2º CICLO DE ESTUDOS
MESTRADO EM SISTEMAS DE INFORMAÇÃO GEOGRÁFICA E ORDENAMENTO DO TERRITÓRIO
Retrieval and multi-year analysis of
Land Surface Temperature in Fogo
Island, Cape Verde – satellite remote
sensing and GIS synergy
Diogo Emanuel Gomes Vieira
M
2019Diogo Emanuel Gomes Vieira
Retrieval and multi-year analysis of Land Surface Temperature in
Fogo Island, Cape Verde – satellite remote sensing and GIS
synergy
Dissertação realizada no âmbito do Mestrado em Sistemas de Informação Geográfica e Ordenamento do Território, orientada pelo Professor Doutor António Alberto Teixeira Gomes
Faculdade de Letras da Universidade do Porto
Retrieval and multi-year analysis of Land Surface
Temperature in Fogo Island, Cape Verde – satellite remote
sensing and GIS synergy
Diogo Emanuel Gomes Vieira
Dissertação realizada no âmbito do Mestrado em Sistemas de Informação Geográfica e Ordenamento do Território, orientada pelo Professor Doutor António Alberto Teixeira
Gomes
Membros do Júri
Professora Doutora Ana Cláudia Moreira Teodoro Faculdade de Ciências - Universidade do Porto
Professor Doutor António Alberto Teixeira Gomes Faculdade de Letras - Universidade do Porto
Professor Doutor José Augusto Alves Teixeira Faculdade de Letras - Universidade do Porto
Contents
List of Figures ... 9 List of Tables ... 11 List of Abbreviations ... 12 Acknowledgments ... xv Abstract ... xvi Resumo ... xvii Introduction ... 2Chapter 1. Conceptual framing and state of the art ... 6
1.1. Remote sensing and volcanic studies ... 6
1.2. Remote Sensing ... 10
1.2.1. Historical Perspective ... 10
1.2.2. Electromagnetic Radiation ... 12
1.3. Thermal Infrared Remote Sensing ... 15
1.4. Land Surface Temperature ... 19
1.4.1. Concepts and retrieval ... 19
1.4.2. Landsat data used in LST studies: a brief synopsis ... 22
Chapter 2. Fogo Island... 30
2.1. Fogo island general setting ... 30
2.2. Fogo 2014-2015 eruption ... 35
Chapter 3. Data ... 40
3.1. MODIS ... 42
3.1.1. MODIS Land Surface Temperature and Emissivity Product ... 44
3.2. ASTER ... 46
3.2.1. ASTER Surface Kinetic Temperature Product... 48
7 3.4. Other Data ... 53 Chapter 4. Methodology ... 54 4.1. Pre-processing ... 54 4.1.1Atmospheric Correction ... 55 4.1.2. Radiometric Calibration ... 56 4.1.3. Cloud Screening ... 57
4.1.4. MODIS and ASTER LST Products ... 57
4.2. LSE retrieval ... 58
4.3. LST retrieval ... 59
4.4. LST products cross-comparison ... 61
4.5. Solar Radiation calculation ... 62
4.6. Regression analysis LST – Solar Radiation ... 62
Chapter 5. Results ... 63
5.1. Volcanic eruption ... 63
5.2. Multi-year analysis ... 73
5.3. Cross-comparison of LST products ... 82
5.4. Regression analysis Land Surface Temperature – Solar Radiation ... 84
Chapter 6. Discussion ... 85
6.1. volcanic eruption ... 85
6.2. Multi-year analysis ... 86
6.3. Cross-comparison of LST products ... 86
6.4. Regression analysis Land Surface Temperature – Solar Radiation ... 87
Conclusion ... 88
References ... 89
8 Appendix 1 ... 101 Appendix 2 ... 103 Appendix 3 ... 104 Appendix 4 ... 105 Appendix 5 ... 106 Appendix 6 ... 107 Appendix 7 ... 107 Appendix 8 ... 108 Appendix 9 ... 109 Appendix 10 ... 110 Appendix 11 ... 111 Appendix 12 ... 112 Appendix 13 ... 114 Appendix 14 ... 116 Appendix 15 ... 117
9
List of Figures
Figure 1. Methodological scheme ... 4
Figure 2. Example of sources of heat emission detectable by a satellite sensor in a volcanic setting. Tground and Tair refer to ground and air temperature. Mrad, Mconv and Mcond are represented as forms of heat loss by active lava forms, corresponding to radiation, convection and conduction, respectively. Adapted from Harris [3]. ... 6
Figure 3. Chronology of thermal satellite remote sensing crucial advances and stages. . 9
Figure 4. Electromagnetic Spectrum domains commonly used in remote sensing ... 13
Figure 5. Diagram of atmospheric windows ... 14
Figure 6. Infrared divisions in the electromagnetic spectrum ... 15
Figure 7. Factors contributing to radiant temperature ... 16
Figure 8. Peak emission wavelength variation related to temperature according to Wein’s Displacement Law ... 18
Figure 9.Number of LST publications using Landsat data over the years ... 24
Figure 10. A Cape Verde archipelago B Fogo Island – Sentinel-2A scene (7 April 2017) C Areas affected by 1951, 1995, and 2014-2015 lava flows ... 32
Figure 11. A - Digital Elevation Model (DEM) of Fogo island; B - N-S cross-section elevation profile; C - W-E cross-section elevation profile ... 34
Figure 12. Google Earth/Digital Globe images of Chã das Caldeiras and Pico do Fogo (A) before the eruption (4 November 2012) and (B) after the eruption (3 February 2016) ... 36
Figure 13. Set of pre- (4 November 2012) and post-eruption (3 February 2016) Google Earth satellite images ... 38
Figure 14. Evolution and occurrences of 2014-2015 Fogo eruption. ... 39
Figure 15. Disposition of Landsat-7, Landsat-8, MODIS and ASTER spectral bands in the electromagnetic spectrum ... 42
Figure 16. MODIS instrument components ... 43
Figure 17. ASTER subsystems: TIR, VNIR, and SWIR ... 46
Figure 18. Landsat-series satellites timeline... 49
10
Figure 20. Flowchart of LST retrieval and cross-comparison of LST products... 55 Figure 21. Flowchart of Landsat-8 cloud mask processing ... 57 Figure 22. Distribution of maximum, minimum, and mean values of LST for Fogo island – ASTER and Landsat data ... 67 Figure 23. Distribution of maximum, minimum, and mean values of LST for the volcanic complex – ASTER and Landsat data ... 68 Figure 24. (A) Landsat-8 LST values – 25-02-2014; (B) Landsat-8 LST values – 29-03-2014; (C) Landsat-8 LST values – 16-05-29-03-2014; (D) Landsat-8 LST values – 01-06-29-03-2014; (E) Landsat-8 LST values – 17-06-2014; (F) Landsat-8 LST values – 19-07-2014. ... 70 Figure 25. (A) Landsat-8 LST values – 23-10-2014; (B) Landsat-8 LST values – 24-11-2014; (C) Pansharpened false color combination (7-6-5); (D) ASTER AST_08 LST values – 21-12-2014; (E) Landsat-8 LST values – 11-01-2015; (F) Landsat-8 LST values – 28-02-2015. Note: Blank spaces in figures A, B, E, and F represent cloud masks, whereas in D represents bad quality pixels. ... 71 Figure 26. (A) Landsat-8 LST values – 01-04-2015; (B) Landsat-8 LST values – 17-04-2015; (C) Landsat-8 LST values – 03-05-17-04-2015; (D) Landsat-8 LST values – 19-05-17-04-2015; (E) ASTER AST_08 LST values – 12-06-2015; (F) Landsat-8 LST values – 20-06-2015. ... 72 Figure 27. Distribution of maximum, minimum, and mean values of LST for Fogo island – ASTER and Landsat data ... 76 Figure 28. Distribution of maximum, minimum, and mean values of LST for the volcanic complex – ASTER and Landsat data ... 77 Figure 29. (A) Landsat-7 LST values – 27-08-2002; (B) ASTER AST_08 LST values – 27-08-2002; (C) Landsat-7 LST values – 17-12-2002; (D) ASTER AST_08 LST values – 17-12-2002... 79 Figure 30. (A) Landsat-8 LST values – 13-12-2015; (B) ASTER AST_08 LST values – 21-12-2015; (C) Landsat-8 LST values – 06-06-2016; (D) ASTER AST_08 LST values – 14-06-2016... 81
11
List of Tables
Table 1. Main divisions of the Electromagnetic Spectrum ... 12
Table 2. Publications of LST studies applying Landsat data ... 23
Table 3. Publications of LST studies applying Landsat data for volcanic areas ... 27
Table 4. Acquisition date of used satellite imagery... 41
Table 5. MODIS MYD11_L2 LST product characteristics ... 46
