PERSPECTIVES ON DROUGHT AND TEMPERATURE VARIABILITY FOR THE
SOUTHWESTERN UNITED STATES FROM A NEW HYDRO-ISOTOPIC NETWORK
by
Max B. Berkelhammer
A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(GEOLOGICAL SCIENCES)
December 2010
Copyright 2010 Max B. Berkelhammer
Epigraph
There is no such thing as a long piece of work, except one that you dare not start.
-Charles Baudelaire
ii
Acknowledgements
I would like to express my profound gratitude to the many individuals who have contributed to the
completion of this work. It has been an honor and pleasure to interact and work intimately with so
many gifted and kind people for the past five years.
Namely, Dr. Lowell Stott, who will be a life-long friend and mentor. Miguel Rincon whose immense
patience allowed me to finish this work despite arduous and frustrating battles with various instru-ments.
Melanie Gerault who’s genius and warmth was my greatest inspiration. My various lab
mates over the years including Andres Martinez, Reetta Saikku, Deborah Khider, Mengfan Zhu and
especially Patrick Horan who was unlucky enough to have to share an office with me for the last 4
years. Ashish Sinha and Mark Bernstein for critical discussion on topics other than what is included
in this thesis.
My dear friends including (but not limited to) Byron Kahr, Ryan Adlaf, Nizar Wattad and John
Nixon. My parents, especially my father who showed enough interest in my work to follow me into
caves in remote corners of India .
Lastly, I need to thank Kei Yoshimura for providing the IsoGSM data and helping me to run my first
climate model simulations, Tom Harlan who taught me about cross-dating, Kevin Anchukaitis who
hosted me at the Tree Ring Lab, Chris Lehmann from the National Atmospheric Deposition pro-gram
for providing the precipitation samples, and my thesis committee, Donal Manahan and Julien
Emile-Geay for providing critical feedback.
iii
Table of Contents
Epigraph ii
Acknowledgements iii
List of Figures vii
Abstract xvi
Chapter 1 Introduction 1
1.1 Climate forecasts for southwestern US . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Tree-ring drought reconstructions . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 Tree-ring temperature reconstructions . . . . . . . . . . . . . . . . . . . . . . . . 5
1.4 Climatic information from hydro-isotopes . . . . . . . . . . . . . . . . . . . . . . 6
1.5 Thesis Outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
Chapter 2 Atmospheric circulation and the isotopic composition of precipitation over
the western US 10
2.1 Introductory note . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.2 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.3 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.3.1 Locations and Samples . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.3.2 Analytical Techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.3.3 Numerical Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.3.4 A Lagrangian Assumption . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.4.1 Model Validation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
2.4.2 Regional and Synoptic Controls . . . . . . . . . . . . . . . . . . . . . . . 23
2.4.3 Mesoscale Controls . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
2.4.4 Water-tagging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
2.4.5 Deuterium-excess gradients . . . . . . . . . . . . . . . . . . . . . . . . . 37
2.5 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
2.5.1 Isotopes and 21st century hydrologic changes . . . . . . . . . . . . . . . . 38
2.5.2 Isoscapes and Proxy reconstructions . . . . . . . . . . . . . . . . . . . . . 42
2.6 Isotopic controls at Inland Sites . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
2.7 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
iv
Chapter 3 Testing models of the mechanistic controls on the isotopic composition of
cellulose using intra-annual sampling 50
3.1 Introductory note . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
3.2 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
3.3 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
3.3.1 Age model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
3.3.2 Analytical methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
3.4 Site descriptions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
3.4.1 White Mountains . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
3.4.2 Almagre Mountains . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
3.5 Cellulose Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
3.5.1 The Source Water Term . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
3.5.2 The leaf water term . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
3.5.3 Model parameters-White Mountains . . . . . . . . . . . . . . . . . . . . . 67
3.5.4 Model parameters-Almagre Mountains . . . . . . . . . . . . . . . . . . . 68
3.6 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
3.6.1 White Mountains . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
3.6.2 Almagre Mountain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
3.7 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
3.7.1 White Mountains . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
3.7.2 Validation of Model Assumptions . . . . . . . . . . . . . . . . . . . . . . 75
3.7.3 Almagre Mountains . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
3.8 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
Chapter 4 A growth-independent temperature reconstruction for the southwestern
United States 81
4.1 Introductory note . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
4.2 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
4.2.1 Tree Cellulose Paleothermometry . . . . . . . . . . . . . . . . . . . . . . 86
4.3 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
4.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
4.5 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
4.5.1 Tree ring response to temperature . . . . . . . . . . . . . . . . . . . . . . 96
4.5.2 Hypothesis testing with a paleo-gcm simulation . . . . . . . . . . . . . . . 100
4.6 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
Chapter 5 Insights into the mechanisms that generate drought in the southwestern
US derived from the isotopic composition of tree-ring cellulose 103
5.1 Introductory note . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
5.2 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
5.2.1 Hydroclimatology of the southwestern US . . . . . . . . . . . . . . . . . . 106
5.3 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
5.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
5.4.1 White Mountain Site . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
5.4.2 Alta Peak Site . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
v
5.4.3 Geochemical modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
5.5 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
5.5.1 Isotope relation to drought . . . . . . . . . . . . . . . . . . . . . . . . . . 117
5.5.2 Low Frequency 20th century trends . . . . . . . . . . . . . . . . . . . . . 124
5.5.3 The 19th Century Transition . . . . . . . . . . . . . . . . . . . . . . . . . 125
5.6 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
Chapter 6 Enigmatic isotopic responses to Greenland Interstadials in caves from the
southwestern US 135
6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
6.2 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
6.3 Fort Stanton and Cave of the Bells . . . . . . . . . . . . . . . . . . . . . . . . . . 140
6.4 Isotopic Modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
6.5 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
6.5.1 Timing and Shape of Events . . . . . . . . . . . . . . . . . . . . . . . . . 143
6.5.2 Response between North Atlantic and Southwestern US . . . . . . . . . . 150
6.6 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
6.7 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
References 157
Appendix 180
vi
List of Figures
1.1 Gridded correlation coefficient (r) between d18O of precipitation and surface tem-perature
and precipitation amount. Data used are from (Yoshimura et al., 2008). . 7
1.2 Photo of White Mountain Bristlecone Pine stands during the June 2008. This photo
serves to provide some visual reference to the reader on the environment at which
the trees discussed in Chapters 3 and 5 grow. . . . . . . . . . . . . . . . . . . . . 9
2.1 Map showing all isotopic monitoring sites referred to in this chapter. . . . . . . . 16
2.2 Relationship of d18O and dD (Local Meteoric Water Line) for all measurements
made for this study. The slope for each site is indicated on the figure. The global
average slope is 8 and the low values as observed at the DV site indicate evaporative
enrichment of the sample. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.3 Storms following a latitudinal transect (top), an altitudinal transect (center) and dxs.
Landfalling storms produce increasingly depleted d18O with increasing latitude and
altitude. dxs does not display a strong altitude effect, but does follow latitude. At
each site a probability distribution function using a normal kernel density estimator
was fitted to the data. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
2.4 Monthly distributions of storm events at all sites for oxygen and dxs. The isotopic
composition of each storm is corrected to the average of the site and then box plots
with quartiles are made for each month. While there is a seasonal cycle where mean
values in the summer are higher than the winter, the distribution of individual storms
overlaps almost completely between months. Deuterium-excess has a more defined
seasonality, with extremely negative values occurring during the summer months.
Outlier values are marked by open circles. . . . . . . . . . . . . . . . . . . . . . . 25
2.5 Comparison between d18O values of measured storm event and their predicted val-ues
based on the IsoGSM simulation. The mean isotopic values were subtracted
from each dataset to correct for the coarse topography in the model simulation (left
panel). The colors are used to denote the different sites. The distribution of the same
storm events shown in the left panel, bins are 1‰ and the line is the best fit Gaussian
distribution (right panel). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
vii
2.6 Time evolution of d18O of precipitable water (left column), precipitation rate (mid-dle
column), and atmospheric specific humidity (right column) during a sequence
of the most enriched (top row) and depleted (bottom row) storm events. The mean
value for the different storms are shown as a bold gray line with circle markers.
All values were taken from the IsoGSM simulation. Time 0 represents an arbitrary
beginning point before precipitation began to fall moving forward in 6-hour time
steps. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
2.7 Difference between latent heat flux (w/m2) for the enriched and depleted compos-ites.
High latent heat flux from just offshore is a common feature of the most
enriched events. Data for latent heat is from the North American Regional Reanal-ysis
dataset (Mesinger et al., 2006). . . . . . . . . . . . . . . . . . . . . . . . . . 29
2.8 Top panels (left and right) show the precipitation rates in six hour time steps as two
isotopically enriched storm events strike the western US. The bottom panels show
vertical cross sections of the isotopic composition of water vapor as the storms pass
over the region. The figure shows how the maximum isotopic changes (on the order
8-10 ) occur between the 800-700 mb levels. . . . . . . . . . . . . . . . . . . . . 29
2.9 A plan view of the average 850 mb wind fields during the most depleted (left) and
enriched (center) events and the difference between the two vector fields (right).
