DESCRIPTION, COMPARISON, AND PREDICTION OF PACIFIC 
LAMPREY (ENTOSPHENUS TRIDENTATUS) AND WESTERN 
BROOK LAMPREY (LAMPETRA AYRESII) HABITAT IN THE 
SOUTHERN OREGON COASTAL RANGE 
 
 
 
 
 
 
 
 
by 
PARKER JUNG 
 
 
 
 
 
 
 
 
 
 
A THESIS 
Presented to the Department of Biology  
and the Robert D. Clark Honors College  
in partial fulfillment of the requirements for the degree of  
Bachelor of Science 
 
May 2024 
 
 
 
An Abstract of the Thesis of 
Parker Jung for the degree of Bachelor of Science 
in the Department of Biology to be taken June 2024 
 
 
Title:  Description, Comparison, and Prediction of Pacific Lamprey (Entosphenus tridentatus) 
and Western Brook Lamprey (Lampetra ayresii) Habitat in the Southern Oregon Coastal Range 
 
 
 
Approved:  Dr. Richard Emlet  
Primary Thesis Advisor 
 
Lampreys are jawless fish that represent one of the most ancient extant lineages on the 
planet, evolving into their current form over 340 million years ago. Ten of the approximate forty 
extant lamprey species are native to Oregon, making the state’s freshwater ecosystems of special 
interest for lamprey conservation. Lamprey conservation is of great interest due to their 
significance not only in their ecological niche, but in Pacific Northwest indigenous culture. Five 
of Oregon’s lamprey species, including the Pacific lamprey (Entosphenus tridentatus) and the 
western brook lamprey (Lampetra ayresii), are listed by the state as sensitive species. The Pacific 
lamprey and western brook lamprey exhibit significant habitat overlap in Southern Oregon 
coastal streams despite major differences in their size and life histories. This study aimed to 
characterize and compare these species’ habitats using presence/absence data produced by an 
eDNA sampling effort conducted by South Slough National Estuarine Research Reserve 
(SSNERR). Environmental variables considered in this assessment include elevation, slope, 
aspect, hillshade, stream density, forest type, canopy cover, and cost distance. Most variables 
were significant above a 70% confidence interval, indicating that they may be found to be 
significant on a 95% confidence level in studies that work with data differently and remediate the 
caveats associated with this analysis. Variables that were found to be significant above a 95% 
2 
 
 
confidence level for at least one species include elevation, stream density, and cost distance. 
Random forest classifiers used to predict suitable habitat for each species had an accuracy score 
of 60% for the Pacific lamprey and 83% for the western brook lamprey, indicating the greater 
specificity of western brook lamprey habitat. Regardless of caveats that resulted from the nature 
of the data and method of analysis, the findings and predictions produced by this project should 
be useful as a supplementary tool to temporally variable conditions in predicting the presence of 
these species. Understanding where these species are likely to be present will further their 
conservation by informing planning for restoration and construction projects and therefore 
decreasing the likelihood of suitable habitat degradation.  
3 
 
 
Acknowledgements 
 
I would like to thank Dr. Shon Schooler, Dr. Lisa Munger, and Dr. Richard Emlet for 
serving as my thesis committee and helping me throughout this process. I would especially like 
to thank Dr. Schooler for his mentorship during my entire time working with South Slough 
NERR and the extensive experience I gained through the opportunities he presented me with. 
I would also like to thank my father, Paul Jung, who offered his own insight during 
difficult aspects of this analysis. The success of this project would not have been possible 
without his support and collaborative troubleshooting. 
 
  
4 
 
 
Table of Contents 
Introduction 8 
Methods 16 
Results 21 
Elevation 21 
Slope 23 
Aspect 25 
Hillshade 27 
Stream Density 29 
Forest Type 32 
Canopy Cover 36 
Cost Distance 38 
Model 40 
Discussion 43 
Pacific Lamprey Habitat 43 
Western Brook Lamprey Habitat 44 
Comparison 46 
Models 48 
Limitations 49 
Conclusion 51 
Appendix A 52 
Variable Summary Table 52 
Appendix B 53 
Variable Correlation Matrix 53 
Bibliography 54 
Supporting Materials  
Raster (.tif): Stream Density 
Raster (.tif): Cost Distance 
Raster (.tif): Pacific Lamprey Predictions 
Raster (.tif): Western Brook Lamprey Predictions 
  
5 
 
 
List of Figures  
Figure 1. Elevation distribution by species 22 
Figure 2. Slope distribution by species 24 
Figure 3. Aspect distribution by species 26 
Figure 4. Hillshade distribution by species 28 
Figure 5. Stream density distribution by species 30 
Figure 6. Forest type distribution by species 33 
Figure 7. Canopy cover distribution by species 37 
Figure 8. Cost distance distribution by species 39 
Figure 9. Mean decrease in impurity (MDI) plots by species 41 
  
6 
 
 
 
List of Tables  
Table 1. Elevation statistics 21 
Table 2. Elevation t-tests 22 
Table 3. Slope statistics 23 
Table 4. Slope t-tests 24 
Table 5. Aspect statistics 25 
Table 6. Aspect t-tests 26 
Table 7. Hillshade statistics 27 
Table 8. Hillshade t-tests 28 
Table 9. Stream density statistics 29 
Table 10. Stream density t-tests 31 
Table 11. Forest type means 32 
Table 12. Conifer forest type t-tests 33 
Table 13. Mixed forest type t-tests 34 
Table 14. Hardwood forest type t-tests 35 
Table 15. Canopy cover statistics 36 
Table 16. Canopy cover t-tests 37 
Table 17. Cost distance statistics 38 
Table 18. Cost distance t-tests 39 
Table 19. Mean decrease in impurity (MDI) 41 
 
 
 
7 
 
 
 
Introduction 
Lampreys are jawless fish that belong to one of the most ancient extant lineages, evolving 
into their current form about 340 million years ago (Wyodski & Whitney 2003). Oregon 
constitutes crucial lamprey habitat and hosts 10 of the roughly 40 lamprey species distributed 
globally (Clemens & Wade 2023). Five of the ten species inhabiting Oregon are listed by the 
state as Sensitive Species, including the Pacific lamprey (Entosphenus tridentatus) and western 
brook lamprey (Lampetra ayresii) (Clemens et al. 2020). 
The western brook lamprey has historically been referred to as Lampetra richardsoni due 
to inconsistencies between its resident lifestyle and the anadromous tendencies of its satellite 
species, the western river lamprey (Lampetra ayresii). Carim et al. 2023 proposes the 
consolidation of these species on the basis of genetic similarity and classification under 
Lampetra ayresii due to its precedence. The present study uses Lampetra ayresii in agreement 
with Carim et al. 2023. 
The Pacific lamprey and western brook lamprey coexist in southern Oregon coastal 
watersheds and exhibit significant habitat overlap. This habitat overlap may be explained by the 
species similarities in environmental requirements during reproduction and early life stages. Both 
species require deep pockets of sediment to build their redds, or nests, to shelter their eggs from 
the water column. After hatching, ammocoetes, or larvae, remain nearby for several years (E. 
tridentatus for 5-7 yrs, L. ayresii for 4 yrs) before metamorphosing (McPhail 2007). During this 
time, ammocoetes bury themselves in pockets of sediment during the day and extend their bodies 
into the water column at night to filter feed.  
Following metamorphosis, there are few similarities between these species. The western 
brook lamprey is a resident species living the duration of its life in freshwater streams. This 
8 
 
