The Egyptian Journal of Remote Sensing and Space Sciences 26 (2023) 861–880 Contents lists available at ScienceDirect The Egyptian Journal of Remote Sensing and Space Sciences journal homepage: www.sciencedirect.com Research Paper Trading greens for heated surfaces: Land surface temperature and perceived health risk in Greater Accra Metropolitan Area, Ghana Ronald Reagan Gyimah a,1, Clement kwang a,1,*, Raymond Agyepong Antwi b,1, Emmanuel Morgan Attua a,1, Alex Barimah Owusu a,1, Eric Kofi Doe a,1 a Department of Geography and Resource Development, University of Ghana, Legon, Accra, Ghana b Department of Urban Forestry and Natural Resources, Southern University and A&M College, USA A R T I C L E I N F O A B S T R A C T Keywords: The unsustainable expansion of cities is generating urban heat islands (UHIs) by exchanging (trading) vegetation Temperature cover (green) for built impervious surfaces which is associated with heat-related health risks, globally. This Urban built-up environment phenomenon is exacerbated by climate change and anthropogenic activities like urban population growth, Land cover change particularly in African cities. This study explores the spatio-temporal trends of land surface temperature (LST), Vegetation cover Urban heat island land use land cover (LULC) and their economic and health risks in the Greater Accra Metropolitan Area (GAMA) Urbanization of Ghana, from 1991 to 2021. We extracted LST/LULC information from Landsat datasets to perform change GAMA analysis, alongside an online survey across 56 communities on how LST relates to economic and human health risks perceptions of residents. The results show urbanization of GAMA is trading greens for heated surfaces, impacting communities’ health risks. While the built environment grew (8.6%), the vegetation cover declined (2.5%) and the mean LST rose (0.8⁰C) in 25 years. A 30⁰C LST corresponds to the point of inflexion of exchanging green vegetative cover for heated built surfaces. The forest community of Kisseman, the populous community of Dansoman and the harbour city of Tema corresponded to the first, fourth and fifth LST quintiles, changing at − 0.05⁰C, 0.06⁰C and 0.164⁰C per year. The common health risks include discomfort from heavy sweating, headaches, dehydration, thirst and skin rashes. These results call for climate action and green spatial planning through urban forestry and environmentalism in GAMA. For urban resilience and sustainable cities, we advocate green-cooling multi-purpose housing, roads, and industrial infrastructure. 1. Introduction heatstroke and other heat-related morbidities and mortalities (Avashia et al., 2021; Halder et al., 2021). Urban heat island (UHI) occurs when With rapidly expanding urban areas, land use land cover change urbanized areas experience warmer temperatures than nearby rural (LULCC) effects on urban surface temperatures and human health con- areas (Fu and Weng, 2016; Meyers et al., 2020). The phenomenon is akin sequences are becoming increasingly significant issues (Jenerette et al., to trading greens (vegetative spaces) for heated surfaces when building 2016, Chen et al., 2022). The global rise in average land surface tem- impervious surfaces. UHI often happens when natural vegetation peratures (LST) is expected to strike urban communities more harshly (green) cover is replaced by dense concentrations of man-made heat- due to population growth and Urban Heat Islands (UHIs) in cities (Fu absorbing materials such as concrete buildings, tared and asphalt roads, and Weng, 2016; IPCC, 2021; Park et al., 2021). Climate Change (CC) in iron roofing sheets, block/concrete pavements and other impervious rapidly urbanizing cities triggers UHIs, economic and health effects surface build-ups in city development process (urbanization) (Sun et al., (Halder et al., 2021; Imhoff et al., 2010). The immediate health risks of 2021; Zhou et al., 2018). These impervious materials absorb and emit environmental hazards like UHIs are common in poor homes and heat energy that increases LST, contributing to a general rise in atmo- neighbourhoods of heavily populated cities (Songsore and McGranahan, spheric air temperature (airT) and global warming (Kim and Brown, 2012; Das and Das 2020). Globally, the common health impacts of UHIs 2021; Meyers et al., 2020). In some instances, the removal of green include heat exhaustion, general discomfort, respiratory difficulties, spaces for heated surfaces exacerbates extreme heat events (EHEs) of * Corresponding author. E-mail address: ckwang@ug.edu.gh (C. kwang). 1 Authors contributed equally to the research and manuscript. https://doi.org/10.1016/j.ejrs.2023.09.004 Received 11 August 2022; Received in revised form 18 September 2023; Accepted 24 September 2023 Available online 7 October 2023 1110-9823/© 2023 National Authority of Remote Sensing & Space Science. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). R.R. Gyimah et al. T h e E g y p t i a n J o u r n a l o f R e m o t e S e n s i n g a n d S p a c e S c i e n c e s 26 (2023) 861–880 greater than 40.6⁰C (Meyers et al., 2020; Nangombe et al., 2019; Chen savannah ecological and dry equatorial climatic zones (Fig. 1). It is et al., 2022). located along the southern coast of the Gulf of Guinea, covering a total LST is defined as the radiative energy released by various types of land area of 1,524.99 km2. GAMA is a low-lying area (below sea level) impervious materials that absorb and release heat within 1.2 m above with an average annual rainfall of about 730 mm. The annual average ground (Fu and Weng, 2016; Meyers et al., 2020). LST and airT were temperature of the area is 25 ◦C with monthly mean temperatures terms used interchangeably in the IPCC 2021 report. According to IPCC ranging from 22 ◦C to 33 ◦C (Wemegah et al., 2020). Coastal lands (2021), the periods of 2001–2011 and 2011–2020 globally, experienced (wetlands and dunes), shrublands and grassland constitute the major warmer airT by 0.99⁰C (0.84–1.10 ◦C) and 1.09⁰C (0.95–1.20 ◦C) vegetation zones of GAMA. respectively. The increase in airT is expected to reach 1.5-2⁰C or more by The GAMA spans 30 Metropolitan, Municipal and Districts (MMDA) 2050 (IPCC, 2021). Easterling et al. (1997) projected Africa to experi- with a total population of 5,454,959. According to the Ghana Statistical ence a 3-5⁰C rise in mean airT measured at 1.5 m above the surface of the Service, (2021), males constitute 48.97% whereas females form 51.03% ground, by the end of the 21st century. Ringard et al. (2016) reported of the population in 2020. GAMA is a relatively high-income city, similar trends in West Africa. These increased temperatures are higher dubbed the “economic hub” of the Greater Accra Region and the entire than those observed in the pre-industrial era of 1850–1900. The rise of country (Songsore et al., 2006). GAMA was chosen for this study due to LST and airT is attributed to the great increase in anthropogenic activ- the urban milieu and ecological dynamics it presents. For instance, there ities (Anthropocene) which is threatening planetary boundaries through is high demand for land for housing, road expansion and business climate change (IPCC, 2021; Rockström et al., 2009; Quan et al., 2022). development to create employment centres and tourism among other The urban population of Ghana grew by 56% within the period be- economic and social infrastructure projects (Oduro et al., 2015). tween 1950 and 2018, mainly in the city of Accra and Kumasi (DESA- UN, 2018; Ampim et al., 2015). Human population growth is impacting 2.2. Study design LULCC in the cities. The human population growth occasions rapid expansion of urban built-up areas characterized by urban sprawl, eco- A hybrid approach of spatio-temporal analytical techniques nomic and health risks (Ampim et al., 2015; Krehbiel et al., 2016; involving remote sensing and geographic information systems and a Saviour Mantey et al., 2014). However, literature on the relationship semi-structured questionnaire survey of 56 communities involving 103 between LULCC, UHIs measured in terms of LST and their associated participants was used. The methodological approach is summarized in economic and health risks implications for the city dwellers in Ghana is Fig. 2. rare. Despite studies on extreme weather events particularly in low- income communities (Wilby et al., 2021; Kayaga et al., 2021), few 2.3. Study data- satellite datasets acquisition and pre-processing have drawn the linkage to health risks (Jenerette et al., 2016; Codjoe et al., 2020). For instance, Jenerette et al., (2016) observed the LST, As shown in Fig. 2, remotely sensed Landsat datasets were down- LULCC, LST and health risk nexus in Pheonix (Arizona, USA) that pro- loaded from the US Geological Survey (USGS) website vided evidence for critical look at heat vulnerability and city planning in (http://earthexplorer.usgs.gov) and pre-processed to determine the that state. However, this type of evidence is patchy in GAMA and thus NDVI and their respective LST. The metadata of the Landsat data series is hinders green spatial planning for health benefits it the city. This study, provided in Table 1. therefore, seeks to explore the linkages of LULCC, LST and its perceived The pre-processing of the Landsat images was done to suppress health risk on residents in the Greater Accra Metropolicatan Area distortion or enhance important features for further processing and (GAMA). According to Gonzalez-Trevizo et al. (2021) and Chapman analysis (Vision, 1993). The image pre-processing includes radiometric et al. (2017), there is overly concentrated literature on LULCC/UHIs in and atmospheric corrections, subsetting and stacking selected bands to most advanced parts of the world while such knowledge in developing form composite images. In this study, the digital numbers (DNs) of the countries remains less known. This dearth of information exists in the features were converted to Top of Atmosphere (TOA) radiance using Greater Accra Metropolitan Area (GAMA) of Ghana where city dwellers equation Eq.1 (Adeyeri et al., 2017). are becoming potentially vulnerable to health risks due to increasing LULCC and UHIs linkages (Acheampong, 2019; Ampim et al., 2015). The Lλ = MLQcal +AL (1) situation of GAMA becomes scarier considering the inability of the city where: planners and health systems to develop contingency plans to address L is the TOA measured in Watts/ m2 λ *srad *μm. ML and AL are band current and future UHIs (Johnson et al., 2012). Assessing the trends of multiplicative and additive rescaling factors, respectively indicated in LST and to suggest ways to reduce its severe human health