Scale bar shows the length of a 10 m/s vector. . . . . . . . . . . . . . . . . . . . . 30
2.10 Isotopic concentrations of d18O of water vapor during depleted (left) and enriched
(right) events. Water vapor is taken for the 850 mb level. Colors show isotopic
anomalies relative to the field in view while contours show absolute isotopic con-centration
relative to VSMOW. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
2.11 Correlation coefficient between annually average vertically integrated meridional
moisture flux and amount weighted d18O of precipitation over the southwestern
US. Contours indicate correlations that are significant at the 95% confidence based
on a Student’s T-test. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
2.12 Correlation coefficients between meridional moisture flux and the d18O of vapor
for a vertical cross section of the Pacific between 120-180oW (bottom). Correlation
coefficients between vertical velocity (omega) and the d18O of vapor for a vertical
cross section of the Pacific between 120-180oW (top). . . . . . . . . . . . . . . . 34
2.13 A cross section across the Pacific basin with contours showing the average isotopic
concentration of water vapor and colors showing vertical velocities (positive values
are dark gray and negative values are orange). . . . . . . . . . . . . . . . . . . . . 35
viii
2.14 The relationship between the relative percentage of tagged water in the atmospheric
column over southwestern US and the isotopic composition of the integrated water
column (left). All days associated with storm events were selected from the figure
on the left showing the coherent relationship between tagged water and the isotopic
composition of water during landfalling frontal storms. . . . . . . . . . . . . . . . 35
2.15 A composite of tagged water concentration for a series of isotopically enriched
(right) and depleted (left) storm events. The concentration of tagged water in the
atmospheric column is taken as the ratio of the mass of tagged water to total water
(e.g. specific humidity). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
2.16 Plan view of the 850mb wind fields and relative percentage of tagged water in the
atmospheric column during the atmospheric river event on January 22, 2005. . . . 36
2.17 Precipitation rate (kg of water/m2) during storms included in the high (left), low
(middle) and average (right) dxs gradient events. The figure emphasizes that the
storms influenced the entire coast. . . . . . . . . . . . . . . . . . . . . . . . . . . 39
2.18 Vertically integrated meridional moisture flux (kg/m) during storms included in the
high (left) and low (middle) dxs gradient events and the difference between the two
composites (right). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
2.19 Latent heat flux (W/m2) during storms included in the high (left) and low (middle)
dxs gradient events and the difference between the two composites (right). . . . . 40
2.20 Average annual 850 mb geopotential height (m) and wind vector anomalies from
NCEP 2 Reanalysis (Kanamitsu et al., 2002) during 1989, 1998 and 2003 (left to
right, top row) and the isotopic anomalies in precipitation associated with these
same years (bottom row). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
2.21 Isotopic composition of 112 individual storm events between 1994-1998 from the
Global Network of Isotopes in Precipitation event station in the Pawnee Grasslands
in Colorado plotted against surface temperatures during the events. The right panel
shows the slopes of the regression after storms have been binned by season and
prevailing trajectory. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
2.22 Slope of monthly integrated d18Op and temperature at the 39oN and 105oW grid
point in the IsoGSM (Yoshimura et al., 2008), Echam4 (Hoffmann et al., 1998) and
GissE (Schmidt et al., 2007) isotope-enabled GCM simulations. The relationship is
highly significant in all three models and the slopes are nearly equivalent though the
intercept is larger in the nudged IsoGSM simulation than observed in the other two
models. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
ix
2.23 Probability Distribution Functions of Dd18O-DT for sliding 5-year windows encom-passing
the entire length of the simulations from preceding Figure. Positive (nega-tive)
values indicate 5-year windows in which the slope was steeper (shallower) than
the mean slope calculated from the entire dataset. The slope stays within 5% of the
mean slope more than 90% of the time. . . . . . . . . . . . . . . . . . . . . . . . 47
2.24 Lagrangian back trajectory analysis for the coastal and inland sites. The latitudinal
gradient depicted for the coastal sites serves as a first order predictor for the isotopic
composition while the more varied trajectory at the inland sites does not serve the
same function. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
3.1 Scan of a Bristlecone Pine tree core used for isotopic analysis. The dark bands are
called ”late wood” and form at the end of the growing season because of increased
cell density. The growth orientation is up. The colored lines are used to show the
sampling strategy used in this chapter. Each ring is approximately 1mm. . . . . . 55
3.2 Photo of the rotary microtome at USC used for slicing samples. The screen on the
left of the photo is being fed from the microscope, which is focussed on the mounted
core. Individual wood cells can be seen on the screen as light circles. In this image
the growth direction is down and to the right. . . . . . . . . . . . . . . . . . . . . 56
3.3 A highly schematicized representation of the leaf and soil systems where movement
of water is denoted by arrows and the process labeled. Steady-state theoretical iso-topic
profiles (enriched to the left) that arise from the phase change and diffusion
processes are included. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
3.4 Parameters input into the geochemical model described above to model the intra-annual
cellulosic cycle. Uncertainty envelope is a 1s window generated from daily
instrumental climate data from the nearby Crooked Creek meteorological station. . 69
3.5 Raw isotopic measurements for each of the three wood sections. The non-growing
season hiatus were removed from the age model to reduce long empty spaces between
years. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
3.6 All measurements included in Figure 3.5, with each datum having been normalized
to a common age model and corrected as an anomaly relative to the mean value for
the entire year. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
3.7 A third order polynomial fit to each of the time periods (left panel). Uncertainty
envelope is one s. Box plots of the annual isotopic standard deviation from each of
the time windows (center panel). Same as in the left panel except each cycle was
corrected to a common variance to compare simply the shape of the cycles (right
panel). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
x
3.8 Modeled isotopic composition at the White Mountain site using the parameters
shown in Figure 3.4. Three simulations with a normal, shortened and elongated
growing season were conducted and 100 random iterations from the three simula-tions
are used to illustrate the results. . . . . . . . . . . . . . . . . . . . . . . . . 73
3.9 Correlation matrix between modeled and measured isotopic cycles. . . . . . . . . 73
3.10 Intra-annual isotopic measurements from the Almagre Mountain site. A spline fit
has been included. Note that because it was not possible to sample all rings at this
resolution some years were excluded and thus the x-axis is not evenly-spaced. . . 74
3.11 The mean isotopic value was removed to generate a composite intra-annual cycle
(left), with a 1s envelope shown in gray. A modeled cycle based on the average
climate conditions during this time interval is shown in red alongside the measured
cycle (grey). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
4.1 Complete time series from the Almagre Mountain site generated from separate
tree cores (blue and green). A low-pass filtered (18 years) timeseries of the mean
between cores is shown in black. The pink region denotes the instrumental period
used to calibrate the response between d18Oc and temperature. . . . . . . . . . . . 89
4.2 A map showing the correlation coefficient (r) between d18Oc and surface temper-atures
from the CRUTEM3 temperature dataset between 1900-2001. The purple
line roughly encompasses the area where the correlation between d18Oc and surface
temperature is significant (p=0.01). The Almagre site is labelled A, B is the frost
ring chronology (Brunstein, 1996), C is Pawnee National Grasslands Global Net-work
of Isotopes in Precipitation site, D is the tree ring temperature reconstruction
from (Salzer & Kipfmueller, 2005), E is the tree ring temperature reconstruction
from (Biondi et al., 1999), F is the isotope-based temperature reconstruction from
(Edwards et al., 2008), and the dotted line encompassing G is the composite tem-perature
reconstruction based on wood density from (Briffa et al., 1992). . . . . . 90
4.3 Correlation coefficients between monthly average temperature and d18Oc from the
CRUTEM3 temperature dataset at the 39oN 105oW grid point. The final column,
shows the correlation coefficient against mean annual temperatures. The p-values
for each of the correlation coefficients are shown in the bottom panel. . . . . . . . 91
4.4 Timeseries showing instrumental temperatures from the nearby Canon City meteo-rological
station (gray) against ring widths (left) and d18Oc (center) all of which are
presented as an anomaly relative to the 20th century mean. The three timeseries are
shown together in the right column. The ring width chronology only extends until
1983, which is why it is truncated relative to the other records. . . . . . . . . . . . 91
xi
4.5 Relationship between surface temperature and the isotopic composition of individ-ual
storm events from the Pawnee Grasslands GNIP station (left). The slope of
the relationship between temperature and the isotopic composition of annual cel-lulose
(green), intra-annual cellulose (red) and precipitation (purple) (center). The
envelope shows the 90% confidence band around the slope. Intra-annual isotopic
measurements (purple) with a 1s error envelope, modeled isotopic composition of
cellulose from Chapter 3 (red) and seasonal temperature cycle (green). . . . . . . 92
4.6 Lomb-scargle periodogram of the d18Oc timeseries based on the methods described
in (Schulz & Mudelsee, 2002) using a Welch window and 3 overlapping segments.
The confidence interval was generated using a 2000 iteration Monte Carlo simula-tion.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
4.7 Wavelet analysis of the d18Oc timeseries using a Morlet window based off the code
from (Torrence & Compo, 1998). The black line shows regions where the power is
significant. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
4.8 Time series of temperature reconstructions from the co-located width (purple) and
isotope-based (green) techniques. Both records are shown as anomalies relative to
the 1960-1990 average temperatures. The gray region highlights the discrepancy
between the two records. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
4.9 Time series of low-pass filtered temperature reconstructions from the co-located
width (purple) and isotope-based (green) techniques alongside the composite global
temperature reconstruction from (Mann et al., 2008) with the uncertainty in the latter
shown as gray shading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
4.10 Frost ring frequency record from (Brunstein, 1996). In this plot an increase in frost
ring frequency is caused by cooler conditions and thus the axis is reversed. The
number of frost rings is reported as an anomaly relative to the average number of
frost rings per decade during the 20th century. . . . . . . . . . . . . . . . . . . . . 96
4.11 A compilation of temperature records, which include tree ring widths (left column)
and do not include tree-ring widths (right column). All records are specific to the
western US except the northern Hemisphere composite record from (Mann et al.,
2008). Each record is reported using the same temperature scale as the original
authors except (Edwards et al., 2008), where I show the uncorrected isotopic time-series.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
4.12 Evolutive response function between weekly temperature and ring widths. The ordi-nate
is time during the growing season with the markers noting 30 day intervals
starting at the bottom with May 1 and ending at the top with September 1. Cor-relation
coefficients were calculated for sliding ten year windows during the 20th
century using daily temperature data from the Canon City meteorological station.
Only positive response functions are shown. . . . . . . . . . . . . . . . . . . . . 99
xii
4.13 The residual between the isotope and width-based (green) and density and width-based
(blue) temperature estimates. The gray line is the difference between MAT
and June temperature anomalies relative to the 1960-1990 temperatures from the
ECHO-G climate simulation, which has been smoothed using a Stineman function. 100
5.1 Photograph of Bristlecone Pine slab highlighting the quasi-exponential decay in ring
size as a function of increasing tree radius. . . . . . . . . . . . . . . . . . . . . . 111
5.2 Timeseries of multiple BcP chronologies spanning modern (right) to AD 1500 (left).
Certain sections were lost either due to instrumentation era of insufficient material. 113
5.3 Correlation coefficients between 3-year running mean of zonal geostrophic winds
derived from the Trenberth SLP dataset and the BcP record (top) and correlation
against the Kaplan SST dataset (bottom). . . . . . . . . . . . . . . . . . . . . . . 114
5.4 Cellulosic d18O from the White Mountain site (A) compared to indices of annually-averaged
Atmospheric (North Pacific Index, B) and Oceanic (Pacific Decadal Oscil-lation,
C) indices of climate variability. Cellulose (red) and PDO (blue) records have
been smoothed with a 3-year moving average filter. The NPI record is shown on a
reverse scale where up denotes an equatorward shift in storm tracks driven by SLP
anomalies in the Aleutian Low region. . . . . . . . . . . . . . . . . . . . . . . . . 115
5.5 Cellulosic d18O from the Alta Peak site for three separate trees. . . . . . . . . . . 116
5.6 Timeseries of measured BcP and Alta (red and green) and modeled (gray) annual
isotopic values. The modeled values were calculated based on the equations and
methodologies discussed in Chapter 2 except done using integrated growing season
values as opposed to daily or weekly time steps. . . . . . . . . . . . . . . . . . . 117
5.7 Cellulosic d18O for the White Mountains (green) and rainfall anomalies averaged
from California Climate Division 4, 5 and 7 during the 20th century (blue). Pre-vious
multi-year droughts are denoted by gray bands while the current drought is
delineated by the pink band. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
5.8 Comparison of SST anomalies from the Kaplan dataset during the 1950’s drought
and the latter stages of the current drought. . . . . . . . . . . . . . . . . . . . . . 121
5.9 PDSI from (Dai et al., 2004) during the two major droughts of the early and mid
20th century compared to the current PDSI anomalies. . . . . . . . . . . . . . . . 122
5.10 The isotopic composition of water vapor over grid points along the North Pacific
Basin from the IsoGSM simulation. The black line is used to show the areal average. 122
5.11 Schematic representation of how the water budget varies from an isotopic perspec-tive
during previous droughts, the current drought and theoretically the situation if
a La Ni ˜ na-like drought were to occur amidst the current state. . . . . . . . . . . . 123
xiii
5.12 Comparison of isotopic trends from sites in the Pacific Northwest (Mt Logan and
Eclipse (Fisher et al., 2004), Jellybean Lake (Anderson et al., 2005), Mica Lake
(Schiff et al., 2009), Lakes (Hu et al., 2001)) and the southwestern US. All Pacific
Northwest sites were normalized and timeseries were interpolated to a common
decadal timescale. Colored dots on the map correspond to the locations of the sites
used in the composite timeseries. . . . . . . . . . . . . . . . . . . . . . . . . . . 126
5.13 Correlation (r) between zonal geostrophic winds and time for the Trenberth and
HadSLP datasets. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
5.14 Cross wavelet coherence using a morlet wave between PDSI reconstruction from
(Cook et al., 2007) and the BcP annual chronology. Periods of significant coherence
(95%) are denoted with the bold black line. The arrows indicate the phase angle of
the relationship with arrows pointing right denoting in-phase coherence, and to the
left indicating anti-phase. Power is reported in normalized units and the veil, shows
the cone of influence, under which power needs to be treated with caution. Analysis
is done using the algorith from (Grinsted et al., 2004). . . . . . . . . . . . . . . . 129
5.15 Timeseries of other circulation proxies from the Pacific Basin. Top is the BcP
chronology from this study, center is the upwelling/wind proxy from the Santa Mon-ica
Basin from (Holsten et al., 2004) and bottom is the ice core d18O chronology
from Mount Logan (Fisher et al., 2004). All records have been smoothed (colored
line) using a running mean filter. . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
6.1 Time series of NGRIP (Svensson et al., 2008), Cave of the Bells (Wagner et al.,
2010), and Fort Stanton (Asmerom et al., 2010). All shown on independent timescales
described in the original publications . . . . . . . . . . . . . . . . . . . . . . . . 141
6.2 Slope between amount-weighted monthly modeled isotopic composition of precipi-tation
and precipitation amount (normalized) at the grid point nearest to the Cave of
the Bells Site for winter months between 1979-2008 (top). Slope between amount-weighted
monthly modeled isotopic composition of precipitation and monthly tem-perature
at the grid point nearest to the Fort Stanton site (bottom). . . . . . . . . . 144
6.3 Shape of the 9 DO events at each site, where each event was centered at its peak and
the mean cycle calculated (color). The well-defined sawtooth in the NGRIP record
erodes quite substantially in the speleothem records. . . . . . . . . . . . . . . . . 146
6.4 The ”average” interstadial event generated from the average of 9 events from the
NGRIP record (blue) alongside the peaks of each of the events at the two sites (red
is CoB and green is Stanton). The error bars are the uncertainty of the timing of
the peak based on the U/Th ages. The height of each of the peaks is based on the
magnitude of the event. This is to show that the relative lead or lag is not a function
of the size of the event. The CoB peaks tend to scatter just before the NGRIP events
while the Stanton events tend to lag the NGRIP events. . . . . . . . . . . . . . . . 147
xiv
6.5 Lomb-scargle periodogram for the three time series using the original age models
with no linear interpolation performed. Analysis was done using a Welch window
and the dotted line denotes the 99% confidence interval based on a 5000 iteration
Monte Carlo simulation against a red noise background (Schulz & Mudelsee, 2002).
