Full download Thermal and Optical Remote Sensing John O Odindi file pdf all chapter on 2024

Thermal and Optical Remote Sensing
John O Odindi
Visit to download the full and correct content document:
https://ebookmass.com/product/thermal-and-optical-remote-sensing-john-o-odindi/
More products digital (pdf, epub, mobi) instant
download maybe you interests ...
Optical Remote Sensing of Land Surface. Techniques and

Methods 1st Edition Nicolas Baghdadi
https://ebookmass.com/product/optical-remote-sensing-of-land-
surface-techniques-and-methods-1st-edition-nicolas-baghdadi/
Bio-optical Modeling and Remote Sensing of Inland

Waters 1st Edition Edition D.R. Mishra
https://ebookmass.com/product/bio-optical-modeling-and-remote-
sensing-of-inland-waters-1st-edition-edition-d-r-mishra/
Remote sensing digital image analysis Sixth Edition

John Alan Richards
https://ebookmass.com/product/remote-sensing-digital-image-
analysis-sixth-edition-john-alan-richards/
Atmospheric Remote Sensing Abhay Kumar Singh
https://ebookmass.com/product/atmospheric-remote-sensing-abhay-
kumar-singh/
Techniques and methods in urban remote sensing Weng
https://ebookmass.com/product/techniques-and-methods-in-urban-
remote-sensing-weng/
Remote Sensing Digital Image Analysis 6th ed. 2022

Edition John A. Richards
https://ebookmass.com/product/remote-sensing-digital-image-
analysis-6th-ed-2022-edition-john-a-richards/
Remote Sensing of Ocean and Coastal Environments Meenu

Rani
https://ebookmass.com/product/remote-sensing-of-ocean-and-
coastal-environments-meenu-rani/
Hyperspectral Remote Sensing: Theory and Applications

Prem Chandra Pandey
https://ebookmass.com/product/hyperspectral-remote-sensing-
theory-and-applications-prem-chandra-pandey/
Atmospheric Remote Sensing: Principles and Applications

Abhay Kumar Singh
https://ebookmass.com/product/atmospheric-remote-sensing-
principles-and-applications-abhay-kumar-singh/
Thermal and Optical
Remote Sensing
Evaluating Urban Green Spaces and Urban
Heat Islands in a Changing Climate
Edited by
John O. Odindi, Elhadi Adam, Elfatih M. Abdel-Rahman and
Yuyu Zhou
Printed Edition of the Special Issue Published in Remote Sensing
www.mdpi.com/journal/remotesensing
Thermal and Optical Remote Sensing:
Evaluating Urban Green Spaces and
Urban Heat Islands in a Changing
Climate
Thermal and Optical Remote Sensing:
Evaluating Urban Green Spaces and
Urban Heat Islands in a Changing
Climate
Editors
John O. Odindi
Elhadi Adam
Elfatih M. Abdel-Rahman
Yuyu Zhou
MDPI • Basel • Beijing • Wuhan • Barcelona • Belgrade • Manchester • Tokyo • Cluj • Tianjin
Editors
John O. Odindi Elhadi Adam Elfatih M. Abdel-Rahman
University of KwaZulu-Natal University of the International Centre of Insect
South Africa Witwatersrand Physiology and Ecology
South Africa (ICIPE)
Kenya
Yuyu Zhou
Iowa State University
USA
Editorial Office
MDPI
St. Alban-Anlage 66
4052 Basel, Switzerland
This is a reprint of articles from the Special Issue published online in the open access journal
Remote Sensing (ISSN 2072-4292) (available at: https://www.mdpi.com/journal/remotesensing/
special issues/RS Urban Green UHI).
For citation purposes, cite each article independently as indicated on the article page online and as
indicated below:
LastName, A.A.; LastName, B.B.; LastName, C.C. Article Title. Journal Name Year, Volume Number,
Page Range.
ISBN 978-3-0365-6275-9 (Hbk)

ISBN 978-3-0365-6276-6 (PDF)
Cover image courtesy of Getty Images
© 2023 by the authors. Articles in this book are Open Access and distributed under the Creative
Commons Attribution (CC BY) license, which allows users to download, copy and build upon
published articles, as long as the author and publisher are properly credited, which ensures maximum
dissemination and a wider impact of our publications.
The book as a whole is distributed by MDPI under the terms and conditions of the Creative Commons
license CC BY-NC-ND.
Contents
About the Editors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii
Preface to ”Thermal and Optical Remote Sensing: Evaluating Urban Green Spaces and
Urban Heat Islands in a Changing Climate” . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ix
Xiangyu Li, Guixin Zhang, Shanyou Zhu and Yongming Xu

Step-By-Step Downscaling of Land Surface Temperature Considering Urban Spatial
Morphological Parameters
Reprinted from: Remote Sens. 2022, 14, 3038, doi:10.3390/rs14133038 . . . . . . . . . . . . . . . . . 1
Wenhao Zhu, Jiabin Sun, Chaobin Yang, Min Liu, Xinliang Xu and Caoxiang Ji
How to Measure the Urban Park Cooling Island? A Perspective of Absolute and Relative
Indicators Using Remote Sensing and Buffer Analysis
Tshilidzi Manyanya, Janne Teerlinck, Ben Somers, Bruno Verbist

and Nthaduleni Nethengwe
Sentinel-Based Adaptation of the Local Climate Zones Framework to a South African Context
Terence Darlington Mushore, Onisimo Mutanga and John Odindi

Determining the Influence of Long Term Urban Growth on Surface Urban Heat Islands Using
Local Climate Zones and Intensity Analysis Techniques
Ali S. Alghamdi, Ahmed Ibrahim Alzhrani and Humud Hadi Alanazi

Local Climate Zones and Thermal Characteristics in Riyadh City, Saudi Arabia
Meiling Zhou, Linlin Lu, Huadong Guo, Qihao Weng, Shisong Cao, Shuangcheng Zhang
and Qingting Li
Urban Sprawl and Changes in Land-Use Efficiency in the Beijing–Tianjin–Hebei Region, China
from 2000 to 2020: A Spatiotemporal Analysis Using Earth Observation Data
Kangning Li, Yunhao Chen and Shengjun Gao

Comparative Analysis of Variations and Patterns between Surface Urban Heat Island Intensity
and Frequency across 305 Chinese Cities
Baohui Mu, Xiang Zhao, Jiacheng Zhao, Naijing Liu, Longping Si, Qian Wang, Na Sun, et al.
Quantitatively Assessing the Impact of Driving Factors on Vegetation Cover Change in China’s
32 Major Cities
Abdelali Gourfi, Aude Nuscia Taı̈bi, Salima Salhi, Mustapha El Hannani and Said Boujrouf
The Surface Urban Heat Island and Key Mitigation Factors in Arid Climate Cities, Case of
Marrakesh, Morocco
Terence Mushore, John Odindi and Onisimo Mutanga

“Cool” Roofs as a Heat-Mitigation Measure in Urban Heat Islands: A Comparative Analysis
Using Sentinel 2 and Landsat Data
v
Thu Thi Nguyen, Harry Eslick, Paul Barber, Richard Harper and Bernard Dell
Cooling Effects of Urban Vegetation: The Role of Golf Courses
vi
About the Editors
John O. Odindi
John O. Odindi holds a PhD in Environmental Geography with focus on terrestrial remote
sensing. He is a full professor and former Head of Discipline of Geography and Environmental
Sciences at the University of KwaZulu-Natal, South Africa. His main interests are urban green spaces
and urban microclimate using optical, thermal and Radar datasets, remote sensing of above ground
biomass, local and regional carbon modelling, remote sensing of rangeland nutritive value and crop
yield, remote sensing of plant invasive species and crop disease modelling. His current projects
include; Drone based high spatial resolution crop phenotyping, Data science for global health
and climate change, Earth observation big data science, Ecosystem/urban heat island nexus using
remotely sensed datasets, The integrated pest management strategy for food security in Eastern
Africa (FAW-IPM) and the African reference laboratory (with satellite stations) for the management
of pollinator bee diseases and pests for food security. Professor Odindi has been previously selected
as one of the 52 African researchers with outstanding contribution to Geo-information and Earth
Observation on the continent. He is a Brown International Advanced Research Institute (BIARI)
fellow of the Watson Institute, Brown University, Rhode Island, USA.
Elhadi Adam
Elhadi Adam (Professor) holds a PhD in Environmental Science with a specific focus on
hyperspectral remote sensing. His expertise lies in the applications of remote sensing in applied
environmental science and precision agriculture. He is interested in vegetation mapping and
monitoring by integrating field-based, airborne and space-borne hyperspectral remote sensors for
detecting subtle patterns in natural vegetation and crops. He focuses on biochemical concentration,
invasive species mapping, wetland health, herbaceous biomass distribution and characterization of
heterogeneous landscapes. His current research project is developing a support information system
to provide a broader range of agronomically-relevant information about crop biophysical status and
leaf biochemistry using remote sensing techniques and machine learning algorithms. Prof Adam’s
current research also focuses on detecting and mapping disease on crops and quantifying forest
fragmentation and its impact on biodiversity and ecosystem services.
Elfatih M. Abdel-Rahman
Elfatih M. Abdel-Rahman (Research Scientist) holds a PhD in Environmental Sciences with
over 14 years of experience in working with multidisciplinary teams of Environmental Scientists,
Agronomists, Entomologists, Plant Pathologists, Vector Scientists, Mathematical Modelers, Computer
Scientists, Engineers, Geographers and Social Scientists. He uses remote sensing, Earth observation
and geospatial modelling tools to improve the management of complex agro-natural and urban
systems and to enhance crop risks monitoring, forecasting and early warning systems. This extensive
experience, coupled with his background as an Agronomist and a university Associate Professor,
gives him an excellent foundation to carry out various remote sensing and geospatial modelling
tasks to develop advisory tools for improving the health of human, plant, animal, and environment.
His current research and development projects focus on assessing the effect of landscape dynamics
and climate change on agricultural pests, malaria and tsetse vectors, crop yield using satellite data,
data science and machine learning algorithms. He is a member of the Editorial Advisory Board of
ISPRS Journal of Photogrammetry and Remote Sensing and the Editorial Boards of the International
vii
Journal of Tropical Sciences, and Frontiers in Multi- and Hyper-Spectral Imaging. Dr. Abdel-Rahman
ranks among the best 10% scientists who apply remote sensing and GIS in agriculture in Africa. He
is a Brown International Advanced Research Institute (BIARI) fellow of the Watson Institute, Brown
University, Rhode Island, USA.
Yuyu Zhou
Yuyu Zhou is an Associate Professor in the Department of Geological and Atmospheric Sciences
at Iowa State University, USA. He holds a Bachelor of Science degree in Geography and Master of
Science degree in Remote Sensing from Beijing Normal University and a Ph.D. in Environmental
Sciences from the University of Rhode Island. Dr Zhou’s general focus is the quantification of
spatiotemporal patterns of environmental change and developing modeling mechanisms to bridge
the driving forces (both natural and socioeconomic factors) and consequences of environmental
change so that the impacts of human activities on environment can be effectively measured, modeled,
and evaluated. Specifically, his research focuses and interest include the application of remote
sensing, GIS, Integrated Assessment Modelling, and spatial analysis in ecosystem, environmental,
and social sciences, including interdisciplinary areas: Energy sources, demand, climate change
impact, high spatial and temporal energy and emissions modelling, hydrology and ecosystem
modelling. He was awarded a NASA Land-Cover Land-Use Change (LCLUC) Research Grant for
Early Career Scientist to study global urban expansion in the context of climate change. He previously
worked as a Geographical Scientist at the Joint Global Change Research Institute, a partnership of the
Pacific Northwest National Laboratory (PNNL) and the University of Maryland.
viii
Preface to ”Thermal and Optical Remote Sensing:
Evaluating Urban Green Spaces and Urban Heat
Islands in a Changing Climate”
Urbanization, typified by land-use-land-cover transformation is a major cause of bio-physical,

