AN EVALUATION OF SPAWNING HABITAT SITE SELECTION AMONG UPPER SOUTH FORK SALMON RIVER CHINOOK SALMON by KARA E COLLIER B.S., University of Wisconsin, Eau Claire, 2006 A thesis submitted to the Faculty of the Graduate School of the University of Colorado in partial fulfillment Of the requirements for the degree of Master of Science Environmental Sciences Program 2014
2014 KARA COLLIER ALL RIGHTS RESERVED
ii This thesis for the Master of Science degree by Kara E Collier has been approved for the Environmental Sciences Program B y Rafael Moreno Sanchez, Chair Peter Anthamatten Michael Wunder November 20, 2014
iii Dedicated to my parents, who raised me to follow my passions in life no matter where those passions take me I am eternally grateful to them for continuing to bel ieve in me and for the l ifetime of l ove, support and tolerance they have provided. Regardless of the adventure s or paths I have taken, they have always been at the other end waiting for me Thank you Mama, for the many hugs and nights of talking, laughter and endless plates of food. Thank you Papap, for putting a roof over my head and food in my be lly while dealing with my tornado of a presence for the past five years Also dedicated to my partner in life who endured many hours of detailed conversations on my progress, excitement or frustrations and who read and offered advice on many of my thesis drafts without hesitation. His love and encouragement lifted me up time and again Thank you Craig, for your surprises to help me smile, your gifts to help me relax a nd your infinite understanding that never ceases to amaze me
iii ACKNOWLEDGEMENTS To Dr. Rafael Moreno Sanchez, I am forever gr ateful for your belief in my abilities and your enthusiasm for my work. Thank you for taking an interest in me and working hard to help guide me in using my own passions and background to create this project. Without your ability to multi task and keep tra ck of my millions of emails, questions and concerns I would have likely been lost in a sea of my own madness. Your suggestions and ideas along the way helped me to keep my head on straight and guided me through one of the toughest challenges I have ever fa ced. I am endlessly inspired by your love for GIS and environmental modeling and my project would not have been the same without you. I look forward to our future projects together and am so very grateful for the continued opportunities you allow me to be a part of. I value you as an advisor and friend To Dr. Peter Anthamatten, your advice and counsel for my many ideas, projects and presentations helped me tend and grow giant beanstalks of cartographic skill and concise thought geared towards understandin g the technical aspects of mapping and writing. Without these elements, the posters, maps and reports that were the forethought of this thesis would have not been as significant a contribution. The enthusiasm you have shown in my work as both a professor a nd friend has give n me the confidence to think outside of the box on more than one occasion and I look forward to our future work together. To Dr. Michael Wunder, your statistical wisdom, patience and counsel was pivotal in the level of success achieved b and I am so very grateful that you agreed to be a part of it; y our knowledge of statistics and biology pulled m e out of cavernous rabbit holes on many occasions I admire the depth of your knowledge and the tenacity with which you approach your teaching, advising and career. Thank you for your guidance and understanding and I hope to one day find us fishing the South Fork as we discuss future research projects that focus on the very fish we are attempting to catch.
iv I am very fortu nate that the past four years of my life have provided me with the relationships, knowledge and resources that made this thesis possible. The spark that lit the motivated fire behind my graduate school conquest began when Gina Bonaminio hired me as an inte rn for the US Forest Service in McCall, Idaho in 2010. For each subsequent summer, she has hired me as a Fisheries Biologist Technician, affording me the ability to explore my interests, work with habitat data and network with other agencies concerned with Chinook salmon management, such as the Idaho Fish and Game and Nez Perce Tribe. Without these experiences and opportunities, I would not have had the building blocks I needed to create this project and see it through to fruition. To Gina, I would like to say thank you and that I believe we have always been our own Lady in Red. Initially, I hesitated to choose the thesis route for my graduate school education and without the encouragement from many people to choose the thesis option, I would likely not hav e completed such a wonderful project of which I am so passionate about. For the many discussions which involved trying to convince me that I would not regret writing a thesis and for his enthusiasm and advice along the way, I want to thank Wesley Keller fo r being the final push that helped me make my decision to jump head first into this project. you to know that I actually meant it at the same time I am grateful to the Idaho Fish and Game and Nez Perce T ribe for their contribution of Chinook spawning data and especially to Mitch Daniel and Laurie Janssen/Kim Apperson for their time and patience in helping me to understand the database spreadsheets; t heir assistance made this project possible. Without the interest and organizational efforts Mitch Daniel provided, I would have spent many grueling hours trying to understand spreadsheets and organize data. I want to thank him for his kindness in helping m e to make
v the very start of my project much less tumultuous and frustrating than it likely would have been. Mitch, I owe you so much more than one. I would also like to thank Caleb Zurstadt with the US Forest Service for allowing me to bounce ideas off of him to try and learn from his repertoire of fisheries management and Chinook salmon knowledge. His thoughtful ideas and words of advice helped me gain perspective in many ways throughout each stage of my project. Caleb, to your ability to see beyond the s urface and into the cavernous depths of just about anything, I would like to say, right arm. To the many friends I have made at the University of Colorado in Denver, the support we provided one another throughout this entire experience helped me maintain whatever sanity I had left in me and I will never forget the times you helped me keep my head above water. Our periods of treading water together kept us from sinking to the depths of the Marianas Trench and I am so lucky to have had each and every one of you. To Dr. Casey Allen, thank you for taking me under your wing during my first year of graduate school; your assistance in helping me attend conferences and acquire teaching assistant and instructor positions helped me to build my experie nce and further my knowledge in the field of Environmental Science Next time we meet, the nachos are on me. For my family and friends who decided to still love me through my disappearance from the face of the social Earth for two straight years, I never stopped thinkin g about you and what you mean to me. I promise I did not grow shockingly pale from the lack of sunlight nor does a mist of fresh air stiffen my library soaked lungs. Next time we meet, first round of club sodas are on me!
vi Collier, Kara E. (M.S., Environmental Science) An Evaluation of Spawning Habitat Site Selection Among Upper South Fork Salmon River Chinook Salmon Thesis directed by Associate Professor Rafael Moreno S anchez. ABSTRACT South Fork Salmon River Chinook salmon have been listed as threatened since 1992. In spite of this, their isolated spawning habitat within the upper South Fork watershed has not previously been geospatially analyzed. Spatial relationships of channel width, stream gradient and sediment size distribution with Chinook spawning habitat were derived through geospatial and geostatistical analyses of spawning site data collected between 2005 and 2012. Heavily spawned areas were found to have (1) a positive correlation with incr easing channel widths and ( 2) an inconsistent, negative correlation with increasing stream gradients. Descriptive analysis of sediment size distribution and Chinook spawning found the most heavily utilized spawning site alon g the South Fork to contain the highest proportion of fine sediment when compared to two other sediment sites. The information generated in this study will be valuable in identifying suitable spawning sites within the South Fork Salmon River and fundamenta l to local fisheries management within the Payette and Boise National Forest. The form and content of this abstract are approved. I recommend its publication. Approved: Rafael Moreno Sanchez
vii TABLE OF CONTENTS CHAPTER I. INTRODUCTION ................................ ................................ ................................ ................................ ............... 1 Chinook Salmon Significance as a Natural Resource ................................ ................................ 1 Policy and Managment ................................ ................................ ................................ ........................... 4 Study Objective ................................ ................................ ................................ ................................ ......... 6 II. BACKGROUND ................................ ................................ ................................ ................................ ................. 8 Study Area: South Fork Salmon River, Idaho ................................ ................................ ............... 8 Spawning Habitat Parameters ................................ ................................ ................................ .......... 10 III. METHODS ................................ ................................ ................................ ................................ ...................... 15 Datasets and Data Processing ................................ ................................ ................................ ........... 15 Analysis of Redd Density ................................ ................................ ................................ .................... 19 D atasets and Data Processing ................................ ................................ ................................ ........... 22 Analysis of Redd Presence and Redd Absence ................................ ................................ ........... 23 Data Sets and Data Processing ................................ ................................ ................................ .......... 25 Analysis of Sediment Size Distribution ................................ ................................ ......................... 28 Characterizing the Top Salmon Spawning Sites ................................ ................................ ........ 29 VI. RESULTS ................................ ................................ ................................ ................................ .......................... 30 Relationships with Redd Density ................................ ................................ ................................ .... 30 Relationships with Redd Presence ................................ ................................ ................................ 34 Sediment Site Comparisons ................................ ................................ ................................ ............... 37 Top 5 Spawning Sites ................................ ................................ ................................ ........................... 44 V. DISCUSSION ................................ ................................ ................................ ................................ ................... 47 Channel Width and Stream Gradient ................................ ................................ ............................. 48 Limitations of Channel Width and Stream Gradient Analyses ................................ ............ 52
viii Sediment Size Distribution ................................ ................................ ................................ ................ 56 Limitations of Sediment Size Distribution Analyses ................................ ............................... 58 VI. CONCLUSIONS AND RECOMMENDATIONS ................................ ................................ ............. 60 Channel Width and Stream Gradient ................................ ................................ ............................. 60 Sediment Distribution ................................ ................................ ................................ .......................... 62 LITERATURE CITED ................................ ................................ ................................ ................................ ......... 64
ix LIST OF FIGURES FIGURE 1 illustrates the locations of the sediment sites and spawning survey reach within the upper South Fork Salmon River watershed of central Idaho. All data deri ved for this study were sampled from these locations or spatially derived from the survey location via geograp hical information systems (GIS) ................................ ................................ .................... 8 2 depicts aerial images covering a 500 m stretch for each sediment site. Lush riparian vegetation can be seen in both Poverty Flats (A) and Stolle Meadows (C) while burn scars are visible around both Dollar Creek (B) and Stolle Meadows. The paved South Fork Salmon River road can be seen following Poverty Flats and Dollar Creek on the right side of each image. Lighter colored regions within the stream channel re veal shallower stream depths and/or higher levels of small gravels and fine sediments. ... 10 3 illustrates quartile spreads of redd counts over eight years (2005 2012) for all patches (Total) and low medium and high/very high density patches. ........................ 30 4 depicts quartile boxplot summaries for stream gradient percentage of redd density classes with minimum, median and maximum values reported for (1) all redd patch densities and (2) low medium and high/very high redd density classes. .................. 32 5 depicts quartile boxplot summaries for channel width in meters for redd density classes with minimum, median and maximum values reported for (1) all redd patch densi ties and (2) low medium and high/very high redd density classes. .................. 33 6 depicts quartile boxplot summaries for stream gradient percentages with minimum, median and maximum values reported for (1) the full survey reach and (2) redd presence and redd absence areas. ................................ ................................ ................................ ...... 34 7 depicts quartile boxplot summaries for channel width in meters with minimum, median and maximum values report ed for (1) the full survey reach and (2) redd presence and redd absence areas. ................................ ................................ ................................ ...... 35 8 illustrates total redd counts per year between 2006 and 2012 (7 years) found within the 5 00 meter stretch of river defined for sediment site. ................................ ......................... 37 9 illus trates quartile boxplot summaries with minimum, median and maximum stream gradient percentages for the 500 m stretches defined for Dollar Creek, Poverty Flats and Stolle Meadows sediment sites. ................................ ................................ ................................ .. 38
x 10 illustrates quartile boxplot summaries with minimum, median and maximum channel width measurements for the 500 m stre tches defined for Dollar Creek, Poverty Flats and Stolle Meadows. ................................ ................................ ................................ ..... 39 11 illustrates median proportion values out of 40 samples at each sediment size per year to demonstrate sediment proportion trends over time. A loess line (purple) and line of best fit (turquoise) are included in each scatterplot and the 95% confidence inter val is represented as dotted lines. ................................ ................................ ................................ .................. 41 12 demonstrates statist ically significant differences in the median proportion of sediment between each pair of years between 2006 and 2012. Poverty Flats is the only sediment site to result in statistically significant differences between years outside of comparisons with 2006. ................................ ................................ ................................ .... 43 13 illustrates the location of the top five spawning sites along the upper South Fork River (right) coupled with aerial views of each of the top five spawning sites captured from Google Earth (right). Each image is overlaid with an outline of the uniquely identified redd patch associated with the location. ................................ ................................ .......................... 45