 
species obtains nutrients solely from filter feeding until they are adults, when feeding ceases 
entirely. In contrast, the Pacific lamprey is anadromous and migrates to the Pacific Ocean as a 
juvenile after living in freshwater for 3-7 years (Ostberg et al. 2019). Pacific lampreys live in the 
ocean for 20-40 months where they become parasitic on marine fish and mammals until they are 
ready to reenter freshwater streams and spawn (Wyodski & Whitney 2003; Ostberg et al. 2019; 
Close 2002). Both species die within several months of spawning (Wyodski & Whitney 2003). 
There is also a notable difference in the species’ maximum size, as the Pacific lamprey can reach 
up to 30 in (850 mm) in length while western brook lampreys are rarely longer than 7 in (180 
mm) (Wyodski & Whitney 2003). Marked differences in the sizes and life histories of these two 
species suggest that attention to variables associated with these differences should allow for 
distinction between their preferred habitats.  
Despite the negative connotation lampreys hold for many Americans as a result of sea 
lamprey invasions in the Great Lakes, they typically balance ecosystems and serve many 
invaluable ecological and cultural functions (Close 2002). Lampreys serve as prey for other 
animals throughout all stages of their lives. Freshwater fish, namely rainbow trout and speckled 
dace, have been observed consuming lamprey eggs that overflow from redds (Close 2002). Coho 
salmon are known to prey on ammocoetes as they first emerge from their eggs (Close 2002). 
Ammocoetes and adults are known prey items of many freshwater fish, birds (including great 
blue herons), terrestrial mammals, and pinnipeds (Dunkle et al. 2020; Close 2002).  
Many of the species that prey on lampreys also hunt salmonids. In comparison with 
salmon, lampreys are easier to capture, have higher caloric value, and often migrate in dense 
schools (Close 2002). Collis et al. 2001 found that 71% by volume of four Columbia River 
larids’ (gulls) diets in May was composed of lampreys. Furthermore, Pacific lampreys have been 
9 
 
 
identified as the most common prey item of seals and sea lions in the Pacific Northwest (Roffe & 
Mate 1984). Because seabirds and pinnipeds are primary predators of salmon, their consumption 
of lampreys can act as a buffer against salmonid predation and relieve pressure on these 
populations. Declining lamprey populations and neglected conservation can be equated with 
increased predation of salmonids (Kalan et al. 2023; Wyodski & Whitney 2003).  
Lampreys continue to feed their ecosystems after death, acting as sustenance for many of 
the same predators listed above, but also for smaller fauna not capable of hunting live lamprey. 
The migratory lifestyle of Pacific lampreys aids in the distribution of marine nutrients, an 
important service considering the disparity between the eutrophic NE Pacific and oligotrophic 
freshwater and riparian ecosystems. Influx of subsidies in lotic ecosystems has proven to be an 
important contributor to primary and secondary production (Dunkle et al. 2020). Lampreys only 
absorb 30-40% of the nutrients from food they ingest, meaning they also support the diets of 
smaller animals and meiofauna. Undigested food is passed in a form finer than when it was 
originally ingested, facilitating the uptake of nutrients by smaller suspension and deposit feeders 
(Close 2002). 
The behavior associated with lamprey redd construction establishes them as ecosystem 
engineers. Physical and chemical properties of the stream bed determine the types of species and 
communities that may thrive there. Redd construction requires lampreys to disturb the substrate, 
facilitating the cycling of nutrients, energy, and oxygen between interstitial habitat and the water 
column (Boeker & Geist 2016). These services, in conjunction with the microhabitats formed 
during burrowing, aid in benthic macroinvertebrate survival and reproduction and are important 
in meiofaunal community composition. 
10 
 
 
Kalan et al. (2023) reported on Pacific lampreys’ contributions to maintaining and 
improving water quality. Suspended coliforms, notably Echerichia coli (E. coli), can enter 
streams from agricultural practices, sewage, and defecation by animals and people, and runoff. 
High concentrations of E. coli threaten water quality, biodiversity, and human and ecological 
health. Kalan et al. (2023) showed that the filter feeding behavior of lampreys, specifically 
ammocoetes, significantly decreased concentrations of E. coli in the water column. With this 
behavior, lampreys not only support the health of other native species but make bodies of 
freshwater safe for human uses including water consumption, recreation, and fishing (Kalan et al. 
2023). 
Lampreys also have significance among humans due to their viability as bait for fishing, 
feed for salmonid cultures, and uses among indigenous tribes. Since time immemorial, 
indigenous groups have captured and processed lampreys for food by smoking, sun-drying, and 
salting the fish. It has been reported that they valued lampreys as highly as salmon and that the 
lampreys also had ceremonial and medicinal purposes for tribes (Wyodski & Whitney 2003). In 
addition to acting as a food source, lampreys were also valued for their rich oil which was often 
applied to the body and hair. The channeling and damming of lotic systems has significantly 
decreased the amount of suitable lamprey habitat accessible for fishing. The decline of lamprey 
is associated with a loss of fishing opportunities and thus a loss of culture, as many young 
indigenous people are no longer familiar with the capture, preparation, and use of lampreys 
(Close 2002). 
Flow regulation of rivers not only changes habitats and the amount of flowing water to 
distribute nutrients to suspension feeders, but also establishes barriers to migration. While the 
Pacific lamprey’s larger body size allows it to overcome some barriers by suctioning to flat 
11 
 
 
surfaces, doing so is nearly impossible for smaller lampreys like the western brook lamprey 
(Clemens & Wade 2023; Wyodski & Whitney 2003). Between 1966 and 2001, the number of 
Pacific lampreys recorded at Winchester Dam in the Umpqua River dropped from 46,785 to 34 
(Close 2002). Lampreys’ reliance on benthic habitat also makes them incredibly vulnerable to 
any changes in hydrogeomorphic features and general habitat degradation (Boeker & Geist 
2016). Regardless of population declines and loss of habitat, lamprey conservation has 
historically received little attention and the western brook and Pacific lampreys were both denied 
for inclusion in the Endangered Species Act in 2004 (Carim et al. 2023). 
One of the most common methods used in lamprey studies is electrofishing which 
requires researchers to use equipment to influence lampreys to leave the sediment where they 
burrow, capture and identify them, and record their metrics. Although this was the standard study 
method in the past, its requirement for specialized equipment, researchers trained in 
identification, and disturbing the species and their environments makes it impractical. Even with 
specialized equipment and team members, errors in taxonomic identification are possible and 
habitats can be hard to reach (Balasingham et al. 2017). These issues are particularly relevant to 
this project because aside from their adult size, the Pacific lamprey and western brook lamprey 
can be exceedingly difficult to identify due to the morphological simplicity of family 
Petromyzontidae. This becomes an issue when attempting to distinguish between the species’ 
ammocoetes, as they are of similar size and only differ phenotypically by the pigmentation in 
their caudal fins. 
The sampling effort that provided the data used in this analysis tested a genetic barcoding 
method that uses environmental DNA (eDNA) collected from streams to determine species 
presence (Schooler et al. 2022). This procedure stands to resolve the issues presented by 
12 
 
 
common study methods such as electrofishing. eDNA sampling is a new method of species 
detection that has been rapidly growing in the field of macrobial aquatic species observation over 
the last decade (Carim et al. 2016). The process of eDNA sampling requires researchers to pass 
stream water through filters, then test for the presence of DNA from the target species. eDNA 
can come from many sources including shed skin cells or exoskeletons, urine, feces, sperm, eggs, 
mucus, and decaying tissues (Balasingham et al. 2017; Seymour et al. 2018). In addition to 
resolving the issues of false identification, habitat accessibility, and the complications of 
capturing individuals, eDNA methods are favorable because of their sensitivity, reliability, and 
efficiency. Each sample requires 15 minutes or less to collect, significantly decreasing sampling 
time and cost (Carim et al. 2016; Balasingham et al. 2017). eDNA is detectable for hours to days 
after the source is removed and can be detected in instances where other methods may 
incorrectly deduce absence. This short persistence time frame offers an additional positive of the 
method: near real-time monitoring (Seymour et al. 2018). 
While results from eDNA can be quantitative to give some indication of population 
density or distance from the sample site, the eDNA results used in this analysis simply indicate 
presence or absence (Balasingham 2017; Seymour et al. 2018). It is also worth noting that most 
past eDNA studies have been undertaken in lentic ecosystems (Seymour et al. 2018). There are 
many factors to consider when trying to understand spatial patterns via eDNA sampling in lotic 
ecosystems. As lentic ecosystems experience very low levels of flow, their conditions are more 
static and have a fairly low rate of change. The flow rate, pH, and temperature of lotic 
ecosystems, on the other hand, can change drastically day by day depending on the season, 
weather, and organic input (Seymour et al. 2018). These factors all influence eDNA persistence 
and detection and thus, understanding the meaning of eDNA results from these systems can be 
13 
 