thermodynamic, surface energy and micro- and macro climate perturbations. These changes
commonly result in environmental deterioration that in turn adversely affects bio-physical processes
and quality of urban life. A major consequence of urbanization is the higher thermal values
compared to the surrounding peri-urban and rural areas, causing the Urban Heat Island (UHI)
effect. In recent decades, above-average heat during summer has become prevalent in global cities,
a trend that is expected to continue. It is anticipated that the intensifying UHI effect, in concert
with increasing anthropogenic activities, will exacerbate the vulnerability of urban landscapes to
climate-related disasters such as floods and heatwaves. Hence, UHIs have become an invaluable
theme in environmental research. A recent proliferation of optical and thermal remotely sensed
datasets offers great potential for understanding the relationship between the urbanization processes
and their respective bio-physical and climatic implications. This book focuses on the theoretical
principles and practical adoption of remote sensing approaches and datasets in understanding the
nexus between urbanization, natural landscapes, urban micro-climate, and climate change. This
book provides a basis for understanding urban ecological and natural patterns, critical for the
management of urban physical, ecological and social processes. Specifically, understanding past,
current, and future Land Surface Temperature (LST) patterns and drivers is critical for, among
others, urban environmental management, urban spatial planning, the optimal and sustainable use of
urban landscapes and climate change mitigation. The book’s first two chapters explore the potential
of downscaling remotely sensed data and improved feature extraction to determine the effect of
urban surface types on thermal characteristics. Chapter one adopts a Step-by-Step Random Forest
Downscaling-Morphological (SSRFD-M) model to relate natural surfaces to LST, while chapter two
proposes the absolute and relative indicators for the detailed derivation of landscape features and
thermal values using Geofen 2 (GF-2) and Landsat 8 Thermal Infra-Red (TIR), respectively. Chapter
three to five adopt the standardized Local Climate Zone (LCZ) typology to relate urban landscape
feature types to thermal characteristics. Chapter three and four use the World Urban Database and
Access Portal Tool (WUDAPT) and the LCZ framework in South Africa (Cape Town, Thohoyandou
and East London) and Zimbabwe (Bulawayo), respectively, while chapter five relates seasonal LCZ
to daytime LST in Riyadh, Saudi Arabia. Chapter six to nine investigate multi-city urban LULC
and the contribution of the climate, urbanization and CO2 to UHI in multiple cities. Chapter six
proposes a Landsat imagery time series approach in Google Earth Engine platform to map built-up
areas in 305 cities, while chapter seven compares Surface Urban Heat Island (SUHI) in relation to the
SUHI fraction’s key drivers in 305 Chinese cities. Chapter eight determines a critical competitive
point of Artificial Surfaces (AS) and Urban Blue-Green Space (UBGS) on LST in 28 cities, while
chapter nine relates the Global Land Surface Satellite (GLASS) Fractional Vegetation Cover (FVC)
to CO2 , urbanization, and climate in 32 major cities. Chapter ten to twelve investigate a range of
UHI-mitigation approaches. Chapter 10 explores the value of urban green spaces in mitigating UHI
in Marrakesh, Morocco, while chapter eleven investigates the role roof colours in assimilating surface
temperature. Using vegetation’s morphological Spatial Patter Analysis (MSPA), chapter twelve
adopts ArborCamTM multispectral high-resolution imagery to determine the role of golf courses in
ix
assimilating urban LST. This book should be of interest to both specialists and generalists interested
in, among others, urban planning, ecological conservation, the urban micro-climate, atmospheric
science, environmental management, and climate change.
John O. Odindi, Elhadi Adam, Elfatih M. Abdel-Rahman, and Yuyu Zhou

Editors
x
remote sensing
Article
Step-By-Step Downscaling of Land Surface Temperature
Considering Urban Spatial Morphological Parameters
Xiangyu Li 1 , Guixin Zhang 2, *, Shanyou Zhu 1 and Yongming Xu 1
1 School of Remote Sensing and Geomatics Engineering, Nanjing University of Information Science
and Technology, Nanjing 210044, China; lxyxzs@nuist.edu.cn (X.L.); zsyzgx@nuist.edu.cn (S.Z.);
xym30@nuist.edu.cn (Y.X.)
2 School of Geographical Sciences, Nanjing University of Information Science and Technology,
Nanjing 210044, China
* Correspondence: 001631@nuist.edu.cn
Abstract: Land surface temperature (LST) is one of the most important parameters in urban thermal
environmental studies. Compared to natural surfaces, the surface of urban areas is more complex, and
the spatial variability of LST is higher. Therefore, it is important to obtain a high-spatial-resolution
LST for urban thermal environmental research. At present, downscaling studies are mostly performed
from a low spatial resolution directly to another high resolution, which often results in lower accuracy
with a larger scale span. First, a step-by-step random forest downscaling LST model (SSRFD) is
proposed in this study. In our work, the 900-m resolution Sentinel-3 LST was sequentially downscaled
to 450 m, 150 m and 30 m by SSRFD. Then, urban spatial morphological parameters were introduced
into SSRFD, abbreviated as SSRFD-M, to compensate for the deficiency of remote-sensing indices as
driving factors in urban downscaling LST. The results showed that the RMSE value of the SSRFD
results was reduced from 2.6 ◦ C to 1.66 ◦ C compared to the direct random forest downscaling model
Citation: Li, X.; Zhang, G.; Zhu, S.;
(DRFD); the RMSE value of the SSRFD-M results in built-up areas, such as Gulou and Qinhuai
Xu, Y. Step-By-Step Downscaling of District, was reduced by approximately 0.5 ◦ C. We also found that the underestimation of LST caused
Land Surface Temperature by considering only remote-sensing indices in places such as flowerbeds and streets was improved in
Considering Urban Spatial the SSRFD-M results.
Morphological Parameters. Remote
Sens. 2022, 14, 3038. https:// Keywords: step-by-step downscaling of LST; land surface temperature; urban spatial morphology
doi.org/10.3390/rs14133038
Academic Editors: Yuyu Zhou,

Elhadi Adam, John Odindi and
Elfatih Abdel-Rahman
1. Introduction
As an important physical variable driving the energy exchange between the surface
Received: 21 May 2022
and the atmosphere, the surface temperature (LST) is one of the key parameters for studying
Accepted: 21 June 2022
the energy balance of the surface at global or regional scales. Currently, LST is widely used to
Published: 24 June 2022
assess surface moisture and drought levels [1–4], calculate urban heat island intensity [5–7]
Publisher’s Note: MDPI stays neutral and simulate surface energy exchange [8–11]. In urban areas, the spatial and temporal
with regard to jurisdictional claims in heterogeneity of urban surface temperature is obvious due to the extremely complex
published maps and institutional affil- surface, the strong differences in three-dimensional spatial geometry and the variety of
iations. surface components and types. Therefore, studies of the urban thermal environment and
other urban-related research fields usually require LST data with a higher spatiotemporal
resolution.
The LST data obtained by thermal infrared remote-sensing technology generally have
Copyright: © 2022 by the authors.
the problem of conflicting spatial and temporal resolutions. For example, the Moderate
Licensee MDPI, Basel, Switzerland.
This article is an open access article
Resolution Imaging Spectroradiometer (MODIS) LST product has a high temporal resolu-
distributed under the terms and
tion, but the spatial resolution is only 1000 m; Sentinel-3 operates through a binary orbit
conditions of the Creative Commons with a temporal resolution of fewer than 0.9 days, but the spatial resolution of the LST
Attribution (CC BY) license (https:// product is also 1000 m; the land surface temperature product retrieved from Landsat TIRS
creativecommons.org/licenses/by/ has a spatial resolution of 100 m, but the revisit period is as long as 16 days. Together with
4.0/). the influence of clouds, the available valid Landsat LST data are even further diminished.
Remote Sens. 2022, 14, 3038. https://doi.org/10.3390/rs14133038 1 https://www.mdpi.com/journal/remotesensing