xi LIST OF TABLES TABLE 1 shows redd density ranges defined for low medium and high/very high redd density classes. ................................ ................................ ................................ ................................ ............ 21 2 shows the range of each sediment size class defined for analyses. ................................ ...... 26 3 summarizes redd density statistics for low medium and high/very high redd density classes both with and without redd density outliers. ................................ ................. 31 4 represents the multiple logistic regression values and equation associated with the dependent variable of redd presence probability and independent variables of stream gradi ent, channel width and [S.Gradient:C.Width]. Because the interaction variable [S.Gradient:C.Width] shows significant contributions to the probability equation, all other variables are not considered. ................................ ................................ ................................ .... 36 5 characterizes instances of statistically significant differences found in median proportions of sediment size wi thin each year and between each pair of sediment sites based on the probability value threshold of 1.157e 04. Each instance is symbolized by a shaded cell. ................................ ................................ ................................ ................. 40 7 characterizes each of the top five redd density patches determined as the top five spawning sites of the upper South Fork. ................................ ................................ .......................... 44
xii LIST OF EQUATIONS EQUATION 1 Order Correlation Coefficient Test ................................ ................................ 20 2 Pairwise Wilcoxon Signed Rank Test ................................ ................................ ................................ 22 3 Multiple Logistic Regression Generalized Linear Model ................................ ........................ 24
1 CHAPTER I INTRODUCTION No matter how intently one studies the hundred little dramas of the woods and meadows, one can never learn all the salient facts abou t any one of them. Aldo Leopold (1948) A s a pioneer in environmental consciousness, Aldo Leopold recognized the vast and dynamic complexity of nature (1948) Decades later, the idea of an ever evolving and intricate natural world remains a fundamental concept of environmental research and resource management Because ecosys tems consist of dynamic objects, organisms and interactions, the continuous development of empirical data is essential for responsible planning and management of natural resources. This is especially applicable when considering spawning habitat suitability for Endangered Species Act (ESA) listed species, such as Chinook salmon. This study brings together datasets from multiple organizations to characterize spawning habitat and designate critical spawning sites for a population of Chinook salmon ( Onchorhyncu s tshawytscha ) in Idaho Chinook Salmon Significance as a Natural Resource Ecological Chinook salmon are a valuable natural resource with ecological, economic and cultural significance As the largest Pacific salmon, they mature at about one meter in le ngth and average around 18 kilograms (40 pounds) (NOAA 2012) Chinook are anadromous, which means they migrate from the sea to their natal (birth) freshwater streams to spawn Research has suggested that salmon will not only migrate to their natal stream, but often home to a specific reach which contains their incubation site (Neville, Isaak, Dunham, Thurow & Rieman, 2006; Quinn, Stewart & Boatright, 2006) Juveniles spend up to two y ears in freshwater and between two and six years feeding and maturing in the open ocean before migrating back upstream As a result, Chinook depend on and contribute to
2 biodiversity and trophic paradigms within multiple ecosystems ( Montgomery 2004; Dodds 2002 ; Cederholm 1999 ). Biodiversity is considered a fundamental natural resource with both eco centric and anthropocentric elements Within the multiple ecosystems Chinook occupy during their lifetime, their presence enhances both species and genetic biodiversity while there interactions wi th the environment contribute to ecosystem diversity Stability, productivity and resilience of ecosystems are positively related to biodiversity (Dodds 2002), which emphasizes the importance of Chinook salmon in both fresh and salt water environments Chinook provide energy and nutrients to terrestrial and aquatic systems through being consumed directly by animals such as killer whales, eagles or bears and also by addi ng nutrients into system s through decomposition (Cederholm 1999). Chinook are semelp arous, meaning they expire after spawning. Decomposition of Chinook carcass es provides marine derived nutrients to freshwater systems, such as nitrogen, phosphorus and energy rich carbon compounds (Hicks, Wipfli, Lang & Lang 2005; Mitchell & Lamberti 200 5; Wipfli 1997) Considering that reproductively isolated populations of Pacific salmonids evolved as far back as 15 20 million ye ars ago (Montgomery 2004 & 2000; Stearley 1992) and salmon carcasses contribute to the freshwater nutrient cycles of their natal streams, it stands to reason that the nutrients and benefits provided by decaying salmon carcasses hav e become especially important to f reshwater environments. Gresh et al. (2000) discovered that decreased numbers of spawning salmon greatly decreas es the amount of marine derived nutrients delivered to streams and may create nutrient deficient aquatic systems. Pacific salmon gain up to 95.0 0 % of their total body mass during maturation in the sea (Groot & Margolis 1991), subsequently contributing con siderable quantities of nutrients to freshwater streams ( Cram, Kiffney, Klett & Edmonds 2011; Schindler et al. 2003 ; Naima n, Bilby, Schindler & Helfield 2002;) S almon carcasses
3 have been found to provide anywhere from 25.0 0 to 75.0 0 % of annual nitrogen loads (Finney, Gregory Eaves, Sweetman, Dougas & Smol 2000) which support riparian and aquatic plants as well as juvenile salmon (Helfield & Nai man 2001; Naiman et al. 2002) Marine derived nutrient loads from salmon carcasses also produce boosts in primary producer biomass and primary consumer populations (Cram et al. 2011; Schuldt & Hershey 1995) Previous research has found that primary producer biomass, such as plankton and algae, provides bottom up support in juvenile Chinook food webs (Maier & Simenstad 2009) Specifically, boosts in primary producer biomass help support macro invertebrates (e.g. primary consumers), which are the dominant food source for juvenile Chinook prior to their migration to the sea and during maturat ion within coastal waters and estuaries. While maturing in the sea, Chinook have been found to primarily feed at higher trophic levels in coastal waters (Johns on & Schindler 2009) helping to support coastal food webs by keeping prey populations in check. Economic. Chinook should also be considered closely connected to the condition and well being of local and national economies They are considered one of the most valuable salmonid species caught by anglers (McKean 2000) The US Fish and Wildlife Servi ce reported that over 90 million people participated in fish and wildlife related activities in 2011, with around US$87 billion paid to hunting and fishing expenditures (USFWS 2012) The average angler spent US$1,262 in 2011, with 91.0 0 % of anglers contri buting to residential state revenue and 21. 0 0% contributing to non residential states by travelling across borders in order to fish (USFWS 2012) According to the 2011 Fisheries of the US (FUS) report, the domestic commercial Chinook salmon fishery yielde d more than US$44 million (NOAA 2012) in 2011 to pay more than $33 million per year for recreational fishing (McKean 2000) The survey
4 also evaluated a potential economic benefit of $544 million annually if salmon fisheries in Idaho were restored to historical abundances (McKean 2000). Cultural. Rest oring salmon fisheries to past population abundance s could also enhance the quality of l ife and economic stability of struggling tribal communities. Historically, Chinook salmon have held spiritua l and cultural significance for the Columb ia River Basin tribes of the US Northwest showing that salmon based econo mies appeared as far back as 300 0 years ago (Lichatowich 1999). Well developed economies were sustained through specialized technology and highly regarded rules geared towards the wise use and respect of salmon. Native Americans based their economies on the concepts of gifts and equalit y in all parts of the earth This meant that even plants, animals and rock s were a part of their community and any animal killed for nourish ment was a gift from the animals to the people Salmon became their principle resource an d stable economies based on annual Chinook salmon runs existed for up to 1500 years before E uropeans appeared in the region, bringing with them the worldview that nature was to be tamed and that land and water resources were inexhaustible (Lichatowich 1999). As a trade item as wel l as a primary source of protein, Chinook had remained a celebrated and pivotal component for tribal communities before European settlement s began to invade their territories A steady decline in Chinook salmon abundance has occurred over the past few centuries with the greatest decline Chinook returns created an economic deficit that destroyed many tribal communities and continues to threaten the livelihood of others Policy and Manag e ment Population v iability In the continental US, salmon have been wiped out of almost half their historic range (Nehlsen et al. 1991). The decline of Chinook populations can be attributed to both natural
5 and anthropogenic influences, including fisheries exploitation, ma jor dam construction an d critical habitat degradation ( Utz, Mesick, Cardinale & Dunne 2013) Critical habitat is defined as any area contained within or outside of a geographical unit that is essential to the conservation of a species. National Marine Fis heries Service (NMFS) identifies salmonids according to evolutionary significant units (ESU) which are defined as reproductively isolated populations that hold evolutionary significance for a species (NOAA 2012) The Snake River spring/summer Chinook salm on ESU has been ESA listed as threatened since 1992 (NOAA 2012). This ESU comprises 28 extant populations, including the four populations of the South Fork Salmon River watershed in Idaho. been referred to as one of the most important sub basins in the entire Colombia River basin relative to salmonid abundance and exploitation (IDFG 1985). Spawning Chinook abundances in the South Fork have been inconsistent, experiencing m ultiple periods of growth and decline over the past 30 years (Leth, Cassinelli and Knipper, unpublished works). Due to inconsistent production and repeated low counts of wild (e.g. non hatchery) Chinook spawners, overall viability ratings that examine abun dance and productivity continue to find South Fork Chinook populations at high risk (NOAA 2012). Management i ndicators A principal limiting factor in salmon recovery is the quality of critical spawning habitat (Honea et al. 2009) which influences adu lt spawning habitat selection ( Johnson, Roni & Pess 2012) and is positively correl ated to embryo and emerging fry survivorship ( Sternecker, Cowley & Geist 2013 ; Utz, Zeug & Cardinale 2012 ; Sternecker & Geist, 2010 ; Armstrong, Kemp, Kennedy, Ladle & Milner 2003 ) Spawning habitat quality depends on many hydrologic factors such as depth, channel width, gradient and substrate distribution An adult female will dig test sites prior to selecting a final location to construct her
6 spawning bed, referred t o as a redd Redds can reach up to 4.0 m in width (Isaak & Thurow 2006) and are typically found in lower stream gradients (Hanrahan 2007; Isaak & Thurow 2006; Mo ntgomery, Beamer, Pess & Quinn 1999; Platts Torquemada, McHenry & Graham 1989) Large cha nnel widths, which are associated with lower stream gradients (Benda et al. 2004) are also preferred by spawning Chinook salmon ( Montgomery 2004 ; Kondolf & Wolman 1993 ). Because Chinook rely on natal streams for spawning and rearing juvenile s their ab undance is strongly dependent on the geomorphology and condition of the natal stream and its watershed. Within a single watershed, variability exists in the landscape dynamics that govern the supply and movement of water and sediment (Montgomery 2004). Du e to watershed and habitat variability, wild salmonid populations can exhibit localized adaptations ( Rollinson & Hutchings, 2011 ; Ramstad, Woody & Allendorf 2010 ; Rosenfeld, 2003) associated with their natal streams; an occurrence supported by the existence of 28 extant genetic lines of Snake River spring/summer run Chinook salmon. The potential for localized adaptations of geographically isolated populations of Chinook creates a n eed for site specific understanding of spawning habitat and should be emphasized in the management of threatened South Fork Chinook salmon. Study Objective Current m anagement policy. Anthropogenic influences, such as timber harvest, road construction, mining and livestock grazing, combined with natural events, such as fires and flooding, have all impacted habitat in the South Fork ( NOAA 2012 ; Nelson 2010). As a consequence, spawning habitat protection and remediation are critical in fisheries manageme nt recovery strategies. The current recovery strategy outlined by NMFS (2012) focuses on a top priority of reducing fine sediment delivery and includes the following actions:
7 (1) Decommissioning roads in selected areas of important Chinook spawning and rearing habitat (2) Improving riparian function in selected locations to reduce soil erosion, especially in areas that have been degraded by noxious weeds and other anthropogenic influences. Though the action items listed above are straightforward, the time and fun ding dedicated to their implementation can be streamlined by determining the stream locations most utilized by spawning South Fork Chinook. Additionally, quantifying critical spawning habitat features associated with South Fork Chinook will make the proces s of designating and prioritizing suitable spawning habitat more efficient. Enhancing policy through quantification. Critical spawning habitat for many ESA listed Chinook populations has been characterized H owever the spawning habitat for the South For k Salmon River major population group has yet to be quantified with the use of empirical, site specific data T his study which is the first of its kind, focused on quantifying three known variables associated with salmon spawning habitat relative to one m ain stem population of summer run South Fork Chinook salmon. Empirical sediment and spawning site data were used along with GIS software and statistical analyses in order to identify critical South Fork spawning habitat. This research also utilize d geographic coordinate data on redd locations in conjunction with stream gradient, channel width and sediment distribution data to generate a detailed classification of the suitability for Chinook salmon spawning sites in the South Fork.
8 CHAPTER II BACKGR OUND Study Area: South Fork Salmon River, Idaho Geomorphological history. The U pper South Fork watershed penetrates both the Payette and Boise National Forests and is embedded in a granitic bedrock formation referred to as (Figure 1; EPA 2013) The Idaho Batholith is characterized by narrow valleys, steep slopes and shallow, unconsolidated soils (Platts et al. 1989) which create highly erodible conditions throughout the South Fork r iver valley. These erodible conditions are amplified by regular periods of fire and flooding, which can disrupt stream habitat, impact riparian vegetation and result in increased sedimentation into the watershed ( Nelson, 2010 ; Benda, Miller, Bigelow & Andra s 2003; Meyer, Pierce, Wood & Jull, 2001 ) Most recently, hig h flow and flood events occurred in 2008 and 2010 with massive fire events occurring in 20 06 and 2007 (Bonaminio & Nelson 2012). Poverty Flats Stolle Meadows Dollar Creek Upper South Fork Salmon River Watershed Flow A B C A. B. C. Figure 1 illustrates the locations of the sediment sites and spawning survey reach within the upper South Fork Salmon River watershed of central Idaho. All data derived for this study were sampled from these locations or spatially derived from the survey location via geographi cal information systems (GIS).
9 Study location. Critical spawning habitat for four ESA listed populations of Chinook salmon is contained within the Upper S outh Fork Watershed (NOAA 2012). Located in the Salmon River Mountain range of central Idaho, elevations for the upper South Fork range from 654 to 2841 m (2146 to 9321 ft) above sea level (USGS 2003). The 138 km long South Fork encompasses spawning habi tat for the threatened summer run population of Chinook that is the principle species of this study (USGS 2003). Though summer run Chinook also utilize tributaries for spawning, breeding activity by the major population group primarily occurs in the upper 50 km of the South Fork main stem (Platts et al. 1989) As a result, spawning habitat analyses did not include data from tributaries and were isolated to a 52 km reach of the South Fork that is surveyed for Chinook spawning sites annually (hereafter refe rred to Though five historic sediment survey sites were chosen in 1965 after major flood events inundated the South Fork with sediment and debris (Bonaminio & Nelson 2012; Nelson 2010), data used for sediment distribution analyses were restricted to three sites : Poverty Flats, Dollar Creek and Stolle Meadows (Figure 1 & 2 ). Sediment site elevations range from 1267 m (4157 ft) at Poverty Flats to 1488 m (4882 ft) at Dollar Creek and 1629 m (5345 ft) at Stolle Me adows. Both Poverty Flats and Stolle Meadows are surrounded by large floodplains covered in lush riparian vegetation (Figure 2) Burn scars encompass the area surrounding Dollar Creek although young trees and other vegetation have begun to grow (Figure 2) Dollar Creek sediment site has a short stretch of riparian vegetation within the floodplain of one bank but consists primarily of steep and rocky slopes (Figure 2) The paved South Fork Salmon River road lies parallel to Poverty Flats and Dollar Creek wh ile Stolle Meadows remains isolated from major forest transportation routes (Figure 2).