The commonly cited 1500 year cycle is the NGRIP record is marked on each of the
records. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
6.6 Coherence and phase angle between the CoB and NGRIP records using the multi-taper
coherence method. The gray bar denotes to 1500 year cycle, which is coherent
in both records and the phase angle suggests the two are nearly in phase. . . . . . 150
6.7 A cross wavelet coherency analysis between the NGRIP and CoB (left) and NGRIP
and Stanton (right). Significance coherency is shown in red tones, with significant
regions denoted by a heavy black line. The arrows are used to show the phasing
with arrows to the right indicating in phase behavior. . . . . . . . . . . . . . . . . 151
6.8 Linear fit between 9 DO events in NGRIP and CoB (red) and Fort Stanton (purple)
caves. The fit is improves if DO event 9 is removed from CoB record. Both fits are
high significant and the slopes are statistically different. . . . . . . . . . . . . . . 152
xv
Abstract
Late Holocene paleoclimatology of the southwestern United States has been reconstructed largely
through the analysis of ring width variability from a network of gridded tree chronologies. Trees
commonly respond to moisture stress in this semi-arid environment providing a spatially coherent
annually-resolved record of PDSI variability. At a handful of high altitude sites, trees are thermally
stressed, providing a record of temperature variability. This thesis addresses two prominent ques-tions
that arise from the tree ring network; 1) The precipitation history highlights that sustained
hydroclimate epochs are an ubiquitous aspect of the regions climate history though it is not fea-sible
with the existing tree ring network alone to partition whether drought (or pluvial) regimes
arise through a singular forcing mechanism or can be delineated into dynamical types. 2) The
response between tree growth and temperature variability is poorly understood and because these
records are some of the few available mid latitude temperature records, an independent assessment
of their response to temperature is required. To address these questions, isotopic chronologies from
Bristlecone and Foxtail Pine trees in the White Mountains, Almagre Mountains and Sierra Nevada
are presented. Because the isotopic composition of cellulose responds not distinctly to moisture
availability but rather to the isotopic composition of precipitation, the isotope records provide infor-mation
on atmospheric circulation and temperature that is independent from the co-located drought
and temperature reconstructions.
To directly test the relationship between atmospheric circulation and the isotopic composition of
precipitation in the southwestern US, I develop a catalog of 120 individual storm events striking
the west coast over a 5-year period. The cause of isotopic variability is assessed using an isotope-enabled
GCM simulation that has been nudged to Reanalysis fields. The results from this analysis
show that changes in meridional moisture flux from the low latitudes leave a tangible mark on pre-cipitation
in the region. This relationship can theoretically be quantified by a linear relationship
xvi
between the modeled isotopic composition of precipitation and the relative percentage of low lati-tude
tagged water that is delivered with the storm system. The controls on the isotopic variability of
precipitation change substantially moving eastward into the North American Monsoon region where
moisture is not delivered by large frontal storms but rather through localized convection. In these
regions, variability of the moisture source is subdued and isotopic variability arises principally as a
function of depth of convection, which leads to a close correlation between temperature and d18O.
Mechanistic constraints are placed on the cause of isotopic variability in the cellulose chronologies
using a forward modeling approach where meteorological data and the isotopic composition of soil
water and vapor from an ensemble of isotope-enabled GCM simulations are fed into a geochemical
model that captures the isotopic fractionation associated with the biogeochemical processes in the
tree prior to cellulose metabolism. At sites where precipitation is predominantly from winter pre-cipitation
the intra-ring isotopic cycles are driven largely by the relative humidity and temperature
at the leaf boundary while the higher amplitude interannual variability arises from changes in the
trees source water. Variations in the shape of the cycle reflect not only differences in growing season
climate but also changes in the length of the growing season. At sites where moisture is predomi-nantly
from summer rains, the cycle directly tracks the isotopic composition of precipitation during
the growing season.
The isotope chronology from the Almagre Mountains in Colorado shares little covariance with the
more western sites, because it relies predominately on monsoonal moisture whose isotopic compo-sition
tracks temperature. The record from this sites provides a 500-year reconstruction of growing
season temperatures for this region. Growing season temperatures in the southern Rocky Moun-tains
display a high degree of multi-decadal variability between the 18th-20th century but only sub-dued
variability prior (15th-18th centuries). The temperature reconstruction presented here differs
markedly from the tree-ring width based temperature reconstruction from the same site but agrees
with regional temperature reconstructions based on instrumental temperature records and tree ring
density. Because the relationship between temperature and the isotopic composition of precipitation
xvii
is stable, I interpret the difference between the two reconstructions to be evidence that the ring-width
based reconstruction is biased by non-stationary relationships between temperature and growth.
Isotopic chronologies from Bristlecone Pines in the White Mountains show distinctive minima dur-ing
each of the multi-year droughts of the 20th century, suggesting that drought is sustained by a
selective loss of low latitude moisture. This is dynamically consistent with an understanding that
each of the major 20th century droughts was driven by cooler conditions in the eastern Tropical
Pacific and a northward shift in the storm track. The exception to the isotope-aridity relationship
is during the current drought, which has been associated with rising isotopic values and therefore
appears dynamically distinct. Rising temperatures in the mid troposphere over the North Pacific
during the current decade has diminished the pressure gradient between the subtropics and extrat-ropics
and has therefore reduced the convergence of the depleted northerly moisture source. Prior to
the 20th century there is a dramatic change in the isotopic composition of precipitation at the White
Mountain site. The cause of this is hypothesized to be an increase in low latitude moisture associ-ated
with cooler Northern Hemispheric conditions, which led to a more southerly storm track that
consistently tapped into the enriched tropical moisture pool. Dramatic isotopic shifts in ice cores
and lake sediments from the North Pacific confirm a major change in the midlatitude storm track at
the terminus of the Little Ice Age.
A brief consideration at the end of this thesis is given to recently published isotopic chronologies
from stalagmites in the southwestern US that span the last Glacial Period. While these records show
high amplitude millennial variability in this region during the last Glacial Period, the enigmatic
discrepancies between two nearby records, are taken as evidence of overwhelming kinetic effects
that likely obscure the exact nature of the climatic changes that occurred in the southwestern US
during Greenland Interstadials.
xviii
Chapter 1
Introduction
1.1 Climate forecasts for southwestern US
In 1994, Scott Stine published the paper, Extreme and Persistent drought in California during
mediaeval time, where he documents the presence of tree stumps rooted in situ deep below the
waters of Mono, Walker and Tenaya Lakes in Eastern California and Nevada (Stine, 1994). The
presence of these buried stumps indicates that lake levels had dropped sufficiently low during
certain periods of recent history, to expose these depths of the lake bottom to the atmosphere. The
alarming aspect of this finding was that some of these stumps had over 200 annual growth rings,
implying that severe drought had gripped the region for multiple centuries. This study serves
as a haunting depiction that the semi-arid southwestern US is vulnerable to stark hydroclimatic
extremes and leads to an obvious concern regarding the impacts that such an event would have on
today’s municipal and agricultural systems in the western US. The ambitious and complex water
distribution infrastructure in California relies heavily on the same snow melt that feeds the lakes of
Stine’s megadroughts, in fact it is specifically the catchment of Mono Lake thatWilliam Mulholland
targeted in the development of the Los Angeles aqueduct. While, there is no adequate recent analog
to consider the impact of a comparable event today, the noted synchrony between megadroughts
and collapse of Puebloan civilizations serves as some indication of the cultural sensitivity in this
region to water availability (MacDonald & Tingstad, 2007; Cook et al., 2007; Benson, 2009).
It is likely that the southwestern US will undergo a long term drying during the upcoming
century (Seager et al., 2007b). Changes in the radiative balance and the associated warming of the
troposphere will lead to a rise in global atmospheric humidity and a poleward shift in the latitude
of the storm track, which acting in concert will bring about the projected drying (Held & Soden,
1
2006; Yin, 2005; Salathe, 2006). The impact of the drying on agricultural and municipal water
resources of the region will be exacerbated by localized feedbacks associated with the warming
trend, including a shift in the proportion of rain and snow and an earlier onset of snowmelt each
year (Stewart et al., 2005; Cayan et al., 2001). These predictions will be slowly borne out over the
coming century amidst additional higher frequency and amplitude variability that will periodically
enhance or alleviate the trend towards a more arid southwestern US.
Transient droughts in the southwestern US are part of a symmetrical hemispheric response
to cooling in the eastern Tropical Pacific, which leads to wave-driven shifts in the locations of the
quasi-permanent high pressure cells. These changes influence the trajectory of the jet stream and
consequently the prevailing pathway of storms. The cool anomalies in the eastern tropical Pacific
can persist for multiple years and have been called upon as the mechanism that triggered each of
the major droughts of the instrumental record (Seager, 2007; Seager et al., 2003; Burgman et al.,
2010; Herweijer et al., 2006). Additional drought forcing is also likely rooted in SST variability in
the North Atlantic (McCabe et al., 2004) though the mechanism that brings about this empirical
relationship is not currently understood. Unlike the long term drying, multi-annual and decadal
hydroclimatic and temperature variability in the western US is not consistently reproduced between
21st century forecasts and it is thus unclear how variability on these timescales will aggravate or
alleviate the predicted long term drying trend. Therefore, further characterization of the natural
modes of variability that bring about the hydroclimatic anomalies, which can be accomplished
through coupling proxy reconstructions and hindcast modeling are beneficial in elucidating the
relevant dynamics (Herweijer et al., 2006; Seager et al., 2007a; Graham & Hughes, 2007).
2
1.2 Tree-ring drought reconstructions
The southwestern US is coincidentally endowed with a widespread occurrence of some of the
world’s oldest living trees (Schulman, 1958; Brunstein & Yamaguchi, 1992; LaMarche & Mooney,
1967; Lamarche, 1978). Because tree growth in semi-arid regions is often limited by the availability
of water, year to year variations in growth enable a depiction of drought over the entire region for
nearly 1,000 years and in some instances such as the White Mountains, many thousands of years
(Fritts, 1969). The gridded network of tree ring chronologies from the western US (North American
Drought Atlas), depicts high amplitude year to year precipitation variability characteristic of both
stochastic processes and the slightly more predictable influence that ENSO events have on the
region (Cook et al., 2007). The gridded drought record also depicts lower frequency decadal and
arguably centennial epochs where the region is characteristically dry or wet (Cook et al., 1999). The
tree rings for example not only depict the major droughts recorded in the Mono Lake tree stumps
(Stine, 1994) but also a general recurrence of numerous severe multidecadal wet and dry intervals.
It is an ongoing effort to use the tree-ring hydroclimate database to understand the mechanisms
responsible for bringing about severe drought. One approach which has been adopted is to use the
spatial pattern of droughts to elucidate the mechanisms responsible. Woodhouse et al. (2009) uses
Empirical Orthogonal Functions of reconstructed drought patterns to clearly distinguish distinctive
drought modes. Many of the large droughts display a classic dipole pattern with an anomalously
wet Pacific Northwest and a dry southwestern US, which is similar to the pattern that emerges
during annual ENSO events (Cook et al., 2007; Cayan et al., 1998). The spatial footprint of a
drought will also indicate drought epicenters, which is useful in distinguishing where the peak
impact should be found. Recently, Cook et al. (2009) distinguishes a unique spatial characteristic
of the Dust Bowl drought, which was linked to feedbacks associated with dust over the continental
US and its impacts on radiative forcing. Spatial comparisons, thus can elucidate subtle dynamics
that distinguish one event from the next.
3
An additional approach to understanding drought mechanisms is through targeted proxy
analysis of teleconnection hotspots. For example, by generating Sea Surface Temperature (SST)
reconstructions in the eastern Tropical Pacific or North Atlantic, it is is possible to fingerprint
if drought was associated with an anomaly in one of these locations. A nice discussion on this
approach is found in Conroy et al. (2009), who distinguish large western North American droughts
that were associated with both Atlantic and Pacific SST anomaly patterns. The results from this
analysis are consistent with the spatial analysis of droughts, in that there appear to be multiple
mechanisms that bring about drought and delineating a single origin is not likely feasible. The
challenge with the teleconnection approach is that it requires SST proxies with nearly absolute
dates. Consider that severe drought may last for 2-3 decades and thus fingerprinting an SST
anomaly that brought about this drying would require a sediment record of near equal resolution
and age control. While a few records of this type are available (Cobb et al., 2003), such records are
decidedly rare. Furthermore, the co-occurrence of an SST anomaly with a drought event, does not
necessarily imply causality.
