Tree mortality during California’s drought

Last summer, after four years of extreme drought, more than 21 million trees died in California.

This figure is based on mortality surveys performed by the U.S. Forest Service, which every summer for the past 10 years has flown a small aircraft over most of the forested area of California and recorded the locations of dead trees. The mortality observed in 2015 was by far the worst ever recorded. Mortality was especially intense in the Southern Sierra Nevada, where, across many large landscapes, the majority of the conifers died.



Tree mortality in the Southern Sierra. Photos: U.S. Forest Service

I’ve been working with the U.S. Forest Service aerial mortality monitoring program and UC Davis and Yale colleagues Jens Stevens, Mason Earles, and Andrew Latimer to analyze the mortality patterns recorded by the Forest Service and to understand the factors that lead certain forests to suffer more during drought. The first step of the analysis was to take the set of polygons that the aerial observers drew around dead trees (here’s an example of those polygons from the Southern Sierra Foothills that you can view in Google Earth) and convert them into a regular grid in which each cell is assigned a value representing the number of dead trees observed inside it. Here’s the resulting grid for the mortality throughout the state in 2015, highlighting the serious situation in the southern Sierra:

View larger map

The map shows the mortality amount observed in each grid cell (adjusted proportionally for mortality patches that overlap multiple grid cells and/or only partially overlap a given grid cell). The map only includes grid cells that fell completely within the plane’s field of view (the Forest Service also reports their flight lines and approximate observation distance) in order to avoid bias in our subsequent analysis of mortality patterns–this explains the linear gaps in the mortality grid.

Our next step was to see how tree mortality rates changed with time, year after year, as the drought progressed. To do this, we converted the mortality survey data from each year into a grid the same way we did for the 2015 data. We can visualize the mortality over time as an animation:


The map makes it painfully clear that although there is always some amount of mortality each year, the mortality in 2015–particularly in the southern Sierra–is far greater than that observed during other years in recent history. It is interesting to note that 2105, the year in which mortality spiked, was the fourth year of extreme drought in the state. This observation highlights the fact that tree mortality can take several years to respond to drought. Such a delayed response is often observed in studies of drought stress, and the existence of this delayed response hints that we are likely to observe high mortality well into 2016 and potentially beyond, especially in Southern California, where the severe drought continues for a fifth year.

My colleagues and I have used the Forest Service aerial mortality survey data, combined with other sources of environmental data–including long-term climate–to evaluate the factors that predispose forests to experience high mortality during drought. Our analysis is currently in review at an academic journal–stay tuned for a description of what we found!

How closely related are adjacent trees?

amriv1What is the likelihood that two adjacent trees of the same species share a parent? Addressing this question is an essential first step in my work to evaluate the adaptive capacity of California trees. Many tree populations exhibit “spatial genetic structure,” in which trees that are located near one another are generally more closely related than trees located farther apart. This pattern is largely a product of the fact that many seeds from each mother tree fall within a relatively restricted radius. When there are few potential parent trees scattered across a landscape (for example, following a severe wildfire), the seedlings that establish in a given area are most likely offspring of the nearest adult tree. However, when adult trees are abundant, a given site often contains offspring from many different parents, and adjacent seedlings may be only distantly related.

I am conducting a study of the relatedness among the Douglas-fir trees in my study sites in order to refine the sampling design for my main study of adaptive potential. For that study, I need to sample trees that are as closely spaced as possible without sharing any parents; the trees must be spaced closely to minimize differences in environmental variables such as topographic slope and soil depth among trees, but they cannot share parents because my sample must represent the range of genetic variation present within the site. If I sample many closely-related trees, I will not know whether certain genotypes are common on a site because (a) they are highly adapted to the site or (b) they all came from the nearest parent tree (or both).

