditions. A pond leveler consists of a
flexible, perforated plastic pipe that has an inflow protected by a large metal cage and is anchored to the bottom of the
pond, and runs through the dam, and is set at the desired water level height. It may be necessary to notch the dam in
order to set the pipe at the desired pond height. Following installation, we recommend placing additional material over
the end of the pipe in order to prevent beaver from clogging the outflow. Examples of a pond leveler installation are
shown in Figure 29.
Figure 29: Pond leveler installation. From left: securing flexible pipe in cage to protect inflow from being clogged; placing pipe into beaver pond;
rebuilding beaver dam after setting pipe into notched dam at desired water height.
Idaho BRAT
52 of 55
Beaver deterrent to prevent culvert/irrigation diversion clogging
As shown above beaver deterrents (Figure 28) can be used pre-emptively in order to prevent beaver from becoming
active in areas that are determined to be high risk. We recommend using beaver deterrents where streams are diverted
for irrigation.
Removal, live trapping and relocation
If beaver activity is having a negative impact and/or posing unacceptable risks, and ‘living with beaver’ strategies have
proved ineffective, then removal, whether by live trapping and relocation or lethal trapping may be required. We strongly
recommend live-trapping and relocation in order to maintain the benefits of beaver activity elsewhere in the watershed,
and further recommend that lethal trapping should be treated as a last resort. While we recognize that beaver activity
may pose a threat at any time of year, we recommend, when possible that trapping and relocation do not take place
during winter months, when their chances of survival are limited. Choosing an appropriate relocation site, with suitable
habitat, and limited threats to infrastructure is also critical.
CONCLUSIONS
With the development of the Idaho BRAT model the scope of what is possible in terms of partnering with beaver for
restoration is now clearly defined and mapped. For example, the ‘easiest: low-hanging fruit’ reaches identified in Figure
9 should be further analyzed for beaver related conservation and restoration. We believe the Idaho BRAT model helps
build realistic expectations about what beaver dam-building may achieve locally on a given stream, and also helps
scale-up those expectations at the watershed level. BRAT model outputs can be used to initialize restoration and
conservation planning and can also support initial conceptual design and siting of specific restoration actions. BRAT
model outputs can also aid with expectation management, and conservation and restoration prioritization.
As we well know, beaver dam building activities can cause conflict where valuable infrastructure and/or land is
impacted. Many conflicts can be managed to minimize damage while ensuring animal welfare and delivering ecosystem
benefits. Understanding the capacity of streams to support dam building and identifying areas of risk and opportunity
is therefore critically important for the effective beaver management. This application of BRAT provides the information
needed to understand actual and potential beaver dam capacities, where human infrastructure is present, where
nuisance beaver can be relocated, where ‘living with beaver’ strategies may be needed and where beaver can be
employed in watershed conservation and restoration efforts to recover degraded streams, meadows and wetlands.
Idaho BRAT
53 of 55
ACKNOWLEDGMENTS
We would like to thank Cory Mosby for his efforts to make this project a reality and for Curtis Hendricks, Matt Proett,
Rex Sallabanks, Rita Dixon, Sonya Knetter, Chris Murphy, Leona Svancara, Casey Pozzanghera, Norm Merz, Laura
Wolf, Joel Sauder, Diane Evans Mack, Becky Abel, Micah Ellstrom, Clay Hickey, Don Jenkins, Brad Lowe, Christopher
Yarbrough, Jamie Utz, Anna Owsiak, Matt Pieron, Rick Ward, Tempe Regan, Regan Berkley, David Dressel, Jeanne
McFall, Art Butts, Lauren Andrews, David Teuscher, Angie Schmidt, and Terry Gregory for their participation in the
workshop, model calibration and beaver dam census efforts.
Idaho BRAT
54 of 55
REFERENCES
BHAT, M. G., HUFFAKER, R. G. & LENHART, S. M. 1993. Controlling Forest Damage by Dispersive Beaver
Populations - Centralized Optimal Management Strategy. Ecological Applications, 3, 518-530.
BIRD, B., O’BRIEN, M. & PETERSON., M. 2011. Beaver and Climate Change Adaptation in North America:
A Simple, Cost-Effective Strategy for the National Forest System. WildEarth Guardians, Grand
Canyon Trust, & The Lands Council.
BOUWES, N., WEBER, N., JORDAN, C. E., SAUNDERS, W. C., TATTAM, I. A., VOLK, C., WHEATON, J. M. &
POLLOCK, M. M. 2016. Ecosystem experiment reveals benefits of natural and simulated beaver
dams to a threatened population of steelhead (Oncorhynchus mykiss). Scientific Reports, 6, 28581.