Table 6. ASTER AST_08 LST product characteristics ... 48
Table 7. Characteristics of Landsat-8, Landsat-7, and Landsat-5 satellites ... 52
Table 8. Descriptive statistics of ASTER and Landsat LST values ... 65
Table 9. Descriptive statistics of ASTER and Landsat LST values ... 73
Table 10. Cross-comparison of MODIS LST product and resampled Landsat LST data for Fogo island ... 82
12
List of Abbreviations
ASTER Advanced Spaceborne Thermal Emission and Reflection Radiometer
AVA ASTER Volcanic Archive
CBEM Classification-Based Emissivity Method
DN Digital Number
DOS Dark-Object Subtraction
EOS-1 Earth Observation One
ETM Enhanced Thematic Mapper
ETM+ Enhanced Thematic Mapper Plus
GHF Geothermal Heat Flux
GSW Generalized Split-Window
HDR Heat Discharge Rate
HRV High Resolution Visible
JERS-1 Japan Earth Resources Satellite 1
LSE Land Surface Emissivity
LST Land Surface Temperature
LWIR Long-Wave Infrared
MIR Midwave Infrared
MISR Multi-Angle Imaging Spectroradiometer
MODIS Moderate Resolution Imaging Spectroradiometer
MODVOLC MODIS Volcano Thermal Alert System MSS Multispectral Scanner
NASA National Aeronautics and Space Administration
NBEM NDVI-Based Emissivity Method
NDVI Normalized Difference Vegetation Index
NIR Near Infrared
OLI Operational Land Imager
RHF Radiative Heat Flux
RBV Return Beam Vidicon
13 SNDVITHM Simplified Normalized Difference Vegetation Index Threshold Method
SPOT Satellite pour l’Observation de la Terre
SW Split-Window
SWIR Short-Wave Infrared
TES Temperature Emissivity Separation
TIR Thermal Infrared
TIRS Thermal Infrared Sensor
TM Thematic Mapper
TOA Top of Atmosphere
USGS United States Geological Survey
VIS Visible
14
Declaração de honra
Declaro que a presente dissertação é da minha autoria e não foi utilizada previamente noutro curso ou unidade curricular, desta ou de outra instituição. As referências a outros autores (afirmações, ideias, pensamentos) respeitam escrupulosamente as regras da atribuição, e encontram-se devidamente indicadas no texto e nas referências bibliográficas, de acordo com as normas de referenciação. Tenho consciência de que a prática de plágio e auto-plágio constitui um ilícito académico.
Porto, 18 de Novembro de 2019
Acknowledgments
This dissertation is far from being an individual effort, this being said I would like to thank those that contributed to it.
First and foremost, I would like to thank my supervisor Professor Alberto Gomes for his support, valuable insights, commitment and companionship. To Professor Ana Cláudia Teodoro for her effort, encouraging words and determination that served as an example. Their combined experience, guidance and suggestions were crucial, without ever constraining my ideas, but rather mature them. I would also like to thank them for the careful revision of the text.
To my parents and sister for all their continuous support, advices, and patience. To Joana for her unconditional support and input. To all my friends, particularly Bruno and Inês for their companionship and ideas. To Professor Ricardo Baptista for his words of advice and comradeship. To António Costa for his incentive and clever insights. And lastly to all teachers I have had the opportunity to share a classroom with through the years, as I am certain they contributed to spice up my curiosity and willingness to learn and further develop my education.
Abstract
Land Surface Temperature (LST) is an important parameter related to land surface energy transfers and balance that changes continuously through time. Studying LST dynamics of areas prone to volcanic and geothermal activity has both environmental and socio-economical interest as it can disclose unknown underground heating activity and sources. Adding to this, volcanic materials deposited throughout an eruption transform the landscape and alter LST measurements. This thesis aims to assess variations of satellite-derived LST and to detect spatial patterns and thermal anomalies in Fogo island by means of a pluriannual analysis spanning from 2001 to 2016 and including 2014-2015 eruption. LST data was retrieved from Landsat images by applying a single-channel algorithm and deriving emissivity values using an NDVI-based methodology. The absence of in situ measurements is compensated by using MODIS and ASTER LST datasets for cross-comparison. Additionally, regression analysis tested the potential influence of solar radiation in LST values. Analysis of retrieved LST data revealed an expected dynamic of temperature values and patterns following the evolution of the eruption, with higher temperature values being located inside the caldera. High temperature values were also founded on the south-facing flank of the caldera. Although spatial patterns observed on the retrieved data remained roughly the same during the time period considered, temperature values changed over time. The observation of thermal anomalies induced by geothermal activity before and after the eruptive event is not evident. Moreover, LST data also suggests that solar radiation can play a substantial role in temperature values. Areas affected by lava flows of previous eruptions are easily recognized due to well-defined lower LST spatial patterns. A multi-year analysis revealed not only that temperatures maintained a regular behaviour throughout the years as well as what appears to be a cyclic temperature trend.
Keywords: Land Surface Temperature, Landsat, Thermal Infrared data, Geothermal
Resumo
A Temperatura da Superfície Terrestre é um importante parâmetro relacionado com transferências e balanco de energia da superfície terrestre que sofre alterações ao longo do tempo. O estudo das dinâmicas de TST em áreas propensas à atividade vulcânica ou geotérmica reveste-se de interesse ambiental e socioeconómico já que pode revelar atividade e fontes de calor subterrâneo desconhecidas. Adicionalmente, os materiais vulcânicos depositados durante erupções transformam a paisagem e alteram os valores de TST. A presente dissertação procura aferir variações de TST derivada de imagens de satélite e detetar padrões espaciais e anomalias térmicas na ilha do Fogo através de uma análise plurianual desde 2001 a 2016, incluindo a erupção de 2014-2015. Os dados de TST foram obtidos a partir de imagens Landsat através da aplicação de um algoritmo single-channel e cálculo dos valores de emissividade usando uma metodologia baseada no NDVI. A ausência de medições in situ é compensada pelo uso de dados TST dos sensores MODIS e ASTER para comparação. Adicionalmente, foi e efetuada uma análise regressiva que testou o grau de influência da radiação solar nos valores de TST. A análise aos dados LST processados revelou uma dinâmica de temperaturas e padrões que seguiu a evolução da erupção, com os valores de temperatura mais elevados a localizarem-se na caldeira. Ainda que os padrões espaciais observados se tenham mantido durante o período considerado, os valores de temperatura alteraram-se com o tempo. A observação de anomalias térmicas induzida por atividade geotérmica não é evidente. Adicionalmente, dados de TST sugerem que a radiação solar influencia de forma determinante as temperaturas. Áreas afetadas por escoadas de lava em erupções passadas são facilmente reconhecidas. Uma análise plurianual revelou não só que as temperaturas mantêm um comportamento regular, como também o que parece ser uma tendência cíclica de aumento de temperatura.
