An Abstract of the Thesis Of

Home , Hayrick Butte, Hogg Rock, Hoodoo Butte

AN ABSTRACT OF THE THESIS OF

Shyam Das-Toke for the degree of Master of Science in Geology presented on April 29, 2019

Title: Creating a Predictive Model of Cinder Cone Ages in Central Oregon

Abstract approved: ______Adam J.R. Kent

Cinder cones are useful geomorphic features for geological analysis because they generally have known initial states and follow a similar pattern of degradation as they are exposed to erosive processes. Characterizing large cinder cone fields in terms of age and composition requires significant resources, so this study aims to understand the parameters that influence cinder cone evolution in order to create a predictive age model of cones in Central Oregon. We utilize newly available high- resolution topography and new estimates of cinder cone age to evaluate a quantitative model of cone ages. The field site allows assessment of the effects of cone composition and precipitation in governing the morphology of Central Oregon cinder cones through time. We anticipate that these results will allow estimation of the age of individual cones which have not yet been dated directly. Statistical analyses show that younger cones have both significantly steeper slopes and more defined cone craters than older ones. Results also show that increased local precipitation does not appear to significantly influence cone morphology. However, composition as determined by silica content does influence initial morphology, as weight percentages vary by 15% and lead to differences of 8° in cone slope. After cone formation though, erosive processes such as mass wasting are more influential in influencing cones than chemical content. The statistical model of cone age that we create can be used to estimate cinder cone ages throughout the Central Oregon region and can be used in time-volume studies of monogenetic volcanism. The model and procedure should also be applicable in other portions of the Cascades and potentially to other volcanic fields.

by Shyam Das-Toke

A THESIS

submitted to

Oregon State University

in partial fulfillment of the requirements for the degree of

Master of Science

Presented April 29, 2019 Commencement June 2019 Master of Science thesis of Shyam Das-Toke presented on April 29, 2019

APPROVED:

Major Professor, representing Geology

Dean of the College of Earth, Ocean, and Atmospheric Sciences

Dean of the Graduate School

I understand that my thesis will become part of the permanent collection of Oregon State University libraries. My signature below authorizes release of my thesis to any reader upon request.

Shyam Das-Toke, Author

ACKNOWLEDGEMENTS

This work would not have been possible without the incredible support of my advisor, Dr. Adam Kent, and my committee members, Dr. Andrew Meigs, Dr. Eric Kirby, and Dr. David Cann, who provided input on my thesis and reviewed drafts several times. Thanks also to the United States Geological Survey and Julie Donnelly-Nolan, who provided age and silica content information for many of the cones in this study. Additional thanks to the Oregon Department of Geology and Mineral Industries, which provided the LiDAR imagery for the morphometric analyses. Without this data, we could not have obtained the detailed measurements of slope and crater transition that we use.

TABLE OF CONTENTS

Page

Background ...... 1 Motivation ...... 8

Methodology ...... 9 Evaluating Morphology ...... 9 Cone Parameter Data ...... 11 Statistical Treatments ...... 13

Results and Discussion ...... 15

Differences in Aspect ...... 16

Overview of Potential Regression Models ...... 22

Linear Models ...... 26 Effect of Precipitation, Outer Slope, and Silica Content on Age ...... 27 Crater Slope ...... 30 Combined Regression ...... 33

Uncertainties ...... 38

Conclusion ...... 43

Appendix A: Script in R ...... 45

Appendix B: Cutoffs for Confidence Levels ...... 46

Appendix C: USGS Silica Weight Percentage Database ...... 47

Appendix D: List of Cones ...... 54 Appendix E: Precipitation and Angle..………………………………………………55

References ...... 59

LIST OF FIGURES

Figure 1 Schematic of Cones...... 11

Figure 2 Measured Parameters...... 13

Figure 3 Cross Section of Cone:...... 13

Figure 4 Comparison of Slopes...... 18

Figure 5 Map of Cone Locations ...... 21

Figure 6 Plot of Assymetry in NVZ...... 24

Figure 7 Plot of Assymetry in all Cones ...... 25

Figure 8 Plot of Crater Assymetry ...... 26

Figure 9 QQ Plots of Regressions ...... 34

Figure 10 Cone Age vs. Crater Slope...... 41

Figure 11 Outer Slope vs Cone Age ...... 41

Figure 12 Predicted Cone Ages...... 44

Figure 13 Measured vs. Predicted Ages….………………………………...………..46

Figure 14 Crater Angle vs. Cone Age………………………………………..……....46

Figure 15 Comparison of Eruptive Style..……………………………………....…...49

LIST OF TABLES

Table Page

Table 1:...... 25

Table 2:...... 25

Table 3:…………………………………………………………………………...… 27

Table 4 …………………..……………………………………………………….….27

Table 5: ...... 43

Table 6: ...... 50

Background

Cinder cones are often understudied volcanic landforms, because larger stratovolcanoes often take up more attention and research. However, recent developments in technology, landscape imagery, and dating methodologies provide new techniques for detailed and accurate cone measurements and age determinations, which can save time and resources in characterizing these landforms. Cinder cones are formed when a volcanic vent erupts and volcanic debris such as tephra and scoria or spatter are deposited around it (Fig 1). The initial slopes of the cone are dictated by the critical angle of repose (Porter et al. 1972, Wood 1980). Because diffusion processes follow similar patterns in these areas, we can create predictive models of cone age based on measurable parameters such as morphology, silica, and precipitation exposure. Doing this allows us to estimate the age of cones without conducting time-consuming and resource intensive dating methodologies. While numerous potential factors can help determine how old a cone is, this study focuses on morphology, precipitation, composition, and initial eruption style. Previous studies show that age, precipitation, and chemical composition have statistically significant relationships with cinder cone morphology (Porter et al. 1972, Taylor et al. 2003, McGuire et al. 2014). However, no study has compiled a comprehensive relationship among all of these variables. Composition affects the initial viscosity of the cone, leading to different initial shapes, while precipitation provides a transport mechanism to move scoria and debris downslope, thus changing morphological characteristics. Age is influential because it allows erosive processes such as mass wasting and precipitation more time to act and degrade the cone. Chemical composition is potentially a key factor in altering cone morphology and hence modeled cone age, as it can influence the initial state of a cinder cone through viscosity and eruption dynamics (Wood 1980, Forniciai 2011). Lava viscosity is primarily controlled by silica content as this controls polymerization of silicon-oxygen tetrahedral within the melt with additional effects related to alkali and water contents that breakdown polymerized networks. Viscosity also plays an important role in eruption dynamics and thus influences the size and distribution of

tephra and scoria which make up the cones. Volatile abundances within the magma also play an important role in eruption dynamics and tephra dispersal, and there can be complex interplay between viscosity, eruption temperature, degassing and eruption style (Koleszar et al. 2011). Lavas that erupt with low silica content (<50% weight percentage) tend to be less viscous, potentially leading to lower cone slope values and gentler initial states (Bemis 2011, Forniciai 2011, McKay et al. 2009). However, cones are largely made up of tephra layers so the viscosity of flowing lava may have limited effects on morphology (Petrovic 2011). Precipitation is also a driver of cinder cone evolution as it is responsible for much for erosion of these cones. It governs direct runoff by overland flow which can transport cinders, and it influence the amount of vegetation of the cone (Taylor et al. 2003, Molotch et al. 2005, Kervyn et al. 2010). Overland flow is most common in areas of high precipitation (Fig. 12) but occurs regularly over the entire area during spring and fall, when precipitation increases across the region (Taylor et al. 2003). It is most relevant in high precipitation events, when soil is saturated and runoff flows along the surface of the cones. Extended amounts of precipitation can also create more conducive environments for vegetation growth. Increased vegetation hinders the erosion and transport of scoria and weathered debris (Cline and Pelletier 2007, Wood 1980). Greater precipitation exposure also helps chemically weather scoria deposits, and can assist mass wasting processes that change the steepness of the cone slopes. The resulting sediment may fill in the cones’ crater or move to the base of the cone, resulting in a flatter topographic profile (Porter et al. 1972, Pelletier and Cline 2007, Forniciai 2011). These geomorphic processes are more influential in shaping the cone over greater time periods. Since precipitation accounts for significant erosion, the effects of precipitation will be amplified when the cones are exposed to precipitation for longer periods of time and provide time for resulting debris to be transported to the base of the cone or the crater. Investigating the relationship between age, precipitation, silica content, and cinder cone morphology will allow us to build a predictive relationship that determines the eruptive ages of cones based on these known variables.

Figure 1: Schematic of cinder cone showing various layers and magma conduit. Cones are mostly composed of layers of tephra. Source: Petrovic 2011.

Previous Studies

Previous research has provided the basis for modeling general hillslope degradation, and several studies have used that to understand cinder cone degradation over time. The studies on cinder cones have also shown the value of measuring morphometric parameters through landscape imagery, but have been limited in their applications and incorporation of explanatory variables (Wood 1980, Bemis 2011, Forniciai 2011, McGuire et al. 2014). Studies focusing on cone morphology have obtained accurate morphology measurements and have previously studied how vegetation and precipitation influence cone morphology. However, their source imagery was of lower resolution than current LIDAR imagery.

The framework for how hillslopes change under slope-dependent transport has its roots in the degradation of scarps. This includes faults, wave-cut lacustrine shorelines, and terrace risers. Andrews and Bucknum (1987) demonstrated an early methodology to calibrate diffusion models for wave cut shorelines, using non-linear slope models to understand the rate of transport. Anderson et al. (1999) built upon this model for marine terraces, using one dimensional numerical models to analyze slope diffusion. Pierce and Coleman (1986) also sought to understand the evolution of terrace scarps in order to quantify the evolution of slopes. These studies helped refine the diffusive slope models that we apply to cinder cones. Dietrich et al. (2003) further quantified geomorphic transport laws in the Oregon Coast Range and found that runoff and overland flow can have limited effects on sediment transport in vegetated environments. The analysis informs our study of the effect of precipitation on cinder cones and modeled age. These studies on general landscape evolution have also informed subsequent research of cinder cones, and shed insight into how climate, time, and tectonic activity affect cone degradation. We use these studies as the basis for our analysis of cinder cone degradation which relies on accurate analysis of morphology. Extracting cone morphology parameters through data in ArcGIS is a well documented and commonly used method among landscape evolution studies. McKay et al. (2009), Forniciai (2011), and McGuire et al. (2014) have all used this method to measure outer cone slope, cone angle, and the crater transition angle in their study of cinder cones. To make the measurements, these studies used digital elevation model imagery (DEM) with resolution of 10 meters, as higher resolution data was not yet available. However, this resolution also limits the accuracy of the morphometric measurements and given that higher resolution data of sub-1 meter resolution is now available, there is new potential for refining and augmenting morphometric measurements.

Figure 2: Image of various parameters that this study uses to measure cinder cones morphology. Cone slope is measured from cone base to crater rim, while crater slope is measured from crater base to crater rim.

Figure 3: Cross section of measured values, showing where various parameters were measured along cone profiles.

Studies that have focused on Central Oregon cinder cones have also been limited in their area of study. For example, studies such as McKay et al. (2009) and

McGuire et al. (2014) limited sampling and morphometric analysis in Oregon to within the Newberry Volcanic Zone (NVZ) and the Newberry Rift Zone. Although these areas have a high density of cinder cones, many of Central Oregon’s cinder cones are located outside the area (Fig. 5) (Taylor et al. 2003). Therefore a more accurate model must account for the wider range of cones by analyzing cones from throughout the region. Compiling data from cones in all parts of Central Oregon allows for a more comprehensive understanding of the effect of precipitation regimes on morphology and takes into account greater silica content variation. In addition, cones within the NVZ may have been subject to certain eruptive styles that influenced their morphology, but this may not have affected the cones outside of the area. Precipitation is an influential factor on cone evolution and in predicting cone age, but only a select number of studies have analyzed its impact on cone degradation (Pelletier and Cline 2007, Forniciai 2011, de Michieli Vitturi and Arrowsmith 2013, McGuire et al. 2014). Because individual cinders weather to clay-like materials over time, precipitation can transport that material downslope via runoff (Wood 1980). High levels of precipitation often result in high levels of hillslope diffusivity because of the faster rate of transport (Whitlock 1992). McGuire et al. (2014) also analyzed cinder cones across the western United States to determine how their morphology is controlled by precipitation. The study shows that differences in microclimate and vegetation result in topographic differences between north and south facing cinder cone slopes. While the study provides explanations for the hillslope processes that govern these cones, it does not focus on the impact of silica content or eruptive style on morphology, thus neglecting crucial aspects of cinder cone evolution. This study also did not try to produce a predictive model of cone age based on morphology. Furthermore, McGuire et al. (2014), and Wood (1980) suggest that precipitation increases with elevation in this region. This may help counteract variables such as age because many of the younger cones tend to surround peaks of the high Cascades, such as the Three Sisters. The higher elevation results in more present-day precipitation than the older cones in drier regimes, potentially leading to greater rates of erosion. Thus, the effect of precipitation may be amplified in these areas.