10 Spawning Habitat Parameters Channel w idth Channel width has been mentioned in the literature regarding Chinook spawning habitat, yet information has mostly been very general Details on channel width preference for spawn ing ha ve been predominantly restricted to simple notes explaining that there may exist a preference for larger channel widths due to the large size of the fish (Hall, Holzer & Beechie 2007; Montgomery 2004 ; Kondolf & Wolman 1993) Only a few specif ic suggestions on channel width preference have been made from actual field data: (1) Evidence suggests that reaches smaller than four meters in width may be inaccessible to Chinook during the low flows that occur during spawning season ( Busch, Sheer, Burnett McElhany & Cooney 2013) (2) Chinook rarely spawn in reaches smaller than three to five meters in width ( Washington Department of Fish and Wildlife 2001 ; Montgomery et al. 1999 ). These minimum thresholds may be due to the large size of Chinook salmon in combination with the space required for redd construction. The average dimensions of Figure 2 depicts aerial images covering a 500 m stretch for each sediment site. Lush riparian vegetation can be seen in both Poverty Flats (A) and Stolle Meadows (C) while burn scars are visible around both Dollar Creek (B) and Stolle Meadows. The paved South Fork Salmon River road can be seen following Poverty Flats and Dollar Creek on the right side of each image. side of each image. Lighter colored regions within the stream channel reveal shallower stream depths and/or higher levels of s mall gravels and fine sediments.
11 Chinook redds vary widely and have been estimated to range between 2.0 and 45.0 m 2 (US Fish and Wildlife Service 1995) with a mean area of anywhere between 6.0 and 17. 0 m 2 (Healey 1991). Larger channel width measurements are also associated with complex channel formations considered to be important for rearing juvenile Chinook, including side and braided channel habitats (Busch et al. 2013; Hall et al. 2007) In an analysis of the Columbia and Snake River redd densities Dauble and Geist (2000) found a statistically significant relationship between larger channel widths and the concentration of spawning Chinook salmon. Though a relationship with channel width w as supported, the range of channel widths associated with spawning Chinook salmon populations in these two rivers are extremely large compared to channel widths in the South Fork. Therefore, minimum and maximum thresholds for channel width preference of So uth Fork Chinook cannot be estimated from previous research and must be quantified through site specific analyses. Stream g radient Stream gradient has been a common and integral variable in the large body of research on Chinook spawning habitat suitabil ity. It is well understood that Chinook redds are typically found in lower stream gradients (Isaak & Thurow 2 006; Montgomery et al. 1999; Platts et al. 1989) though Chinook spawning has been recorded in gradients up to 7.0 0 % (Sheer & Steel 2006) Daub le and Geist (2000) found that Chinook salmon redds were typically located in stream gradients of 4.0 0 % or smaller within the Columbia River, while an unpublished study on Chinook salmon in Alaska identified a preference of smaller than 1.0 0 % stream gradie nts when spawning (B idlack, Benda, Miewald, Reeves & McMahan 2014) Further research completed in Idaho suggests Chinook salmon prefer stream gradients of less than 2.0 0 % (Isaak & Thurow, 2006; Platts et al., 1989) As evident by the various geographical locations having a large range of reported stream gradients preferred
12 by Chinook salmon, localized adaptations are a potential explanation which supports this specific characterization. A review of the literature suggests that th e geomorphological features associated with lower stream gradients are critical features preferred by spawning Chinook. Lower gradients are associated with wider stream channels (Benda et al., 2004) and pool riffle morphology along longitudinal stream prof iles (Powell, Laronne, Reid & Barzilai, 2012) Pool riffle stream morphologies result from low stream gradients and have previously been shown to be correlated with spawning Chinook distributions ( Tonina & Buffington, 2009 ; Montgomery et al., 1999 ) Hanrahan (2007) found that 84 .00 % of the variation in spawning habitat choice for Chinook salmon in the Snake River, Idaho, was explained by the presence of riffles (p<0.001). Riffle features are constrained by many factors including channel width and st ream gradient and are considered features of low depth, relatively high flow velocities and coarse substrate, which are also preferred by spawning Chinook salmon (Brown & Pasternack, 2008 & 2009; Bjornn & Resier 1991). Pool riffle channel types also show increased hyporheic exchange, which is critical during egg incubation ( Geist et al., 2002 ; Dauble & Geist, 2000 ) The hyporheic zone is a transitional ecotone between groundwater and river water and the interaction between the two (e.g. hyporheic exchange) can increase dissolved oxygen flow to egg pockets and help regulate water temperature during seasonal fluctuations (USGS 2013). Other features associated with lower stream gradients that have been shown to be important predictors in Chinook spawning ar e sinuosity (Fukushima, 2001) and complex channel features such as side channels and islands (Busch et al. 2013 ; Dauble & Watson 1990). Overall, the variable features considered associated with spawning Chinook salmon habitat are numerous and many of the m are associated with lower stream gradients. Consequently the general idea that Chinook salmon prefer lower stream gradients has been
13 considered universal across all populations of Chinook regardless of geographic location even though ranges of accepta ble stream gradient s may vary due to local geomorphology and localized adaptations. Therefore, site specific characterization of stream gradients preferred by South Fork Chinook could reveal information about suitable habitats for this geographically isola ted species. Sediment size d istribution Sediment size distribution is a well documented limiting factor in spawning habitat suitability for Chinook ( Sternecker et al., 2013 ; Johnson et al., 2012 ; Louhi Ovaska, Maki Petays, Erkinaro & Muotka, 2011 ; Sternecker & Geist, 2010 ; Hanrahan, Geist, & Arntzen, 2005 ; Coulombe Pontbriand & LaPointe, 2004). Sediment distribution and the variation in fine sediment volumes are directly associated with stream gradient (Anlauf et al., 2011) and sediment permeability (Hanrahan et al., 2005) Sediment permeability is positively correlated to embryo and fry survival (young s upon emerging from their redd ) and negatively correlated with percentage of fines (Hanrahan et al., 2005; Shirazi & S eim, 1981 ) Well sorted, coarse sediment increases porosity, and therefore permeability, of a stream bed, allowing for oxygenated water to flow through channels and into egg pockets more effectively. Greater permeability allows large r volumes of oxygen ric h water to be delivered to embryos during incubation (Hanrahan et al., 2005) ultimately increasing embryo survival rates (Sternecker et al., 2013) Permeability is also positively linked to fry survival by providing channels of passage for emergence from substratum (Sternecker & Geist, 2010) High f ine sediment volumes, ranging from smaller than 0.85 to smaller than 6.4 mm, are negatively linked to embryo and fry survival rates ( Sternecker et al., 2013; Utz et al., 2013 ; Hanrahan et al., 2005; Shirazi & Seim, 1981 ) Fine sediment can impact survival rates by filling in egg pockets which could block oxygenated water flow and suffocate embryos or
1 4 block channels for emerging fry. A published review of the research available for embryo survival rates found th at the odds of survival for Chinook decreased by 6.7 0 % with every 1.0 0 % increase above 20 .00 % volume in fine sediment smaller than 4.8 mm in diameter (Jensen, Steel, Fullerton & Pess, 2009) The Payette National Forest is a key agency involved in the mana gement of upper South Fork Chinook habitat Guidelines designated by the Payette National Forest Monitoring and Evaluation Report (2012) and as determined by Burns and Nelson (2005) will be used as an additional reference for habitat risk. As stated by the report, spawning habitat is functioning at risk (FR) when the five year mean volume of sediment smaller than 6.33 mm in size is (1) between 28 .00 and 36 .00 % with no more than two years greater than 36 .00 %, or (2) between 28 .00 and 36 .00 % with an increasin g trend over the last 10 years or (3) 36 .00 % or more with a decreasing trend over the last 10 years. Spawning habitat is functioning at unacceptable risk (FUR) when the five year mean of sediment smaller than 6.33 mm in size is (1) 36 .00 % or more or (2) 36 .00 % or more with an increasing trend over the past 10 years. Though fine sediments have been defined usi ng various scales ranging from <0.85 to < 8.4 mm ( Sternecker & Geist, 2010 ; Jensen et al., 2009) this study will define fine sediment s as anything smaller than 6.4 mm in diameter t o correspond with the Payette National Forest Mo nitoring and Evaluation Report. An additional sub division of super fine sediment as anything smaller than 2.3 mm in d iameter will also be considered to correspon d with previous Forest Service documentation.
15 CHAPTER III METHODS Redd Density Datasets and Data Processing To examine the relationships of redd density with stream gradient and channel width, a dataset of 545 uniquely identified redd patches was created with values of redd density (redds/m 2 ), stream gradient (in % slope) and channel width (in meters). The following sections outline the steps and processes involved in creating the dataset. Redd location c oordinates. Idaho Fish and Game (IDFG) and Nez Perce Tribe (NPT) Fisheries annually complete ground redd surveys to support research related to uncertainties of hatchery supplementation effects on natural Chinook populations in Idaho (IDFG 2012). Survey reaches represent all probable spawning hab itat along the upper South Fork with a total length of around 52 kilometers Surveys are completed between July and September and include either multiple passes throughout the spawning season or a single pass during peak spawning time to achieve increased accuracy of total redd counts (IDFG 2012) Each observed redd is georeferenced using a hand held global position ing system (GPS) and flagged to avoid duplication; neither in field averaging nor post processing techniques are completed for the geographical coordinates. To standardize measurements, all surveyors used the World Geodetic System (WGS) 1984 geographic coordi nate system in marking redd locations Two data sets one from NPT Fisheries and the other from IDFG were obtained and contained a combined total of 5388 redds observed between 2005 and 2012. Redd location coordinates along the South Fork river were imp orted as point features into ArcGIS 10.1 with an attached value representing the number of redds at each point location. Most points represented an individual redd with a few exceptions representing multiple redds.
16 The tot al number of redds utilized in this study was narrowed down to 4856 in three stages: (1) Ground survey techniques have previously been shown to provide better accuracy of redd counts compared to aerial surveys due to multiple reasons, including flight speed and limited visibility (USFWS 19 98). Limited visibility results from canopy cover, overhanging vegetation and the view angle of the observer relative to the suns reflectivity off the surface of the water. O nly survey extents with consistent ground survey techniques were used, decreasing the total survey reach length to 49 kilometers. (2) A locational error of 15 m is assumed to account for the multiple year extent of the study and the use of different GPS devices. A ll redds beyond 15 m from the stream bank were excluded from the study b ecause accurate stream channel locations for redds would be difficult to interpret (3) Only main channel redds were considered due to the difficulty in accurately processing side channels and braided river systems. Specifically, GIS methods applied to spawning habi tat characterization could not differentiate between redd locations where multiple channels were in close proximity to one another, causi ng overlap in redd patch analyse s. Defining redd patches and redd density. Redd density represented the total count o f individual redds present over eight years at the same location. Therefore, redd density is partially characterized by the continued spawning in a single area over multiple years and can be interpreted as an indication of stable, high quality spawning hab itat within the upper South Fork. Defining clusters of redds as distinct patches was beneficial in maintaining standard units of comparison for both redd presence and redd density. Considering the large size and elliptical shape of
17 Chinook redds with the a ssumed 15 m locational error of GPS coordinate readings, any redds within 15 m of one another were considered associated and existing within similar habitat. A 7.5 m buffer was created for each of the 4856 selected redds, enabling redds within 15 m of one another to have intersecting boundaries. Dissolving boundaries at overlapping buffers resulted in 545 redd patches. Redd density was calculated by dividing the total number of redds within each patch by patch area. Calculation of stream gradient. A Digital Elevation Model (DEM) with 8.87 m resolution was downloaded from United States Geological Survey, National Map website (USGS 2013). The National Map 1/3 arc second DEMs are derived from either digital photogrammetry or Interfe rometric Synthetic Aperture Radar (IFSAR) and contain elevation values that do not penetrate water body surfaces (USGS 2013). To produce a smooth water surface without depressions, h ydrological analysis tools in ArcGIS 10.1 were utilized to fill sinks (e rroneous, local depressions) present in the DEM To facilitate processing and analysis, the DEM was then resampled to a 10 m resolution. A line was manually digitized via visual estimation using a 1 meter resolution, georeferenced image from the Digital O rthorectified Aerial Imagery Series of Idaho (2009) to represent the thalweg (main flow) of the upper South Fork. Prior to finalizing the stream thalweg line, segments corresponding to seven bridge locations were deleted in order to avoid skewed stream gra dient calculations. The resulting stream line was rasterized to 10 m resolution with each cell reclassified to a value of one. Elevation values from the DEM were extracted using the stream raster through a GIS raster calculator multiplication operation. Th is resulted in a stream raster with attached elevation values. The ArcGIS 10.1 Slope function was used to calculate gradient in percent per 10 m along the Upper South Fork.
