ThesisPDF Available

Vegetation Response and Recovery in the 20 years following the 1982 eruption of El Chichón volcano: A Remote Sensing Approach

Authors:

Abstract and Figures

This study aims to use remote sensing to investigate vegetation response and recovery after the 1982 eruption of El Chichón volcano. In the past, studies into the effects of volcanic eruptions on vegetation have been limited in number and scope, due to the intensive field work required and the inaccessibility of many volcanoes. This study will therefore evaluate the effectiveness of remote sensing for this task. The study utilised five Landsat images acquired between 1980 and 2002. From these, normalised difference vegetation index (NDVI) was derived and image differencing was carried out to evaluate patterns of vegetation change over time. These results were analysed in terms of the eruption characteristics, in order to investigate the specific impact of different forms of volcanic activity, and regression analysis was carried out to identify the most important factors controlling vegetation recovery. The results showed that areas affected by both pyroclastic surges and tephra deposition experienced the greatest change in vegetation and took the longest time to recover. Areas affected by tephra deposition alone showed the least vegetation change and recovered within a period of a few years. The regression analysis found factors related to initial vegetation recovery to be important at first, but then factors relating to vegetation growth became more important over time. The results indicated that full vegetation coverage was reached within 20 years of the eruption, showing that recovery is rapid at El Chichón. Findings showed both similarities and differences with those from other volcanoes, highlighting the importance of carrying out individual case studies. Finally, the study highlighted some weaknesses of using remote sensing alone. Without ground observations, the study was unable to make inferences about absolute vegetation coverage or vegetation composition on the ground. Despite this, the methods used are useful for deriving spatial and temporal patterns of change, as well as investigating reasons for these, and so remote sensing should be seen as a valuable tool for further research in the field.
Content may be subject to copyright.
FACULTY OF SCIENCE,
ENGINEERING AND COMPUTING
School of Geography, Geology and the
Environment
MSc DEGREE
IN
Geographical Information Systems & Science
Name: Jennifer Rozier
ID Number: K1400667
Project Title: Vegetation Response and Recovery in the 20
years following the 1982 eruption of El Chichón volcano:
A Remote Sensing Approach
Date: 7th September 2015
Supervisor: Dr. Mike Smith
WARRANTY STATEMENT
This is a student project. Therefore, neither the student nor Kingston University makes any
warranty, express or implied, as to the accuracy of the data or conclusion of the work
performed in the project and will not be held responsible for any consequences arising out
of any inaccuracies or omissions therein.
1
ABSTRACT
This study aims to use remote sensing to investigate vegetation response and
recovery after the 1982 eruption of El Chichón volcano. In the past, studies
into the effects of volcanic eruptions on vegetation have been limited in
number and scope, due to the intensive field work required and the
inaccessibility of many volcanoes. This study will therefore evaluate the
effectiveness of remote sensing for this task. The study utilised five Landsat
images acquired between 1980 and 2002. From these, normalised difference
vegetation index (NDVI) was derived and image differencing was carried
out to evaluate patterns of vegetation change over time. These results were
analysed in terms of the eruption characteristics, in order to investigate the
specific impact of different forms of volcanic activity, and regression analysis
was carried out to identify the most important factors controlling vegetation
recovery. The results showed that areas affected by both pyroclastic surges
and tephra deposition experienced the greatest change in vegetation and
took the longest time to recover. Areas affected by tephra deposition alone
showed the least vegetation change and recovered within a period of a
few years. The regression analysis found factors related to initial vegetation
recovery to be important at first, but then factors relating to vegetation
growth became more important over time. The results indicated that full
vegetation coverage was reached within 20 years of the eruption, showing
that recovery is rapid at El Chichón. Findings showed both similarities and
differences with those from other volcanoes, highlighting the importance of
carrying out individual case studies. Finally, the study highlighted some
weaknesses of using remote sensing alone. Without ground observations, the
study was unable to make inferences about absolute vegetation coverage
or vegetation composition on the ground. Despite this, the methods used are
useful for deriving spatial and temporal patterns of change, as well as
investigating reasons for these, and so remote sensing should be seen as a
valuable tool for further research in the field.
2
CONTENTS
1 Introduction ....................................................................................................... 6
1.1 Overview ...................................................................................................... 6
1.2 Theoretical Background ............................................................................ 6
1.2.1 Volcanoes and Vegetation ............................................................... 6
1.2.2 Potential of Remote Sensing .............................................................. 7
1.2.3 El Chichón ............................................................................................. 8
1.3 Research Aims ............................................................................................ 9
2 Literature Review ............................................................................................ 11
2.1 Remote Sensing Principles ...................................................................... 11
2.2 Remote Sensing & Volcanology ............................................................ 11
2.2.1 Current Applications of Remote Sensing in Volcanology ........... 12
2.2.2 Utility of Current Applications .......................................................... 15
2.3 Studying the Impact of Volcanic Eruptions on Vegetation .............. 16
2.3.1 Ground-Based Studies ...................................................................... 16
2.3.2 Remote Sensing Studies .................................................................... 17
2.3.3 Utility of Remote Sensing for Vegetation Studies .......................... 18
2.4 Review Conclusions ................................................................................. 20
3 Research Methods ......................................................................................... 21
3.1 Study Area ................................................................................................. 21
3.1.1 Location and Background Information ......................................... 21
3.1.2 1982 Eruption of El Chichón ............................................................. 22
3.2 Image Analysis .......................................................................................... 24
3.2.1 Sources ................................................................................................ 24
3.2.2 Image Pre-Processing ....................................................................... 26
3.2.3 Image Analysis ................................................................................... 29
3
3.3 Ancillary Data ........................................................................................... 30
3.3.1 Direct Effects of the Eruption ........................................................... 31
3.3.2 Post-Eruption Ecological Driving Factors ........................................ 33
3.4 Statistical Analysis ..................................................................................... 37
3.4.1 Scatter Plots and Bar Graphs ........................................................... 37
3.4.2 Regression Analysis ............................................................................ 38
4 Results ............................................................................................................... 40
4.1 NDVI ............................................................................................................ 40
4.2 Change Detection ................................................................................... 43
4.3 Regression Analysis ................................................................................... 49
4.3.1 Tephra Only Zone .............................................................................. 49
4.3.2 Surges Only Zone ............................................................................... 50
4.3.3 Surges & Tephra Zone ....................................................................... 52
5 Discussion ......................................................................................................... 54
5.1 Initial Eruption Impacts ............................................................................ 54
5.2 Patterns of Revegetation ........................................................................ 55
5.3 Factors Influencing Revegetation ......................................................... 57
6 Conclusions ..................................................................................................... 60
7 References ....................................................................................................... 62
8 Appendices ..................................................................................................... 68
8.1 Appendix 1: False Colour Images .......................................................... 68
8.2 Appendix 2: Moran’s I Test for Spatial Autocorrelation ...................... 69
4
LIST OF FIGURES
Figure 2.1- Radiation absorption spectrum of a leaf ................................................... 18
Figure 2.2- Radiation reflectance spectrum of a leaf .................................................. 18
Figure 3.1- Study area location map. ............................................................................. 21
Figure 3.2- Tephra thickness isopach map .................................................................... 23
Figure 3.3- Map of pyroclastic flows and surges. .......................................................... 24
Figure 3.4- Flowchart of image pre-processing steps .................................................. 26
Figure 3.5- Map of number of surges affecting areas. ................................................ 31
Figure 3.6- Map of the three zones of interest. .............................................................. 32
Figure 3.7- DEM of the study area. .................................................................................. 33
Figure 3.8- Map showing aspect in the study area ...................................................... 34
Figure 3.9- Map showing slope of the study area. ....................................................... 35
Figure 3.10- Map showing distance to the summit crater ........................................... 36
Figure 3.11- Map showing distance to surviving vegetation. ..................................... 37
Figure 4.1- NDVI of the study area in 1980 ..................................................................... 40
Figure 4.2- NDVI of the study area in 1984 ..................................................................... 40
Figure 4.3- NDVI of the study area in 1990 ..................................................................... 42
Figure 4.4- NDVI of the study area in 1996 ..................................................................... 42
Figure 4.5- NDVI of the study area in 2002 ..................................................................... 43
Figure 4.6- NDVI change between 1980 and 1984 ...................................................... 43
Figure 4.7- NDVI change between 1984 and 1990 ...................................................... 44
Figure 4.8- NDVI change between 1990 and 1996 ...................................................... 44
Figure 4.9- NDVI change between 1996 and 2002 ...................................................... 45
Figure 4.10- Mean NDVI over time by zone ................................................................... 46
Figure 4.11- Mean NDVI difference between images by zone .................................. 46
Figure 4.12- Mean NDVI by tephra thickness ('Tephra Only' zone) ........................... 47
Figure 4.13- Mean NDVI difference by tephra thickness (‘Tephra Only’ zone) ........ 47
Figure 4.14- Mean NDVI by tephra thickness (‘Surges & Tephra’ zone) .................... 47
Figure 4.15- Mean NDVI difference by tephra thickness (‘Surges & Tephra’ zone) . 47
Figure 4.16- Mean NDVI by number of surges (‘Surges Only’ zone) .......................... 48
Figure 4.17- Mean NDVI difference by number of surges (‘Surges Only’ zone) ....... 48
Figure 4.18- Mean NDVI by number of surges (‘Surges & Tephra’ zone) .................. 48
Figure 4.19- Mean NDVI difference by number of surges (‘Surges & Tephra’ zone)48
5
LIST OF TABLES
Table 3.1- Band designations for Landsat MSS .............................................................. 25
Table 3.2 - Band designations for Landsat ETM+........................................................... 25
Table 3.3 - Band designations for Landsat TM ............................................................... 25
Table 3.4 - Image acquisition dates and characteristics ............................................ 26
Table 3.5 - Values used for Dark Object Subtraction ................................................... 28
Table 3.6 - RMS errors for geometric correction ............................................................ 29
Table 4.1 - First regression results for the ‘Tephra Only’ zone ...................................... 49
Table 4.2 - Second regression results for the ‘Tephra Only’ zone ............................... 50
Table 4.3 - First regression results for the ‘Surges Only’ zone ....................................... 50
Table 4.4 - Second regression results for the ‘Surges Only’ zone ................................ 51
Table 4.5 - First regression results for the ‘Surges & Tephra’ zone ............................... 52
Table 4.6 - Second regression results for the ‘Surges & Tephra’ zone ........................ 53
LIST OF EQUATIONS
Equation 1 - NDVI calculation ......................................................................................... 19
Equation 2 - Conversion of DNs to radiance ................................................................. 27
Equation 3 - Conversion radiance to TOA reflectance ............................................... 27
Equation 4 - Spatial Error Model Regression equation ................................................. 38
Equation 5 - Standardisation of variables equation ..................................................... 39
Equation 6 - Regression equation for 1984-1990 in the ‘Tephra Only’ zone ............. 49
Equation 7 - Regression equation for 1990-1996 in the ‘Tephra Only’ zone ............. 49
Equation 8 - Regression equation for 1996-2002 in the ‘Tephra Only’ zone ............. 49
Equation 9 - Regression equation for 1984-1990 in the ‘Surges Only’ zone .............. 51
Equation 10- Regression equation for 1990-1996 in the ‘Surges Only’ zone ............. 51
Equation 11 - Regression equation for 1996-2002 in the ‘Surges Only’ zone ............ 51
Equation 12 - Regression equation for 1980-1984 in the ‘Surges & Tephra’ zone .... 52
Equation 13 - Regression equation for 1984-1990 in the ‘Surges & Tephra’ zone .... 52
Equation 14 - Regression equation for 1990-1996 in the ‘Surges & Tephra’ zone .... 52
Equation 15 - Regression equation for 1996-2002 in the ‘Surges & Tephra’ zone .... 53
6
1 INTRODUCTION
1.1 OVERVIEW
Volcanic activity can add large volumes of ejected materials and elements
to the atmosphere and environment that can have dramatic effects on
vegetation (Dale et al, 2005). However, because of the large number of
potentially active volcanoes worldwide (estimated to be around 1500)
(USGS, 2015a), the specific impacts of eruptions on vegetation has remained
largely unstudied at most locations. With 10% of the world’s population living
on volcanic soils and potentially using volcanic soils as agricultural land (Floor
et al, 2011), understanding the impacts that eruptions could have on local
vegetation is becoming increasingly important. This study aims to utilise
remote sensing to examine the specific example of vegetation response and
recovery following the 1982 eruption of El Chichón, Mexico.
1.2 THEORETICAL BACKGROUND
1.2.1 Volcanoes and Vegetation
The effects that volcanic activity can have on vegetation are highly
variable, depending upon the type of activity, distance from the summit,
and chemical composition of ejected material (Dale et al, 2005). For
relatively small eruptions, the input of volcanic material may improve soil
fertility and vegetation health through the addition of elements present in
ash, such as calcium and magnesium, which are beneficial for plant growth
(Benson, 2005). As such, soil can be rejuvenated where ashfall amounts to
less than a few centimetres, as this is not enough to cause significant
damage to existing vegetation (Lockwood & Hazlett, 2007). Conversely,
larger eruptions can be devastating, potentially wiping out all vegetation in
the local area and creating a barren volcanic desert (Oppenheimer, 2011).
Thick tephra deposits may bury and kill vegetation, potentially taking
7
decades to recover; while hot lava flows can have an even more severe
effect, burying and burning vegetation and potentially taking centuries for
full recovery (Dale et al, 2005). Differences in recovery times have to do with
the degree of vegetation damage. For large eruptions, completely new
volcanic surfaces are created on the ejected material and primary
succession is therefore the dominant processes of recovery. Whereas, where
some biota and soil are able to survive under thin deposits, secondary
succession may occur, taking less time to reach full recovery (del Moral &
Grishin, 1999).
As well as eruption style, vegetation response is related to a multitude of
environmental factors and local conditions (Oppenheimer, 2011). These
include: climate (temperature and precipitation), through its effects on
growing conditions and weathering rates (Oppenheimer, 2011); type of
vegetation present (for instance, some types of plant are better able to
withstand the effects of eruptions, such as shorter plants protected by
topography) (del Moral & Grishin, 1999); rate of erosion (Inbar et al, 2001);
the presence of birds and animals which may distribute seeds and mix soil to
the surface (Oppenheimer, 2011); distance to surviving vegetation and
influence of human management (Lawrence, 2006); and chance factors,
such as the time of year at which the eruption occurs. Because of the
number of varied and interwoven influences, the pattern of vegetation
damage and recovery following a particular eruption will be very difficult to
predict, highlighting the importance of individual case studies (del Moral &
Grishin, 1999).
1.2.2 Potential of Remote Sensing
The kind of ground-based field work that is necessary to examine vegetation
response to volcanic eruptions and long-term patterns of recovery can be
extremely time-consuming and expensive. Remote sensing, however, is
8
commonly applied to vegetation mapping and monitoring (Skidmore, 2003),
and in recent decades, the use of remote sensing within the field of
volcanology has increased dramatically, providing valuable information for
both hazard monitoring and furthering of scientific understanding
(Oppenheimer, 1998). As such, remote sensing offers the possibility to study
both the eruptive behaviour of volcanoes and the impact on vegetation. As
well as this, remote sensing offers the unique ability to carry out wide area
studies, from a distance, with the possibility of studying volcanoes in