Last week I received an initial, draft set of set of molecular genetic data from a microsatellite assay performed for me by the National Forest Genetics Lab (NFGEL). The lab analyzed DNA from 127 Douglas-fir trees from 9 of my study sites in the central Sierra. Within each site, I sampled multiple clusters of adjacent trees so that I could analyze relatedness across a range of distances from very near (adjacent trees) to very far (different study sites). The sampling design is illustrated in the map below.

spatial structure map

Map of sampling design. The left panel shows the locations of the study sites, and the right panels depict the spatial arrangement of sampled trees at three example sites.

By comparing the genotypes of any two trees, it is possible to estimate how closely related they are; trees that have more DNA in common are more closely related, and the exact amount they have in common roughly translates into a familial relationship (e.g., full-siblings, half-sibling, cousins, etc.). I used a software package called SPAGeDi (Spatial Pattern Analysis of Genetic Diversity) to estimate relatedness among every pair of sampled individuals and determine how the degree of relatedness varies with spatial distance among the trees.

relatedness plot

The preliminary analysis across all study sites reveals that trees that are closer together are generally more closely related, which in genetics jargon means that there is “spatial genetic structure” within populations. In the figure above, a relatedness coefficient of 0.125 represents the relatedness of half-siblings (trees that share a mother but have different fathers). Relatedness near the half-sibling level appears to be common at a distance of about 0 to 50 meters (~150 feet). Above 50 meters, relatedness is much lower–below the relatedness of cousins. Based on this analysis, an appropriate sampling distance for my climate adaptation project is between 50 and 100 meters. This analysis is based on draft data; NFGEL is performing a second run to fill in data gaps, and once I have the final data I will repeat this analysis and evaluate each study site separately to determine if the patterns of relatedness vs. distance vary among sites.

A big thank you to the Neale Lab at UC Davis, which extracted DNA from my Douglas-fir needle samples, and NFGEL, which performed the microsatellite assay.

Can California’s trees adapt to climate change? Part II: Initial sample collection & processing

Douglas-fir trees in the Yosemite area may eventually be a thing of the past if they are unable to adapt to increasing drought stress under climate change. I have begun a project to evaluate the adaptive potential of Yosemite trees in collaboration with the National Park Service, and this summer we conducted initial fieldwork. The field season had its fair share of frustration–a huge wildfire, two medium wildfires, and a government furlough, to name a few–but all told, it was a huge success.

We collected tree cores, needle samples, and size measurements from over 300 Douglas-fir trees from 12 sites in and around Yosemite in just 20 days of fieldwork (not counting site scouting, break days, and travel days). And this was just a warm-up for the real work planned for next year.

The tree cores offer a window into each tree’s relationship with climate. It is often possible to find correlations between the width of each annual growth ring and each year’s climate (e.g., drought intensity). For instance, the amount of growth reduction observed in drought years can indicate a tree’s drought tolerance and help predict how it will respond to anticipated increases in drought stress given ongoing climate change. We extract cores from trees using a hand-powered borer–essentially a long, hollow drill bit.


To get data from the cores, we mount them on grooved boards, sand them, scan them with a high-resolution scanner, and load the images into a program that automatically identifies ring boundaries and measures ring widths.

YOSE1-004coreTo make sure we identified the rings correctly, we “cross-date” the cores by comparing ring widths across cores. In the graph below, the horizontal axis represents year (going back in time from 2012) and the vertical axis represents relative ring width. Each color represents a different core. In this example, the two cores show good agreement in growth response from one year to the next.

CaptureThe next step is to correlate ring widths with climate variables to determine each tree’s sensitivity to climate. Among other factors, differences in climate-sensitivity from one tree to the next may be due to genetic differences among trees. By relating each tree’s genetic composition to its drought tolerance, we can identify mutations that increase drought tolerance and use that information to infer a population’s capacity to adapt to future increases in drought stress. The first step in obtaining genetic data is collecting tissue–in this case, needles. The task–my favorite–requires a very specialized skill that involves slinging a weight (tied to a thin rope) over a lower canopy branch (sometimes 20 meters up!) and shaking until a twig falls down. We then clip the needles off the twig and store them with desiccant in a fancy barcoded vial for inventorying and storage in the Forest Tree Genetic Stock Center until we are ready for DNA extraction and sequencing.


