Quantitative Ecology & Environmental Science

I earned my PhD in the Department of Environmental Science, Policy, and Management at UC Berkeley.  I was part of the de Valpine group, which focuses on sophisticated statistical modeling of complex ecological data. I also had the good fortune to work with a variety of other teams on various projects, some of which are ongoing.

Sierra Mixed-Conifer Growth, Survival, Fecundity, and Population Modeling

My dissertation features growth and survival modeling for mixed-conifer forests in the Sierra Nevada.  I work with the reserve compartments at Blodgett Forest Research Station, which have seen no management in approximately 100 years other than fire exclusion.  Ultimately I hope to create fecundity models as well and then combine all three demographic models in an Integral Projection Model for these forests (IPM, the continuous version of a matrix projection model – not to be confused with Integrated Population Modeling, or Integrated Pest Management).  These models are capable of handling complex population dynamics, incorporating environmental variability, and differences in individual quality (some trees might grow, survive, or reproduce better than others).  I have published a model for white fir growth in Ecological Applications, and the model for survival is in press at Ecosphere. White fir growth is affected by the size of the tree (larger trees grow faster) and the basal area of the plot (more crowded plots had slower growth) – but additional variation was important for both individual trees and plots. See a concise summary on the Sierra Nevada Adaptive Management Project webpage.  Survival for the seven species we modeled at the site was controlled by different variables for different species.

(Note: I have corrected a database error which affected both of these papers.  Please see the Errata for the Ecological Applications paper and the Ecosphere paper.)

Ideally these models can eventually be made for different treatments at Blodgett, shedding light on how management (fire, thinning, or other treatments) affects the population dynamics of these trees.  In addition, the model for reserve plots can be used for simulations.  For example, artificially inflating mortality for sugar pine from white pine blister rust might give an idea of how this disease will affect sugar pine population dynamics.

I also attended a short course on IPMs at the Max Planck Institute for Demographic Research in Rostock, Germany and contributed to an R package on IPM analysis, “IPMpack.”

Conifer Encroachment into Oak Woodlands in Mendocino and Humboldt Counties

I was also a part of a team investigating the rate, extent, and ecological effects of conifer encroachment into oak woodlands in the North Coast of California.  When we keep fire out of these woodland and forest ecosystems, Douglas-fir in particular is able to grow into dense forests where there were previously open oak woodland stands.  Because conifers (and several other hardwoods) are shade-tolerant, the seedlings recruit and grow under the oak canopies, and then eventually grow through the oak canopy and overtop them, shading them out and killing them.  This is a widespread problem with many ecological impacts, including the obvious effects on oak trees, but also on the understory vegatation community.  The field crews recorded information on age and size structure of the trees as well as the diversity of understory vegetation.  My role in this project was to analyze historical photos from the 1940s and current imagery from the National Agricultural Imagery Program to determine the rate and extent of conversion from oak to conifer.  I used Leica Photogrammetry Suite to georegister and orthorectify the historical imagery, ArcGIS to manipulate the images, create maps, and calculate topographic variables from digital elevation models, and eCognition for object-based classification of these high-resolution images. I presented this analysis in a conference paper at the Seventh California Oak Symposium: Managing Oak Woodlands in a Dynamic World. I then modeled the change in woody cover as a function of topographic proxies for moisture and light availability, using Generalized Additive Models  to account for spatial autocorrelation. These models appeared in the third chapter of my dissertation and are in press at Ecological Informatics. The most straightforward and important conclusion from this study was that the results regarding topographic controls on woody cover growth were highly dependent on the spatial scale of the remote sensing data: for different pixel sizes, the results could change completely. The larger team is in the process of creating a publication through UC Agriculture and Natural Resources describing these issues of conifer encroachment and their threats to oak conservation.

Fire History in the Santa Cruz Mountains

Fire has been an important process for all California ecosystems.  Native peoples frequently burned landscapes to maintain open tree stands and healthy understory vegetation suitable for basket-making, medicine, and food.  Lightning strikes were often ignition sources as well, so Chuck Striplen led a study on the fire history in the Santa Cruz mountains, where there are few lightning ignitions.  He obtained tree cores for Coast Redwood near the traditional lands of the Quiroste, and determined from fire scars in the wood how frequently fires occurred in that landscape.  Some of these cores, however, had no bark and therefore no way to determine specifically what year a fire happened – these are called ‘floating chronologies.’  I assisted Dr. Striplen with an innovative statistical model treating each year in the tree core as a binary (yes/no) trial, with ‘yes’ indicating that a fire happened in that year, and ‘no’ that there was no fire.  We were able in this way to use these cores to investigate some of the spatial determinants of the fire history.  The results can be found in a Joint Fire Science Program report.

Habitat Preferences of the Western Spadefoot in Orange County

With Kathy Baumberger, I modeled the upland habitat preferences of the western spadefoot, a fossorial (burrowing) amphibian closely related to toads. The spadefoot is dependent on vernal pools for breeding, and therefore much of the conservation efforts are focused on preserving existing pools or mitigating their destruction. While these are important steps in supporting amphibian species in California, upland habitat use is key as well.  We radio-tagged adult spadefoot and measured their locations throughout most of a year, and then modeled their preferences for burrow sites and aestivation (dry-season hibernation) sites, as well as what determined their movements. This work is currently under review at the Journal of Herpetology.