dangerous and remote locations. Moreover, while ground studies will be
limited to collecting samples in the present, remotely sensed imagery is now
available spanning decades, providing the ability to monitor how
vegetation changes and adapts to volcanic activity over time (Lillesand et
al, 2008).
1.2.3 El Chichón
The 1982 eruption of El Chichón was one of the largest volcanic events of the
20th Century, causing an estimated $55million of damage in Mexico (Global
Volcanism Program, 1982) and costing over 2000 people their lives (Bonasia
et al, 2012). The eruption has come to be known for the large quantity of
sulphur dioxide it released into the atmosphere, causing global climate
change (Sigurdsson et al, 1987), but it also had significant local impacts.
Devastating pyroclastic surges and flows, which are extremely hot, fast-
moving and high-density mixtures of gases and rock fragments (USGS,
2014a), were an exceptional feature of the eruption. The surges destroyed
nine villages and carbonised vegetation (Sigurdsson et al, 1987), while the
flows dammed a river, generating a destructive flood when it later failed
(Global Volcanism Program, 1982). As well as this, the eruption generated
large quantities of tephra (rock and ash fragments) that blanketed the
landscape up to hundreds of kilometres from the vent (Inbar et al, 2001). The
eruption has been extremely well studied, with research constraining tephra
9
deposition (e.g. Carey & Sigurdsson, 1986; Bonasia et al, 2012), pyroclastic
surge and flow deposits (e.g. Sigurdsson et al, 1987; Scolamacchia & Macías,
2005), and the distribution and effects of the aerosol plume (e.g. Kerr, 1983;
Robock & Matson, 1983). However, although some studies have
incorporated observations of vegetation response to the eruption (e.g. Inbar
et al, 2001), no detailed and long-term evaluations have been carried out,
and none have utilised remote sensing.
1.3 RESEARCH AIMS
Given the difficulty of predicting vegetation response to eruptions at
individual volcanoes, and the relative lack of existing case studies, this study
aims to investigate vegetation response and recovery following the 1982
eruption of El Chichón using remote sensing techniques. The study will use a
vegetation index derived from Landsat satellite imagery to measure
vegetation immediately before the eruption and to quantify vegetation
change over the 20 years following the eruption. Eruption parameters
derived from other studies and a series of datasets relating to ecological
driving factors will also be used to draw conclusions about the relative
impact of different kinds of eruptive activity and the most important factors
controlling vegetation recovery. Throughout, the results will be examined to
identify spatial and temporal patterns of change, and where possible,
findings will be compared to other studies of El Chichón and similar studies
at other volcanoes. The study therefore has three main aims:
(1) To investigate the immediate effect of the eruption on surrounding
vegetation
(2) To investigate the spatial pattern and rates of revegetation in the 20
years following the eruption
(3) To investigate the factors influencing vegetation recovery
10
As well as remote sensing, the study will also make use of geographical
information systems (GIS) in order to combine datasets for analysis and
present maps and results. Finally, the utility of remote sensing and the
methods used will be evaluated and conclusions will be drawn about the
applicability of the methods and results to other volcanoes.
11
2 LITERATURE REVIEW
2.1 REMOTE SENSING PRINCIPLES
Remote sensing is the process of obtaining information about an object or
an area using a device not in contact with it (Lillesand et al, 2008). One of
the most widely used forms is the measurement of electromagnetic energy
emanating from the Earth’s surface. Sensors mounted on a platform above
the Earth’s surface, such as a satellite or aeroplane, can be used to record
the electromagnetic energy emanating at different wavelengths (Boyd,
2007). This form of remote sensing is based upon the premise that objects all
reflect, absorb and emit electromagnetic energy in different ways,
depending on the composition of the feature, as well as the wavelength of
energy. As such, “spectral signatures” can be detected in remotely sensed
imagery which allow the characterisation of features, based upon their
spectral reflectance (Lillesand et al, 2008). Over the past few decades, use
of remote sensing has increased across many areas of science, government
and business (Lillesand et al, 2008), providing the unique ability to collect
data over a spatial and temporal scale that is difficult to achieve through
ground-based data collection methods (Boyd, 2000).
2.2 REMOTE SENSING & VOLCANOLOGY
Volcanology is one such field that has made increased use of remote sensing
in recent decades (Oppenheimer, 1998), with Tralli et al (2005) even going
so far as to claim that remote sensing is “defining a new paradigm for
volcanological observations” (pg. 190). In volcanology, the dominant form
of remote sensing used is spaceborne satellite remote sensing of
electromagnetic energy, but airborne remote sensing is also often used
(Head, Maclean & Carn, 2012). Various aspects of volcanic activity, such as
lava and plume emissions, thermal anomalies and ground deformation, all
have spectral signatures that can be detected (Tralli et al, 2005). Therefore,
12
remote sensing provides a powerful tool for volcano monitoring, particularly
for volcanoes that are located in remote areas or countries with limited
financial resources, or at volcanoes where ground-based monitoring is
unsafe due to frequent activity (Oppenheimer, 1998).
2.2.1 Current Applications of Remote Sensing in Volcanology
Over the past two decades, the number of published studies applying
remote sensing to volcanoes has been increasing steadily (Pyle et al, 2013)
and potential applications are multiplying due to advancements in
technology and the Internet providing vast amounts of low-cost and easily
obtainable data (Carn & Oppenheimer, 2000). However, remote sensing
among volcano observatories is still far from universal, with many volcanoes
still going unmonitored and unstudied. As well as this, the range of potential
applications is still developing alongside the technology itself, meaning that
there are some application areas that remain under-exploited (Pyle et al,
2013). Despite these challenges, remote sensing provides valuable
information for hazard monitoring and furthering scientific understanding of
eruptions (Oppenheimer, 1998). The mapping of lava flows, the detection of
thermal anomalies, and the study of plume dispersion and composition have
become the main application areas.
2.2.1.1 Lava Flow Mapping
Maps of past lava flows can be used to estimate future flow extents, study
magma-supply dynamics, and update geologic maps (Lu et al, 2004).
However, lava mapping from field studies is often expensive and time-
consuming, and can be impossible at inaccessible volcanoes (Lu et al, 2004;
Head, Maclean & Carn, 2012). Therefore, use of remote sensing has
increased and lava mapping studies have been carried out at a number of
volcanoes for many years, albeit without the development of a standard
methodology (Zarah et al, 2008). Some studies use optical data in the visible,
13
near-infrared and thermal portions of the electromagnetic spectrum to map
lava flows based on differences in spectral signatures that emerge as flows
weather with age and revegetation occurs (Lu et al, 2004). For example,
Head, Maclean & Carn’s (2012) study of Nyamuragira volcano using Landsat
imagery, and Abrams et al’s (1991, 1996) studies of Mauna Loa and Mount
Etna using NASA’s NS-001 and TMS scanners, respectively. Conversely,
mapping based on radar remote sensing (such as Carn, 1999; Colclough,
2005; and Dean et al, 2002) identifies differences in the degree of radar
backscatter caused by weathering and vegetation growth over time (Carn,
1999). In recent years, studies have utilised both optical and radar imagery
in tandem, acknowledging the shortfalls of each alone. For instance Lu et
al’s (2004) study of Westdahl volcano and Smets et al’s (2010) study of
Nyamuragira have proven much more successful at distinguishing flows
along their entire length; using radar where optical imagery is obscured by
cloud or struggles to distinguish between new flows emplaced upon other
recent ones (Head, Maclean & Carn, 2012), and utilising optical imagery
where radar has difficulty distinguishing the edges of distal flows (Lu et al,
2004). Commonly, such studies utilise manual digitisation to detect flows
(Joyce et al, 2009), but some attempts to apply unsupervised (e.g. Head,
Maclean & Carn, 2012) and supervised (e.g. Abrams et al, 1996)
classification techniques have been attempted.
2.2.1.2 Thermal Surveillance
Detecting volcanic thermal signals has been one of the most common
volcanological applications of remote sensing since the 1980s (Joyce et al,
2009). Thermal anomalies such as increases in temperature at fumaroles or
detection of “hotspots” under volcanoes, can indicate heat transfer from
the interior of the Earth and eruption likelihood (Oppenheimer, 1998).
Thermal surveillance can also be used to track eruptive materials, such as
lava flows, and activate “thermal alerts” for eruption onsets (Rose & Kimberly,
14
1995). The development of more sophisticated thermal sensors in recent
years has led to increased applications (Joyce et al, 2009). Comparisons with
background temperature is the most common means of detecting
temperature anomalies, using shortwave and mid-Infrared channels
(Oppenheimer, 1998). Complications in detection can arise because the
temperature of the feature, or anomalous surrounding features, may
saturate the sensor if too high. Moreover, detection of thermal features such
as lava flows, lakes and domes can be difficult if sensors have an
instantaneous field of view (IFOV) coarser than 1km2 (Oppenheimer, 1998).
Joyce et al (2009) therefore warns that thermal alerts, which may be freely
available on the Internet, should be viewed with caution. Regardless,
applications of thermal surveillance using remote sensing is extremely useful
at volcanoes where ground-based monitoring is not available. For example,
Carn & Oppenheimer (2000) used AVHRR data to detect lava flows and
small episodes of explosive activity at Indonesian volcanoes, providing
evidence for cyclic behaviour and aiding hazard prediction.
2.2.1.3 Plume Surveillance
The third major volcanological application of remote sensing is that of plume
surveillance. Most commonly, research has focused upon plumes associated
with large explosive eruptions (Oppenheimer, 1998). Detecting and tracking
plumes is important for hazard mitigation, in terms of risk to aviation,
infrastructure stability, and human and environmental health (Joyce et al,
2009). Typically, plumes are distinguished by their characteristic wedge-
shape (Oppenheimer, 1998) and unique spectral characteristics. Prata et al
(1989) found that volcanic clouds exhibit lower reflectance in AVHRR band
4 than band 5, while meteorological clouds display the opposite. Robock &
Matson (1983) successfully used these methods to track the plume of the
1982 El Chichón eruption as it encircled the globe in 3 months, and Glaze et
al (1989) identified two separate pulses of explosive activity during the 1986
15
eruption of Láscar, Chile, from the beaded shape of the plume. However,
the huge variety of eruption styles means that spectral signatures and shapes
of plumes can vary enormously, and for less powerful eruptions, plumes can
be very difficult to distinguish from meteorological clouds (Oppenheimer,
1998). This means that many plumes go unmonitored, although the advent
of higher resolution sensors, such as ASTER, has improved detection
capabilities for smaller plumes (Joyce et al, 2009). Greater characterisation
of the constant, yet barely visible, plumes of quiescent volcanoes is gaining
increasing attention in the literature (Joyce et al, 2009). For instance,
volcanic sulphur dioxide can be measured by examining its absorption
effects of ultraviolet radiation (Krueger, 1983).
2.2.2 Utility of Current Applications
Numerous examples of real-world applications of the findings of
volcanological remote sensing research can be identified in the literature.
Hazard maps are often created by combining remotely sensed data with
data on local populations and infrastructure within GIS, as this provides a
powerful tool for combining the raster layers of remotely sensed imagery with
other spatial layers (Pareschi et al, 2000). The USGS, for instance, use ESRI’s
ArcGIS to forecast volcanic hazards, produce hazard maps and create
evacuation plans for dangerous volcanoes, such as Merapi, Indonesia
(Griswold, 2009). These maps can be created by directly inputting the results
of the remotely sensed analysis, or by first using the data in mathematical
models to simulate potential eruption scenarios (Pareschi et al, 2000). One
common form are models predicting lava flow paths based on the
assumption that they will follow the maximum slope. Using digital elevation
models (DEMs) to obtain slope values and remotely sensed data on known
lava flow paths to constrain speed and distance parameters, prediction
models can be created (Tralli et al, 2005). More complex models are also
being developed, such as Felpato’s (2007) GIS-based modelling package
16
for predicting lava flow extent, ash fallout, and pyroclastic density currents.
Complex mathematical models like this are difficult to create, as they require
complex algorithms and extensive ground-truth data alongside the remotely
sensed data (Rose & Kimberly, 1995). However, the development of such
models is gaining increasing ground because the actual processes that
cause volcanic eruptions are largely unobservable.
2.3 STUDYING THE IMPACT OF VOLCANIC ERUPTIONS ON VEGETATION
Remote sensing could offer a valuable tool for furthering research in this
area, however, as Oppenheimer noted in 1998, there are very few published
studies that make use of these capabilities. Examination of the literature to
date reveals that this is still an underdeveloped field of research.
2.3.1 Ground-Based Studies
Examining landscape and ecology recovery after volcanic eruptions has
been the subject of much work across the fields of volcanology, biology and
environmental studies (Oppenheimer, 2011). Such studies have been
important for identifying the factors that are important for vegetation
recovery (detailed in section 1.2.1). Individual studies will not be examined in
detail, but it is important to note that they tend to be relatively limited in
scope to volcanoes that are easily accessible, for instance many studies
have been carried out at Mount St. Helens in the USA (e.g. del Moral 1983,
1993; Halpern & Harmon, 1983; and Dale, 1989, 1991; among others). They
tend to also be limited in scale to specific plant species or the effect of
individual elements, such as Martin et al’s (2009) detailed study of sweet
chestnut trees or Floor et al’s (2011) study of selenium mobilisation, both at
Mount Etna. The limited nature of such studies is in large part due to the
intensive, time-consuming and costly nature of the fieldwork required.
17
2.3.2 Remote Sensing Studies
Remote sensing studies, on the other hand, have been limited in number or
have only briefly touched upon the topic while undertaking other
investigations. For instance, lava flow mapping studies, such as Head,
Maclean & Carn (2012), utilise the spectral signatures of revegetation to
distinguish flows, but do not specifically investigate the spatial or temporal
patterns of revegetation. Other studies have simply looked at vegetation
impact in terms of the area of crop fields or forests destroyed (e.g. Grishin et
al, 2011; Yulianto et al, 2013). There have been a few studies, however, which
have made more extensive use of remotely sensed data, predominantly
using the normalised difference vegetation index (NDVI: see section 2.3.3.1).
One such study is Lawrence’s (2006) temporal study of vegetation response
at Mount St. Helens in the 20 years following its 1980 eruption. Lawrence
(2006) measured the NDVI of vegetation surrounding the volcano and
observed how it changed over time in relation to various factors, including
the type of volcanic hazard they were initially exposed to, elevation,
distance to the summit and human intervention. Similarly, De Rose et al
(2011) used NDVI derived from ASTER imagery in the 10-16years following the
1991 eruption of Mount Pinatubo to map patterns of NDVI change around
the volcano, estimate a trajectory of vegetation recovery and identify
factors that influenced vegetation recovery. De Schutter et al (2015) also
used remote sensing and the NDVI to delineate the area most affected by
ash fallout following the 2007-08 eruption of Oldoinyo Lengai and to examine
whether recovery correlated with ash thickness. These, and the few other
studies in this area (e.g. Takahata & Miyama, 1980; Xu et al, 2009), highlight
that techniques are being developed to make use of remote sensing for
these kinds of studies, but they remain far from widely applied.
18
2.3.3 Utility of Remote Sensing for Vegetation Studies
Multispectral imagery is particularly suited for vegetation studies because
plants reflect different wavelengths of radiation to varying degrees. As Figure
2.1and Figure 2.2 show, blue and red light are preferentially absorbed for use
in photosynthesis, while green light and, to an even greater extent, near-
infrared radiation are scattered and reflected more by the mesophyll tissue
and cavities within the leaf (Campbell & Wynne, 2011). Other surface
features do not exhibit the same spectral response. For instance, soils tend
to have low reflectance in all portions of the spectrum, with a slight increase
in the near-infrared, while water has peak reflectance in the visible portion,
and low in the near-infrared (Boyd, 2007). Vegetation will therefore appear
much brighter in the near-infrared portion of the spectrum than surrounding
areas in multispectral imagery (Campbell & Wynne, 2007). Moreover, it is
even possible to differentiate between types of vegetation as different
plants display different spectral properties due to factors such as water
content, leaf area, and stress (Boyd, 2000). Many methods have been built
around these spectral characteristics of vegetation in order to use remote
sensing to map and monitor vegetation, with vegetation indices and
change detection being amongst the most common.
Figure 2.1- Electromagnetic radiation
absorption spectrum of a typical leaf. Source:
Campbell & Wynne (2011; figure 17.3, pg. 471)
Figure 2.2- Electromagnetic radiation
reflectance spectrum of a typical leaf.
Source: Campbell & Wynn (2011; figure 17.4,
pg. 473)
19
2.3.3.1 NDVI
Numerous vegetation indices have been developed which combine several
spectral values and yield a single result relating to biomass and vegetation
vigour within an image. One of the simplest and arguably the most widely
used is the NDVI (Campbell & Wynne, 2011): a simple quotient between red
and near-infrared reflectance, designed to highlight the difference between
these bands (Liu & Mason, 2009). The NDVI equation (Equation 1) yields a
value between -1 and 1, with high positive values indicating healthy and
dense vegetation, due to the large difference between red and near-
infrared reflectance this exhibits. Negative values are associated with
clouds, water and snow because they have greater reflectance in the visible
spectrum, and bare soil has an NDVI value near 0 due to similar reflectances
in the red and near-infrared (Singh et al, 2003). The NDVI has been found to
be highly correlated with numerous vegetation parameters (e.g. vegetation
cover, biomass and leaf area) and therefore forms an integral part of many
studies aiming to map and monitor vegetation where ground data is
unavailable (e.g. Singh et al, 2003; Lawrence, 2006; De Rose et al, 2011; De
Schutter et al, 2015).
2.3.3.2 Change Detection
The procedure of change detection can be applied to conduct
multitemporal studies and gain important information about vegetation
change over time (Skidmore, 2003). The term “change detection”
encompasses a wide variety of techniques which vary in complexity and
   