 
difficult. It is also important to realize that sampling a flowing system entails sampling upstream 
habitat as well as the sample location. For example, in a study of a river in eastern Canada, 
Balasingham et al. (2017) found that residual eDNA was still detected 11.5 hours after the source 
was removed at the source site and the residual eDNA signal strength decreased as sample 
distance increased downstream. In a meta-analysis of eDNA downstream transport Jo and 
Yamanaka (2022) found that most eDNA particles travel less than 2km under ordinary 
hydrogeographic conditions.   
Due to the difficulties associated with elucidating the extent of upstream habitat that is 
included in an eDNA sample, especially when using presence/absence rather than quantitative 
data, the GIS analysis used in this study treats each sample as a discrete sample site. It is 
acknowledged that the specific sample site identified as suitable based on a positive PCR may 
not actually contain lampreys, but rather be located downstream from the detected suitable 
habitat. Therefore, this caveat should be considered when assessing the results of the analysis. 
While this may produce misleading results for variables like aspect, general trends in elevation, 
for example, should still be accurate. For some variables, such as elevation limits, the properties 
of eDNA sampling are a benefit because the sample integrates over a much longer stream 
distance than a discrete sample would, so determinations of absence above the sample location 
are more robust.  
This study aimed to analyze presence data collected using eDNA methods to elucidate 
the specific environmental variables that define Pacific lamprey and western brook lamprey 
habitats. Lampreys face many threats, many of which (e.g. habitat degradation and 
simplification, loss of riparian cover, artificial barriers) are a direct result of human activities 
(Clemens & Wade 2023). Lampreys serve countless ecological and cultural functions, so proper 
14 
 
 
assessment of their presence and threats is crucial to informing management and restoration 
practices, as well as construction projects (Schooler et al. 2022). Understanding defining habitat 
characteristics would allow for quick assessment of risk to lampreys before undergoing invasive 
projects. 
The products of this analysis should be useful in understanding the likelihood of lamprey 
presence in a given stream based on variables that prove significant. While the predictions made 
by the culminating/culminant/cumulative models cannot stand on their own due to exclusion of 
temporally variable factors like flow and temperature, they should be informative preliminary 
tools for determining whether presence of these species is likely or at all possible. 
 
15 
 
 
Methods 
The dataset used in this analysis was produced by a sampling effort conducted by South 
Slough National Estuarine Research Reserve (SSNERR) throughout the southern Oregon Coastal 
Range. Sampling occurred at 71 sites between May and December of 2019, 52 sites between 
June and August of 2021, and 39 sites between May and September of 2022. Nine of the total 
162 samples were taken in lakes to test the accuracy of eDNA sampling in lentic ecosystems to 
detect species presence in the watershed. Because lakes with input streams positive for lamprey 
DNA did not test positive for lamprey DNA, it was concluded that sampling lakes was not 
accurate for this type of study. Thus, the nine lake samples were excluded from the analyzed 
dataset. Four additional samples represented experimental sampling (drainage ditches, for 
example), and were not used in the final analysis. Therefore, 149 total data points were used to 
assess the species’ preferred habitat. 
At each site, 5 L of water was pumped through a filter which was then kept separate from 
filters from other sites to avoid contamination (using USFS methods, Carim et al. 2016). The 
filters were dried and sent to the USFS Genomics Lab for sequencing to determine which, if 
either, of the two species was present (Carim et al. 2016). It is important to recognize that eDNA 
analysis cannot currently distinguish between Lampetra species, so while Pacific lamprey DNA 
is a sure indicator of Pacific lamprey presence, detected Lampetra DNA could indicate the 
presence of any Lampetra species. Despite this caveat, this analysis assumes that detected 
Lampetra presence refers to the western brook lamprey, as no other Lampetra species have been 
found in the southern Oregon coastal watershed. 
Elevation, slope, aspect, hillshade, stream density, forest type, canopy cover, and cost 
distance served as explanatory variables for defining preferred lamprey habitat.  
16 
 
 
The first seven variables will help define the general topography and ecology of the 
habitat. Slope is measured as the angle above horizontal, ranging from 0 to 90 degrees. Aspect 
describes the direction of the downhill slope and is measured in degrees. This measurement 
begins with 0 degrees at north and continues clockwise back to north, meaning east, south, and 
west would be located at 90, 180, and 270 degrees, respectively. Hillshade is a terrain surface 
shaded based on the sun’s relative position where raster cells are given a value between 0 (no 
shade) and 255 (complete shade) shade units. This analysis uses a traditional, as opposed to 
multi-directional hillshade, with an azimuth (the sun’s relative position in degrees, measured in 
the same manner as aspect) of 315 degrees and an altitude of 45 degrees above the horizon. Cost 
distance measures how difficult it is to travel a distance by assigning a cost to cross each raster 
cell based on slope. Cost distance is measured in units of cost. This measurement should also 
help distinguish between the species because the Pacific lamprey must travel upstream between 
each generation while the western brook lamprey may have a harder time traveling upstream due 
to its smaller body size.  
All data layers were projected to the Oregon-Lambert conformal conic projection, a 
projected coordinate system designed for spatial data analysis conducted in the state of Oregon. 
This coordinate system uses feet as a unit, so all layers with units of length use feet by default. 
Metric conversions are provided in the text. A cell size of 32.808 ft (10 m) was maintained for all 
rasters. 
Raster layers for elevation, hillshade, slope, and aspect were derived from a 10 m Digital 
Elevation Model (DEM) obtained from Oregon Spatial Data Library using the associated tools in 
ArcGIS Pro (2.8.0). These layers were then clipped to the extent of the Oregon portion of the 
Oregon-Washington Coastal zone polygon contained in the Oregon Watersheds layer from 
17 
 
 
Oregon Spatial Data Library. The resulting layers were used as snap rasters and clipping extents 
for the remaining layers to ensure the data was of the same resolution and comparable for 
subsequent analysis. 
 Raster data for forest type and percent canopy cover were acquired from Oregon 
State University’s Landscape Ecology, Modeling, Mapping and Analysis (LEMMA) research 
group. The raster layer for forest type had values ranging from 0 to 100 with metadata 
categorizing 0 - >30 as conifer forest, ≤30 - 70 as mixed forest, and ≤70 as hardwood forest. 
These forest types were separated into three raster layers where cells of the given forest type had 
a value of 1 and all other cells had a value of 0. The extent of these layers did not align with that 
of the Oregon Spatial Data Library watershed polygon, so the final study area was determined by 
the overlap of the two. 
 The layers for stream density and cost distance originated from the National 
Hydrography Dataset (NHD) Flowline dataset. This dataset provided high-definition stream 
polylines that were essential to characterize habitat in the smaller creeks that were sampled. To 
calculate layers for stream density and cost distance, all stream polylines were given the same 
value before they were used as an input layer. Stream density was generated by calculating the 
polyline density of this layer in ArcGIS Pro. Due to the scale of this analysis, a search radius of 
just 5280 ft (1 mile) was used to calculate density. Before cost distance was run, the NHD vector 
streams were rasterized and a coastline polyline from the flowline layer was buffered by 30 ft to 
intersect stream polylines that didn’t already meet the coastline. The final cost distance layer was 
created using the ArcGIS Pro Cost Distance tool using the rasterized NHD streams as the input 
cost raster and the buffered coastline as the input feature source data (Travel direction: 
FROM_SOURCE). 
18 
 
 
 A spatial join was performed to connect the sample sites with their associated 
values in the raster layers of the explanatory variables. The significance of the variables and their 
mean values in relation to species presence were assessed via descriptive statistics and t tests. T 
tests were run in four ways: 1) to assess the significance of variables in determining presence 
versus absence of Lampetra ayresii, 2) to assess the significance of variables in determining 
presence versus absence of Entosphenus tridentatus, 3) to assess the significance of variables in 
determining presence Lampetra ayresii versus presence of Entosphenus tridentatus, and 4) to 
assess the significance of variables in determining absence Lampetra ayresii versus absence of 
Entosphenus tridentatus.  The first two sets of t tests aimed to define each species’ preferred 
habitat and deduce which variables are relevant in this definition. The last two sets of t tests were 
geared towards comparing the habitats of the two species to describe the differences between 
their preferred habitats and the habitats they do not occupy. 
While only relationships proven significant on a 95% confidence level will be regarded 
as truly significant, confidence levels down to 70% are reported. The caveats involved in this 
analysis, including the nature of eDNA sampling and slight errors that may be associated with 
geographic location in high resolution data may influence results. Thus, reporting results at lower 
confidence intervals will provide context to future studies that may find these variables 
significant. Furthermore, these confidence levels are in many cases related to the relative 
importance of variables in influencing the models, and thus provide an opportunity for further 
analysis in comparing these two metrics. 
The effectiveness of these data and explanatory variables in predicting lamprey presence 
was then tested by training a machine learning model. The data were used to train a logistic 
regression, decision tree classifier, and random forest classifier (300 estimators) for each species 
19 
 
 
which were then vetted based on accuracy score. The random forest classifier proved most 
accurate in predicting western brook lamprey habitat. While the decision tree classifier (63%) 
was more accurate than the random forest classifier (60%) in predicting Pacific lamprey habitat, 
the difference was small, and the random forest classifier was ultimately used for both species 
for the purpose of consistency. These models were used to write predicted suitability raster 
layers within the study area for each species. 
20 
 