Remote Sens. 2022, 14, 3038
High-temporal-resolution data are difficult to generate for refined surface temperature

studies at an urban scale due to their low spatial resolution, while the high spatial resolution
LST data are unable to study the variation pattern of LST in time due to their low temporal
resolution. To solve the contradiction of spatial and temporal resolutions of remote-sensing
thermal data, scholars have proposed a large number of technical methods from various
perspectives, such as image processing and statistical regression, to obtain land surface
temperature data with high spatial and temporal resolutions.
The statistical regression method has gained wide application in LST downscaling
studies due to its low computational complexity and satisfactory downscaling results. The
application of this method has become relatively mature in suburban and mountainous ar-
eas with simple land covers at a large spatial scale [12,13]. However, there are two problems
that cannot be ignored when applying the statistical method to urban areas with complex
land surface types. Firstly, the larger the spatial resolution span of the downscaling, the
lower the accuracy tends to be. From the available thermal infrared remote-sensing data,
there are lots of LST products with a higher temporal resolution at about 1000-m spatial
resolution. When they are downscaled to the 100-m level or even the 10-m level, the spatial
resolution of the downscaled LST differs from the original resolution by a factor of 10 or
even 100, and the downscaled accuracy decreases as the spatial resolution increases. The
main reason for this problem is that the assumption of a “constant spatial scale relationship”
between LST and the driving factor does not hold when the resolution difference is large.
Secondly, the traditional two-dimensional remote-sensing spectral indices and surface
parameters are not sufficient to accurately describe the spatial pattern of an urban surface.
Currently, commonly used remote-sensing indices for downscaling models, such as the
normalized difference vegetation index (NDVI), normalized difference moisture index
(NDMI), normalized difference water index (NDWI) and normalized difference building
index (NDBI) [14] use surface parameters including the DEM, slope, slope direction, lati-
tude, longitude and surface cover type [15,16], as well as multispectral data [17] describing
the vegetation cover, moisture status and topographic relief of the land surface from the
two-dimensional perspective. In contrast, cities are dominated by buildings and imper-
vious pavements, but the influence of the three-dimensional morphological structure on
local land surface temperature is less considered. In fact, a large number of studies have
demonstrated that urban spatial morphological parameters such as the sky view factor
(SVF), frontal area index (FAI) and building density (BD) are closely related to LST [18–21],
meaning they need to be considered in downscaling models.
To address the above two problems, this study aimed to develop a step-by-step LST
downscaling method by further considering urban spatial morphological parameters to
obtain the urban land surface temperature at a spatial resolution of 30 m with high accuracy.
The paper takes the central urban area of Nanjing, Jiangsu Province, China as the study area,
and selects multi-source remote-sensing data, three-dimensional spatial distribution data
of urban buildings to downscale the Sentinel-3 LST with the spatial resolution of 1000 m
to the resolutions of 450, 150 and 30 m step-by-step using surface 2D and 3D parameters
as driving factors. The downscaling results are evaluated by Landsat TIRS LST at the
resolution of 30 m. Then, the influence of urban spatial morphological parameters on land
surface temperature downscaling is discussed. The step-by-step LST downscaling method
changes the traditional studies that directly downscale LST from a low spatial resolution to
a high one, selecting several spatial resolutions for the intermediate downscaling process.
The intermediate downscaling process is equivalent to supplementing the model with the
land surface information and reducing the difference in spatial resolution, thus ensuring
that the statistical regression model varies less with the spatial scale.
2. Research Review
A lot of research has been carried out on land surface temperature downscaling by
scholars around the world. The main methods used for LST downscaling can be divided into
two categories: image-based spatiotemporal fusion and kernel-driven downscaling methods.
2
Remote Sens. 2022, 14, 3038
The image fusion method obtains a high spatial and temporal resolution land of surface
temperature by constructing a model to generate fused images, based on the combined
weight of spectral, temporal and spatial information, by selecting similar images in the
spatiotemporal neighborhood to be fused. Unlike the statistical downscaling method, the
image fusion method does not directly model the relationship between land surface temper-
ature and influencing parameters at low-spatial-resolution scales. Classical algorithms are as
follows. Weng et al. [22] improved the STARFM model to establish the relationship between
MODIS and TM radiometric brightness by linear spectral mixing analysis, and proposed a
spatiotemporal adaptive fusion algorithm (SADFAT) for land surface temperature downscal-
ing. Wu et al. [23] proposed a spatiotemporal integrated temperature fusion model (STITFM)
for estimating high-spatiotemporal-resolution LST from multi-scale polar and geostationary
orbiting satellite observations. The image fusion-based approaches assume that the radiative
brightness for similar pixels behaves consistently at any spatial resolution, while in practice,
the radiative brightness will inevitably vary in space and time. So, this approach generally
performs poorly in urban areas with high-spatial-heterogeneity characteristics.
Kernel-driven downscaling methods can be classified into physical models and sta-
tistical regression methods according to whether the model is physically meaningful or
not. Physical downscaling methods establish the relationship between LST and auxiliary
data by using the physical mechanism of a mixture pixel and the thermal radiation prin-
ciple. In this way, low-spatial-resolution pixels are decomposed to multiple subpixels to
obtain the high-spatial-resolution LST. For example, L.J and Moore [24] developed the
pixel block intensity modulation (PBIM) method to improve the spatial information in
the low-spatial-resolution thermal infrared band by using multispectral data with a high
spatial resolution. Nichol [25] proposed the emissivity modulation (EM) model to improve
the spatial resolution of thermal radiation by using land surface emissivity and land cover
data. Wang et al. [26] downscaled MODIS LST to a 30-m resolution based on the thermal
decomposition equation. However, the design of physical models is usually difficult and
the models are computationally complex and time-consuming.
The basic principle of a statistical regression method is to assume that the relationship
between land surface temperature and driving factors does not change with the spatial
scale. A statistical regression model is built using the low-spatial-resolution LST and the
drivers, after which the high-spatial-resolution drivers are added to the model to predict the
high-spatial-resolution LSTs. Up to now, the statistical regression method is the most widely
used method in LST spatial downscaling studies. Based on the number of driving factors,
statistical regression methods can be divided into single- and multi-factor models. For
example, Distrad models used NDVI as the driver [27], and the TsHARP model used vege-
tation cover instead of NDVI [28]. In addition, Anthony et al. [29] developed a high-resolution
urban thermal fusion (HUTS) technique to downscale Landsat TIRS to 30 m based on NDVI
and surface albedo. Lacerda et al. [30] used the TsHARP model to downscale the MODIS
LST to a high spatial resolution of 10 m. Vaculik et al. [31] downscaled the GOES-R LST
with the resolution of 2000 m to 30 m by establishing a linear relationship between NDVI
and LST. J.M. et al. [32] modified the TsHARP algorithm to downscale MODIS LST data
covering one Spanish farm. However, single-factor models are only applicable to a region
with high vegetation cover; they do not perform well in urban or arid areas, limited by the
predictor variables. Multi-factor models with multiple remote-sensing indices and land
surface parameters as driving factors were gradually proposed and applied. For example,
Liu et al. proposed the G_Distrad model by adding NDBI and NDWI to the traditional
Distrad model [14]. Pereira et al. [33] proposed a geographically weighted regression model
(GWRK) by using NDVI and multispectral data to downscale the ASTER thermal infrared
data for the natural regions and urban areas of Pantanal, Brazil. Considering the complex
nonlinear relationships between land surface temperature and various geophysical parame-
ters in urban areas [12,13], the three-layer structural (TLC) model [34], neural network [35],
support vector machines [6], random forests [36] and other multivariate nonlinear statistical
models have been continuously developed and applied to urban LST downscaling studies.
3
Remote Sens. 2022, 14, 3038
Random forest models have been widely used in urban LST downscaling studies in recent
years because of their low model complexity, fast training speed and ability to effectively
avoid overfitting problems. Li et al. [36] compared three machine learning algorithms,
random forest (RF), support vector machine (SVM) and artificial neural network (ANN), to
the traditional TsHARP method in both urban and suburban areas of Beijing, and found
that the LST downscaling accuracy of the machine learning algorithm was higher than
that of the TsHARP algorithm. Wang et al. [37] compared the downscaling results from
a multiple linear regression model (MRL), TsHARP model and random forest (RF), and
found that the RF model is more suitable for heterogeneous surfaces such as urban areas.
Ebrahimy et al. [38] used an adaptive random forest regression method to downscale
MODIS LST over Iran to 30 m in the GEE platform. Njuk et al. [39] proposed an improved
downscaling method for low-resolution thermal data based on minimizing the spatial
mean bias of random forest, and the results demonstrated that the method reduces the
inherent mean bias in the LST downscaling process and is more suitable for LST down-
scaling applications in complex environments. Here, we comprehensively analyzed most
land surface temperature downscaling methods and built global models and assumed that
the statistical relationships were spatially invariant, however, global models may produce
large errors in local area applications. In recent years, many scholars have devoted their
work to using local models to capture the spatial non-stationary characteristics of land
surface variables, and then established the local relationships between LST and influencing
factors to improve the accuracy of LST downscaling [15].
3. Materials and Methods