An additional approach is through hindcast modeling, where boundary conditions are
constrained through proxies and it is tested whether the atmospheric dynamics resulting from
these initialized conditions are sufficient to trigger the spatial extent and severity of the proxy
reconstructed drought (Graham & Hughes, 2007). This is an emerging approach, whose challenge
is the appropriate delineation of paleo-boundary conditions, which are often only constrained
through a few available records. Feng et al. (2008) for example, show that in order to replicate
the medieval megadroughts, consideration must be given to both the Atlantic and Pacific SST
patterns, while Seager et al. (2007a); Graham & Hughes (2007) find it is possible to replicate
drought intensity only through defining conditions in the tropical Pacific. The choice of where to
develop new proxy records for developing boundary conditions in hindcast models is guided by
4
instrumental observations of locations with important teleconnections, but still it is possible that
influential regions may have sparse or non-existent proxy data.
1.3 Tree-ring temperature reconstructions
While much of the focus on tree ring width in the region has been on reconstructing precipitation
patterns, tree-ring width variability has also been used to depict temperature variability in the
southwestern US (Scuderi, 1993; Graumlich, 1993). Drought reconstructions are more ubiquitous
because trees are principally water-stressed in semi-arid regions, though at some rare sites, tree
growth has been shown to be more sensitive to temperature variability (LaMarche & Stockton,
1974). Temperature-sensitive sites appear to emerge only under rare environmental stress that
occur along the upper treeline where minimum tenmeratures remain near to the minimum threshold
required for xylogenesis (LaMarche & Stockton, 1974; Rossi et al., 2008). Salzer et al. (2009),
show very clearly how stands of trees in the White Mountains of California, which may only be
separated by a few 100 meters have entirely unique responses to temperature. These sorts of rather
fortuitous and highly localized relationships between temperature and growth have led to varying
degrees of skepticism regarding the use of this proxy to reconstruct temperature. Details of this
debate will be provided in Chapter 4. Despite the inherent uncertainties that may arise from using
tree rings to reconstruct temperature, they are one of the few available temperature proxies, outside
of the polar regions and thus are an important source of information on understanding temperature
trends prior to the instrumental period.
5
1.4 Climatic information from hydro-isotopes
The accumulation and flux of the isotopologues of water throughout the ocean-atmosphere system
provides a useful tracer for key processes in the hydrological cycle and have been used for decades
to understand meteorological or climatological phenomenon (Craig & Gordon, 1965; Dansgaard,
1964; Rozanski et al., 1993; Gonfiantini et al., 2001; Lawrence et al., 1982). If we trace an air parcel
from its origin at the marine surface, the vapor will have an isotopic signature characteristic of the
temperature and humidity that prevailed when the moisture evaporated following the Craig-Gordon
Model (Craig & Gordon, 1965; Merlivat & Jouzel, 1979) As this air parcel moves through the
atmosphere, the isotopic composition of its vapor evolves as a function of condensation, which
progressively depletes the air mass in heavy isotopes and through entrainment of newly evaporated
water or mixing with other air masses. The isotopic composition of an air mass at a given moment
in time therefore bears information on numerous intreacting processes that have affected this air
mass. From a global perspective, evaporation of water into the atmosphere occurs principally in
the tropics. As air moves poleward as part of the global overturning circulation, the air becomes
progressively depleted leading to a latitudinal gradient in the isotopic composition of vapor,
that largely tracks the meridional temperature gradient. This zonal isotopic gradient is the most
prominent spatial feature of the global distribution of water isotopes and provides an important
depiction of the global hydrological cycle.
The spatial patterns of isotopic variability are useful in highlighting global atmospheric pro-cesses
but in order to use isotopic variability to reconstruct past oceanic or atmospheric processes,
it is critical to understand what influences the isotopic composition at one location in the time
domain. In Figure 1.1, I show the results of a point-by-point correlation (r) between the isotopic
composition of precipitation and temperature (right) and precipitation amount (left) during a thirty-year
simulation of a global climate model that includes isotopic tracers (Yoshimura et al., 2008).
In the left hand panel, we observe that throughout much of the low latitudes, there is a negative
relationship between the isotopic composition of precipitation and precipitation amount. This is
referred to generally as the ”amount effect” and implies that over time, changes in the isotopic
6
Figure 1.1: Gridded correlation coefficient (r) between d18O of precipitation and surface temperature
and precipitation amount. Data used are from (Yoshimura et al., 2008).
composition of precipitation at a given grid point will reflect changes in how much precipitation has
fallen there (Lee et al., 2008a). On the opposing panel, I show the correlation between the isotopic
composition of precipitation and surface temperature. Here we see that the positive relationships
between surface temperature and the isotopic composition of precipitation are restricted to the high
latitudes (Rozanski et al., 1993). Therefore time series of the isotopic composition of precipitation
would contain information on temperature variability at this given location.
The ”amount effect” and ”temperature effect” have proven to be incredibly useful relationships
in understanding past climatic variability. Ice core records, from Greenland and Antarctica depict
systematic shifts in Earth’s temperature over the recent 800 thousand years while corresponding
isotopic records from low latitude sites derived, from Loess sequences and speleothems, show cor-responding
changes in hydrology that accompanied these temperature changes (Wang et al., 1999;
Jouzel et al., 1987). What is also obvious from Figure 1.1, is that the southwestern US falls between
the regions where the isotopic composition of precipitation characteristically responds to either
precipitation amount or surface temperatures. Indeed, the lack of a systematic relationship between
the isotopic composition of precipitation and surface conditions throughout the midlatitudes has led
to a lack of quantitative paleoclimatic information from these regions.
7
Therefore one of the first challenges in this thesis is filling in the ”correlation gap” in the mid-latitudes
that would enable meaningful climatic information to be generated from isotopic recon-structions
from these regions. The analysis relies heavily on a series of fundamental changes which
are currently underway in the field of isotope hydrology. Firstly, there is an increasing effort to
include isotope tracers in global climate simulations (Noone & Simmonds, 2002; Hoffmann et al.,
1998; Lee et al., 2007). This opens up opportunities to explore the causes of isotopic variability that
are too complex to explore with linear regression models. Secondly, there has been a fundamental a
change in the instrumentation of isotope hydrology, which has led to increasingly routine measure-ments.
Particularly, the development of commercial cavity ring down spectroscopy has opened up
the door to make increasingly large numbers of measurements of both water and vapor samples at
relatively low cost, which means the number of measurements available to understand the climatic
controls on mid latitude precipitation are growing exponentially.
1.5 Thesis Outline
In this thesis, I will explore two primary questions. Firstly, I will attempt to shed light on the mech-anisms
that drive drought in this region, which is motivated by a hope to improve hydroclimatic
forecasts for the region. Secondly, I will provide a test of the relationship between surface tempera-ture
and tree growth at sites where growth is believed to be ”temperature-sensitive”. This is relevant
both from an improved understanding regional temperature variability as it pertains to for example
the timing of snow melt, but also because of the significance there records play in global temper-ature
reconstructions (Mann et al., 1998, 1999, 2008, 2009). In order to address these questions,
I revisit some of the ancient Bristlecone Pine tree ring chronologies of California and Colorado
(LaMarche & Stockton, 1974) but as opposed to measurements of ring width, I use variations in the
isotopic composition of cellulose from these trees. This complements the existing climatic informa-tion
derived from growth variability but is inherently independent. The thesis is a compilation of
mostly published work or work that has been submitted for publication and so at times the chapters
8
Figure 1.2: Photo of White Mountain Bristlecone Pine stands during the June 2008. This photo serves to
provide some visual reference to the reader on the environment at which the trees discussed in Chapters
3 and 5 grow.
will come across as independent entities. However, the common theme is the utility of isotope prox-ies
in improving our understanding of the mechanisms that drive climatic variability in this region
and perhaps across the mid latitudes. Chapters 2 and 3 are calibration studies where I visit the
mechanistic controls on the isotopic composition of precipitation in the western US and the isotope
systematics of tree ring cellulose in Bristlecone Pine trees. Chapters 4 and 5 are case studies where
I present isotopic time series which shed light on temperature and drought variability respectively.
In Chapter 6, I explore isotopic records from speleothems in the southwestern US to refine climatic
information that can be garnered from these proxies by utilizing the information presented in Chap-ters
2-5. The work presented here in many ways is intended to motivate extending the network of
isotopic records presented further back in time and include other sites to improve the spatial density.
So as opposed to being a ”closed book”, each chapter intends to plant seeds for continued studies.
The long term goal of this project is the development of a true network akin to the Isonet project
currently underway in Europe (Treydte et al., 2007).
9
Chapter 2
Atmospheric circulation and the isotopic
composition of precipitation over the western US
SUMMARY
In this chapter I present a description of the controls on the isotopic composition of precip-itation
along the west coast of the United States. Previous efforts to delineate the dominant
climatological influences on the isotopic composition of precipitation for this region have been
hampered by both a lack of empirical observations and a model that can be used to parse the
complexities associated with moisture transport in extratropical cyclones. To address the for-mer
issue, I have developed a network of sites spanning a latitudinal transect of the west coast
of the US. I use this network to validate the robustness of an isotope-enabled General Cir-culation
Model with a spectral nudging routine, which allows for direct comparison between
individual storm events and and their modeled equivalents. The model shows that efficient
poleward transport of tropical moisture within certain frontal storm systems is responsible
for driving precipitation towards extremely enriched isotopic values. The mass of low lati-tude
moisture that precipitates over the western US is a cumulative function of both seasonal
and interannual changes in the rigor of meridional circulation and the dynamics of individual
synoptic storm systems. A theoretical test of this hypothesis is conducted through a numer-ical
painted water exercise where water evaporated from the tropical Pacific is released into
extratropical cyclones and its evolution traced. An additional test is presented where a third
isotopic variable called deuterium excess, which describes the relative proportion of d18O to
dD in a water mass, is used as a proxy for where moisture in the storm initially evaporated
from. The deuterium excess analysis is used to delineate two families of storms, one group
which is dominated by moisture convergence along the storm’s trajectory and another group
10
which behaves with river-like characteristics, which is to say that it conserves the isotopic
composition of its source. It is this latter family of storms that are principally responsible for
the convergence of tropical moisture to the western US and consequently enriched isotopic
values. The complexity of moisture source controls on the isotopic composition of precipi-tation
at the coastal sites is contrasted against isotopic measurements from central Colorado,
which are overwhelmingly dominated by conditions that prevailed during condensation and
can be described with a simple linear relationship between surface temperature and d18O.
2.1 Introductory note
The contents of this chapter have been published in:
Berkelhammer, M., Stott, L., Yoshimura, K., Johnson, K., and Sinha, A. (submitted to Cli-mate
Dynamics). Synoptic and mesoscale controls on the isotopic composition of precipitation in
the southwestern US.
2.2 Introduction
The isotopic composition of precipitation at a given site reflects a summation of remote and local
processes that can affect contributions of moisture from different source locations, rainout along
the storm trajectory and conditions that prevail during condensation (Dansgaard, 1964). Isotopic
records thus capture an integrated signal of synoptic and mesoscale atmospheric processes. In
tropical and high latitude settings, the complex multivariate signal can often be reduced to a simple
univariate linear regression model leading to climate reconstructions from archived precipitation
that largely reflect a single climate variable. In the polar regions for example, the isotopic
composition of precipitation tracks latitudinal variations in the moisture source, which varies
with hemispheric temperatures and consequently allows for reconstructions of broadly regional
or even global temperatures from high latitude ice cores (Noone, 2008; Jouzel et al., 1997). The
11
robustness of high latitude ice core temperature reconstructions is not simply a product of a direct
physical relationship between local conditions and the isotopic composition of precipitation but
rather arises because the atmospheric overturning circulation, which drives the isotopic variability,
is tightly linked with hemispheric temperatures (Hendricks et al., 2000; Kavanaugh & Cuffey, 2003).
In mid latitude and subtropical locations, isotopes also track large-scale atmospheric circula-tion
patterns. However, circulation in the subtropics has a characteristically more complex
relationship with local and regional surface climate and thus, direct univariate regression models
between the local climate and isotope ratios are typically not robust (Alley & Cuffey, 2001; Fricke
& O’Neil, 1999). The difficulties in calibrating isotopic records from mid latitude and subtropical
locations has resulted in a lack of quantitative paleoclimate information from these regions that
would be a valuable asset in studies that attempt to link low and high latitude ocean and atmospheric
dynamics. Therefore, partitioning the controls on isotopic variability would provide opportunities
to develop new records and reinterpret existing ones.