18 Slope function was not performed on the entire DEM to avoid false, steep calculatio ns of gradient within the river channel. False gradient calculations could occur due to the slope three by three window (i.e. 30 m x 30 m) inevitably taking into account elevation values located outside of the stream channel Gradient estimati ons per 10 m segments would have portrayed accuracy that was not representative of the actual elevation data used Specifically, it is difficult for any DEM to characterize exact values along the entire upper South Fork because stretches of the river can v ary considerably in width and sinuosity Consequently, s tream gradient was averaged every 20 m to generate more conservative estimates. The full upper South Fork line was split into 20 m segments and then buffered to create polygons 10 m wide by 20 m long. ArcGIS 10.1 Zonal Statistics as a Table function was run using the stream raster and buffered polygons as inputs to calculate the average stream gradient every 20 meters. Resulting values were spatially joined with their corresponding stream polygons for later use Calculation of stream gradient for redd patches. The 49 from the upper South Fork thalweg line using coordinates provided by the IDFG and NPT. This resulted in a 49 km survey reach containing segments with stream gradient in percent slope attached to them. Polygons to mask the stretches of river where no redd patches were present ng parallel to and width running perpendicular to the survey reach thalweg line Stream were created by erasing survey reach segments where redd absence polygons overlapped with the
19 survey reach. This resulted in a total redd presence stream length of 20 km that included attached gradient values. Redd presence stream segments were buffered to create 80 m wide polygons that ensured an intersection with every redd patch. Six issues described in the discussion arose due to meandering bends of the stream and the effects of overlap from the 80 m wide buffer. Stream gradient percentages were spatially joined to the corresponding buffered polygons. A spatial join of buffered polygons t o redd patches resulted in multiple gradient values being attached to each redd patch which were used to calculate mean gradient fo r each of the 545 redd patches. Calculation of channel width for redd patches. Channel width is defined as the wetted width (length is measured from wetted edge to wetted edge) of a stream channel. Three channel width lines were manually digitized via visual estimation for each redd patch using the same 1 meter resolution digital ort hoimagery ( Digital Orthorectified Aerial Imagery 2009). C hannel width measurements were placed at equal intervals across the longitudinal length of the patch in order to better represent an average measurement. Each channel width line was drawn perpendicu lar to the direction of the thalweg line. If longitudinal length of a redd patch exceeded 50 m, channel width lines were doubled in order to obtain a more representative measure. Channel width lines were spatially joined in ArcGIS 10.1 to redd patches and the mean value calculated, resulting in 545 redd patches with mean stream gradient (previously calculated) and mean channel width values attached to each of them. Analysis of Redd Density R elationships to channel width and stream gradient Prior to stati stical examination of relationships and upon plotting redd density with channel width and stream gradient, it was realized that (1) two extreme redd density
20 outliers existed, and (2) the data did not meet assumptions of normality, linearity and homoscedast icity (constant variance). The two extreme values of redd density were confirmed as influential outliers using an outlier test in R statistical software and we re excluded from further analyse s involving density. Because data distribution assumptions were n ot met, non parametric methods of analysis were chosen. order correlation coefficient test. dependence between the rank ordered variables of stream gradient and channel width to redd density (Equation 1), resulting in a variables and a probability value in dicating statistical dependence between variables. The test also reported on the relationship of stream gradient to channel width. Due to a significant correlation between them (p=0.03), an interaction term for [S.Gradient : C.Width] was added to the an alyse s Testing with this interaction term means that the association between stream gradient and channel width will be taken into account when testing for relationships with redd density. Equation 1 Order Correlation Coefficient Test
21 L ow, medium and high/very high redd density classes are easily app licable in management processes and serve as valuable indicators for Chinook spawning habitat occurrence along the South Fork River. Upon viewing the redd density dataset noticeable breaks in the values were evident, allowing redd densities to easily be d ivided into the three classes (Table 1). Pairwise Wilcoxon signed rank test. T he Pairwise Wilcoxon test is a paired difference test that compares the median p opulation ( n ) rank values of two samples, resulting in a probability value that estimates whether the distributions of [ x n ] are identical between the paired samples of redd den sity classes (Equation 2) Pairwise Wilxocon tests with Bonferroni p value adjustments were completed between paired redd density classes to determine whether statistical differences exist in population mean rank values for channel width, stream gradient and the interaction term. Table 1 shows redd density ranges defined for low medium and high/very high redd density classes.
22 E quation 2 Pairwis e Wilcoxon Signed Rank Test Redd Presence Datasets and Data P rocessing In order to examine the relationships between spawning (redd presence) and non spawning (redd absence) habitat, a dataset of 545 redd absence stream segments was created. Each of these stream segments has attached values of stream gradient (% slope) and channel width ( m ). This dataset was used in combination with the previously identified redd presence data to anal yze relations hips with stream gradient and channel width. Redd absence and stream gradient. Redd absence stream segments were clipped from the survey reach using the redd absence polygons previously created, resulting in a total redd absence stream lengt h of about 29 kilometers. Because redd absence stream segment lengths could extend up to 1000 m, determined by the distance between actual redd patches, it was necessary to define a uniform segment length that would enable comparisons with redd presence ha bitat data. After the longitudinal stream length (via the thalweg line) was determined for each redd patch, normality assumptions were not found to be met within the stream length data
23 distribution. Thus, median length of redd patches (17 m) was used to cr eate 1708 redd absence stream segments for comparison. Redd absence stream segments were spatially joined in ArcGIS 10.1 wit h the survey reach to calculate mean gradient for each 17 m segment prior to exporting the attribute table to excel where 545 random redd absence stream segments were selected for statistical analyses. Redd absence and channel width. The randomly selected 545 redd absence segments wer e isolated from the survey reach in ArcGIS 10.1 Calculation of mean channel width per redd absence segment mirrored methods previously used to calculate redd patch channel width, resulting in 545 segments with stream gradient and channel width values. An alysis of Redd Presence and Redd Absence C hannel width and stream gradient. Upon plotting redd presence and absence with channel width and stream gradient, it was again realized that assumptions of normality, linearity an d homoscedasticity were not met Pairwise Wilxocon tests with Bonferroni p value adjustments were completed to determine whether statistical difference s exist between redd presence and absence relative to median values of channel width, stream gradient and the inter a ction term.
24 Red d pre sence probability based on channel width and stream gradient. Logistic regression was used to measure the relationship between the categorical depende nt variables of redd presence and redd absence and the independent variables stream gradient, channel wi dth and the interaction term, resulting in a measure of the probability that the independent variables contribute to redd presence (Equation 3). Equation 3 Multiple Logistic Regression Generalized Linear Model A generalized linear model following a binomial distribution was fit with stream gradient, channel width and the interaction term for redd presence. The binomial distribution family was chosen because over dispersion of the binomial linear model was not found to be an i ssue (p=0.49), which means observed variance of redd presence was not larger than what would be expected from a binomial distribution. Resulting regression coefficients show direction of the relationship between all three independent variables and redd pre sence. Exponentiation of regression coefficients was completed to assess the odds of redd presence based on values for stream gradient, channel width and the interaction term. The coefficients that are calculated represent the change in odds of finding re dd presence based on a one unit change in the independent variables. A prediction test of the multiple logistic regression model was completed to assess the probability of redd presence based on given values of stream gradient and channel
25 width. Specifically, the analyses provide probability values of (1) finding a redd at varying levels of s tream gradient while holding channel width constant and (2) the probability of finding a redd at varying levels of channel width while holding stream g radient constant. Constant measures were held at median values of str eam gradient and channel width. Data Sets and Data Processing To examine relationships between three sediment monitoring sites, the following dataset was created: t hree sediment site pol ygons with (1) total redds per year (2006 2012), (2) average stream gradient, (3) average channel width and (4) median sediment size distribution of petite gravel, fines and super fines (Table 2) Sediment monitoring. Sediment monitoring is completed an nually or biennially in collaboration with Payette and Boise National Forests and as required by the Payette Land and Resource Management Plan (Nelson 2007) Sediment monitoring sites were established in 1966, following major flooding and mud slide events that dumped massive amounts of sediment into the river (Nelson 2010). Sites were chosen to represent South Fork Salmon River habitat based on known Chinook spawning occurrence (Nelson 2010). Monitoring provides data on sub surface sediment size distribu tion and is measured to ensure habitat quality is adequate for continued Chinook spawning and juvenile survival. Sub surface sediments are defined as any sediment found from the streambed surface to about 250 mm in depth. Monitoring sub surface sediments requires 40 core samples to be examined per site each year. Core sample locations within each site are randomly selected within superi mposed, rectangular grids. A 300 mm diameter core is driven into the gravel to an estimated depth of 250 mm and substrate is extracted, filling a 15 qt metal bucket to a standardized level. Each sample is processed in the field by straining sediment through metal sieves to separate substrate by particle size in millimeters. Contents of each metal
26 sieve are emptied into previ ously calibrated five gallon buckets filled with water. Displaced water is caught in measuring buckets, quantified with graduated cylinders and recorded as volume per individual particle size class Calculation of sediment size distribution. Sediment mo nitoring data was obtained from the United States Forest Service (2013) for all historic sediment monitoring sites surveyed between 2005 and 2012 Poverty Flats, Stolle Meadows and Dollar Creek monitoring sites (Figure 1) were chosen for analyses based on consistent methods of annual data collection from 2006 to 2012 (7 years). Although redd location data are available for evaluation from 2005, an apparent change in units of measurement from liters to milliliters during sediment monitoring occurred in 2006. Variation of measurements that can occur when rounding from milliliters to liters led to the decision of using only seven years of sediment data measured exclusively in milliliters. Five sediment size categories were organized into three classes commonl y discussed in sediment size distribution literature relative to Chinook spawni ng habitat (Table 2). The proportions of each sediment size class were calculated per sample per year for each of the three sediment sites within Excel Table 2 shows the range of each sediment size class defined for analyses.
27 Sediment site habitat and redd presence. Sediment site habitat was considered any habitat within 250 m of thalweg stream length from the initial site location coordinates, totaling 500 m of stream. Mask polygons were created for each sediment site to encompass all stream habitat within the 500 m stream length. Redd location coordinates, represented a s points, were isolated by year and spatially joined with corre sponding site polygons. Prior to determining whether sediment site habitat (i.e. stream gradient, channel width and sediment size distribution) could be related to sediment site redd presence, differences in median number of redds over seven years and b etween all sites was established via pairwise Wilcoxon tests with Holm p value adjustments (p<0.05). Sediment site habitat and stream gradient. Previously created s urvey reach segments conta ining mean gradient per 20 m were spatially joined to each sedim ent site polygon to calculate mean stream gradient. This resulted in three, 500 m site polygons with mean stream gradient attached to each. Sediment site habitat and channel width. Channel width lines were digitized manually every 50 m via visual est imat ion using 1 meter, digital orthoimagery ( Digital Orthorectified Aerial Imagery 2009), resulting in ten channel width measurements within each 500 m polygon. Channel width lines were spatially joined to site polygons, resulting in 500 m polygons with mean stream gradient (previously calculated) and mean channel width attached to each.
28 Analysis of Sediment Size Distribution S ediment site channel width and stream gradient. After establishing that differences in median redd counts exist between all sediment monitoring sites over seven years, non parametric comparisons between sediment site stream gradients and channel widths were completed via Pairwise Wilxocon tests with Holm p value adjustments. S ediment proportions between sediment sites each year. To d etermine trends and differences between sediment sites in relation to proportions of petite gravel, fine and super fine sediment, Pairwise Wilcoxon analyses were completed with a B onferroni pr obability value adjustment to 1.157e 04. This threshold was dete rmined by dividing the original alpha probability value of 0.05 by the 432 sediment size distribution tests completed for analysis.
29 S ediment proportions between years per sediment site. P airwise Wilxocon tests were completed with a B onferroni pr obabili ty value adjustment to 1.157e 04 to determine meaningful differences in median values of petite gravel, fine and super fine sediment proportions between individual years within each site. Characterizing the Top Salmon Spawning Sites Identification of the top five spawning s ites The five highest redd density patches were identified as the top five Chinook salmon spawning sites of the upper South Fork over eight years. Each site was identified by coordinate location, mapped and then characterized by redd density, elevation, average stream gradient and average channel width. Google Earth and digital orthoimagery ( Digital Orthorectified Aerial Imagery 2009) were used to describe noticeable stream bank and riparian features along with other fluvial character istics
30 CHAPTER VI RESULTS Relationships with Redd Density Summary of redd d ensit y Over eight years, total redd count per redd patch ranged from 1 to 335, with patch areas ranging from 1 76 to 7647 square meters Redd density calculations for the 545 redd patches ranged from 0.0057 to 0.3034 redds/m 2 with a median of 0.0072 redds per square meter Redd density class es are summarized in Figure 3 L ow density patches represented 65.32% of th e total and only 13.41% of the total redds used for analysis; raw redd counts ranged from 1 to 12 redds per patch. Medium density patches characterized 28.07% of the total number of redd patches and 38.47% of the total number of redds with raw redd counts ranging from 2 to 72 redds per patch. High/very high density patches represented only 7.00% of total redd patches but accounted for 48.13% of total redds; raw redd counts ranged from 8 to 355 redds per patch. T OTAL R EDD C OUNT BY D ENSITY C LASS Redd Density Redd Patch Density (redds/m2) Figure 3 illustrates quartile spreads of redd counts over eight years (2005 2012) for all patches (Total) and low medium and high/very high density patches.
31 Two redd patches with extremely high densities were identified as outliers and excluded from the high/very high density class due to their potential influence on results (refer to section 2.2.2 ) Exclusion of these two outliers resulted in the analyse s of 543 total patches with maximum redd density decreasing from 0.3034 to 0.0424 redds/m 2 ; median density for all patches and median redd counts for all redd density classes remained unchanged (Table 3 ) Density outliers included 356 redds, reducing redds used in density analyses from 4856 to 4500. Though redd density outliers did not contribute to density related statistical analyses, they were included in redd presence vs. redd absence analyses and are characterized as top spawning sites of the upper South Fo rk Table 3 summarizes r edd den sity statistics for low medium and high/very high redd density classes both with and without redd density outliers
32 Relationships with stream gradient and channel w idth When considering only raw, unclassified redd density values, a test revealed a non significant statistical relationship between stream gradient and redd density ( s (541) = 0.01, P=0.80), preventing null hypothesis rejection and indicating that the vari ables are not associated There was a statistically significant positive correlation between channel width and redd density ( s (541) = 0.13, P=0.01) C han nel width and s (541) = 0.11, P=0.03) were also associated supporting the importa nce of an interaction variable in P airwise Wilcoxon comparisons of redd density classes. Low and medium redd density classes were similar in stream gradient range (9.98 and 9.16 % respectively) and median stream gradient percentages (0.48 and 0.40 % respectively) The h igh/very high redd density class exhibited the smallest range in stream gradient percentages (6.35 % ) and the highest median stream gradient (0.65 % ) when compared to low and medium redd density classes (Figure 4) There were no statisti cally Redd Density by Class (redds/m 2 ) All Patches Low Medium High/Very High Stream Gradient (%) Figure 4 depicts quartile boxplot summaries for stream gradient percentage of redd density classes with minimum, median and maximum values reported for (1) all redd patch densities and (2) low medium and high/very high redd density classes.