CURRAN, J. C. & CANNATELLI, K. M. 2014. The impact of beaver dams on the morphology of a river in the
eastern United States with implications for river restoration. Earth Surface Processes and
Landforms, 39, 1236-1244.
FRETWELL, S. D. & LUCAS, H. L. 1970. On territorial behavior and other factors influencing habitat
distribution in birds. I. Theoretical Development. Acta Biotheoretica, 19, 16-36.
GOLDFARB, B. 2018. Eager: The Surprising, Secret Life of Beavers and Why They Matter, Chelsea Green
Publishing.
GURNELL, A. M. 1998. The hydrogeomorphological effects of beaver dam-building activity. Progress in
Physical Geography, 22, 167-189.
HETER, E. W. 1950. Transplanting Beaver by Airplane and Parachute. Journal of Wildlife Management, 14,
143-147.
HILL, E. P. Control methods for nuisance beaver in the southeastern United States. 1976. 25.
LANDFIRE. 2014. Existing Vegetation Type (EVT) layer [Online]. Landscape Fire and Resource Management
Planning Tools Project. Available: http://www.landfire.gov/NationalProductDescriptions21.php
[Accessed June-July 2014].
MACFARLANE, W. W., MCGINTY, C. M., LAUB, B. G. & GIFFORD, S. J. 2016. High-resolution riparian
vegetation mapping to prioritize conservation and restoration in an impaired desert river.
Restoration Ecology, n/a-n/a.
MACFARLANE, W. W., WHEATON, J. M., BOUWES, N., JENSEN, M. L., GILBERT, J. T., HOUGH-SNEE, N. &
SHIVIK, J. A. 2017. Modeling the capacity of riverscapes to support beaver dams. Geomorphology,
277, 72-99.
MCKINSTRY, M. C. & ANDERSON, S. H. 1999. Attitudes of private- and public-land managers in Wyoming,
USA, toward beaver. Environmental Management, 23, 95-101.
NAIMAN, R. J., MELILLO, J. M. & HOBBIE, J. E. 1986. Ecosystem alteation of boreal forest streams by beaver
(Castor canadensis). Ecology, 67, 1254-1269.
PASTERNACK, G. B. 2011. 2D Modeling and Ecohydraulic Analysis, Seattle, WA, Createspace.
PAULSEN, S., STODDARD, J., HOLDSWORTH, S., MAYIO, A. & TARQUINIO, E. 2006. Wadeable streams
assessment: A collaborative survey of the nation’s streams. Report No. EPA.
POLLOCK, M. M., BEECHIE, T. J., WHEATON, J. M., JORDAN, C. E., BOUWES, N., WEBER, N. & VOLK, C. 2014.
Using Beaver Dams to Restore Incised Stream Ecosystems. Bioscience, 64, 279-290.
ROSELL, F., BOZSER, O., COLLEN, P. & PARKER, H. 2005. Ecological impact of beavers Castor fiber and
Castor canadensis and their ability to modify ecosystems. Mammal review, 35, 248-276.
SILVERMAN, N. L., ALLRED, B. W., DONNELLY, J. P., CHAPMAN, T. B., MAESTAS, J. D., WHEATON, J. M.,
WHITE, J. & NAUGLE, D. E. 2018. Low-tech riparian and wet meadow restoration increases
vegetation productivity and resilience across semiarid rangelands. Restoration Ecology, 0.
Idaho BRAT
55 of 55
WHEATON, J. M., BENNETT, S. N., BOUWES, N., MAESTAS, J. D. & SHAHVERDIAN, S. M. 2019. Low-Tech
Process-Based Restoration of Riverscapes: Design Manual. Version 1.0., Logan, Utah, Utah State
University Restoration Consortium.
WRIGHT, J. P., JONES, C. G. & FLECKER, A. S. 2002. An ecosystem engineer, the beaver, increases species
richness at the landscape scale. Oecologia, 132, 96-101.
... Specifically, BRAT is a spatially explicit network tool that predicts where along streams and rivers beaver may be useful as a restoration tool and where they may be a nuisance, in which case their impacts can be mitigated or the nuisance beaver can serve as a source population for live-trapping and relocation to areas where they can help achieve restoration and conservation objectives. The BRAT model has been run for the entire state of Utah (Macfarlane et al. 2017), the entire state of Idaho (Macfarlane et al. 2019), and large portions of Wyoming and Oregon. Additionally, BRAT has been run for smaller portions of New York, Washington and the United Kingdom. ...