In a seminal study of the isotopic composition of precipitation from the southwestern United
States, Friedman et al. (1992) suggested that storm trajectories were likely the leading cause
of isotopic variability in precipitation. Their conclusions were based largely on conjecture (by
their own admission) because their analysis utilized 6-month integrated precipitation samples and
thus lacked sufficient resolution to properly explore direct relationships between individual storm
trajectories and the isotopic composition of precipitation. Benson & Klieforth (1989) working in
the Yucca Mountain region of Nevada and Friedman et al. (2002) working in Utah and Nevada
both presented event-scale isotopic values to address the shortcomings of seasonal and monthly
sampling. Consistent between these studies was the finding that there is a range of storm to
storm variations that is on the order of 20‰ for d18O and 180‰ for dD. The wide range of
isotopic variability was attributed to two primary factors; trajectory, and the depth of atmospheric
convection. Friedman et al. (2002) addressed the latter effect (depth of convection), showing that
storms associated with high vertical wind shear generated isotopically more-depleted rainfall.
12
Their data loosely fit a Rayleigh curve where the heavier isotopologues of water are preferentially
distilled from the vapor phase as the air cools adiabatically upon ascent to higher altitudes. Recently,
(Coplen et al., 2008) provided a more rigorous test of this idea by making isotopic measurements
every 15 minutes during a single storm that struck the coast of California. They argued that
during air mass ascent isotopic values dropped and subsequently rose as the air mass descended to
warmer atmospheric levels. This is consistent with the closed-system pseudo-adiabatic Rayleigh
distillation model of (Gonfiantini et al., 2001) where the isotopic composition of the precipitate is
a function of condensation temperature (equivalent in this case with height) and the extent of rainout.
The findings of Coplen et al. (2008), suggest that each storm is a discrete closed system that
has been initialized with a different integrated water vapor (IWV) content. In this chapter, I consider
whether or not the most important influence on isotopic variability between storms is attributable
to differences in the isotopic composition of the moisture source incorporated into the large frontal
storms that deliver most of the annual precipitation budget to the western US. There have been
other event-scale studies for the western US including those by Benson & Klieforth (1989) and
Friedman et al. (2002), which have considered this question but their studies were conducted in
the Great Basin, which Ingraham & Taylor (1991) have argued represents a quasi-closed isotopic
system. This would imply that changes in the isotopic composition of water vapor over the Pacific
that arise from variations in sea surface temperatures or from atmospheric circulation changes
have only secondary influence on the isotope hydrology within the basin because of large dilution
effects and mixing with recycled evapotranspired water (Ingraham & Taylor, 1991; Eltahir & Bras,
1996). Therefore, the present study is unique because data is presented from locations that are
situated to provide a more direct conduit to moisture advected from the Pacific, which allows
for the examination of how changes in vapor source are manifest in the isotopic composition of
precipitation that falls along the west coast of the US.
The isotopic composition of IWV has typically been estimated by assuming a source region
for the vapor using trajectory analysis and then calculating the isotopic composition of vapor that
13
would have evaporated from seawater in those source regions, if values for sea surface temperature
and humidity are known and an assumption is made that evaporation takes place in isotopic equi-librium
with the sea surface and the vapor diffuses through an unsaturated boundary layer (Craig
& Gordon, 1965; Wright et al., 2001; Yamanaka et al., 2002). The initial isotopic composition
of vapor would be modified following a Rayleigh model, as the vapor both loses and gains water
through processes of mixing and rainout (Hendricks et al., 2000). Recent satellite estimates of
the isotopic composition of water vapor provide valuable validations of this theoretical model
(Frankenberg et al., 2009; Worden et al., 2007). Given the inherent spatial and temporal complexi-ties
associated with the evolution of the vapor source, isotope-enabled General Circulation Models
(GCM) provide a tool to interpret the cause of isotopic variability that minimizes the reliance on
a priori assumptions about the source region (Henderson-Sellers et al., 2006; Hoffmann et al., 2000).
The efficacy of a GCM model simulation to capture the distribution of stable water isotopo-logues
throughout the atmosphere depends greatly on its ability to accurately depict the isotopic
fractionations that occur during phase changes and then to capture how the water vapor is moved
through the atmosphere, both vertically and horizontally. The former task has been well constrained
and validated with numerous model-empirical inter-comparisons (Hoffmann et al., 2000; Noone
& Simmonds, 2002). Yoshimura et al. (2008) were able to improve the representation of atmo-spheric
circulation by prescribing small-scale atmospheric processes with an Atmospheric General
Circulation Model while synoptic-scale patterns were constrained by the lower-resolution NCEP
Reanalysis II dataset, which captures the large synoptic scale system behavior that is associated
with precipitation over the western US (Neiman et al., 2008; Kanamitsu et al., 2002).
In the first section of this chapter I present 240 event-scale stable isotope measurements
(d18O and dD) for precipitated water from 96 individual storm events that struck the southwest
coast of the Unites States between 2001-2005. A basic description of this dataset is provided to
orient the reader to characteristics of storm to storm isotopic variability in the region. The catalog is
sufficient in size to produce a validation of the IsoGSM simulations representation of precipitation
14
over the region and consequently allow for an investigation of vapor source evolution (e.g. mixing
and rainout) prior to storms making landfall. From this analysis, a hypothesis is presented that
the isotopic composition of a precipitation event can be predicted based on the latitudinal origin
of a storm and the dynamics of a storm, which determine how much of the initial moisture source
is transported within the system (hereafter storm efficiency). This hypothesis is tested with a
numerical experiment, in which passive water molecules that do not fractionate during phase
changes are included in a nudged GCM simulation, a painted water experiment. An additional test
of this hypothesis is done using a third isotopic parameter called deuterium excess. To conduct this
test, a catalog of isotopic measurements from individual storm events striking northern California
and Washington are presented. In this analysis, a series of storms that swept along the entire west
coast (e.g. southern California to northernWashington) are selected from the complete catalog. It is
shown that a certain population of these storms produce the same deuterium excess values along the
entire west coast. This serves as empirical evidence of a high degree of efficiency in these systems.
The chapter concludes with a short discussion on published isotopic values from an inland station
in Colorado. The isotopic composition of precipitation from the central United States is shown
to be almost exclusively controlled by the temperature during condensation and can therefore be
described with a simple univariate regression model between surface temperature and the isotopic
composition of precipitation. The contrasting isotopic controls on coastal and inland sites is then
discussed.
2.3 Methods
2.3.1 Locations and Samples
Precipitation samples used in the study were provided from the National Atmospheric Deposition
Program (NADP) archives. Previous work has shown that the collection and archiving protocol used
by the NADP is adequate for isotopic analysis (Harvey, 2001, 2005; Vachon et al., 2007; Welker,
2000). The samples were collected from 2001-2005 at six sites shown in Figure 2.1.
15
-130
-130
-125
-125
-120
-120
-115
-115
-110
-110
-105
-105
30 30
35 35
40 40
45 45
50 50
Hopland
Pinnacles
Sequoia
Joshua Tree
Death Valley
Destruction Island
Olympic
Pawnee
Santa Maria
Figure 2.1: Map showing all isotopic monitoring sites referred to in this chapter.
16
2.3.2 Analytical Techniques
All water samples from the Joshua Tree, Pinnacles, Sequoia and Death Valley sites were analyzed
by a continuous flow method using a Thermofinnigan TC/EA and Delta Plus XP mass spectrometer.
The 0.5ml water samples were injected into He carrier gas and carried in vapor form to the TC/EA
reduction furnace where the water undergoes a pyrolysis reaction at 1400oC (H2O + C ) H2 +
CO). The reaction products are separated by a Gas Chromatograph column and analyzed directly.
Data are corrected using 5 in-house and certified standards. The precision is  0.2‰ for d18O and
 2.0‰ for dD. The methods are very similar to those described in detail in (Sharp et al., 2001).
All water samples from the Washington and Hopland sites were measured using Cavity Ringdown
Spectroscopy with a Picarro L1102. Approximately 0.5ml of water is injected into a vaporizer that
is maintained at 140oC. The vapor is introduced into a cavity on a stream of dry nitrogen gas where
a tuned laser is passed through the vapor sample. Because the different isotopologues of water have
unique absorption spectra, the relative abundance of the different water isotopes will affect how
much of the energy from the laser is absorbed. This is translated into a molecular abundance, which
is corrected to a standard ‰ isotope ratio through comparison against repeated measurements of
known isotopic standards (International Atomic Energy Agency’s, VSMOW, GNIP and VSLAP).
Each sample is measured 8 consecutive times. The first 3 samples are rejected because of slight
memory effects between samples and the average of the last 5 samples are used as the reported
result. After every 8 samples, an in-house isotopic standard is measured to ensure that stability of
the measurements. The uncertainty is less than  0.2‰ for oxygen and  1.0‰ for hydrogen. A
complete list of results is included in the Appendix.
2.3.3 Numerical Model
The principal tool used to interpret the isotopic variability of precipitation is the IsoGSM model
outputs from Yoshimura et al. (2008), which provide the isotopic composition for surface waters
and atmospheric vapor at 6-hour resolution on a 2.5x2.5o global grid from 1979-2008. The model
simulation was generated by fitting isotope tracers into the Experimental Climate Prediction Centers
17
Global Spectral Model with prescribed SSTs. The simulation was nudged to historical reanalysis
data, which allows for direct comparison between model outputs and historical isotopic observa-tions
on event-timescales, analogous to previous model validations using monthly integrated Global
Network of Isotopes in Precipitation samples or ice core data (Yoshimura et al., 2008; Schneider
& Noone, 2007; Lee et al., 2007). Further details of the methodology used to generate the model
outputs can be found in the original publication (Yoshimura et al., 2008). An additional experiment
was conducted for this study where I included passive water tracers in the same nudged simula-tion.
In these experiments the evaporative flux of water from a specified region is labelled following
an approach that is computationally analogous to the isotope-enabled simulations, except in these
experiments labelled water is assigned a fractionation factor (a) of 1. Therefore, the molecules
behave the same as regular water during phase changes and diffusion. These experiments thus
enable water to be conservatively traced through the hydrological cycle. The method is similar to
that described by (Kelley, 2003).
2.3.4 A Lagrangian Assumption
Implicit in the following discussion is that isotopic variability between storm events can be described
from a Lagrangian perspective where the vapor mixture at a given point point can be traced back
to a probabilistic source region. In its pure form, the Lagrangian equations track the movement of
an infinitesimally small particle through a three-dimensional fluid field (Draxler & Rolph, 2003).
The particle is theoretically non-reactive (conservative) and therefore its position in time will be
determined exclusively by the potential flow of the fluid matrix. There is a considerable literature
on the use of Lagrangian physics to describe the isotopic composition of atmospheric moisture
(for example: Pfahl & Wernli (2008, 2009); Sodemann et al. (2008) and references therein) but
nonetheless its use still requires a certain appreciation of some fundamental assumptions regarding
the chemistry that underly the reactions of water molecules in the atmosphere and the appropriate
choice in defining the state of the atmosphere. It should be noted that for discussing the isotopic
variability in the atmosphere, the most common alternative to the Lagrangian framework is the Eule-rian
coordinate system, where molecules are not tracked through the atmosphere but rather the flow
18
potential is calculated at fixed points in time. This approach has been discussed in Noone (2008).
The choice of coordinate system will not lead to any substantial differences in the interpretation of
isotopic variability but rather the two have distinct utilities depending on the scale of the question.
Because I am attempting to make an analysis of specific storm events with time scales of hours to
days, it is simpler to discuss isotopic variability from a Lagrangian perspective. This same approach
would become increasingly difficult if the goal was trying to define the isotopic composition of the
state of the atmosphere on long time scales, because the magnitude of the probable source region
would expand rapidly with time. For such studies, a Eulerian coordinate system is clearly preferable.
The choice of the Lagrangian coordinate system comes with uncertainties that are rooted in
three general locations 1) a proper definition of the state of the atmosphere on synoptic scales, 2)
the depiction of subgrid-scale physics and 3) the kinetic fractionation between water isotopologues
that occurs during phase changes. The atmospheric state was defined for this exercise using
the Global Spectral Model, where the primitive equations are solved using spherical harmonics
and then transformed onto a Gaussian grid (Yulaeva et al., 2008). It is generally accepted by
quasi-geostrophic theory that this approach will result in an accurate depiction of atmospheric flow
at the synoptic scale if boundary conditions are properly defined (Peixoto & Oort, 1992). The
atmospheric state in this model is subsequently ”nudged” at each time-step to temperature and
wind fields based on the closest approximation of the actual atmospheric state, which is defined
by Reanalysis fields that incorporate instrumental measurements into a fully realized atmospheric
circulation model (Kanamitsu et al., 2002). This technique effectively addresses the definition
of proper boundary conditions. While this procedure will produce fields that are correct from a
synoptic scale, it is possible that smaller-scale processes that play an active role within the storm
track are not properly represented.
In addition, it is not possible to define many of the critical sub-grid processes which require
parametrization schemes. Perhaps most important for understanding of water vapor transport
would be the convection scheme (a relaxed Arakawa-Schubert scheme) and cloud parametrization.