3.1. Study Area
The central urban region of Nanjing, Jiangsu Province, China was chosen as the study
area because it contains a variety of underlying surface types such as buildings, vegetation
and water bodies, which helps to carry out land surface temperature downscaling studies of
complex ground cover types. Figure 1 presents a true-color image and building distribution
map of the study area at a spatial resolution of 10 m.
Figure 1. Sentinel-2 true-color composite image and building data of the study area (blue line
represents the main urban area of Nanjing; red lines represent the study area boundaries).
4
Remote Sens. 2022, 14, 3038
The study area includes several urban administrative districts, including the Qixia,
Xuanwu, Gulou, Qinhuai and Jianye Districts, with an area of approximately 18 × 18 km2 .
The study area is located in the central region of the lower Yangtze River, with geographic
coordinates between 31◦ 14 N and 32◦ 37 N and 118◦ 22 E and 119◦ 14”E. The total built-up
area is approximately 823 km2 . Although the study area is near a hilly area, the topography
is flat, and there are many low hills and gentle hills. Nanjing has a humid subtropical
climate with four distinct seasons, abundant rainfall and significant temperature differences
between winter and summer. The average annual precipitation is 1106 mm, and the average
annual temperature is 15.4 ◦ C. Nanjing had a population of 10,312,200 at the end of 2019,
with a resident population of 8.5 million, including 7.072 million in urban areas, and an
urbanization rate of 83.2%. Nanjing is one of the economic-center cities in China, with a
regional GDP of $1.6 billion in 2021.
3.2. Data
The Sentinel-3 LST product at a 1000-m spatial resolution for downscaling and Sentinel-2
multispectral image data were downloaded from the ESA website (https://scihub.copernicus.
eu/dhus/#/home, accessed on 14 May 2022). The Landsat LST product at a 30-m spatial
resolution for validation of downscaling results was downloaded from the USGS website
(https://earthexplorer.usgs.gov/, accessed on 30 December 2021). Sentinel-2 visible light
and shortwave infrared bands (B2–B4, B8, B11 and B12) were used to calculate normalized
remote-sensing spectral indices. Nanjing downtown building and wind data were used to
calculate urban spatial morphological parameters. Nanjing wind data from 2016 to 2020
were used to calculate the wind direction frequency, which were downloaded from the
China Air Quality Online Monitoring and Analysis Platform (https://www.aqistudy.cn/
historydata/, accessed on 14 May 2022). The details of all the data involved in this study
are presented as follows.
3.2.1. Landsat LST Product

Landsat LST products were generated by EROS based on a single-channel inversion
algorithm, by using the Landsat C2L1 thermal infrared band and other ancillary data [40,41].
Landsat LST products were resampled from 100 to 30 m for release to users by EROS using
the nearest-neighbor resampling method. The Landsat LST images used in this study were
imaged at 10:37 a.m. on 4 October 2021, with orbital row/column numbers 120/038 [42].
3.2.2. Sentinel-3 LST Product

The Sentinel series is an Earth observation satellite mission of the European Copernicus
program. Sentinel-3 carries a variety of payloads, such as OLCI (sea and land colorimeter)
and SLSTR (sea and land surface temperature radiometer), which are mainly used for
high-precision measurements of the sea surface topography, sea and surface temperatures,
ocean water color and soil properties [43]. Both 3A and 3B satellites in orbit have revisit
periods of less than one day for areas within 30◦ latitude of the equator. SLSTR has six
solar reflection bands (S1–S6) and four thermal infrared bands (S7–S9, F1, F2) with spatial
resolutions of 500 and 1000 m, respectively. The Sentinel-3 LST products are produced
by a split-window algorithm using three bands of S7–S9 and other auxiliary data, and
the products are accurate to 1 K. The LST product of Sentinel-3B SLSTR was selected for
this study, with an imaging time of 10:04 am on 4 October 2021. The Sentinel LST was
resampled to a spatial resolution of 900 m by using the bilinear interpolation method for
downscaling in this study to match the reference LST with a spatial resolution of 30 m.
3.2.3. Sentinel-2 Multispectral Data

Sentinel-2 is a multispectral high-resolution imaging satellite with two satellites in
orbit, 2A and 2B, with a revisit period of five days [44]. Each satellite carries a multispectral
imager (MSI), which can cover 13 spectral bands with ground resolutions of 10, 20 and
60 m and an amplitude of 290 km. The blue, green, red, and near-infrared bands needed for
5
Remote Sens. 2022, 14, 3038
this study are the B2, B3, B4 and B8 bands of the Sentinel-2 satellite, each with a resolution
of 10 m. B11 and B12 are shortwave infrared bands with a resolution of 20 m, resampled to
10 m to match the visible bands.
3.2.4. Urban Building Data

The building data used in this study were provided by Urban Data Corps, obtained in
around 2012. Urban Data Corps is ranked in the top 10 in the big-data field according to the
2017 China Big Data Development Report published by the National Information Center of
China. Urban Data Corps can provide a variety of high-precision data for urban research.
The building vector data contain the polygon of the building distribution, building
floor data and building height data in a shapefile format with the WGS-84 coordinate
system. Comparing the urban building distribution data with satellite images in 2012,
we found that the building location and outline highly overlap with the satellite images,
and the building floor number is also very consistent with the field survey results, which
indicates the high accuracy of the building distribution data. The field survey found that
the ground cover types in most of the study areas, such as Gulou District and Qinhuai
District, did not change significantly between 2012 and 2021, except for Qixia District.
In this study, the building vector data in the shapefile format were firstly converted to
raster data in the TIF format with a spatial resolution of 10 m. After that, the urban spatial
morphological parameters were calculated based on the building raster data and other
auxiliary data using corresponding formulas.
3.3. Calculation of the Downscaling Driving Factor

3.3.1. Calculation of the Remote-Sensing Spectral Index
The L2A-level surface reflectance data from Sentinel-2B were selected for this study to
calculate remotely sensed spectral indices that are closely related to surface temperature,
including the modified normalized difference water index (MNDWI), normalized differ-
ence building index (NDBI), normalized difference built-up and soil index (NDBSI) [28],
normalized difference moisture index (NDMI), normalized difference vegetation index
(NDVI) and soil adjusted vegetation index (SAVI). The calculation process was performed
using SNAP 8.0, a professional piece of software for data-processing in the Sentinel series.
The calculation formula is shown in Table 1.
Table 1. Remote-sensing spectral indices required for downscaling and calculation formulas.
Var. Description Equations

Improve the normalized difference water body index to highlight water ρ3−ρ12
MNDWI
body information MNDW I = ρ3+ρ12
Normalize the difference building index to highlight building ρ12−ρ8

NDBI
information NDBI = ρ12+ρ8

2×ρ11 ρ8 ρ3
− +
(ρ11+ρ8) (ρ8+ρ4) ρ3+ρ11
IBI = 2×ρ11 ρ8 ρ3
(ρ11+ρ8)
+ (ρ8+ρ4) + ρ3+ρ11
NDBSI Indicate the degree of dryness of the ground surface [45] (ρ11+ρ4)−(ρ8+ρ2)
SI = (ρ11+ρ4)+(ρ8+ρ2)
IBI +SI
NDBSI = 2
ρ8−ρ11
NDMI Indicate the vegetation moisture NDMI = ρ8+ρ11
ρ8−ρ4
NDVI Highlight vegetation information NDV I = ρ8+ρ4
ρ8−ρ4
Reduce the sensitivity of vegetation indices to changes in reflectance of SAV I = ρ8+ρ4+ L × (1 + L )
SAVI
different soils L = 0.5
Notes: ρ1–ρ12 denote the surface reflectance of Sentinel-2 bands 1–12.
3.3.2. Calculation of Urban Spatial Morphological Parameters

The building vector data of Nanjing were converted to raster data with a 10-m resolu-
tion to calculate urban spatial morphological parameters including the building density
(BD), frontal area density (FAD), floor area ratio (FAR), mean height (MH) and sky view
6
Remote Sens. 2022, 14, 3038
factor (SVF). The SVF was calculated by the raster algorithm. The influence of buildings
within a radius of 100 m was considered when calculating the SVF of each pixel. FAD was
calculated using the building raster data and the wind direction frequency data by a self-
developed raster algorithm with a plot area of 100 × 100 m2 . Other spatial morphological
parameters, such as BD, FAR and MH, were calculated using 10-m building raster data and
the corresponding equations in Table 2 with a plot area of 100 × 100 m2 .
Table 2. Urban spatial morphological parameters required for LST downscaling.
Var. Description Equations

n
BD = ∑i=A1T i
A
BD Building Density
Ai indicates the ith building area and AT indicates the calculated plot area
n A(θ )
proj(z)
λ f (z) = ∑ AT × Pθ,i
i =1
λf(z) indicatestheweightedfrontalareadensity(FAD);
FAD Frontal Area Density
A(θ ) proj(z) indicates the projected area at a certain height z
in the wind direction θ, Pθ,i indicates the frequency of the wind direction θ,
i = 1,. . . , 16
n
F ×A
FAR Floor Area Ratio FAR = ∑i=1ATi i
Fi indicatestheithbuildingfloornumberandAi indicates the ith building area
n
MH = ∑i=n1 i
H
MH Mean Height
Hi indicates the ith building height and n indicates the number of buildings
360/α
Ψsky = 1 − ∑ sin2 β × (α/360)
i =1
SVF Sky View Factor β = tan−1 ( H/X )
ψsky indicates SVF, β indicates the building height angle, H indicates the building
height, X indicates the calculated radius and is set to 100 m in this study
Note: Except for SVF, the plot area AT calculated by all other parameters takes the value of 100 × 100 m.
3.4. Downscaling LST Method Based on Random Forest

The core idea of surface temperature downscaling is the invariance of the relationship
between LST and driving factors at different spatial resolutions so that the statistical
relationship between LST and the regression kernel at a low resolution can be applied to
a high spatial resolution to complete the downscaling process. The specific formulas are
as follows:
Tc = f (varc ) (1)
ΔT = Tc − Tc (2)

