Advanced

Analyzing Vegetation Trends with Sensor Data from Earth Observation Satellites

Jamali, Sadegh LU (2014)
Abstract (Swedish)
Popular Abstract in Swedish

Sammanfattning



Mänskliga aktiviteter och klimatförändringar påverkar vegetationen över hela

jorden, idag och i framtiden. Denna påverkan kan studeras och kvantifieras med hjälp av observationer från satelliter vilka förmedlar information om

vegetationsförändringar i tid och rum. I denna doktorsavhandling har metoder för analys av vegetationsförändringar utvecklats och utvärderats. Lämpligheten hos olika typer av tidsserieanalys och statistiska metoder har testats och utvärderats i områden med kända vegetationsförändringar. Vidare har användarvänliga gränssnitt utvecklats för att underlätta för andra användare av satellitdata att utnyttja de utvecklade... (More)
Popular Abstract in Swedish

Sammanfattning



Mänskliga aktiviteter och klimatförändringar påverkar vegetationen över hela

jorden, idag och i framtiden. Denna påverkan kan studeras och kvantifieras med hjälp av observationer från satelliter vilka förmedlar information om

vegetationsförändringar i tid och rum. I denna doktorsavhandling har metoder för analys av vegetationsförändringar utvecklats och utvärderats. Lämpligheten hos olika typer av tidsserieanalys och statistiska metoder har testats och utvärderats i områden med kända vegetationsförändringar. Vidare har användarvänliga gränssnitt utvecklats för att underlätta för andra användare av satellitdata att utnyttja de utvecklade metoderna. Metoderna bidrar till förenklade studier, kvantifiering och bättre förståelse av vegetationsförändringar och deras relation till miljöförändringar och mänsklig påverkan. (Less)
Abstract
Abstract



This thesis aims to advance the analysis of nonlinear trends in time series of vegetation data from Earth observation satellite sensors. This is accomplished by developing fast, efficient methods suitable for large volumes of data. A set of methods, tools, and a framework are developed and verified using data from regions containing vegetation change hotspots.



First, a polynomial-fitting scheme is tested and applied to annual Global Inventory Modeling and Mapping Studies (GIMMS)–Normalized Difference Vegetation Index (NDVI) observations for North Africa for the period 1982–2006. Changes in annual observations are divided between linear and nonlinear (cubic, quadratic, and concealed) trend... (More)
Abstract



This thesis aims to advance the analysis of nonlinear trends in time series of vegetation data from Earth observation satellite sensors. This is accomplished by developing fast, efficient methods suitable for large volumes of data. A set of methods, tools, and a framework are developed and verified using data from regions containing vegetation change hotspots.



First, a polynomial-fitting scheme is tested and applied to annual Global Inventory Modeling and Mapping Studies (GIMMS)–Normalized Difference Vegetation Index (NDVI) observations for North Africa for the period 1982–2006. Changes in annual observations are divided between linear and nonlinear (cubic, quadratic, and concealed) trend behaviors. A concealed trend is a nonlinear change which does not result in a net change in the amount of vegetation over the period.



Second, a systematic comparison between parametric and non-parametric

techniques for analyzing trends in annual GIMMS-NDVI data is performed at

fifteen sites (located in Africa, Spain, Italy, and Iraq) to compare how trend type and departure from normality assumptions affect each method’s accuracy in detecting long-term change.



Third, a user-friendly program (Detecting Breakpoints and Estimating Segments in Trend, DBEST) has been developed which generalizes vegetation trends to main features, and characterizes vegetation trend changes. The outputs of DBEST are the simplified trend, the change type (abrupt or non-abrupt), and estimates for the characteristics (time and magnitude) of the change. DBEST is tested and evaluated using both simulated NDVI data, and actual NDVI time series for Iraq for the period 1982-2006.



Finally, a decision-making framework is presented to help analysts perform

comprehensive analyses of trend/change in time series of satellite sensor data. The framework is based on a conceptual model of the main aspects of trend analyses, including identification of the research question, the required data, the appropriate variables, and selection of efficient analysis methods. To verify the framework, it is applied to four case studies (located in Burkina Faso, Spain, Sweden, and Senegal) using Moderate-resolution Imaging Spectroradiometer (MODIS)–NDVI data for the period 2000–2013. Each of the case studies successfully achieved its

research aim(s), showing that the framework can achieve the main goal of the

study which is to advance the analysis of nonlinear changes in vegetation.