19
Consider an air parcel moving through the atmosphere, its water content will change as convection
entrains moisture into the air and will lose moisture during the formation of clouds. These processes
are therefore clearly important but at the current time must be represented only with parametriza-tion.
The choice of how to represent these processes is typically made in order to maximize the
relationship between the model outputs and observed surface conditions. Because mid and upper
tropospheric moisture in the extratropics is not well-defined instrumentally (i.e. there are few
physical measurements of tropospheric moisture content over the ocean and there are questions
regarding the appropriate algorithm for satellite data in both cloudy and low moisture regions), it is
not clear what the consequence of different parameterization schemes are on accurately depicting
moisture fields in this region, which lies in the pathway of the mid latitude storm track. Recent
discussion on this topic from both an isotopic and non-isotopic perspective can be found in Lee
et al. (2009a) and Sherwood et al. (2010) respectively.
The final consideration must be given to kinetic fractionation that occurs during phase change,
particularly with respect to the behavior at the sea-air interface. The Craig-Gordon model (Craig &
Gordon, 1965), has been shown to be immensely effective at depicting the isotopic composition of
the vapor flux from the ocean surface. However, the model is sensitive to two principal parameters
that are not well constrained, which are 1) the sensitivity of the kinetic fractionation factor to wind
speed, which evolves non-linearly according to (Merlivat & Jouzel, 1979) and 2) the difference
between the ocean skin temperature (where evaporation actually occurs) and SST measurements,
which integrate some depth into the ocean. For example, the IsoGSM model utilizes the relation-ship
between wind speed and kinetic fractionation according to Merlivat & Jouzel (1979), while
Pfahl & Wernli (2009) showed recently that by removing the influence of wind speed on kinetic
fractionation, one may more accurately capture the isotopic composition of the evaporative flux.
This study is indeed being pursued amidst active research on many of the topics discussed
in the above paragraphs. Improvements in instrumentation will undoubtedly improve the current
isotopic and non-isotopic depiction of the atmospheric state. This will allow for opportunities to
20
make more meaningful model benchmarks and lead to enhanced depictions of poorly understood
physical processes. However, it is also important to state that it is the presence of these very sources
of uncertainty that motivate this work. Particularly, this study is driven by the dearth of knowledge
on how moisture is transported in the mid latitude storm track.
2.4 Results
The local meteoric water line (LMWL) for each of the sites is consistent with that of previous
studies from the western US (Benson & Klieforth, 1989; Friedman et al., 2002; Ingraham &
Taylor, 1991; Smith et al., 1992) (Figure 2.2). The slopes of the LMWL are not statistically
different between the Sequoia (SEKI), Pinnacles (Pinn), Joshua Tree (JT), Hopland (HO) or
Olympic (Ol) sites however, the slope is notably smaller at the Death Valley (DV) site, which
is indicative of evaporation that occurs either while the precipitation fell or after it reached
the collector (Clark & Fritz, 1997). This is not unexpected given the extreme aridity at this
site. On the basis of these results we infer that post-precipitation evaporation did not affect the
integrity of samples from the SEKI, PINN, HO , OL or JT sites. Because evaporation bias cannot be
ruled out for the DV data, the results from this location are not included in the subsequent discussion.
Although this study is concerned principally with the cause of temporal variations in the
isotopic composition of precipitation, spatial patterns can be informative in highlighting the
mechanisms that drive isotopic variability (Bowen & Revenaugh, 2003). In Figure 2.3, probability
distribution functions are shown for each of the sites, highlighting how average isotopic values
decrease with latitude and altitude consistent with (Bowen & Revenaugh, 2003).
While precipitation is heavily weighted towards the winter months at these sites, seasonal
isotopic variability is also informative in highlighting the mechanisms that drive isotopic variability.
For example, Feng et al. (2009) using a global dataset argue that isotopic seasonality reflect shifts
in the latitude of the subtropical high pressure zones. There is a seasonal isotopic cycle at each of
21
the sites, consistent with Friedman et al. (1992), but it is also noted that there is a wide distribution
of isotopic values within each month (Figure 2.4). This result illustrates that although the mean
isotopic value for all storms during the winter months are more depleted than the summer months,
the isotopic intra-storm range of values during any single month encompasses values during any
other month of the year. Therefore substantial isotopic variability arises from mechanisms other
than that which drives the seasonal cycle. The seasonal deuterium-excess (dxs) cycle (Figure 2.4)
defined as:
dxs = dD􀀀d18O  8 (2.1)
shows stable values for most of the year with a sharp decline during the summer months.
The sharp decline in dxs during the summer months arises from continental-sourced moisture from
localized summer convective storms, which have anomalously low dxs values (Welker, 2000).
2.4.1 Model Validation
In order to validate the IsoGSM model’s ability to accurately depict isotopic variability at an event
scale, I compared the d18O of each measured value from a storm event with their model-predicted
d18O value. If a measured isotopic value represented precipitation from three consecutive days
of rainfall, I identified these three model days and calculated the amount-weighted isotopic value.
Event-scale sampling produces an episodic dataset and thus this approach is only effective if the
model is able to produce isotopic estimates at the correct time storms make landfall. Previous
work has demonstrated that NCEP Reanalysis data is effective in reproducing accurate estimates of
precipitation within the western US (Neiman et al., 2008) and there are very few instances when an
isotopic measurement was made, which had not been simulated by the model. The only necessary
adjustment to directly compare the model-simulated values with the measured values arose because
of the effects of elevation, which the model’s coarse topographic resolution does not resolve well
in complex orographic regions like the western US. As a consequence, the modeled values are
positively offset from measured values by an average of 1.8‰, at the higher altitude sites (SEKI and
JT). Despite this, the comparison between measured isotopic values and the corresponding IsoGSM
22
-150
-100
-50
-20 -15 -10 -5 0
δ18O
HO (m=8.0)
DV (m=6.7)
JT (m=7.8)
OL (m=8.3)
Pinn (m=7.7)
SEKI (m=7.5)
-10
δD
Figure 2.2: Relationship of d18O and dD (Local Meteoric Water Line) for all measurements made for
this study. The slope for each site is indicated on the figure. The global average slope is 8 and the low
values as observed at the DV site indicate evaporative enrichment of the sample.
isotopic values shown in Figure 2.5 demonstrates how well the model simulates the storm-to-storm
isotopic variability. The correlation between the isotopic composition of 96 measured and modeled
storm events is high (adjusted r2=0.50), providing confidence that IsoGSM captures the critical
processes associated with synoptic moisture transport in the mid-latitude storm track.
2.4.2 Regional and Synoptic Controls
To determine what mechanisms control the storm to storm isotopic variability, I selected a suite of
storm events from the observational and model simulated data base that were associated with the
most isotopically enriched and depleted precipitation (selected storms are denoted in the Appendix).
This subset includes only storms that made landfall and passed over the region shown in Figure 2.1.
Timeseries of areally-averaged precipitation rate, precipitable water and the isotopic composition
of precipitation are generated for the entire region encompassing the sample sites for each of
23
0.00
0.05
0.10
0.15
0.20
-20 -15 -10 -5 0
HO (39oN)
OL (48oN)
Pinn (36oN
0.00
0.05
0.10
0.15
0.20
-25 -20 -15 -10 -5 0
Pinn (317 m)
SEKI (1902 m)
δ18O
0.00
0.02
0.04
0.06
0.08
0.10
0.12
-10 -5 0 5 10 15 20 25 30
HO
Pinn
SEKI
OL
dxs
δ18O as a function of latitude
δ18O as a function of altitude
dxs as a function of latitude
Figure 2.3: Storms following a latitudinal transect (top), an altitudinal transect (center) and dxs. Land-falling
storms produce increasingly depleted d18O with increasing latitude and altitude. dxs does not
display a strong altitude effect, but does follow latitude. At each site a probability distribution function
using a normal kernel density estimator was fitted to the data.
the selected storms and then pooled together to generate a composite sequence of atmospheric
conditions that occur when enriched and depleted systems made landfall. As each storm system
approached the western US, there was a sharp increase in the precipitable water content, which
coincided with marked changes in the isotopic composition of the moisture in the atmospheric
column (Figure 2.6 panels A and D). In the model simulation, the isotopic composition of the water
column rises or falls by as much as 8‰ relative to the background moisture (i.e. prior to and after
the storm system passes through). As the precipitation rate fell to zero, the amount of water in the
atmospheric column and its isotopic composition returned to their background values (Figure 2.6).
(Coplen et al., 2008) argue that the isotopic variability within precipitation events reflects changes
in the temperature at which condensation occurs and thus isotopic values could be driven positive
24
-15
-10
-5.0
0.0
5.0
10
15
1 2 3 4 5 6 7 8 9 10 11 12 1 2 3 4 5 6 7 8 9 10 11 12
Month Month
Anomaly (‰)
Monthly δ18O distribution Monthly dxs distribution
Figure 2.4: Monthly distributions of storm events at all sites for oxygen and dxs. The isotopic compo-sition
of each storm is corrected to the average of the site and then box plots with quartiles are made for
each month. While there is a seasonal cycle where mean values in the summer are higher than the winter,
the distribution of individual storms overlaps almost completely between months. Deuterium-excess has
a more defined seasonality, with extremely negative values occurring during the summer months. Outlier
values are marked by open circles.
-10
-5
0
5
10
Modeled δ18O (anomaly)
Measured δ18O (anomaly)
Slope: 0.97
Intercept: 0.01
R: 0.72
-10 -5 0 5 10
0
10
20
30
40
50
Measured
IsoGSM
Frequency
Isotopic Anomaly
-12 -8 -4 0 4 8 12
Figure 2.5: Comparison between d18O values of measured storm event and their predicted values based
on the IsoGSM simulation. The mean isotopic values were subtracted from each dataset to correct for
the coarse topography in the model simulation (left panel). The colors are used to denote the different
sites. The distribution of the same storm events shown in the left panel, bins are 1‰ and the line is the
best fit Gaussian distribution (right panel).
25
-27
-24
-21
-18
-15
0 6 12 18 24 30 36 40
0
0.0001
0.0002
0.0003
0.0004
0.0005
0.0006
0.0007
0.0008
0 6 12 18 24 30 36 40
-30
-27
-24
-21
-18
-15
0 6 12 18 24 30 36 40
0
0.0001
0.0002
0.0003
0.0004
0.0005
0.0006
0 6 12 18 24 30 36 40
5
10
15
20
25
30
35
0 6 12 18 24 30 36 40
10
15
20
25
30
35
0 6 12 18 24 30 36 40
δ18O(vapor) δ18O(vapor)
Specific Humidity (kg/m2) Specific Humidity (kg/m2)
Precipitation Rate (kg/m2/s2) Precipitation Rate (kg/m2/s2)
Columnar Water Vapor Precipitation Rate Precipitable Water
Time (hrs) Time (hrs) Time (hrs)
A B C
D E F
Figure 2.6: Time evolution of d18O of precipitable water (left column), precipitation rate (middle col-umn),
and atmospheric specific humidity (right column) during a sequence of the most enriched (top
row) and depleted (bottom row) storm events. The mean value for the different storms are shown as a
bold gray line with circle markers. All values were taken from the IsoGSM simulation. Time 0 represents
an arbitrary beginning point before precipitation began to fall moving forward in 6-hour time steps.
or negative with negligible antecedent changes in the isotopic composition of the integrated water
column. The results of the analysis shown in Figure 2.6 imply that the range of storm to storm
variability actually reflects comprehensive and heterogeneous changes in the isotopic composition
of the entire water column and not variations in the magnitude of isotopic fractionation during
condensation.
The convergence of moisture that saturates the atmospheric column includes vapor derived
from a wide radius surrounding the frontal storm center. The isotopic changes in the atmospheric
column (Figure 2.6, left column) thus track the evolving mixture of water that is entrained locally
26
and from remote moisture sources that has been transported by the storm system (Trenberth et al.,
2003). It is not feasible to isolate the proportional contributions of the local and remote moisture
to the integrated water vapor column. Hence, I consider whether there is a difference in local
moisture fluxes associated with the isotopically enriched and the isotopically depleted storms and
then consider how wind fields and moisture transport influences the source and relative contribution
of remote moisture. In figure 2.7, it is shown that the enriched storm events are associated with
higher surface latent heat flux (taken as a proxy for evaporation) just offshore. Thus, a portion of
the isotopic difference between the enriched and depleted columns can be attributed to increased
evaporation from the underlying local coastal water prior to landfall. This is consistent with what
would be predicted using a Rayleigh distillation model where local (i.e. younger) waters are
predictably more isotopically enriched (Feng et al., 2009; Noone, 2008; Clark & Fritz, 1997) and
the increased flux of this source would thus shift the column towards more isotopically enriched
values. A higher latent heat flux is indicative of warmer SST conditions, which additionally
reduces the magnitude of isotopic fractionation during surface evaporation (Majoube, 1971). The
fractionation effects at the ocean-atmosphere interface coupled with simply the increase in this
local moisture source appear therefore to shift the isotopic composition of the near-surface vapor
and the subsequent precipitation towards more enriched values.