OLYMPUS DIGITAL CAMERA             IMG_4602-001

We also map the precise location of each tree by designating a “reference point” (e.g., a tall, dead tree), obtaining an accurate GPS reading for it, and measuring the distance and bearing from each sampled tree to the reference using a compass and laser rangefinder. I wrote some custom computer scripts to process the bearing-and-distance records and compute UTM coordinates for each tree. Here is a map sampled trees in one plot, with circle size proportional to tree diameter. The points don’t line up perfectly with trees from the aerial photograph, but the locations are precise enough to re-locate the trees in the field on future visits.

bigcreekAfter we finished most of our sampling in and around the northern part of the park, the Rim Fire tore through and burned more than half of our sampled plots. We’ll have to return next spring to assess mortality of the sampled trees. If mortality was high, we’ll face an interesting decision: what do we do with the DNA from a bunch of dead trees? Jurassic Park schemes aside, there are a few options. 1) Proceed with the analysis because it will provide general insights into the dynamics of climate adaptation in the Sierra Nevada conifers despite the fact that the specific sampled stands no longer exist; 2) Select new sampling sites with lower mortality to replace those that burned at high severity–this will ensure managers have data on real-world stands they are interested in managing; or 3) Return to the high-mortality sites in a few years to sample regenerating seedlings for an inter-cohort comparison–does fire promote adaptation to changing climate by shortening generation time, or does it reduce adaptive potential by depleting the pool of potential parent trees? I’d be quite excited to address the latter question, but of course it all comes down to funding.

yose map-001

Growing seedlings in the Sierra = playing with fire

Last fall, I installed a series of six “common gardens” along an elevation gradient in the Sierra Nevada near Foresthill. At each site, I planted Jeffrey pine seeds collected from four different elevations–from 4000 ft to 7500 ft–in the northern Sierra. The goal of the experiment is to see whether seeds perform best in the climate where they are produced (that is, they are “locally adapted”) or in a different climate (that is, they are “maladapted” to their site). Because the reproductive trees in any given site established long ago (trees take decades to reach maturity and can survive for hundreds of years), it is possible that the climate today is quite different than it was when those trees established. If locally adapted populations are unable to track changes in climate, we can expect a significant decline in forest tree performance under climate change.

I installed the plots in areas that burned at moderately high severity during the 2008 American River fire, as Jeffrey pine seedlings usually establish soon after a disturbance–usually fire–kills mature trees and frees up important resources such as light and water. Working in burned patches can be quite unpleasant; due to the lack of overstory, there is little shade, and the shrubs that often establish soon after fire can be thorny and dense. But burned patches can also be strikingly beautiful.


Mule’s ears and whitethorn ceanothus near the highest-elevation common garden site.

This August, the American Fire started burning near my plots. I watched the fire perimeter expand each day, enveloping one, then two, then three of my common gardens. Last week, the main road to my plots was finally re-opened, so I went  to survey the damage.  Given the Sierras are a naturally fire-prone system, I understood the risks of planting seedlings here, and I established numerous plots over a relatively wide area to hedge my bets. The lowest- and highest-elevation plots were spared, but the intermediate plots got toasted. Here is (was) one of my common gardens:

IMG_4534 IMG_4536All that remained were some metal stakes and screws and the fried circuit board of a climate datalogger.

My plots were all within the perimeter of the 2008 American River fire, so this summer’s American Fire cleared away the remaining litter and woody debris, leaving behind the proverbial “moonscape.” Other areas nearby hadn’t burned in 2008 and contained abundant mature trees. This year’s fire scorched most of those trees right to the crown, leaving behind some eerie scenery.

IMG_4558 IMG_4573Fall colors canceled.

