UC Berkeley

ABSTRACT

Complexity in Climatic Controls on Plant Species Distribution: Satellite Data Reveal Unique Climate for Giant Sequoia in the California Sierra Nevada

by

Eric Kindseth Waller

Doctor of Philosophy in Environmental Science, Policy, and Management

University of California, Berkeley

Professor Dennis D. Baldocchi, Chair

A better understanding of the environmental controls on current plant species distribution is essential if the impacts of such diverse challenges as invasive species, changing fire regimes, and global climate change are to be predicted and important diversity conserved. Climate, soil, hydrology, various biotic factors (e.g., competition, disease, dispersal), fire, history, and chance can all play a role, but disentangling these factors is a daunting task. Increasingly sophisticated statistical models relying on existing distributions and mapped climatic variables, among others, have been developed to try to answer these questions. Any failure to explain pattern with existing mapped climatic variables is often taken as a referendum on climate as a whole, rather than on the limitations of the particular maps or models. Every location has a unique and constantly changing climate (challenging our definition of climate), so that any distribution could be explained by some aspect of climate. This is not a rationale for doing so - such models would certainly be overly fit to the given conditions and not generalizable. It is an argument for a more complete evaluation of possible climatic controls and the interaction of climate with other important variables (e.g., soil). A lack of adequate maps for various climate and climate-derived variables plays a large role in the failure of modelers to sufficiently address climatic complexity. In particular, site water balance, radiation, humidity, wind, and temporal variability in all of these factors may be poorly understood factors in controlling distributions. Weakness in the mapping of these variables is well recognized in the water balance modeling field, but is less emphasized in the species distribution modeling field, despite the fact that variables that affect the water balance are also likely to play a major role in species distribution. In this dissertation, I 1) improve the mapping of cloud cover from satellite imagery, in order to generate accurate monthly cloud frequency variables; 2) use these cloud frequency variables to demonstrate the possible importance of cloud cover, a previously overlooked climate variable, to the distribution of a particularly charismatic plant species, giant sequoia; and 3) further investigate the climate associated with the frequent cloud cover in the vicinity of giant sequoia, to identify other climatic factors that may also be important to the tree's distribution.

Chapter 1 of this dissertation reviews some of the major flaws in species distribution modeling (with existing climate data) and addresses concerns that climate may therefore not be predictive of, or even relevant to, species distributions. Despite problems with climate-based models, climate and climate-derived variables still have substantial merit for explaining species distribution patterns. Additional generation of relevant climate variables and improvements in other climate and climate-derived variables are still needed to demonstrate this more effectively. Satellite data have a long history of being used for vegetation mapping and even species distribution mapping. They have great potential for being used for additional climatic information, and for improved mapping of other climate and climate-derived variables.

Improving the characterization of cloud cover frequency with satellite data is one way in which the mapping of important climate and climate-derived variables (e.g., water balance) can be improved. An important input to water balance models, solar radiation maps could be vastly improved with a better mapping of spatial and temporal patterns in cloud cover. Chapter 2 of this dissertation describes the generation of custom daily cloud cover maps from Advanced Very High Resolution Radiometer (AVHRR) satellite data from 1981-1999 at ~5 km resolution and Moderate Resolution Imagine Spectroradiomter (MODIS) satellite reflectance data at ~500 meter resolution for much of the western U.S., from 2000 to 2012. Intensive comparisons of reflectance spectra from a variety of cloud and snow-covered scenes from the southwestern United States allowed the generation of new rules for the classification of clouds and snow in both the AVHRR and MODIS data. The resulting products avoid many of the problems that plague other cloud mapping efforts, such as the tendency for snow cover and bright desert soils to be mapped as cloud. This consistency in classification across cover types is critically important for any distribution modeling of a plant species that might be dependent on cloud cover.

In Chapter 3, monthly cloud frequencies derived from the daily classifications were used directly in species distribution models for giant sequoia (Sequoiadendron giganteum (Lindley) Buchholz) and were found to be the strongest predictors of giant sequoia distribution. A high frequency of cloud cover, especially in the spring, differentiated the climate of the west slope of the southern Sierra Nevada, where giant sequoia are prolific, from central and northern parts of the range, where the tree is rare and generally absent. Other mapped cloud products, contaminated by confusion with high elevation snow, would likely not have found this important result. The result illustrates the importance of accuracy in mapping as well as the importance of previously overlooked aspects of climate for species distribution modeling. But it also raises new questions about why the clouds form where they do and whether they might be associated with other aspects of climate important to giant sequoia distribution. What are the exact climatic mechanisms governing the distribution? Detailed aspects of the local climate warranted more investigation.

Chapter 4 investigates the climate associated with the frequent cloud formation over the western slopes of the southern Sierra Nevada: the "sequoia belt". This region is climatically distinct in a number of ways, all of which could be factors in influencing the distribution of giant sequoia and other species. Satellite and micrometeorological flux tower data reveal characteristics of the sequoia belt that were not evident with surface climate measurements and maps derived from them. Results have implications for species distributions everywhere, but especially in rugged mountains, where climates are complex and poorly mapped.

Chapter 5 summarizes some of the main conclusions from the work and suggests directions for related future research.