To distinguish the relative influences that locally entrained and remotely advected moisture
have on storm to storm changes in the integrated water column, I deconvolve the atmospheric
column into discrete pressure horizons and identify the height in the atmospheric column at which
isotopic changes were most pronounced during the composite enriched and depleted events (Figure
2.8). The most pronounced isotopic changes occur not along the surface but rather between
the 800-700hPa pressure surfaces. The large isotopic changes at these heights cannot likely be
attributed to the entrainment of moisture from below. Instead, they reflect moisture that is being
advected into the region by winds aloft. This observation is consistent with theoretical discussions
in Trenberth et al. (2007) who suggest that up to 70% of the moisture associated with major
27
landfalling storms in the subtropics and mid-latitudes is composed of vapor from distant sources.
Similarly, experiments with tagged water molecules in an ensemble of GCM simulations, show
that midlatitude cyclones, characteristic of the major rainfall events in the western US, may be
drawing on water from a radius of over 25o latitude (Kelley, 2003). Bao et al. (2006) test this idea
by looking at moisture convergence and moisture trajectories in a subset of land-falling storms and
provide more empirical evidence for the presence of remote moisture sources in certain systems that
make landfall along the west coast of the US. Thus, based on the magnitude and height at which
maximum isotopic changes occur within the atmospheric column and the independent tagging
experiments (Kelley, 2003) as well as the empirical moisture convergence studies (Bao et al., 2006),
it is concluded that the principal mechanism affecting the storm to storm water column isotope
budget is convergence of moisture from distant sources rather than variable moisture flux from local
terrestrial or marine surfaces.
In figures 2.9 and 2.10 I show both the mean prevailing wind fields and d18O of integrated
water vapor over the Pacific during the composite (6 events) of the most enriched and depleted
storm events from Figure 2.6. Isotopically depleted storms are associated with a high pressure
anomaly that is centered near 30oN over the central Pacific (Figure 2.9). This generates strong
low level northerly flow along the west coast of the US, which advects isotopically depleted
moisture from the Aleutian Low region over the southwestern US. This is in clear contrast with the
isotopically enriched storms that are characterized by more zonal flow and a high pressure center
that is located further south near 25oN (Figure 2.9). By subtracting the composite wind vectors
from one another it is evident that anomalous southwesterly flow is characteristic of the most
enriched isotopic storm events (Figure 2.9). The prevailing southwesterly winds lead to a wide
swath of isotopically enriched moisture that stretches from the central Pacific, near Hawaii, to the
southwestern US (Figure 2.10). The isotopes appear to provide a tracer of the poleward flux of low
latitude moisture and confirm the suggestion of both Dettinger et al. (2004) and Bao et al. (2006)
that southwesterly storms do indeed tap into and export tropical water vapor to the western US.
28
Figure 2.7: Difference between latent heat flux (w/m2) for the enriched and depleted composites. High
latent heat flux from just offshore is a common feature of the most enriched events. Data for latent heat
is from the North American Regional Reanalysis dataset (Mesinger et al., 2006).
Figure 2.8: Top panels (left and right) show the precipitation rates in six hour time steps as two iso-topically
enriched storm events strike the western US. The bottom panels show vertical cross sections of
the isotopic composition of water vapor as the storms pass over the region. The figure shows how the
maximum isotopic changes (on the order 8-10 ) occur between the 800-700 mb levels.
29
Depleted storms Enriched storms Enriched-Depleted
Figure 2.9: A plan view of the average 850 mb wind fields during the most depleted (left) and enriched
(center) events and the difference between the two vector fields (right). Scale bar shows the length of a
10 m/s vector.
Figure 2.10: Isotopic concentrations of d18O of water vapor during depleted (left) and enriched (right)
events. Water vapor is taken for the 850 mb level. Colors show isotopic anomalies relative to the field in
view while contours show absolute isotopic concentration relative to VSMOW.
The isotopic plume portrayed in the right panel of Figure 2.10 suggests that the vapor trans-ported
by the most enriched storms have the capacity to conserve the isotopic composition of mois-ture
from their low latitude source region. However, water vapor is typically short-lived in the
30
atmosphere (9 days) and thus a high rate of turnover from entrainment and condensation would typ-ically
obscure the capacity of isotopes to behave as a conservative tracer. Because the analysis on
individual storms was based on only a relatively small population whose dynamical behavior may
well be aberrant, further validation of the isotopic influence of distant moisture sources is accom-plished
by calculating the correlation coefficient between annually-averaged d18O of precipitation
over southern California and gridded meridional moisture flux. This analysis takes advantage of the
longer timeseries available from the IsoGSM simulation and allows for an assessment of whether
the relationships inferred on an event-scale are stable on timescales relevant to proxy records (e.g.
annually-resolved). I find that meridional moisture flux from across the tropical Pacific is correlated
with the isotopic composition of precipitation in southern California (Figure 2.11), which suggests
this region would be particularly sensitive to ocean-atmosphere changes in this region. In addition,
increased meridional moisture flux along the southwesterly storm track, leads to enriched isotopic
values. I also note negative correlations between d18O of precipitation and meridional moisture flux
on the west side of the Pacific Basin consistent with the enhanced southerly flow west of 150oW
during depleted storm events as presented in the left panel of Figure 2.9.
2.4.3 Mesoscale Controls
Although the event scale analysis helps to to delineate specific moisture source regions that have
direct influences on the isotopic composition of landfalling precipitation, because of the wide
radius that mid latitude cyclones draw moisture from, it is critical also to consider how mesoscale
circulation with seasonal to interannual timescales influence the distribution of water isotopologues
across the basin. To assess this, I regress 6-hourly values for the isotopic composition of water vapor
from the IsoGSM simulation against meridional and vertical moisture flux (Figure 2.12, top and
bottom panels respectively). In the lower troposphere between 10oN and 40oN there is a strongly
positive correlation between meridional moisture flux and the isotopic composition of vapor. In both
model and satellite-derived estimates of lower tropospheric water vapor, there is a steep latitudinal
gradient in the isotopic composition of water vapor across this region, with d18Ovapor dropping by
approximately 10‰ over these latitudes. The positive correlation between meridional flow and
31
Figure 2.11: Correlation coefficient between annually average vertically integrated meridional moisture
flux and amount weighted d18O of precipitation over the southwestern US. Contours indicate correla-tions
that are significant at the 95% confidence based on a Student’s T-test.
the isotopic composition of water vapor is thus interpreted to largely reflect poleward transport of
enriched vapor by transient eddies, which are the principal mechanism to drive meridional moisture
flux at these latitudes (Peixoto & Oort, 1992). Poleward of 40o, the relationship weakens until it is
no longer significant and eventually reverses sign. An increase in meridional moisture flux is thus
actually associated with depleted isotopic values in the high northern latitudes of the Pacific Basin.
The lack of a positive correlation could arise because of an absence of any measurable isotopic
gradient in the high latitudes of the north Pacific (Frankenberg et al., 2009; Craig & Gordon, 1965)
and from the fact that meridional moisture transport is minimal north of the subtropics (Peixoto &
Oort, 1992). It is not clear why the relationship actually reverses sign and does not simply dissipate.
This may result from the influence that humidity or skin temperature changes have on the isotopic
composition of the evaporative flux, which do directly influence the meridional moisture transport.
32
Figures 2.12 and 2.13 illustrate how overturning circulation influences the isotopic composi-tion
of water vapor over the Pacific basin. Vertical velocity is considered to exert an important
control on the isotopic composition of atmospheric moisture by influencing ascent (subsidence)
of isotopically enriched (depleted) moisture (Feng et al., 2009; Noone, 2008). In the IsoGSM
simulation, this circulation behavior is manifest in isotopic contours that loosely follow the
boundaries between rising and subsiding air masses (Figure 2.13). To characterize the influence
that vertical mixing has on the isotopic composition of vapor, I calculate point-by-point correlations
between vertical velocity and the isotopic composition of precipitation (Figure 2.12). Regions of
dominantly rising air (0-10oN and 45-60oN) display a positive correlation between the isotopic
composition of water vapor and vertical velocity. A similar relationship was noted by (Feng et al.,
2009) and can be explained as arising from the fact that by suppressing convection and rainout, the
moisture is on average more enriched. The opposite relationship can be observed in the subtropics
where increased subsidence leads to isotopic values that are depleted (e.g. a negative relationship
between vertical velocity and vapor). Therefore, more vigorous overturning (e.g. anomalous ascent
at the low latitudes and descent in the subtropics) would collectively lead to depleted isotopic values
across much of the Pacific basin.
2.4.4 Water-tagging
A further analysis of the source controls on the d18O of precipitation is conducted by tagging all
waters evaporated from the tropical Pacific between 0-20oN and 120-180oW. In this way water
molecules can be conservatively tracked through the atmosphere. In order to quantify the relation-ship
between the percentage of low latitude moisture during a precipitation and its isotopic com-position,
the isotopic composition of the water column is plotted against the percentage of tagged
water in the column (Figure 2.14). There is a complex and poorly defined relationship between the
two on a 6-hourly basis, suggesting the the isotopic composition of the atmospheric column cannot
be described as a simple product of moisture source origin. However, when days in which storm
events strike are selected from the simulation, a far more coherent picture emerges. The absence or
presence of low latitude moisture accounts for greater than 50% of the isotopic variability. Although
33
Figure 2.12: Correlation coefficients between meridional moisture flux and the d18O of vapor for a
vertical cross section of the Pacific between 120-180oW (bottom). Correlation coefficients between
vertical velocity (omega) and the d18O of vapor for a vertical cross section of the Pacific between 120-
180oW (top).
the slope of this relationship is presented as only a tentative result, it suggests that a loss of 20%
of the tagged moisture would result in a 2‰ depletion of the average isotopic composition of the
water column (Figure 2.14). In Figure 2.15, I show a plan view of the fraction of tagged water in the
atmospheric column during the composite of enriched (right) and depleted (left) storm events. This
figure is comparable to Figure 2.10 except showing tagged water as opposed to the isotopic com-position
of the water column. The results depict that up to 40-50% of the moisture in the enriched
storm events striking the western US are derived from a remote southwesterly source (Figure 2.15).
For illustrative purposes the atmospheric river event from January 22, 2005 is shown (Neiman et al.,
2008), where the moisture making landfall is composed of approximately 70% tagged moisture.
34
Figure 2.13: A cross section across the Pacific basin with contours showing the average isotopic concen-tration
of water vapor and colors showing vertical velocities (positive values are dark gray and negative
values are orange).
70%
0% 10% 20% 30% 40% 50% 60% 70%
-10
-20
-30
-40
-50
δ18O
-22
-20
-18
-16
-14
-12
-10
0% 10% 20% 30% 40% 50% 60% 70%
Percentage of tagged Water
R2=0.52
Percentage of tagged Water
δ18O Figure 2.14: The relationship between the relative percentage of tagged water in the atmospheric column
over southwestern US and the isotopic composition of the integrated water column (left). All days
associated with storm events were selected from the figure on the left showing the coherent relationship
between tagged water and the isotopic composition of water during landfalling frontal storms.
35
Fraction of tracer
Figure 2.15: A composite of tagged water concentration for a series of isotopically enriched (right) and
depleted (left) storm events. The concentration of tagged water in the atmospheric column is taken as
the ratio of the mass of tagged water to total water (e.g. specific humidity).
Figure 2.16: Plan view of the 850mb wind fields and relative percentage of tagged water in the atmo-spheric
column during the atmospheric river event on January 22, 2005.
36
2.4.5 Deuterium-excess gradients
Deuterium-excess (equation 2.1), is considered a more robust tracer of moisture source region than
either d18O or dD individually. This is because the deuterium excess (dxs) value of a water mass
is effected most strongly by the differing degrees of kinetic enrichment between the isotopologues
of water during evaporation (Merlivat & Jouzel, 1979). The moisture evaporated from the source
region is thus marked by the temperature and humidity of the marine atmosphere from where
the moisture originated. Processes occurring during water mass evolution (i.e. condensation) are
considered to occur in near equilibrium conditions and therefore impose only a small change in the
deuterium excess characteristic of the water mass (Merlivat & Jouzel, 1979). On average there is
an approximately 4‰ gradient in dxs values between northern Washington and central California
based on monitoring from the Global Network of Isotopes in Precipitation sites in Santa Maria, CA
and Destruction Island off the coast of Washington. This gradient reflects not surprisingly, a more
southerly source for the moisture in southern California. As an empirical test of the results of the
tagged water simulation, which suggested a highly efficient transport of remote moisture in certain
storm systems, a series of 20 storm events from 2001-2003 that swept along the entire west of the
US are selected from the catalog and the dxs gradient is calculated (dxsCali f ornia-dxsWashington). The
mean dxs gradient of all storms ( 3.8‰) is found to be similar to the gradient which was calculated
from the monthly mean gradient GNIP values (4‰). From a storm to storm perspective, the dxs
gradient can differ quite substantially from the mean, with some storms displaying no measurable
dxs difference between the northern and southern sites while other storms producing values that
are a factor of 2 greater than the mean. To interpret the cause of the variability of storm to storm
dxs gradients, I explore the atmospheric conditions that prevailed during the high and low gradient
storms using the North American Regional Reanalysis dataset (Mesinger et al., 2006).