Naturally, fires in this type of forest, known as “mixed-conifer,” are frequent but quite mild–rarely are they so severe as to scorch needles all the way to the crown over huge swaths of land. However, a century of fire suppression by the Forest Service has allowed substantial fuel (fallen needles and woody debris) to accumulate so that when fires do occur, they are very severe and can cause high mortality, even in these fire-adapted species.

I collected seedling mortality data from the plots that didn’t burn. As long as some seedlings survive, those that die are also data points!

IMG_4496A one-year-old Jeffrey pine seedling.IMG_4503A dead Douglas-fir seedling.

I haven’t thoroughly analyzed the data yet, but some interesting trends are already evident.


These figures are from the lowest-elevation common garden at 4800 ft. The figure on the left shows the proportion of planted seeds that germinated in the spring, emerged, and survived through June 10th. The plot on the right shows the proportion of seedlings alive on June 10th that survived to October 2nd. Interestingly, the higher-elevation seedlots initially performed better than the local seeds at this low-elevation site. However, later survival showed the expected pattern of local adaptation in which the seeds sourced from nearest the planting site performed better than those from higher elevations. This suggests that seedlings require different conditions at different stages of development and that overall adaptation to climate is not as straightforward as often assumed. Stay tuned for an integrated analysis of first-year survival at all sites.

Can California’s trees adapt to climate change? Part I: Prospects for Douglas-fir in Yosemite

Douglas-fir (Pseudotsuga menziesii) is one of the world’s most important timber species. It is often considered the emblematic tree of low-elevation Pacific Northwest forests, but its full native range spans an impressive 36 degrees of latitude (about 2,500 miles) from central British Columbia into central Mexico. Within the Sierra Nevada, Douglas-fir occurs as far south as Yosemite—the southernmost population in the Sierra is found just 25 miles south of the park. This local range limit is intriguing, especially because the conifer species Douglas-fir occurs with in the northern Sierra (think ponderosa pine, incense cedar, and white fir) thrive well south of Yosemite. What is unique about Douglas-fir? An exploratory analysis I have conducted using climate data from PRISM and forest inventory data from FIA and VTM hints at an explanation—one that doesn’t bode well for the fate of these trees as climate continues to change.


This map shows the northern and central Sierra Nevada, and filled symbols represent occurrences of Douglas-fir based on forest inventories and my own observations around Yosemite. The fill color indicates the annual heat moisture index (AHM), which roughly approximates the degree of water limitation at each stand. AHM is calculated as weighted annual average temperature divided by total annual precipitation. It is readily apparent in the map that the most southerly populations of Douglas-fir experience among the highest AHM of all Sierra Nevada Douglas-fir. If the southern populations are at the limits of their drought tolerance, then increased climate warming and drying might be expected to drive Douglas-fir out of the Yosemite region.

But there is hope! Even though the Douglas-fir populations around Yosemite experience high moisture limitation, they may be capable of tolerating drought stress beyond the levels they currently experience. Unfortunately, a recent high-profile global study concluded that most tree populations exist very near the limits of their drought tolerance and have little wiggle room. Whether or not this generalization applies specifically to Douglas-fir in California, however, remains to be explored. Another hope lies in the possibility that tree populations will be able to “migrate,” or disperse seeds into newly-suitable sites outside the species’ historical range, as climate continues to change. A recent study of natural migratory ability in trees, however, concluded that migratory ability is weak and unpredictable; other studies have also drawn very mixed conclusions.

Even if the trees in Yosemite today have little climatic wiggle room and poor migratory ability, there is a potential that their offspring, though evolutionary adaptation, will be able to tolerate more intense drought, perhaps to such a degree (no pun intended) that they will be capable of persistence through continued changes in climate. The potential for evolutionary adaptation depends on the amount of genetic variation present in the populations around Yosemite. Populations at range limits generally have low levels of genetic variation. Whether or not this generalization applies to the trees in Yosemite is completely unknown, despite its critical importance for predicting and responding to the impacts that inevitable climate change will have on the region. And it is just what I have proposed to study in a grant proposal to the National Park Service.