The methods developed in this thesis can help to contribute more accurate information about vegetation dynamics to the field of land cover change research. (Less)
Please use this url to cite or link to this publication:
author
supervisor
opponent
  • Professor Eastman, Ronald, Clark University, USA.
organization
publishing date
type
Thesis
publication status
published
subject
keywords
Change detection, Satellite sensor data, Time series analysis, Vegetation dynamics, Vegetation index
pages
160 pages
publisher
Department of Physical Geography and Ecosystem Science, Lund University
defense location
Världen auditorium, Geocentrum I
defense date
2014-10-24 10:00
ISBN
978-91-85793-42-6
language
English
LU publication?
yes
id
bd2c0c53-e64a-4f86-ba47-7e32b465e1c1 (old id 4689926)
date added to LUP
2014-10-02 12:28:23
date last changed
2016-09-19 08:45:13
@phdthesis{bd2c0c53-e64a-4f86-ba47-7e32b465e1c1,
  abstract     = {Abstract<br/><br>
<br/><br>
This thesis aims to advance the analysis of nonlinear trends in time series of vegetation data from Earth observation satellite sensors. This is accomplished by developing fast, efficient methods suitable for large volumes of data. A set of methods, tools, and a framework are developed and verified using data from regions containing vegetation change hotspots.<br/><br>
<br/><br>
First, a polynomial-fitting scheme is tested and applied to annual Global Inventory Modeling and Mapping Studies (GIMMS)–Normalized Difference Vegetation Index (NDVI) observations for North Africa for the period 1982–2006. Changes in annual observations are divided between linear and nonlinear (cubic, quadratic, and concealed) trend behaviors. A concealed trend is a nonlinear change which does not result in a net change in the amount of vegetation over the period.<br/><br>
<br/><br>
Second, a systematic comparison between parametric and non-parametric<br/><br>
techniques for analyzing trends in annual GIMMS-NDVI data is performed at<br/><br>
fifteen sites (located in Africa, Spain, Italy, and Iraq) to compare how trend type and departure from normality assumptions affect each method’s accuracy in detecting long-term change.<br/><br>
<br/><br>
Third, a user-friendly program (Detecting Breakpoints and Estimating Segments in Trend, DBEST) has been developed which generalizes vegetation trends to main features, and characterizes vegetation trend changes. The outputs of DBEST are the simplified trend, the change type (abrupt or non-abrupt), and estimates for the characteristics (time and magnitude) of the change. DBEST is tested and evaluated using both simulated NDVI data, and actual NDVI time series for Iraq for the period 1982-2006.<br/><br>
<br/><br>
Finally, a decision-making framework is presented to help analysts perform<br/><br>
comprehensive analyses of trend/change in time series of satellite sensor data. The framework is based on a conceptual model of the main aspects of trend analyses, including identification of the research question, the required data, the appropriate variables, and selection of efficient analysis methods. To verify the framework, it is applied to four case studies (located in Burkina Faso, Spain, Sweden, and Senegal) using Moderate-resolution Imaging Spectroradiometer (MODIS)–NDVI data for the period 2000–2013. Each of the case studies successfully achieved its<br/><br>
research aim(s), showing that the framework can achieve the main goal of the<br/><br>
study which is to advance the analysis of nonlinear changes in vegetation. <br/><br>
<br/><br>
The methods developed in this thesis can help to contribute more accurate information about vegetation dynamics to the field of land cover change research.},
  author       = {Jamali, Sadegh},
  isbn         = {978-91-85793-42-6},
  keyword      = {Change detection,Satellite sensor data,Time series analysis,Vegetation dynamics,Vegetation index},
  language     = {eng},
  pages        = {160},
  publisher    = {Department of Physical Geography and Ecosystem Science, Lund University},
  school       = {Lund University},
  title        = {Analyzing Vegetation Trends with Sensor Data from Earth Observation Satellites},
  year         = {2014},
}