33 significant differences between redd density classes w hen considering stream gradient alone (p=1.00). Low and medium redd density classes were also similar in channel width range (33.44 and 36.39 m respectively) and median channel width (15.00 and 15.70 m respectively; Figure 5 ) High/very high redd density class had a smaller channel width range (31.67 m) and the largest median channel width (19.80 m) compared to both low and medium density classes (Figure 5) When considering channel width as an independe nt variable, statistically significant differences were found between high/very high redd density and both low and medium redd density classes (p=0.003 and p=0.049 respecitively) There were no differences between low and medium redd density classes (p=0.6 92). Pairwise Wilcoxon testing also indicated that when accounting for the effects of stream gradient on channel width and the effects of channel width on stream gradient, statistically significant relationships were found between redd density classes W ith inclusion of the interaction variable [S.Gradient : C.Width], differences appeared in median Figure 5 depicts quartile boxplot summaries for channel width in meters for redd density classes with minimum, median and maximum values reported for (1) all redd patch densities and (2) low medium and high/very high redd density classes. Redd Density Clas s (redds/m 2 ) All Patches Low Medium High/Very High Channel Width (m)
34 stream gradient and median channel width between high /very high redd density patches compared to both low (p=0.002) and medium (p= 0.049) density patches Low and medium density patches were not different in median channel width or median stream gradient (p=0.793) Relationships with Redd Presence Relationships with stream gradient and channel w idth Stream gradient for t he full survey reach ranged from 0.00 to 17.31 % with a median of 0.49 percent Areas of redd presence showed a drastically smaller stream gradient range (9.98 % ) and slightly smaller median stream gradient (0.46 % ) than areas of redd absence (1 7.31 and 0.51 % respectively; Figure 6 ) The maximum stream gradient percentage for areas of redd absence was much larger than areas of redd presence (Figure 6 ) Full Survey Reach Redd Presenc e Redd Absence Stream Gradient (%) Figure 6 depicts quartile boxplot summaries for stream gradient percentages with minimum, median and maximum values reported for (1) the full survey reach and (2) redd presence and redd absence areas.
35 Channel width for the full survey reach ranged from 1.40 to 41.55 m with a median of 14.78 meters Redd presence areas exhibited a larger channel width range (37.97 m) than areas with redd absence (33.09 m) and a larger median channel width (15.75 m) than both the full survey reach (14.78 m) and a reas of redd absence (14.07 m; Figure 7 ) Minimum and maximum channel width measurements were also larger for areas of redd presence compared to areas of redd absence (Figure 7) When considering the interaction va riable [S.Gradient : C.Width], P airwis e Wilcoxon tests revealed statistically significant differences in median stream gradient and median channel width between areas of redd presence and areas of redd absence (p=0.0099) Probability of redd p resence Relative to logistical probability for redd presence, the interaction variable [S.Gradient : C.Width] was found to be a significant part of the equation ( = 0.0297, p <0.001 ), meaning that the effects of both stream gradient and channel width on the Full Survey Reach Redd Presence Redd Absence Channel Width (m) Figure 7 depicts quartile boxplot summaries for channel width in meters with minimum, median and maximum values reported for (1) the full survey reach and (2) redd presence and redd absence areas.
36 probability of redd presence are not random ( T able 4 ). Additionally, the effect of channel width on redd presence probability depend s on the value of stream gradient and the effect of stream gradient on redd presence probability depends on the value of channel width This illustrates substantive significance for both channel width and stream gradient relative to the pro bability of redd presence. Raw probability values of redd presence based o n both channel width and stream gradient ranged from 1.26 to 95.78 percent. The m aximum probability value of 95.78 % was estimated for a channel width of 36.19 m combined with a str eam gradient of 5.90 percent Minimum probability of 1.26 % was predicted for a channel width of 11.84 m with a stream gradient of 10.54 percent Probability falls below 50.00 % when median channel width drops below 10.00 m in combination with stream gradients rang ing from 0.0 0 to 17.31 percent. Higher redd presence p robability values remain consistently above 65.00 % after Table 4 represents the multiple logistic regression values and equation associated with the dependent variable of redd presence probabi lity and independent variables of stream gradient, channel width and [S.Gradient:C.Width]. Because the interaction variable [S.Gradient:C.Width] shows significant contributions to the probability equation, all other variables are not considered.
37 channel width s increase beyond 24.00 m in combination with stream gradients ranging anywher e between 0.75 and 9.16 percent. Sedimen t Site Comparisons Redd c ounts Within th e 500 m stretch of river analyzed for each sediment site between 2005 and 2012, yearly counts ranged from four to 103 redds The extremely influential outlier of 103 redds was found in 2005 at Dollar Creek and was removed decreasing years evaluated to seven (2006 2012; Figure 8 ) Differences in median redd count between all pairs of sediment sites over seven years were found to be statistically significant (p<0.03). Median redd counts for Poverty Flats, Dollar Creek and Stolle Meadows between 2006 and 2012 were 48, 28 and 7, respectively Poverty Flats redd counts were consistently higher than Dollar Creek and Stolle Meadows with the exception of 2006 when Dollar Creek had f ive more redds present (Figure 8 ) R EDD C OUNTS O VER Y EARS BY S EDIMENT S ITE Poverty Flats Dollar Creek Stolle Meadows Redd Count Year Figure 8 illustrates total redd counts per year between 2006 and 2012 (7 years) found within the 500 meter stretch of river defined for sediment site.
38 Poverty Flats and Dollar Creek show similar patterns of increasing and decreasing redd counts over time, with a substantial decrease in 2009 followed by an equally substa ntial increase in 2010 Sharp decreases in redd count between 2010 and 2012 are also present for both Poverty Flats and Dollar Creek Though Stolle Meadows has similar and corresponding trends over the years, redd counts a re consistently low and differences are minimal compared to Poverty Flats and Dollar Creek. Differences in median redd count between all pairs of sediment sites over seven years were found to be statistically significant (p<0.03) Stream g radient and c han nel w idth Differences in median stream gradient and channel width between all pairs of sediment sites are statistically significant when consi dering the interaction variable (p< 0.001 ) This means that the differences found in median stream gradient between sediment sites are dependent on the value of channel width and v ice versa. Stream gradient estimates for the three sediment sites ranged from 0.00 to 14.61 percent (Figure 9 ) Stolle Meadows had the smallest range for stream gradients, which showed 1.90 % difference between the smallest and largest estimates Dollar C reek had the largest range, showing a difference of 14.61 % while Poverty Flats range showed a Poverty Flats Dollar Creek Stolle Meadows Stream Gradient (%) Figure 9 illustrates quartile boxplot summaries with minimum, median and maximum stream gradient percentages for the 500 m stretches defined for Dollar Creek, Poverty Flats and Stolle Meadows sediment sites.
39 difference of 10.10 % between smallest and largest stream gradient estimates (Figure 9) Stolle Meadows had the largest median stream gradient of 0.80 % and Poverty Flats had the smallest median stream gradient of 0.51 percent. Channel widths for all sediment sites ranged from 5.08 m at Stolle Mead ows to 41.69 m at Poverty Flats (Figure 10 ) Median channel width at Poverty Flats was more than four times that of Stolle Meadows and only 1. 5 times that of Dollar Creek Stolle Meadows had the smallest range of channel widths, showing a difference of 13.70 m between the smallest and largest width measurement s Poverty Flats had the largest range of channel widths with a difference of 17.47 meters Additionally, the minimum channel wid th measurement at Poverty Flats was larger than the maximum channel w idth s at Dollar Creek and Stolle Meadows (Figure 10) Dollar Creek channel width measurements spanned a range of 14.13 m, which was slightly larger than the range found at Stolle Meadows. Channel Width (m) Poverty Flats Dollar Creek Stolle Meadows Figure 10 illustrates quartile boxplot summaries with minimum, median and maximum channel width measurements for the 500 m stretches defined for Dollar Creek, Poverty Flats and Stolle Meadows.
40 Sediment size d istribution Relative to sediment site comparisons, Poverty Flats showed the most differences in median proportions of sediment classes over seven years ( Table 5 ) From 2007 to 2011, Poverty Flats shows statistically significant differences in median proportion of super fine sediment compared to Dollar Creek (p< 1.157e 04 ) Compared with Stolle Meadows, statistically significant differences in super fine sediment proportions are found from 2007 to 2010 and again in 2012 (p< 1.157e 04 ) Median proportion of fine sediment was found to be different between Poverty Flats and Stolle Meadows in three different years, 2006 and 2008 2009 (p< 1.157e 04 ) The only statistically significant difference in median proportion of petite gravel for pairs of sediment site s was found in 2011 between Dollar C reek and Stolle Meadows (p< 1.157e 04 ) Stolle Meadows and Dollar Creek were not different in median proportion of any size of sediment over seven years except for fine sediment in 2006. Table 5 char acterizes instances of statistically significant differences found in median proportions of sediment size within each year and between each pair of sediment sites based on the probability value threshold of 1.157e 04. Each instance is symbolized by a shade d cell.
41 Figure 11 illustrates median proportion values out of 40 samples at each sediment size per year to demonstrate sediment proportion trends over time. A loess line (purple) and line of best fit (turquoise) are included in each scatte rplot and the 95% confidence interval is represented as dotted lines.
42 Petite Gravel trends over time do not show much overall variation, though a slight upward trend is present for Poverty Flats with very slight downward trends seen in both Dollar Cre ek and Stolle Meadows (Figure 11 ). Over seven years, a ll three sediment sites show decreasing overall trends in fine sediment and modest increasing trends fo r super fine sediment (Figure 11 ) Slight increases in fine sediment occurred between 2010 and 2011 fo llowed by slight decreases between 2011 and 2012 across all sediment sites. In opposition of fine sediment trends, super fine sediment shows overall increasing trends over seven years, with the steepest increases found between 2006 and 2007 Poverty Flat s shows extremely sharp increases in super fine sediment between 2006 and 2008 followed by a modest decrease between 2008 and 2009 Poverty Flats and Stolle Meadows show similar trends between 2006 and 2011 but deviate in direction in 2012 when Poverty Fla ts continues a steady decrease in super fine sediment proportion as Stolle Meadows increases.
43 Statistically Significant Difference (p<1.157e 04) Figure 12 demonstrates statistically significant differences in the median pro portion of sediment between each pair of years between 2006 and 2012. Poverty Flats is the only sediment site to result in statistically significant differences between years outside of comparisons with 2006. The majority of the statistically significant differences within all sediment sites and between years involved the year 2006 (Figure 12) Both fine and super fine sediment proportions in 2006 were found to be statistically different compared to all years between 2007 and 2012 for both Poverty Flats and Doll ar Creek (p<1.157e 04; Figure 12 ) Poverty Flats showed 2011 was statistically different in median proportion of petite gravel relative to all years between 2007 and 2012 while also showing differences in fine sediment propo rtions bet ween 2012 and four other years Dollar Creek was found to have three years with statistically different median petite gravel proportions. Stolle Meadows showed no changed in petite gravel over seven years with statistical differences in fine and super fine proportions in relation with 2006 and 5 other various years.
44 Top 5 Spawning Sites The top five spawning sites of the upper South Fork over eight years (2005 2012) were identified by the top five redd patch densities out of 545 total redd pa tches (Table 6, Figure 13 ) Table 6 characterizes each redd patch, or spawning site, and provides the geographic coordinates in NAD83 UTM zone 11N for the downstream and upstream locations of each site The longitudinal length of stream covered by each spa wning site was determined with a GIS by measuring the thalweg length from the downstream to the upstream location of each site. Table 6 characterizes each of the top five redd density patches determined as the top five spawnin g sit es of the upper South Fork Site one is found within the 500 m region defined for Povert y Flats sediment site (Figure 13 ) Relative to the tot al redd count over eight years it is the most heavily utilized spawning habitat of the entire upper South Fork study area Site one also has the longest thalweg length, highest average stream gradient and largest average channel width out of the top five spawning sites (Table 6) Site two is located the furth est upstream out of all five spawning sites and has the smallest average stream gradient percentage and one of the smallest thalweg lengths (Table 6) It contains the most sinuous (curvy or meandering) stream habitat compared to the other top spawning site s (F igure 13 ) Site three has the second h ighest average stream gradient with the smallest channel width, thalweg length Redd Density (redds/m 2 ) x y x y 1 335 0.3034 37.77 1.67 340 602375 4965001 602355 4964668 2 97 0.0424 19.67 0.29 82 603052 4946718 603132 4946720 3 73 0.0387 16.33 1.43 75 602902 4950982 602879 4950915 4 110 0.0370 34.18 0.35 92 602368 4965131 602376 4965039 5 264 0.0345 19.93 0.62 280 603367 4952941 603445 4952720 Downstream UTM's Upstream UTM's Site Redd Count Channel Width (m) Stream Gradient (%) Thalweg Length (m)
45 and redd count over eight years Site four is the furthermost downstream site compared to the oth er top spawning sites (Figure 13 ) It is also located within the 500 m region defined as Poverty Flats sediment site Site four has one of the smallest average stream gradient Figure 13 illustrates the location of the top five spawning sites along the upper South Fork River (right) coupled with aerial views of each of the top five spawning sites captured from Google Earth (right). Each image is overlaid with an outline of the uniquely id entified redd patch associated with the location. Flow
46 percentages with the second largest average channel width (Table 6) Site five is found within the 500 m region defined as Dolla r Creek sediment site (Figure 13 ) and contains the second most utilized spawning habitat relative to raw redd counts over eight years It has the second longest thalweg length and one of the smallest averag e stream gradients compared to the other top spawning sites.