Location Advantages

Our study area encompasses over 6475 km2 in Central Oregon, and includes cones as far north as the Mt. Jefferson wilderness, and as far south as the Newberry Volcanic Zone (Fig 5). This area provides several distinct advantages for cone analysis. There are hundreds of cones spread out among five precipitation regimes in the area, allowing a broad examination of the influence of different geomorphic environments on morphology (NACSE 2011). Additionally, there is readily available high resolution LiDAR data for this area, which was not available when prior cinder cone morphology studies were conducted. The eruptive ages and chemical composition of cones in the region are also well documented in the literature, allowing for a stronger focus on the relationships between the variables, rather than on collecting the data. Recent developments in isotope analysis and LiDAR have also provided more accurate methods to analyze cinder cones morphometry and rock samples. Previous studies that have analyzed the ages of Central Oregon cinder cones have often used K-Ar dating methods, which can produce large uncertainties in the resulting age values. A more accurate method is the 40Ar-39Ar dating method, which relies on irradiation of samples to allow measurement of the ratio of Ar-40 and Ar-39 in a given sample to determine the age (Bemis et al. 2011). This method tends to produce more reliable and accurate age values than K-Ar dating (Kuiper et al. 2008), and thus many of the age values for the study are collected in this way. Additionally, previous studies that have employed GIS to analyze and measure morphometric parameters of cinder cones used 10 meter DEM imagery (Ma 2005, Kneissl et al. 2011, Gilichinsky et al. 2010, Fornaciai 2011, McGuire et al. 2014), meaning the imagery has a resolution of approximately 10 meters on the ground. Recent technological developments in landscape imagery have allowed higher resolution imagery and more detailed analysis of morphology through LiDAR imagery, which can produce resolution of less than 1 meter (Nieto-Torres et al. 2019). While this study employs methods similar to those in previous research to extract morphological characteristics in ArcGIS, it uses LiDAR data to acquire higher resolution and allow more precise measurements of topographic features that are not possible with 10 meter DEMs.

Additionally, many previous morphometric studies do not emphasize the importance of crater rims in cinder cone evolution. This part of the cone is most exposed to weathering processes, since it is the most prominent feature. While it initially begins as a sharp transition into the crater, the rim erodes over time, leading to sediment infilling within the crater itself (Wood 1980, Gilichinsky et al. 2010). Geomorphic processes tend towards topographic balance, eroding prominent features and filling depressions such as the crater. This investigation considers cones with varying crater transition angles, and measures parameters such as crater slope in order to better incorporate this crucial aspect of cinder cones into a model of cone age (Fig 2).

Motivation Analyzing the patterns between cinder cone evolution and external variables allow for a predictive age model for cinder cone ages in the region. Given morphology, silica content, and precipitation values for a particular cone, the model can provide an estimate of the age of eruption of the cone. This is predicated on the duration of eruption being relatively short with respect to the age of the cone, which is reasonable given the short duration of most monogenetic cone- producing eruptions (Wood 1980, Forniciai 2011). In any respect the age estimate is strictly for the most recent event at that cone. We hypothesize that morphology will be most influential variable in estimating cone age, but we also believe that the effects of composition and precipitation may serve to refine the age estimates. Using a number of different morphometric measurements may provide a more accurate age estimate than the use of outer cone slope alone, which is a common approach in existing studies. There is also the possibility of bias or greater uncertainty associated with differences in the initial state of cones resulting from volcanological variations during eruption. To test this we make use of a natural experiment where cones similar in age from the NVZ can be used to determine potential uncertainties in predicted ages that result from morphology differences related to cones produced through short explosions, versus those that erupt in longer lived lava flow eruptions.

Additionally, there may be directional differences in the cones’ morphology and in the effects of prevailing winds. The north-facing sides receive less sunlight and therefore may host snowfields for longer periods of time than the warmer south- facing sides (Pierce and Coleman 1986, West et al. 2014, McGuire et al. 2014). This may result in less frequent snowmelt that would otherwise degrade the cones’ north slopes. Thus, the investigation will determine the extent of these morphological differences and whether they are confined to certain areas or age ranges.

Methodology

The variables that we compiled through this study are crater slope, outer cone slope, precipitation, cone age, and silica content. We collected the data from previous studies, databases, and aerial imagery (McGuire et al. 2014, McKay et al. 2009, Schilling 1975, Sherrod 1997), and analyzed which statistical model best fits the observed data. We compared the models’ precision, uncertainty, and accuracy to determine the best fit for our data. We then compiled several predictive equations that allow us to determine how the measured variables estimate age of eruption.

Evaluating Morphology We evaluated the morphology of the cinder cones through measurements of the cone’s height, mean outer slope, crater slope, and crater rim angle, which we obtained in ArcGIS through LiDAR imagery provided by the Oregon Department of Geology and Mineral Industries (DOGAMI). Within the ArcGIS environment, we calculated the average slope of each outer surface using the Slope function within the Spatial Analyst Toolbox. This provided a mean value of the entire slope from the cone’s basal height to the cone’s rim. Areas that have been quarried or otherwise altered through human activity were not included in the measurements. We conducted measurements of crater slope in a similar manner by setting the basal height as the base of the crater, and calculating the average slope to the crater rim. This method incorporates the entire circumference of the crater. We determined cone height by subtracting the basal elevation of the cone from the summit height of the rim. This

provided the value of topographic relief that the cone exhibits. Finally we use the Surface function in ArcGIS to calculate the average curvature of crater rims, among cones that had intact craters throughout their circumference. To ensure our slope method produced reliable results, we compared data values collected through ArcGIS in this study to slope values from McGuire et al. (2014), which analyzed cinder cones in the Newberry Volcanic Zone through 10 meter DEM layers in ArcGIS. Slope values from that study have approximately a 1 to 1 linear relationship with slope values obtained in this study (Fig. 4), indicating our slope measurement methodology is reliable and accurate.

McGuire et al. vs Das-Toke Slope values

50 ) ° 45 y = 0.99x + 3E-13

40 Toke 2018 ( 2018 Toke - 35

25 Slope from Das Slopefrom

20 20 25 30 35 40 45 50 Slopes from McGuire et al. 2014 (°)

Figure 4: Comparison of cone slopes acquired from McGuire (2014), to this study. With a slope of 0.99, the line of best fit indicates slopes from the same cinder cones are relatively consistent between the two studies. Slopes are provided in degrees, as this is what McGuire provided values as.

To conduct a comparison of north facing and south facing cone slopes, we divided each cone into two 180° regions. The dividing line between these areas runs directly east to west, and is parallel to the equator to ensure accurate comparisons between each cone. We calculated the slopes of each area using the same method as for the overall cone. Then we plotted the data and conduct statistical tests of

significance to determine whether there is a statistical difference between north and south slopes.

Cone Parameter Data

We obtain silica content values, precipitation levels, and cone ages from previously published sources and databases (Appendix C). For most of our data, values are corroborated in the literature to ensure that the data we use is as accurate and accepted as possible. Because much of the data we utilize is openly available, we are able to focus on the analysis and numerical models, rather than data collection. We have a full dataset for a total of 92 cinder cones (Appendix D), although there are many cones in Central Oregon for which we do not have data. Age values for the cones come from previous literature, including USGS publications and field guides (Appendix D). The ages are categorized based on the method in which they were obtained – Carbon-14 dating, Ar-Ar dating, or K-Ar dating. The values that we obtain though K-Ar dating are potentially unreliable due to the large uncertainty surrounding the ages and the relatively outdated nature of the method (Sherrod 1997, Kuiper et al. 2008). Therefore, we also utilize newly published USGS age values calculated through Ar-Ar dating to supplement the unreliable values. Overall our study includes cinder cones with ages that range between 1500 and 610,000 years old. We obtain silica content values through lava chemistry databases compiled by previous studies, as well as by the USGS (Appendix C). Previous studies have analyzed the chemistry of many of the cones that we use, and the database contains the chemical contents as well as the coordinates and relative locations for the samples listed (Schilling 1975, Sherrod 1997). The values represent silica weight percentages in analyzed samples. We are able to match our cones to the corresponding silica content for material in that cone based on latitude and longitude, as well as with qualitative information about cone names within each database. Because the silica content data for some samples vary based on location within the cone, we calculate average silica weight percentages for cones with multiple weight percentage values.

We also use current local precipitation data to analyze the effect of precipitation on cones. We obtained precipitation data from the PRISM Climate Group based at Oregon State University (NACSE 2011). The group provided a map layer in ArcGIS that categorizes regions in Central Oregon based on the amount of current precipitation experienced. There are five precipitation intervals that bound the regions in the study area (Fig. 12). To determine the amount of annual precipitation each cinder cone received, we overlay this layer on a location map of each of the cinder cones used in this study. The precipitation region that encompasses each cone determined which value we record for that cone.

Figure 5: Locations of cinder cones used in study, for which there was obtainable data on silica weight percentage, age, and precipitation.

Statistical Treatments To create a statistical model that accurately describes the relationship between the cone age, morphology, precipitation, and composition, we conduct several statistical analyses and evaluate their precision and accuracy. In the R statistical program, we fit a multivariable linear regression model, a 3rd order polynomial regression, and a logarithmic model to the data. Each model produced the coefficients for an equation that represent the relative influence of each variable in predicting cone age. Higher coefficients represent greater influence, while the sign of the

coefficient determines whether the variable has a direct or inverse relationship to cone age. To determine which of these models provides the best predictive relationship between age, silica, precipitation, and morphology, we evaluate the resulting residuals and uncertainty values. The model that provides the lowest uncertainties and residuals, and the best overall fit is the model we chose. To compare the values of residuals and uncertainties for each model, we calculate the average mean value for each model – multivariable linear regression, polynomial regression, and logarithmic regression. The mathematical mean value incorporates all uncertainty and residual values, and this provides a clear picture of how well each model fits the data. Uncertainties are derived from regression outputs and represent the standard error in our statistical analysis. This is the difference between observed values and estimated true values from the population of cones. Residuals represent the difference between the observed cone age and the predicted age from the statistical model. Thus, models that produce lower residual magnitudes more accurately predict the cone age. We also evaluate the models’ R2 value. This parameter indicates the proportion of variation within the age values that is predictable from the independent variables. The higher the proportion, the better the model is at predicting the age values. Thus, the value is a statistical measure of how close the data are to the fitted regression line. The equation with the highest R2 value most closely matches the data, therefore this parameter further determines the final model to use. Finally, as confirmation of our selection we use the Akaike information criterion (AIC). This statistical indicator provides a numerical value to provide a relative comparison of statistical models for a given data set. It determines how well the models fit the data relative to the other models. We compared the multivariable linear regression, the polynomial linear regression, and the logarithmic data using the AIC test in R (Table 1).

Evaluating Other Influences To determine whether initial asymmetries have a significant influence on Oregon cinder cones, we compare the youngest cone profiles from our study to

modern cinder cones that are less than 50 years old in Mexico and eastern Russia. We evaluate the north, south, east, and west slopes of each of these cones, and compare the resulting averages to the average for cones younger than 3000 years in Oregon. We use a t-test in the R statistical program to obtain p-values that indicate the probability of statistical difference between each respective side, which may indicate a systematic influence on cone morphology. We analyze young Oregon cones to avoid skewed results from extended cinder cone degradation, and we use modern cones elsewhere because they have experienced very little degradation, and thus represent the initial cone shape.

Results and Discussion

The goals of this study are to understand how cone morphology depends on time, how composition and precipitation influence this, and then create a predictive model based on that information. We first look at potential asymmetries between north facing and south cone slopes to determine how to incorporate slope into our model. We then determine which statistical model best fits our data and use that model to build a quantitative relationship between outer cone slope, silica content, precipitation, and the age of cone eruption. We also investigate whether the slope of a cone’s crater can replace outer cone slope in our statistical model, and integrate crater slope, outer cone slope, and the angle of the crater rim into a third regression model to predict age. Finally, we evaluate potential uncertainties in our model.

