Advanced

Automated temporal NDVI analysis over the Middle East for the period 1982 - 2010

Tomov, Hristo LU (2016) In Master Thesis in Geographical Information Science GISM01 20161
Dept of Physical Geography and Ecosystem Science
Abstract
The NDVI time-series consist of trend, season and noise. Changes in the season component are related to climate factors and they happen gradually over long period of time. The changes in the trend component are often due to human activities, fires and etc. This paper implements two algorithms (PolyTrend and DBEST) in R language, in order to examine the vegetation changes in the Middle East and to give more possibilities in the hands of the remote sensing communities. DBEST can analyse the gradual and the abrupt changes by decomposing the data, while PolyTrend classifies the inter-annual change between the picks of the green season.

PolyTrend and DBEST were adapted for R language environment. Two additional algorithms were developed to... (More)
The NDVI time-series consist of trend, season and noise. Changes in the season component are related to climate factors and they happen gradually over long period of time. The changes in the trend component are often due to human activities, fires and etc. This paper implements two algorithms (PolyTrend and DBEST) in R language, in order to examine the vegetation changes in the Middle East and to give more possibilities in the hands of the remote sensing communities. DBEST can analyse the gradual and the abrupt changes by decomposing the data, while PolyTrend classifies the inter-annual change between the picks of the green season.

PolyTrend and DBEST were adapted for R language environment. Two additional algorithms were developed to apply both algorithms over NDVI3g data set of the Middle East. A third algorithm discovered the affected land-cover through an overlay analysis by the use of the UMD land-cover classification data set.

PolyTrend showed linear (4%), quadratic (2%) and cubic (3%) trends. The different trend types were often found to be grouped in clusters. The largest clusters of trends were found near the south-eastern corner of the Arabian Peninsula and in the central regions of Saudi Arabia. More than 10% of all mixed forests were affected by these trends, most of which were in negative direction.

DBEST showed that 1% of the vegetation experienced a higher magnitude of change. Clusters of these changes were mainly located in the south-eastern and the western part of Turkey, the northern regions of Iraq and Syria, as well as along the coastlines of the Black Sea and the Caspian Sea. The changes were mainly related to the cropland and the grassland and were more in positive directions.

The study demonstrated the potential of PolyTrend and DBEST in R language for the remote sensing. It concludes that probably climatic factors affected the forests in Turkey and Iran. The high magnitude of changes of the cropland and grassland indicates that in some regions the agriculture expanded, while in others it declined. (Less)
Popular Abstract
The constant Earth observations from space allow monitoring of the vegetation changes on a regional scale. The changes in vegetation can be long and gradual (e.g. due to climatic factors) or more sudden and abrupt (e.g. due to fires, diseases and etc.). In order to estimate the changes of the vegetation, researchers use algorithms that decompose the observed data to seasonal, trend and remainder (e.g. noise). The algorithms that can distinguish these changes are of limited number, often hard to be accessed and most of the existing ones could be applied only to specific situations.

This paper implements two such algorithms (PolyTrend and DBEST) in R language, in order to give more possibilities in the hands of the remote sensing... (More)
The constant Earth observations from space allow monitoring of the vegetation changes on a regional scale. The changes in vegetation can be long and gradual (e.g. due to climatic factors) or more sudden and abrupt (e.g. due to fires, diseases and etc.). In order to estimate the changes of the vegetation, researchers use algorithms that decompose the observed data to seasonal, trend and remainder (e.g. noise). The algorithms that can distinguish these changes are of limited number, often hard to be accessed and most of the existing ones could be applied only to specific situations.

This paper implements two such algorithms (PolyTrend and DBEST) in R language, in order to give more possibilities in the hands of the remote sensing communities, and both are used to examine the vegetation changes in the Middle East. DBEST can analyse the gradual and the abrupt changes by decomposing the data, while PolyTrend classifies the inter-annual change between the picks of the green season.