T f = f var f + ΔT (3)
where varc denotes the low-resolution explanatory variable, varf denotes the high-resolution
explanatory variable, Tc denotes the low-resolution LST, Tc ’ denotes the predicted low-
resolution LST, ΔT denotes the simulation residual and Tf ’ denotes the predicted high-
resolution LST.
This study used the random forest algorithm to construct the LST downscaling model.
Random forest is an integrated decision tree-based learning algorithm proposed by Breiman
in 2001 as a supervised learning algorithm [32]. The algorithm uses the bootstrap resam-
pling method to randomly select samples from the training sample set. The extracted
bootstrap samples are trained separately for each decision tree, and an algorithm for ran-
domly selecting a subset of features is introduced in the process of splitting the nodes of
the decision tree. The prediction results of each decision tree are counted and voted on
to obtain the final prediction results of the input data. The random forest algorithm has
stronger noise immunity than other algorithms because of the introduction of randomly
selected training samples and randomly selected feature subsets that make the correlation
7
Remote Sens. 2022, 14, 3038
among each decision tree smaller. The random forest algorithm is better at handling non-
linear problems than traditional statistical regression algorithms. As long as the number of
decision trees is sufficient, the random forest algorithm can effectively avoid the overfitting
problem, and the training speed is faster. The random forest algorithm can examine the
interaction between features during the training process and output the feature importance,
which is a reference for analyzing the degree of influence of features. In this study, the
experimental dataset was divided into training and validation datasets according to the
ratio of 8:2. The main parameters that need to be set manually to build a random forest
downscaling model using the Pycharm platform include the number of decision trees
(n_estimators) and the maximum number of features (max_features). n_estimators refers to
the number of decision trees built in the random forest, which was set to 700 after testing in
this study. max_feature refers to the maximum number of features selected when building
each decision tree, which was set to log(n_estimators) in this study. Other parameters were
set to default values.
3.5. Step-By-Step Random Forest Downscaling Method (SSRFD)

When the spatial resolution spans a larger range, the downscaling results cannot
accurately characterize the spatial distribution of LST, which tends to underestimate the
surface temperature in building areas and overestimate it for water bodies and vegetated
areas. This study proposed a step-by-step downscaling LST method based on the random
forest model (SSRFD), which achieves a significant increase in the spatial resolution of
LST without excessive loss of accuracy through multiple, small-scale spatial resolution
downscaling processes. During the SSRFD model’s work, each intermediate downscaling
adds additional and finer surface information to the model. In this way, models are trained
to accurately express the relationship between land surface temperature and driving factors.
In this study, the direct random forest downscaling (DRFD) method was performed to
directly downscale 900-m Sentinel-3 LST to a 30-m resolution, and then two downscaling
methods were conducted using the proposed SSRFD method. The first method, named
SSRFD, downscaled the 900-m Sentinel-3 LST to 30 m after 450 m and 150 m sequentially,
where the SSRFD was driven by the six remote-sensing indices mentioned above. The sec-
ond method, named SSRFD-M, downscaled 900-m Sentinel-3 LST to 30 m by the same
process through SSRFD, where five additional urban spatial morphological parameters
were added as the SSRFD driver factors. After that, LST downscaling results of DRFD,
SSRFD and SSRFD-M were compared at a 30-m spatial resolution, which were all evalu-
ated by using the 30-m Landsat-8 LST as the reference. Moreover, the influence of urban
spatial morphological parameters on LST was analyzed based on SSRFD-M at a 30-m
spatial resolution.
3.6. Accuracy Evaluation Methods