Differences in Aspect

Difference in North vs South Facing Slopes at Newberry

0.16 Volcanic Zone

0.14

0.12

0.1 y = 4E-05x - 0.2275

0.08 Difference 0.06 Trendline 0.04

Difference Difference inSlope (m/m) 0.02

0 5800 6000 6200 6400 6600 6800 7000 7200 Age (years)

North vs South Facing Outer Slopes at Newberry Volcanic 0.800 Zone 0.700 0.600 0.500 0.400 0.300 North

Slope(m/m) 0.200 South 0.100 0.000 5800 6000 6200 6400 6600 6800 7000 7200 b. Age (years)

Figure 6: a. Plot of the difference in north facing slopes vs. south facing outer slopes at young cones in the Newberry Volcanic Zone. In all cases, the slopes are steeper on the north side, and the divergence increases with time. b. Plot of individual north slopes and south slopes. Distance between vertically stacked points represents slope divergence for a given cone.

Differences in North-South Slopes in Central Oregon Cones

0.2

0.15

0.1

0.05 y = -9E-08x + 0.0408 0 1000 10000 100000 1000000

-0.05 Difference Difference inSlope (m/m) -0.1 Age (years)

Figure 7: a. Plot of difference in north-facing and south-facing outer slopes across all cones in the study, showing that the divergence between these sides decrease as age increases over broad timescales. The p-value for this data is 0.3, indicating little evidence of statistical difference. The trendline represents a linear fit, and shows curvature because of the logarithmic x-axis. b. Plot of individual north facing slopes and south facing slopes. Orange squares represent south slopes, while blue diamonds represent north slopes. Larger disparities between points of the same age represent larger differences in cone slopes.

Figure 8: a. Plot of difference in slopes between southern slopes and northern slopes in the crater. The p-value for the difference in these slopes is 0.19. b. Plot of individual north slopes and south slopes within craters. Results show that the south side of the crater is steeper than the north side, which faces the sun.

Evaluating topographic asymmetries helps determine if we must separate north and south facing slopes in our analysis to more accurately represent morphology in our statistical models. Previous literature suggests there may be a difference between these sides as a result of geomorphic and climactic feedbacks

(Porter 1972, Pierce and Coleman 1986, McGuire et al. 2014). We investigate if this applies to all cones in our study.

The data show that among young cones less than ~7000 years old, there is a divergence in outer cinder cone slope between north facing and south facing sides, but this difference does not appear to apply to longer time scales. Thus the divergence decreases when analyzing the 610,000 year history of cinder cones in Central Oregon.

Figure 7 shows that the average difference in north and south slopes steadily decreases from an average of 0.05 to approximately 0. However, the divergence is prevalent for cones in the Newberry Volcanic Zone, where cones are approximately

5000-7000 years old (McKay et al. 2009). The resulting asymmetry in cone slope suggest there are external geomorphic feedbacks that influence cone evolution and cause differences in morphology for young cones, but that these variables are not as significant for long-term cone evolution.

For the cones in the Newberry Volcanic Zone (NVZ), north-facing slopes are steeper than their south facing counterparts and the trend increases with age (Fig 6).

A t-test of statistical difference for north and south facing slopes in the Newberry

Volcanic Zone produces a p-value of 0.009 (Fig 6). This provides strong evidence that there is indeed a significant statistical difference between the north facing and south facing sides, and that the difference did not occur by chance. McGuire et al.

(2014) also found this relationship when studying cones within the Newberry

Volcanic Zone. While that study focused on more mature cones within the NVZ, it found that among cones ~7000 years old, north and south facing slopes diverge when the cone angle is greater than 20°. This supports our findings and suggests the difference applies when cones are still in early stages of evolution with steep slopes.

The divergence in cone aspect among young cones is likely a combination of geomorphic feedbacks and differences in microclimate. In the northern hemisphere, north-facing slopes are less exposed to sunlight, meaning that snowfields and frost that accumulate on the cones do not melt as frequently on the north side as they do on the more exposed south facing slopes (Pierce and Coleman 1986, Zappa et al. 2003,

West et al 2014). Frequent snowfield and frost melting produce greater runoff, which can accelerate geomorphic degradation. In young cones such as those in the NVZ, there is also generally low vegetation cover to stabilize the slopes, so the exposed south sides experience more cycles of degrading melt, leading to the gentler profiles

(Cline and Pelletier 2007). During times of glaciation, however, these cycles would not have induced degradation.

The significance in topographic asymmetry decreases as cones become older.

When analyzing all the cones in our study, the difference between north and south facing slopes decreases over time (Figure 7). A test of statistical significance for this larger sample of cones produces a p-value of 0.3, indicating little evidence of a statistical difference between north facing and south facing slopes over the entire range of cone ages. Because cones across Central Oregon are generally older than

cones in the NVZ, we can analyze broader patterns in topographic asymmetry by looking at the entire range of cinder cones.

Broader timescales allow for an additional feedback loop that results from steeper north facing slopes and helps balance out the initial topographic differences.

As north facing slopes maintain their steep profiles and south facing slopes rapidly degrade, the precipitation that falls may affect each side differently. Although north and south facing slopes experience the same amount of precipitation, the precipitation that falls on the steeper north-facing slope may move downhill quickly and is less likely to infiltrate the cone material (Howard 1997). On the other hand, precipitation is more likely to infiltrate the south-facing sides because these areas have more gentle profiles and runoff may move downslope more slowly and have more opportunity to percolate through soil or rocks (Stewart 2009). This difference in precipitation runoff may explain why the divergence in cone slope decreases on broad timescales.

Because runoff is less likely to infiltrate soil on north-slopes, plants that grow on these sides are drier than plants on the south-facing side, where water can infiltrate more easily (Taylor et al. 2003). The healthy, moist vegetation that grows on the south side is more effective in increasing cone cohesive strength, so when these plants have time to develop deeper roots, degradation slows in these areas. However, it takes time for cones to develop soil profiles. In younger cones there are not as many plants as on older cones because the soil has only recently developed (Bemis 2011). Without many plants to stabilize the slope, the degradation that occurs from increased snowmelt is amplified on these south sides. As cones become older, more healthy

plants develop and are able to slow cone degradation, thus allowing differences from slopes on north-facing sides to become less significant.

The results show a similar pattern of divergence in crater slope, albeit reversed. Slopes on the north side of craters face south and thus receive more sunlight than southern crater slopes, which face away from the sun. Thus, the same phenomenon occurs when frost and snow melts on the northern crater slope and degrade those areas more frequently. However, this divergence decreases with time

(Fig 8), just as it does with outer cone slope. This occurs because of increase in vegetation and reduction in erosion that also applies to outer cone slope (Pelletier and

Cline 2007). We observe a p-value of 0.19 for differences in crater slope, indicating little evidence of a statistical difference between north and south crater slopes across all time scales.

Because the topographic asymmetry between north and south slopes is statistically insignificant across the timescales we are considering, we do not distinguish between these two regions in the models of cinder cone ages. Instead, we use the average slopes for the entire cone when creating our quantitative models.

Overview of Potential Regression Models

We test three statistical models to determine which model fits out data and can best quantify the relationship between morphology, silica content, precipitation, and cone age. We evaluate the results of a multivariable linear regression, a polynomial regression, and a logarithmic regression. Of the three statistical regression models we use to predict cone ages, the multivariable linear regression model produced the most accurate values and had the lowest associated uncertainty (Table 1). Results from the

three models are shown in Table 1. Each model can be evaluated by residual magnitudes and uncertainties associated with predicting cone age. They also indicate the relative influence of silica content, morphology, and precipitation, on cone age.

To evaluate each model we analyze residuals, uncertainty, R2 values, and

Akaike information criterions. Residual values from each model indicate how accurate the analysis is and evaluate goodness of fit. Of the three models we use, the linear regression model produces the lowest magnitude of residuals, with an average of 9690 years across all cones (Table 1). This result means that there is an average

9690 year difference between the predicted age-value and the actual observed ages in the linear model. The residual averages from the polynomial and logarithmic models are significantly higher, at 33400 and 45640, respectively. Residual plots from the linear regression model closely fit a normally distributed dataset (Figure 9) whereas residuals from the polynomial and logarithmic models significantly deviate from a normally distributed dataset.

The average uncertainty associated with each model also helps determine which of our tested relationships best fit our data. The multivariable linear regression model again has the lowest uncertainty average with a mean of 5290, while the polynomial regression has an uncertainty of 6210 (Table 1). The logarithmic regression model produces an uncertainty of 15480 indicating the model has a broad range of potential error. These values represent the range within which the predicted eruptive age may fall when estimated by our model. Again, the linear regression model produces the least amount of potential error, resulting in more precise and accurate age values.

Each model also produces a unique R2 value, allowing us to evaluate the models’ ability to explain cone age variation. The multivariable linear model has the highest R2 value at 0.38, meaning it explains 38% of the observed variability among cone ages (Table 1). The polynomial model has a slightly lower R2 value, at 0.36, and the logarithmic model only has a R2 of 0.34. Thus, the linear regression model is slightly better at explaining the variation among observed age values than the other models. However, the R2 values show that much of the variation is not explained by any of the models.

Finally, the Akaike information criterion (AIC) can also be used to assess relative goodness of fit. This statistical indicator provides a relative comparison of statistical models for a given data set and determines how well the models fit the data compared to the other models. Table 1 shows that the AIC value for the linear model is the lowest of the three (539). The polynomial model is of lower quality with an

AIC of 10390, while the logarithmic model poorly fits our data, with an extremely large AIC of 100340. Lower AIC values indicate better relative quality of the model, and the results again show that the linear model is the highest quality model among our choices.

Because the linear regression model has the lowest residual values, the lowest uncertainty, the highest R2 value, and the lowest AIC value, we conclude that this type of model best illustrates the relationship between the cone parameters of interest.

The age values it produces most closely match the actual observed age values, and the model provides the smallest standard error associated with its results. However, we note that even our best fit explains only 38% of the observed variations.

This model implies that slope, precipitation, and silica content have

proportional linear relationships to cone age, as opposed to polynomial or logarithmic

relationships. As these parameters change, age increases or decreases by a set amount

as determined by the model’s coefficients. The nature of erosive processes helps

explain this. Over time, erosive processes work to transport material from the top of

the cone to lower elevations at a consistent rate. The linear relationship indicates that

variables do not exhibit a cumulative effect on cones age that might be apparent

through a non-linear model.

Table 1: Comparison of statistical models and resulting predictive equations. X is outer cone slope (m/m), P is precipitation (cm), Y is silica content (weight percentage), and A is cone age (years).

Type Equation R2 Average Average AIC Uncertainty Residual Value Magnitude Multivariable A= -453392.2X-368.4P- 0.38 5290 9680 539 Linear 5670y+643524.0 3rd Order A=-193X3+131P2+3060y-3944 0.36 6210 33400 10390 Polynomial Logarithmic A =-1001ln(X)+1456ln(P)+256ln(y)- 0.34 15480 45640 100340 342

Table 2: a. T-values from the multivariable linear regression model, indicating the relative influence of each of the three independent variables. Greater magnitudes indicate greater influence on the predicted cone age. Negative values indicate an inverse relationship with predicted cone age. Cutoffs for confidence levels are found in Appendix A. b. T-values from regression where slope is the dependent variable, and age, precipitation, and silica content are independent variables. a.

Variable T-value

Outer Slope -4.398

Precipitation -1.204

Silica -1.295

b. Variable T-value

Age -4.404

Precipitation -1.107

Silica -1.154

Figure 9: Normal QQ plot of residuals from each statistical model. Solid lines represent an idealized, normally distributed dataset, while each point represents a residual, showing the disparity between the predicted value from the model and the actual observed value. The results from the multivariable linear regression model most closely fit a normal distribution.

Linear Models Using the multivariable linear model, we compiled three models with various permutations of silica content, morphology, and precipitation, in order to estimate cone age. Our first model relates outer cone slope, silica weight percentage, and

precipitation levels to the age of the cone. We then create a second model that substitutes crater slope for outer cone slope. Finally, our third model includes crater slope, outer cone slope, and the angle of the crater rim, in estimating cone age.

Effect of Precipitation, Outer Slope, and Silica Content on Age Using the information from the multivariable linear regression model, we can determine how changes in each of variables affect predicted cone age. Outer slope, precipitation, and chemistry influence the predicted cone age by varying degrees as represented by the magnitude and sign of the coefficients in the model. The t-value results from the linear regression provide further evidence of each variables’ relative influence.