Technical Report
Full-text available
The purpose of this design manual is to provide restoration practitioners with guidelines for implementing a subset of low-tech tools —namely beaver dam analogues (BDAs) and post-assisted log structures (PALS)—for initiating process-based restoration in structurally-starved riverscapes. While the concept of process-based restoration in riverscapes has been advocated for at least two decades, details and specific examples on how to implement it remain sparse. Here, we describe ‘low-tech process-based restoration’ (LT-PBR) as a practice of using simple, low unit-cost, structural additions (e.g. wood and beaver dams) to riverscapes to mimic functions and initiate specific processes. Hallmarks of this approach include: - An explicit focus on the processes that a low-tech restoration intervention is meant to promote - A conscious effort to use cost-effective, low-tech treatments (e.g. hand-built, natural materials, non-engineered, short-term design life-spans) because of the need to efficiently scale-up application. - ‘Letting the system do the work’ which defers critical decision making to riverscapes and nature’s ecosystem engineers. Other resources available at: http://lowtechpbr.restoration.usu.edu
Article
Full-text available
Restoration of riparian and wet meadow ecosystems in semi‐arid rangelands of the western U.S. is a high priority given their ecological and hydrological importance in the region. However, traditional restoration approaches are often intensive and costly, limiting the extent over which they can be applied. Practitioners are increasingly trying new restoration techniques that are more cost effective, less intensive, and can more practically scale up to the scope of degradation. Unfortunately, practitioners typically lack resources to undertake outcome‐based evaluations necessary to judge the efficacy of these techniques. In this study, we use freely‐available, satellite remote sensing to explore changes in vegetation productivity (NDVI) of three distinct, low‐tech riparian and wet meadow restoration projects. Case studies are presented that range in geographic location (Colorado, Oregon, and Nevada), restoration practice (Zeedyk structures, beaver dam analogs, and grazing management), and time since implementation. Restoration practices resulted in increased vegetation productivity of up to 25% and increased annual persistence of productive vegetation. Improvements in productivity with time since restoration suggest that elevated resilience may further enhance wildlife habitat and increase forage production. Long‐term, documented outcomes of conservation are rare; we hope our findings empower practitioners to further monitor and explore the use of low‐tech methods for restoration of ecohydrologic processes at meaningful spatial scales. This article is protected by copyright. All rights reserved.
Article
Full-text available
In highly impaired watersheds, it is critical to identify both areas with desirable habitat as conservation zones and impaired areas with the highest likelihood of improvement as restoration zones. We present how detailed riparian vegetation mapping can be used to prioritize conservation and restoration sites within a riparian and instream habitat restoration program targeting 3 native fish species on the San Rafael River, a desert river in southeastern Utah, United States. We classified vegetation using a combination of object-based image analysis (OBIA) on high-resolution (0.5 m), multispectral, satellite imagery with oblique aerial photography and field-based data collection. The OBIA approach is objective, repeatable, and applicable to large areas. The overall accuracy of the classification was 80% (Cohen's κ = 0.77). We used this high-resolution vegetation classification alongside existing data on habitat condition and aquatic species' distributions to identify reaches' conservation value and restoration potential to guide management actions. Specifically, cottonwood (Populus fremontii) and tamarisk (Tamarix ramosissima) density layers helped to establish broad restoration and conservation reach classes. The high-resolution vegetation mapping precisely identified individual cottonwood trees and tamarisk thickets, which were used to determine specific locations for restoration activities such as beaver dam analogue structures in cottonwood restoration areas, or strategic tamarisk removal in high-density tamarisk sites. The site prioritization method presented here is effective for planning large-scale river restoration and is transferable to other desert river systems elsewhere in the world.
Article
Full-text available
Beaver have been referred to as ecosystem engineers because of the large impacts their dam building activities have on the landscape; however, the benefits they may provide to fluvial fish species has been debated. We conducted a watershed-scale experiment to test how increasing beaver dam and colony persistence in a highly degraded incised stream affects the freshwater production of steelhead (Oncorhynchus mykiss). Following the installation of beaver dam analogs (BDAs), we observed significant increases in the density, survival, and production of juvenile steelhead without impacting upstream and downstream migrations. The steelhead response occurred as the quantity and complexity of their habitat increased. This study is the first large-scale experiment to quantify the benefits of beavers and BDAs to a fish population and its habitat. Beaver mediated restoration may be a viable and efficient strategy to recover ecosystem function of previously incised streams and to increase the production of imperiled fish populations.
Article
Full-text available
Biogenic features such as beaver dams, large wood, and live vegetation are essential to the maintenance of complex stream ecosystems, but these features are largely absent from models of how streams change over time. Many streams have incised because of changing climate or land-use practices. Because incised streams provide limited benefits to biota, they are a common focus of restoration efforts. Contemporary models of long-term change in streams are focused primarily on physical characteristics, and most restoration efforts are also focused on manipulating physical rather than ecological processes. We present an alternative view, that stream restoration is an ecosystem process, and suggest that the recovery of incised streams is largely dependent on the interaction of biogenic structures with physical fluvial processes. In particular, we propose that live vegetation and beaver dams or beaver dam analogues can substantially accelerate the recovery of incised streams and can help create and maintain complex fluvial ecosystems.