Palavras-chave: Temperatura da superfície terrestre, Landsat, Dados Infravermelhos
Introduction
Nowadays remote sensing is commonly used for different purposes such as disaster relief, military intelligence, security, natural resources management, climate change studies, urban planning, and precision agriculture, having greatly impacted both commercial and scientific uses [2, 5]. Despite not being solely dedicated to Earth observation, remote sensing is often seen and referred to as only having such purpose. This common mistake speaks volumes about the importance of Earth observation and how it influenced the evolution of remote sensing, its techniques and applications, central to a large number of scientific fields [1]. Scientific studies benefit from satellite remote sensing as it allows scientists and researchers to study the Earth’s surface, its dynamics and phenomena from a vantage point, regularly, enabling continuous monitoring at different scales of almost any site on the planet, regardless of how remote or inaccessible they are [2].
Volcanology and associated fields of study have benefited greatly from remote sensing and its advances, being particularly associated with thermal infrared (TIR) remote sensing, a fairly recent incrementation to the study volcanoes and eruptive episodes which has progressed in the last couple of decades with the development of new methodologies and broader data accessibility [3, 6]. TIR satellite-driven data is used in volcanological studies for monitoring eruptions and the analysis of a myriad of volcanic features and phenomena such as ground deformation, ash and gas clouds, thermal monitoring of lava flows and ground heat, volcanic deposits, heat flux, and thermal anomalies [3, 7, 8]. Understanding how complex volcanic systems work and their dynamics are not only of interest for science, but also for the populations established in volcanic settings as it could help mitigate risks related with eruptive episodes [9].
Realizing the importance of remote sensing for the study of volcanic eruptions, patent in various studies conducted in numerous volcanic regions across the planet, the present thesis sets to study the latest eruption of Pico do Fogo volcano (Cape Verde), dated from 2014-2015, by analysing multi-year Land Surface Temperature (LST) Landsat-derived data. Due to the lack of in-situ and ground-based measurements for validation purposes, ASTER and MODIS LST-products, AST_08 and MYD11_L2 respectively, were used for cross-comparison and cross-validation, as in Qin, et al. [10]. Additionally, solar radiation values were calculated and used as input for regression
analysis, in combination with LST data, in order to establish the degree to which these variables were related, and to determine to what extent solar radiation acted as an input and influenced surface temperature readings, in an effort to facilitate the distinction between possible sources of heat (solar radiation or geothermal/volcanic activity).
This approach provides insights into LST values spanning a volcanic context, helping to detect thermal hotspots and lava flows, constituting a volcanic activity monitoring tool and an efficient way to monitor thermal anomaly areas on volcanic islands.
Fogo island and the Cape Verde archipelago have been studied throughout the years with various publications focusing on past eruptions [11, 12], a large flank collapse [13-15], seismic activity [16-18], and volcanic characteristics and origins [19-22]. The latest eruptive event was studied in respect to lava flow hazard modelling, measuring, and estimating predictive scenarios [23-25], as well as reporting the evolution of the eruption and describing the impacts caused by it in the surrounding landscape [26, 27] While the temperature of vents and fumaroles is referred in Worsley [26] and hotspot detection in MODIS and SEVIRI images is used in Cappello, et al. [23], only Vieira, et al. [28] mentions LST values and how it evolved throughout the eruptive event. The main drives to study Fogo island and its volcanic setting were continue and further develop the work already produced in Vieira, et al. [28], analysing a broader range of data, the readiness and accessibility of satellite data, and ultimately the recent eruption which in itself provides a rare opportunity for researchers to study the dynamics of the volcanic setting both before, during, and after an eruptive episode. Regarding the objectives, this thesis aims to answer the following questions:
1. Assessing the utility of Landsat TIR data to monitor volcanic settings through the retrieval of LST values.
2. Can thermal anomalies and LST readings act as a proxy for geothermal activity? 3. To what extent can satellite images help explain the evolution and dynamics of a
volcanic eruption or support the characterisation of a volcanic setting?
The collection and consultation of bibliography, either books and scientific papers, on TIR remote sensing, its applications to volcanological studies, LST-retrieval algorithms, and regarding Fogo island characteristics and previous studies preceded the manipulation and processing of any satellite image (Figure 1). Having decided what
imagery to use and after collecting it, the workflow regarding the retrieval of LST values from Landsat images was defined and incorporated in a dedicated ArcGIS toolbox allowing for a more expedite processing of satellite data. Following the processing of all LST data, including ASTER and MODIS imagery for cross-comparison and validation purposes, solar radiation data was derived. Afterwards, both LST and solar radiation values were extracted and used as input in regression analysis.
Different software programs were used to process all collected and derived data. While ArcGIS was used more extensively to calculate Land Surface Emissivity (LSE) and LST data, derive a detailed Digital Elevation Model (DEM) for Fogo island, and compute solar radiation values, both QGIS and MODIS Reprojection Tool Swath (MRTSwath) were used in ancillary processing of satellite images. Regression analysis was computed in SPSS Statistics software package.
Figure 1. Methodological scheme
Considering the thematic of this dissertation and the methodology used the document is divided in seven chapters. Chapter 1 introduces concepts regarding remote sensing and TIR remote sensing, volcanic studies, and LST studies and algorithms. Additionally, presents the current state of the art and advances on volcanological studies and LST applications. Chapter 2 focuses on the general description and origins of Cape Verde archipelago, the physical and geological characterisation of Fogo island and its
1. Bibliography 2. Satellite Imagery 3. Processing
Volcanological studies TIR remote sensing LST algorithms Landsat-5 Landsat-7 Landsat-8 MODIS (Aqua) ASTER (Terra) Pre-processing LSE retrieval - CBEM LST retrieval - SNDVITHM Solar Radiation Regression analysis 4. Results LST cross-comparison
Cape Verde and Fogo island
volcanic features, ending with a detailed report of the latest eruption. Chapter 3 is dedicated to all data used, describing with particular detail satellite imagery used. Chapter 4 showcases thoroughly the various steps of the methodology applied. Chapter 5 presents LST, solar radiation and regression analysis results, as well as cross-comparison and validation figures, being all discussed in Chapter 6. Final conclusions and considerations about future work are presented in Chapter 7.
Chapter 1. Conceptual framing and state of the art
1.1. Remote sensing and volcanic studiesVolcanic eruptions and magmatic activity are precursors and responsible for a variety of other geophysical phenomena, such as earthquakes swarms, terrain deformation, and produce considerable amounts of gaseous outputs and radiation in the form of heat, which ultimately interact and affect both the surface and the atmosphere [3]. Possible sources of radiating heat during an eruption are displayed in Figure 2Erro! A
origem da referência não foi encontrada.. In this sense, volcanic areas are studied and
monitored by a myriad of devices, ranging from seismometers to spectrometers and radiometers. The advent of TIR remote sensing added satellites to this list, despite no satellite mission has ever been exclusively oriented to volcanic exploration, which poses some constrains in terms of the sensitivity and capabilities of instruments [3, 29]. Aside from the obvious scientific purpose of studying volcanoes, negative impacts and the danger posed by eruptions often affecting large swaths of population globally add importance to their study and reinforce the importance of active monitoring [30].
Figure 2. Example of sources of heat emission detectable by a satellite sensor in a volcanic setting. Tground and Tair refer
to ground and air temperature. Mrad, Mconv and Mcond are represented as forms of heat loss by active lava forms,
corresponding to radiation, convection and conduction, respectively. Adapted from Harris [3].