Based on the results of the regression test, outer cone slope has a stronger influencing when estimating cone age than silica content or current precipitation. The regression coefficient for slope is -453392, indicating a strong inverse relationship with age (Table 1). If cone slope increases by a factor of 1 m/m, the predicted age for the corresponding cone would decrease by 453,392 years according to our model.

Estimated cone age is thus extremely sensitive to changes in slope. This regression test shows that slope produces a t-value of -4.398 (Table 2) meaning we can project with 99% confidence that this variable is a statistically significant driver of predicted cone age. This large influence on the predicted age is due to the fact that degraded morphology is the most obvious consequence of cinder cone evolution. Over time, erosive processes such as mass wasting and runoff from precipitation are better able to flatten cone profiles and thus decrease cone slope (Wood 1980, Cline and Pelletier

2007, McGuire et al. 2014).

This multivariable linear regression model indicates current precipitation levels have a smaller influence in estimating cone age, although the two variables still have an inverse relationship with one another. The coefficient of -368.4 indicates that if precipitation increases by 1 cm, the predicted cone age decreases by only 368 years, with all other variables held constant. This is only a slight inverse correlation compared to the influence that changes in slope have on age. The t-value of -1.204 for precipitation provides further evidence of its limited influence (Table 2), as it shows that we can only project 70% confidence that precipitation has an influence on predicted cone age.

Explanations for this slight inverse relationship vary, but center on the spatial distribution of precipitation in this region. The cones experiencing the highest current annual precipitation levels (254 cm+) are located around the high Cascades (Fig. 12).

The Cascade Mountains have greater exposure to precipitation and host some of the youngest cinder cones since the area is currently more volcanically active than the rest of our study area (Taylor et al. 2003). The mountains also act as a rainshadow, reducing the amount of precipitation that falls on cinder cones to the east. As a result, the older cinder cones on the eastern reaches of our study area experience less precipitation than those at higher elevations. Thus, spatial distribution in this area influences the effect of precipitation on morphology.

Precipitation exerts slight influence on morphology and hence the model- predicted age because these are current precipitation levels, and precipitation likely also varied over the history of the cones. The oldest cone we analyze is 610,000 years old and climactic trends may have resulted in different levels of snow and rainfall

across the region since then. Whitlock et al. (1992) shows that precipitation decreased from 10,000 to 4000 years ago, particularly along the western area of our study site.

Thus, areas that experience high current precipitation levels may have experienced much lower precipitation levels previously. Prior to 10,000 years ago, precipitation increased consistently along the entire study area (Zappa et al. 2003). This variation would skew our results because the current precipitation levels may not have resulted in significant degradation, while historical precipitation levels may have had more of an effect on erosion and weathering. However, we do not have access to detailed quantitative Quaternary precipitation levels for each site in our study and thus cannot quantify how historic precipitation influenced individual sites in our study.

The third variable in our initial model, silica content, has a similarly small influence on our model. Although the composition of erupted magmas affect viscosity and eruption style, this influence does not appear significant in determining the cone’s age. With a coefficient of -5670, silica weight percentage has a slight inverse relationship with the age of eruption. A one percent increase in silica content indicates a reduction in modeled cone age by 5670 years, holding slope and precipitation constant. As Table 2 shows, the regression t-value of -1.295 with an alpha level of 0.3 allows 70% confidence that silica content statistically influences predicted cone age. This influence is slightly larger than current precipitation, but significantly lower than outer cone slope.

We attribute this slight inverse relationship between silica content and cone age to the initial lava viscosity during cone formation and differences in subsequent weathering. Higher silica content in cones correlates with more viscous eruptions.

Although cinder cones are mostly composed of tephra (Wood 1980, Fornaciai 2011), erupted scoria is more viscous with higher silica weight percentages. Thus the erupted lavas flow much slower than non-viscous lavas, meaning that they cool closer to the original eruption source (Porter 1972, Kervyn et al. 2010). This generally results in steeper cone slopes as cones build themselves upward. Less viscous eruptions are associated with more gentle slopes as any erupted lava can spread out more freely

(Wood 1980). Thus, silica content may be used with other parameters such as morphology to estimate the cones’ age, but it is likely not a good stand-alone indicator. Furthermore, layers of erupted tephra are not significantly affected by the viscosity changes, so silica does not contribute to the variations in tephra angle. The silica weight percentage ranges by ~15% among our cones, and this variation leads to differences of approximately 10° in cone slope. Holding slope and precipitation constant, this range in chemical composition leads to differences of just 1700 years in predicted cone age, further illustrating the negligible effect of silica content on predicted age. It influences the initial shape of the cone more than it affects the estimated age of eruption.

We also conduct a regression test where morphology is the dependent variable, and age, silica content, and precipitation are the independent variables. We do this because in cinder cones, slope depends on age, silica content, and precipitation, not the other way around. Our statistical output (Table 2b) shows the relationships among the respective variables is consistent between each version of the regression. Age has a t-value magnitude above 4, showing it has an influential inverse relationship with slope. Precipitation and silica content have small inverse

relationships as indicated by the previous regression permutation. They again have t- value magnitudes just above 1. Not only do they have small influences on age, the results show they have small influences on slope and may not be the most accurate variables for our age model.

Crater Slope To determine if crater slope exerts a similar influence on predicted cone age as outer cone slope does, we replace outer slope with the slope of the crater in a second model. This is possible because the high-resolution LiDAR data we accessed allows detailed measurements of craters and crater slopes. Previous studies of Central

Oregon cinder cones and cones elsewhere around the world have utilized 10m DEMs, but LIDAR offers new options. When replacing the average outer cone slope with the average crater slope in our linear regression model, the model produces similar results in estimating cone age and in the relative influence of morphology in a multivariable linear regression. Steeper, more defined craters are indicative of young cones, while shallow, gradual crater slopes are indicative of mature, older cinder cones (Fig. 10).

The results of this second model show that crater slope is more influential in making estimates of cone age than current precipitation and silica content are. As

Table 3 shows, the coefficient of -413427 from a linear regression model again indicates an inverse relationship between crater slope and predicted cone age, and a much larger influence than silica and precipitation, as represented by their coefficients. The coefficient and t-value of -3.928 also indicate a very high degree of certainty that crater slope is a statistically significant influence in predicting cone age, just as outer cone slope is (Table 4). These results mirror the results of our first

equation, which shows outer slope is more significant in age estimates than silica or precipitation values.

Furthermore, replacement of outer slope with crater slope results in similar levels of accuracy in our model. The regression model with crater slope produces an average residual value of 9932, meaning the average difference between predicted cone age and actual cone age is approximately 9932 years (Table 3). Considering that our model successfully predicts cone ages up to 610,000 years, this residual value is fairly low. It is also comparable to the average residual value from the linear model that uses outer slope (9680). Both models are relatively accurate.

Sediment infilling in the crater over time explains the reduction in crater slope and the inverse relationship between crater slope and cone age. Figure 10 shows that crater slope generally decreases as age increases and after 100,000 years, crater slope rapidly degrades. However, there are no instances in our study where the crater completely disappeared because of degradation. Degradation occurs because gravity and runoff from precipitation help move material from the top of the crater rim down into the crater itself just as is the case for cones’ outer slopes (de Michieli Vitturi and

Arrowsmith 2007, Wood 1980). As the sediment fills the crater and the cone rim erodes, the crater slope gradually decreases. Because the results mirror the influence of outer cone slope, we can infer that cone degradation affects the crater morphology to the same degree as cone’s outer morphology.

Table 3: Parameters of accuracy and uncertainty for multivariable linear regression model that includes crater slope. X is crater slope (m/m), P is precipitation (cm), Y is silica content (weight percentage), and A is cone age (years).

Type Equation R2 Uncertainty Residuals Multivariable A= -413427X-358.4P-5970Y+643524 0.37 5290 9930 Linear With Crater Slope

Table 4: T-values from the regression model, substituting crater slope for outer cone slope. Variable T-value Crater Slope -3.928 Precipitation -1.104 Silica -1.191

Age vs. Crater Slope

0.8 0.7 0.6 y = -5E-07x + 0.4692 0.5 0.4 0.3 0.2

Crater Crater Slope(m/m) R² = 0.3048 0.1 0 1000 100000 Cone Age (years)

Figure 10:Plot of observed cone age to measured crater slope. Results show that crater slope decreases with time, especially for cones older than 100,000 years. The ages are shown on a logarithmic scale.

Age vs. Outer Slope

0.700 0.600 0.500 R² = 0.3977 0.400 0.300

0.200 OuterSlope (m/m) 0.100 y = -7E-07x + 0.5238 0.000 1000 100000 Cone Age (years)

Figure 11: Plot of outer slope vs cone age. Slopes rapidly decreases after about 50,000 years, but consistently decrease throughout all timescales Combined Regression Because crater slope and outer cone slope have shown to be the most influential variables and because we show precipitation and silica content to be negligible factors in predicting cone age, we compile a linear regression model using crater slope and outer slope, and add in crater transition angle. The resulting model explains more of the variance in ages than previous regression models and has similar uncertainty and accuracy estimates. The equation explains 65% of the variance in age compared to 38% in the previous linear regression models (Table 5). The uncertainty values of 5030 and residual value of 9310 are similar to the respective values from the previous models as well. Because this model incorporates more significant variables in crater slope and outer slope, we believe it to be a more accurate model than models that include precipitation and silica content. The coefficients again represent an inverse relationship between crater slope average outer slope, and cone age. Both crater slope and outer cone slope have significant influences on cone age as indicated by their weighted coefficients and large t-values (Table 5). Because the magnitudes of both t-values are well above 4.0, we can project greater than 99% confidence that each variable is a significant influence on predicted cone age (Appendix B). Furthermore, the large coefficients for

both crater slope and outer slope indicate that slight changes in these variables produce large differences in cone age, again illustrating the sensitivity of cone age to the variables. Crater rim angle shows to another influential variable in estimating cone age. Rim angles vary from 93° to 136° (Appendix E), and have a positive correlation with cone age. As cones get older, the rim transition angle becomes larger. This is again a result of the degradation in cones over time. The large t-value of 4.65 for this variable further illustrates its weighted influence, and is comparable to those of crater slope and outer cone slope. (Table 5). This part of cinder cone is the most susceptible to erosion (Kneissl et al. 2011) which explains the large variation in angles over time. The results from the combined equation illustrate linear hillslope diffusivity among our slopes. Hillslope diffusivity refers to the movement and spreading of materials downslope (Cline and Pelletier 2007). Slopes with linear diffusion have proportional relationships to the rate of transport, which is typical in areas of consistent curvature without anomalies such as gullies or rills (Anderson et al. 1999). The slopes in our study mainly exhibit linear diffusivity that can attributed to the abundance of young cones and lack of gullies and drainage patterns. We also assume equal diffusivity among our cones, which is a sound assumption in young Central Oregon landforms (Taylor et al. 2003). Most of our cones have similar rates of curvature based on age, thus leading to equal diffusivity. Some of the more mature cones do not exhibit linear diffusion because of irregularities in slopes, but these cones are outliers and do not affect the overall trend of proportional rates of transport.

Table 5: Results of combined regression, with crater slope. outer cone slope, and crater rom angle as the input variables. “A” again represents cone age, while crater slope (m/m) is symbolized by “X”, and outer slope (m/m) is symbolized by “Y”. Crater rim angle is symbolized by “C”. T-values for these parameters are provided.

Type Equation R2 Avg. Residuals Uncertainty Crater Slope A=-201971Y+8367-248326X-657599 0.65 5030 9310 +Outer Slope+Crater Rim Angle

Crater Slope Crater Rim Angle Outer Slope T-value -6.48 4.65 -5.52

Figure 12: Predicted cone ages from the combined linear regression model. This includes cones with irregular morphology that we did not use in calibrating our model.

Estimated Age vs Predicted Age 600000 R² = 0.9655 500000 y = 0.9533x + 6378.9 400000

300000

200000

Estimated Age (Years) 100000

0 0 100000 200000 300000 400000 500000 600000 700000 Observed Age (years)

Figure 13: Plot of actual observed cone ages vs. model-estimated cone ages. Estimates are shown to be consistent with age values across all time frames. Clustering of ages younger than 100,000 years is due to abundance of cones in that time range.

Crater Angle vs Cone Age

700000

600000

500000

400000

(Years) 300000 Cone Age Age Cone 200000

100000

0 90 100 110 120 130 140 Crater Rim Angle (°)

Figure 14: Plot of crater angle vs cone age for calibration cones, with ‘good’ morphology that have intact crater rims.