 
Results 
Across the 149 total samples, 80 indicated presence of the western brook lamprey and 92 
indicated presence of the Pacific lamprey. 59 samples reflected the presence of both species and 
36 samples did not contain DNA from either species. 
Elevation  
 
Table 1. Elevation statistics  
Mean, minimum, and maximum elevations at sample sites where Pacific lampreys (PL) and 
western brook lampreys (WBL) were present (P) and absent (A). 
Sample sites ranged in elevation from 1.19 ft to 2248.14 ft (0.36 to 685.23 m)(Table 1). 
Pacific lampreys were present between 1.19 ft and 1219.34 ft (0.36 to 371.65 m) (Fig. 1a). The 
mean elevation of sample sites where Pacific lampreys were present was 121.68 ± 235.32 ft 
(37.09 ± 71.73 m). The elevations at which Pacific lampreys were absent ranged from 3.22 ft to 
2248.14 ft (0.98 to 685.23 m) with a mean of 348.12 ± 573.71 ft (106.11 ±174.87 m). 
Western brook lampreys were present between 1.19 and 438.16 ft (0.36 to 133.55 m)(Fig 
1b). The mean elevation of sample sites where western brook lampreys were present was 47.00 ± 
61.79 ft (14.33±18.83 m). The elevations at which western brook lampreys were absent also 
ranged from 3.22 ft to 2248.14 ft and had a mean of 395.33 ± 548.79 ft (120.50 ±167.27 m). 
21 
 
 
a. b. 
   
Figure 1. Elevation distribution by species 
Distribution of elevations at sample sites where a. Pacific lampreys (PL) and b. western brook 
lampreys (WBL) were present and not present. 
 
Table 2. Elevation t-tests 
Degrees of freedom (df), t-statistic, p-value, and confidence level of t-tests run to determine the 
role of elevation in differentiating between habitats where Pacific lampreys are present and absent 
(PL P/A), western brook lampreys are present and absent (WBL P/A), Pacific lampreys and 
western brook lampreys are present (PL/WBL P), and Pacific lampreys and western brook 
lampreys are absent (PL/WBL A). 
The t test comparing the mean elevations where Pacific lampreys were present and absent 
found the elevations at which the species were present to be significantly lower than those at 
which they were absent at a 95% confidence level; t=-2.84, df=147, p=0.006 (Table 2). The t test 
comparing the mean elevations where western brook lampreys were present and absent also 
22 
 
 
found the elevations at which the species present to be significantly lower than those at which 
they were absent at a 95% confidence level; t=-5.24, df=147, p=1.61E-06. 
 The t test comparing the mean elevation where Pacific lampreys were present and 
the mean elevation where western brook lampreys were present indicated the mean elevation at 
which Pacific lampreys were present to be significantly higher than the mean elevation at which 
western brook lampreys were present at a 95% confidence level; t=2.93, df=170, p=0.004. The t 
test comparing the elevations where the species were absent did not find a significant difference 
between the means; t=-0.47, df=124, p=0.64. 
Slope 
 
Table 3. Slope statistics  
Mean, minimum, and maximum slopes at sample sites where Pacific lampreys (PL) and western 
brook lampreys (WBL) were present (P) and absent (A). 
The slope at sample sites ranged from 0.02 to 43.17 (Table 3). Pacific lampreys were 
present at slopes between 0.02 and 23.72 (Fig 2a). The mean slope at sample sites where Pacific 
lampreys were present was 4.96 ± 4.99. The slopes at which Pacific lampreys were absent ranged 
from 0.13 to 43.17 with a mean of 6.69 ± 8.54. 
 Western brook lampreys were present at slopes between 0.02 and 23.72 (Fig 2b). The 
mean slope at sample sites where western brook lampreys were present was 4.61 ± 4.77. The 
23 
 
 
slopes at which western brook lampreys were absent ranged from 0.13 to 43.17 and had a mean 
of 6.80 ± 8.12. 
a. b. 
   
Figure 2. Slope distribution by species 
Distribution of slopes at sample sites where a. Pacific lampreys (PL) and b. western brook 
lampreys (WBL) were present and not present. 
 
Table 4. Slope t-tests 
Degrees of freedom (df), t-statistic, p-value, and confidence level of t-tests run to determine the 
role of slope in differentiating between habitats where Pacific lampreys are present and absent (PL 
P/A), western brook lampreys are present and absent (WBL P/A), Pacific lampreys and western 
brook lampreys are present (PL/WBL P), and Pacific lampreys and western brook lampreys are 
absent (PL/WBL A). 
The t test comparing the mean slopes where Pacific lampreys were present and absent 
found the slopes where the species were present to be significantly smaller than those at which 
24 
 
 
they were absent on an 80% confidence level; t=-1.39, df=147, p=0.17 (Table 4). The t test 
comparing the mean slopes where western brook lampreys were present and absent found the 
slopes at which the species were present to be significantly smaller than those at which they were 
absent at a 90% confidence level; t=-1.97, df=147, p=0.051. 
 The t test comparing the mean slope at sites where Pacific lampreys were present 
and the mean slope where western brook lampreys were present did not indicate a significant 
difference between the means; t=0.47, df=170, p=0.64. The t test comparing the slopes where the 
species were absent also did not find a significant difference between the means; t=-0.07, 
df=124, p=0.94. 
Aspect 
 
Table 5. Aspect statistics  
Mean, minimum, and maximum aspects at sample sites where Pacific lampreys (PL) and western 
brook lampreys (WBL) were present (P) and absent (A). 
The aspect of sample sites ranged from 1 to 357 (Table 5). Pacific lampreys were present 
at aspects between 14 and 357 (Fig 3a). The mean aspect at sample sites where Pacific lampreys 
were present was 146.50 ± 94.25. The aspects at which Pacific lampreys were absent ranged 
from 1 to 347 with a mean of 147.16 ± 78.47. 
25 
 
 
 Western brook lampreys were present at aspects between 2 and 347 (Fig 3b). The mean 
aspect at sample sites where western brook lampreys were present was 134.88 ± 81.50. The 
aspects at which western brook lampreys were absent ranged from 1 to 357 and had a mean of 
160.52 ± 94.25. 
 
a. b. 
   
Figure 3. Aspect distribution by species 
Distribution of aspects at sample sites where a. Pacific lampreys (PL) and b. western brook 
lampreys (WBL) were present and not present. 
 
Table 6. Aspect t-tests 
Degrees of freedom (df), t-statistic, p-value, and confidence level of t-tests run to determine the 
role of aspect in differentiating between habitats where Pacific lampreys are present and absent 
(PL P/A), western brook lampreys are present and absent (WBL P/A), Pacific lampreys and 
western brook lampreys are present (PL/WBL P), and Pacific lampreys and western brook 
lampreys are absent (PL/WBL A). 
26 
 
 
The t test comparing the mean aspects where Pacific lampreys were present and absent 
did not find a significant difference between the means; t=-0.05, df=147, p=0.96 (Table 6). The t 
test comparing the mean aspects where western brook lampreys were present and absent found 
the aspects at which the species were present to be significantly smaller than those at which they 
were absent at a 90% confidence level; t=-1.76, df=147, p=0.08. 
The t test comparing the mean aspect at sites where Pacific lampreys were present and 
the mean aspect where western brook lampreys were present did not indicate a significant 
difference between the means; t=0.87, df=170, p=0.39. The t test comparing aspect where the 
species were absent also did not find a significant difference between the means; t=-0.87, 
df=124, p=0.39. 
Hillshade 
 
Table 7. Hillshade statistics  
Mean, minimum, and maximum hillshades at sample sites where Pacific lampreys (PL) and 
western brook lampreys (WBL) were present (P) and absent (A). 
The hillshade at sample sites ranged from 113 to 237 (Table 7). Pacific lampreys were 
present at hillshades between 129 and 237 (Fig 4a). The mean hillshade at sample sites where 
Pacific lampreys were present was 216.74 ± 19.07. The hillshades at which Pacific lampreys 
were absent ranged from 113 to 236 with a mean of 210.86 ± 28.40. 
27 
 
 
 Western brook lampreys were present at hillshades between 129 and 232 (Fig 4b). The 
mean hillshade at sample sites where western brook lampreys were present was 216.58 ± 18.36. 
The hillshades where western brook lampreys were absent ranged from 113 to 237 and had a 
mean of 212.07 ± 27.68. 
a. b. 
 