Oak Population Modeling in the Sierra Foothills

I am working with Katharine Suding’s group on modeling blue oak populations at Sierra Foothills Research and Extension Center.  California tree oaks (valley oak, blue oak, coast live oak and interior live oak, canyon live oak, Englemann oak, and black oak) are ecological and cultural keystone species, meaning that they are critical to the rest of their ecosystems (providing food and structure for many animals and shelter for plants) as well as to the cultural life of many Californians.  Their acorns were a food staple for native peoples and their gnarled, majestic shapes are still iconic and dear to many of us.

Most of these species are thought to be declining because we see many seedlings and many adults but nothing in between.  If seedlings never make it to be adults, then we are looking at forests and woodlands made up of “standing dead.”  But this assertion has rarely been checked with a population model, which is essential because in long-lived species it is normal to have fewer adults, many offspring which may not survive, and infrequent transitions between juvenile and adult. Each adult tree only needs to replace itself approximately once every 100 years.

In addition, oak populations may be very spatially variable as well as history dependent (some locations may have been grazed or burned 100 years ago with effects that persist to the present), and recruitment of seedlings to saplings may be stochastic (random!) or pulsed (happens all at once very infrequently).  These possible population dynamics are very challenging to model, but I believe it is critical to start with the population model in order to better understand how oak populations change over time.  While at first it may seem that we have little information to make a population model, there are more data out there than one might think.  In particular, at UC reserves and research properties, researchers have been taking data on tree oaks for years, and in some cases conducting grazing experiments as well.  We are starting with blue oak at SFREC because there is enough data to set up the model.

Hospital Data from Lindsay Wildlife in Walnut Creek

Finally, I have worked on two projects involving data from the wildlife hospital at Lindsay Wildlife in Walnut Creek.  First, I looked in detail at the hospital records for all mourning doves treated from 1993-2010.  As the most abundant species in the database, mourning doves are a good species to study to see how survival may depend on the length of stay in the hospital, the bird’s weight, or the time of year it was brought in. Weight and age category have a strong relationship, implying that age categories are well defined and consistently used by intake volunteers.  The hospital has treated mourning doves slightly more effectively over time: the length of stay in the hospital for survivors is 6 days shorter. Different age classes show different seasonality: youngest age classes peak in July, while immature classes peak around August.  This reflects the life history of the species in the wild.  Mention of “cat” in the dove’s record is not independent of whether the animal dies in the first 24 hours, supporting the experience of rehabilitators that cats can be quite lethal to wildlife, even when the wounded animal is brought to the wildlife hospital. Some of the results of this project will enable the wildlife hospital to make record-keeping more consistent and to improve care, while some results reflect the natural history of the species.

Second, I am looking at the wildlife hospital’s records over the last 20 years to see if trends expected of urbanization are seen.  In rapidly urbanizing cities like Brentwood and Pleasanton, the populations of birds which adapt to urban landscapes such as crows and house finches are likely to increase, while bird species that prefer to avoid urban environments such as blue-gray gnatcatchers and Steller’s jays are likely to decrease.  In a similar way, certain bird guilds, such as insect-eaters, tend to decrease, while others (such as grain-eaters) tend to increase.  I am analyzing the hospital database to see if these trends are reflected in the hospital’s intake of different species.  I am also using U.S. census data to account for cities which are developing more rapidly than others.  Because retaining higher biodiversity is an important conservation goal and urbanization tends to homogenize species composition, testing this hypotheses about urban adapters and avoiders can be important in city planning.

Agricultural Management Practices and Vegetated Buffers

In between my Master’s degree and my PhD, I worked with Minghua Zhang’s Agricultural GIS lab at UC Davis.  I was insturmental in helping to collect and synthesize information for an agricultural management practices report for the Central Valley Regional Water Quality Control Board.  This agency is responsible for calculating Total Maximum Daily Loads (TMDLs) for various pollutants, including sediment, nitrogen, and pesticides, among others.  In their regulatory capacity they wanted to have suggestions for people to mitigate agricultural runoff pollutants in order to meet TMDLs.

During the time I was working on the literature review on this topic, I also developed a mathematical theoretical background section for a meta-analysis of vegetated buffer efficacy.  Vegetated buffers are strips around the edges of agricultural fields which are planted with vegetation intended to slow down overland flow of water, and therefore allow pollutants and sediment to settle out on the edge of the field rather than traveling into a nearby waterway.  My colleagues had created statistical models with results from a collection of studies which evaluated how wide a buffer needed to be in order to capture a certain percentage of the pollutant.  I added a section describing mathematically why the pollutant removal should be an exponential (saturating) function of buffer width: a zero-width buffer removes zero pollutant, but as the buffer widens, it removes more and more, until eventually it has removed nearly all the pollutant.  This paper was published in the Journal of Environmental Quality.