We recommend the combined slope-rim angle model as the most accurate way to estimate cone age, based on the results of our statistical analyses. We show that this equation accounts for significantly more of the variation in cone ages based

on its R2 value, and has lower uncertainties and residual values than the equations incorporating silica content and precipitation. Morphology appears to directly relate to the age of the cone and slope values change as a function of cone age. Thus, these parameters are highly sensitive to the stage of cone evolution and are indicative of cone age. Figure 13 shows that we can use this model to estimate cone ages with consistent accuracy compared to observed ages.

Uncertainties:

Our analysis of the morphology-age relationship also shows the effect of various sources of uncertainty which we evaluate in detail below. We look at the effects of eruptive style and prevailing wind direction and their respective impacts on model-predicted cone age.

Eruptive Style

Because even our most accurate regression model explains only 52% of the variation in cone ages, there may be other variables we did not include that may introduce uncertainties into our equation relating slope, precipitation, and chemistry to predicted cone age. Eruptive style and prevailing wind direction during cone eruptions may account for some of these uncertainties. Both variables influence the initial shape and morphology of the cinder cone, and thus may influence the predicted cone age.

To test the influence on initial cone morphology further, we analyze the influence of two main eruptive styles in the Central Oregon cinder cone field. Cones in this area can erupt in explosive strombolian type eruptions or more viscous

Hawaiian style eruptions (McKay et al. 2009; Kervyn et al. 2010). The former often results in repeated effusive blasts that produce voluminous pyroclastic material, while the latter style forms low, steep sided cones known as spatter cones through eruptions of lava (Porter 1972). Though eruptive style is related to silica content, we examine differences in these types of cones because they may have unforeseen effects on our model. Since cones in the NVZ have very similar ages, we can use age as a control to determine how eruptive style influences cone morphology (McKay et al. 2009).

The data show that in the NVZ, there is a slight difference in the profiles between cones that experienced strombolian eruptions and those that experienced less viscous Hawaiian eruptions. Cones with explosive strombolian eruptions tend to have steeper slopes in this area, while less viscous Hawaiian eruptions produce cones with slightly gentler slopes (Fig 14). The difference between these two kinds of eruptions produces a p-value of 0.07, indicating moderate evidence against the theory that the difference in slopes occurred by chance.

This disparity in slope between strombolian and Hawaiian eruptions is a result of viscosity, silica composition, and particle size during the initial eruption. Spatter cones eject small magma bodies that often fall near their source, while less viscous

Hawaiian eruptions allow the erupted lava to spread out from its source, producing muted profiles (Porter 1972). Higher silica content in strombolian eruptions also increases the viscosity of erupted material, thus decreasing the distance that the material flows (Taylor et al. 2003). This allows cones to build upwards and results in steeper profiles. Because there is moderate evidence against the significance of eruptive style, we acknowledge that it may introduce errors into our model. It may

help account for the average uncertainty of 5840 years in the combined crater slope- outer slope regression model and explain why our most accurate model only explains half the observed age value variation. According to our primary regression model that incorporates silica content, the average difference between strombolian and Hawaiian eruptions produce a disparity of 3300 years, holding all other variables constant.

Strombolian vs. Hawaiian Style Eruption 45

30 Strombolian Hawaiian 25

Vertical Relief Vertical Relief (meters) 10

0 0 200 400 600 800 1000 1200 Horizontal Distance (meters)

Figure 15: Comparison of profiles between strombolian eruption (Graham Butte) and Hawaiian style eruptions (Bessie Butte). Strombolian eruptions result in steeper slopes, while lava from Hawaiian eruptions is more spread out, producing more gentle slopes. Wind Direction

Another potential uncertainty is the effect of prevailing wind direction during the eruptions of Oregon cinder cones. During the cone eruption and debris deposition, strong winds could potentially transport tephra or other erupted material and deposit them on a particular side of the cone (Wood 1980, Gilichinsky et al. 2010, Forniciai

2011). This leads to asymmetries in slope, as gradual deposition on one side may

result in gentle topography at those locations but steeper slopes on sides without wind deposition. Our comparison of Oregon cinder cones to cones elsewhere in the world can provide insight into the potential effect of wind.

When comparing Oregon cinder cones younger than 3000 years old to modern day cinder cones in Kamchatka, Russia and Central Mexico, we found no significant statistical difference between the north, south, east, or west slopes of these cones

(Table 6). A significant difference between modern cones younger than 50 years old and those in our study could imply outside interference on morphology as a result of variables such as wind. Because there is no statistical difference between any one side in our comparison, it is less likely that wind had significant effects on Oregon cones.

However, the results do not preclude the influence of wind direction on Oregon cones. Prevailing winds may have simply added more material onto a given side, resulting in thicker depositions but not necessarily changing slope values.

Furthermore, there is a very small window of time for wind to influence the initial shape of the cinder cones. It is most likely to have an influence during the time of eruption, where ash and erupted debris are airborne. Since cone eruptions are generally short-lived (Wood 1980, Howard 1997, Forniciai 2011), a wind gust of only several minutes or less may have significant influences on initial morphology by creating larger asymmetries in cone slope. These wind gusts may have also originated from a different direction than the overall prevailing winds, making analysis more complex.

We conclude that external variables such as wind direction and eruptive style introduce only small uncertainties into our model. The data show that the difference

between strombolian and Hawaiian eruptions is slight, and that there is no statistical difference between young Oregon cones and modern cones from other regions that may indicate the influence of wind. Thus, we conclude that these variables do not significantly impact our model of cinder cone evolution nor do they significantly influence predicted cone age. Broader analysis and comparison to more modern cones outside Oregon may provide more insight into this.

Table 6: Slopes of modern cinder cones (<50 years old), and slopes of young Oregon cones (<3000 years)

Cone Average North South East West Slope (Percentage Ratio) External Paracutin 0.392 0.394 0.395 0.384 0.373 Cones Bezymianny 0.431 0.453 0.410 0.443 0.430 Kame 0.436 0.442 0.431 0.429 0.447 Cono del 0.475 0.476 0.471 0.469 0.463 Laghetto Young Nash 0.394 0.413 0.382 0.399 0.382 Oregon Le Conte 0.419 0.429 0.420 0.403 0.402 Cones Collier 0.462 0.469 0.443 0.459 0.449 Four in One 0.442 0.463 0.445 0.437 0.453 Yapoah 0.453 0.471 0.442 0.459 0.442 Crater Kelsey 0.382 0.415 0.379 0.399 0.390

Table 7: P-values representing the statistical differences between average slopes of modern cones less than 50 years old, and young Oregon cones less than 3000 years old. The data show little evidence of a statistical difference between the average of these cones since all p-values are above 0.1 and are considered high.

North South East West

P-value 0.20 0.19 0.184 0.16

Errors

All measures were taken to ensure morphology data is as accurate as possible, but there is still the possibility of unforeseen errors in the measurements. When measuring crater slope and outer slopes ten times each for a sampling of 20 cones, we achieved a slope result within 0.03 m/m each iteration. Any differences between iterations are a result of selecting slightly offset boundary areas to limit the slope calculations in ArcGIS. Because our results match the slope values from McGuire et al. (2014), we consider these values to be accurate (Fig. 4).

Conclusion We assemble several quantitative models to illustrate the variables that impact predicted cone age. Of the variables we consider, morphological parameters are the best predictors of cone age and accurately explain the variation in observed cone ages. We find that silica content and current precipitation levels do not provide significant insight into eruptive age, although they do influence cone morphology.

We also find that there is asymmetry in cone slopes based on geomorphic and climatic feedback effects, but because the divergence decreases over time it does not significantly affect our analysis. In younger cones, north facing slopes tend to be steeper than south facing slopes because they experience fewer cycles of erosion and snowmelt. However, once these cones are mature enough to develop soil profiles, degradation slows as vegetation is better able to stabilize the cone and prevent erosion.

Among our multivariable models, cone slope and crater slope have much greater impacts in modeling the eruptive age of the cone than precipitation or silica

weight percentage do. As a result, morphology has a strong inversely proportional relationship to cone age, while precipitation and silica content have only weakly inverse relationships with cone age. While precipitation is still an influential factor in cone evolution, current precipitation levels may not have significant impacts on cone degradation. Historical precipitation may have had greater influence due to the much longer time frames. Furthermore, silica composition influences the initial shape of cinder cones, but erosive processes are more influential on cone degradation over a cone’s lifespan.

We find that the best way to relate variables to predicted cone age is through a multivariable linear regression. The linear relationship illustrates that influential variables such as cone and crater morphology degrade steadily over time as precipitation and gravity-driven processes drive materials to the base of the cone and crater. There is no cumulative effect from these processes, rather erosion moves material downslope relatively consistently. Because this is somewhat unexpected, we reason that our data is heavy on young cones less than 10,000 years old. Thus our data is young compared to the efficiency of hillslope transport processes.

However, our models do not explain all the relationships among the data. Our most accurate model, incorporating crater slope, crater rim angle, and outer cone slope, explains 65% of the variation in cone age values. Thus, we examine other variables that could introduce uncertainties into our model. Prevailing wind direction and eruptive style have definite impacts on initial morphology, although our analysis shows wind direction has no statistical impact on our age model. However, there are moderate differences between morphology due to eruptive style, thus influencing our

numerical models of cone age. Future work may entail expanding the analysis of these uncertainties to cinder cones on a more global scale.

Appendix A Script for Statistical Analysis attach(Conedata) list(Conedata) fit<-lm(Age~Slope+Precipitation+Silica) # linear model summary(fit) qqnorm(resid(fit)) qqline(resid(fit))

attach(test) pfit<-lm(Age ~ polym(Slope, Precipitation, Silica, degree=3, raw=TRUE)) #polynomial model y ~ x1 + x2 + I(x1^2) + I(x2^2) + x1:x2 summary(pfit) qqnorm(resid(pfit)) qqline(resid(pfit)) p+labs(title="Linear Regression”)

x <- 1:n > set.parameters(10) > x <- 1*log(Silica+ Precipitation +Slope)-6+rnorm(n) > > #plot the data > plot(Age~x) > > #fit log model > fitlog <- lm(Age~log(x)) > #Results of the model > summary(fit)

AIC(fit) #Aic linear stopifnot(all.equal(AIC(fit), AIC(logLik(fit))))

var<-lm(Age~Slope) summary(var) qqnorm(resid(var)) qqline(resid(var))

AIC(pfit) #AIC for polynomial stopifnot(all.equal(AIC(pfit), AIC(logLik(pfit))))

summary(pfit) qqnorm(resid(pfit)) qqline(resid(pfit))

AIC(pfit) #AIC for logarithmic stopifnot(all.equal(AIC(fitlog AIC(logLik(pfit))))

summary(fitlog) qqnorm(resid(fitlog)) qqline(resid(fitlog))

attach(Cone) pcrater<-lm(Age ~ Crater.Slope+Precipitation+Silica)) #crater replacement summary(pcrater) qqnorm(resid(pfit)) qqline(resid(pfit))

pcombined<-lm(Age~Crater.Slope+Cone.Slope+Crater.rim))#combined equation summary(pcombined)

Appendix B: Cutoffs for Confidence Levels Table A: Cutoff levels for t-value significance levels for samples of greater than 30 observations. Ross and Chanceman 2015).