Figure 4. Hillshade distribution by species 
Distribution of hillshades at sample sites where a. Pacific lampreys (PL) and b. western brook 
lampreys (WBL) were present and not present. 
 
Table 8. Hillshade t-tests 
Degrees of freedom (df), t-statistic, p-value, and confidence level of t-tests run to determine the 
role of hillshade in differentiating between habitats where Pacific lampreys are present and absent 
(PL P/A), western brook lampreys are present and absent (WBL P/A), Pacific lampreys and 
western brook lampreys are present (PL/WBL P), and Pacific lampreys and western brook 
lampreys are absent (PL/WBL A). 
28 
 
 
 
The t test comparing the mean hillshades where Pacific lampreys were present and absent 
found the mean hillshade where they were present to be significantly greater than where they 
were absent on an 80% confidence level; t=1.38, df=147, p=0.17 (Table 8). The t test comparing 
the mean aspects where western brook lampreys were present and absent found the hillshades 
where the species was present to be significantly greater than those at which they were absent at 
a 70% confidence level; t=1.15, df=147, p=0.25. 
 The t test comparing the mean aspect at sites where Pacific lampreys were present 
and the mean hillshade where western brook lampreys were present did not indicate a significant 
difference between the means; t=0.06, df=170, p=0.95. The t test comparing hillshade where the 
species were absent also did not find a significant difference between the means; t=-0.24, 
df=124, p=0.81. 
Stream Density 
 
Table 9. Stream density statistics  
Mean, minimum, and maximum stream densities at sample sites where Pacific lampreys (PL) and 
western brook lampreys (WBL) were present (P) and absent (A). 
Stream density ranged from 0.000159 streams/ft2 to 0.001890 streams/ft2 (0.001711 
streams/m2 to 0.02034 streams/m2) across all sample sites (Table 9). Pacific lampreys were 
present at stream densities between 0.000159 streams/ft2 and 0.001890 streams/ft2 (0.001711 
29 
 
 
streams/m2 to 0.02034 streams/m2)(Figure 5a). The mean stream density at sample sites where 
Pacific lampreys were present was 0.001072 ± 0.000355 streams/ft2 (0.01154 ± 0.003821 
streams/m2). The stream densities at which Pacific lampreys were absent ranged from 0.000159 
streams/ft2 to 0.001724 streams/ft2 (0.001711 streams/m2 to 0.01856 streams/m2) with a mean of 
0.001038 ± 0.000333 streams/ft2 (0.01117 ± 0.003584 streams/m2). 
Western brook lampreys were present at stream densities between 0.000350 streams/ft2 
and 0.001890 streams/ft2 (0.003767 streams/m2 to 0.02034 streams/m2)(Figure 5b). The mean 
stream density at sample sites where western brook lampreys were present was 0.001130 ± 
0.000318 streams/ft2 (0.01216 ± 0.003423 streams/m2). The stream densities where western 
brook lampreys were absent ranged from 0.000159 streams/ft2 to 0.001759 streams/ft2 (0.001711 
to 0.01893 streams/m2) and had a mean of 0.000977 ± 0.000361 streams/ft2 (0.01052 ± 0.003996 
streams/m2). 
a. b. 
 
Figure 5. Stream density distribution by species 
Distribution of stream densities at sample sites where a. Pacific lampreys (PL) and b. western 
brook lampreys (WBL) were present and not present. 
 
30 
 
 
 
Table 10. Stream density t-tests 
Degrees of freedom (df), t-statistic, p-value, and confidence level of t-tests run to determine the 
role of stream density in differentiating between habitats where Pacific lampreys are present and 
absent (PL P/A), western brook lampreys are present and absent (WBL P/A), Pacific lampreys and 
western brook lampreys are present (PL/WBL P), and Pacific lampreys and western brook 
lampreys are absent (PL/WBL A). 
The t test comparing the mean stream density where Pacific lampreys were present and 
absent did not find a significant difference between the means; t=0.58, df=147, p=0.56 (Table 
10). The t test comparing mean stream densities where western brook lampreys were present and 
absent found the stream densities where the species was present to be significantly greater than 
those at which they were absent at a 95% confidence level; t=2.72, df=147, p=0.007. 
 The t test comparing the mean stream densities at sites where each species was 
present found that western brook lampreys were present at a significantly higher mean stream 
density at a 70% confidence level; t=-1.13, df=170, p=0.26. The t test comparing stream 
densities where the species were absent did not find a significant difference between the means; 
t=0.99, df=124, p=0.32. 
31 
 
 
Forest Type 
 
Table 11. Forest type means 
Mean of each forest type (classified 1 as the target forest type, 0 as any other forest type) at sample 
sites where Pacific lampreys (PL) and western brook lampreys (WBL) were present (P) and absent 
(A). 
 The mean values of each forest type indicated the percentage of samples that fell under 
that category. Conifer forest had a mean value of 0.413 (41.3%) where Pacific lampreys were 
present, 0.421 (42.1%) where Pacific lampreys were absent, 0.463 (46.3%) where western brook 
lampreys were present, and 0.362 (36.2%) where western brook lampreys were absent (Table 11; 
Fig. 6). Mixed forest had a mean value of 0.261 (26.1%) where Pacific lampreys were present, 
0.386 (38.6%) where Pacific lampreys were absent, 0.225 (22.5%) where western brook 
lampreys were present, and 0.406 (40.6%) where western brook lampreys were absent. 
Hardwood forest had a mean value of 0.326 (32.6%) where Pacific lampreys were present, 0.193 
(19.3%) where Pacific lampreys were absent, 0.313 (31.3%) where western brook lampreys were 
present, and 0.232 (23.2%) where western brook lampreys were absent. 
32 
 
 
 
a. b.  
Figure 6. Forest type distribution by species 
Distribution of forest type at sample sites where a. Pacific lampreys (PL) and b. western brook 
lampreys (WBL) were present and not present. 
 
Table 12. Conifer forest type t-tests  
Degrees of freedom (df), t-statistic, p-value, and confidence level of t-tests run to determine the 
role of conifer forest type in differentiating between habitats where Pacific lampreys are present 
and absent (PL P/A), western brook lampreys are present and absent (WBL P/A), Pacific lampreys 
and western brook lampreys are present (PL/WBL P), and Pacific lampreys and western brook 
lampreys are absent (PL/WBL A). 
The t test did not find that a habitat being in a conifer dominant forest was significant in 
determining the presence of the Pacific lamprey; t=-0.10, df=147, p=0.92 (Table 12). The t test 
assessing the role of conifer forest type in determining western brook lamprey presence found 
that this forest type was significant on a 70% confidence level; t=1.24, df=147, p=0.22. Conifer 
33 
 
 
forest was not significant in differentiating between habitats where the two species were present 
(t=-0.65, df=170, p=0.52) or where the two species were absent (t=0.67, df=124, p=0.51). 
The t test found that a habitat being located in a mixed forest was significant in 
determining the presence of the Pacific lamprey on an 80% confidence level; t=-1.57, df=147, 
p=0.12 (Table 13). The t test assessing the role of mixed forest type in determining western 
brook lamprey presence found that this forest type was significant on a 95% confidence level; t=-
2.38, df=147, p=0.02. Mixed forest type was not significant in differentiating between habitats 
where the two species were present (t=0.55, df=170, p=0.59) or where the two species were 
absent (t=-0.22, df=124, p=0.82).  
 