risks are the metadata file of the Landsat data, and Qcal is quantized and cali- concerns for city dwellers, researchers and world policymakers. brated digital number (DN). Unearthing the trends of LST/LULCC as scientific evidence is useful for After converting DNs to TOA, image calibration was performed with advocating the need for building climate-resilient and liveable cities dark object subtraction (DOS) equation to get the images in surface (UN-HABITAT, 2020, Ampim et al., 2015; Puplampu and Boafo, 2021). reflectance using equation Eq.2. It is relevant for regional and urban spatial planners and for providing heat-related health risks information, adaptation and coping strategies Rc = Rs +Rsi (2) against high LST and UHIs. Since public health risks of UHI epitomized where: by remotely sensed LST data have been marginal in Ghana’s research Rc is an atmospherically rectified image, Rs is the TOA and Rsi is space, the current study seeks to explore the phenomenon of trading given by the formula Mean Rw–(2*Standard Deviation Rw), where Rw is green spaces for heated surfaces (UHI) in GAMA, from 1991 to 2021. defined as the spectral value being considered as an offset. The DOS Specifically, the study seeks to determine the nexus between urban LST, method has the advantage of producing an atmospherically corrected built-up expansion and vegetation loss changes and their health impli- image (Kane et al., 2016). cations for urban community dwellers in GAMA. 2.4. LULC categorizations and change detection analysis 2. Methods LULC analysis is an essential tool for understanding and managing 2.1. Study area changes within the urban spaces (Xu et al., 2022). Image classification and LULC categorizations were performed after the pre-processing. The The study was conducted in GAMA, which lies within the coastal image classification automatically grouped the pixels into their 862 R.R. Gyimah et al. T h e E g y p t i a n J o u r n a l o f R e m o t e S e n s i n g a n d S p a c e S c i e n c e s 26 (2023) 861–880 Fig. 1. Study Area Map showing MMDA boundaries and sampled communities. Fig. 2. Methodological Workflow. respective land cover types with the main objective of creating thematic the main reasons it was used in this study. Authors that used SVM in LULC maps (Saha et al., 2021). The support vector machine (SVM) is a LULC studies include Liu et al., (2022); Nyamekye et al., (2021); Tian non-parametric classifier that produces superior results relative to the et al., (2020); Shao and Lunetta, (2012). The GAMA LULC maps were maximum likelihood classifier, neural network classifier and decision classified into built-up, waterbody and vegetation classes using the SVM tree classifier (Kavzoglu and Colkesen, 2009; Huang et al., 2002). The algorithm in QGIS software. The descriptions of the LULC classes are SVM algorithm provides higher stability and accuracy and these were explained (Table 2). 863 R.R. Gyimah et al. T h e E g y p t i a n J o u r n a l o f R e m o t e S e n s i n g a n d S p a c e S c i e n c e s 26 (2023) 861–880 Table 1 Metadata of Landsat data series. Path/ Sensor Scene ID Acquisition date Time (GMT) Season Row 193/056 Landsat 4 TM LT41930561991010XXX03 1991-01-10 09:40:09 Dry 193/056 Landsat 7 ETM+ LE71930562002024EDC00 2002/01/24 10:04:20 Dry 193/056 Landsat 7 ETM+ LE71930562011017ASN00 2011-01-17 10:09:00 Dry 193/056 Landsat 8 OLI &TIRS LC81930562021004LGN00 2021/01/04 09:33:24 Dry Table 2 LULC class categorization and description. LULC type Description Vegetation Grassland, dense forest and urban vegetation such as forest reserves, parks, lawns and vegetated recreational areas Built-up Urbanized and industrial areas, open lands such as barren land and wasteland Waterbody Wetlands, rivers, lakes, canals, lagoons, riparian and ponds The spatio-temporal changes in the LULC types between two 2.6. Derivation of NDBI and NDVI from Landsat images different years were analyzed using the algebra-based change detection approach. The algebra-based approach was chosen because it was easy The Normalized Difference Built-up Index (NDBI) and Normalized to implement and detect different scales of changes in the LULC types Difference Vegetation Index (NDVI) were used to establish the nexus of (Afaq and Manocha, 2021). Post-classification image differencing LST, built-up expansion and vegetation cover loss, following Jenerette method was used to determine the changing pattern and acreage for et al., (2016) and Guha et al., (2021). NDBI denotes built-up areas which each LULC type from 1991 to 2021. The post-classification image dif- correlate positively with LST. NDVI reflects vegetation greenness and is ferencing method assesses the spatial variation in the two images by inversely related to LST. In this study, the NDBI and NDVI of 1991, 2002, analyzing the intensity of each pixel colour (Minu and Shetty, 2015). 