  
Where:
NIR = spectral reflectance measurement in the near-infrared region
RED = spectral reflectance measurement in the red region
Equation 1 NDVI calculation
20
applicability. Amongst the simplest are algebraic methods: simply
differencing one image from another. This can be performed on the raw
imagery, on vegetation indices, or on already classified images, providing a
simple indicator of change over time. More complex techniques, such
Principal Components Analysis (PCA) and Tasselled Cap, tend to involve
complex calculations and transformations of the data, and are more widely
applied to large datasets because they reduce data redundancy (Liu &
Mason, 2009). Change detection forms a very important part of vegetation
studies, and is especially important in studies of vegetation regrowth after
volcanic eruptions, because of the unpredictability of the process
(Oppenheimer, 2011). For example, Lawrence (2006) utilised image
differencing between a series of NDVI images in order to document patterns
of regrowth at Mount St. Helens.
2.4 REVIEW CONCLUSIONS
This literature review has discussed some of the current volcanological
applications of remote sensing and the gap in the literature with regard to
utilising remote sensing to study the effects of volcanic eruptions on
vegetation. It has highlighted the current ways in which remote sensing is
used to map and monitor vegetation change and pointed the way towards
applying these methods within the field of volcanology.
21
3 RESEARCH METHODS
3.1 STUDY AREA
3.1.1 Location and Background Information
El Chichón is a small, but powerful, volcano located in the northwest of the
State of Chiapas, Mexico (Figure 3.1) (Smithsonian Institution, 2013). It consists
of a symmetrical andesitic tuff cone and lava dome complex, covering an
area of approximately 10km in diameter at its base and extending to 1150m
above sea level (Smithsonian Institution, 2013; Volcano Discovery, 2015).
3.1.1.1 Climate and Vegetation
El Chichón’s climate is humid, tropical and wet. Rains are experienced year-
round, exceeding totals of 4000mm annually, with peaks in the rainy season
(June October) (Inbar et al, 2001). Temperatures are fairly constant
Figure 3.1- The true-colour Landsat ETM+ image, acquired in 2002, and inset map show the location
of El Chichón volcano. The image shows a 300km2 study area around the volcano, with the summit
and town and villages highlighted.
22
throughout the year, but vary with elevation on the volcano’s flanks (Nations
Encyclopedia, 2015). Before the 1982 eruption, large parts of the natural
tropical vegetation had been cleared, with just 17% of the land in the region
remaining as tropical forest in 1970. The majority of the land (80%) was under
pasture for cattle, while crops of mainly coffee, cocoa and beans made up
1.5% of land cover (Cervantes-Borja et al, 1983 cited in Inbar et al, 2001).
3.1.1.2 Eruptive History
Before 1982, El Chichón was relatively unknown; with no record of major
eruptive activity since ca.1360 (Smithsonian Institution, 2013). Examination of
the stratigraphic record revealed that at least 11 explosive eruptions
occurred at the volcano in the 8000 years prior to 1982 (Scolamacchia &
Macías, 2005; Macías et al, 2008; Smithsonian Institution, 2013). The 1360
event was given a rating of 5 on the Volcanic Explosivity Index (VEI)
(Smithsonian Institution, 2013): a logarithmic index for measuring the size of
volcanic eruptions, where a 1-level increase indicates a 10-times increase in
amount of material ejected and a score of 5 indicates that at least 1km3 of
material was ejected (USGS, 2009).
3.1.2 1982 Eruption of El Chichón
The eruption began on the 29th March 1982 and lasted 7 days, during which
time three large, explosive (‘Plinian’) eruptions occurred, at: 0532
Coordinated Universal Time (UTC) on 29th March, 0135 UTC on 4th April, and
1122 UTC 4th April (Rose & Durant, 2009). The second event produced the
largest eruption column, with an estimated height of 32km and a peak mass
eruption rate of 1.9x108 kg/s. The first and third events are estimated to have
had column heights of 27km and 29km, and peak mass eruption rates of
1.1x108 kg/s and 1.3x108 kg/s, respectively (Carey & Sigurdsson, 1986). In all
three events, the eruption cloud reached the stratosphere, causing
bidirectional plume dispersal: in the east-northeast direction in the
23
troposphere and the west-southwest direction in the stratosphere (Carey &
Sigurdsson, 1986). The eruption received a VEI score of 5 (Smithsonian
Instituion, 2013) and ejected a total estimated 1.1km3 of material (dense rock
equivalent) (Macías et al, 2003).
3.1.2.1 Tephra Deposits
Three distinct layers of tephra were deposited, each associated with one
phase of the eruption (Rose & Durant, 2009). Combined, tephra fallout
covered an estimated area of approximately 50,000km2, extending mostly in
the east due to deposition from the tropospheric plume (Varekamp et al,
1984). Although fine ash was deposited hundreds of kilometres from the
volcano, the furthest deposits were thin and therefore quickly removed by
rain and wind. As such, no major damage or landform effects were noticed
in these areas (Inbar et al, 2001). It was therefore decided to limit this analysis
to the areas with the thickest deposits (>5cm) (Figure 3.2).
Figure 3.2- Map shows tephra isopach lines for the 1982 eruption of El Chichón volcano, up
to 5cm in ash thickness (after Inbar et al, 2001), displayed over a true-colour Landsat ETM+
image acquired in 2002.
24
3.1.2.2 Pyroclastic Surge and Flow Deposits
Three devastating pyroclastic surges were generated, each associated with
one phase of the eruption. They covered areas of 93km2, 104km2 and 39km2
for surges ‘S-1’, ‘S-2’ and ‘S-3’, respectively. Minor pyroclastic flows
accompanied the first and second eruptions (Sigurdsson et al, 1987). The
distribution of these can be seen in Figure 3.3.
3.2 IMAGE ANALYSIS
3.2.1 Sources
The study utilised Landsat Multispectral Scanner (MSS), Landsat Thematic
Mapper (TM) and Landsat Enhanced Thematic Mapper (ETM+) imagery,
which is free to download from the USGS (2015b) ‘EarthExplorer’ website.
Band designations for these sensors are given in Table 3.1, Table 3.3 and
Figure 3.3- Map shows the locations of pyroclastic flow and surge deposits from the 1982
eruption of El Chichón volcano (after Sigurdsson et al, 1987), displayed over a true-colour
Landsat ETM+ image acquired in 2002.
25
Table 3.2. Landsat imagery dates back to 1972 and has a recurrence cycle
of approximately 16 days, making it extremely useful for long-term studies
(Bruce & Hilbert, 2006).
Landsat 4-5
Wavelength
(micrometres)
Band 1
0.5-0.6
Band 2
0.6-0.7
Band 3
0.7-0.8
Band 4
0.8-1.1
Table 3.1- Band designations for Landsat MSS (USGS, 2014b)
Landsat 4-5
Wavelength
(micrometres)
Band 1
0.45-0.52
Band 2
0.52-0.60
Band 3
0.63-0.69
Band 4
0.76-0.90
Band 5
1.55-1.75
Band 6
10.40-12.50
Band 7
2.08-2.35
Table 3.3- Band designations for Landsat TM (USGS,
2014b)
Five images were downloaded for this study (Table 3.4). The 1980 and 1984
images are the closest available before and after the eruption. Vegetation
is unlikely to have changed significantly in the two years before the eruption,
so the pre-eruption image is not problematic. However, significant amounts
of change may have occurred in the two years post-eruption that will go
unobserved. For the remaining images, attempts were made to pick images
equally spaced apart and acquired at the same time of year, to minimise
distortion due to differences in sun elevation and azimuth. Cloud-free
January images are available for 1984, 1996 and 2002, but for 1990, the
closest in terms of time of the year is a December image. This means that
although the images are spaced at 6-year intervals, there are an unequal
number of growing seasons in each period. An image from March 1990 was
Landsat 7
Wavelength
(micrometres)
Band 1
0.45-0.52
Band 2
0.52-0.60
Band 3
0.63-0.69
Band 4
0.76-0.90
Band 5
1.55-1.75
Band 6
10.40-12.50
Band 7
2.09-2.35
Band 8
0.52-0.90
Table 3.2- Band designations for Landsat
ETM+ (USGS, 2014b)
26
also available, but it was considered more important to minimise the effects
of the time of year of acquisition.
3.2.2 Image Pre-Processing
The flowchart in Figure 3.4 outlines the pre-processing stages used in this
study to correct the images prior to analysis.
3.2.2.1 Conversion of DNs to TOA Reflectance
Reflectance recorded by the sensor is scaled to integer digital numbers
(DNs) between 0 and 255 within Landsat images. DNs are therefore only
useful for expressing relative brightness between pixels in a single image, not
between images of different dates (Campbell & Wynne, 2011). To normalise
the scale of measurement across multitemporal imagery, DNs must be
converted to Top of Atmosphere (TOA) Reflectance units (Bruce & Hilbert,
Image
Acquisition
Date
Sensor
Resolution (m)
Path/Row
Sun Azimuth
(o)
Nov 17 1980
Landsat 3 MSS
60
023/048
136
Jan 09 1984
Landsat 4 MSS
60
022/048
138.6528466
Dec 19 1990
Landsat 5 MSS
60
022/048
139.9567095
Jan 02 1996
Landsat 5 TM
30
022/048
134.7699485
Jan 10 2002
Landsat 7 ETM+
30
022/048
142.965432
Table 3.4- Image acquisition dates and characteristics
Figure 3.4- Flowchart showing the image pre-processing steps
27
2006). ERDAS Imagine’s ‘Model Maker’ and Equation 2 and Equation 3 were
used for this purpose (Campbell & Wynne, 2011). The first equation converts
DNs to radiance, then the second converts radiance to TOA reflectance. For
Equation 2 the values for , ,  and  are supplied in the
header file of each Landsat image. For Equation 3 both  and can be
found in Chander et al’s (2009) summary of radiometric calibration
coefficients, while is obtained by taking the sun elevation angle found in
the image header file away from 90o (and converting to radians for use in
ERDAS Imagine).
   