Pearson’s correlation coefficient (r), the mean absolute error (MAE) and root mean
square error (RMSE) were used to comprehensively evaluate the downscaling results.
Meanwhile, the maximum/minimum, mean (Mean) and standard deviation (SD) were
calculated to evaluate the spatial variability characteristics of LST images before and after
downscaling. The SD can reflect the spatial variability of thermal features.
4. Results
4.1. Comparison of the Results Obtained with SSRFD and DRFD
To reduce data differences caused by different sensors and LST inversion algorithms
and increase the comparability and verifiability between Landsat-8 and Sentinel-3 LST
products, a simple linear correction was applied to Sentinel-3 LST before the downscaling
work. After performing the linearity correction, the maximum, minimum and mean values
of Sentinel-3 LST changed from 38.80, 27.57 and 35.18 ◦ C to 41.25, 30.84 and 37.90 ◦ C,
respectively, which were closer to Landsat LST in the range of values. The RMSE of the
8
Remote Sens. 2022, 14, 3038
two LST products changed from 3.22 to 1.59 ◦ C, indicating that the systematic differences
between Sentinel-3 LST and Landsat LST were reduced to some extent.
Comparative plots of the downscaling results are given in Figure 2, where the 900-m
Sentinel-3 LST was downscaled to 30 m using the DRFD method and SSRFD method. Com-
paring Figure 2b,c with Figure 2a, both results captured finer spatial discrepancy character-
istics and texture features, and the resulting LST distribution is basically consistent with
Landsat LST. However, according to Figure 2b, there are large areas of high-temperature
regions in the study area. The results obtained with DRFD as a whole are significantly
overestimated. For example, the regions along the northwestern coast of the Yangtze
River, Xuanwu Lake and Zijinshan Mountain show obvious temperature overestimation
errors. The difference characteristics between the high-temperature region and the sub-high-
temperature region are less clearly expressed than Landsat LST in Figure 2a. In contrast to
Figure 2b, the results obtained with SSRFD (Figure 2c) captured the spatial distribution
differences and textural characteristics more accurately. The distribution characteristics of
both the building high-temperature zone and the water and vegetation low-temperature
zone are in good agreement with Figure 2a. Overall, the results obtained with DRFD show
an overall underestimation of the high-temperature region and an overestimation of the
low-temperature region, which cannot accurately depict the spatial distribution pattern of
LST in the study area.
(a) (b) (c)
Figure 2. Spatial distribution of LST at 30-m spatial resolution. (a) Landsat reference LST. (b) Down-
scaled LST results of DRFD. (c) Downscaled LST results of SSRFD.
The statistics (Table 3) show that the results from DRFD, with an SD of 2.64, are 0.82
lower than Landsat LST, but their mean value is 1.41 ◦ C higher than Landsat LST. This is
consistent with the performance of the DRFD results in Figure 2b, which further illustrates
the overall high surface temperature predicted by DRFD. In comparison, the maximum,
mean and SD of the results from SSRFD are 50.76, 36.78 and 3.73 ◦ C, respectively, which
only differ from the corresponding index of Landsat LST by approximately 0.3 ◦ C. In
summary, the downscaling results of SSRFD are more accurate, which is also demonstrated
in Figure 3.
Table 3. Statistical values of LST downscaled from DRFD, SSRFD methods and Landsat reference
LST data at a 30-m resolution.
Statistical
Reference LST/◦ C DRFD LST/◦ C SSRFD LST/◦ C
Variables
Maximum 51.06 42.80 50.76
Minimum 13.94 28.95 24.36
Mean 36.50 37.91 36.78
Standard deviation 3.46 2.64 3.73
9
Another random document with
no related content on Scribd:
excursions in the neighbourhood. The country around is beautiful
and very fertile. There are a number of traders established there,
and a large assemblage of native huts and houses. There is also a
fine old church, the only remaining evidence of the existence of the
old missionaries. Both this and the church at Muxima contain great
numbers of small bats. The roof inside is completely covered with
them. I have noticed a very curious circumstance in Angola with
regard to these bats, and that is the way they issue at dusk from any
window or crevice communicating with the interior of the roofs or
other dark places that they occupy during the day.
At regular intervals of about thirty seconds to one minute, a small
puff or cloud of these bats is seen to issue together, and so they
continue till all are out. This strange habit of leaving their hiding-
places intermittently, and not continuously as might be expected,
cannot easily be explained, nor why they assemble together to go
forth in distinct batches only. Whether they return in the same
manner I cannot say. Once out, they seem to spread apart
immediately, and fly away in all directions, and do not appear in the
least to keep in flocks like birds, though they may roost together in
communities.
The town of Dondo is about twenty miles from Massangano. It
stands in a small, triangular, level plain surrounded by hills on all
sides, the base of the triangle being the River Quanza and the low
line of hills on its southern bank. From the configuration of the
ground, shut in on all sides from winds, it is perhaps, as might be
imagined, the hottest corner of Angola. The heat in a calm summer’s
day is almost stifling, and the nights, generally cool everywhere else,
are not less oppressive. Formerly the town was on the high land
above, at Cambambe, as the town itself was called, but the
exigencies of trade have peopled the present town of Dondo.
It is a growing and flourishing place, where a number of traders
and agents of Loanda houses are established, and is the receiving
port for embarkation of the produce and trade of the neighbouring
districts and of those of the interior. Thousands of tons of ground-
nuts, coffee, wax, palm-oil, ivory, &c., are shipped yearly at Dondo
for Loanda by the steamers. There is a fine large square in the
middle of the town, where a fair or market is held every day, and to
this the natives resort from all parts around with produce and
provisions. Many different tribes from the interior are to be seen at
Dondo, both from the northern and southern banks of the river, who
have brought produce for sale to the white men. The “residencia” of
the “chefe” is on a hill to the south of the town, and the view from it is
truly magnificent. As far as the eye can reach it is one gorgeous
scene of mountains, dark palms, and forests, range after range, till
lost to view in the horizon. There are two views in Angola that would
alone almost repay the trouble of travelling there. One is that just
described, and the other from the hill at Tuco on the road from
Ambriz to Bembe.
About six or eight miles from Dondo up the river are the first
cataracts of Cambambe. Immediately on leaving Dondo the river is
enclosed by high hills or cliffs on both sides, and winds a good deal,
so that a succession of fresh and seemingly more beautiful pictures
is constantly presented to the traveller’s admiration as he ascends
the river in a boat. The river is wide and deep, and the slopes and
perpendicular sides of the hilly walls on either side are of endless
variety of colour, both of rock, moss and lichen, plant and tree. Deep
red iron-stained sandstone, conglomerate, blue clay slate, huge
white-stemmed baobabs, dark masses of palm-trees, plots of large-
leaved plantains, masses of trees overgrown with creepers, meet the
eye in ever varying combination, and gradually the wide valley worn
by the water becomes narrower and narrower, until at last it is a
deep gorge with almost upright walls of clay slate, and the passage
for the great body of water is barred right across by vast rocky
ledges and peaks, over which, in the rainy season, it rushes and
dashes with a deafening wild roar and mad flinging up into the air of
showers of water and foam. The last time I saw these rapids I was
accompanied by my wife, and we landed on a bank of gravel a little
below the cataract and walked and scrambled on the rocks, till we
were on a great ledge quite close to, and but little over the level of
the waters; but, it being the end of the dry season, they were so far
reduced as to run between the rocks in a swift dark oily-looking
mass, and at such a considerable inclination and speed as to give
the idea of vast and irresistible force. On the rocks covered over and
splashed by the water, were growing masses of a curious semi-
transparent plant with thick stems, and bearing minute white flowers.
The singular appearance of this plant, so exactly like a sea-weed,
attracted our attention, and as I had never before observed it
anywhere in my travels in Angola, we secured specimens, which we
dried and preserved, and on forwarding them to Kew it was found to
be a new genus of Podostemaceæ, and has been described by Dr.
Weddell in the ‘Journal of the Linnean Society,’ xiv. 210, † 13, as the
Angolæa fluitans.
It is said that coal has lately been discovered near the river on its
southern bank, and not far from Dondo.
Of the old town of Cambambe, situated on the high ground over
the cataract, but little else remains than the church and a few
houses. The River Quanza is not navigable beyond the rapids,
except perhaps for short distances and for small canoes. At
Nhangui-a-pepi it is only a broad shallow stream, but with a very
rocky bed; its source, however, is far beyond Pungo Andongo.
About the higher parts of the river a gigantic species of the “Bagre”
(Bagrus) is found. This is a siluroid fish, and my attention was first
called to its extraordinary dimensions by seeing a black using the flat
top of the skull of one as a plate or dish. I was then in the province of
Cambambe, and several times had an opportunity of asking natives
from Pungo Andongo the size of these fish: one man told me that
they were captured so large that two men were required to carry one
fish slung on a pole on their shoulders by passing the pole through
the gills, and that the tail then drags on the ground. Another black,
who was a river fisherman, explained to me that the “bagres” were
caught with an iron hook made on purpose by the native smiths and
baited with a piece of meat; he gave me an idea of the thickness and
size of the hook with a piece of twig, and it was as large as an
ordinary shark-hook; he further drew on the ground the size of this
large fish, and it was six feet long. Other natives who were with him
joined in the description and corroborated him, and I am perfectly
convinced that they spoke the truth, and were not exaggerating
much. I wrote to some Portuguese traders at Pungo Andongo,
asking them to send me the head of one in spirits, but of course I
never got it.
Only the northern bank of the Quanza is subject to the Portuguese
(with the exception of the town of Muxima), and the natives
inhabiting it are greatly civilized and well behaved, and very civil.
They are, of course, not industrious, the women cultivating the usual
mandioca and other produce for food, and manufacturing palm-oil. A
little tobacco is also cultivated by them, and the leaves when fully
grown are gathered, and a string passed through the stem. This
string of leaves is stretched round their huts to dry, and the large
leaves thus hanging give them a curious appearance.
The first trips of the steamers caused the natives on either bank
the most intense astonishment; they would race them along the
banks, shouting and yelling; and when the steamers stopped at any
place, crowds would flock round and come on board to stare at the
machinery, which was universally pronounced to be a white man’s
great “fetish.”
All natives are very fond of the “batuco” or dance, of which there
are two kinds. The Ambriz blacks and those of the Congo country
dance it in the following manner:—A ring is formed of the performers
and spectators; “marimbas” are twanged and drums beaten
vigorously, and all assembled clap their hands in time with the
thumping of the drums, and shout a kind of chorus. The dancers,
both men and women, jump with a yell into the ring, two or three at a
time, and commence dancing. This consists almost exclusively of
swaying the body about with only a slight movement of the feet,
head, and arms, but at the same time the muscles of the shoulders,
back, and hams, are violently twitched and convulsed. The greatest
applause is given to those who can most strongly shake their flesh
all over in this way. It is difficult to do, and appears to require
considerable practice, and seems very fatiguing, for in a few minutes
the dancers are streaming with perspiration and retire for others to
take their places, and so they will often continue for a whole night
long, or in dark nights, as long as the great heap of dry grass that
they have provided lasts—the illumination being obtained by burning
wisps of this grass, two or three blacks generally having the care of
that part of the performance. The natives at these dances are
dressed as usual in the ordinary waistcloth, the men arranging theirs
so as to allow the ends to trail on the ground. There is nothing
whatever indecent in them.
The “batuco” of the Bunda-speaking natives of Loanda and the
interior is different. The ring of spectators and dancers, the
illuminations, the “marimbas” and drums are the same, but only two
performers jump into the ring at a time, a man and a girl or woman;
they shuffle their feet with great rapidity, passing one another
backwards and forwards, then retreat facing one another, and
suddenly advancing, bring their stomachs together with a whack.
They then retire, and another couple instantly take their places. This
performance might be called somewhat indecent, but I do not believe
that the natives attach the most remote idea of harm to the “batuco.”
The “marimba” is the musical instrument par excellence of the
natives of Angola. In Plate XI. is represented the better made ones.
It consists of a flat piece of wood, generally hollowed out, and with a
number of thin iron tongues secured on it by cross bits, but so as to
allow them to be pulled out more or less for the purpose of tuning. In
front is affixed a wire, on which some glass beads are loosely strung
that jangle when the instrument is played, which is done by holding it
between both hands and twanging the tongues with the thumbs. The
light wood of which the “marimbas” are made is that of the cotton-
wood tree. They are also made with slips of the hard cuticle of the
stem of the palm-leaf instead of the iron tongues, but these are the
commoner instruments. Others are made smaller and with fewer
tongues.
The more complete ones have an empty gourd attached to the
under part, which is said to give them greater sonorousness. The
blacks are excessively fond of these instruments everywhere in
Angola, playing them as they walk along or rest, and by day or night
a “marimba” is at all hours heard twanging somewhere. The music
played on it is of a very primitive description, consisting only of a few
notes constantly repeated.
Another common instrument of noise, much used to accompany
the “marimbas” and drums at the “batucos,” is made by splitting a
short piece of palm stem about four or five feet long down one side,
and scooping out the soft centre. The hard cuticle is then cut into
little grooves across the slit, and these, energetically rubbed with a
stick, produce a loud, twanging, rattling kind of noise.
A musical instrument sometimes seen is made by stretching a thin
string to a bent bow, about three feet long, passed through half a
gourd, the open end of which rests against the performer’s bare
stomach. The string is struck with a thin slip of cane or palm-leaf
stem held in the right hand, and a finger of the left, which holds the
instrument, is laid occasionally on the string, and in this way, with
occasional gentle blows of the open gourd against the stomach, very
pleasing sounds and modulations are obtained.
Another very noisy instrument with which the drums and
“marimbas” are sometimes accompanied at the “batucos,” is made
by covering one end of a small powder-barrel or hollow wooden
cylinder (open at both ends), with a piece of sheepskin tied tightly
round it. A short piece of round wood, about six or seven inches
long, is pushed through a hole in the middle of the sheepskin cover,
a knob at the end preventing it from slipping quite in. The hand of the
performer is then wetted and inserted into the cylinder and the piece
of wood is lightly grasped and pulled, allowing it to slip a little, the
result being a most hideous, booming sound.
I would strongly recommend my youthful readers, if they would like
to create a sensation in a quiet household, to manufacture one of
these simple and efficacious musical instruments, and I would
suggest the application of a little powdered resin instead of water to
the hand, to produce a full tone.
I once saw at a town near Bembe a musical instrument which I
thought rather ingenious. A small rectangular pit had been dug in the
ground, and over its mouth two strings, about six inches apart, were
stretched with pegs driven in the ground. Across these strings ten or
twelve staves from a small powder-barrel were fastened. These were
struck with a couple of sticks, on the end of which was a little knob or
lump of india-rubber, and an agreeable sound was produced. I have
seen two Kroomen from the West Coast, at the River Zaire, playing
on a similar kind of instrument, but the flat pieces were laid across
two small plantain-stems, and were of different sizes and thickness,
so as to produce a kind of scale when struck by the performers with
a couple of sticks each. The rapidity with which this instrument was
played was really marvellous, and the music sounded like variations
of their usual plaintive song, always in a minor key, and one seemed
to be playing bass to the other’s rapidly-executed treble. This air is
played or sung ad infinitum, and the second bar is often repeated.
The Kroomen on board the steamers on the coast always sing it and
harmonize it prettily, when it has a very pleasing effect indeed.
The southern bank of the Quanza, from its mouth to opposite