Thermal remote sensing experienced several advances as a discipline [4]. It was only during the last couple of decades of the twentieth century that satellite remote
sensing was used to study active volcanoes, associated structures and processes [30]. Volcano monitoring has evolved significantly, both in terms of available data, parameters and techniques. While initially being constrained by coarser weather satellite data, presently high-resolution Short-Wave Infrared (SWIR) and TIR data is commonly used [31]. Progressing from elementary detection of thermal features to the retrieval of various parameters, temporal analysis, monitoring of thermal dynamic of volcanic features, mineral mapping, ground deformation analysis, and advanced dynamic modelling [3]. First uses were related to the identification of thermal anomalies linked with volcanic activity using weather satellites in the 1960’s and 1970’s [32, 33]. Only more than a decade later studies with new insights and purposes other than the basic detection of hot spots and thermal anomalies were published [34-37]. By deriving parameters related to volcanic activity, such as temperature and heat flux, from satellite imagery, these works were paramount to establish remote sensing as a valid technique for volcano monitoring and influence other researchers to further develop their previous advances.
In the early 1990’s retrieval of new parameters related to heat interactions was published. Estimation of thermal budgets [38], lava flow temperatures [39], and crust thickness and surface temperature variation [40] were the new advances, increasing the complexity of parameters being derived. Later, realizing the importance of continuous observation of volcanoes due to their dynamic during eruptions researchers started using meteorological satellites due to their fairly immediate temporal resolution, while applying and redesigning techniques previously used in satellites with finer spatial resolution. Data spanning an entire eruptive event started to be used, taking advantage of daily imagery acquisitions by some satellites, enabling retrospective analyses [41] and advances on live monitoring and automated detection and monitoring of volcanoes were made preceding later developed automated algorithms like MODIS Volcano Thermal Alert System (MODVOLC) [42, 43], archives like ASTER Volcano Archive (AVA) [44] and networks of sensor systems as used on Earth Observation One (EO-1) satellite [45].
The turn of the century brought with it an influx of thermal data and new platforms with enhanced TIR capabilities, such as multiple thermal bands (i.e. MODIS and ASTER), that could be used for volcanological studies. Terra, Aqua, 7, Landsat-8, EO-1 and the recently launched Sentinel-3 form a suite of valid instruments that can be used in synergy with ground-based measurements and other data. The evolution of satellites, sensors, data and techniques contributed to the use of parameters in advanced
modelling of volcanological processes [23-25] in recent years [3]. The constant incrementation of technological advances and maturation of methodologies led to an increase in the number of scientific publications concerning volcanic studies using satellite thermal remote sensing techniques [6]. An extensive listing of satellite sensors used in volcanological studies is presented in Appendix 1.
As described in Harris [3] the majority of volcanological studies are applied to four main themes:
1. Study of the thermal dynamic of volcanic features.
2. Evaluating surface roughness, deformation and volume changes. 3. Mapping and spectral analysis of volcanic deposits.
4. Study of volcanic clouds/plumes.
This work focused on the study of the thermal dynamics of Pico do Fogo, Cape Verde, and its adjacent areas by using satellite-derived LST as its driving parameter. When compared to more conventional volcanic studies, the use of remote sensing techniques offers researchers various advantages. Satellites enable the continuous collection of data, not only spanning entire eruptive events but also contributing to reliable uninterrupted monitoring of volcanic areas with daily observations from low resolution and geostationary satellites, complemented by less frequent polar orbiting satellites overpasses. Such broad and consistent data collection, backed by an extensive range of acquisition sources, contributes to creation of standardized time-series archives [3]. Satellite observations can also overcome the visibility constrains imposed by ash and gas plumes or systematic cloud cover that affect the optical region of the electromagnetic spectrum using as an alternative radar images [30]. Other relevant benefits are both the convenience of studying the entire extent of a volcanic area within a single image and having the opportunity to use multispectral or even hyperspectral data, corresponding to various regions of the electromagnetic spectrum, ranging from the visible (VIS), to the Near Infrared (NIR), Medium Infrared (MIR), Thermal Infrared (TIR) domains [3]. The increasing use of satellite data is also favoured by easily accessible calibrated data and user-friendly formats.
Having presented the origins and progress of remote sensing of volcanic settings and related features the following sections will resume the development of remote sensing and cover the principles of thermal remote sensing and LST satellite-retrieval. Additionally, a collation of scientific works using Landsat-derived LST data is analysed.
Figure 3. Chronology of thermal satellite remote sensing crucial advances and stages. Adapted from Harris [3].
1.2. Remote Sensing
1.2.1. Historical Perspective
Remote sensing can be simply defined as “…the gathering of information at a distance…” [2] although a more complex definition tends to be more correct and precise such as “… the technique of obtaining information about objects through the analysis of data collected by special instruments that are not in physical contact with the objects of investigation.” [46]. Having numerous applications it is essential to state in advance that the type of remote sensing being addressed in this work is the one related with Earth’s observation, as so the definition given in Campbell and Wynne [2] is more accurate to this case “Remote sensing is the practice of deriving information about the Earth’s land and water surfaces using images acquired from an overhead perspective, using electromagnetic radiation in one or more regions of the electromagnetic spectrum, reflected or emitted from the Earth’s surface.”.
As it is acknowledged today remote sensing has evolved expressively since the twentieth century with the advent of hyperspectral imagery, new acquisition and image processing techniques and more advanced satellites and sensors being designed [5]. In Campbell and Wynne [2] a list with different definitions can be found. As for the term itself it was coined by Evelyn Pruitt, U.S. Navy’s Office of Naval Research, as she acknowledged that the term aerial photography was insufficient to describe imagery acquired out of the visible domain of the electromagnetic spectrum [2, 46].
Some of the advantages of using remote sensing are the large areas covered in each acquired image, compared to traditional methods, regular acquisition of new data and the practicality and commodity of the acquisition in itself since data is acquired autonomously [5]. The surge of new technologies and further development of others improves data acquisition and processing time.
Remote sensing roots can be traced back to the origins of both photography and aviation. Successive evolutions on these two fields provided the conditions for the proliferation of airborne cameras and the evolution of aerial photography, often seen as a precursor of Earth observation remote sensing [46].
After the World War II, in 1947, March 7, the first image was taken from space making use of a V-2 rocket, being this moment seen as the foundation of space-based Earth-observations [46]. These efforts further extended in the following years with the
launch of several satellites which provided constant breakthroughs, shaping today’s satellites. The first satellite image was captured by Explorer-6 in 1959, in the same year Explorer-7 was launched as an effort to study the Earth’s radiation balance. TIROS-1, the first meteorological satellite was launched in 1960. During the Cold War U.S. military used CORONA strategic reconnaissance satellite, a platform that may well be understood has an important precursor to current satellites [2].
In the following decades, numerous satellites were launched into orbit by several countries in order to study the Earth’s processes and systems. The lifetime of both satellites and sensors grew, their spectral, radiometric, temporal and spatial resolution improved and missions carried on for longer periods of time [46]. Active sensors capable of emitting radiation and recapture the signal reflected from sensed objects were introduced, no longer being dependent of reflected radiation these new instruments fomented the advent of new fields of study and paved the way for innovative applications. Another important event for Earth-observation and collection of imagery was the launch of the first Landsat satellite in 1972, establishing the Landsat Program as one of the most important satellite missions [5]. The Landsat mission will be addressed in coming chapters. Landsat and other satellite programs showed that space-based observations are of great value to the study of both Earth and ocean surface, its dynamics and interaction with other means, contributing to the establishment of new missions in different countries, the introduction of commercial satellites, some with submeter resolutions, and the creation of partnerships between different international institutions and corporations (e.g. NASA – ESA) [47].
Nowadays a multiplicity of satellites with different characteristics orbits the Earth, some with a payload of multiple sensors, either active or passive, and hundreds of spectral bands since the wake of the 1980’s with the evolution of hyperspectral remote sensing, enabling new applications related to specific well-defined electromagnetic domains with narrow bandwidths, allowing for further specialized applications.