  
Where: 2 sr m)
 = the quantized calibrated pixel value (DN)
 = the minimum quantized calibrated pixel value corresponding to 
 = the maximum quantized calibrated pixel value corresponding to 
 = the spectral at-sensor radiance that is scaled to 
 = the spectral at-sensor radiance that is scaled to 


Where: = at sensor, unitless in-band reflectance
= at-sensor, in-band radiance
 = the band- and sensor-specific, mean solar exoatmpospheric irradiance
= the solar zenith angle
= the Earth-Sun distance in astronomical units for the data in question
Equation 3- Conversion radiance to TOA reflectance
Equation 2- Conversion of DNs to radiance
28
3.2.2.2 Atmospheric Correction
Atmospheric scattering and absorption of radiation can distort the scale of
measurement between images (Song et al, 2001) and contaminate NDVI
measurements (Verstraete, 1994). To correct for these effects, the dark
object subtraction (DOS) method was used. This method is based on the
assumption that dark features in images (such as shadows) that should have
zero reflectance will appear brighter due to atmospheric scattering
(Chavez, 1989). The correction was applied by identifying the reflectance of
a shadowed pixel and subtracting this from all other values in the image
using ERDAS Imagine’s ‘Model Maker’ on a band-by-band basis. Table 3.5
shows the correction values used. The result does not provide an accurate
measure of actual surface reflectance because it unrealistically assumes
uniform atmosphere across a scene and only corrects for the additive effects
of scattering, not the multiplicative effects of absorption (Bruce & Hilbert,
2006). While more complex methods do exist, they require in-situ
measurements of atmospheric composition, which were not available.
Moreover, Song et al (2001) found that DOS was as effective as more
complex methods for aligning multitemporal data onto a common
radiometric scale, making it adequate for the purposes of this study.
Image
Dark Pixel Value
Band 1
Band 2
Band 3
Band 4
Band 5
Band 6
Band 7
1980
N/A
N/A
N/A
0.026932
0.012680
0.014958
0.008229
1984
0.037369
0.019798
0.025874
0.020303
N/A
N/A
N/A
1990
0.037856
0.018330
0.017564
0.016146
N/A
N/A
N/A
1996
0.053755
0.029981
0.010953
0.013661
0.002916
N/A
0.003
2002
0.056862
0.031161
0.014330
0.013603
0.002417
N/A
0.002285
Table 3.5- Values used for each spectral band to correct the Landsat images using Dark Object
Subtraction
3.2.2.3 Geometric Correction
The rotation and curvature of the Earth, movement of the sensor and
atmospheric refraction of radiation can geometrically distort Landsat
images (Campbell & Wynne, 2011). To correct for these distortions between
29
dates, all images were registered to the 2002 image (in the WGS-84 UTM zone
15N projection) using ERDAS Imagine. 25 ground control points (GCPs) were
used for each pair of images and a third order polynomial transformation. In
each case, the root mean square (RMS) error was less than 0.5 (Table 3.6)
and the nearest neighbour technique resampling technique was used to.
This resampling method was deemed better than alternatives such as
bilinear interpolation and cubic convolution, because it preserves the values
in the original image by assigning the corrected pixel the value of the nearest
pixel in the input image (Campbell & Wynne, 2011).
Image Pair
X Error
Y Error
Total RMS Error
1980 and 2002
0.1951
0.1631
0.2542
1984 and 2002
0.4019
0.2534
0.4751
1990 and 2002
0.1732
0.1640
0.2385
1996 and 2002
0.2472
0.1299
0.2792
Table 3.6- Root mean square error values obtained through the process of geographically registering
each Landsat image to the 2002 image
3.2.2.4 Selecting Subsets
To limit analysis to the study area, ERDAS Image was used to select a subset
for each image consisting of a 16x18.75km rectangle around the volcano
that incorporated both the extents of the pyroclastic flows and surges, and
up to the 5cm tephra isopach, producing a 300km2 study area. The resulting
images can be seen in Appendix 1.
3.2.3 Image Analysis
3.2.3.1 NDVI
After pre-processing, ERDAS Imagine was used to obtain the NDVI for each
image (see Equation 1, section 2.3.3.1). For both the TM and ETM+ images,
bands 4 and 5 were used, as these correspond directly to the red and near-
infrared wavelengths, respectively, with equal bandwidths across both
sensors. However, the MSS images have two near-infrared bands (bands 6
30
and 7 for 1980, bands 3 and 4 for 1984 and 1990) and the red band (band 5
for 1980, band 2 for 1984 and 1990), has a different bandwidth to TM/ETM+.
The first near-infrared band was used for NDVI calculation with MSS images
as Huete et al (2002) find this combination to have smaller differences with
TM/ETM+ calculated NDVIs. However, they note mean deviations of NDVI as
high as 0.086 between TM and MSS images, so the potential effects of this
should be noted during analysis of results.
3.2.3.2 Change Detection
To facilitate change detection, the 1996 and 2002 NDVI images were first
resampled to 60x60m cell size using the nearest neighbour technique, so that
all images had the same resolution. ERDAS Imagine was then used to create
NDVI difference images for four periods: 1980-1984, 1984-1990, 1990-1996
and 1996-2002. In each case, the NDVI pixel values of the first image in the
period were subtracted from the geographically corresponding pixels in the
last image. Negative results indicate a decrease in NDVI (and therefore
vegetation cover), positive results indicate increases in NDVI and zeros
indicate no change.
3.3 ANCILLARY DATA
Ancillary data related to the direct effects of the eruption and post-eruption
ecological driving factors were collated in order to evaluate their impact on
vegetation response and regrowth. As no field data collection was carried
out, in most cases surrogate variables and secondary data had to be used.
GIS was used to compile the necessary datasets so that they could easily be
integrated with the change detection results.
31
3.3.1 Direct Effects of the Eruption
3.3.1.1 Tephra Thickness
The extent of tephra deposits (Figure 3.2) were manually digitised in ArcMap
from measurements previously published by Inbar et al (2001; fig.1, pg.177).
Tephra thickness is an important variable as it is related to the ability of buried
vegetation to survive and re-establish, and it determines whether colonisers
are able to reach organic soil below (Lawrence, 2006).
3.3.1.2 Pyroclastic Surges
The extents of pyroclastic surges and flows were also digitised manually in
ArcMap (Figure 3.3) using the findings of Sigurdsson et al (1987; fig.2, pg.470).
For the purposes of analysis, the extent of surge deposits were transformed
to represent the number of surges affecting each area (Figure 3.5). The
effects of pyroclastic flows are difficult to quantify and distinguish from those
of surges, and so they were not included in analysis. The number of surges
will be related to vegetation destruction and ability to survive.
Figure 3.5- Map showing distribution of areas with equal number of surges affecting them,
displayed over a true-colour Landsat ETM+ image acquired in 2002.
32
3.3.1.3 Delineating Zones of Interest
The study area was delineated into three zones (Figure 3.6) relating to the
type of eruptive activity experienced: a ‘Surges Only’ zone where land was
only exposed to pyroclastic surges, not significant tephra fallout; a ‘Tephra
Only’ zone where land was only exposed to tephra deposition; and a ‘Surges
& Tephra’ zone which was exposed to both types of eruptive products. The
purpose of these zones is to act as surrogates for the types of successional
processes occurring. Tephra deposits are more likely to be associated with
secondary succession, as there is a higher chance of vegetation and soil
survival. Conversely, pyroclastic surges and flows create new volcanic
surfaces and are therefore more likely to be associated with processes of
primary succession (del Moral & Grishin, 1999). The Boolean operators ‘clip’
and ‘erase’ were used on the extents of surge and tephra deposits in
ArcMap to delineate the zones.
Figure 3.6- Map showing the three zones of interest, displayed over a true-colour Landsat
ETM+ image acquired in 2002.
33
3.3.2 Post-Eruption Ecological Driving Factors
3.3.2.1 Elevation
Elevation data was used as a surrogate variable for growing conditions, due
to its negative correlation with temperature (Lawrence, 2006). 1-Arc Second
Shuttle Radar Topography Mission (SRTM) elevation data was downloaded
from the USGS (2015b) ‘EarthExplorer’ website. Arcmap was used to clip the
data to the study area and fill small voids of missing data using the mean of
the five pixels within a circular neighbourhood of each missing pixel. The
result was resampled to 60x60m cell size using the cubic convolution
technique (which uses a weighted average of the nearest 16 input pixels to
calculate the value of the output pixel) (Campbell & Wynne, 2011) and
projected to the WGS-84 UTM zone 15N projection to match the Landsat
data, producing the DEM in Figure 3.7.
Figure 3.7- 1-Arc Second SRTM digital elevation model (resampled to 6ox60m pixel size),
acquired in 2000, displayed over a hillshade for the study area.
34
3.3.2.2 Aspect
Aspect was also used as a surrogate for growing conditions, as it determines
the amount of solar radiation a surface receives and is therefore also related
to temperature (Lawrence, 2006). ArcMap was used to calculate the aspect
(Figure 3.8) from the DEM, with 60x60m cell sizes.
3.3.2.3 Slope
Slope angle was used as a surrogate for rates of erosion, with the
expectation that eruption products will erode quicker on steeper slopes
(Lawrence et al, 2006). Slope was derived from the DEM in ArcMap with
60x60m cells (Figure 3.9).
Figure 3.8- Map showing the surface aspect (slope direction) of the study area, created
using 1-Arc Second SRTM elevation data (resampled to 6ox60m pixel size) acquired in 2000.
35
3.3.2.4 Intensity of eruption impacts
Distance from the summit crater was used as a surrogate measure of
eruption intensity, with the expectation that intensity is greatest close to the
summit (Lawrence, 2006). The ‘Euclidean Distance’ tool in ArcMap was used
to create a raster dataset showing the distance from the summit crater
(Figure 3.10).
Figure 3.9- Map showing the angle of the surface slope for the study area, created using
1-Arc Second SRTM elevation data (resampled to 6ox60m pixel size) acquired in 2000.
36
3.3.2.5 Potential Seed Dispersal
Distance to surviving vegetation was used as a surrogate measure of the
potential for seed dispersal, with an increased likelihood of animal- and
wind-borne seed dispersal expected closer to vegetation (Lawrence, 2006).
The ‘Euclidean Distance’ tool in ArcMap was also used to create the raster
map showing distance from vegetation (figure 3.11). Areas of surviving
vegetation were defined as those with NDVI scores of at least 0.2 in 1984, as
this is generally agreed to be the minimum value for presence of vegetation
(Weier & Herring, 2000). However, this definition should be viewed as no more
than an estimate. With no imagery closer to the eruption date it cannot be
ruled out that areas with an NDVI of 0.2 in 1984 actually had no vegetation
immediately following the eruption, and with a lack of ground
measurements, it is impossible to tell whether 0.2 is an accurate measure of
the presence of vegetation.
Figure 3.10- Map showing distance to the summit crater, displayed over a true-colour
Landsat ETM+ image acquired in 2002.
37
3.4 STATISTICAL ANALYSIS
Arcmap was used to extract values of NDVI at each date, NDVI difference
for each period, the number of surges, tephra thickness, and all ancillary
variables (elevation, slope, aspect, distance from the summit and distance
from surviving vegetation) for every pixel within the three zones of interest.
Analysis was therefore carried out using a full census rather than just a
sample.
3.4.1 Scatter Plots and Bar Graphs
Microsoft Excel was used to create both scatter plots of mean NDVI over
time and bar charts of mean NDVI difference between periods within each
zone. These provide a visual indication of the initial eruption impacts and
changes in vegetation over time in relation to the initial eruption activity
experienced.
Figure 3.11- Map showing distance to surviving vegetation, displayed over a true-
colour Landsat ETM+ image acquired in 2002.
38
3.4.2 Regression Analysis
In order to evaluate the relative influence of ancillary variables on NDVI
difference, linear regression analysis was carried out in GeoDa. Ordinary
Least Squares (OLS) regression was initially trialled, but examination of the
residuals using Moran’s Index revealed significant spatial autocorrelation,
thereby violating the assumption of independent observations (Lawrence,
2006) (Appendix 2). The Maximum Likelihood Spatial Error Model of
Regression was therefore used, producing a regression equation in the form
of Equation 4, which includes an additional term to the OLS equation
accounting for spatial autocorrelation in the residuals (Ward & Gleditsch,
2008). Prior to running the regression, all independent variables were
standardised using Equation 5 and NDVI difference was scaled to 0-255 (8-
bit data range), in order to remove the effects of different units of
measurement and negative values in the dependent variable, which might
otherwise hinder the evaluation of relative influence (Bayramov et al, 2011).
A spatial weights matrix for the dataset was created using the threshold
distance of one-and-a-half pixels (90m) and this was input into the model.
Regression was carried out by zone with NDVI difference (‘NDVIdiff’) as the
dependent variable. For the first period (1980 to 1984), the independent
variables were number of surges (‘SUR’) and/or tephra thickness (‘TEPH’), in
order to evaluate the response of vegetation to the eruption. For the
remaining three periods, elevation (‘ELEV’), aspect (‘ASP’), slope (‘SLP’),
distance from the summit (‘SUM’) and distance from surviving vegetation
 
Where: = the dependent variable
= the independent variable
= the coefficient for the independent variable
= a spatially uncorrelated error term
 = the extent of spatial correlation in errors based on the spatial weights matrix
= a spatially correlated error term
Equation 4- Spatial Error Model Regression equation
39
(‘VEG’) were also included as independent variables, to evaluate the
factors affecting regrowth. After running each regression model, the p-
values of the results were analysed for significance. If coefficients were not
significant at the p<0.05 level, the variables were removed and the model
was re-run to provide a better fit. With the variables standardised, the
magnitude of the variable coefficients in the model output can be
interpreted as relating to its influence on the dependent variable.