Article
Strychnine alkaloid baits were consumed by both captive and wild beaver without any apparent hesitation. An approximate minimal acute lethal dose of sodium monofluoro-acetate to beaver of mixed ages and sex was 0.202 mg/kg. Trapping beaver on four study area watersheds in Alabama with No. 330 conibear traps for approximately two weeks in winter during two successive years essentially eliminated beaver. Older individuals were trapped the first year, maturing juveniles and the remaining few adults were trapped the second year, and there was very little reproduction between the trapping periods. Trapping, with its recreational appeal, and income and food potential seems the better and more prudent approach to control of nuisance beaver than others being considered.
Article
Abstract The construction of beaver dams facilitates a suite of hydrologic, hydraulic, geomorphic, and ecological feedbacks that increase stream complexity and channel–floodplain connectivity that benefit aquatic and terrestrial biota. Depending on where beaver build dams within a drainage network, they impact lateral and longitudinal connectivity by introducing roughness elements that fundamentally change the timing, delivery, and storage of water, sediment, nutrients, and organic matter. While the local effects of beaver dams on streams are well understood, broader coverage network models that predict where beaver dams can be built and highlight their impacts on connectivity across diverse drainage networks are lacking. Here we present a capacity model to assess the limits of riverscapes to support dam-building activities by beaver across physiographically diverse landscapes. We estimated dam capacity with freely and nationally-available inputs to evaluate seven lines of evidence: (1) reliable water source, (2) riparian vegetation conducive to foraging and dam building, (3) vegetation within 100 m of edge of stream to support expansion of dam complexes and maintain large colonies, (4) likelihood that channel-spanning dams could be built during low flows, (5) the likelihood that a beaver dam is likely to withstand typical floods, (6) a suitable stream gradient that is neither too low to limit dam density nor too high to preclude the building or persistence of dams, and (7) a suitable river that is not too large to restrict dam building or persistence. Fuzzy inference systems were used to combine these controlling factors in a framework that explicitly also accounts for model uncertainty. The model was run for 40,561 km of streams in Utah, USA, and portions of surrounding states, predicting an overall network capacity of 356,294 dams at an average capacity of 8.8 dams/km. We validated model performance using 2852 observed dams across 1947 km of streams. The model showed excellent agreement with observed dam densities where beaver dams were present. Model performance was spatially coherent and logical, with electivity indices that effectively segregated capacity categories. That is, beaver dams were not found where the model predicted no dams could be supported, beaver avoided segments that were predicted to support rare or occasional densities, and beaver preferentially occupied and built dams in areas predicted to have pervasive dam densities. The resulting spatially explicit reach-scale (250 m long reaches) data identifies where dam-building activity is sustainable, and at what densities dams can occur across a landscape. As such, model outputs can be used to determine where channel–floodplain and wetland connectivity are likely to persist or expand by promoting increases in beaver dam densities.
Article
The US Castor canadensis population has caused severe damage to valuable timberland through dam-building and flooding of bottomland forest. Since beaver populations are mobile, beaver extermination in controlled parcels results in beaver immigration from neighbouring less controlled parcels. This study develops a bioeconomic model that incorporates dispersive population dynamics of beavers into the design of a cost-minimizing trapping strategy. The optimality system for this problem is solved numerically. The validity of the theoretical model is examined using the bioeconomic data collected for the Wildlife Management Regions of the New York State Department of Environmental Conservation. The optimal trapping program causes the initially uneven population to eventually smooth out across the habitat. -from Authors
Article
Through dam building and feeding activities, beaver act as a keystone species to alter hydrology, channel geomorphology, biogeochemical pathways and community productivity. In Quebec, density of dams on the small streams (= or <4th order) studied averages 10.6 dams/km; the streams retain up to 6500 m3 of sediment per dam, and the wetted surface area of the channel is increased up to several hundredfold. Beaver are also active in larger order streams (= or >5th order), but their effects are most noticeable along riverbanks and in floodplains. Comparative carbon budgets per unit area for a riffle on 2nd order Beaver Creek and a beaver pond downstream show the pond receives only 42% of the carbon acquired by the riffle annually, but because the pond has a surface area 7 times greater than the riffle, it receives nearly twice as much carbon as the riffle per unit of channel length. Carbon in the pond has an estimated turnover time of 161 yr compared to 24 yr for the riffle. Beaver ponds are important sites for organic matter processing. -from Authors