To put this evolution into perspective and to understand how far modern Earth observation is from its humble origins, ESA estimates that since the launch of Sputnik in 1957 and January 1, 2008, around 5600 satellites were launched into the Earth’s orbit [5]. The majority of which contributed in some way to the Earth’s study and showed the importance of remote sensing to it.
1.2.2. Electromagnetic Radiation
Electromagnetic radiation (EMR) is crucial to remote sensing as it is used in remote sensing as a way of gathering information about objects or large surfaces, assessing its changes [48]. Sensors detect EMR that reaches their lenses, being this either emitted and/or reflected by objects or the surface, which is ultimately dependent of the inherent characteristics of the object/surface, emissivity or reflectivity respectively [1]. As objects interact differently with radiation, each has a particular spectral signature which eases their identification through satellite images [5]. Radiation travel in form of waves and integrate the electromagnetic spectrum (Figure 4). Its associated values of wavelength and frequency vary accordingly to its position in the spectrum, establishing different categories of EMR (Table 1Erro! A origem da referência não foi encontrada.) [2, 5].
The visible (VIS), infrared (IR) and microwave are the most commonly used wavelengths in remote sensing. The IR domain of the electromagnetic spectrum will be further discussed in following sections.
Table 1. Main divisions of the Electromagnetic Spectrum Adapted from Harris [3] and Campbell and Wynne [2]
Division Wavelength limits
Gamma rays < 0.03 nm X-rays 0.03-300 nm Ultraviolet radiation 0.30-0.38 µm Visible light 0.38-0.72 µm Infrared radiation | Near infrared 0.72-1.10 µm | Shortwave infrared 1.10-3.00 µm | Midwave infrared 3.00-5.00 µm | Longwave infrared 5.00-20.00 µm | Far infrared 20.00-1.000 µm Microwave radiation 1 mm-30 cm Radiowaves ≥ 30 cm
Figure 4. Electromagnetic Spectrum domains commonly used in remote sensing Adapted from Rees [1]
Sensors can be categorized as active or passive. This labelling is related to what is being measured. Active sensors – i.e. radars – emit radiation and capture the reflected signal from its target feature. Whereas passive sensors record naturally emitted or reflected radiation.
Independently of being emitted or reflected EMR interacts with the atmosphere, particles and aerosols present in it, Earth’s surface and other objects as it cruises the atmosphere both to and from the Earth’s surface [5]. Possible interactions include absorption, reflection, refraction, transmission, and scattering [2, 49]. These interactions can cause the signal to be altered, often producing an additive effect, and are wavelength dependent, as so the importance of radiometric calibration and atmospheric correction, both proving to have a crucial role in preserving the quality of collected data and its consistency [50]. The variable concentration of gases present in the atmosphere, suspended particle size, incidence angle, and surface’s spectral characteristics and roughness influence the behaviour of radiation and both how and at which scale it interacts when in contact with surfaces [2].
Other important aspect of EMR is atmosphere’s permeability, as it can seriously interfere or even stop radiation from penetrating it and reaching the surface due to absorption by gases and suspended particles present in its composition. Such interactions with the atmosphere can hamper the effectiveness of remote sensing for some uses, while being of interest for others such as atmospheric studies. The regions of the spectrum where the atmosphere is more permeable, and radiation is easily transmitted are known as atmospheric windows, with some of the most important ones being in the VIS, NIR, and TIR domains [2]. The position of some atmospheric windows in the electromagnetic spectrum is depicted in Figure 5.
Figure 5. Diagram of atmospheric windows
1.3. Thermal Infrared Remote Sensing
The use of infrared and thermal data has been increasing, with more processing techniques being developed, an increasing number of sensors incorporating such capabilities being built and new applications to thermal data being discovered [4]. The key aspect that differentiates TIR remote sensing from optical remote sensing is what is being measured, as remote sensing in other domains of the electromagnetic spectrum measures radiation reflected by a body’s surface, while TIR remote sensing measures radiation emitted by objects and surfaces [8]. This is only possible as every object or surface with temperature above absolute zero (0 K or -273.15 °C) emits thermal radiation [1]. Lacking a strict definition of the TIR domain in the electromagnetic spectrum several authors classify it with slight discrepancies between the 5.0-20.0 μm interval [2, 3, 8]. Also referred to as Longwave Infrared (LWIR), TIR radiation in conjunction with NIR, SWIR and MIR regions of the electromagnetic spectrum are part of the infrared domain [3]. Far infrared is also referred by some authors as part of the infrared region, either including or not TIR radiation [2, 48, 51]. Figure 6 depicts some divisions within the infrared domain of the electromagnetic spectrum. Though TIR radiation assumes a preponderant role in volcanological remote sensing, both NIR, SWIR and MIR bandwidths are often used as well in order to sense higher temperature ranges following well-defined physical principles discussed below. Despite not being affected by scattering as other regions with narrower wavelengths TIR radiation is still constrained by atmospheric absorption, thus atmospheric windows between 3-5 µm and 8-14 µm ranges enable the use of both MIR and TIR radiation to sense Earth’s surface [2, 4].
TIR data applications range from the study of coal and forest fires, to urban heat islands, mapping of geologic deposits and materials, energy balance studies, energy
Figure 6. Infrared divisions in the electromagnetic spectrum Adapted from Campbell and Wynne [2]
transfers dynamics, calculation of land and sea surface temperature, soil moisture monitoring, forest fires detection, and a multiplicity of volcanological studies, such as ground deformation, detection of thermal anomalies, mapping of volcanic deposits, analysis of volcanic plumes, monitoring of thermal dynamics, radiative and geothermal heat flux measurements and thermal analysis of volcanic features [8, 10].
Thermal sensors measure emitted radiation of ground objects or surfaces, which translates to radiant temperature, and is influenced by both kinetic temperature, also known as true temperature, and emissivity (ε) [4]. Figure 7 shows factors modelling radiant temperature. To better understand and characterise emitted radiation some considerations regarding theorical background of TIR remote sensing will be presented.
The basis of thermal remote sensing is the Planck Function, also known as Planck’s blackbody radiation law [52]. Developed by Max Planck, it describes how electromagnetic radiation emitted by a blackbody in thermal equilibrium behaves considering a defined wavelength and temperature. A blackbody is a theoretical inexistent object or surface that absorbs and then emits all incident radiation. Spectral radiance is given by:
Figure 7. Factors contributing to radiant temperature Adapted from Prakash [4]
𝐵𝜆(𝑇) = 𝑐1 𝜆5[𝑒𝑥𝑝(𝑐2
𝜆𝑇)]−1
(Eq. 1.1)
Where 𝐵𝜆(𝑇) corresponds to spectral radiant exitance (W m-2 μm-1 sr-1) of a
blackbody at defined wavelength 𝜆 (μm) and temperature 𝑇 (K), 𝑐1 and 𝑐2 are physical
constants (1.19104 x 108 W μm4 m-2 sr-1; 14387.7 μm K).
Values of spectral radiance are higher for blackbodies with higher temperatures increasing with temperature increase, no matter the wavelength. However, as temperatures increase, wavelength at which peak spectral radiance occurs decreases. This variation, known as Wein’s Displacement Law, illustrated in Figure 8, is given by:
𝜆𝑚 = 𝐴
𝑇 (Eq 1.2)
Where 𝜆𝑚 represents peak radiation wavelength (μm), 𝑇 blackbody temperature
(K), and 𝐴 is Wien’s constant (2897.9 μm K).
As a result, peak spectral radiance for regular surfaces on Earth, approximately 288 K on average, occurs in the thermal infrared region of the electromagnetic spectrum, whereas for crusted lava surfaces with temperatures between 250 °C and 800 °C, or for bodies at magmatic temperatures (higher than 1000 °C), peak spectral radiance is registered in MIR and SWIR regions respectively [3].