Where:
= the standardised variable
= the original value the variable
= the mean of the variable
= the standard deviation of the variable
Equation 5- Standardisation of variables equation
40
4 RESULTS
4.1 NDVI
Figure 4.1- NDVI calculated from Landsat MSS data for 1980
Figure 4.2- NDVI calculated from Landsat MSS data for 1984
41
Figure 4.1 shows that prior to the eruption, the area around the volcano
predominantly had positive NDVI values, largely greater than 0.6.
Examination of pixel values shows that less than 1% have NDVI less than 0.2
and comparison with the 1980 Landsat image (Appendix 1) reveals these
areas to correspond with small clouds and with a river to the west of the
summit. Figure 4.2 shows that by 1984, the majority of the volcanic edifice
has NDVI values close to zero (-0.1 - 0.2), and 21% of image pixels have NDVI
less than 0.2. The area of low NDVI extends almost radially around the
summit, but with slight elongation in the northwest and southeast. The river
previously identified now appears much wider and further channels can be
identified to the north and south of the summit. Comparing Figure 4.3, Figure
4.4 and Figure 4.5 with Figure 4.2 shows a contraction of the area of low and
negative NDVI between 1984 and 2002. Over time, increases in NDVI appear
to move inwards towards the summit, although areas on the eastern flanks
of the volcano show consistently lower NDVI than surroundings from 1990
onwards. By 1996, most of the edifice has NDVI greater than 0.2 and by 2002,
only the summit crater has lower values. However, NDVI values on the edifice
remain lower than they were pre-eruption, predominantly between 0.4 and
0.6, rather than greater than 0.6. Conversely, areas in the rest of the 1996 and
2002 images display higher NDVI values than before 1990. The widths of the
identified river channels appear to contract over time and by 2002.
42
Figure 4.3- NDVI calculated from Landsat MSS data for 1990
Figure 4.4- NDVI calculated from Landsat TM data for 1996
43
4.2 CHANGE DETECTION
Figure 4.5- NDVI calculated from Landsat ETM+ data for 2002
Figure 4.6- NDVI change between 1980 and 1984
44
Figure 4.7- NDVI change between 1984 and 1990
Figure 4.8- NDVI change between 1990 and 1996
45
The NDVI differencing results in Figure 4.6, Figure 4.7, Figure 4.8 and Figure 4.9
further support the patterns observed from the NDVI images. The majority of
change is negative 1980-1984, indicating a decrease NDVI throughout. On
the volcanic edifice, the negative difference is predominantly in the region
of 0.79-0.6, exhibiting the elongated shape previously observed. After 1984,
change is predominantly positive, suggesting increasing NDVI values over
time. Between A large area around the summit of the volcano experiences
positive change in 1984-1990, mostly in the region of 0.21-0.4, and some
areas with >0.41. The area experiencing positive change has decreased by
1990-1996 and consists mostly of change in the region of 0.21-0.4. It becomes
impossible to distinguish a distinct region of NDVI change around the
volcano by 1996-2002. Instead, the majority of the area experiences minor
positive change in the region of 0.01-0.2.
Figure 4.9- NDVI change between 1996 and 2002
46
Figure 4.10- Mean NDVI over time by zone Figure 4.11- Mean NDVI difference between
images by zone
Figure 4.10 and Figure 4.11 show that NDVI decreased in all zones between
1980 and 1984, with the largest change occuring in the ‘Surges & Tephra’
zone (-0.479), the smallest in the ‘Tephra Only’ zone (-0.093) and
intermediate in the ‘Surges Only’ zone (-0.224). These differences occurred
despite mean pre-eruption NDVI being roughly equal (approximately 0.6).
This pattern continued throughout the study period, with the ‘Surges &
Tephra’ zone consistently exhibiting the lowest mean NDVI and greatest
amount of change. In all three zones, mean NDVI showed an increasing
trend after 1984, but the decreasing amounts of positive change between
periods shown in Figure 4.11 indicates a slowing of the rate of increase. After
1984, mean NDVI values in the three groups became more similar and
differences in amount of change experienced between groups decreased
over time. Although they do not fully converge, by 2002 all three zones have
mean NDVI between 0.684 and 0.798 and positive change between 0.046
and 0.087 for 1996-2002, further indicating a slowing rate of change. Finally,
Figure 4.10 shows that in all zones, mean NDVI eventually surpasses initial
values with the ‘Tephra Only’ zone reaching this point in 1990, ‘Surges Only’
in 1996 and ‘Surges & Tephra’ in 2002.
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1980 1985 1990 1995 2000
Mean NDVI
Year
Mean NDVI Over Time by Zone
Tephra Only Surges Only Surges & Tephra
-0.6
-0.5
-0.4
-0.3
-0.2
-0.1
0
0.1
0.2
0.3
1984-1980 1990-1984 1996-1990 2002-1996
Mean NDVI Difference
Time Period
Mean NDVI Differences Between Images
by Zone
Tephra Only Surges Only Surges & Tephra
47
Figure 4.12- Mean NDVI over time in each
tephra isopach area for the 'Tephra Only' zone
Figure 4.13- Mean NDVI difference between
images by tephra thickness in the ‘Tephra
Only’ zone
Figure 4.14- Mean NDVI over time by tephra
thickness for the ‘Surges & Tephra’ zone
Figure 4.15- Mean NDVI difference between
images by tephra thickness in the ‘Surges &
Tephra’ zone
Figure 4.12 and Figure 4.13 show that in the ‘Tephra Only’ zone, very similar
mean NDVIs and NDVI differences are recorded throughout the study period
for both 10-20cm and 5-10cm thicknesses. Conversely, Figure 4.14 and Figure
4.15 show that in the ‘Surges & Tephra’ zone, there is the greatest negative
change in NDVI within the >20cm category, intermediate for 10-20cm and
smallest for the 5-10cm category. The subsequent shape of the scattergraph
and relatively equal size of the bars for remaining periods indicates that the
amount of NDVI increase over time is fairly equal among the three
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1980 1985 1990 1995 2000
Mean NDVI
Year
Mean NDVI Over Time in each Tephra
Isopach Area for the 'Tephra Only' Zone
5-10cm 10-20cm
-0.15
-0.1
-0.05
0
0.05
0.1
0.15
1984-1980 1990-1984 1996-1990 2002-1996
Mean NDVI Difference
Time Period
Mean NDVI Differences Between Images
by Tephra Thickness in the 'Tephra Only'
Zone 5-10cm 10-20cm
-0.1
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
1980 1985 1990 1995 2000
Mean NDVI
Year
Mean NDVI Over Time by Tephra Thickness
for the 'Surges & Tephra' Zone
5-10cm 10-20cm <20cm
-0.7
-0.6
-0.5
-0.4
-0.3
-0.2
-0.1
0
0.1
0.2
0.3
0.4
1984-1980 1990-1984 1996-1990 2002-1996
Mean NDVI Difference
Time Period
Mean NDVI Differences Between Images
by Tephra Thickness in the 'Surges &
Tephra' Zone
5-10cm 10-20cm >20cm
48
categories despite these initial differences, albeit with a slightly smaller initial
increase in the >20cm category. Figure 4.16, Figure 4.17, Figure 4.18 and
Figure 4.19 show that in both the ‘Surges Only’ and ‘Surges & Tephra’ zones,
distinct mean NDVI values can be identified post-eruption based on the
number of surges affecting the area, with areas affected by three surges
experiencing the greatest initial NDVI decrease and areas affected by one
surge experiencing the least.
Figure 4.16- Mean NDVI over time by number
of surges in the ‘Surges Only’ zone
Figure 4.17- Mean NDVI difference between
images by number of surges in the ‘Surges
Only’ zone
Figure 4.18- Mean NDVI over time by number
of surges in the ‘Surges & Tephra’ zone
Figure 4.19- Mean NDVI difference between
images by number of surges in the ‘Surges &
Tephra’ zone
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1980 1985 1990 1995 2000
Mean NDVI
Year
Mean NDVI Over Time by Number of Surges
in the 'Surges Only' Zone
1 Surge 2 Surges 3 Surges
-0.4
-0.3
-0.2
-0.1
0
0.1
0.2
0.3
1984-1980 1990-1984 1996-1990 2002-1996
Mean NDVI Difference
Time Period
Mean NDVI Differences Between Images
by Number of Surges in the 'Surges Only'
Zone
1 Surge 2 Surges 3 Surges
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
1980 1985 1990 1995 2000
Mean NDVI
Year
Mean NDVI Over Time by Number of Surges
in the 'Surges & Tephra' Zone
1 Surge 2 Surges 3 Surges
-0.7
-0.6
-0.5
-0.4
-0.3
-0.2
-0.1
0
0.1
0.2
0.3
0.4
1984-1980 1990-1984 1996-1990 2002-1996
Mean NDVI Difference
Time Period
Mean NDVI Differences Between Images
by Number of Surges in the 'Surges &
Tephra' Zone
1 Surge 2 Surges 3 Surges
49
4.3 REGRESSION ANALYSIS
4.3.1 Tephra Only Zone
Table 4.1 indicates that there is
no significant influence of tephra
thickness on NDVI difference in
the ‘Tephra Only’ zone. It shows
that SLP is not significant for 1984-
1990 and TEPH, SUM and VEG
are not significant for 1990-1996
or 1996-2002. Re-running the
models without the insignificant
variables (Table 4.2) results in the
significant regression models
shown in Equation 6, Equation 7
and Equation 8.
Equation 6- Regression equation for 1984-1990 NDVI difference in the ‘Tephra Only’ zone
Equation 7- Regression equation for 1990-1996 NDVI difference in the ‘Tephra Only’ zone
Equation 8- Regression equation for 1996-2002 NDVI difference in the ‘Tephra Only’ zone
Regression
Coefficients
1980-
1984
1984-
1990
1990-
1996
1996-
2002
TEPH
p-value
-0.305
0.714
-2.060
0.043
-0.145
0.827
1.019
0.224
ELEV
p-value
- - -
-8.015
0.000
-2.316
0.001
2.538
0.008
ASP
p-value
- - -
1.128
0.004
-1.474
0.000
1.594
0.000
SLP
p-value
- - -
-0.486
0.276
-1.577
0.000
1.902
0.000
SUM
p-value
- - -
-5.246
0.005
-1.015
0.210
0.392
0.718
VEG
p-value
- - -
3.640
0.000
-0.184
0.425
0.431
0.104
Constant
p-value
140.01
0.000
101.70
0.000
134.48
0.000
116.21
0.000
Spatial Error
p-value
0.892
0.000
0.891
0.000
0.749
0.000
0.788
0.000
Table 4.1- Results of the first run of regression analysis for
the ‘Tephra Only’ zone. Non-significant values (p>0.05)
are highlighted in yellow
    
 
    