Dondo, is called the Quissama country, and is inhabited by the
peculiar race or tribe of negroes of the same name. They have not
been subjected to the Portuguese in modern times, and I apprehend
that in former years, when the Portuguese were in great strength on
the River Quanza, they were never considered worth the trouble of
subjugating, as they certainly are not now. The former missionaries
also do not appear to have been able to do anything with them, as
not a trace exists there of the habits of civilization they so
successfully introduced, and which are so apparent in the natives of
the greater part of Angola, where they were formerly established.
Their greatest stations were on the Quanza, where their efforts were
most successful; and there can be no reason to suppose that any
other obstacle existed to the Quissama natives participating in their
teaching or example than the resistance due to the very low type,
both physical and mental, of this tribe, so apparent at the present
day. The missionaries must have had a station of some importance
at Muxima, in their own country, as shown by the very fine church
still existing there, besides those at Calumbo, Massangano,
Cambambe, and perhaps at Bruto.
The Quissama negroes are very black in colour, undersized,
exceedingly dirty, and have a remarkably ugly cast of countenance.
With the exception of the tribe of Muquandos, south of Benguella,
the Quissama blacks are the most miserable-looking race in Angola.
They have a wild, savage look, not seen in the faces of any other
tribe, and have not the free mien or attitude of perfect ease of other
blacks, but appear frightened and very suspicious. Everywhere on
the river they cross over daily with their produce for sale to the
different houses of the white traders, as well as to the petty native
grog-shops and traders on the river from Calumbo to Dondo, but
they never drop their distrustful behaviour, and cross over to their
own side without delaying more than necessary. They will not allow
traders to establish on their side of the river;—one or two that are
said to have done so were robbed, and their houses burnt. They are
on terms of perfect friendship with the natives and white men of the
northern bank, and at Calumbo and a few other places their land is
cultivated by the natives of those places.
The greater part of the Quissama country is very barren, and
perfectly destitute of water except on the banks of the river itself, the
Quissamas employing baobab trees, hollowed out for the purpose,
as reservoirs for the rain-water falling in the wet season. It is very
scantily populated except near Massangano and Dondo, where their
chief towns are, and where they manufacture a considerable quantity
of palm-oil.
When the oil season is approaching, the white traders go over
from Dondo to their principal towns to settle with the “sobas,” or
kings, the price per measure at which the oil is to be sold. Though so
wild in appearance, they are most inoffensive and peaceable, and
are in the greatest fear lest the Portuguese should take it into their
heads to annex their territory, which they could most easily do if they
thought it advisable.
The Libollos, or natives of the Libollo country, are a very much
finer and cleaner race than their neighbours the Quissamas, and
their country (according to the accounts I received from several
Portuguese and natives) most beautiful and fertile, and covered in
great part with palm-trees.
The manner in which they have lately cultivated large quantities of
ground-nuts, and increased the production of palm-oil, proves that
they are an industrious race.
They are antagonistic to the Quissamas, and very favourable to
the Portuguese, and have often offered to reduce them to the power
of the latter, if these will give them leave and supply them with guns
and ammunition. On one occasion, when I was at Dondo, an
embassy arrived, through the Libollo country, from some powerful
tribes of the Bailundo district—the most warlike of the tribes of the
interior—also offering to conquer the country for the Portuguese on
condition of being allowed to plunder and carry off prisoners as
slaves.
These warlike tribes to the south of the Libollo and Quissama
countries are known in Angola by the name of Quinbundos, and are
the handsomest natives of any, being all very tall and well formed.
They come in caravans to Dondo, principally laden with beeswax,
singing on the march, and at night when assembled together, a song
with chorus with great effect. They put up at the Libollo and
Quissama towns, and come over to the northern bank to trade with
the white men. They plait their hair in thin strings all round their
heads, and in each plait they put several beads, mostly made of red
paste in imitation of coral. The Quissamas are not cannibals, as
described by an English tourist, whose sensational story about them,
written after a few days’ trip in the steamers on the Quanza, is full of
gross inaccuracies.
The dress of the men is a waistcloth of fine matting or cotton
cloths, obtained from the traders; that of the women is very singular,
being the soft, beaten inner bark of the baobab-tree made into a
thick sort of skirt, which is fastened round the waist, and has extra
layers at the back to puff it out still more, something in the manner of
the “dress-improvers” worn by the fair sex in our own country. (Plate
XII.). Nature has provided the Quissama ladies with an abundant
development of what the Spaniards call “enthusiasm,” and on this
account the use of the extra thicknesses on the back of their skirts is
really quite unnecessary; but they are not satisfied, and consider this
fashion an improvement. Their appearance, therefore, is very
comical, particularly when they run, as the thick short skirt moves up
and down, and swings round with every motion of the body.
Plate XII.
Maxilla, and Barber’s Shop.—Carrying Corpse to Burial.—
Quissama Women, and manner of pounding and sifting Meal in
Angola.
To face page 147.
They are very dirty in their habits, never appearing to make use of
water for washing, and their baobab skirts are always black with
grease, smoke, and perspiration. I had to order the two dresses in
my possession, one for a married woman and the other for a girl or
young woman, to be made on purpose, as I could not touch any of
those offered to me for sale.
The women, when they come over to the northern bank,
sometimes wear a handkerchief or other cloth over the breast, and
even over the baobab skirt, but this latter they must wear, according
to the custom of their country. They carry the produce of their
plantations in large conical baskets on their backs, secured by a
band round the forehead (Plate XII.). It is, as far as I know, the only
tribe in Angola that carries a load in this manner.
The Quissama blacks are extremely poor, their arid country
producing hardly anything besides the food necessary for their bare
subsistence, and a little beeswax. The principal food of those near
the river is fish.
There is a deposit of rock-salt in the Quissama country
somewhere between Muxima and Calumbo (said to be south of the
former), and at some distance from the river. It has never to my
knowledge been visited by any white man, nor would the Quissamas
readily allow one to go to the place; but the most curious thing
connected with this salt is that they cut it into little bars with five or
six sides or facets, about eight or nine inches long and about an inch
thick, tapering slightly to the ends, and closely encased in cane-
work. These pass as money, not only on the river, but in the interior,
where they are at last perhaps consumed.
During the Abyssinian war, some of the correspondents described
exactly the same shaped pieces of rock-salt encased in similar
wicker-work, as being obtained and employed in that country for the
same purpose. This is extremely interesting, and opens several
questions as to a possible common origin for the custom in the far
and dim past; and the case of the bellows already described is
another similar instance.
Many of the native words mentioned by the same correspondents
are identical with those used in different places in Angola. I am very
sorry now that I did not devote more attention to the investigation of
the languages of the natives of Angola, and in particular that of the
Quissama tribe, which is different to the Bunda language, and is also
said to be different to that of Benguella Velha and Novo Redondo
farther south. The number of distinct languages and dialects in
Angola is very curious, and a similar multiplicity of tongues has been
noted by travellers in other parts of Tropical Africa. None of the
languages in Angola are guttural, or spoken with a “click.”
There is a great deal of most interesting detail to be worked out in
Angola in every branch of natural history and ethnology.
My chapters are little more than an indication of the wealth that
lies there buried for future explorers, and of the success that will
attend their investigations.
CHAPTER VI.
COUNTRY SOUTH OF THE RIVER QUANZA—
CASSANZA—NOVO REDONDO—CELIS—
CANNIBALS—LIONS—HOT SPRINGS—BEES—
EGITO—SCORPIONS—RIVER ANHA—
CATUMBELLA.
The country south of the River Quanza is very different from that
to the north of it, just described, not only in its physical aspect, but
also in the tribes of natives inhabiting it. The evidences of a former
degree of civilization, and of the good work of the old missionaries,
are not here visible, and I should almost imagine that this part of
Angola was not under their care to anything like the same extent.
From June 1861 to the end of 1863, I was engaged in working two
copper deposits at Cuio and Benguella, and in exploring the coast
from Cassanza, about eighty miles from the River Quanza, as far as
and including Mossamedes or Little Fish Bay.
In these explorations I did not go inland a greater distance than
about thirty or forty miles at Mossamedes, and forty or fifty at Novo
Redondo. I cannot, therefore, speak from personal knowledge of
those most interesting places in the interior, Bihé and Bailundo, or
the Portuguese districts of Caconda, Quillengues, Huilla,
Capangombe, &c.
The geological character of the coast-line from the Quanza to
Mossamedes is gneiss, mostly very quartzose, then with a good deal
of hornblende and mica near Cuio, passing to a fine-grained
porphyry and fine granite with large, distinct feldspar about
Mossamedes. Close to the sea these primary rocks are joined by a
line of tertiary deposits, principally massive gypsum, and sandstones
of different thicknesses curiously separated by layers of the finest
dust. Farther south, between the River San Nicolao in 14° S. lat. and
Mossamedes, there is a strip of columnar basalt and trap-rock of
only a few miles in width.
The character of these rocks is sufficient to account for the very
sterile nature of the country; in fact, most of it is completely a rocky
desert, without a drop of water, and covered with but little grass, and
frightfully thorny bushes. Although this is the general character, there
are numerous places of the greatest beauty, particularly at a
distance of twenty to thirty miles from the coast, where the first
elevation is reached, and where the vegetation, as in the rest of
Angola, changes to a luxuriant character.
The country about Cassanza is level and well covered with grass,
and the natives appeared inoffensive and quiet. They have a
considerable quantity of fine cattle, and what is rare amongst the
natives of Angola, they milk the cows regularly twice a day, the milk
being a principal article of food with them. The few days that I was
there in 1863, I enjoyed the abundance of beautiful milk immensely.
The Portuguese with whom I was staying was then engaged in
cotton planting, but the ground did not appear very suitable for its
cultivation. He also had a beautiful cotton and sugar-cane plantation
at Benguella Velha, and at a pretty place called Cuvo, where there is
a small river and good ground near its mouth.
On that occasion I had come up in a sailing barge from Benguella
to Novo Redondo, to explore that district for copper, specimens of
the ore having been found in several places. The river at Novo
Redondo had overflowed its banks, and the road we had to follow
was under water for some miles, and whilst waiting for the river to
subside, I started to Cuvo and Cassanza to see the country and my
friend. On returning to Novo Redondo I obtained for guide the
services of a jovial and useful black named David, who had been
educated at Benguella. He could read and write Portuguese, which
language he spoke perfectly, and was a man of great importance in
the Novo Redondo country, as he was the hereditary king of the
place, and was to be proclaimed as such as soon as he could make
up his mind to eat a man’s head and heart, roasted or stewed, as he
should fancy. David was not at all inclined either to forego his
kingship, or to eat any part of one of his fellow-creatures, which by
the custom of his country it was imperative he should do to be
proclaimed king.
He had been putting off the disagreeable ceremony for some two
years, if I remember right, but his people were getting impatient at
not having a king, and were threatening to elect another. How he got
over the difficulty, or if he at last submitted to overcome his
repugnance to roast or stewed negro, I never heard.
The “Mucelis,” or natives of Novo Redondo and of the country
inland called “Celis,” are cannibals, and, as far as I could ascertain,
there are no others in Angola.
The Portuguese have no stations inland on that part of the coast,
that of Caconda, to the interior and south of Benguella, being the
first; and they do not allow the practice of cannibalism at the town of
Novo Redondo itself, as they strictly prohibit and punish there, as in
the rest of Angola, any fetish rite or custom, but I found that at
Cuacra, the second large town I passed on my way inland, human
flesh was eaten, and in several other towns I passed I saw
evidences of this custom in a heap of skulls of the blacks that had
been eaten in the centre of the towns, and on the trees were also the
clay pots in which the flesh was cooked, and which, according to
their laws, can only be used for that purpose.
One night I walked out of my hut at a town where I was sleeping,
and seeing that no one was about, I chose a nice skull from the
heap, and brought it home and presented it to my friend Professor
Huxley, who exhibited it at a meeting of the Anthropological Society. I
had previously asked whether I might take one of these skulls, but
had been told that it would be considered a great “fetish” if I did, and
David begged me not to do so, as there would be a great
disturbance, so I was obliged to steal one in the way I have
described, and hide it carefully in my portmanteau.
It is only natives who are killed for “fetish” or witchcraft that are
eaten, and the “soba” or king of the town where they are executed
has the head and heart as his share.
I was informed that at these feasts every particle of the body was
eaten, even to the entrails. At the principal towns of Ambuin and
Sanga (said to be the capital) I was told that as many as six and
seven blacks were eaten every month, and that the “sobas” of those
two towns, and their wives, only used human fat to anoint their
bodies with.
I was shown at one of the towns the little axe with which the poor
wretches were decapitated, and which was distinguished from others
used by the natives by having a lozenge-shaped hole in the blade.
I was very much surprised to find that, notwithstanding their
cannibal propensities, the natives of Novo Redondo were such an
extremely fine race; in fact, they are the finest race of blacks, in
every way, that I have met with in Africa.
Cannibalism may possibly be one reason of their superiority, from
this custom supplying them with a certain amount of animal food
more than other tribes make use of, or it may be due to their usual
food, which is principally a mixture in equal parts of haricot beans
and indian-corn, being very much more nutritious than the diet of
mandioca meal, of almost pure starch, that supplies the staple food
of other tribes. Whatever the reason may be, there can be no
question of the superior physique and qualities of this cannibal tribe.
When about to start on my journey, I saw that only four carriers
had been provided for my hammock, and I refused to start with less
than six or eight, as I made sure, judging from every other place in
Africa I had travelled in, that I should have to walk a great deal, as
four men, even in Ambriz, where I had found the best carriers, would
not be able to carry me, day after day, on a long journey. I was
assured that it was never customary to have more than four, that two
would carry me from daybreak till noon, and the other two from noon
till sunset, and that I might have six or more, but that four alone
would carry me every day. This I found was the case, not only in that
journey, but also when returning overland from Novo Redondo to
Benguella, a distance of about ninety miles.
Another extremely curious feature, distinguishing them favourably
from all other negro races, is their degree of honesty and honour.
Any white or other trader going into the interior agrees to pay the
“soba” of a town the customary dues, and he provides the trader with
a clean hut, and is responsible for the goods in it. The trader may go
away farther inland, and he is perfectly certain that on coming back
he will find his property untouched, exactly as he left it. Whilst I was
at Novo Redondo, an embassy arrived from a town in the interior,
where a Portuguese had established himself to trade in palm-oil and
beeswax, and where he had died, bringing every scrap of produce
and goods belonging to him to deliver the same to the “chefe.”
They were paid and rewarded for their honesty, and I was told that
it was the usual thing for these natives to do, on the death of a trader
in their country. I do not know of any other part of Africa where such
an example would be imitated, certainly not by the Christian negroes
at Sierra Leone.
There is a magnificent palm forest on the banks of the river at
Novo Redondo. This river is small, but brings down a considerable
body of water in the rainy season.
I crossed it on the second day of my journey inland by means of a
curiously constructed suspension-bridge attached to the high trees
on either side. This bridge was made entirely of the stems of a very
tough tree-creeper, growing in great lengths, and about the thickness
of an ordinary walking-stick. From two parallel ropes made of this
creeper, right across the stream and about two feet apart, hung a
frame of open, large-meshed basket-work about three feet deep,
forming a kind of flexible net or trough open at the top. The bottom or
floor of this trough was made of the same creeper, woven roughly
and openly in the same manner as the sides, and when walking in it,
I found it necessary to be careful to tread on the network, or my feet
would have slipped through, and to help myself along by holding on
to the guys or ropes at the top, which reached up to about my waist.
The length of the bridge must have been some thirty paces. Near
it I noticed, on a flat-topped tree of no great height, a large bird of the
eagle species sitting on its eggs in an open nest, and the male bird
on a branch near his mate; this tree was quite close to the road or
path, and though numerous natives passed under it to and fro,
neither they nor the birds seemed to heed one another in the least.
The Celis country is infested with lions, but I was not so fortunate
as to see one, though one morning we came upon the fresh
footprints of what the natives affirmed to be a family consisting of a
full-grown male and female and three half-grown young ones. To my
inexperienced eyes there appeared to have been more, so
numerous seemed the plainly-marked footprints in the moist sand of
the bottom of a small ravine.
We escaped an encounter with one the day we started on our trip.
About half-past four in the afternoon we arrived at a pretty clump of
trees round a pool of deliciously cool water, and near a low line of
rather bare-looking hills. David would not allow our carriers to tarry at
this pool, as he knew it to be the evening drinking-place for the lions
living in the low hills near. We went on, and shortly after we met an
old Cabinda man on his way to Novo Redondo, carrying a letter tied
in a cleftstick (the usual way to send a native with a letter in Angola).
He was an acquaintance of David’s, who had a talk with him, and we
went on our several ways. Next morning we heard that the poor
fellow had been caught by a lion not more than an hour after, and at
the very pool of water where David had warned us not to stop long.
The lion had evidently eaten part of the body at the pool itself, and
had carried off the rest to its lair in the hills.
I went to several places where indications of copper had been
found, but was disappointed in finding any worth exploring. They
were all in the recent beds at the junction or near the primary rock of
the country, and consisted of indications of blue and green carbonate
of copper in the fine sedimentary mud and sandstone beds. These
indications are most abundant everywhere in that district, and
curiously enough the plantain-eaters are also most abundant, more
so than in any other part of Angola I have been in. I went as far as a
range of very quartzose schist rock or gneiss mountains, called
Ngello, which I suppose to be between forty and fifty miles from the
sea; and at a pass called Tocota on the road to some important town
in the interior, named Dongo, I visited a hot-water spring about half
way up the mountain side. I had no thermometer with me, but the
water, as it issued from a crevice in the rock, was so hot that I could
only keep my hand in it for a few seconds.
The direction of the mountain range was about N.N.E by S.S.W.;
the rock composing it was nearly vertical, inclining slightly to the
west, and with a strike about north and south. There is a most
picturesque little town of huts stuck on a rocky ledge, and the natives
use the water from the hot springs to drink, but first allow it to stand a
day to cool. It has a very pleasant taste when cold, with just the
slightest ferruginous flavour. From this range of mountains
magnificent views are obtained, the scenery and vegetation
reminding me strongly of Cazengo; and there can be no question
that it is likewise capable of growing the coffee-plant to perfection.
Some sugar-cane I saw growing there was as fine as I have ever
seen it, and the native plantations were most luxuriant. I do not know
whether trade at Novo Redondo has increased in the same ratio as
on the Quanza and Ambriz, but that it is destined to be a very rich
country I have no doubt. There is a great deal of white gum in the
country, collected from a tree of which whole forests are said to be
found.
The principal article of trade at Novo Redondo when I was there
was palm-oil, which was mostly bought in exchange for rum,
measure for measure, and I often saw the very gourds and pots in
which the natives brought the palm-oil filled up with the rum in
exchange without any more cleansing than allowing the vessels to
well drain off the oil.
I noticed a great variety of birds, and I am sure the country would
well repay a collector’s trouble. In the middle of a small cultivated
valley I saw a low, flat-topped baobab, which had been taken
possession of by a flock of eight or ten birds about the size of a
thrush, of a black colour, with smoky-white feathers on the wings.
They had built a common nest on the flat top of the tree, and were all