Stefan-Boltzmann Law is also part of the theorical background of thermal remote sensing. It defines the total radiation emitted by a blackbody considering its absolute temperature:
𝑀𝑟𝑎𝑑 = 𝜎𝑇4 (Eq 1.3)
𝑀𝑟𝑎𝑑 corresponds to radiant flux density (W m-2), 𝜎 to the Stefan-Boltzmann constant (5.6697 x 10-8 W m-2 K-4), and 𝑇 is blackbody absolute temperature.
It demonstrates that radiating surfaces with higher temperatures emit radiation at a higher magnitude that cooler ones, according to a power-law distribution.
Inversion of the Planck Function allows for the calculation of temperature values for a blackbody:
𝑇 = 𝑐2
𝜆 ln(𝑐1𝜆−5 𝐵𝜆(𝑇)+1)
Conversely to the theorized assumptions of Planck’s law, no surface or object in nature acts as a blackbody, this is, emit all the radiation they absorb. Emissivity (ε) is the intrinsic capacity a material has to emit radiation compared with the same capacity of a blackbody at the same wavelength and temperature [4].
𝜀(𝜆) = 𝐵𝜆(𝑇)
𝐵𝜆(𝑇)𝐵𝐵 (Eq 1.5.)
Where 𝜀(𝜆) is emissivity at a determined wavelength, 𝐵𝜆(𝑇) spectral radiance, and 𝐵𝜆(𝑇)𝐵𝐵 blackbody spectral radiance.
Ranging from 0 to 1, emissivity values vary with wavelength, viewing angle, and surface roughness [53].
In order to calculate the real temperature of a surface, kinetic temperature, and as materials do not act as perfect emitters (ε=1), emissivity values have to be taken into account: 𝑇𝑘𝑖𝑛= 𝑐2 𝜆 ln(𝜀(𝜆)𝑐1𝜆−5 𝑀(𝜆,𝑇) +1) (Eq 1.6)
Albeit, as atmosphere influences the signal reaching the sensor such effects have to be taken into consideration in order to retrieve the real temperature of the surface. This is only possible when both emissivity and atmospheric corrections are accounted for [54].
Figure 8. Peak emission wavelength variation related to temperature according to Wein’s Displacement Law Adapted from Harris [3]
1.4. Land Surface Temperature
1.4.1. Concepts and retrieval
Earth is constantly being heated either by solar radiation or by radioactively produced heat in the Earth´s core that reaches the surface due to both conductive and convective processes [7]. Such heat translates to LST.
LST is a crucial physical parameter part of land surface processes that is related with interactions between the surface and the atmosphere and energy balance [55, 56]. LST values change continuously through time and space, assessing its variation can reveal unknown processes and is of great importance to numerous applications ranging from vegetation monitoring, to climate change, environmental studies, monitoring of geothermal areas, thermal studies of volcanic sites, among others [9, 10, 54].
LST measurements can be affected by external factors, such as solar radiation, vegetation, soil or terrain characteristics [10, 54]. The growing interest in thermal-infrared (TIR) remote sensing and in LST retrieval has increased over the years and numerous techniques to retrieve LST were published [55]. As a result, various categories of algorithms to retrieve LST were developed ranging from single-channel methods, to multi-channel and multi-angle approaches. This interest led to the creation of projects and algorithms of autonomous eruption detection previously referred in this work, as the Volcano Sensor Web Project [45] and MODVOLC [43]. Albeit, LST retrieval from space is still considered to be complex and problematic due to the corrections of both atmospheric and emissivity effects in order to achieve accurate measurements [54].
The radiative transfer equation (RTE) is the connecting link between satellite-derived TIR data and LST [54].
RTE is given by:
𝐿𝑠𝑒𝑛,𝜆= [𝜀𝜆𝐵𝜆(𝑇𝑠) + (1 − 𝜀𝜆)𝐿↓𝑎𝑡𝑚,𝜆]𝜏𝜆+ 𝐿↑𝑎𝑡𝑚,𝜆 (Eq 1.7)
Where 𝐿𝑠𝑒𝑛,𝜆 is at-sensor radiance for a defined wavelength, ε is land surface emissivity, 𝐵(𝑇𝑠) corresponds to blackbody spectral radiance obtained from Planck’s
function (Eq 1.1), 𝑇𝑠 is LST; 𝜏, 𝐿↑𝑎𝑡𝑚, and 𝐿↓𝑎𝑡𝑚 are atmospheric parameters, total
atmospheric transmissivity, upwelling atmospheric radiance, and downwelling atmospheric radiance, respectively.
𝐵𝜆(𝑇𝑠) =𝐿𝑠𝑒𝑛,𝜆−𝐿↑𝑎𝑡𝑚,𝜆−𝜏𝜆(1−𝜀𝜆)𝐿↓𝑎𝑡𝑚,𝜆
𝜏𝜆𝜀𝜆 (Eq 1.8)
Though relatively simple, inversion of RTE implies the use of three atmospheric parameters (𝜏, 𝐿↑𝑎𝑡𝑚, and 𝐿
𝑎𝑡𝑚
↓ ), which forces the need for atmospheric soundings close
to the study area at the time of satellite overpass, or the use of complex models of atmospheric profiles [55].
Constrains to LST retrieval are closely related with both emissivity effects and atmospheric corrections [56]. Atmospheric correction methodologies are often complex and difficult to apply. Errors in atmospheric correction implementation and noise affecting sensors hinder the accuracy of LST retrieval [54]. Spectral and angular variation of emissivity values adds to the uncertainties of LST retrieval [53]. Interpretation of LST values can be challenging as it often depends of the methodology applied and can also be affected by the cell size of acquired images as pixels can cover heterogeneous areas with different emissivity and temperature values [54]. Validation of LST values presents its own hurdles as well. Likewise, in-situ measurements are complex to perform as LST can vary greatly both spatially and temporally, only allowing ground measurements to be used effectively in validation efforts over homogeneous land surfaces, such as deserts and lakes. Additionally, downscaling satellite imagery to match ground measurements is troublesome and, in most cases, ineffective [54].
Development of LST retrieval methodologies has evolved over the past decades, with some algorithms being purposely designed to match sensors characteristics. Algorithms can be grouped according to whether LSE values are known beforehand or not, with single-channel methods, multi-channel methods, and multi-angle methods integrating the first category, whereas stepwise retrieval methods, simultaneous retrieval of LST, LSE, and atmospheric profiles being part of the latter [54].
Single-channel methods use radiance values measured by a sensor to derive LST through the inversion of RTE (Eq 1.8), in the condition that emissivity values are known in advance. Atmospheric profile data serve as input for atmospheric transmittance/radiance code, used to perform atmospheric correction. In order to derive accurate LST measurements using single-channel algorithms both LSE and atmospheric profile data must be precise, and the influence of topography should be taken into account [57]. Such requirements in some cases increase the complexity of implementation of these algorithms [54].
Multi-channel methodologies use the differential atmospheric absorption between more than one channel in the EM spectrum, usually adjacent channels in the TIR domain between 10 and 12 μm [58]. A myriad of multi-channel algorithms has been developed to retrieve LST values. Various approaches were considered to deal with differences between air temperature (Ta) and LST, spectral and spatio-temporal variations of LSE,
viewing zenith angle (VZA), and total column water vapour (WV) in the atmosphere [54]. Multi-angle methods make use of the differences of atmospheric absorption verified when the signal emitted by a surface is analysed via multiple viewing angles [54]. It is only applicable to data generated by platforms with biangular capabilities, e.g. ASTER nadir and off-nadir modes. Some issues arise from this fact as emissivity values on rough surfaces are affected by the viewing angle [53] and topography can distort the area covered by pixels misrepresenting surfaces and objects leading to faulty LST retrieval. As so, multi-angle methods should be applied to homogeneous areas [54].