Magnitude of T-value Confidence Level 4.0 99% 2.0 95% 1.68 90% 1.29 80%

Appendix C: USGS Silica Weight Percentage Database

Lookout Point, High Cascades 48.40 Garrison Butte chain of cinder cones 48.30 Belknap Crater, east flow 50.70 SW of Black Butte, highway roadcut 49.70 SW of Trout Creek Butte 50.00 from jct, Hwy 20 and road to Hoodoo ski area 51.90 Olivine andesite (welded tuff?), highway cut south of Tamolitch falls 52.70 collected near Craig Monumnet, McKenzie Pass 52.00 predates North Sister; c;ollected from roadcut between Snow and Squaw Creeks 52.00 west base, middle sister 52.50 Two Butte, near summit 51.00 Near Beaver Marsh, basal High Cascades 51.50 Scott Mountain, Olallie Road turnoff near base of cliffs 51.10 From Sawyers Cave (early Nash crater flow) 51.70 north shoulder, little Brother 52.50 from Cayuse Crater, collected along Century Drive 51.20 from high cliffs near Payne Creek 51.70 Aubrey Butte, near radio tower 53.00 From Broken Top, knob at south end of Green Lakes 53.80 Sand Mountain cones, south group 52.10 Little Brother 53.20 southeast ridge of Husband 54.20 Sims Butte flow 52.60 West base, middle sister, collected along skyline trail 52.40 quarry near Scott Lake 52.50 Bunchgrass Ridge 54.20 from Bluegrass Butte 52.80 Tam McArthur Ridge, near eastern end 54.40 Little Nash Crater flow 52.60 outcrop north of Little Brother 52.50 north base, bachelor butte 52.50 Lava, highway cut near Tamolitch Falls overlook 52.70 West lobe of a Belknap crater flow 52.80 Graham butte in roadcut 53.20 Skylight Cave, Sixmile Butte flow 53.50 halfway down northern slope, Tam McArthur Ridge 54.20 Belknap Crater lava on McKenzie Pass 53.40 North side of Black Butte 53.20

Tumalo Mtn, just east of Bachelor Butte turnoff 54.70 Madness Ridge, lower western flank Little Brother, near Skyline Trail 53.30 Outcrop along Squaw Creek, rev. polarity 53.50 Clear Lake flow from Sand Mountain 54.40 underlies flow from Four-in-One, collected near terminus 53.80 Windy Point, McKenzie Pass Highway 55.70 from Deer Butte 54.00 West Lava Camp flow from Belknap Crater 54.10 Nash Crater, Fish Lake flow 54.50 Sims butte blow, highway roadcut 54.40 south plug of Husband 56.00 Le Conte Crater flow, along Century Drive 54.70 Hoodoo Bowl lava, se corner, mesa above ski area 54.40 Broken Top, at Fall Creke falls 56.70 from Trout Creek Butte lookout 54.80 Jack Pine road, potato Hill 54.90 Hwy 20 cut at west end of Suttle Lake, rev. polarity 54.80 above Squaw Creek; rev. polarity 55.00 Bachelor Butte, road to ski area 55.00 Dike, south end of Eileen Lake 55.50 Husband, margin of north plug 54.90 Collier Cone lava, near center of flow where Frog Camp trail crosses lava field 57.70 west Base, middle Sister, along Skyline trail 55.90 South sister, summit cindery pyroxene andesite 58.40 Yapoah flow, just east of Dee Wright observatory 56.30 Todd Lake volcano 56.10 Montague Memorial, skyline trail 56.40 Maxwell butte lava, west of Little Nash Crater 57.10 Broken Top lava, south end of Green Lakes 59.00 Collier Cone west flow, terminus 60.90 Collier Cone, collected just west of cone 61.10 South Sister, along trail to summit from Green lakes, at about 8,000 ft elev. 61.40 Four in One flow, terminus 59.30 Hogg Rock 60.40 Middle Sister, east of Skyline Trail above sunshine shelter 62.00 Collier Cone, collected in crater 62.50 Hayrick Butte, collected south of Hoodoo ski area 60.90 Same flow as 77, middle sister, collected northeast of Linton lake 62.50 South end of Linton Lake 61.60 Middle Sister, from large black hump on NW side 64.80 Kokostick Butte 63.40 Broken top lava, east of southeast lobe of Newberry rhyodacite flow 62.10 Middle Sister, collected 4.5 mi se of Linton Lake 65.20 From Middle Sister, high bluff along skyline trail above Sunshine Shelter 63.70 Lane Mesa, glassy dacite from Middle Sister 66.80 Kathleen Butte 72.30 Devils Lake rhyodacite 71.80 E base South sister, just west of Green Lake 71.50

Newberry rhyodacite 71.90 Rock Mesa 73.10 Obsidian Cliffs, collected near vent at west foot of middle sister 76.20 Obsidian Cliffs 75.50 Obsidian Cliffs flow, collected at midpoint 75.60 Map No. 78, West end of Potato Hill, no chem 44,66 Belknap, at Yapoah-Belknap-Little Belknap triple contact site 49.10 Little Belknap, at Yapoah-Belknap-Little Belknap triple contact site 52.24 Kipuka, north of Yapoah flow 51.08 Kipuka, rim of closed depression 51.41 Kipuka, location approximate 50.53 Kipuka, location approximate 51.20 Kipuka, vegetated kipuka visible on air photos 50.03 Dugout, from top of butte 51.84 Dugout, from hill 5245' 52.50 Kipuka, low-relief outcrop east of road 50.71 Bluegrass, outcrop next to road 52.73 Dugout, from north side of Belknap flow lobe 51.65 Bluegrass, prominent outcrop 52.18 Belknap, next to road (not on some maps) 48.80 Map No. 56, "Tma" N of Scar Mt 47.25 Map No. 32, Tp Dogami-Coffin-Buck Mt 50.27 Map No. 31, Tp Dogami -W of Fisher Pt 49.71 Map No. 36, Tp Dogami-nr Trappers Bt 51.15 Map No. 11, "Jorn Lk" lava S of Hoodoo Bt 51.09 Map No. 10, Turp Pk lava in Marion Ck cyn; = Conrey's RCTP-8 52.62 Map No. 23, Tp Dogami-E of Echo Mt 51.58 Map No. 27, Tp Dogami -S Pyramid 50.91 Map No. 25, Tp Dogami -S Pyramid 49.68 Map No. 26, Tp Dogami -S Pyramid 52.00 Map No. 52, "Tmb" SW of S Pyramid 48.79 Map No. 57, "Tmbl" up Straight Ck-prob Tma equiv 49.25 Map No. 21, Tp Dogami-E of Echo Mt-6.3 Ma 52.32 Map No. 1, Qhb Dogami NW Little Nash Crater 52.14 Map No. 34, Tp Dogami-Coffin-Buck Mt 52.35 Map No. 7, Ridge N of Marion Pk 52.50 Map No. 6, Marion Mt 53.35 Map No. 30, Tp Dogami-Crescent Mt 52.84 Map No. 24, Tp Dogami-NE Crescent Mt 52.53 Map No. 12, "Jorn Lk" lava SE of Maxwell Bt 53.51 Map No. 16, QTb-cap nr Fisher Pt 53.36 Map No. 51, "Tmb" W of Three Pyramids 53.02 Map No. 20, Tp Dogami-S of S Pyramid 53.31 Map No. 40, Tp Dogami-E Coffin Mt-5.8 Ma 54.18 Map No. 69, Basaltic andesite of Hoodoo Butte 55.07 Map No. 15, QTb-Caps bench SW of Marion Fks; poss older 55.08 Map No. 9, Turpentine Peak 55.60 Map No. 55, Tp on Scar Mt 53.71

Map No. 29, Tp Dogami -S Pyramid-6.3 Ma 54.80 Map No. 8, Turpentine Peak 55.65 Map No. 33, Tp Dogami-Coffin-Buck Mt 55.14 Map No. 22, Tp Dogami-E Echo Mt-7.2 Ma 55.15 Map No. 19, QTb 1.8 Ma intcyn bench in N Santiam 55.23 Map No. 4, Maxwell Butte lava 55.83 Map No. 62, Tp on Coffin Mt 54.86 Map No. 35, Tp Dogami-Coffin-Buck Mt 55.54 Map No. 28, Tp Dogami -S Pyramid 54.92 Map No. 5, Maxwell Butte lava 55.95 Map No. 58, "Tmbi" intrsn W of S Pyramid-elev. suggests Tp eqiv. 56.51 Map No. 47, Ta (rdgcap) -ScarMt-Trappers Bt 57.97 Map No. 65, Qmb-Maxwell Bt but W of N Santiam Hwy 56.80 Map No. 53, "Tmb" E of S Pyramid-prob Tma 58.00 Map No. 54, "Tma" in NW corner Crsnt Mt Qd 58.48 Map No. 48, Ta (rdgcap) -ScarMt-Trappers Bt 59.17 Map No. 63, "Tmai" intrsn W of Fisher Pt-age unk; poss=rdgcp 61.77 Map No. 59, "Tma" NW of 3 Pyramids 63.05 Map No. 50, Td (rdgcap) S of S Pyramid 63.20 Map No. 49, Td (rdgcap) S of S Pyramid 64.53 Map No. 64, Tbp pyr tuff NW crnr Crs Mt Qd 64.82 Red Creek Butte 64.82 Map No. 61, "Tma" SE of Coffin Mt-poss ridgecap seq 64.84 Map No. 14, Minto Ck streamcut = Crag Ck or Desch equiv 67.51 Map No. 60, "Tma" on Straight Ck-prob= landslid dacite 70.55 Map No. 17, QTb -Smith Prairie Bas (=HCK?) 48.03 e Prospect porph bas; xEProspect 47.7 thol w Prospect; xWProspect 47.9 Bedpan Ctr; bcp 47.9 bas s of 1530 cone; xSEBaldCrater 48.4 bas above 1525; xSEBaldCrater 48.5 Castle Pt cone, flow in gulch; bcp 49.4 thol L Cas Ck; bcc 49.7 HAOT E Castle Pt; bcp 49.4 Vent S of Bedpan Ctr; bcp 49.8 SSE Crater Pk woods; bsw 51.1 SSE Crater Pk woods; bsw 51.1 Diktytax S Varmit Camp; buc 51.3 Dike SE Varmit Camp; buci 51.3 N end Cas Pt lowest flow; bacp 52.0 Kimball Park; xKimballStPk 52.1 Thol W of Prospect; xWProspect 48.0 Sun Pass near top; xSunPass 52.5 SW of Boundary Butte; xWSWBndryB 52.0 Boundary Butte; bbp 52.8 Sand Ridge cone on rhd; bsr 52.1 E Annie Ck; bafe 52.4 Castle Ck; bcc 47.6

L Castle Ck; blccnp 52.6 Rogue at Farewell Bend; xRogue River 47.9 Rogue at Bybee Ck; xRogue River 47.8 Williams Crater inclusion; mw 52.9 Williams Crater scoria; bwp 50.7 Williams Crater scoria; bwp 51.5 Williams Crater inclusion; bwp 51.7 Williams Crater flow W ; bw 51.3 Williams Crater inclusion; mw 51.2 Williams Crater flow E; bw 51.5 clot basalt at E cone; bwn 52.6 Clot bas E spruce Lk; bwn 52.7 Steel Bay wall; bsbp 52.9 S of Desert Cone (op); brn 52.3 NW of Red Cone (ao); brc 52.4 Thol? NW of Red Cone; xSEBaldCrater 48.7 NW of Red Cone (opb); xSEBaldCrater 48.5 Sm Cn E of Bald Ctr (opb); xEBaldCrater 50.0 Bot thol E Bald Ctr; xEBaldCrater 48.4 Top thol E Bald Ctr; xEBaldCrater 48.4 Mid thol E Bald Ctr; xEBaldCrater 48.5 Top fr(W) thol E Bald Ctr; xEBaldCrater 48.6 2nd fr(N) thol below 1539; xEBaldCrater 48.8 E of Bald Crater; xEBald Ctr 48.6 Top fr(N) thol E Bald Ctr; xEBaldCrater 48.9 Hill 6062, N Des R Mg-ba; xNDesertRidge 52.0 W of 1740, skel ol; bbw 52.9 Oasis Butte ol and; bob 53.0 NW of Red Cone (ao); brc 53.1 Prob. vent older op Pole Br Ck; bbw 53.1 Cone E Cavern Ck; bce 52.7 Rock Concert hill; bob 53.2 N of Lookout Butte; blbtp 53.2 Scarp E of Bald Ctr (ao?); x-brc 53.3 ol bas n Ctr Pk; bcrnp 53.4 S. Oasis Butte ol rich; bob 53.5 Intrus SW Varmit Camp; buci 52.9 N of Hwy 138; xDiamond Lk 53.6 Hill 5955 N Gaywas Pk; xNGaywasPeak 53.6 Klamath Ridge; xKlamathRidge 53.6 Maklaks Crater; bmcr 53.6 Vent S Crater Pk; bcrsp 53.7 Cone S Oasis Butte; bob 53.7 Quarry E Rocktop Butte; buci 53.7 Great dike; bwci 53.7 NW of Red Cone (op); brw 53.7 Red Cone bomb; brcp 54.0 Southern cone Sand Ridge; xSand Ridge 53.8