Table 13. Mixed forest type t-tests  
Degrees of freedom (df), t-statistic, p-value, and confidence level of t-tests run to determine the 
role of mixed forest type in differentiating between habitats where Pacific lampreys are present 
and absent (PL P/A), western brook lampreys are present and absent (WBL P/A), Pacific lampreys 
and western brook lampreys are present (PL/WBL P), and Pacific lampreys and western brook 
lampreys are absent (PL/WBL A). 
34 
 
 
 
Table 14. Hardwood forest type t-tests  
Degrees of freedom (df), t-statistic, p-value, and confidence level of t-tests run to determine the 
role of hardwood forest type in differentiating between habitats where Pacific lampreys are present 
and absent (PL P/A), western brook lampreys are present and absent (WBL P/A), Pacific lampreys 
and western brook lampreys are present (PL/WBL P), and Pacific lampreys and western brook 
lampreys are absent (PL/WBL A). 
The t test found that a habitat being located in a hardwood dominant forest was 
significant in determining the presence of the Pacific lamprey on a 90% confidence level; t=1.85, 
df=147, p=0.07 (Table 14). The t test assessing the role of hardwood forest in determining 
western brook lamprey presence found that this forest type was significant on a 70% confidence 
level; t=1.10, df=147, p=0.27. Mixed forest type was not significant in differentiating between 
habitats where the two species were present (t=0.19, df=170, p=0.85) or where the two species 
were absent (t=-0.53, df=124, p=0.60). 
35 
 
 
Canopy Cover 
 
Table 15. Canopy cover statistics  
Mean, minimum, and maximum canopy covers at sample sites where Pacific lampreys (PL) and 
western brook lampreys (WBL) were present (P) and absent (A). 
Canopy cover ranged from 0% to 89.19% across all sample sites (Table 15). Pacific 
lampreys were present between canopy covers of 0% and 89.19% (Fig. 7a). The mean canopy 
cover at sample sites where Pacific lampreys were present was 35.47 ± 30.40%. The canopy 
covers at which Pacific lampreys were absent ranged from 0% to 85.31% with a mean of 39.44 ± 
31.03%. 
 Western brook lampreys were present at canopy covers between 0% and 86.19% (Fig. 
7b). The mean canopy cover at sample sites where western brook lampreys were present was 
33.46 ± 30.13%. Canopy cover where western brook lampreys were absent ranged from 0 to 
89.19% and had a mean of 41.09 ± 30.86%. 
 
36 
 
 
 
a. b. 
 
Figure 7. Canopy cover distribution by species 
Distribution of canopy cover at sample sites where a. Pacific lampreys (PL) and b. western brook 
lampreys (WBL) were present and not present. 
 
Table 16. Canopy cover t-tests  
Degrees of freedom (df), t-statistic, p-value, and confidence level of t-tests run to determine the 
role of canopy cover in differentiating between habitats where Pacific lampreys are present and 
absent (PL P/A), western brook lampreys are present and absent (WBL P/A), Pacific lampreys and 
western brook lampreys are present (PL/WBL P), and Pacific lampreys and western brook 
lampreys are absent (PL/WBL A). 
The t test comparing the mean canopy cover where Pacific lampreys were present and 
absent did not find a significant difference between the means; t=-0.76, df=147, p=0.45 (Table 
16). The t test comparing mean canopy cover where western brook lampreys were present and 
37 
 
 
absent found the canopy covers where the species was present to be significantly less than those 
at which they were absent at a 80% confidence level; t=-1.52, df=147, p=0.13. 
 The t test comparing the mean canopy covers at sites where each species was 
present did not find a significant difference between the means; t=0.44, df=170, p=0.66. The t 
test comparing canopy covers where the species were absent did not find a significant difference 
between the means either; t=-0.30, df=124, p=0.77. 
Cost Distance 
 
Table 17. Cost distance statistics  
Mean, minimum, and maximum cost distances at sample sites where Pacific lampreys (PL) and 
western brook lampreys (WBL) were present (P) and absent (A). 
Cost distance is a measure of the cost of traveling a given distance by considering slope 
and elevation gain. The value of each cell in cost units is calculated based on the cell’s length 
and the cost or energy required to cross it. For example, a cell that is 10 m wide and has a cost 
value of 10 would have a final value of 100 in a cost distance raster. 
 Cost distance ranged from 1.00x105 to 6.13x108 across all sample sites (Table 17). 
Pacific lampreys were present at cost distances between 1.00x105 to 2.51x108 (Fig 8a). The mean 
cost distance at sample sites where Pacific lampreys were present was 6.01x107 ± 6.11x107. The 
cost distances at which Pacific lampreys were absent ranged from 1.88x105 to 6.13x108 with a 
mean of 1.04x108 ± 1.21x108. 
38 
 
 
 Western brook lampreys were present at cost distances between 1.00x105 and 6.13x108 
(Fig 8b). The mean cost distance at sample sites where western brook lampreys were present was 
4.95x107 ± 4.05x107. The cost distances where western brook lampreys were absent ranged from 
1.00x105 to 6.13x108 and had a mean of 1.09x108 to 1.19x108. 
a. b. 
 
Figure 8. Cost distance distribution by species 
Distribution of cost distance at sample sites where a. Pacific lampreys (PL) and b. western brook 
lampreys (WBL) were present and not present. 
 
Table 18. Cost distance t-tests  
Degrees of freedom (df), t-statistic, p-value, and confidence level of t-tests run to determine the 
role of cost distance in differentiating between habitats where Pacific lampreys are present and 
absent (PL P/A), western brook lampreys are present and absent (WBL P/A), Pacific lampreys and 
western brook lampreys are present (PL/WBL P), and Pacific lampreys and western brook 
lampreys are absent (PL/WBL A). 
39 
 
 
The t test comparing the mean cost distance where Pacific lampreys were present and 
absent found that the mean cost distance where the species was present to be significantly less 
than the mean where it was absent on a 95% confidence level; t=-2.54, df=147, p=0.01 (Table 
18). The t test comparing mean cost distances where western brook lampreys were present and 
absent found the mean cost distance where the species was present to be significantly less than 
those at which they were absent at a 95% confidence level; t=-3.92, df=147, p=0.0002. 
 The t test comparing the mean cost distances at sites where each species was 
present found that Pacific lampreys were present at a significantly higher mean cost distance on 
an 80% confidence level; t=1.35, df=170, p=0.18. The t test comparing cost distances where the 
species were absent did not find a significant difference between the means; t=-0.21, df=124, 
p=0.83. 
Models 
 The logistic regression model for the Pacific lamprey had an accuracy of 43%, the 
decision tree classifier had an accuracy of 63% and the random forest classifier had an accuracy 
of 60%. The logistic regression model for the western brook lamprey had an accuracy of 47%, 
the decision tree classifier had an accuracy of 77%, and the random forest classifier had an 
accuracy of 83%. 
The random forest classifier for the Pacific lamprey data recalled 31% of the absences 
correctly and 82% of the presences correctly. The random forest classifier for the western brook 
lamprey data recalled 73% of the absences correctly and 89% of the presences correctly. 
 
40 
 
 
 
Figure 9. Mean decrease in impurity (MDI) plots by species 
Mean decrease in impurity (MDI) due to each variable for the a. Pacific lamprey and b. western 
brook lamprey habitat random forest classifiers 
 
Table 19. Mean decrease in impurity (MDI) 
Mean decrease in impurity (MDI) values by variable  
In the Pacific lamprey model the MDI due to elevation was 0.19, the MDI due to slope 
was 0.15, the MDI due to aspect was 0.12, the MDI due to hillshade was 0.10, the MDI due to 
stream density was 0.11, the MDI due to conifer forest was 0.02, the MDI due to mixed forest 
was 0.03, the MDI due to hardwood forest was 0.01, the MDI due to canopy cover was 0.13, and 
the MDI due to cost distance was 0.15 (Fig. 9a; Table 19).  
41 
 
 
In the western brook lamprey model, the mean decrease in impurity (MDI) due to 
elevation was 0.24, the MDI due to slope was 0.10, the MDI due to aspect was 0.10, the MDI 
due to hillshade was 0.07, the MDI due to stream density was 0.16, the MDI due to conifer forest 
was 0.01, the MDI due to mixed forest was 0.01, the MDI due to hardwood forest was 0.01, the 
MDI due to canopy cover was 0.07, and the MDI due to cost distance was 0.24 (Fig. 9b; Table 
19). 
42 
 