2011 and 2021 were estimated using atmospherically rectified Landsat The change in acreage for each LULC type was further analyzed using series of 30 m spatial resolution. We calculated the NDBI and NDVI Microsoft Excel to explore the trends and patterns of the spatio-temporal following Jones and Vaughan, (2010), Huang et al., (2021) and Zheng changes. The following equations were applied in determining the et al., (2021) using Eq. (6) and Eq. (7) respectively. change in LULC types in percentage and rate of change in LULC. ShortwaveNearinfrared − Nearinfraredband(SWNIR − NIR) LULC NDBI = (6) ChangeinLULC ha Currentyear − LULCPastyear (3) ShortwaveNearinfrared + Nearinfraredband(SWNIR + NIR)( ) = LULCPastyear (Near − infraredband − Redband) LULC − LULC NDVI = (7) %ChangeinLULC Currentyear Pastyear(ha) = x100% (4) (Near − infraredband + Redband) LULCPastyear Both NDVI and NDBI values range from − 1 to + 1. Positive NDVI [( ) ] LULC − LULC RateofChangeinLULCperyear Currentyear Pastyear= x100% Ã⋅30year (5) LULCPastyear values indicate vegetated areas, while negative values are non-vegetated areas. NDVI value close to 0.00–0.25 indicates a bare surface, 1.00–0.99 2.5. Accuracy assessment for water and 0.30–1.00 for vegetation (Jones and Vaughan, 2010). High positive NDVI values show dense vegetation cover such as forest An accuracy assessment was performed on each of the LULC category (Huang et al., 2021). Positive NDBI values depict built urban land areas, maps to determine their level of accuracy in predicting each LULC type. and negative values represent non-built urban land areas (Zheng et al., Five hundred and thirty (530) accuracy assessment points were gener- 2021). ated for each LULC category map by using the stratified random sam- pling technique (Ranagalage et al., 2019) in ArcGIS software. Google 2.7. Retrieval of land surface temperature for 1991, 2000, 2010 and Earth historical images were then used as reference data for determining 2020 the truth points for the 530 accuracy assessment points of 1991, 2002, 2011 and 2021 LULC maps. The confusion matrix was created by using Retrieving the LST data from NDVI followed a mono-window algo- the accuracy assessment points for each of the LULC maps in order to rithm that estimates LST from Landsat thermal bands (Kumari et al., determine the accuracies (user’s accuracy, producer’s accuracy and 2021, Kafy et al., 2021). This involves four steps; (1) DN-spectral radi- overall accuracy) and Kappa coefficient. The user’s accuracy accounts ance conversion, (2) converting spectral radiance to brightness tem- for commission errors while the producer’s accuracy deals with omis- perature (Zhao et al., 2021), (3) application of atmospheric correction sion errors (Ranagalage et al., 2019). The kappa coefficient is used in on the thermal bands using dark object subtraction (4) a proportion of remote sensing image classifications to express the accuracy of LULC vegetation to correct emissivity and (5) estimating LST (Yao et al., 2021) categorized maps (Foody, 2020; Xu et al., 2022). as illustrated in the steps below. In step 1, the DN values of thermal bands were converted to spectral radiance using eq. (6). 864 R.R. Gyimah et al. T h e E g y p t i a n J o u r n a l o f R e m o t e S e n s i n g a n d S p a c e S c i e n c e s 26 (2023) 861–880 DN L L L L (8) 2.10. Assessing the relationship between NDVI and LST λ = min +( max + min) × 255 where, L is the spectral radiance; For Landsat 5; L 1.238; L NDVI has a direct relationship with LST even though the correlation λ min= max = 15.30; for Landsat 7; Lmin = 0.0; L 17.04; For Landsat 8; L would be negative. NDVI-LST were established using statistical software max= min= 0.10033; Lmax = 22.00180 (Kafy et al., 2021). such as Excel spreadsheets. The correlation results were displayed After converting the radiometric correction, the spectral radiance graphically as a scatter plot chart. was changed to the atmosphere brightness temperature (T ) in Kelvin Some 56 communities in the study were characterized according to B with eq. (9). their LST quintiles and income classes. Each quintile (q) corresponds to one-fifth (≃11/56) of the 56 communities as well as one-fifth (20th %) K T 2= (9) of LST estimated. This characterization was done by creating a 1 km B ln((K1/Lλ)+1) buffer from the observed community centre and extracting the LST where, K is Calibration constant 1; K Calibration constant 2; For values within the buffer to estimate the minimum, maximum and mean 1 2 Landsat 4, K1 = 671.62; K2=1284.30; For Landsat 7, K1 = 666.09; LST at the community level for 1991, 2001, 2011 and 2021. This K2=1260.56; For Landsat 8, K1 = 480.88; K =1201.14 (Kafy et a l., enabled a trend analysis of the community LST from 1991 to 2021. 2 2021; Zhao et al., 2021). Determination of the community income class was done using the Accra We converted LST measured in Kelvin to degree Celsius (C) using Eq. Metropolitan Assembly (AMA) community income rankings as low in- (10). come, middle income and high income. AMA does this classification for tax collection purposes and the main variables used in this classification TB(C) = TB(K) − 273.15 (10) are building quality, building structuring and socio-economic infra- This enabled estimation of the LST from the T in degree Celsius (C) structure available in the communities. An analysis of variance B using the following equation Eq10. (ANOVA) and Kruskal-Wallis test of difference in community LST be- tween the years were also conducted to determine any statistical dif- LST = TB/(1+(λ × TB/ρ) × ln(E ) ) (11) ferences in the LST estimates over the years and across income classes. Where, λ is wavelength of emitted radiance = 11.5 µm, ρ = 1.438 × 102mK (Kafy et al., 2021); 2.11. Spatial correlation of community NDVI, income status and LST in GAMA Surface emissivity, E = 0.004PV+ 0.986 (12) (Kafy et al., 2021). A spatial correlation map of the community NDVI, income status and LST quintile was created using inverse difference weighting (IDW) and a The proportion of Vegetation, PV = (NDVI − NDVImin/NDVImax − NDVImin) raster calculator in ArcMap. It was done by coding the community NDVI (13) as 1 = vegetation and 0 = no vegetation. The community income status was coded as 100 = low income, 200 = middle income, and 300 = high- Where NDVImax and NDVImin are highest and lowest NDVI values income communities. The codes for the community LST quintiles were respectively (Kafy et al., 2021). 