    
50
For 1984-1990, 65% of variance in NDVI
difference is explained by the regression
model (R2=0.652), with elevation
displaying the greatest influence.
Distance from the summit is also
important, while aspect is least
important. For 1990-1996, 42% of
variance is explained by the model
(R2=0.420). Elevation still displays the
most influence, but slope and aspect
have also become important. For 1996-
2002, the same distribution of influence is exhibited, but the model accounts
for a greater percentage of variance (49% or R2=0.488).
4.3.2 Surges Only Zone
Table 4.3 indicates that number
of surges does not exert
significant influence on NDVI
difference in the ‘Surges Only’
zone. It shows that SUR, ELEV and
ASP are not significant for 1984-
1990 and ELEV and SUM are not
significant for 1996-2002. Re-
running the models without the
insignificant variables (Table 4.4)
results in the significant
regression models shown in
Equation 9, Equation 10 and
Equation 11.
Regression
Coefficients
1984-
1990
1990-
1996
1996-
2002
TEPH
p-value
-2.060
0.044
- - -
- - -
ELEV
p-value
-8.048
0.000
-2.106
0.002
2.372
0.010
ASP
p-value
1.148
0.000
-1.470
0.000
1.612
0.000
SLP
p-value
- - -
-1.570
0.000
1.941
0.000
SUM
p-value
-5.210
0.005
- - -
- - -
VEG
p-value
3.630
0.000
- - -
- - -
Constant
p-value
101.70
0.000
134.48
0.000
116.21
0.000
Spatial Error
p-value
0.891
0.000
0.750
0.000
0.789
0.000
Table 4.2- Results of the second run of
regression analysis for the ‘Tephra Only’ zone.
Regression
Coefficients
1980-
1984
1984-
1990
1990-
1996
1996-
2002
SUR
p-value
-0.933
0.259
-0.107
0.900
-1.639
0.011
1.062
0.029
ELEV
p-value
- - -
-0.156
0.921
-4.317
0.000
0.156
0.770
ASP
p-value
- - -
-0.317
0.139
-0.667
0.002
1.177
0.000
SLP
p-value
- - -
-0.958
0.000
0.910
0.000
1.206
0.000
SUM
p-value
- - -
-5.337
0.026
-6.105
0.000
-0.610
0.250
VEG
p-value
- - -
14.988
0.000
-2.640
0.000
3.588
0.000
Constant
p-value
115.30
0.000
134.41
0.000
141.86
0.000
109.98
0.000
Spatial Error
p-value
0.962
0.000
0.948
0.000
0.837
0.000
0.736
0.000
Table 4.3- Results of the first run of regression analysis for
the Surges Only’ zone. Non-significant values (p>0.05)
are highlighted in yellow
51
Equation 9- Regression equation for 1984-1990 NDVI difference in the ‘Surges Only’ zone
Equation 10- Regression equation for 1990-1996 NDVI difference in the ‘Surges Only’ zone
Equation 11- Regression equation for 1996-2002 NDVI difference in the ‘Surges Only’ zone
76% of variance in NDVI difference is
explained by the regression model for
1984-1990 (R2=0.763). Distance to
vegetation clearly exerts the greatest
influence, distance to summit exerts
moderate influence and slope is of minor
importance. For 1990-1996, 53% of
variance is explained by the model
(R2=0.530). Distance from summit is now
most important, while elevation becomes
the second most important. Distance to
vegetation has become less influenctial
but is still more important than all remaining variables, which exhibit minor
influence. For 1996-2002, distance from vegetation exhibits the greatest
influence and all remaining variable exhibit similarly small values. The model
accounts for a smaller percentage of variance (41% or R2=0.414).
Regression
Coefficients
1984-
1990
1996-
2002
Number of Surges
p-value
- - -
1.118
0.012
Elevation
p-value
- - -
- - -
Aspect
p-value
- - -
1.184
0.000
Slope
p-value
-0.958
0.000
1.224
0.000
Summit Distance
p-value
-5.255
0.027
- - -
Vegetation Distance
p-value
14.997
0.000
3.635
0.000
Constant
p-value
134.41
0.000
109.98
0.000
Spatial Error Term
p-value
0.948
0.000
0.736
0.000
Table 4.4- Results of the second run of
regression analysis for the ‘Surges Only’
zone
   
   
 
    
52
4.3.3 Surges & Tephra Zone
Table 4.5 shows that: TEPH is not
significant for 1980-1984; SUR and
ELEV are not significant for 1984-
1990; SUR, SLP and VEG are not
significant for 1990-1996; and SUR,
ELEV and VEG are not significant
for 1996-2002. Re-running the
models without the insignificant
variables (Table 4.6) results in the
regression significant models
shown in Equation 12, Equation
13, Equation 14 and Equation 15.
Equation 12- Regression equation for 1980-1984 NDVI difference in the ‘Surges & Tephra’ zone
Equation 13- Regression equation for 1984-1990 NDVI difference in the ‘Surges & Tephra’ zone
Equation 14- Regression equation for 1990-1996 NDVI difference in the ‘Surges & Tephra’ zone
Regression
Coefficients
1980-
1984
1984-
1990
1990-
1996
1996-
2002
TEPH
p-value
-0.987
0.064
-1.865
0.019
2.881
0.000
1.619
0.000
SUR
p-value
-1.795
0.000
1.071
0.152
-0.735
0.152
-0.337
0.309
ELEV
p-value
- - -
3.044
0.092
-5.833
0.000
-0.587
0.137
ASP
p-value
- - -
0.365
0.025
-1.867
0.000
2.207
0.000
SLP
p-value
- - -
-0.959
0.000
0.050
0.811
1.361
0.000
SUM
p-value
- - -
61.661
0.000
-3.280
0.000
-2.889
0.000
VEG
p-value
- - -
136.62
0.000
1.172
0.191
-0.385
0.484
Constant
p-value
104.49
0.000
146.57
0.000
140.65
0.000
125.04
0.000
Spatial Error
p-value
0.964
0.000
0.993
0.000
0.767
0.000
0.672
0.000
Table 4.5- Results of the first run of regression analysis for
the ‘Surges & Tephra’ zone. Non-significant values
(p>0.05) are highlighted in yellow
   
    

    

53
Equation 15- Regression equation for 1996-2002 NDVI difference in the ‘Surges & Tephra’ zone
For 1980-1984, the model
accounts for 88% of the variance
(R2=0.878). For 1984-1990, the
regression model accounts for
74% of variance in NDVI
difference (R2=0.740), with
distance to vegetation clearly
the most imporant factor and
distance from the summit also
exerting significant influence.
Tephra thickness, aspect and
slope are of minor importance. By
1990 to 1996, the regression
model accounts for 43% of variance (R2=0.426), with elevation having
become significant and the most important influence. Distance to the
summit remains second most influential, and the importance of tephra
thickness and aspect have remained low. Only 35% of variance in NDVI
difference is accounted for by the regression model for the final period
(R2=0.345). Distance to summit and aspect appear slightly more important,
but the influences of tephra thickness and slope are only slightly lower.
Regression
Coefficients
1980-
1984
1984-
1990
1990-
1996
1996-
2002
TEPH
p-value
- - -
-1.860
0.019
2.958
0.000
1.090
0.000
SUR
p-value
-1.809
0.000
- - -
- - -
- - -
ELEV
p-value
- - -
- - -
-5.336
0.000
- - -
ASP
p-value
- - -
0.364
0.026
-1.867
0.000
2.211
0.000
SLP
p-value
- - -
-0.948
0.000
- - -
1.347
0.000
SUM
p-value
- - -
60.892
0.000
-3.465
0.000
-2.439
0.000
VEG
p-value
- - -
136.34
0.000
- - -
- - -
Constant
p-value
104.50
0.000
146.70
0.000
140.66
0.000
125.04
0.000
Spatial Error
p-value
0.964
0.000
0.993
0.000
0.768
0.000
0.674
0.000
Table 4.6- Results of the second run of regression
analysis for the ‘Surges & Tephra’ zone
    