 
Discussion 
Pacific Lamprey Habitat 
Elevation and cost distance were the only variables significant in determining Pacific 
lamprey presence at or above a 95% confidence level. Slope, hillshade, mixed forest, and 
hardwood forest were deemed significant in determining Pacific lamprey presence on a 70%-
90% confidence level. The potential reasoning behind significance will be discussed for all of the 
listed variables due to the potential impact of chosen methods on results, however only variables 
on the 95% confidence level have been accepted as significant. 
The comparatively small number of significant variables in addition to their low 
confidence levels suggest that Pacific lamprey habitat is not easily defined. While the majority of 
sample sites where Pacific lampreys were present were at elevations under 225 ft (68.5 m), 
Pacific lampreys were consistently detected up to 1220 ft (371.8 m). It is reasonable to predict 
that they are fairly common at elevations above 225 ft (68.5 m) and that the low number of high 
elevation presences in this dataset is a result of the elevation distribution of sample sites. 
Pacific lampreys were present at relatively small slopes in the context of all slopes 
sampled, which could be explained by their poor swimming ability in relation to other freshwater 
fishes. This could also be associated with their greater presence at lower elevations. The 
significance of hillshade in Pacific lamprey habitat may be associated with the significance of 
slope, as hillshade values closer to zero often coincide with steep slopes, both of which are not 
preferred by this species (Appendix B). 
Presence and absence were both most common in conifer forests which is due to the 
distribution of sampled forest types and explains why this variable was not significant. Pacific 
43 
 
 
lampreys were more often absent in mixed forests and more frequently present in hardwood 
forests, showing a potential preference for hardwood forests over mixed forests. 
The highest cost distance at which Pacific lampreys were present was only about half of 
the maximum cost distance sampled. This is reasonable in the context of the species’ maximum 
elevation, as cost distance considers elevation and slope. This value also presents a limit for how 
far this species migrates upstream before settling to spawn. 
Western Brook Lamprey Habitat 
All ten of the assessed variables (elevation, slope, aspect, hillshade, stream density, 
conifer forest, mixed forest, hardwood forest, canopy cover, cost distance) were significant at a 
confidence level of 70% or higher in determining whether the western brook lamprey was 
present. The significance of such a wide array of variables would indicate that western brook 
lampreys have specific habitat preferences. However, only elevation, stream density, mixed 
forest type, and cost distance were significant on a 95% confidence level, meaning these are the 
only variables accepted as truly significant in the context of this paper. As with the Pacific 
lamprey, all variables will be discussed for the purpose of their potential implications in future 
studies. 
While the maximum elevation at which western brook lampreys were present was about 
438 ft (133.55 m), only one sample above 225 ft (68.5 m) indicated presence of western brook 
lamprey DNA. Furthermore, the mean elevation where western brook lampreys were present was 
47 ft (14.33 m), indicating that this species is most common at elevations below 225 ft (68.5 m), 
but can be found up to elevations of about 500 ft (152 m).  
The mean slope at sample sites where western brook lampreys were absent was 
significantly higher than slopes where they were present, indicating the preference for gentler 
44 
 
 
slopes. This preference could be related to their small body size, stream flow rate, or impacts of 
slope and flow on substrate type. 
The mean aspect where western brook lampreys were present was significantly smaller 
than where they were absent. Aspect is measured clockwise in degrees starting from true north, 
meaning the habitat western brook lampreys were not occupying was slightly more south-facing 
than their preferred habitat. 
The mean hillshade value where western brook lampreys were present was higher than 
that where they were absent. Hillshade is expressed by integers ranging from 1 to 255, with 255 
representing areas that are completely shaded.  Because the difference between hillshade values 
where western brook lampreys were present and absent was so small and significance could only 
be suggested with 70%, it is likely that hillshade is not a good indicator of suitable western brook 
lamprey habitat.  
Stream density was a significant indicator of western brook lamprey presence on a 
confidence level of 95%. Sample sites with stream densities between 0.001 and 0.00125 
streams/ft2 (0.0108 and 0.0135 streams/m2) had a considerable number of presences, indicating 
this range of stream densities to be preferred habitat for western brook lampreys. This may have 
to do with western brook lamprey life history, as these lampreys extend their range by migrating 
short distances from the redd from which they hatched. A more condensed stream network may 
also be equated with greater available habitat and because western brook lampreys do not 
migrate long distances from their origin, this may give populations greater opportunity to 
establish themselves with ample resources. 
The mean values for forest types suggest that western brook lampreys most prefer conifer 
forests and tend to inhabit mixed forests least often. While t-tests show that all three forest types 
45 
 
 
are significant in determining western brook lamprey presence versus absence, the fact that 
lampreys are more often present in strictly hardwood forest or conifer forest implies that this 
significance is likely due to other factors. 
Canopy cover was only a significant indicator of lamprey presence on a 70% confidence 
level, but the means where the species was present and absent indicate that it prefers canopy 
cover around 33%. There is significant overlap between the two distributions, though, which 
would suggest that canopy cover is not a reliable predictor of presence.  
The highest cost distance at which western brook lampreys were present was about one 
third of the maximum cost distance sampled. Cost distance does not have the same significance 
for western brook lampreys as it does for Pacific lampreys because this species does not migrate 
to their habitat from the ocean. Therefore, western brook lamprey presence at a high cost 
distance is not necessarily related to ability to travel that distance, but rather speaks to the 
location of the habitat in the stream and how it compares to the location of Pacific lampreys. 
Comparison 
Ten of the explanatory variables were significant in defining differences between habitats 
where western brook lampreys were present and absent, while only six were significant for 
Pacific lampreys. This would suggest that Pacific lampreys more readily adapt to different 
environments while western brook lampreys have a more specific set of requirements for suitable 
habitat. The adaptability of Pacific lampreys may be attributed to their migratory lifestyle as they 
must be able to survive in every environment they pass through while migrating down the length 
of streams towards the ocean. Western brook lampreys, on the other hand, exhibit minimal 
migration and their small body size may make them more susceptible to changing conditions. 
46 
 
 
For example, changes in canopy cover and slope could influence water temperature and flow 
rate, two variables that would more readily affect smaller animals. 
Elevation, stream density, and cost distance were the only explanatory variables that 
proved significant in differentiating between the two species’ preferred habitats. It was originally 
expected that western brook lampreys would occupy higher elevations as their resident life 
history and small size would allow them to penetrate further up stream networks over time. 
While western brook lampreys were generally found at elevations under 225 ft (the site at 438 ft 
was the only presence above 225 ft), Pacific lampreys were found at elevations up to 1220 ft. 
The significantly higher mean elevation at which Pacific lampreys were present is likely a result 
of their body size and thus greater ability to overcome passage barriers. 
Western brook lampreys were generally found at higher stream densities than Pacific 
lampreys. Higher stream densities may be associated with more complex stream networks of 
small tributaries. Western brook lampreys may be more prevalent in these environments because 
their smaller body size allows them to occupy smaller streams. In addition, these complex stream 
networks may be further from the ocean and more complicated to reach, so western brook 
lampreys may have a higher chance of reaching them by migrating from nearby source 
populations rather than from the ocean.  
The significance of cost distance can be attributed to many of the same factors as 
elevation. Cost distance gives insight into how far from the ocean a habitat is and how difficult it 
would be to travel there from the coast, so Pacific lampreys’ presence at higher cost distance 
speaks to their stronger swimming abilities and success at overcoming certain barriers. 
None of the explanatory variables proved significant in differentiating between habitats 
where each species was absent, suggesting that the habitats in which the species were absent are 
47 
 
 
similar in terms of these variables. Therefore, it may be easier to define habitats where they are 
absent. For example, you should not expect to find either species above 1220 ft based on these 
data. 
The two species showed some similarities for preferred habitat in terms of canopy cover, 
slope, hillshade. While canopy cover had large standard deviations for all groups, habitats where 
the species were present averaged about 34% while the habitats where they were absent had a 
mean of about 40%. The species also had very similar values for slope (sites where they were 
present had means of about 5) and hillshade (sites where they were present had means of about 
216). While this study does not have any official findings regarding the similarities between the 
habitats of these two species, the similarities between the means of these variables may indicate 
their applicability to lampreys overall even if they are not species specific. 
Models 
The accuracy scores for the random forest classifiers reiterate the specificity and 
predictability of western brook lamprey habitat in comparison with the more variable Pacific 
lamprey habitat. 
It is worth noting that the model more accurately predicts presence than absence, 
meaning it is more likely to produce a false positive than a false negative. This is ideal for 
management applications, as it would be favorable to assume lampreys are present and take 
precautions so as not to risk habitat degradation or threats to the animals.  
Elevation, cost distance, and stream density were the most important variables in 
predicting western brook lamprey presence. These leading variables agree with the t test results, 
indicating that they are good indicators of western brook lamprey presence. 
48 
 