N of Castle Ck lowest; bacp 53.8 W hill W of 1740; bbw 53.8 Desert Cone; xDesert Cone 53.8 E. Oasis Butte ol bas; brc 53.9 n Diamond Lk Jct; xNDiaLkJct97 53.9 S of Bald Crater; xBald Ctr 53.9 Bald Crater bomb; xBald Ctr 53.9 Older Castle Pt lava; bwc 54.0 S of Bedpan Ctr; bacp 53.4 S of Red cone; brc 54.0 SW ridge below Union Pk; bupi 53.5 Sun Pass - Sand Ridge; xNW Sun Pass 54.0 Great dike; bwci 53.9 N Castle Pt top; bacp 54.0 Hill NNW hill 6589 NNE GP; xNGaywasPeak 54.1 Great dike; bwci 54.1 Hill 6256 N Gaywas Pk; xNGaywasPeak 54.2 Hill 6509, N Desert Ridge; xNDesertRidge 54.2 Intrus in Union Ck valley; bupi 54.2 Buckeye Butte; xBuckeye Bte 54.2 Hill NE hill 6589 NNE GP; xNGaywasPeak 54.3 Vent N of Bald Ctr; xNBaldCrater 54.3 Hill 6428, N hill 6712; bwbp 54.3 W Union Pk pa; bupi 54.4 Scoria Cone agglutinate; bsc 54.4 N of Castle Ck cliff; blcnw 54.4 Great dike; bwci 54.4 Bomb Whitehorse Bluff; bwcp 54.5 Great dike; bwci 54.5 Bomb cone ENE Boundry Butte; bbnp 54.6 Annie Falls; bsc 54.7 Conelet N TC, S of road; xNTimberCrtr 54.8 SW of Varmit Camp (PEW); buc 54.2 W hill 6577; bsc 54.9 N of pass btw PumFl & PCT; bpf 55.0 Un Pk lava, l of 2 flows; bup 55.1 Un Pk lava, u of 2 flows; bup 55.1 Top NE wall glac valley Union Pk; bup 55.2 Bomb cinder ridge S of hill 6712; buep 55.3 Hill 6712, W Arant Pt; baw 55.3 Dike N base hill 6926; bpfi 55.4 S Union Pk (clots); bup 55.4 W Arant Pt (clots); bpf 55.5 Int ESE Union Pk; bpwi 55.5 W Stuart Falls trail; bsf 55.6 SW of Arant Pt base; baw 55.6 Whitehorse Bluff lwr; bmcw 55.7 N of Castle Ck top; blcnw 55.7

Union Pk int core; bupi 55.8 SE of Castle Pt; bws 55.8 Cone E Sand Ck; bdep 55.9 E of hill 6315; bpw 56.0 Hill betw Whthrse Bl & Un Pk; bwsi 56.0 Sm hill NNE hill 6256; xNGaywasPeak 56.1 S entrance scarp top; baw 56.2 w Welch butte; xWWelchButte 56.3 Bomb sm cone N of 1618 cone; bbnp 56.3 Whitehorse Bluff W of hwy62; bwb 56.3 E Whitehorse Bluff; bwb 56.3 Rim flow glac valley S of Castle Pt; bup 56.4 Hill 6315; bup 56.5 Crater Pk bomb; acrp 56.9 Lower Sun Ck; acr 57.0 N of Sand Ck; bsd 57.1 Ctr Pk; acr 57.2 Arant Pt; apt 57.3 Arant Pt; apt 57.2 Scott Ck; asc 57.3 Quarry S Quillwort Pond; apt 57.4 SE Crater Pk and; acr 57.4 Dike S of Applegate; aap 57.4 E. Annie Ck; aafn 57.4 Arant Pt chill; aptg 57.6 Annie Ck near E fork; acr 57.6 Lower Sun Ck; acr 57.6 Annie Ck near E fork; acr 57.6 Welch Butte; xWelchButte 57.7 Crater Pk lava; acr 57.7 Arant Pt; apt 58.0 Arant Pt; apt 57.8 Scott Ck; asc 58.0 N Pum Fl, SW hill 6442; abb 58.1 Scott Ck cone; asc 58.1 Timber Ctr ; xTimberCtr 58.2 Timber Ctr high E; xTimberCtr 58.6 Timber Crater near summit; xTimber Ctr 58.6 N of L Castle Ck; accnp 58.7 Scott Ck; asc 58.8 W slope Timber Crater; xTimber Ctr 58.8 Timber Ctr ol rich; xTimberCtr 58.6 Quarry S Sand Ck; asq 58.8 N T Ctr, from NNW vent; xTimberCrtr 58.9 Timber Ctr SE; xTimberCtr 58.8 N of Pothole Ck; albt 59.0 Hill 6366, SW Desert Rdge; xSWDesertRdg 59.3 Gaywas Pk; xGaywasPeak 59.3

NW summit Desert Ridge; xDesertRidge 59.4 Summit Desert Ridge; xDesertRidge 59.5 E Bear Bluff; abb 59.5 Hill 6589, NNE Gaywas Pk; xNGaywasPeak 59.9 Whitehorse Bluff uppr; bmcw 59.9 Sand Ck xtl poor and; ascs 61.3 SE Mt Scott; acnp (akn) 62.1 Aphyc below 1229 Sand Ck; ascs? 62.5

Appendix D: Table of Cones

Outer Silica Slope Inner Content (m/m) - Age Slope (%) Silica Source ArcGIS (years) Age Source Method (m/m)

Wampus 54.08 Sherrod 1997 0.490 43000 (USGS) K-Ar 0.430 Gilcrest 56.30 Sherrod 1997 0.250 510000 (USGS) K-Ar 0.362 Graham 53.02 Schilling 1975 0.330 120000 (USGS) K-Ar 0.330 Pilot 52.900 Sherrod 1997 0.592 500000 (USGS) K-Ar 0.301 Lava 47.59 Sherrod 1997 0.603 7000 O'Connell 1967 C-14 0.450 Nash 52.60 Sherrod 1997 0.604 2590 (USGS) C-14 0.446 Le Conte 55.50 Sherrod 1997 0.590 2300 (USGS) C-14 0.465 Forked 53.31 Sherrod 1997 0.576 5600 (USGS) C-14 0.472 Egan 51.10 Sherrod 1997 0.557 7600 (USGS) C-14 0.491 Triangle 59.03 Sherrod 1997 0.326 310000 (USGS) Ar-Ar 0.210 Pilpil 54.97 Sherrod 1997 0.394 7000 O'Connell 1967 C-14 0.630 Lava top 49.90 Schilling 1975 0.356 72000 (USGS) Ar-Ar 0.545 Wickiup 55.61 Schilling 1975 0.268 590,000 Wood 1980 K-Ar 0.390 Wanoga 55.61 Sherrod 1997 0.304 60,000 Wood 1980 K-Ar 0.587 McKinney 62.20 Sherrod 1997 0.368 320,000 Sherrod 1997 K-Ar 0.413

Mokst 55.50 Sherrod 1997 0.514 7000 Donnelly-Nolan Ar-Ar 0.524 Cayuse Crater 56.36 Schilling 1975 0.625 9,100 C-14 O'Connell 1967 0.424 Talapus 51.20 Schilling 1980 0.540 5,300 USGS C-14 0.496 Katsuk 54.30 Sherrod 1997 0.594 34,000 USGS Ar-Ar 0.413 Powell 54.30 Sherrod 1997 0.550 41,500 Duane 2004 Ar-Ar 0.725 Davis 48.50 Schilling 1981 0.599 2250 McGuire 2014 Ar-Ar 0.448 Red Crater 56.40 Sherrod 1997 0.387 118000 McGuire 2014 Ar-Ar 0.450 Twin Butte 51.30 Sherrod 1997 0.213 300000 McGuire 2014 Ar-Ar 0.230 Sand Mountain 51.08 Schilling 1982 0.556 3000 Cashman 2012 C-14 0.487 Garrison Butte 51.41 Sherrod 1997 0.430 40390 McGuire 2014 Ar-Ar 0.326 Collier 50.53 Sherrod 1997 0.600 1500 Sherrod 1997 C-14 0.458 Four in One 51.2 Schilling 1983 0.612 1980 Sherrod 1997 C-14 0.435 Kokostick Butte 50.03 Sherrod 1997 0.331 24500 Donnely-Nolan Ar-Ar 0.320 Union Butte 51.84 Sherrod 1997 0.485 45000 Duane 2004 K-Ar 0.464 Lookout 52.50 Schilling 1984 0.308 211000 Duane 2004 K-Ar 0.314 Horse Butte 50.71 Sherrod 1997 0.434 125000 Donnely-Nolan Ar-Ar 0.311 Tipton 52.73 Sherrod 1997 0.601 7600 McGuire 2014 C-14 0.453 Forest Road 51.65 Schilling 1985 0.594 6500 Donnely-Nolan C-14 0.456 Hidden 52.18 Sherrod 1997 0.602 5500 Donnely-Nolan C-14 0.456 Gas Line 48.80 Sherrod 1997 0.590 7200 Shilling 1975 C-14 0.454 North Sugarpine 47.25 Schilling 1986 0.595 7300 Sherrod 1997 C-14 0.444 Indian Heaven 50.27 Sherrod 1997 0.591 6890 (USGS) C-14 0.453 Green Mt 49.71 Sherrod 1997 0.502 53000 (USGS) Ar-Ar 0.487 Picque Butte 51.15 Schilling 1987 0.549 26000 (USGS) Ar-Ar 0.456 Black Hill 51.09 Sherrod 1997 0.562 56300 (USGS) Ar-Ar 0.409 Candlestick 52.62 Sherrod 1997 0.600 84000 Schilling 1975 K-Ar 0.297 Iris 51.58 Schilling 1988 0.554 11200 (USGS) K-Ar 0.477 Bessie 50.91 Sherrod 1997 0.416 520000 (USGS) Ar-Ar 0.354 North Line 49.68 Sherrod 1997 0.592 8000 (USGS) C-14 0.440 Graywas Peak 52.00 Schilling 1989 0.539 17000 (USGS) Ar-Ar 0.464 Yapoah Crater 48.79 Sherrod 1997 0.596 1200 (USGS) C-14 0.453 Rocky Butte 49.25 Sherrod 1997 0.323 8300 O'Connell 1967 C-14 0.640 Twin Lakes 52.32 Schilling 1990 0.542 42000 (USGS) Ar-Ar 0.450 Sand Mountain 52.14 Sherrod 1997 0.565 2000 Wood 1980 C-14 0.454 Garrison Butte 52.35 Sherrod 1997 0.565 55000 Wood 1980 K-Ar 0.433 Lookout 52.50 Schilling 1991 0.594 7800 Sherrod 1997 C-14 0.450 Four in One 53.35 Sherrod 1997 0.564 23000 Sherrod 1997 K-Ar 0.340 Desert Cone 52.84 Sherrod 1997 0.235 41000 O'Connell 1967 C-14 0.670 Klawhop 52.53 Schilling 1992 0.354 35000 USGS Ar-Ar 0.550 Kelsey 53.51 Sherrod 1997 0.543 57000 USGS Ar-Ar 0.401 Kimball Peak 53.36 Sherrod 1997 0.452 19000 Duane 2004 Ar-Ar 0.474 Yapoah Crater 53.02 Schilling 1993 0.595 1130 USGS C-14 0.454 Bunchgrass 53.31 Sherrod 1997 0.4543 102000 CVO K-Ar 0.345