54
5 DISCUSSION
5.1 INITIAL ERUPTION IMPACTS
The dramatic extent of vegetation change caused by the eruption can be
observed from the large amount of negative difference displayed in Figure
4.6. It indicates that large areas of vegetation were damaged or killed by
the eruption, especially in the area up to about 6km from the summit. This is
consistent with other ground-based studies of the area, such as Inbar et al
(2001) who noted that there was no vegetation in the immediate vicinity of
the summit post-eruption, and Sigurdsson et al (1987) who noted the
pyroclastic surge deposits to be free of vegetation in early 1983. As well as
this, the shape of the area of negative change in NDVI is revealing of
eruption characteristics. Although the slightly elongated area of NDVI
difference between -0.79 and -0.6 does extend further east of the summit
than west, corresponding with the direction of maximum plume deposition
(Rose & Durant, 2009), it does not exhibit the bidirectional dispersal pattern
of the plume observed by Carey & Sigurdsson et al (1986). It extends more
to the north-northwest and east-southeast than south-southwest and east-
northeast, suggesting deposition from the plume is not the main cause of the
vegetation change.
Figure 4.10 and Figure 4.11 show that all zones experienced a decrease in
NDVI in 1980-1984, implying that both tephra and surges have at least some
negative impact on vegetation. However, greater amounts of change are
experienced in zones affected by surges, suggesting these are more
damaging to vegetation than tephra deposition. This conclusion is
supported by evidence from other volcanoes. At Mount St. Helens, for
example, the worst affected zones in terms of vegetation loss were those
where vegetation was flattened and carbonised by hot pyroclastic surges
(Oppenheimer, 2011). The results also show that the greatest impacts are
experienced in the ‘Surges & Tephra’ zone. This also is to be expected as not
55
only does this area experience both forms of volcanic activity, it is also the
area closest to the summit and will therefore experience the greatest
intensity of activity (Lawrence, 2006).
In the ‘Tephra Only’ zone, differences in vegetation response are not
distinguishable between 5-10cm and 10-20cm tephra thickness, as shown by
Figure 4.12 and Figure 4.13 and the lack of a significant result in the regression
analysis. This result could be caused by the fact that the closest post-eruption
image was two years after the event. As Inbar et al (2001) note, erosion rates
at El Chichón were several orders of magnitude higher than normal in the
first rainy season after the eruption, and are typically high anyway due to the
tropical climate. As such, by 1984 large amounts of the tephra deposits may
have already been removed, making it difficult to distinguish the effects. The
regression analysis also reveals that number of surges is only a significant
controlling variable in 1980-1984 for the ‘Surges & Tephra’ zone. The lack of
significance in the ‘Surges Only’ zone could also be a result of the late image
acquisition date as this zone is comprised of the distal parts of the surge
deposits, which will have dissipated in destructive force and leave thinner
deposits (Sigurdsson et al, 1987), making it easier to recover quickly. Despite
this, however, the ability to distinguish mean NDVI and NDVI difference by
number of surges in both the ‘Surges Only’ and ‘Surges & Tephra’ zones
reveals that one surge alone was not enough to completely destroy
vegetation, but rather successive surges added to the damage. Moreover,
the fact that number of surges was a significant factor and tephra thickness
was not in the ‘Surges & Tephra’ zone provides further support for surges
being more destructive to vegetation.
5.2 PATTERNS OF REVEGETATION
The initial high NDVI increase in the first period post-eruption (Figure 4.11)
suggests that there is not a large lag at the beginning of the recovery process
56
that may be expected on other kinds of eruption deposits. Thick lava flows,
for instance, can take decades to weather sufficiently to allow vegetation
to colonise (del Moral & Grishin, 1999). If there was a lag before revegetation
began, the time periods used in this study were too coarse to capture this.
The subsequent steady decline in amount of change by period,
approaching values close to zero in 1996-2002, could suggest that full
recovery is reached in this last period. This interpretation is supported by Inbar
et al’s (2001) study which found vegetation cover on the flanks of the
volcano to be complete by 1997. They do, however, point out that this
definition of full coverage relates to coverage of shrubs and grasses, not
necessarily the full tropical vegetation observed before the eruption. This
definition is supported by Figure 4.5 which shows the area covered by low
positive NDVI values, rather than high ones. As NDVI values in each zone are
not equal by the final period, like they were prior to the eruption (Figure 4.10),
and there are still small amounts of increase occurring (Figure 4.11), this
suggests vegetation growth is still occurring, supporting Cervantes-Borja et
al’s (1983; cited in Inbar et al, 2001) estimation that full recovery of tropical
forests would take over 20 years.
In the ‘Tephra Only’ zone, it can be concluded that full recovery occurred
by 1990 because by this point, mean NDVI surpassed the pre-eruption value
(Figure 4.10). This is significant because all NDVI values were calculated from
Landsat MSS data up to 1990 and are therefore comparable, and it is
supported by Inbar et al’s (2001) identification of recovery rates of only a few
years in tephra-affected areas. This result therefore provides support for the
assertion made by Dale et al’s (2005) evaluation of multiple eruptions that
the effect on vegetation from tephra is only short-term (years to decades)
and that tephra deposition does not alter vegetation recovery trajectory.
Such a conclusion cannot be made for the zones affected by surges,
however. Initial NDVI values were not surpassed until later in the study period,
when Landsat TM/ETM+ imagery was used. The different wavelengths used
57
to calculate NDVI for these sensors means the results are not directly
comparable. NDVI estimates are likely to be higher for these sensors (Huete
et al, 2002) and evidence for this can be seen in Figure 4.4 and Figure 4.5
which show higher values recorded for areas unaffected by the eruption
than in previous years. Although these higher values could be the result of
increased vegetation growth due to other factors such as climate or the
fertilising effect of ash deposition (Lockwood & Hazlett, 2007), without ground
observations to normalise the NDVI scale, it is impossible to conclude when
or if vegetation reached pre-eruption levels in these zones. The results do,
however, provide support for the theory that revegetation occurs over a
longer time period on pyroclastic surge and flow deposits, compared to on
tephra deposits (Dale et al, 2005).
Spatial patterns can be identified over time that relate to the eruption
characteristics. Figure 4.3, Figure 4.4 and Figure 4.5 show an area to the east
of the summit consistently exhibiting lower NDVI values. This is likely related to
the fact that tropospheric winds mainly deposited volcanic material to the
east leading to increased burial of vegetation and longer recovery periods
(Dale et al, 2005). As well as this, the changing shape of the river channels
over time could suggest that the channels were initially choked with volcanic
material, which was then gradually removed and eroded over time. This
interpretation is supported by observations from Inbar et al (2001) who found
20m of volcanic material deposited in the drainage channels after the
eruption.
5.3 FACTORS INFLUENCING REVEGETATION
For the first period 1984-1990, the regression analysis found that distance to
surviving vegetation and distance to the summit were the most important
factors controlling NDVI change in the ‘Surges Only’ and the ‘Surges &
Tephra’ zone. These factors are consistent with what would be expected in
58
primary-succession zones and with found in corresponding zones at Mount
St. Helens by Lawrence (2006). Because vegetation survival is expected to
be much lower in these zones, revegetation will depend upon the ability of
colonisers to reach the area and take hold. Therefore, rates of vegetation
increase are highest in those areas closest to surviving vegetation, where
seed dispersal is easier, and in the areas furthest from the summit where
eruption intensity is lower and the chances of vegetation taking hold are
higher. Conversely, in the ‘Tephra Only’ zone, where response is expected to
be dominated by secondary succession, the regression analysis result did not
correspond with Lawrence (2006). They found tephra thickness and slope to
be most important in the first period. This would be expected as these factors
relate to the rate of erosion of material. This study, however, found elevation
to be most important. Whilst this does not correspond with Lawrence’s (2006)
findings, it does support the idea posited in section 5.1 that the rapid erosion
rates at El Chichón (Inbar et al, 2001) had already caused much of the
deposits in this zone to be removed by 1984. As such, the factors relating to
initial recovery are no longer important, but those relating to sustained
growth are. Elevation would therefore be expected to be important as areas
of low elevation experience warmer temperatures and are more amenable
to continued growth.
The increase in importance of factors related to growing conditions is visible
in the regression results for all zones in the second period (1990-1996). In the
‘Surges Only’ and ‘Surges & Tephra’ zones, the importance of elevation
increases, while the importance of distance from vegetation decreased. This
is consistent with Lawrence’s (2006) findings at Mount St. Helens, where
distance to vegetation decreases in importance as colonisers have already
established and require conditions favouring vegetation survival instead. In
the ‘Tephra Only’ zone, elevation continues to be important and aspect
becomes important, which is also related to growing conditions through
temperature and solar radiation. The low R2 value for this zone (42%) however
59
indicates that the factors examined do not explain a great deal of the
variance. The same result was found by Lawrence (2006) for tephra zones in
this period. This could indicate that there are other factors not examined in
either study that account for change during this period.
In the final period (1996-2002), the R2 values for regression models in all zones
are very low (49%, 41% and 35%). Again this could suggest the importance
of an unexamined factor, but it is more likely to be further indication that by
the final period, vegetation levels had either reached or were nearing full
recovery because very little change in NDVI needs to be explained. In the
‘Tephra Only’ zone, the most important factors continue to relate to growing
conditions, but in the ‘Surges Only’ and ‘Surges & Tephra’ zones, these
factors seem to become unimportant. Instead, the most important factors
(distance to the summit, distance to vegetation and initial eruption
characteristics) seem to show an inverse relationship with vegetation
change, highlighting the increased vegetation change occurring in the
areas closest to the summit in the latter stages of recovery observed by Inbar
et al (2001). In contrast, at Mount St. Helens, Lawrence (2006) found factors
relating to growing conditions (e.g. elevation and aspect) to be still be
significant 20 years after the eruption in all zones, as large areas were still
undergoing processes of recovery. Inbar et al (2001) and Cervantes-Borja et
al (1983) note full vegetation coverage of the area surrounding El Chichón
within 20 years of the eruption, while del Moral & Lacher (2005) (among
others) note processes of recovery still occurring at Mount St Helens at least
25 years after the eruption. These difference in regression results therefore
highlight the importance of studying volcanic eruptions and environments
individually, because the comparison makes it clear that the differing local
conditions at each volcano, relating to the eruption parameters and local
climate, lead to differences in rates and patterns of vegetation recovery.
60
6 CONCLUSIONS
In conclusion, this study has shown that the area of greatest vegetation
change at El Chichón was experienced closest to the summit, where
eruption intensity was greatest and where vegetation was affected by
pyroclastic surges, flows, and tephra deposition. In more distal locations,
areas subjected only to tephra deposition experienced the least amount of
change, while areas of surge impacts experienced intermediate change.
Moreover, the results have shown this pattern to continue over the 20 years
following the eruption, with areas affected by tephra deposition recovering
the quickest and areas affected by both kinds of eruptive activity taking the
longest. This supports findings from other volcanoes, which suggest tephra is
less destructive to vegetation than pyroclastic surges and flows (Dale et al,
2005), and ground observations at El Chichón (Inbar et al, 2001). Evaluation
of vegetation change and regression analysis has shown that recovery at El
Chichón was a rapid process, with complete vegetation coverage and
decreasing amounts of change by the end of the study period. This is
consistent with Inbar et al’s (2001) observations and with results from other
volcanoes that suggest tropical environments provide conditions for rapid
vegetation recovery (e.g. De Rose et al’s [2010] study of Mount Pinatubo).
However, although many of the results have proven consistent with findings
from other volcanoes, there have also been important differences.
Regression analysis found factors relating to initial vegetation recovery to be
important in the first period after the eruption, and growing conditions to be
important in the second, in agreement with results from Mount St. Helens
(Lawrence, 2006). However, a notably different set of factors were found to
be important in the third period. Moreover, the regression equations for the
final period explained a much lower portion of variance, suggesting
recovery at El Chichón was quicker than at Mount St. Helens. These
differences highlight the importance of studying volcanic eruptions and
61
environments individually, instead of making generalisations, as the course
and rate of succession at different volcanoes is extremely complex,
unpredictable and unique (del Moral & Grishin, 1999).
Finally, this study has highlighted weaknesses in the methods used,
predominantly relating to the lack of ground data available to support the
remote sensing. Because the methods of atmospheric correction used
precluded derivation of surface reflectance, and because NDVI can exhibit
a varying relationship with vegetation between places (Campbell & Wynne,
2011), the ability to draw accurate conclusions about the type and absolute
amount of vegetation present in this study is limited. As well as this, ground
observations may have allowed better characterisation of the factors
affecting vegetation change. Inbar et al (2001), for instance, suggested that
human management of vegetation was important in some areas around El
Chichón, while at Mount Pinatubo, De Rose et al (2010) found geology and
watershed morphology to be important. These factors were impossible to
quantify without ground data and may therefore have accounted for some
of the unexplained variance in NDVI. It is suggested here that in the future,
research should make use of higher resolution imagery, which may allow
better characterisation of vegetation, and ground-truth data where
available. But, despite these weaknesses, this study has permitted the
identification and analysis of patterns of vegetation response and recovery
through remote sensing at a previously unstudied location. With vegetation
response to volcanic eruptions being a very complex and unpredictable
process (del Moral & Grishin, 1999), and with millions of people worldwide
living on the flanks of volcanoes, it is important to understand what the
potential impacts of volcanic eruptions on vegetation may be. With studies
currently limited due to the intensive nature of field work required and the
remote locations of many volcanoes, remote sensing is valuable for
expanding research in this field to previously unstudied volcanoes and as a
supplementary tool to reduce field work requirements.
62
7 REFERENCES
ABRAMS, M., ABBOTT, E. and KAHLE, A., 1991. Combined Use of Visible, Reflected
Infrared, and Thermal Infrared Images for Mapping Hawaiian Lava Flows. Journal
of Geophysical Research, 96(B1), pp. 475-485.
ABRAMS, M., BIANCHI, R. and PIERI, D., 1996. Revised Mapping of Lava Flows on