1000 = very low LSTq1, 2000 = low LSTq2, 3000 = moderate LSTq3, The approach used in estimating the LST in this study is one of the 4000 = high LSTq4 and 5000 = very high LSTq5. These codes were most widely used methods in determining LST using remote sensing interpolated using IDW and then reclassified to their corresponding data. Authors such as Athukorala and Murayama, 2020; Dissanayake codes. Each of these IDW interpolations was resampled to the same et al., 2019; Ranagalage et al., 2019 and Simwanda et al., 2019 have resolution of 30 m × 30 m and then layer stacked by adding them using successfully applied the approach in estimating LST. the raster calculator tool. Our priority expectation was that high-income communities are likely to have more impervious surfaces replacing the 2.8. Time series analysis of land surface temperature over 25 years vegetative cover. The low-income communities are likely to be more vegetative and have fewer impervious surfaces. The aforementioned LST approach requires downloading several large Landsat data over 25 years for a time trend analysis (Ermida et al., 3. Results 2020). Since LST fluctuates frequently over short periods, it necessitates the use of several Landsat datasets to estimate the long-term mean, 3.1. Relationship of land surface temperature change, built-up expansion minimum and maximum LST. Therefore, we employed Ermida et al, and vegetation cover loss (2020)’s automated Google Earth Engine (GEE) java scripts, to simplify the extraction of the LST information from 341 Landsat datasets over 16 3.1.1. Spatio-temporal trends of annual and seasonal land surface to 25 years. We then applied ordinary least square (OLS) regression to temperature in GAMA estimate the trend and the absolute (ab) change (±) in the LST over the The survey reveals that most of the respondents (93%) believe that years (n) observed. See supplementary file 1.1. The 341 Landsat dataset there is a surge in GAMA LST, attributable mainly to climate change. The enabled us to cross-validate our initial LST estimates for 1991, 2000, physical evidence for the temperature rise is presented in Fig. 3. Fig. 3 2010 and 2020 datasets. Pelta and Chudnovsky, (2017)’s augmented illustrates the spatio-temporal distribution of the observed LST as of airT predicted a higher accuracy of temperature brightness (TB) using 1991 (Fig. 3a), 2002 (Fig. 3b), 2011 (Fig. 3c) and 2021 (Fig. 3d) based meteorological data. We employed ERA-5 atmospheric data (supple- on the four dry seasons Landsat datasets. The highest maximum LST of mentary file 1.1) to cross-validate our LST estimates. 49.0⁰C (mean = 33⁰C) occurred in 2021 and the lowest maximum ever experienced, 30.0⁰C (mean = 27⁰C), took place in 1991. The mean LST 2.9. Assessing the health risks of land surface temperature increased by 6⁰C (22.2%) from 27⁰C to 33⁰C during the period. Table 3 shows the annual mean, minimum and maximum LST sta- We designed a semi-structured questionnaire using Google Forms tistics were 31.9⁰C, 27.3⁰C and 33.9⁰C, respectively, based on the 341 and administered it online through social media platforms to conduct a Landsat time series dataset. The dry (Dec-Mar) season mean LST was cross-sectional assessment of socioeconomic and health risks of urban 32.6 ± 1.4⁰C. While the wet season 1 (Apr-Jul) LST was 31.4 ± 3.3⁰C, it LST in GAMA. The questions elicited perceived risks from 103 re- was 32.1 ± 2.2⁰C in the wet season 2 (Aug-Nov). spondents living in 56 GAMA communities. The data were analysed The time trends of the annual and seasonal LST are shown in Fig. 4. using STATA. The annual mean LST increased at 0.03⁰C per year (Fig. 4a), culminating 865 R.R. Gyimah et al. T h e E g y p t i a n J o u r n a l o f R e m o t e S e n s i n g a n d S p a c e S c i e n c e s 26 (2023) 861–880 Fig. 3. Patterns of spatio-temporal land surface temperature as of January (a) 1991, (b) 2002, (c) 2011 and (d) 2021 in GAMA. Table 3 Descriptive statistics of annual and seasonal LST in GAMA. Variable Period Number of Years Mean (⁰C) Std. Dev. (⁰C) Min (⁰C) Max (⁰C) GAMA LST (annual) 1991, 1998–2021 25 31.9 1.5 27.3 33.9 Dec-Mar (Dry season) 1991, 1999–2021 24 32.6 1.4 29.6 35.1 Apr-Jul (Wet season 1) 2000–2020* 16 31.4 3.3 26.0 37.4 Aug Nov (Wet season 2) 1998-2020† 22 32.1 2.2 27.3 35.2 Missing data in 2001, 2008–2011 (*) and 2010 (†) in a 0.8⁰C rise over 25 years. While the trend of the dry season LST was January 2021. The participants (N = 103) of the survey lived in the 56 constant (Fig. 4b), it took downward and upward paths in both wet communities within GAMA as indicated by the circles and boundaries in seasons 1 (Fig. 4c) and 2 (Fig. 4d). Fig. 