PolyTrend and DBEST were re-coded and adapted for R language environment. Two other algorithms were developed to apply both algorithms over imagery data of the Middle East for the period between 1982 and 2010. A third algorithm related the results to a specific class of vegetation by comparing the results from the last two and a land-cover data set.

PolyTrend showed linear (4%), quadratic (2%) and cubic (3%) trends. The different trend types were often found to be grouped in clusters. The largest clusters of trends were found near the south-eastern corner of the Arabian Peninsula and in the central regions of Saudi Arabia. More than 10% of all mixed forests were affected by these trends, most of which were in negative direction.

DBEST showed that 1% of the vegetation experienced a higher magnitude of change. Clusters of these changes were mainly located in the south-eastern and the western part of Turkey, the northern regions of Iraq and Syria, as well as along the coastlines of the Black Sea and the Caspian Sea. The changes were mainly related to the cropland and the grassland and were more in positive directions.

The study demonstrated the potential of PolyTrend and DBEST in R language for the remote sensing. The obtained results showed that long gradual inter-annual changes affected the forests in Turkey and Iran. The reasons for these changes should be further investigated, but are probably related to climatic factors. The land-cover associated with high magnitude of more sudden changes was related to grassland and cropland. This leads to the suggestion that in some regions the agriculture expanded, while in others it declined. (Less)
Please use this url to cite or link to this publication:
author
Tomov, Hristo LU
supervisor
organization
alternative title
Detection of the vegetation changes in the Middle East between 1982 and 2010
course
GISM01 20161
year
type
H2 - Master's Degree (Two Years)
subject
keywords
trend analysis, NDVI Time series, satellite imagery, change detection, Physical Geography and Ecosystem analysis, GIS, PolyTrend, DBEST, R language, Middle East
publication/series
Master Thesis in Geographical Information Science
report number
49
language
English
id
8871893
date added to LUP
2016-05-03 12:39:29
date last changed
2016-05-03 12:39:29
@misc{8871893,
  abstract     = {The NDVI time-series consist of trend, season and noise. Changes in the season component are related to climate factors and they happen gradually over long period of time. The changes in the trend component are often due to human activities, fires and etc. This paper implements two algorithms (PolyTrend and DBEST) in R language, in order to examine the vegetation changes in the Middle East and to give more possibilities in the hands of the remote sensing communities. DBEST can analyse the gradual and the abrupt changes by decomposing the data, while PolyTrend classifies the inter-annual change between the picks of the green season.

PolyTrend and DBEST were adapted for R language environment. Two additional algorithms were developed to apply both algorithms over NDVI3g data set of the Middle East. A third algorithm discovered the affected land-cover through an overlay analysis by the use of the UMD land-cover classification data set.

PolyTrend showed linear (4%), quadratic (2%) and cubic (3%) trends. The different trend types were often found to be grouped in clusters. The largest clusters of trends were found near the south-eastern corner of the Arabian Peninsula and in the central regions of Saudi Arabia. More than 10% of all mixed forests were affected by these trends, most of which were in negative direction. 

DBEST showed that 1% of the vegetation experienced a higher magnitude of change. Clusters of these changes were mainly located in the south-eastern and the western part of Turkey, the northern regions of Iraq and Syria, as well as along the coastlines of the Black Sea and the Caspian Sea. The changes were mainly related to the cropland and the grassland and were more in positive directions. 

The study demonstrated the potential of PolyTrend and DBEST in R language for the remote sensing. It concludes that probably climatic factors affected the forests in Turkey and Iran. The high magnitude of changes of the cropland and grassland indicates that in some regions the agriculture expanded, while in others it declined.},
  author       = {Tomov, Hristo},
  keyword      = {trend analysis,NDVI Time series,satellite imagery,change detection,Physical Geography and Ecosystem analysis,GIS,PolyTrend,DBEST,R language,Middle East},
  language     = {eng},
  note         = {Student Paper},
  series       = {Master Thesis in Geographical Information Science},
  title        = {Automated temporal NDVI analysis over the Middle East for the period 1982 - 2010},
  year         = {2016},
}