History
Share |

Project Detail

CMG: Functional Data Modeling of Climate-Ecosystem Dynamics

Project Number: 0934739 Source: National Science Foundation
Principal Investigator: Surajit Ray Organization: Trustees of Boston University
Project Duration: 09/15/2009 - 08/31/2012 Fiscal Year: 2009
Recovery Act: No   No Award: $356,651


Abstract
The goal of this project is to develop better methods for analyzing the response of vegetation to changes in climate. Satellite observations of the earth's surface can be used to track changes in vegetation over the last three decades, and the task of relating these changes to changes in climate is both important and challenging. Among the significant changes are the earlier onset of the growing season and "greening" and "browning" trends occurring in high northern latitudes, apparently in response to increases in surface temperature. Research conducted in this project will attempt to quantify climate-vegetation relationships using "functional data analysis", a form of analysis in which a set of functions is used to represent variations of vegetation in space and time. Once these functions are defined, their characteristics can be related to trends and fluctuations in climate. Changes in the date of spring onset, the length of the growing season, and other phenological changes can be studied by associating them with properties of the functions and their derivatives. For example, the onset and termination of growing seasons may be defined by points of inflexion, or zero-crossings of the second derivatives of the functions, while the peak of the growing season occurs at a zero-crossing of the first derivative. A particular challenge in relating remotely-sensed vegetation to climate variations is the heterogeneity of the land surface, as surface characteristics can vary tremendously over short distances. The challenges posed by land surface heterogeneity will be addressed by "mixture modeling". In this modeling framework the different land surface types found in satellite observations are represented by a linear superposition of probability density functions, which can be characterized through cluster analysis. The work will lead to improvements in our ability to identify and quantify the vegetation response to climatic changes in regions where land surface type is highly variable. Research conducted under this grant will address a question which is both scientifically and societally important. The question of how vegetation is affected by climate change is important from a conservation standpoint, since climate change can threaten ecosystems and biodiversity. Human welfare also depends on ecosystem services which may be interrupted by changes in climate. In addition, vegetation changes can act as a feedback on climate change, since vegetation can affect climate by changing surface albedo, regulating terrestrial uptake of carbon dioxide, and modulating surface evapotranspiration. In addition, the tools developed in this project will be applicable to a variety of problems at the intersection of statistic and earth system science. Methods developed in the project will be disseminated to students in courses on statistics and natural sciences, and software and sample datasets will be made available to the scientific community to encourage adoption of new methodologies.