In Figure 2.17, I show the precipitation rate during storms that displayed a high (left), low
(right) and normal (left) dxs gradient. In Figure 2.18, I show the vertically integrated meridional
moisture flux for the high (left) and low (middle) gradient storms. The difference between the two
depict the dominant poleward moisture flux that prevails during the low gradient storms. This is
37
interpreted to reflect that low dxs gradients arise because of a common moisture source along the
entire west coast as opposed to more localized moisture from the nearby coast. In Figure 2.19, a
similar analysis is done to look at latent heat flux as a means to identify different moisture source
regions. The analysis does not present a clear depiction of moisture source locations but suggests
that low dxs gradient are associated with increased evaporation from an equatorward source while
high dxs gradients are associated with high latent heat flux north of 35o.
The analysis presented suffers a bit from a small dataset used to generate composites and
also added analytical uncertainty because dxs is a second order parameter and thus accumulates
uncertainty from both the d18O and dD measurements. The analysis of both meridional moisture
flux and latent heat flux suggest that storms with an absence of any measurable dxs gradient arise
from a common low latitude moisture source. This would be akin to storms along the entire west
coast showing high amounts of tagged water as depicted in Figures 2.15 and 2.16. The contrary can
be stated of storms with high dxs gradient where entrainment of moisture leads to storms having a
dxs signature that reflects their latitude and corresponding zonal dxs value Although this analysis is
based on a small data set, it portrays a satisfying test of the results from the numerical modeling
exercises.
2.5 Discussion
2.5.1 Isotopes and 21st century hydrologic changes
A complete description of how ocean and atmospheric variability influences the stable isotope
composition of precipitation over the western US has been hampered by a lack of isotopic
measurements from discrete storm events and an analysis of how rainout and mixing from source to
sink influences the isotopic composition of the moisture. The strong correlation between observed
and simulated isotopic values for storm water provides an important validation of IsoGSMs (and
by corollary NCEP 2 Reanalysis’) ability to simulate a moisture budget for southern California.
The correlation between observed and modeled isotopic variability is robust despite the fact that
38
Figure 2.17: Precipitation rate (kg of water/m2) during storms included in the high (left), low (middle)
and average (right) dxs gradient events. The figure emphasizes that the storms influenced the entire
coast.
convective processes that entrain water into the storm system and raindrop physics are not well
depicted in this type of model (Risi et al., 2008; Lee et al., 2009b). Continued efforts to refine the
analytical representation of these processes will undoubtedly improve isotope simulations of this
kind, though for regions where moisture convergence is predominately the result of large-scale
frontal systems, the major processes are adequately represented. This finding is important because
accurate depictions of moisture transport associated with storms in coupled models is seen as an
obstacle in efforts to improve 21st century hydrologic forecasts for semi-arid subtropical regions
(Seager et al., 2007b). While isotopic tracers have proven important in the representation of the
hydrological cycles in GCM simulations at seasonal or intra-annual timescales, the results presented
here illustrate the utility of using isotope tracers for simulations of event-scale processes that
involve complex condensation and mixing processes.
39
Figure 2.18: Vertically integrated meridional moisture flux (kg/m) during storms included in the high
(left) and low (middle) dxs gradient events and the difference between the two composites (right).
Figure 2.19: Latent heat flux (W/m2) during storms included in the high (left) and low (middle) dxs
gradient events and the difference between the two composites (right).
40
Not surprisingly, there are indeed important regional controls on the isotopic variability of
individual storms that arise because of nearshore sea surface temperature variability (Figure 2.7).
Warmer SSTs reduce the fractionation between ocean water and vapor and also increase the
entrainment of locally evaporated waters (Majoube, 1971; Craig & Gordon, 1965). The local effects
however are not adequate to explain why the isotopic composition of precipitation is driven to such
extreme values. To account for the full range of variability, consideration of the influences associ-ated
with basin-scale atmospheric circulation are required. The large isotopic variability between
storms arises principally from the mechanism initially proposed but not tested by Friedman et al.
(1992), which is the advection of moisture from source regions of distinct isotopic composition.
I find that precipitation over California that is isotopically enriched arises from storms that tap
into subtropical and tropical moisture sources and have trajectories that are more direct than those
systems which take a circuitous route around the middle latitude quasi-permanent high pressure
cells (Dettinger et al., 2004).
Large cyclonic storms draw on moisture from a wide radius, and thus a simple delineation
of the source region neglects the complex mixing that occurs as these storms evolve. On a
basin-wide scale, low-level meridional moisture flux will enrich the vapor sources across the
subtropics and into the mid latitudes. Low frequency changes in eddy-driven poleward moisture
transport would consequently shift the average isotopic value of storms towards values that are
enriched relative to time of reduced overturning. In contrast, an increased vigor of overturning
circulation, would deplete the vapor fields across the region most storms pass through and shift the
average isotopic value of precipitation towards more depleted values.
Simulations of how the concentration of water vapor in the atmosphere responds to radiative
forcing has been investigated by numerous authors (Schneider et al., 2010; Hall & Manabe, 1999;
Held & Soden, 2006) and provide a test bed to consider the regional isotopic response to a global
climate perturbation. Climate projections for the 21st century for example, predict a northward
41
shift in the mid-latitude storm track with a slight reduction in the latitudinal temperature gradient
(Rind et al., 2001; Yin, 2005; Vecchi et al., 2006). If these changes in circulation were acting
alone, a poleward shift in the storm track would lead to a reduction in the isotopic composition of
precipitation during the 21st century for the midlatitudes by increasing (decreasing) the relative
contributions of water from northerly (southerly) storms. However, an increase in the water
holding capacity of warmer air over the tropics could enhance moisture transport from the tropics
to the extratropics albeit with a slope slightly shallower than the global rise in humidity (Lorenz
& DeWeaver, 2007), which would in principal offset the isotopic impact that would arise from
the poleward shift in the storm tracks. To the extent that these changes have already begun, a
compilation of isotopic records from midlatitude sites would aid in assessing if the hydrological
changes in the subtropical high pressure zones as predicted by GCMs driven with rising greenhouse
gas concentration are indeed taking place (Seager et al., 2007b; Hoerling & Kumar, 2003). For
example, rising isotopic values of precipitation in the subtropics and mid latitudes, would indicate
that the poleward shift in the storm track is being offset by changes in the water holding capacity of
the tropical atmosphere. On the contrary, if the isotopic values document a long term decline, this
would provide evidence of a reduced presence of low latitude moisture to the region. Delineating,
which of these two scenarios is underway is consequential not only to regional aridity forecasts but
also to assessing the response of the global hydrological cycle to increased radiative forcing.
2.5.2 Isoscapes and Proxy reconstructions
When the multiple competing influences on the isotopic composition of precipitation along the west
coast of the US are considered, it is clear that caution must be exercised in interpreting isotopic
variability from this region in terms of a single climate variable. I offer a preliminary discussion,
which will be elaborated on more fully in Chapter 5, which is to consider spatial isotopic patterns
across the west coast by integrating multiple records into isotopic networks or isoscapes (Bowen
et al., 2009), which can be used to deconvolve controls associated with local to synoptic variability
42
from basin-scale changes. An example of this is illustrated in Figure 2.20, where enriched
(depleted) isotopic values occur across the entire west coast of the US in years with a strong (weak)
and westward shifted Aleutian Low and anomalous southerly (northerly) flow (Figure 2.20, left and
right panels) while an isotopic dipole pattern emerges, with enriched values south of 45o during
years when there is a strong eastward-shifted Aleutian Low and anomalously strongWesterly winds
(Figure 2.20, center panel). Therefore, a latitudinal transect of proxy reconstructions from along the
western US could be used to delineate between paleo-circulation patterns that generate basinwide
or dipole-like patterns. The interpretation of isotopic variability in western North America in terms
of specific atmospheric modes was suggested by (Birks & Edwards, 2009) who find that the Pacific
North American pattern (a leading mode of low frequency pressure anomalies over the North
Pacific) is a good predictor of isotopic values in central Canada. Their approach attempts to fit
isotopic variability into an established climate mode with well-documented teleconnection patterns.
Longer simulations would allow for more rigorous approaches to defining spatial isotopic patterns
(i.e. Empirical Orthogonal Function analysis of isotopic fields) and test whether isotopic modes are
truly analogous to climate modes defined through SLP or SST patterns. For example, (Field, 2010)
show that the isotopic composition of precipitation over Europe appears related to a NAO-like mode
but with centers of action that are distinct than those defined strictly through SLP patterns.
2.6 Isotopic controls at Inland Sites
Discussion thus far has focussed on near coastal sites which are more sensitive to moisture advected
directly from the Pacific. Sites further inland have been shown to have a strongly linear relationship
with surface temperatures (Vachon et al., 2007; Harvey, 2001). This is because multiple marine
and continental moisture sources mix together and tend to subdue variability that would arise form
changes in the moisture source (Friedman et al., 2002). Therefore changes in the temperature at
which condensation occur serve as the principal control of isotopic varibaility. Using the GNIP
event station in the Pawnee Grasslands of Colorado, I show the presence of a dominant relationship
43
between surface temperature and the isotopic composition of 102 precipitation events that fell
between 1994-1998 (Figure 2.21). To test if this relationship is influenced by source region, I cluster
storms by their prevailing source region using Lagrangian Trajectory analysis (Draxler & Rolph,
2003). 72-hour back trajectories that were initialized at a height of 1000 m above the ground surface
were run with NCEP Reanalysis atmospheric boundary conditions were calculated for each storm
(Figure 2.24). Storm were classified as either North, Northeast, Northwest, Southeast, Southwest,
or South. The slope and goodness of fit between temperature and the isotopic composition of
precipitation (Dd18O-T) is not degraded between storms of different origins (Figure 2.21), which
suggests fundamentally that the isotopic controls are dominated by local processes.
The event data spans only 4 years, so as a test of the long term stability of the relationship
between the isotopic composition of precipitation and surface temperature, I utilize the isotopic
composition of monthly isotopic values from the ECHAM 4 (Hoffmann et al., 1998), GissE
(Schmidt et al., 2007) and IsoGSM (Yoshimura et al., 2008) model outputs. The ECHAM4 and
GissE simulations were not nudged but were run with prescribed 20th century SSTs, which permits
a test of how sensitive Dd18O-T is to perturbations in SST patterns. The Dd18O-T was calculated
at the Pawnee Grid Point using the complete time series of each model (Figure 2.22) and then
Dd18O-T was calculated for randomly shuffled 5-year windows during the 20th century using
a Monte Carlo simulation to assess the stability of the Dd18O-T relationship. The mean slope
was subtracted from the slope calculated for each 5 year window and probability distribution
functions of the slope anomalies were calculated for each of the 3 models. In 90% of the iterations,
the calculated Dd18O-T is within 5% of the cumulative 20th century slope (Figure 2.23), which
suggests an exceptionally stable relationship between temperature and the isotopic composition of
precipitation at this site.
2.7 Conclusions
This chapter presents a new catalog of 367 isotopic measurements of storm events striking sites
along the west coast of the US. The isotope data is used to provide a test of the representation
44
Figure 2.20: Average annual 850 mb geopotential height (m) and wind vector anomalies from NCEP 2
Reanalysis (Kanamitsu et al., 2002) during 1989, 1998 and 2003 (left to right, top row) and the isotopic
anomalies in precipitation associated with these same years (bottom row).
of water vapor transport in the Experimental Climate Prediction Center’s Global Spectral Model,
which has been isotope-enabled (IsoGSM). The model is able to reproduce the large range of
storm-to-storm stable isotope variability in precipitation, suggesting that it accurately captures
the moisture transport that is associated with large frontal storms that constitute most of the
annual water budget across the southwestern United States. I document large shifts in the isotopic
composition of atmospheric water vapor when storms make landfall, which points to convergence
of remote moisture sources that are isotopically distinct from moisture evaporated from the adjacent
coastal ocean. Southwesterly storms are associated with the most enriched isotope values, as they
tap into isotopically enriched low latitude moisture sources that is transported poleward. This is
45
-30