Our research includes the methodological and applied aspects of geographical information science. On the methodological front, we mainly focus on geocomputation, remote sensing techniques, spatial analysis and spatial data accuracy. On the applied front, we are interested in application of geospatial techniques in solving large-scale ecological and geographical problems, with emphasis on the effects of invasive species, climate change, and human disturbance on terrestrial ecosystems.

Current Research Projects:

1) Spatial Uncertainty

It has been estimated that there are more than 2500 million specimens in natural history collections. With the increasing interest in understanding changes in environmental, biological, and cultural resources due to human disturbance and climate change, specimen collections have become ever more important, since they can provide baseline information on the environment and the factors driving change.  Before the advent of geographical information systems (GISs) and global positioning systems (GPSs), occurrence information for most specimens was stored as textual descriptions without explicit geographic coordinates. This is a major obstacle for managing and analyzing specimen data in a GIS. Meanwhile, assessing and recording these uncertainties during the georeferencing process is arguably as important as determining coordinates for the locality, because only with the uncertainty can one determine if the location information is suitable for a particular analysis. In this project, we are interested in developing methods to georeference textual descriptions as well as estimate locality uncertainties. In addition, we are also interested in studying the impact of locality uncertainties and other spatial data uncertainties on environmental modeling and the process of decision making.

2) Object based remote sensing classification methods

Object based image analysis (frequently called OBIA) is an approach to classifying high resolution remotely sensed imagery that is currently seeing increased use due to the proliferation of imagery with small pixel sizes, such as those provided by digitized aerial photographs and the IKONOS and QuickBird satellites. Conventional pixel-based methods, while useful in classifying coarse-scale remotely sensed imagery, are less suitable for classifying high resolution images. Generally, an object based image classifier includes two major steps: first an image is segmented into similar image objects (or segments) and then the objects are classified based on attributes of and interrelations between segmented objects. We are interested in developing methods and theory that will help us better apply OBIA method in classifying high resolution images.

3) Adaptive management of Sierra Nevada forests

The goal of the research proposed here is to learn how to use an adaptive management and monitoring system to understand ecosystem behavior, incorporate stakeholder participation, and inform the implementation of adaptive management for Forest Service lands in the Sierra Nevada of California. This project is a multi-campus collaborated project. My lab focuses on extracting the topography and vegetation information from numerous available existing and remotely sensed geodatasets augmented with field measures acquired using GPS, leaf area meters, and laser rangefinders. We will acquire higher resolution remote sensing images (e.g. IKONOS) and LiDAR data to map in detail the vegetation and topography of the sites, the change of the crown closure, leaf area index, and other vegetation parameters before and after the treatments. These data will improve our understanding of the spatial relationships between canopy and forest characteristics and changes in habitat, fire, and hydrologic responses (e.g. soil moisture and thus stream response). We also plan on scaling up these relationships from the catchment to the forest level using coarser resolution remotely sensed imagery and topography. For more information, please visit: http://snamp.cnr.berkeley.edu.

4) California real-time solar irradiance mapping

The goal of this project is to determine the feasibility of mapping global horizontal (GHI) and direct normal (DNI) irradiances for the entire state of California using a combination of satellite image processing and high-quality, redundant ground solar monitoring experiments. We will use GOES-WEST satellite data, ozone, water vapor, and aerosol optical thickness GIS layers to build the solar model that be used to generate near real time global and directed normal irradiance for the entire state of California. The final products will be distributed on the internet using webGIS technology to ensure timely data delivery and to facilitate the use of solar energy in California.

5) Environmental niche modelling

With the increasing availability of digital ecological data, environmental niche modeling has gained much attention for a range of ecological applications. However, one of the major challenges for the use of niche models is the lack of absence data in ecological observation datasets (e.g. those collected by museums, or on wildlife surveys). In this research, we are interesting developing methods in solving the presence-only data problem in ecological modeling. In addition, we are also developing a freely available and user-friendly software (ModEco) that enables researchers and students to fully and easily explore the rapidly increasing wealth of species distribution data. ModEco will consist of four major components: data management and visualization, feature analysis, model training and prediction, and accuracy assessment.

6) Species distribution, climate change, and land use / land cover change

We are interesting in understand the relationship between species distribution, climate change and land use/land cover change at the landscape and continental scales. Currently, my lab is working on two related projects: 1) studying the change of vegetation in California over the past 100 year and its relationship with climate change, and land use and land cover change; and 2) statistically downscaling multiple GCM models output to a finer spatial resolution, which will then be used to study the impact of climate change on species distribution.

Acknowledgment