Unlike the aforementioned methods where LSE values are known prior to the retrieval of LST, other approaches take into account the difficulties of assessing LSE from space in advance with fine accuracy as a result of surface roughness and heterogeneity, atmospheric effects, and both angular and spectral variation of emissivity for each material [53].
Stepwise retrieval methodologies act as a two-step process first determining LSE values from either the VNIR domain of the electromagnetic spectrum or pairs of MIR and TIR atmospherically corrected radiances [54]. Afterwards, LST is derived using either single-channel, multi-channel, or multi-angle algorithms mentioned previously.
Methods for the simultaneous retrieval of LST and LSE values were developed to ensure higher retrieval accuracy considering the close dependence of accurate LST values with the accuracy of predetermined LSE. Such methodologies can be divided in two main distinguishable classes, multi-temporal and multi-spectral retrieval methods [54]. While the first category use measurements with different acquisition timestamps to simultaneous retrieve LST and LSE in the condition that LSE values are constant, the latter focus on the spectral behaviour of LSE.
As precise atmospheric parameters are seldom obtained at the same period of TIR measurements, simultaneous retrieval of LST and LSE values can be affected if atmospheric corrections are not applied correctly. To reinforce data precision some methods that retrieve both LST, LSE, and atmospheric profiles were created using the
capabilities of hyperspectral sensors with precise spectral resolution allowing the retrieval of more accurate atmospheric profiles and surface data due to improved vertical resolution [54].
The variety of methods described above reinforces the fact that LST retrieval is in many cases complex. Data quality is crucial to obtain accurate reliable results, with the application of methodologies being dependent of data availability, sensor characteristics, and other constraints. Consequently, it is incorrect to assume that some algorithms ultimately excel others as there is no method that can be used universally [54].
Validation of LST data is essential as it determines the credibility and accuracy of measured records which helps foster the use of thermal data in a growing number of applications. The temperature-based, radiance-based and cross validation methods are commonly used. Whereas the temperature-based method compares satellite-derived LST records with in situ measurements coincident with the overpass of the satellite [59], the radiance-based method uses both in situ and satellite derived LSE data, and atmospheric profiles of the validation area synchronized with the satellite overpass [60]. The cross validation method is more straightforward as it simply uses validated LST data obtained from other satellite comparing it with the derived LST measurements [61]. Only the lack of validated LST data can hinder the application of this technique, as the absence of field measurements adds to its simplicity.
1.4.2. Landsat data used in LST studies: a brief synopsis
Landsat satellites with TIR capabilities have been used to support thermal infrared remote sensing by tracking the activity of active volcanoes, making use of its high spatial resolution to the thermal study of lava flows [62], geothermal exploration [10], calculation of radiative and geothermal heat flux [63] and to detect thermal anomaly areas [64]. Flynn, et al. [9] lists several other works where Landsat data was used in support of thermal studies of volcanic features and products. Landsat TIR data can also be used to efficiently retrieve LST values for geothermal areas [10]. Although, due to low temporal resolutions (16 days) Landsat data are normally confined to thermal studies and retrieval of LST values [9]. Landsat extensive catalogue supports the study of energy balance and energy transfers dynamics and its evolution.
To assess the quantity and evolution of previous LST studies applying Landsat data Web of Science database was used. A simple search to collect data referring to scientific
publications was conducted in late January of 2019. Restricted to works published until December 2018, it used different search combinations, whether searching for ‘Landsat’ and ‘Land Surface Temperature’ terms in the search fields ‘Title’ and ‘Topic’1 or combining both. An additional query to the same database was performed to search LST studies using Landsat data for volcanic areas. The term ‘Volcan*’2 was combined with
the previous two, the search field ‘Topic’ was used for all. Table 2 shows the results for each approach. The results are in most cases papers from scientific journals. Conference proceedings, book sections and books are also represented. The outcome concerning publications number is diverse, whereas the time range covered by said publications is very similar. Such differences are due to the use of both restrict and broad search fields (‘Title’ or ‘Topic’).
Table 2. Publications of LST studies applying Landsat data
Search Terms LST (T) Landsat (T) LST (T) Landsat (t) LST (t) Landsat (T) LST (t) Landsat (t) LST (t) Landsat (t) Volcan* (t) Publications 131 362 410 1511 36 Time range 1993-2018 1993-2018 1992-2018 1990-2018 1993-2018 Search fields: T- Title; t- Topic.
Taking a closer look to the search results including both ‘Land Surface Temperature’ and ‘Landsat’ terms in the search field ‘Topic’ it is possible to observe in Figure 9 that the first publication addressing an LST study using Landsat data was published in 1990. In that same decade (1990-1999) the number of publications published per year, despite its inconstancy grew ever slowly, peaking in 1998 (11) and registering an average of 4.8 publications/year. By the end of the following decade (2000-2009) the total number of publications had more than quadrupled to 46 publications (2009) averaging 23.9 publications/year. From then on (2010-2018) the number of publications experienced a continued incremental growth, passing the mark of 100 publications in
1 Search field ‘Topic’ is used as an aggregate search tool that encloses ‘Title’, ‘Abstract’, and
‘Author Keywords’ search fields.
2 The search term ‘Volcan*’ was used as the asterisk (*) allows for the search of suffixes of the term
2014-2016 and 200 publications in 2017-2018. The mean number of publications per year for this time-period was 136.
Further analysis of this group of publications detailed the use and main applications observed. For practicality reasons only publications with more than one hundred citations at the time of the search were considered. In this subset of 56 works the majority applied Landsat-derived LST data to the study of urban heat islands and land cover changes [65-68]. Other applications include surface energy balance [69-71], evapotranspiration and water balance [72-75], vegetation distribution [76, 77], and volcanic studies [78]. Note to the presence of several publications regarding LST algorithms among the most cited [79-82].
Figure 9.Number of LST publications using Landsat data over the years
As addressed earlier, some of the reasons that explain the growth of scientific literature dedicated not only to satellite remote sensing, but also LST studies in general are as well behind the rise of such studies where Landsat data is applied. The constant development of new methodologies and techniques, new instruments, and increasing power and better computational capabilities to name a few, were and still are fundamental to this evolution. In addition to this, the availability and readiness of data are another fundamental driver that might explain the progress depicted in Figure 9. Users of Landsat data saw this specific setting change when in 2008 United States Geological Survey (USGS) decided to implement an open data policy making its Landsat-data archive available at no-costs, resulting in the increase of downloads and consequently broadening the field of applications and scientific studies [83]. From that year onwards is evident the progression in the number of publications.
In respect to the search concerning publications using Landsat data to retrieve LST in volcanic areas, the number (36) is significantly lower as expected since the query to
the database is more restrictive. Moreover, it demonstrates that the application of LST to study volcanic areas using Landsat data is to some extent unexplored. Table 3 presents the listing of publications matching this search.
Analysing all records, it was possible to deem some as outliers as the studies either do not specifically use LST or were not conducted in or near areas with volcanic/geothermal elements only referring the potential application of LST to volcanic studies. These publications are identified as outliers in Table 3.
Examining the remainder of publications, a myriad of parameters and applications can be identified. Beginning with the works published in the last decade of the twentieth century, Ohkura, et al. [84] refers the use of Landsat thermal data for the assessment of surface temperature of volcanic features, Oppenheimer [85] applies Landsat TIR data for mapping volcanic crater lakes and estimate heat flux, whereas Harris, et al. [78] reports the calculation of lava effusion rates and temperature estimation of active lava flows. The absence of the term ‘LST’ in favour of ‘surface temperature’ may correspond to the lack of maturity or scarce implementation of algorithms for the retrieval of LST at the time.
Some publications refer the use of LST to support prospection and exploration of geothermal areas. Peng, et al. [86] coupled it with the analysis of geological data, Sircar, et al. [87] and Tian, et al. [88] focused on the detection of thermal anomalies, whereas Sukojo and Mardiana [89] combined elevation data, vegetation index and LST. Additionally, monitorization of thermal activity and derivation of geothermal parameters such as radiative heat flux (RHF), geothermal heat flux (GHF), and heat discharge rate (HDR) were used in Mia, et al. [90], Mia, et al. [91], Mia, et al. [92] and Eskandari, et al. [64] supporting the study of geothermal areas.