47 CHAPTER V DISCUSSION The goals of this study were to complete site specific analyses and characterizations of main channel spawning habitat preferences for the geographically isolated and threa tened population of upper South Fork Chinook salmon. The resulting information on channel width, stream gradient and sediment size distribution could aid in decision making regarding suitability and prioritization of critical spawning habitat in the upper South Fork. Listed below are three pri mary conclusions coupled with a brief overview of their evidentiary results. Details in the following sub sections include explanations of evidentiary results, potential management implications, limitations and directi ons for future research. (1) The quality of u pper South Fork Chinook salmon spawning habitat is positively correlated with larger channel widths. When considering channel width either as an independent variable or part of the interaction term, it was consiste ntly found to be related to redd presence and redd density. When analyzing the interaction term, calculations for the probability of redd presence show that (1) probability of redd presence decreases below 50.00% with median channel widths smaller than 1 0. 00 m and (2) probability remains consistently above 65.00 % for channel widths measuring 24 m or more. (2) Upper South Fork Chinook salmon may not be optimizing by smaller stream gradients. Stream gradient as an independent variable was not related to spawning habitat and only upon taking into consideration the interaction term was stream gradient found to be significantly associated. When including the interaction term, stream gradients associated with higher redd presence probability values (>65 .00 % ) ranged an ywhere between 0.75 and 9.16 % with no discernible trends.
48 (3) Upper South Fork Chinook spawn in river locations with fine sediment volumes beyond recommended thresholds for egg and fry survival. Poverty Flats contains the most heavily utilized spawning habita t of the entire upper South Fork and therefore contains some of the most important spawning habitat for Chinook salmon. Though Chinook continuously spawn in large numbers at Poverty Flats, super fine sediment volumes are consistently above 20. 0 0 % (Figure 1 1 ), which corresponds to decreased survival rates for egg and fry according to previous research (Jensen et al. 2009) Moreover, Poverty Flats is the only sediment site continually functioning at risk or unacceptable risk (between 2006 and 2012) according to the fine sediment standards set by the Payette National Forest Monitoring and Evaluation Report (2011). Channel Width and Stream Gradient Chan nel width and chinook spawning h abitat In this study, channel width consistently exhib its associations with both redd presence and redd density, supporting the idea that channel width is an important factor in spawning site preference for upper South Fork Chinook salmon. Overall, areas with redds had significantly larger measures of channel width than areas without them (p<0.05), supporting previous assumptions about Chinook salmon preferences for larger channel a positive correlation with unclassified values of redd density ( s (541) = 0.13), p= 0.01), showing that the total number of redds per redd patch increased with channel width. In other words, larger channel widths are associated with repeated spawning in the same locations over the eight year study period,
49 suggesting that large channel wid th is an important factor for spawning habitat suitability in the South Fork. Because redd density is defined here as the total number of redds over eight years at the same location, this study suggests that the consistent use of a relatively few specifi c locations by the majority of spawning Chino ok is related to habitat suitability. Only 7.00% of the total redd patches defined by this study were characterized as high/very high redd density ( 0.0200 redds /m 2 ), yet this small percentage of redd patches e ncompassed almost 50.00 % of the total redds analyzed (Table 3) Median channel width for low and medium redd density areas were comparable to the median channel width of the entire survey reach while median channel width for high/very high redd density ar eas was much larger ( Figure 7 ). Considering that high/very high redd density patches contain almost half of the redds over eight years and channel widths that are (1) substantially larger than median channel width for the full survey reach and (2) signific antly larger than low and medium redd density patches (p<0.05), the results suggest that upper South Fork Chinook salmon are consistently optimizing with large channel widths as a primary factor. Stream gradient as an independent v ariable In this study, stream gradient did not show consistent patterns of association with Chinook spawning habitat. As previously noted, stream gradient is a critical variable in most characterizations of Chinook spawning habitat, which generally state that smaller stream gra dients are preferable (Isaak & Thurow 2006; Montgomery et al. 1999; Platts et al. 1989) Median stream gradient percentages for spawning habitat were be low 1. 0 0 % and while these percentages agree with previous literatures that state Chinook prefer smaller gradients, a significant relationship with spawning habitat was not supported. With stream significant statistical
50 s (541) = 0.0 1, P=0.80) and no relationship was revealed between redd density classes (p=1.00). A likely explanation for the lack of association with stream gradient as an independent variable is that the median stream gradient for the entire survey reach was also cal culated below 1. 0 0 percent (Figure 6). This might suggest that the stable and consistently low stream gradient of the South Fork already provides an optimal gradient for spawning allowing the upper South Fork Chinook to optimize by other factors, such as channel width. Stream gradient and channel width i nteraction Considering the intimate relationship between upper South Fork stream gradient s and channel width s (p<0.05) and the well supported idea that as stream gradients decrease, channel widths widen, it was important to incorporate an i nteraction term into analyse s. There was a statistically significant relationships between redd presence and redd density after the interaction term was incorpora ted. The relationships for channel width did not change, but there were significant associations found with stream gradient which is in contrast to the model without the interaction term. When considering interaction, differences were found in median strea m gradient and median channel width between (1) areas of redd presence and redd absence (p=0.001), and (2) high/very high redd density patches compared to both low (p=0.002) and medium (p=0.049) redd density patches. The logistic regression analysis suppo rts the idea that upper South Fork Chinook have a strong preference for wider channel widths for spawning sites. The model revealed that the probability of redd presence depends on values of both stream grad ient and channel width (p<.001). H owever, the max imum probability value calculated for redd presence was 95.78 % which was associated with a median channel width of 36.19 m, combined with a
51 median stream gradient of 5.90 percent. This stream gradien t is remarkably larger than 1 .00 % which is contrary to traditional notions about Chinook salmon optimizing largely by stream gradient The probability of redd presence dropped consistently below 50.00 % with median channel widths of 10 m or less These low probability values were associated with median stream gradients between 0.40 and 17.31 percent. The probability of redd presence remained consistently above 65.00 % among channel widths larger than 24.00 m and median stream gradients associated with these h igh percentages varied from 0.75 to 9.16 percent Ther e was a strong and steady relationship between spawning sites and channel width, while varying degrees of stream gradient did not show consistent patterns relative to the probability of redd presence. Channel width and stream g radient at s ediment monitori ng s ites Within this study, s tream gradient and channel width have been shown to affect redd density and the probability of redd presence in the same locations over time when considered jointly through an interaction term (p<0.05). Observed differences i n median redd count over time combined with differences found in median stream gradient and median channel width between sediment sites (p<0.001) suggest s partial influence from these two factors on the prevalence of Chinook spawning. Within the upper So uth Fork survey reach, Poverty Flats and Dollar Creek sediment sites contain the most highly utilized spawning habitat, and therefore the most important habitat for Chinook salmon. Relative to the 500 m reach defined for each sediment site, Poverty Flats c ontains the redd patch with the largest total redd count (335) and highest redd density (0.3034 redds/m 2 ) out of 545 total redd patches identified over eight years. It also contains the highest median redd count over seven years compared to all three sedim ent sites analyzed for this study (p<0.03). Dollar Creek sediment site contains the
52 redd patch with the second largest total redd count (264) and fifth largest redd density (0.0345 redds/m 2 ) out of 545 total redd patches over eight years. Dollar Creek also comprises the second highest median redd count over seven years compared with all three sediment sites (p<0.03). Median stream gradient for each sediment site was smaller than 1. 0 0 % though the largest median percentage was estimated for Stolle Meadows, which had consistent and extremely low redd counts each year. However, because the range of stream gradients for both Poverty Flats and Dollar Creek were substantially larger than the range of stream gradients for Stolle Meadows, it can again be inferred that the Chinook salmon of the upper South Fork are optimizing by factors other than stream gradient. Poverty Flats sediment site encompassed channel width measurements consistently larger than any channel width measurements in Dollar Creek or Stolle Me adows, resulting in a substantially larger median channel width of almost 31 meters. channel width that was smaller than 10 meters. These channel width relationships, co upled with the quantification of spawning incidence occurring within each site, emphasizes this previous finding that probability of redd presence consistently remains above 65. 0 0 % with channel widths larger than 24 m while dropping below 50. 0 0 % wi th channel widths smaller than 10 m after accounting for interaction. Limitations of Channel Width and Stream Gradient A nalyses Data collection and analyses. Datasets used for this study were invaluable, as they provided extensive and rich information allowing for a sound characterization of spawning habitat. In spite of this, analyses could have been slightly skewed due to the absence of in field averaging of GPS
53 locations during redd surveys The 15 m locational error that was assumed for this study is of in field averaging could increase the positional accuracy of each re dd and allow for a finer scaled characterization for upper South Fork Chinook spawning habitat. The lack of an association with stream gradient for spawning South Fork Chinook could be attributed to the biased selection of areas surveyed annually by the IDFG and NPT. Some stretches along the upper South Fork are not surveyed due to access difficulty and an assumption that the habitat is unsuitable for spawning Chinook salmon. The absence of data from these reaches may have skewed stream gradient results. Inconsistent relationships with stream gradient could have also resulted from the low resolution of the DEM. M edian stream gradient for the e ntire survey reach was based on DEM derived stream gradient calculations along a thalweg line which was created via visual estimation from the 1 m resolution digital orthoimagery The aerial photography georeferenced by the NPT Fisheries was cross checked with Google Earth and found to provide an accurate representation of the South Fork Salmon River Valley, suggest ing accurate thalweg line placement in this study. However, the 10 m resolution DEM provided by the USGS is part of a historical series created from remotely sensed data combined with ground triangulation from field crews DEMs created in this way do not p rovide as much vertical accuracy as other technologies such as light detection and ranging (LIDAR). Elevation models created from LIDAR would have higher resolutions which would be comparable to the digital orthoimagery and a better match for processing a nd analyses. D ue to issues with vertical accuracy during processing of stream gradient vertices along the initial thalweg line were manually adjusted to follow the smallest gradients alon g the DEM
54 The low resolution of the DEM combined with manual adjustments and manipulations cessing of stream gradient certainly created potential for error. D ue to naturally existing variability in channel width along the upper South Fork, there are stretches that are narrower th an 10 m, which means steeper bank elevations may have also skewed gradient estimates. S tream gradient estimates from a lower resolution DEM could result in unrealistically high stream gra dient percentages due to stream bank influences. In addition, l arge b oulders, spanners (fallen tree s that span the width of the river channel), islands and log jams that sit above the surface of the stream are frequently found along the South Fork The existence of such features would cause DEM elevation values within the s surface altering stream gradient estimates. Though extremely high stream gradients are unrealistic for the upper South Fork river channel, estimates based on DEMs provide criti cal clues into potential rela tionships with spawning habitat There may be enough evidence to support funding more accurate elevation data collection (i.e. LIDAR) given the critical relationships commonly demonstrated in previous research of salmon spawnin g habitat and stream gradients, as possible large r range of tolerance for stream gradients among upper South Fork Chinook salmon. The attempt to track down meta data for the image failed, as it was purchased by the NPT a nd meta data information was misplaced in the process. This means that channel widths estimated from the aerial imagery could have been either larger or smaller depending on seasonal flow fluctuations. Without knowing the precise time of year that the imag e was taken, it is important to consider that the wetted channel in the image may or may not be what is characteristic of the stream channel during peak spawning time.