7. The circles represent a 1 km radius from the community centroid (GPS point). Fig. 7a depicts the general pattern of variability in the LST. 3.1.2. GAMA LST, NDVI, NBVI and LULC change relationship In general, the mean LST (Fig. 7b) was 26.95⁰C (ranging from 24.53⁰C Fig. 5 shows the built environment in GAMA grew by 258.7% (849.6 mean minimum to 29.48⁰C mean maximum) at the community level. km2) at an annual rate of 8.6% in the 30 years while the vegetation cover Fig. 7c and Fig. 7d demonstrate the spatial distribution of the minimum experienced a 74.6% (846.3 km2) decline at 2.5% per annum. The water range (19.5⁰C-27.4⁰C) and maximum range (27.1⁰C-47.8⁰C) LST for all bodies also decreased by 3.3 km2. The built-up area increased drastically communities respectively. to overtake the vegetation cover after 2002 and continued to 2021 at an Tema was among the fifth (highest) community LSTq5 category, annual rate of 8.6%. Table 4 indicates that the accuracy of the LULC followed by Dansoman in the fourth (LST4q) category and Ridge in the classifications was high. The overall classification accuracy was 97.4% third (LSTq3) category. Legon was in the second LSTq2 while Kisseman (kappa = 0.93) for 1991 (Fig. 5a), 94.3% (kappa = 0.85) for 2002 was among the first (lowest) LSTq category. (Fig. 5b), 94.2% (kappa = 0.88) for 2011 (Fig. 5c) and 95.8% (kappa = 0.88) for 2021 (Fig. 5d) LULC classes. 3.1.4. Temporal trends and changes in LST within study communities The dwindling vegetation cover and increasing built-up (heated Descriptive statistics of the temporal trends and changes in the impervious) surfaces intermediate the surge in LST change with GAMA. community land surface temperature quintile (LSTq) are presented in As shown in Fig. 6, the LST, NDVI and NBVI relationship suggests that at Table 5. Tema was among the highest LSTq5 (≤100th %) of 35.0 ± a mean LST greater than 30⁰C, the built-up area exceeds the vegetation 2.4⁰C, while Kisseman was in the lowest LSTq1 (≤20th %) at 31.0 ± cover and vice versa. 2.3⁰C. Furthermore, 8a shows there was a 0.164⁰C rise in the mean LST per 3.1.3. Spatial pattern of community land surface temperature in quintiles annum in Tema. This implies a 4.1⁰C temperature increase over the 25 Fig. 7 characterizes the community-level LST in quintile (q) as of years. On the contrary, Fig. 8e revealed a 0.05⁰C decrease in the mean 866 R.R. Gyimah et al. T h e E g y p t i a n J o u r n a l o f R e m o t e S e n s i n g a n d S p a c e S c i e n c e s 26 (2023) 861–880 Kisseman community LST per annum, which implies a 1.2⁰C increase and were not statistically different (Kruskal-Wallis χ2 = 5.453, p-value over the 24 years. = 0.142, df = 3). Concerning the minimum community LST in 1991 (μ = The bar graphs in Fig. 9 exhibit variations of community-level LSTq 25.5 ± 0.63, median = 25.2), 2001 (μ = 23.5 ± 1.8, median = 23.8), in GAMA from 1991 to 2021 based on the four dry seasons Landsat 2011 (μ = 24.2 ± 1.6, median = 24.4) and 2021 (μ = 23.5 ± 0.63, dataset. There were upward trends in the Very high LSTq5 (Fig. 9a), median = 23.8), the test showed at least one of the rank sums was sta- High LSTq4 (Fig. 9b) and Moderate LSTq3 (Fig. 9c) quintile commu- tistically different from the others (Kruskal-Wallis χ2 = 48.419, p-value nities. The trends were downward in the Low LSTq2 (Fig. 9d) and Very = 0.000, df = 3). Similarly, a significant statistical difference was low LSTq1 (Fig. 9e) quintile communities. observed for the maximum community LST (Kruskal-Wallis χ2 = The ANOVA and Kruskal-Wallis tests of the significance of the dif- 81.845, p-value = 0.000, df = 3) for 1991 (μ = 28.1 ± 0.5, median = ference in the minimum, maximum and mean LST (⁰C) observed in 1991, 28.1), 2001 (μ = 30.0 ± 2.7, median = 29.7), 2011 (μ = 29.4 ± 1.2, 2001, 2011 and 2021. The mean community LST in 1991 (μ = 26.9 ± median = 29.0) and 2021 (μ = 30.0 ± 2.7, median = 29.7). 0.6, median = 26.9), 2001 (μ = 26.9 ± 1.3, median = 27.3), 2011 (μ = 27.2 ± 1.0, median = 27.5) and 2021 (μ = 26.9 ± 1.3, median = 27.3), Fig. 4. Trends of annual (a) and dry season (b), wet season 1 (c) and wet season 2 (d) LST in GAMA (n = number of years; ab = absolute change ± ). 867 R.R. Gyimah et al. T h e E g y p t i a n J o u r n a l o f R e m o t e S e n s i n g a n d S p a c e S c i e n c e s 26 (2023) 861–880 Fig. 4. (continued). 3.2. Economic implications for exchanging greens for heated surfaces 3.3. Perceived health implications of exchanging greens for heated surfaces The LST was categorized according to the income status of each community (Fig. 10). The one-way ANOVA test revealed no statistically 3.3.1. Factors contributing to community heat exposure significant differences (see appendix 2.2 supplementary file 2.0 The perceived factors contributing to the LST heat exposure risk in attached) in the minimum, maximum and mean LST across income the communities are mainly the presence of impervious surfaces, the status of the observed communities. This implies that the mean LST for probability of being exposed to extreme heat exposure and the hours of the High (μ = 27.1 ± 0.8), Middle (μ = 26.7 ± 1.2) and Low (μ = 27.1 ± extreme heat exposure during the day and night. 