Mount Etna, Sicily. Photogrammetric Engineering & Remote Sensing, 62(12), pp.
1353-1359.
BAYRAMOV, E., BUCHROITHNER, M.F. and MCGURTY, E., 2011. Determination of
main climate and ground factors controlling vegetation cover regrowth along oil
and gas pipelines using multiple, spatial and geographically weighted regression
procedures. Environmental Earth Sciences, 66(7), pp. 2047-2062.
BENSON, C., 2005. Volcanoes and the Economy. In: J. MARTÍ and G.G.J. ERNST,
eds, Volcanoes and the Environment. New York: Cambridge University Press, pp.
440-467.
BONASIA, R., COSTA, A., FOLCH, A., MACEDONIO, G. and CAPRA, L., 2012.
Numerical simulation of tephra transport and deposition of the 1982 El Chichón
eruption and implications for hazard assessment. Journal of Volcanology and
Geothermal Research, 231-232, pp. 39-49.
BOYD, D.S., 2007. Physical Basis of Remote Sensing. In: P.M. ATKINSON, ed,
Geoinformatics Volume I. EOLSS Publishers.
BRUCE, C.M. and HIBERT, D.W., 2006. Pre-processing Methodology for Application
to Landsat TM/ETM+ Imagery of the Wet Tropics. 44. Cairns: Rainforest CRC.
CAMPBELL, J.B. and WYNNE, R.H., 2011. Introduction to remote sensing. 5th edn.
New York: The Guilford Press.
CAREY, S. and SIGURDSSON, H., 1986. The 1982 eruptions of El Chichón volcano,
Mexico (2): Observations and numerical modelling of tephra-fall distribution.
Bulletin of Volcanology, 48(2), pp. 127-141.
CARN, S.A., 1999. Application of synthetic aperture radar (SAR) imagery to
volcano mapping in the humid tropics: a case study in East Java, Indonesia.
Bulletin of Volcanology, 61(1), pp. 92-105.
CARN, S.A. and OPPENHEIMER, C., 2000. Remote monitoring of Indonesian
volcanoes using satellite data from the Internet. International Journal of Remote
Sensing, 21(5), pp. 873-910.
CHANDER, G., MARKHAM, B.L. and HELDER, D.L., 2009. Summary of current
radiometric calibration coefficients for Landsat MSS, TM, ETM+, and EO-1 ALI
sensors. Remote Sensing of Environment, 113(5), pp. 893-903.
COLCOUGH, S., 2005. Investigations of Nyamuragira and Nyiragongo Volcanoes
(Democratic Republic of the Congo) Using InSAR, University of Cambridge.
63
DALE, V.D., 1989. Wind dispersed seeds and plant recovery on Mount St. Helens
debris avalanche. Canadian Journal of Botany, 67, pp. 1434-1441.
DALE, V.D., 1991. Revegetation of Mount St. Helens debris avalanche 10 years
posteruptive. National Geographic Research and Exploration, 7, pp. 328-341.
DALE, V.H., DELGADO-ACEVEDO, J. and MACMAHON, J., 2005. Effects of modern
volcanic eruptions on vegetation. In: J. MARTÍ and G.G.J. ERNST, eds, Volcanoes
and the Environment. New York: Cambridge University Press, pp. 227-249.
DE ROSE, R.C., OGUCHI, T., MORISHIMA, W. and COLLADO, M., 2011. Land cover
change on Mt. Pinatubo, the Philippines, monitored using ASTER VNIR.
International Journal of Remote Sensing, 32(24), pp. 9279-9305.
DE SCHUTTER, A., KERVYN, M., CANTERS, F., BOSSHARD-STADLIN, S., SONGO, M. and
MATTSSON, H., 2015. Ash fall impact on vegetation: a remote sensing approach of
the Oldoinyo Lengai 200708 eruption. Journal of Applied Volcanology, 4(1), pp.
1-18.
DEAN, K.G., ENGLE, K., GROVES, J., DEHN, J. and PARTINGTON, K., 2002. Analysis of
surface processes using SAR data: Westdahl Volcano, Alaska. International Journal
of Remote Sensing, 23(21), pp. 4529-4550.
DEL MORAL, R., 1983. Initial recovery of subalpine vegetation on Mount St. Helens.
American Midland Naturalist, 109, pp. 72-80.
DEL MORAL, R., 1993. Mechanisms of primary succession on volcanoes: a view
from Mount St. Helens. In: J. MILES and D.H. WALTON, eds, Primary Succession on
Land. London: Blackwell Scientific Publications, pp. 79-100.
DEL MORAL, R. and GRISHIN, S., 1999. Volcanic Disturbances and Ecosystem
Recovery. In: L.R. WALKER, ed, Ecosystems of Disturbed Ground. Amsterdam:
Elsevier Science, pp. 137-160.
FELPETO, A., MARTÍ, J. and ORTIZ, R., 2007. Automatic GIS- based system for
volcanic hazard assessment. Journal of Volcanology and Geothermal Research,
166(2), pp. 106-116.
FLOOR, G.H., IGLESIAS, M., ROMAN-ROSS, G., CORVINI, P.F.X. and LENZ, M., 2011.
Selenium speciation in acidic environmental samples: Application to acid rain-soil
interaction at Mount Etna volcano.(Report). Chemosphere, 85(1), pp. 1664.
GLAZE, L., FRANCIS, P., SELF, S. and ROTHERY, D., 1989. The 16 September 1986
eruption of Lascar volcano, north Chile: satellite investigations. Bulletin of
Volcanology, 51(3), pp. 149-160.
GLOBAL VOLCANISM PROGRAM, 1982, Report on El Chichón (Mexico)
[Homepage of Smithsonian Institution], [Online]. Available:
http://dx.doi.org/10.5479/si.GVP.SEAN198205-341120 [July 1, 2015].
64
GRISHIN, S., 2011. Environmental impact of the powerful eruption of Sarychev Peak
volcano (Kuril Islands, 2009) according to satellite imagery. Izvestiya, Atmospheric
and Oceanic Physics, 47(9), pp. 1028-1031.
GRISWOLD, J., 2009, GIS Versus the Volcano [Homepage of ArcNews Online],
[Online]. Available: http://www.esri.com/news/arcnews/spring09articles/gis-
versus-the-volcano.html [February 12, 2015].
HALPERN, C.B. and HARMON, M.E., 1983. Early plant succession on the Muddy
River mudflow, Mount St. Helens. American Midland Naturalist, 110, pp. 97-106.
HEAD, E.M., MACLEAN, A.L. and CARN, S.A., 2013. Mapping lava flows from
Nyamuragira volcano (19672011) with satellite data and automated
classification methods. Geomatics, Natural Hazards and Risk, 4(2), pp. 119-144.
HUETE, A., YOSHIOKA, H., MIURA, T., KIM, H.J. and GAO, X., 2002. Inter-sensor
calibration of vegetation indices for monitoring and continuity studies of
ecosystem variability. In: J.A. SOBRINO, ed, Recent Advances in Quantitative
Remote Sensing. València: Publicacions de la Universitat de València, pp. 21-30.
INBAR, M., REYES ENRIQUEZ, A. and GRANIEL , J.H., 2001. Morphological changes
and erosion processes following the 1982 eruption of El Chichón volcano, Chiapas,
Mexico/Modifications géomorphologiques et processus d'érosion consécutifs à
l'éruption du volcan El Chichón, Chiapas, Mexico, en 1982. Géomorphologie:
relief, processus, environnement, 7(3), pp. 175-183.
JOYCE, K.E., BELLISS, S.E., SAMSONOV, S.V., MCNEILL, S.J. and GLASSEY, P.J., 2009.
A review of the status of satellite remote sensing and image processing techniques
for mapping natural hazards and disasters. Progress in Physical Geography, 33(2),
pp. 183-207.
KERR, R.A., 1983. El Chichón Climate Effect Estimated. Science, 219(4581), pp. 157-
157.
KRUEGER, A.J., 1983. Sighting of El Chichón eruption clouds with the Nimbus 3 Total
Ozone Mapping Spectrometer. Science, 220, pp. 1377-1379.
LAWRENCE, R., 2006. Remote Sensing of Vegetation Responses During the First 20
Years Following the 1980 Eruption of Mount St. Helens: A Spatially and Temporally
Stratified Analysis. In: V.H. DALE, F.J. SWANSON and C.M. CRISAFULLI, eds,
Ecological Responses to the 1980 Eruption of Mount St. Helen. New York: Dordrecht
Springer-Verlag, pp. 111-123.
LILLESAND, T.M., 2008. Remote sensing and image interpretation. 6th ed. / by
Thomas M. Lillesand, Ralph W. Kiefer, Jonathan W. Chipman.. edn. Hoboken, N.J. :
Chichester: Hoboken, N.J. : Wiley ; Chichester : John Wiley distributor.
LIU, J.G. and MASON, P.J., 2009. Essential Image Processing and GIS for Remote
Sensing. John Wiley & Sons.
LOCKWOOD, J.P. and HAZLETT, R.W., 2010. Volcanoes: Global Perspectives.
Chichester: Wiley & Sons.
65
LU, Z., RYKHUS, R., MASTERLARK, T. and DEAN, K.G., 2004. Mapping recent lava
flows at Westdahl Volcano, Alaska, using radar and optical satellite imagery.
Remote Sensing of Environment, 91(3-4), pp. 345-353.
MACÍAS, J.L., ARCE, J.L., MORA, J.C., ESPÍNDOLA, J.M., SAUCEDO, R. and MANETTI,
P., 2003. A 550-year-old Plinian eruption at El Chichón Volcano, Chiapas, Mexico:
Explosive volcanism linked to reheating of the magma reservoir. Journal of
Geophysical Research: Solid Earth, 108(B12).
MACÍAS, J.L., CAPRA, L., ARCE, J.L., ESPÍNDOLA, J.M., GARCÍA-PALOMO, A. and
SHERIDAN, M.F., 2008. Hazard map of El Chichón volcano, Chiapas, México:
Constraints posed by eruptive history and computer simulations. Journal of
Volcanology and Geothermal Research, 175(4), pp. 444-458.
MARTIN, R.S., MATHER, T.A., PYLE, D.M., WATT, S., DAY, J.A., COLLINS, S.J., WRIGHT,
T.E., AIUPPA, A. and CALABRESE, S., 2009. Sweet chestnut (Castanea sativa) leaves
as a bio-indicator of volcanic gas, aerosol and ash deposition onto the flanks of
Mt Etna in 2005-2007. Journal of Volcanology and Geothermal Research, 179(1-2),
pp. 107-119.
NATIONS ENCYCLOPEDIA, 2015, Encyclopedia of the Nations - Chiapas
[Homepage of Advameg Inc.], [Online]. Available:
http://www.nationsencyclopedia.com/mexico/Aguascalientes-M-
xico/Chiapas.html [June 30, 2015].
OPPENHEIMER, C., 2011. Eruptions that Shook the World. Cambridge: Cambridge
University Press.
OPPENHEIMER, C., 1998. Volcanological Applications of Meteorological Satellite
Data. International Journal of Remote Sensing, 19(5), pp. 2829-2864.
PARESCHI, M., CAVARRA, L., FAVALLI, M., GIANNINI, F. and MERIGGI, A., 2000. GIS
and Volcanic Risk Management. Natural Hazards, 21(2), pp. 361-379.
PRATA, A., 1989. Observations of volcanic ash clouds in the 10-12 μm window using
AVHRR/2 data. International Journal of Remote Sensing, 10(4-5), pp. 751-761.
PYLE, D.M., MATHER, T.A. and BIGGS, J., 2013. Remote Sensing of Volcanoes and
Volcanic Processes: integrating observation and modelling - introduction.
Geological Society, London, Special Publications, 380, pp. 1-13.
ROBOCK, A. and MATSON, M., 1983. Circumglobal transport of the el chichon
volcanic dust cloud. Science (New York, N.Y.), 221(4606), pp. 195-197.
ROSE, W.I. and KIMBERLY, P.G., 1995, Santa María: A Remote Sensing Perspective.
Available: http://www.geo.mtu.edu/volcanoes/santamaria/sanmar-remote/
[January 12, 2015].
ROSE, W. and DURANT, A., 2009. El Chichón volcano, April 4, 1982: volcanic cloud
history and fine ash fallout. Natural Hazards, 51(2), pp. 363-374.
66
SCOLAMACCHIA, T. and MACÍAS, J.L., 2005. Distribution and stratigraphy of
deposits produced by diluted pyroclastic density currents of the 1982 eruption of El
Chichón volcano, Chiapas, Mexico. Revista mexicana de ciencias geológicas,
22(2), pp. 159-180.
SIGURDSSON, H., CAREY, S. and FISHER, R., 1987. The 1982 eruptions of El Chichon
volcano, Mexico (3): Physical properties of pyroclastic surges. Bulletin of
Volcanology, 49(2), pp. 467-488.
SINGH, R.P., ROY, S. and KOGAN, F., 2003. Vegetation and temperature condition
indices from NOAA AVHRR data for drought monitoring over India. International
Journal of Remote Sensing, 24(22), pp. 4393-4402.
SKIDMORE, A.K., 2002. Environmental modelling with GIS and remote sensing.
London: London: Taylor & Francis.
SMETS, B., WAUTHIER, C. and D’OREYE, N., 2010. A new map of the lava flow field
of Nyamulagira (D.R. Congo) from satellite imagery. Journal of African Earth
Sciences, 58(5), pp. 778-786.
SMITHSONIAN INSTITUTION, 2013, Global Volcanism Program - El Chichón
[Homepage of Smithsonian Institution National Museum of Natural History Global
Volcanism Program], [Online]. Available:
http://volcano.si.edu/volcano.cfm?vn=341120 [June 25, 2015].
SONG, C., WOODCOCK, C.E., SETO, K.C., LENNEY, M.P. and MACOMBER, S.A.,
2001. Classification and Change Detection Using Landsat TM Data: When and
How to Correct Atmospheric Effects? Remote Sensing of Environment, 75(2), pp.
230-244.
TAKAHATA, S. and MIYAMA, K., 1980. Study on Plant Damage Caused by Mt. Usu
Eruption Using Remote Sensing , F. ACKERMANN, H. BAUER, G. KONECNY and G.
KUPFER, eds. In: ISPRS Congress Hamburg 1980 - Technical Commission VII:
Interpretation of Data, July 13-25 1980 1980, International Society for
Photogrammetry and Remote Sensing, pp. 907-915.
TRALLI, D.M., BLOM, R.G., ZLOTNICKI, V., DONNELLAN, A. and EVANS, D.L., 2005.
Satellite remote sensing of earthquake, volcano, flood, landslide and coastal
inundation hazards. ISPRS Journal of Photogrammetry and Remote Sensing, 59(4),
pp. 185-198.
USGS, 2009, VHP Photo Glossary: VEI [Homepage of U.S. Geological Survey],
[Online]. Available: http://volcanoes.usgs.gov/images/pglossary/vei.php [June 30,
2015].
USGS, 2014a, Pyroclastic flows and their effects [Homepage of U.S. Geological
Society], [Online]. Available: http://volcanoes.usgs.gov/hazards/pyroclasticflow/
[August 3, 2015].
67
USGS, 2014b, Frequently Asked Questions about the Landsat Missions. Available:
http://landsat.usgs.gov/band_designations_landsat_satellites.php [August 10,
2015].
USGS, March 18, 2015a, Volcanoes General FAQ. Available:
http://www.usgs.gov/faq/categories/9819/2689 [April 16, 2015].
USGS, 2015b, EarthExplorer [Homepage of U.S. Geological Society], [Online].
Available: http://earthexplorer.usgs.gov/ [June 24, 2015].
VAREKAMP, J.C., LUHR, J.F. and PRESTEGAARD, K.L., 1984. The 1982 eruptions of El
Chichón Volcano (Chiapas, Mexico): Character of the eruptions, ash- fall
deposits, and gasphase. Journal of Volcanology and Geothermal Research, 23(1),
pp. 39-68.
VERSTRAETE, M., 1994. The contribution of remote sensing to monitor vegetation
and to evaluate its dynamic aspects. In: F. VEROUSTRAETE and R. CEULEMANS,
eds, Vegetation, Modeling and Climatic Change Effects. The Hague: SPB
Academic Publishing, pp. 207-212.
VOLCANO DISCOVERY, 2015, El Chichón Volcano [Homepage of Volcano
Discovery], [Online]. Available: http://www.volcanodiscovery.com/el-
chichon.html [June 25, 2015].
WARD, M.D. and GLEDITSCH, K.S., 2008. Spatial Regression Models. California: Sage
Publications.
WEIER, J. and HERRING, D., 2000, Measuring Vegetation (NDVI & EVI) [Homepage
of NASA Earth Observatory], [Online]. Available:
http://earthobservatory.nasa.gov/Features/MeasuringVegetation/measuring_veg
etation_1.php [25 July, 2015].
XU, J., LIANG, Y. and WANG, Y., 2009. Examining effects of past volcanic eruptions
on forests of the Changbai Mountain with in situ and remote sensing observations.
YULIANTO, F., SOFAN, P., KHOMARUDIN, M. and HAIDAR, M., 2013. Extracting the
damaging effects of the 2010 eruption of Merapi volcano in Central Java,
Indonesia. Natural Hazards, 66(2), pp. 229-247.
ZARAH, S., NICOLAS, V., MINORU, U., THOMAS, S. and ASTRID, G., 2008. Preliminary
results of lava flow mapping using remote sensing in Piton de la Fournaise, La
Réunion island.
68
8 APPENDICES
8.1 APPENDIX 1: FALSE COLOUR IMAGES
Appendix 1 figures show the fully
pre-processed Landsat images
displayed in false-colour (green
band as blue, red band as green,
and near-infrared band as red).
Vegetation is highlighted in red.
69
8.2 APPENDIX 2: MORANS I TEST FOR SPATIAL AUTOCORRELATION
Appendix 2- Table shows the results of the Moran’s I test on the residuals from
ordinary least squares regression. All results indicate the variables are
significantly spatially autocorrelated.
Regression Equation
I-Value
Z-Value
p-value
1980-84 NDVIdiff
‘Tephra Only’ zone
0.673218
80.00
0.000
1984-90 NDVIdiff
‘Tephra Only’ zone
0.599687
71.26
0.000
1990-96 NDVIdiff
‘Tephra Only’ zone
0.487733
57.97
0.000
1996-2002 NDVIdiff
‘Tephra Only’ zone
0.537227
63.84
0.000
1980-84 NDVIdiff
Surges Only’ zone
0.821789
130.71
0.000
1984-90 NDVIdiff
Surges Only’ zone
0.716533
113.97
0.000
1990-96 NDVIdiff
Surges Only’ zone
0.555081
88.31
0.000
1996-2002 NDVIdiff
Surges Only’ zone
0.459762
73.15
0.000
1980-1984 NDVIdiff
Surges & Tephra’ zone
0.819986
167.52
0.000
1984-90 NDVIdiff
Surges & Tephra’ zone
0.670401
136.96
0.000
1990-1996 NDVIdiff
Surges & Tephra’ zone
0.481017
98.28
0.000
1996-2002 NDVIdiff
Surges & Tephra’ zone
0.390439
79.78
0.000
... AnnDe Schutter et.al (2015) noticed that 2007 eruption of Oldoinyo Lengai, North Tanzania, was severely affecting the vegetation by volcanic ash fall and they noted that the estimated recovery time varies from more than 5 years to less than 6 months with increasing distance from the volcano. JenniferRozier (2015) observed the patterns of vegetation change over time on eruption of El Chichón volcano. Her result is: indication of vegetation coverage that was reached within 20 years of the eruption, is showing rapid recovery. ...
Chapter
Full-text available
In any one year, approximately 60 volcanoes erupt on the Earth. Even though about 80% of these eruptions occur under the oceans, the terrestrial volcanic events are common enough to have major impacts on nearby vegetation, often over large areas (e.g., Bilderback, 1987). Volcanic activity both destroys or modifies existing vegetation and creates new geological substrates upon which vegetation can re-establish. The types of plants surviving and recovering after volcanic activity largely depend upon the type of activity that takes place, the nutrient content of material ejected or moved by the volcano, the distance from the volcanic activity, and the types of vegetation propagules that survive in place or are transported from adjacent areas. The resulting changes in the vegetation abundance and patterning can have dramatic effects on the social and economic conditions of the humans in the areas surrounding volcanoes.