55 As mentioned in the methods, six problems arose within the 49 km survey reach due to me andering curves in the stream and the effects of an 80 m analysis buffer. Overlapping stream gradient values slightly influence d stream gradient calculations with four low density redd patches and two medium density patches. Manual calculations of stream g radients showed that the four low density redd patch stream gradients were only skewed between 0.10 and 0.15 % while the two medium density patches were skewed by 0.04 and 0.20 percent. Given that the total number of skewed stream gradient calculations are slight compared to the total number of patches analyzed, the effects of this were negligible for the overall study results. The complexity of nature. The weak rho value resulting from the relationship between redd density and channel width ( s (541) = 0.13) does not mean that the influence of channel width is inconsequential. There is a vast range of interconnected variables that influence salmon spawning potential, such as conditions of fry emer gence, sea bound migration, oceanic maturation an d the return back to natal streams for reproduction. These analyses constitute a coarse ly scaled examination of the effects of two variables that are easil y measured with GIS software o ve r landscape level distances Consequently, other naturally occurring variables thought to influence spawning habitat preference for Chinook salmon were not accounted for. These variables include a variety of factors such stream flow, sediment size distribution and the presence of large woody debris for shade and shelter. Ot her variables potentially e ffecting spawning habitat preference that are not taken into account for this study include (1) healthy riparian vegetation, which is associated with water temperature regulation, water quality and bank stability (Dosskey et al., 2010; Osborne & Kovacic, 1993) (2) variability in sinuosity and riffle pool development (Fukushima, 2001;
56 Montgomery et al., 1999; Tonina & Buffington, 2009) and (3) nutrient availability for future emerging fry in the form of primary producer abundance (Cram et al., 2011; Maier & Simenstad, 2009; Schuldt & Hershey, 1995) provided small clues into the otherwise boundless realm of influences on Chinook salmon abundances and productivity. The presence of hatcher y Chinook salmon in the upper South Fork may have influenced redd patch density estimates. Hatchery Chinook salmon on the South Fork are introduced and imprinted upstream of Stolle Meadows and tend to return to the same area. Returns can number over 1,000 each year which could influence the density of spawning Chinook in that location. Although this artificial introduction into the stream network may influence density, salmon should still spawn within the habitat parameters that are optimal for survival an d persistence of the species and the spawning site preferences of hatchery salmon should be similar to wild salmon. Sediment Size Distribution Spawning habitat and fine s ediment It is understood among fisheries biologists and managers that there is a sig nificant negative correlation between fine sediment presence and egg/fry survival ( Sternecker et al., 2013 ; Johnson et al., 2012 ; Louhi Ovaska, Maki Petays, Erkinaro & Muotka, 2011 ; Sternecker & Geist, 2010 ; Hanrahan, Geist, & Arntzen, 2005 ; Coulombe Pont briand & LaPointe, 2004 ). According to Jensen et al. (2009) and the limits set in the Payette National Forest Monitoring and Evaluation Report (2011), Poverty Flats sediment site is the only historic site that has been functioning either at risk or unaccep table risk over the past seven years. It not only has significantly larger volumes of fine and super fine sediment when compared with Stolle Meadows and Dollar Creek, but it is also the most highly utilized
57 spawning habitat of the entire upper South Fork. Existing as the most important habitat for spawning Chinook salmon and also the most unacceptable site relative to fine sediment volumes appears contrary to previous literatures which claim f ine sediment s degrade spawning habitat suitability This could me an upper South Fork Chinook have a larger range of fine sediment tolerance than previously considered or that they may be spawning within higher fine sediment areas in order to optimize by other factors, such as channel width. F urther research would be nec essary to support either theory. It is likely that the South Fork River Valley naturally exists with higher levels of fine sediment due to its shallow, unconsolidated soils and long history of frequent flooding and fires. Both of these events are known to increase erosion and therefore increase presence of fine sediment within river channels ( Benda, Miller, Bigelow & Andras, 2003; Me yer, Pierce, Wood & Jull, 2001 ). Thousands of ye ars of flooding and fire in the South Fork River Valley would likely create a myriad of site specific evolutionary trai ts to resist high fine sediment pr esence, ultimately driving specialized adapta tions for local Chinook salmon. Potential support for this can be seen in 2008 when Poverty Flats had the highest volume of super fin e sediment and the second highest spawning occurrence over the seven year study period (Figure 11). Again, this theory is not conclusive by any means and the recomme ndations for further research are described in the next section. Consistent trends in redd frequency over time were not observed in relation to sediment distribution (Figures 11 and 12) However, noticeable decreases in redd presence occurred at both Poverty Flats and Dollar Creek between 2010 and 2011 (Figure s 8 and 11 ). A flooding event cause d by an isolated rain on snow event occurred in 2010, likely causing the fine sediment increases seen across all sediment sites between 2010 and 2011 (Figure 11). Though the Payette Nation Forest did not find deleterious intra gravel effects to have
58 occurr ed, increased flow and erosion could have altered other stream features considered important for spawning habitat suitability, such as undercut bank frequency or pool and riffle structure. Limitations of Sediment Size Distribution A nalyses Data collection and analyses. Locations for sediment monitoring sites were established to represent South Fork Salmon River habitat based on locations of known Chinook spawning occurrence (Nelson 2010). This means that data collection does not provide a random sample, necessary for describing spawning habitat sediment distribution for the entire upper South Fork survey reach. Additionally, grids for core sampling may migrate over the years up to tens of meters in distance, depending on visual estimation of intra gravel conditions according to veteran fisheries managers. Further sampling bias may occur within each sediment site because core samples are randomly taken from grids that are purposefully placed in locations that seem ide al for Chinook salmon spawning. Regardless of location al bias concern s two of the sediment sites are the most highly utilized spawning areas on the upper South Fork and therefore contribute critical information associated with habitat suitability. In addition, t he resulting information convey s insights about Chinook spawning which may be used as a basis for future research regarding egg/fry survival and fine sediment tolerance thresholds. The complexity of nature. The South Fork river valley is naturally characterized by narrow valleys steep slopes and shallow, unconsolidated soils (Platts et al. 1989), which create highly erodible conditions. Frequent fire and flooding events can further amplify susceptibility of erosion and introduce greater proportions of fine sediment into the Sou th Fork than what is naturally present. In addition to the flooding events of 2008 and 2010, multiple fires
59 occurred over the study period. About 53 .00 % (441,771 acres) of the South Fork watershed burned in 2006 and 2007 (Bonaminio & Nelson 2012). The ext reme loss of vegetation due to these events created extreme conditions conducive to increased soil erosion that likely altered fine sediment volumes in the stream, especially at the lowest sediment site ( Po verty Flats) However re vegetation has been stea dy and intra gravel conditions in the watershed during post fire monitoring were not shown to have been negatively affected (Bonaminio & Nelson, 2012). F actors described in section 5.1.5, such as survival during migration and maturation, could have also a ttributed to the changes in frequency of redds at sediment sites over time
60 CHAPTER VI CONCLUSIONS AND RECOMMENDATIONS the numbers are abstracted from the living salmon, the salmon are abstracted from their habitat, the habitat is abstracted from the river, and the river is abst racted from the ecosystem. Jim Lichatowich (1999) Imagine a female Chinook as she glides effortlessly from here to there and back again. Eac h movement graceful and purposeful as she tests the streambed for a suitable location to create her redd as is demanded by her inherent instinct to reproduce. As she glides a little further upstream, two male Chinook begin to fight for their place, bumpin g her a little which forces her to alter her current position. As she continues to test from here to there, another male migrates into the reach and upon attempting to swim up to her, is chased away by one of the other males currently swarming around the lone female. As with any scientific study regarding the elements of nature, it is important to maintain awareness of the abstractions that occur during each step in the process. From observation and initial data collection to defining values as belonging t o specific bins for statistical model analyses and interpretation. These abstractions do not perfectly exemplify the reality of a female allow for the true complexity of nature to be considered. Nevertheless, essential informa tion about spawning habitat suitability can be extracted through crit ical examination of data analyse s results, as this study has done. Channel Width and Stream Gradient Channel width as an indicator. An association with channel width was consistently dem onstrated between Chinook presence and redd density, while stream gradient was not observed to be a significant
61 factor until it was included as part of an interaction term. Considering s tream gradient does not vary drastically within the upper South Fork, the lack of an association with spawning habita t preference is not surprising as the low median stream gradient may already be considered optimal for spawning However, stream gradient results may have been skewed by limitations resulting from the low reso lution of the DEM. A better option for future research into the stream gradient of the South Fork would include finer resolution elevation data from LIDAR technology and the ability to ground truth stream gradient calculations made in a GIS. Ground truthin g is especially necessary when erroneously high stream gradient measurements are calculated which are not typically found within a watershed, such as what was experienced in this study. Whether considered independently or as part of the interaction variable, there is no consistent association with stream gradient on spawning habitat for upper South Fork Chinook salmon. Though the median st ream gradient is less than 1.00 %, the range of str eam gradients associated with high values of redd presence probability could imply a larger range of acceptable stream gradients for spawning habitat on the upper South Fork than previously considered. Further consideration of the potential presence of a l arger range of tolerance for stream gradients could guide management decisions that focus on more influential factors for this area, such as channel width. As expected, c hannel width was consistently found to be a significant factor in spawning habi tat su itability. This suggests that South Fork Chinook salmon populations are optimizing by channel width as a primary factor As a recommendation, i t may benefit stakeholders involved in South Fork Chinook salmon management to consider emphasizing channel width as a principle component when identifying the top critical spawning sites for remediation.
62 Sediment Distribution Further research into egg/fry survival. One of the most surprising findings in this study was the fact that Poverty Flats contains the most heavily utilized spawning habitat of the entire upper South Fork while consistently containing super fine sediment volumes which correspond to decreased egg and fry survival. Considering the fine sediment levels present at this location also correspond wit h unacceptable risk levels for the Payette Land and Resource Management Plan and that the current recovery strategy outlined by NMFS (2012) has the top priority of reducing fine sediment delivery into the upper South Fork f urther research into the actual survival rates of egg and fry is recommended to make sure current management o bjectives are accurate for this population of Chinook Egg and fry survival r esearch would be necessary for supporting an association between fine sediment tolerance and upper South Fork Chinook. If future research finds acceptable rates of egg and emerging fry survival at Poverty Flats, regardless of fine sediment persi stence, a potential increase in sediment threshold limits previously designated for upper South Fork Chinook m ay be warranted These findings would support the possibility that upper South Fork Chinook salmon have evolved a tolerance for fine sediment, specific to the geomorphological history and characteristics of the South Fork Salmon River Valley. However, if f urther research finds decreased rates of egg and fry survival in correlation to the persistence of large volumes of fine sediment at Poverty Flats, immediate implementation of actions to remediate Poverty Flats sediment site would be a crucial step to supp ort an increase in upper South Fork Chinook abundance
63 Enhancing core sample procedures. Recommendations for future placement of core sample grids at sediment monitoring sites include increasing the distribution of sediment monitoring sites and utilizing redd coordinate data to identify the most highly utilized spawning areas within the stream channel By completing a fuzzy analysis on the redd locations, a probability map could be created which would show the likelihood of a redd being construct ed at locations a long the river. This would help fisheries managers understand the connectivity of spawning habitats, as well as allow for refined placement of core sample grids. This is especially important considering that a single stretch of the South F ork can contain a drastically dynamic sediment distribution along its longitudinal profile, as well as from left bank to right bank. For example, fine sediments are typically present in larger proportions along the inside bend of a meander when compared to the outside bend. This same difference in sediment distribution can also occur due to other fluvial features not related to sinuosity, such as large boulders or backwater channels. Increasing the random distribution of sediment sampling while also s amplin g at heavily utilized spawning locations within the stream channel could improve accuracy of data and help to reinforce management d ecisions based on fine sediment levels in the entire upper South Fork.
64 LITERATURE CITED Anlauf, K. J., Jensen, D. W., Burn ett, K. M., Steel, E. A., Christiansen, K., Firman, J. C., . Larsen, D. P. (2011). Explaining spatial variability in stream habitats using both natural and management influenced landscape predictors. Aquatic Conservation Marine and Freshwater Ecosystem s, 21 (7), 704 714. doi: 10.1002/aqc.1221 Armstrong, J. D., Kemp, P. S., Kennedy, G. J. A., Ladle, M., & Milner, N. J. (2003). Habitat requirements of Atlantic salmon and brown trout in rivers and streams. Fisheries Research, 62 (2), 143 170. doi: 10.1016/s0 165 7836(02)00160 1 Benda, L., Miller, D., Bigelow, P., & Andras, K. (2003). Effects of post wildfire erosion on channel environments, Boise River, Idaho. Forest Ecology and Management, 178 (1 2), 105 119. doi: 10.1016/s0378 1127(03)00056 2 Benda, L., Poff, N. L., Miller, D., Dunne, T., Reeves, G., Pess, G., & Pollock, M. (2004). The network dynamics hypothesis: How channel networks structure riverine habitats. Bioscience, 54 (5), 413 427. doi: 10.1641/0006 568(2004)054[0413:tndhhc]2.0.co;2 Bidlack, A. L., Be nda, L. E., Miewald, T., Reeves, G. H., & McMahan, G. (2014). Identifying Suitable Habitat for Chinook Salmon across a Large, Glaciated Watershed. Transactions of the American Fisheries Society, 143 (3), 689 699. doi: 10.1080/00028487.2014.880739 Bjornn, T .C. and Resier, D.W. (1991). Habitat requirements of salmonids in streams. American Fisheries Society Spcial Publication 19:83 138. Brown, R. A., & Pasternack, G. B. (2008). Engineered channel controls limiting spawning habitat rehabilitation success on regulated gravel bed rivers. Geomorphology, 97 (3 4), 631 654. doi: 10.1016/j.geomorph.2007.09.012 Brown, R. A., & Pasternack, G. B. (2009). Compariso n of methods for analysing salmon habitat rehabilitiation designs for regulated rivers. River Research and Applications, 25 (6), 745 772. doi: 10.1002/rra.1189 Burns, D.C. and R.L. Nelson ( 2005 ) Developing appropriate sediment related watershed condition i ndicators for national environmental policy act analyses and biological assessments in the South Fork Salmon River Basin. Unpublished Report. McCall, ID: US Department of Agriculture, Forest Service, Payette National Forest. 101p. Busch, D. S., Sheer, M., Burnett, K., McElhany, P., & Cooney, T. (2013). Landscape level model to predict spawning habitat for lower columbia river fall Chinook salmon ( Oncorhynchus tshawytscha). River Research and Applications, 29 (3), 297 312. doi: 10.1002/rra.1597 Cederholm, C. Jeff, Matt D. Kunze, Takeshi Murota, and Atuhiro Sibatani (1999) Salmon Carcasses: Essential Contributions of Nutrients and Energy for Aquatic and Fisheries. 24, no. 10 : 6 15.