1.0) income communities, is random (ANOVA Fcal = 3.13, p-value = shows that a large proportion (73.0%, n = 65) of the respondents 0.045, df = 2). The same applies to the minimum and maximum com- either agree (38.2% n = 34) or “strongly agree” (34.8%, n = 31) that munity LST. This means that there is no significant difference in the impervious surfaces contribute to rising LST. About 18.0% (n = 16) spatial pattern of LST between high, middle and low-income were, however, indifferent while 3.4% (n = 3) and 5.6% (n = 5) were neighbourhoods. either “highly disagree” or “disagree” respectively. The results in 868 R.R. Gyimah et al. T h e E g y p t i a n J o u r n a l o f R e m o t e S e n s i n g a n d S p a c e S c i e n c e s 26 (2023) 861–880 Fig. 5. Spatiotemporal pattern of land use land cover change in GAMA as of (a) 1991, (b) 2002, (c) 2011 and (d) 2021. (21.4%, n = 22), lab diagnosis (7.8%, n = 8) and other sources such as Table 4 friends and news. As shown in Fig. 13b, there was no statistically sig- Accuracy assessments of GAMA LULC classifications. nificant association of these health risks with the observed LST quintiles Accuracy LULC Type 1991 2002 2011 2021 (Pearson chi-square = 49.2767 Pr = 0.837). User’s accuracy (%) Vegetation 99.8 96.0 97.8 95.0 Built-up 91.4 89.7 92.2 96.2 3.3.4. Common adaptation strategies to health risks of rising LST Water Body 78.6 85.7 91.7 92.3 Use of air conditioning devices (32.0%, n = 33), electric fans (31.1%, Producer’s accuracy (%) Vegetation 97.1 96.5 88.6 85.0 Built-up 99.1 87.4 98.7 99.0 n = 32), opening of doors and windows and folding of curtains for more Water Body 91.7 92.3 73.3 92.3 ventilation (25.2%, n = 26) are the common domestic adaptation stra- Overall accuracy (%) 97.4 94.3 94.2 95.8 tegies at home. Other adaptation strategies to rising LST (11.7%, n = 12) Kappa Coefficient 0.93 0.85 0.88 0.88 intimated by the respondents include taking a freshwater bath, using shade/shade trees, partially undressed, wearing light clothes, drinking Fig. 11b indicate that increase in the number of concrete buildings more water, laying on a cold cement floor and staying indoors to (86.5%, n = 77), concrete floors (66.3%, n = 59), tiled floors (52.8%, n minimize the health risk of high LST. = 47), tarred roads (51.7%, n = 46) and concrete pavements (49.4%, n = 44) are the common impervious surfaces in the observed 4. Discussion communities. 4.1. The extent of rising land surface temperature 3.3.2. The extent of impervious surfacing in GAMA communities As shown in Fig. 12, a large proportion (66.9%, n = 69) of the par- The results indicate that people living in the GAMA believe there are ticipants indicated they were either “exposed” (41.7%, n = 43) or changes in the climate with rising LST in their communities, which “highly exposed” (25.8%, n = 26) to extreme heat conditions where they impacts health and community economic status. The impact of Climate spent most of the time. Only 28.2% (n = 29) were indifferent while 4.9% Change (CC) was confirmed by the LST analysis, which revealed that the (5) said they were “highly unexposed”. highest maximum LST (49.0 ⁰C) in 2021 and the lowest maximum (30.0 ⁰C) in 1991 occurred at an increasing rate of 0.63⁰C per year. The mean 3.3.3. Residents perceived health risk of exposure to high LST LST rise (6.0⁰C) from 27.0⁰C to 33.0⁰C over the 30 years at a rate of Residents of the study communities recount their perceived exposed 0.20⁰C per annum was based on four Landsat datasets. These findings health risks due to high LST in their community. As shown in Fig. 13a, are consistent with some previous global studies (Easterling et al., 1997; heavy sweating that causes discomfort (69.9%, n = 72) was expressed as Nangombe et al., 2019; Roy et al., 2020; Saha et al., 2021). Roy et al. the dominant perceived health risk of extreme heat, followed by head- (2020) reported an increase of 5.6⁰C in the mean LST (20.2⁰C to 25.8⁰C) ache (49.5%, n = 51), dehydration (48.5%, n = 50) and thirst (48.5%, n between 1990 and 2018 in the Chatogram Metropolitan Area in = 50). Informants indicated that their choices of answers were based on Bangladesh. Nangombe et al. (2019) indicated a similar rise of LST in personal experience and perception (56.3% n = 58), medical advice, South and West Africa while Easterling et al. (1997) anticipated a 3-5⁰C 869 R.R. Gyimah et al. T h e E g y p t i a n J o u r n a l o f R e m o t e S e n s i n g a n d S p a c e S c i e n c e s 26 (2023) 861–880 Fig. 6. The relationship between land surface temperature (LST), built-up (NDBI) and vegetation (NDVI) indexes of GAMA. Fig. 7. Spatial variability (a), maximum (b), minimum (c) and mean (d) land surface temperature of GAMA communities as of 2021. Table 5 Descriptive statistics of the long-term community-level annual LST quintile. Community LST quintile (LSTq) Period Years Mean (⁰C) Std. Dev. (⁰C) Min (⁰C) Max (⁰C) Tema q5 (≤100th %) 1991, 1998–2021 25 35.0 2.4 26.1 37.7 Dansoman q4 (≤80th %) 1991, 1998–2021 25 34.8 1.4 31.9 37.6 Ridge q3 (≤60th %) 1991, 1999–2021 24 33.6 1.7 30.3 36.8 Legon q2 (≤40th %) 1991, 1999–2021 24 32.8 2.3 25.4 35.5 Kisseman q1 (≤ 20th %) 1991, 1999–2021 24 31.0 2.3 23.3 34.8 Missing data in 1992–1998 rise in airT of tropical regions in Africa by the end of the 21st century. English Bazaar Urban Agglomeration. The increase in GAMA’s mean LST This pattern of rising LST is consistent with other authors (Saha et al., of 4.0⁰C from 2002 