 
Variables were much more even in their influence on the Pacific lamprey suitability 
model. Elevation was the only clear lead in MDI, but cost distance and slope didn’t fall far 
behind. The degree of influence from all other variables, excluding forest type, agreed with the t 
test results as well.  
Although t tests proposed that forest types were significant to the habitat of both species, 
they had the least influence of all variables in both models. This may be evidence that they are 
not truly important and that their significance in the t tests was due to chance. 
Limitations 
Observational data mentions that western brook lampreys are more prevalent in third 
order or smaller tributaries, suggesting that stream order should be a clear defining variable 
between Pacific and western brook lamprey preferred habitat (Schooler et al. 2022; Clemens & 
Wade 2023). While stream order was originally included in the list of explanatory variables for 
this analysis, the high definition and small size of the sampled creeks combined with the large 
study area made it exceedingly difficult to create an accurate stream order layer. The NHD 
Flowline dataset has a stream order field, however many of these classifications proved 
inaccurate and therefore could not support statistical analysis. The extensive observational data 
supporting the significance of stream order in specific lamprey habitat should be sufficient to 
argue the importance of this variable, however statistical analysis could be undertaken by 
deriving stream order manually. 
Correlations between variables presents a potential issue by deducing significance in 
variables that may not be truly important themselves (Appendix B). For example, the values of 
hillshade and slope at the sample sites used in this study had a correlation of -0.93. While it is 
clear to see why slope may be significant in lamprey habitat, it is likely that any significant 
49 
 
 
relationship between lamprey presence and hillshade is due to the slope at that site rather than the 
influence of hillshade. Elevation and cost distance also shared a relatively high correlation 
(0.78), but this correlation results from similarities between the measurements and thus should 
not produce any misleading results. 
As mentioned in the introduction, the nature of the eDNA data collection means analysis 
of sample sites may not produce results true to the lamprey habitat. In treating the sample sites as 
the source of the DNA as this study did, there is risk that DNA originated upstream and lampreys 
are not truly present at the habitat treated as suitable. On the other hand, DNA detected in 
sampling could originate 2 m upstream from the sample site, so averaging variables over several 
km could incorporate some habitat that is not suitable for the species. 
Because it is impossible to know the exact origin of the DNA detected at the sample site, 
it is also impossible to generalize habitat on a stream reach basis without some degree of 
speculation. While it is possible that lampreys were not truly present at some sample sites where 
their DNA was detected, many of the variables considered in this analysis and their associated 
conclusions are still viable. For example, while lampreys may have been present in the habitat 
above the maximum of 438 ft of elevation where they were detected, 438 ft should still be close 
to their maximum elevation. In addition, the significant difference between the elevations where 
each species was present should still stand. The values of most other variables should be similar 
enough in the upstream habitat to the values at sample sites to generally apply these conclusions.  
Use of quantitative PCR in lotic ecosystems may be one option for more reliably 
predicting the location of the source population. Alternatively, presence/absence sampling over 
smaller set of streams in segments would provide set areas in which to average variables. This 
would set a more consistent framework for characterizing upstream habitat while also potentially 
50 
 
 
identifying a population’s upstream limit. In situations where neither of these alternatives is 
possible, drawing upon eDNA persistence models, such as those described in Jo & Yamanaka 
(2020) should provide a good compromise for considering upstream habitat without too much 
risk of incorporating unsuitable habitat. 
Conclusion 
Western brook lamprey habitat is much more specific than Pacific lamprey habitat in the 
context of the chosen variables. The most definitive difference between the two habitats is 
elevation, as western brook lampreys were detected at a maximum of 438 ft and Pacific lampreys 
were detected up to 1220 ft. The significant difference between the maximum elevations at 
which each species was detected is also the most readily applicable finding from this analysis. As 
previously mentioned, the findings from this paper are not sufficient to predict the presence or 
absence of a species with 100% certainty, but the predictive file can be used to understand 
whether lamprey presence is likely prior to initiating restoration and construction projects. 
 
51 
 
 
Appendix A 
Variable summary table for sites where species were present 
 
52 
 
 
Appendix B 
Variable Correlation Matrix 
 
53 
 
 
Bibliography 
 
Balasingham, Walter, R. P., & Heath, D. D. (2017). Residual eDNA detection sensitivity 
assessed by quantitative real‐time PCR in a river ecosystem. Molecular Ecology 
Resources, 17(3), 523–532. 
Boeker, C. & Geist, J. (2016). Lampreys as ecosystem engineers: Burrows of Eudontomyzon sp. 
and their impact on physical, chemical, and microbial properties in freshwater substrates. 
Hydrobiologia, 777. 
Carim, K. J., McKelvey, K. S., Young, M. K., Wilcox, T. M. & Schwartz, M.K. (2016). A  
Protocol for Collecting Environmental DNA Samples From Streams. U.S. 
Department of Agriculture, Forest Service, Rocky Mountain Research Station. 
Carim, K. J., Larson, D. C., Helstab, J. M., et al. (2023). A revised taxonomy and estimate of 
species diversity for western North American Lampetra. Environmental Biology of 
Fishes.  
Clemens, B., Anlauf-Dunn, K., Weeber, M., Stahl, T. (2020). Coastal, Columbia, and Snake 
Conservation Plan for Lampreys in Oregon. Oregon Department of Fish and Wildlife. 
Clemens, B. & Wade, J. (2023). Conservation Biology of the Lampetra Satellite Species of 
Western North America, with a Focus on Western Brook Lamprey (L. richardsoni) 
Close, D. Fitzpatrick, M., Li, H. (2002). The Ecological and Cultural Importance of a Species at 
Risk of Extinction, Pacific Lamprey. Fisheries. 27. 
Collis, K., Roby, D. D. , Craig, D. P., Ryan, B. A., & Ledgerwood, R. D. (2001). Colonial 
waterbird predation on juvenile salmonids tagged with passive integrated transponders in 
the Columbia River Estuary: vulnerability of different salmonid species, stocks and 
rearing types. Transactions of the American Fisheries Society, 130, 385-396. 
Dunkle, M., Lampman, R., Jackson, A., & Caudill, Ch. (2020). Bring out Your Dead: Factors 
affecting fate of Pacific lamprey carcasses and resource transport to riparian and stream 
macrohabitats. Freshwater Biology. 65. 
Jo, Toshiaki & Yamanaka, Hiroki. (2022). Meta‐analyses of environmental DNA downstream 
transport and deposition in relation to hydrogeography in riverine environments. 
Freshwater Biology. 67. 
Kalan, Steinbeck J., Otte, F., Lema, S. C., & White, C. (2023). Filter-Feeding Pacific Lamprey 
(Entosphenus tridentatus) Ammocetes Can Reduce Suspended Concentrations of E. 
coli Bacteria. Fishes, 8(2), 101. 
 
54 
 
 
McPhail, D. L. (2007). The freshwater fishes of British Columbia. University of Alberta Press. 
Ostberg, Chase, D. M., Hoy, M. S., Duda, J. J., Hayes, M. C., Jolley, J. C., Silver, G. S., & 
Cook‐Tabor, C. (2019). Evaluation of environmental DNA surveys for identifying 
occupancy and spatial distribution of Pacific Lamprey (Entosphenus tridentatus) 
and Lampetra spp. in a Washington coast watershed. Environmental DNA (Hoboken, 
N.J.), 1(2), 131–143. 
Roffe, T.J., & Mate, B.R. (1984). Abundances and feeding habits of pinnipeds in the Rough 
River, Oregon. Journal of Wildlife Management, 48(4), 1262-1274. 
Schooler, S., Schmitt, J., Rudd, D., Flitcroft, R., Rodger, I. (2022). Mapping the Distributions of 
Pacific and Western Brook Lampreys Along the Oregon South Coast using eDNA and 
Community Science: 2019-2022 Report. South Slough National Estuarine Research 
Reserve. 
Seymour, Durance, I., Cosby, B. J., Ransom-Jones, E., Deiner, K., Ormerod, S. J., Colbourne, J. 
K., Wilgar, G., Carvalho, G. R., de Bruyn, M., Edwards, F., Emmett, B. A., Bik, H. M., 
& Creer, S. (2018). Acidity promotes degradation of multi-species environmental DNA 
in lotic mesocosms. Communications Biology, 1(1), 4. 
Wydoski, R. R., & Whitney, R. R. (2003). Inland fishes of Washington (2nd ed., rev. and 
expanded.). American Fisheries Society in association with University of Washington 
Press. 
 
 
55