Full download Thermal and Optical Remote Sensing John O Odindi file pdf all chapter on 2024

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Full download Thermal and Optical Remote Sensing John O Odindi file pdf all chapter on 2024

Uploaded by

Copyright:

Available Formats

Thermal and Optical Remote Sensing

Optical Remote Sensing of Land Surface. Techniques and

Bio-optical Modeling and Remote Sensing of Inland

Remote sensing digital image analysis Sixth Edition

Atmospheric Remote Sensing Abhay Kumar Singh

Remote Sensing Digital Image Analysis 6th ed. 2022

Remote Sensing of Ocean and Coastal Environments Meenu

Hyperspectral Remote Sensing: Theory and Applications

Atmospheric Remote Sensing: Principles and Applications

ISBN 978-3-0365-6275-9 (Hbk)

Cover image courtesy of Getty Images

About the Editors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii

Xiangyu Li, Guixin Zhang, Shanyou Zhu and Yongming Xu

Tshilidzi Manyanya, Janne Teerlinck, Ben Somers, Bruno Verbist

Terence Darlington Mushore, Onisimo Mutanga and John Odindi

Ali S. Alghamdi, Ahmed Ibrahim Alzhrani and Humud Hadi Alanazi

Kangning Li, Yunhao Chen and Shengjun Gao

Terence Mushore, John Odindi and Onisimo Mutanga

Urbanization, typiﬁed by land-use-land-cover transformation is a major cause of bio-physical,

John O. Odindi, Elhadi Adam, Elfatih M. Abdel-Rahman, and Yuyu Zhou

Academic Editors: Yuyu Zhou,

Remote Sens. 2022, 14, 3038. https://doi.org/10.3390/rs14133038 1 https://www.mdpi.com/journal/remotesensing

High-temporal-resolution data are difﬁcult to generate for reﬁned surface temperature

3. Materials and Methods

3.2.1. Landsat LST Product

3.2.2. Sentinel-3 LST Product

3.2.3. Sentinel-2 Multispectral Data

3.2.4. Urban Building Data

3.3. Calculation of the Downscaling Driving Factor

Var. Description Equations

Normalize the difference building index to highlight building ρ12−ρ8

3.3.2. Calculation of Urban Spatial Morphological Parameters

Table 2. Urban spatial morphological parameters required for LST downscaling.

Var. Description Equations

3.4. Downscaling LST Method Based on Random Forest

3.5. Step-By-Step Random Forest Downscaling Method (SSRFD)

3.6. Accuracy Evaluation Methods

(a) (b) (c)

The southern bank of the Quanza, from its mouth to opposite

You might also like