Mia, et al. [93], Mia, et al. [94], Mia and Fujimitsu [95] and Mia and Fujimitsu [96] focused on monitoring heat losses calculating RHF and HDR after retrieving LST values, correlating both parameters. Mia, et al. [97] and Mia, et al. [98] also derived RHF and HDR from LST measurements while using multi-source satellite data coupling Landsat thermal data with ASTER data, while Chan and Chang [99] added MODIS 8-day composite data. Seward, et al. [100] conducted a study to assess surface heat-loss monitoring capabilities at geothermally active areas comparing Landsat thermal data with aerial TIR data and a terrestrial calorimetry survey.
Some publications assumed a more exploratory approach comparing ground and satellite data sources. Lagios, et al. [101] compared LST values retrieved from
Landsat-7 with TIR data obtained from a thermal infrared camera of a fumarolic field, Ganas and Lagios [102] used Landsat-7 night imagery and field measurements to map and analyse the distribution of LST in a crater surface, where Isa, et al. [103] used Landsat-5 thermal data comparing it to in-situ data.
Emetere [104] and Suwarsono, et al. [105] studied volcanic dynamics before eruptive events with the aim of monitoring and identifying precursive eruptive signals. A termographic model and various thermal analysis parameters were used in the former, while a more straightforward approach was used in the latter with the direct calculation of brightness temperature from Landsat radiance values. Murphy, et al. [106] used Landsat TIR data to assess and monitor both temperature and colour of crater lakes which can also be indicative of unrest in the volcanic system.
Tampubolon and Yanti [107] used LST measurements retrieved from Landsat data for disaster mitigation purposes concerning eruptions, while Kayiranga, et al. [108] assessed the combined impact of topography, LST and other climate factors on vegetation growth and distribution in a volcanic massif comprising datasets from ASTER and MODIS sensors.
Vieira, et al. [28] occupies a central space among all other publications as it acts as the predecessor of the present work. In Vieira, et al. [28] LST values were retrieved from Landsat-8 data in order to monitor surface temperature dynamics during Pico do Fogo 2014-2015 eruption in Fogo island, Cape Verde. MODIS LST products were used to perform cross-comparison of temperature records retrieved from Landsat data. Sharing the same thematic and subject of analysis what differentiates the present study, which can be regarded as an extension of the study developed previously, is the analysis of a larger period of time, the use of both ASTER and MODIS LST data for cross-comparison, the processing of solar radiation values and assessment of eventual relationships established between LST and solar radiation via regression analysis.
Table 3. Publications of LST studies applying Landsat data for volcanic areas
Year Authors Title
1993 Ohkura, H; Uehara, S; Yazaki, S; Kumagai, T Application of Multiple Satellite Data to The Study of Natural Disasters
1995 Kasturirangan, K * Indian Experience In-Ground Based Experiments in Support of Satellite Data Applications 1996 Oppenheimer, C Crater Lake Heat Losses Estimated by Remote Sensing
1998 Harris, AJL; Flynn, LP; Keszthelyi, L; Mouginis-Mark, PJ; Rowland, SK; Resing, JA
Calculation of Lava Effusion Rates from Landsat TM Data 2003 Suga, Y; Ogawa, H; Ohno, K; Yamada, K * Detection of Surface Temperature from LANDSAT-7/ETM+ 2003 Ganas, A; Lagios, E Landsat 7 Night Imaging of The Nissyros Volcano, Greece 2007 Lagios, E.; Vassilopoulou, S.; Sakkas, V.; Dietrich, V.;
Damiata, B. N.; Ganas, A.
Testing Satellite and Ground Thermal Imaging of Low-Temperature Fumarolic Fields: The Dormant Nisyros Volcano (Greece)
2008 Din, Saif Ud; Al Dousari, Ahmad; Literathy, Peter * Evidence of Hydrocarbon Contamination from The Burgan Oil Field, Kuwait - Interpretations from Thermal Remote Sensing Data
2012 Mia, Md. Bodruddoza; Fujimitsu, Yasuhiro; Bromley, Chris J.
Estimation and Monitoring Heat Discharge Rates Using Landsat ETM Plus Thermal Infrared Data: A Case Study in Unzen Geothermal Field, Kyushu, Japan
2013 Mia, Md. Bodruddoza; Bromley, Chris J.; Fujimitsu, Yasuhiro
Monitoring Heat Losses Using Landsat ETM Plus Thermal Infrared Data: A Case Study in Unzen Geothermal Field, Kyushu, Japan
2013 Mia, Md. Bodruddoza; Fujimitsu, Yasuhiro Monitoring Heat Losses Using Landsat ETM Plus Thermal Infrared Data - A Case Study at Kuju Fumarolic Area in Japan
2013 Mia, Md. Bodruddoza; Fujimitsu, Yasuhiro Landsat Thermal Infrared Based Monitoring of Heat Losses from Kuju Fumaroles Area in Japan 2013 Isa, M.; Jafri, M. Z. Mat; Lim, H. S. Comparison of Field Temperature Versus Satellite Temperature Thermal Band in Geothermal Area
Year Authors Title
2013 Peng, Fen; Xiong, Yongzhu; Cheng, Yuxiang; Fan, Qicheng; Huang, Shaopeng
Towards Application of Remote Sensing Technology in Geothermal Prospecting in Xilingol In Eastern Inner Mongolia, NE China
2014 Ghosh, Aniruddha; Joshi, P. K. * Hyperspectral Imagery for Disaggregation of Land Surface Temperature with Selected Regression Algorithms Over Different Land Use Land Cover Scenes
2014 Mia, Md. Bodruddoza; Nishijima, Jun; Fujimitsu, Yasuhiro
Exploration and Monitoring Geothermal Activity Using Landsat ETM Plus Images
2014 Jung, Hyung-Sup; Park, Sung-Whan * Multi-Sensor Fusion of Landsat 8 Thermal Infrared (TIR) and Panchromatic (PAN) Images 2015 Eskandari, Amir; De Rosa, Rosanna; Amini, Sadraddin Remote Sensing of Damavand Volcano (Iran) Using Landsat Imagery: Implications for the
Volcano Dynamics 2015 Tian, Bingwei; Wang, Ling; Kashiwaya, Koki; Koike,
Katsuaki
Combination of Well-Logging Temperature and Thermal Remote Sensing for Characterization of Geothermal Resources in Hokkaido, Northern Japan
2015 Uysal, Murat; Polat, Nizar * An Investigation of the Relationship Between Land Surface Temperatures and Biophysical Indices Retrieved from Landsat TM in Afyonkarahisar (Turkey)
2015 Sircar, Anirbid; Shah, Manan; Sahajpal, Shreya; Vaidya, Dwijen; Dhale, Shubhra; Chaudhary, Anjali
Geothermal Exploration in Gujarat: Case Study from Dholera 2015 Mia, Md. Bodruddoza; Nishijima, Jun; Fujimitsu,
Yasuhiro
Monitoring Heat Flow Before and After Eruption of Kuju Fumaroles in 1995 Using Landsat TIR Images
2016 Fadlillah, Lintang N.; Widyastuti, M. * Water Balance and Irrigation Water Pumping of Lake Merdada for Potato Farming in Dieng Highland, Indonesia
2016 Vieira, D.; Teodoro, A.; Gomes, A. Analysing Land Surface Temperature Variations During Fogo Island (Cape Verde) 2014-2015 Eruption with Landsat 8 Images
2016 Tampubolon, T.; Yanti, J. Remote Sensing for Disaster Mitigation of Sinabung