65 Cooney, T., McClure, M., Baldwin, C., Carmichael, R., Hassemer, P. Howell, P., McCullough, D.,Schaller, H., Spruell, P., Petrosky, C., and Utter, F. ( 2007 ) Viability Criteria for Application to Interior Columbia Basin Salmonid ESUs. U.S. Department of Com merce. 90 p. http://www.nwfsc.noaa.gov.ezproxy.library.uvic.ca/trt/trt_documents/ictrt_viability_criter ia_reviewdraft _2007_complete.pdf Coulombe Pontbriand, M., & LaPointe, M. (2004). Geomorphic controls, riffle substrate quality, and spawning site selection in two semi alluvial salmon rivers in the Gaspe Peninsula, Canada. River Research and Applications, 20 (5), 577 590. doi: 10.1002/rra.768 Cram, J. M., Kiffney, P. M., Klett, R., & Edmonds, R. L. (2011). Do fall additions of salmon carcasses benefit food webs in experimental streams? Hydrobiologia, 675 (1), 197 209. doi: 10.1007/s10750 011 0819 9 Dauble, D. D., & Geist, D. R. (2000). Comparison of mainstem spawning habitats for two populations of fall chinook salmon in the Columbia River basin. Regulated Rivers Research & Management, 16 (4), 345 361. doi: 10.1002/1099 1646(200007/08)16:4<345::aid rrr577>3.3.co;2 i D auble, D. D., & Geist, D. R. (2000). Comparison of mainstream spawning habitats for two populations of fall Chinook salmon in the Columbia River basin. Regulated Rivers: Research & Management, 16 (4), 345 361. doi: 10.1002/1099 1646(200007/08)16:4<345::AID RRR577>3.0.CO;2 R Dauble, D.D. and D.G. Watson. 1990. Spawning and Abundance of Fall Chinook Salmon (Oncorhynchus tshawytscha) in the Hanford Reach of the Columbia River, 1948 1988. PNL 7289, Pacific Northwest Laboratory, Richland, Washington. Dodds, Walte r K (2002) ., Freshwater Ecology: Concepts and Environmental Applications: Academic Press 452 455. Dosskey, M. G., Vidon, P., Gurwick, N. P., Allan, C. J., Duval, T. P., & Lowrance, R. (2010). The Role of Riparian Vegetation in Protecting and Improving Che mical Water Quality in Streams1. Journal of the American Water Resources Association, 46 (2), 261 277. doi: 10.1111/j.1752 1688.2010.00419.x ESRI. (2013). ArcGIS Desktop: Release 10.1. Redlands, CA: Environmental Systems Research Institute. Finney, B. P., Gregory Eaves, I., Sweetman, J., Dougas, M. S. V., & Smol, J. P. (2000). Impacts of climatic change and fishing on Pacific salmon abundance over the past 300 years. Science, 290 (5492), 795 799. doi: 10.1126/science.290.5492.795 Fukushima, M. (2001). Salmonid habitat geomorphology relationships in low gradient streams. Ecology, 82 (5), 1238 1246. Geist, D. R., Hanrahan, T. P., Arntzen, E. V., McMichael, G. A., Murray, C. J., & Chien, Y. J. (2002). Physicochemical characteristics of the hyporhe ic zone affect redd site selection by
66 chum salmon and fall chinook salmon in the Columbia River. North American Journal of Fisheries Management, 22 (4), 1077 1085. doi: 10.1577/1548 8675(2002)022<1077:pcothz>2.0.co;2 Gresh, T., Lichatowich, J., & Schoonmake r, P. (2000). An estimation of historic and current levels of salmon production in the Northeast Pacific ecosystem: Evidence of a nutrient deficit in the freshwater systems of the Pacific Northwest. Fisheries, 25 (1), 15 21. doi: 10.1577/1548 8446(2000)025< 0015:aeohac>2.0.co;2 British Columbia Press. Hall, J. E., Holzer, D. M., & Beechie, T. J. (2007). Predicting river floodplain and lateral channel migration fo r salmon habitat conservation. Journal of the American Water Resources Association, 43 (3), 786 797. doi: 10.1111/j.1752 1688.2007.00063.x Hanrahan, T. P. (2007). Bedform morphology of salmon spawning areas in a large gravel bed river. Geomorphology, 86 (3 4 ), 529 536. doi: 10.1016/j.geomorph.2006.09.017 Hanrahan, T. P., Geist, D. R., & Arntzen, E. V. (2005). Habitat quality of historic snake river fall Chinook salmon spawning locations and implications for incubation survival. part 1: Substrate quality. Rive r Research and Applications, 21 (5), 455 467. doi: 10.1002/rra.823 Helfield, J. M., & Naiman, R. J. (2001). Effects of salmon derived nitrogen on riparian forest growth and implications for stream productivity. Ecology, 82 (9), 2403 2409. doi: 10.1890/0012 9 658(2001)082[2403:eosdno]2.0.co;2 Healey, M. C. ( 1991 ) Life History of Chinook Salmon (Oncorhynchus tshawytscha) in Pacific Salmon Life Histories. C. Groot and L. Margolis. (ed.), Vancouver B.C.: UBC Press, 311 393. Hicks, B. J., Wipfli, M. S., Lang, D. W ., & Lang, M. E. (2005). Marine derived nitrogen and carbon in freshwater riparian food webs of the Copper River Delta, southcentral Alaska. Oecologia, 144 (4), 558 569. doi: 10.1007/s00442 005 0035 2 Honea, J. M., Jorgensen, J. C., McClure, M. M., Cooney, T. D., Engie, K., Holzer, D. M., & Hilborn, R. (2009). Evaluating habitat effects on population status: influence of habitat restoration on spring run Chinook salmon. Freshwater Biology, 54 (7), 1576 1592. doi: 10.1111/j.1365 2427.2009.02208.x Isaak, D. J., & Thurow, R. F. (2006). Network scale spatial and temporal variation in Chinook salmon (Oncorhynchus tshawytscha) redd distributions: patterns inferred from spatially continuous replicate surveys. Canadian Journal of Fisheries and Aquatic Sciences, 63 (2), 285 296. doi: 10.1139/f05 214 Idaho Fish and Game. (1985). Anadromous fish management plan 1985 1990. IDFG, Boise. Available: https://research.idfg.idaho.gov (February 2012)
67 Jensen, D. W., Steel, E. A., Fu llerton, A. H., & Pess, G. R. (2009). Impact of Fine Sediment on Egg To Fry Survival of Pacific Salmon: A Meta Analysis of Published Studies. Reviews in Fisheries Science, 17 (3), 348 359. doi: 10.1080/10641260902716954 Johnson, C. L., Roni, P., & Pess, G. R. (2012). Parental Effect as a Primary Factor Limiting Egg to Fry Survival of Spring Chinook Salmon in the Upper Yakima River Basin. Transactions of the American Fisheries Society, 141 (5), 1295 1309. doi: 10.1080/00028487.2012.690815 Johnson, S. P., & Sch indler, D. E. (2009). Trophic ecology of Pacific salmon (Oncorhynchus spp.) in the ocean: a synthesis of stable isotope research. Ecological Research, 24 (4), 855 863. doi: 10.1007/s11284 008 0559 0 Kondolf, G. M., & Wolman, M. G. (1993). The sizes of salmo nid spawning gravels. Water Resources Research, 29 (7), 2275 2285. doi: 10.1029/93wr00402 Leopold Aldo. (1948). A Sand County Almanac and Sketches Here and There. Oxford University Press, New York, 1987. Leth, Cassinelli and Knipper, unpublished works. South Fork Salmon River Summer Chinook Salmon. Idaho Department of Fish and Game, Nampa, ID. h ttp://www.fws.gov/lsnakecomplan/Meetings/2010%20Spring%20Chinook%20Symposi um/Written%20Reports%20&%20ISRP%20Comm ents/South%20Fork%20Salmon%20Ri ver%20Summer%20Chinook%20Salmon.pdf Louhi, P., Ovaska, M., Maki Petays, A., Erkinaro, J., & Muotka, T. (2011). Does fine sediment constrain salmonid alevin development and survival? Canadian Journal of Fisheries and Aquatic S ciences, 68 (10), 1819 1826. doi: 10.1139/f2011 106 Maier, G. O., & Simenstad, C. A. (2009). The Role of Marsh Derived Macrodetritus to the Food Webs of Juvenile Chinook Salmon in a Large Altered Estuary. Estuaries and Coasts, 32 (5), 984 998. doi: 10.1007/s 12237 009 9197 1 Meyer, G. A., Pierce, J. L., Wood, S. H., & Jull, A. J. T. (2001). Fire, storms, and erosional events in the Idaho batholith. Hydrological Processes, 15 (15), 3025 3038. doi: 10.1002/hyp.389 McKean, John R. and R.G. Taylor. (2000) Fi shing Use and Value: Snake River Basin of Idaho Experiment Station Bulletin Univ ersity of Idaho, Moscow, Idaho Mitchell, N. L., & Lamberti, G. A. (2005). Responses in dissolved nutrients and epilithon abundance to spawning salmon in southeast Alaska streams. Limnology and Oceanography, 50 (1), 217 227. Montgomery, D. R., Beamer, E. M., Pess, G. R., & Quinn, T. P. (1999). Channel typ e and salmonid spawning distribution and abundance. Canadian Journal of Fisheries and Aquatic Sciences, 56 (3), 377 387. doi: 10.1139/cjfas 56 3 377 Montgomery, D.R. (2000) Geology : 28, no 12.
68 Montgomery, D.R. (2004). GSA Today 14, no.11 Naiman, R. J., Bilby, R. E., Schindler, D. E., & Helfield, J. M. (2002). Pacific salmon, nutrients, and the dynamics of freshwater and riparian ecosystems. Ecosystems, 5 (4), 399 417. doi: 10.1007/s10021 001 0083 3 Nelson, Roger L.(2010). "Deposition of Fine Sediment in the Salmon River Watershed, Payett e and Boise National Forests, Idaho Statistical Summary of Intragravel Monitoring, 1975 2009." edited by United States Department of Agriculture (USDA). McCall, Idaho: Payette National Forest. Neville, H. M., Isaak, D. J., Dunham, J. B., Thurow, R. F., & R ieman, B. E. (2006). Fine scale natal homing and localized movement as shaped by sex and spawning habitat in Chinook salmon: insights from spatial autocorrelation analysis of individual genotypes. Molecular Ecology, 15 (14), 4589 4602. doi: 10.1111/j.1365 2 94X.2006.03082.x Nez Perce Tribe Department of Fisheries Management. (2012). Nez Perce Tribe: McCall, Idaho. Year Review: Summary & Evalua tion of Snake River Sockeye, Snake River Spring Summer Chinook, Snake River Fall Run Chinook, Snake River Basin Portaland Oregon. Osborne, L. L., & Kovacic, D. A. (1993). Rip arian vegetated buffer strips in water quality restoration and stream management. Freshwater Biology, 29 (2), 243 258. doi: 10.1111/j.1365 2427.1993.tb00761.x Platts, W. S., Torquemada, R. J., McHenry, M. L., & Graham, C. K. (1989). Changes in salmon spawni ng and rearing habitat from increased delivery of fine sediment to south fork salmon river, Idaho. Transactions of the American Fisheries Society, 118 (3). doi: 10.1577/1548 8659(1989)118<0274:cissar>2.3.co;2 Powell, D. M., Laronne, J. B., Reid, I., & Barzi lai, R. (2012). The bed morphology of upland single thread channels in semi arid environments: evidence of repeating bedforms and their wider implications for gravel bed rivers. Earth Surface Processes and Landforms, 37 (7), 741 753. doi: 10.1002/esp.3199 Q uinn, T. P., Stewart, I. J., & Boatright, C. P. (2006). Experimental evidence of homing to site of incubation by mature sockeye salmon, Oncorhynchus nerka. Animal Behaviour, 72 941 949. doi: 10.1016/j.anbehav.2006.03.003
69 Ramstad, K. M., Woody, C. A., & Al lendorf, F. W. (2010). Recent local adaptation of sockeye salmon to glacial spawning habitats. Evolutionary Ecology, 24 (2), 391 411. doi: 10.1007/s10682 009 9313 5 Rollinson, N., & Hutchings, J. A. (2011). Body size specific maternal effects on the offspri ng environment shape juvenile phenotypes in Atlantic salmon. Oecologia, 166 (4), 889 898. doi: 10.1007/s00442 011 1945 9 Rosenfeld, J. (2003). Assessing the habitat requirements of stream fishes: An overview and evaluation of different approaches. Transacti ons of the American Fisheries Society, 132 (5), 953 968. doi: 10.1577/t01 126 Schindler, D. E., Scheuerell, M. D., Moore, J. W., Gende, S. M., Francis, T. B., & Palen, W. J. (2003). Pacific salmon and the ecology of coastal ecosystems. Frontiers in Ecology and the Environment, 1 (1), 31 37. doi: 10.2307/3867962 Schuldt, J. A., & Hershey, A. E. (1995). Effect of salmon carcass decomposition on lake superior tributary streams. Journal of the North American Benthological Society, 14 (2), 259 268. doi: 10.2307/146 7778 Sheer, M. B., & Steel, E. A. (2006). Lost watersheds: Barriers, aquatic habitat connectivity, and salmon persistence in the Willamette and Lower Columbia River basins. Transactions of the American Fisheries Society, 135 (6), 1654 1669. doi: 10.1577/t05 221.1 Shirazi, M. A., & Seim, W. K. (1981). Stream system evaluation with emphasis on spawning habitat for Salmonids. Water Resources Research, 17 (3), 592 594. doi: 10.1029/WR017i003p00592 Stearley, R.F., 1992, Historical ecology of Salmoninae, with special reference of Oncorhynchus, in Mayden, R.L., ed., Systematics, historical ecology, and North American freshwater fishes: Standford, California, Stanford University Press, p. 622 658. Sternecker, K., Cowley, D. E., & Geist, J. (2013). Factors influencing the success of salmonid egg development in river substratum. Ecology of Freshwater Fish, 22 (2), 322 333. doi: 10.1111/eff.12020 Sternecker, K., & Geist, J. (2010). The effects of stream substratum composition on the e mergence of salmonid fry. Ecology of Freshwater Fish, 19 (4), 537 544. doi: 10.1111/j.1600 0633.2010.00432.x Tonina, D., & Buffington, J. M. (2009). A three dimensional model for analyzing the effects of salmon redds on hyporheic exchange and egg pocket hab itat. Canadian Journal of Fisheries and Aquatic Sciences, 66 (12), 2157 2173. doi: 10.1139/f09 146 US Department of Interior (2013) US Geological Survey The National Map Viewer. US Forest Service (2012) US Department of Agriculture. Natural Resources M anagement: Payette National Forest: McCall, Idaho
70 Utz, R. M., Mesick, C. F., Cardinale, B. J., & Dunne, T. (2013). How does coarse gravel augmentation affect early stage Chinook salmon Oncorhynchus tshawytscha embryonic survivorship? Journal of Fish Biolo gy, 82 (5), 1484 1496. doi: 10.1111/jfb.12085 Utz, R. M., Zeug, S. C., & Cardinale, B. J. (2012). Juvenile Chinook salmon, Oncorhynchus tshawytscha, growth and diet in riverine habitat engineered to improve conditions for spawning. Fisheries Management and Ecology, 19 (5), 375 388. doi: 10.1111/j.1365 2400.2012.00849.x Wipfli, M. S. (1997). Terrestrial invertebrates as salmonid prey and nitrogen sources in streams: contrasting old growth and young growth riparian forests in southeastern Alaska, USA. Canadian Journal of Fisheries and Aquatic Sciences, 54 (6), 1259 1269. doi: 10.1139/cjfas 54 6 1259