Twin Lakes 54.18 Sherrod 1997 0.335 4500 CVO C-14 0.653 Mount Talbert 55.07 Schilling 1994 0.489 57600 Schilling 1975 K-Ar 0.454 Hoodoo 55.08 Sherrod 1997 0.587 2000 CVO C-14 0.425 Kostal 55.60 Sherrod 1997 0.113 250000 CVO Ar-Ar 0.382 Pine 53.71 Schilling 1995 0.356 37000 Sherrod 1997 K-Ar 0.544 Bluegrass 54.80 Sherrod 1997 0.4332 140000 Sherrod 1997 K-Ar 0.345 Delaney 55.65 Sherrod 1997 0.5645 18000 Duane 2004 Ar-Ar 0.457 Eagle Crest 55.14 Schilling 1996 0.5785 30000 Schilling 1975 K-Ar 0.458 Dimple Hill 55.15 Sherrod 1997 0.331 130000 Schilling 1975 K-Ar 0.432 Castle Creek 55.23 Sherrod 1997 0.485 65000 Schilling 1975 K-Ar 0.394 Welch Butte 55.83 Schilling 1997 0.578 82000 Schilling 1975 K-Ar 0.354 Bald Crater 54.40 Sherrod 1997 0.434 102000 Schilling 1975 K-Ar 0.325 Bedpan Crater 52.40 Sherrod 1997 0.554 89000 Schilling 1975 K-Ar 0.350 Arant Point 55.50 Schilling 1998 0.499 134000 Schilling 1975 K-Ar 0.303 Sun Pass 54.90 Sherrod 1997 0.505 293000 Schilling 1975 K-Ar 0.430 Whitehorse Bluff 55.70 Sherrod 1997 0.552 19700 Schilling 1975 K-Ar 0.453 Annie Butte 53.50 Schilling 1999 0.591 41000 Schilling 1975 K-Ar 0.393 Boundary Butte 56.70 Sherrod 1997 0.610 14000 Schilling 1975 K-Ar 0.416 Williams Crater 51.70 Sherrod 1997 0.540 84000 Schilling 1975 K-Ar 0.343 Kathleen Butte 54.30 Schilling 2000 0.562 13700 Schilling 1975 K-Ar 0.454 Hogg Butte 62.10 Sherrod 1997 0.600 30000 Schilling 1975 K-Ar 0.402 Prospect Ridge 54.18 Sherrod 1997 0.554 23000 Schilling 1975 K-Ar 0.454 Maxwell 55.07 Schilling 2001 0.516 18000 Schilling 1975 K-Ar 0.504 Pyramid 55.08 Sherrod 1997 0.492 133000 Schilling 1975 K-Ar 0.343 Marion Butte 55.60 Sherrod 1997 0.439 80140 Schilling 1975 K-Ar 0.445 Turpentine Ridge 53.71 Schilling 2002 0.564 20000 Sherrod 1997 K-Ar 0.453 Coffin Butte 54.80 Sherrod 1997 0.323 250500 USGS Ar-Ar 0.438 Scar Peak 55.65 Sherrod 1997 0.542 31000 USGS Ar-Ar 0.454 Trappers Butte 55.14 Schilling 2003 0.453 127000 O'Connell 1967 K-Ar 0.382 Castle Point 55.15 Sherrod 1997 0.653 38000 O'Connell 1967 K-Ar 0.342 Rogue Butte 55.23 Sherrod 1997 0.343 79000 O'Connell 1967 K-Ar 0.550 Red Creek 55.83 Schilling 2004 0.564 200000 O'Connell 1967 K-Ar 0.453 Buck Butte 54.40 Sherrod 1997 0.590 4700 O'Connell 1967 K-Ar 0.453

Appendix E Precipitation and Rim Angle

Cone Precipitation Crater Range Rim (cm/year)- angle PRISM Data (°) Wampus 0-57.15 115.0 Gilcrest 0-57.15 130.0 Graham 0-57.15 120.0 Pilot 57.15-101.6 135.0 Lava 57.15-101.6 111.5 Nash 57.15-101.6 95.0 Le Conte 57.15-101.6 95.0 Forked 57.15-101.6 101.0 Egan 57.15-101.6 108.0 Triangle 254+ 129.9 Pilpil 152-254 109.0 Lava top 154-254 118.6 Wickiup 0-57.15 136.0 Wanoga 0-57.15 118.0 McKinney 0-57.15 128.0 Mokst 57.15-101.6 112.5 Cayuse Crater 57.15-101.6 109.0 Talapus 0-57.15 103.0 Katsuk 0-57.15 115.6 Powell 152-254 116.0 Davis 152-254 96.0 Red Crater 152-254 119.0 Twin Butte 101.6-152 134.0 Sand Mountain 101.6-152.4 93.0 Garrison Butte 152-254 116.6 Collier 0-57.15 94.0

Four in One 57.15-101.6 97.0 Kokostick Butte 101.6-152.4 119.4 Union Butte 254+ 117.3 Lookout 154-254 121.6 Horse Butte 57.15-101.6 119.5 Tipton 57.15-101.6 105.0 Forest Road 57.15-101.6 106.0 Hidden 0-57.15 105.0 Gas Line 0-57.15 109.0 North Sugarpine 0-57.15 110.0 Indian Heaven 0-57.15 111.0 Green Mt 0-57.15 123.0 Picque Butte 57.15-101.6 113.0 Black Hill 57.15-101.6 115.0 Candlestick 0-57.15 114.0 Iris 0-57.15 114.0 Bessie 152-254 131.0 North Line 152-254 110.0 Graywas Peak 154-254 113.0 Yapoah Crater 57.15-101.6 101.0 Rocky Butte 57.15-101.6 104.0 Twin Lakes 57.15-101.6 108.0 Sand Mountain 0-57.15 98.0 Garrison Butte 0-57.15 113.0 Lookout 0-57.15 109.0 Four in One 0-57.15 111.0 Desert Cone 0-57.15 113.0 Klawhop 57.15-101.6 112.0 Kelsey 57.15-101.6 113.0 Kimball Peak 0-57.15 111.0 Yapoah Crater 0-57.15 101.0 Bunchgrass 152-254 119.0 Twin Lakes 152-254 101.0 Mount Talbert 152-254 113.0 Hoodoo 101.6-152 99.0 Kostal 101.6-152.4 121.0 Pine 154-254 113.0 Bluegrass 57.15-101.6 114.0 Delaney 57.15-101.6 110.0 Eagle Crest 57.15-101.6 110.0 Dimple Hill 0-57.15 115.0

Castle Creek 0-57.15 114.0 Welch Butte 0-57.15 116.0 Bald Crater 0-57.15 118.0 Bedpan Crater 0-57.15 114.0 Arant Point 57.15-101.6 118.0 Sun Pass 57.15-101.6 129.0 Whitehorse Bluff 0-57.15 112.0 Annie Butte 0-57.15 111.0 Boundary Butte 152-254 110.0 Williams Crater 152-254 116.0 Kathleen Butte 152-254 113.0 Hogg Butte 154-254 114.0 Prospect Ridge 57.15-101.6 114.0 Maxwell 57.15-101.6 115.0 Pyramid 57.15-101.6 118.0 Marion Butte 0-57.15 116.0 Turpentine Ridge 0-57.15 121.0 Coffin Butte 0-57.15 116.0 Scar Peak 0-57.15 119.0 Trappers Butte 0-57.15 116.0 Castle Point 57.15-101.6 116.0 Rogue Butte 57.15-101.6 119.0 Red 0-57.15 118.8 Buck 0-57.15 115.6

References Anderson, R., Densmore, A., and Ellis, M. 1999. The generation and degradation of marine terraces. Basin Research. 11, 7–19. Andrews, D.J., and Bucknam, R.C., 1987, Fitting degradation of shoreline scarps by a nonlinear diffusion model: Journal of Geophysical Research, v. 92, p. 12,857–12,867. Armstrong, A., 1976, A three-dimensional simulation of slope forms: Zeitschrift fur Geomorphologie, Supplementband, v. 25, p. 20-28. Bemis, K. 2011. The growth and erosion of cinder cones in Guatemala and El Salvador: Models and statistics, J. Volcanol. Geotherm. Res. de Michieli Vitturi, M. and Arrowsmith, J R. 2013. Two dimensional nonlinear

diffusive numerical simulation of geomorphic modifications to cinder cones, Earth Surface Processes and Landforms, doi:10.1002esp.3423. Dietrich, W. E., Bellugi, D. G., Sklar, L. S., Stock, J. D., Heimsath, A. M., and Roering, J. J. 2003, Geomorphic transport laws for predicting landscape form and dynamics, in Wilcock, P. R., and Iverson, R. M., eds., Prediction in Geomorphology: Washington, DC, Geophysical Monograph Series, v. 135, American Geophysical Union, p. 103-132 Forniciai, A. 2011. Morphometry of scoria cones, and their relationship to geodynamic setting: A DEM based analysis. Journal of Volcanology and Geothermal Research. Gilichinsky, M., D. Melnikov, I. Melekestsev, N. Zaretskaya, and M. Inbar. 2010. Morphometric measurements of cinder cones from digital elevation models of Tolbachik volcanic ﬁeld, central Kamchatka, Can. J.Remote Sens., 36, 287– 300. Howard, A. D. 1997, Badland morphology and evolution: Interpretation using a simulation model, Earth Surf. Processes Landforms, 22, 211–227. Koleszar, A. M., Kent, A. J. R., Wallace, P. J. & Scott, W. E. 2012. Controls on long- term low explosivity at andesitic arc volcanoes: insights from Mount Hood, Oregon. J. Volcanol. Geotherm. Res. 219–220, 1–14. Kneissl, T., S. van Gasselt, and G. Neukum. 2011. Map-projection-independent crater size-frequency determination in GIS environments – new soft-ware tool for ArcGIS, Planet. Space Sci., 59,1243–1254, doi:10.1016/j.pss.2010.03.015 Kervyn, M., Ernst, G., Carracedo, J.-C., Jacobs, P., 2010. Geomorphology of “monogenetic” volcanic cones, submitted to Geomorphology. Kuiper, K. F., Deino, A., Hilgen, F. J., Krijgsman,W., Renne, P. R., Wijbrans, J. R., 2008. Synchronizing the Rock Clocks of Earth history. Science 320 (5875), 500–504. Ma, R., 2005. DEM Generation and Building Detection from Lidar Data. Photogram Eng Remote Sensing. 71 (7) 847-854. McGuire, L. A., J. D. Pelletier, and J. J Roering. 2014. Development of topographic asymmetry: Insights from dated cinder cones in the western United States, J.

Geophys. Res. Earth Surf., 119(8), 1725–1750, doi:10.1002/2014JF003081. Mckay, D., J. Donnelly Nolan, R. Jensen, and D. Champion. 2009. The post- Mazama northwest rift zone eruption at Newberry Volcano, Oregon, in Volcanoes to Vineyards: Geologic Field Trips Through the Dynamic Landscape of the Pacific Northwest, Field Guide, vol. 15, edited by J. E. O’Connor et al., pp. 91–110, Geol. Soc. of Am., Boulder, Colo., doi:10.1130/2009.fld015(05). Mitchell, J.G. 1968. The argon-40 argon-39 method for potassium-argon age determination. Geochimica et Cosmochimica Acta,Volume 32, Issue 7. Pages 781-790, https://doi.org/10.1016/0016-7037(68)90012-4. Molotch, N. P., M. T. Colee, R. C. Bales, and J. Dozier. 2005. Estimating the spatial distribution of snow water equivalent in an alpine basin using binary regression tree models: The impact of digital elevation data and independent variable selection, Hydrol. Processes, 19(7), 1459–1479, doi:10.1002/hyp.5586.. Nieto-Torres, A., Lillian, M, 2019. Spatio-temporal hazard assessment of a monogenetic volcanic field, near México City. Journal of Volcanology and Geothermal Research. 371 46-58. Northwest Alliance for Computational Science & Engineering (NACSE). 2011. Average Annual Precipitation for Oregon (1980-2010). USDA Risk Management Agency. Pelletier, J. D., and M. L. Cline. 2007. Nonlinear slope dependent sediment transport in cinder cone evolution, Geology, 35, 1067–1070. Pierce, K., Coleman, S. 1986. Effect of height and orientation (microclimate) on geomorphic degradation rates and processes, late-glacial terrace scarps in central Idaho. Article in Geological Society of America Bulletin 97(7). Porter, S.C., 1972. Distribution, morphology, and size distribution of cinder cones on Mauna Lea Volcano, Hawaii. Geological Society of America Bulletin 83, 3607–3612. Ross and Chanceman. 2015. Investigating Statistical Concepts, Applications, and Methods. Taylor, S. B., J. H. Templeton, and D. E. L. Giles. 2003. Cinder cone morphometry

and volume distribution at Newberry Volcano, Oregon: Implications for age relations and structural control on eruptive process [Abstracts with Programs], Geol. Soc. Am., 35(6), 421. Sherrod, D.R., Mastin, L.G., Scott, W.E., and Schilling, S.P., 1997, Volcano hazards at Newberry Volcano, Oregon: U. S. Geological Survey, Open-file Report 97- 513, 14. Schilling, J.H. 1975. Potassium-Argon Ages of Volcanic Rocks, Central Cascade Range of Oregon. In Isochron/West: A Bulletin of Isotopic Geochronology. Volume 13, Issue 2. Stewart, I. T. 2009. Changes in snowpack and snowmelt runoff for key mountain regions, Hydrol. Proc., 23, 78–94. Whitlock. 1992. Vegetational and climatic history of the Pacific Northwest during the last 20,000 years: The Northwest Environmental Journal 8, 5-28. West, N., Kirby, E., Bierman, P.R., and Clarke, B.A., 2014, Aspect-dependent variations in regolith creep revealed by meteoric 10Be: Geology. Wood, C.A., 1980a. Morphometric evolution of cinder cones. Volcanol. Geotherm. Res., 7: 387--413. Zappa, M., Pos, F., Strasser, U., Warmerdam, P., and Gurtz, J. 2003. Seasonal water balance of an alpine catchment as evaluated by different methods for spatially distributed snowmelt modelling, Nord. Hydrol., 34, 179–202.