to 2011 is far higher than the projected global es- 2021), who estimated a 1.5 to 3.5⁰C increase over the last 30 years for timate of 0.99⁰C in the recent AR6 report by IPCC (2021). However, the 870 R.R. Gyimah et al. T h e E g y p t i a n J o u r n a l o f R e m o t e S e n s i n g a n d S p a c e S c i e n c e s 26 (2023) 861–880 observation for 2011–2021 in GAMA falls short by 2.0⁰C when 2020; Songsore and McGranahan, 2012). The expansion of the built compared to the IPCC’s estimate of 1.09⁰C for 2011–2020. The 6.0⁰C environment is also largely due to human population growth and was based on only four Landsat data sets and thus watered down when rural–urban migrants trying to settle in the city and its outskirts, in low- cross-validated using more data in using the GEE approach. The rise cost or underdeveloped green spaces. Songsore (2020) noted this as turned out to be 0.8⁰C at an annual rate of 0.03 ⁰C in 25 years. None- demographic urbanization rather than economic urbanization. Poverty- theless, this 0.8⁰C mean LST rise has important livability implications for driven rural–urban migration is often characterized in communities by GAMA. The IPCC’s alarm that many urban areas would experience more crowding of houses, poor environmental sanitation management and than a 1.5–2.0⁰C rise in temperature by 2050. This therefore should be of heat effects (Songsore et al., 2006). concern to the GAMA communities, especially in LSTq5 communities A major consequence of unregulated (unplanned) spatial expansion like Tema to plan for a climate-resilient city. This is because it would of the built environment is the loss of vegetation cover. This includes the only take massive and immediate effort to avert the projected 1.5–2.0⁰C loss of green spaces which are useful for controlling air pollution, increase in temperature (IPCC, 2021) for sustainable living. providing walkable, recreational and playable spaces for adults and children (Adjei-Boadi et al., 2022). A changing dynamic is demonstrated by the findings from LULCC influencing UHI. To make cities nature- 4.2. The extent of increasing built-up, decreasing vegetation and water friendly, livable and resilient, there must be considerable urban revi- body talization through urban green development, open space and landscape management strategies (Dissanayake et al., 2019). As proposed by Ojeh We found that the built environment in GAMA grew by 258.7% 2 et al. (2016) it is necessary to design climate-smart cities with nature- (849.6 km ) at an annual rate of 8.6% in the 30 years. On the contrary, 2 friendly urban planning systems and this may include city deconges-vegetation cover experienced a decline of 74.6% (846.3 km ) at 2.5% tion, open space management and decentralised economic activities that per annum. The water bodies also decreased by 3.3 km2. All these are separate from the city centre (Mantey et al., 2014; Puplampu and findings are consistent with Owusu (2018) and Addae and Oppelt, Boafo, 2021). Spatial development planning needs to be employed to (2019), who reported that the built environment in GAMA grew by control the growth of built-ups with green open spaces (Ningrum, 2018). 277% in 24 years while the forest cover decreased alongside the Also, to improve the thermal environment and mitigate UHI, there is the waterbody in 25 years (1991 and 2015). Several socioeconomic factors need for the use of lower absorptive material, higher reflective and have contributed to the rapid expansion of the built environment (Doe larger thermal conductivity of buildings (Ningrum, 2018), restoration et al., 2018; Owusu, 2018). Apart from human population growth, and protection of wetlands (Völker et al., 2013) and introduction of cool GAMA witnessed significant private capital investment in the housing roof and green roof (Malley et al., 2015; Zhang et al., 2017) as well as sector through real estate development from both local and foreign in- promoting urban forestry within GAMA. As iterated by Jenerette et al. vestors (Addae and Oppelt, 2019; Mantey et al., 2014). These in- (2016), the cooling of hotter neighbourhood is more effective with vestments were aimed at meeting the housing needs of expatriates and vegetation. Planting of shade trees supports the arguments of Chen et al. expected diaspora returnees living abroad. This real estate development (2022) and Li (2020), who indicated varied cooling approaches for is conspicuous in the Accra Metropolitan Area (AMA), Tema, Korley mitigating temperature rise and urban microclimate. These cooling ap- Klotey, Tseaddo among other MMDAs (Addae and Oppelt, 2019). proaches involve green, blue and grey infrastructures for cooling the In contrast to high-income countries like Germany, UK or USA, microclimate of different urban facilities across various urban land use expansion of the built environment in GAMA is rarely shaped by eco- types. nomic development and spatial planning. While increased productivity The green infrastructure is strategically a network of high-quality and industrialization are notable attributes of urbanization in the natural and semi-natural areas with other environmental features, Western world, the experience in Africa is often not the same (Songsore, Fig. 8. Time Trends of annual LST in Tema (a), Dansoman (b), Ridge (c), Legon (d) and Dansoman (e), (n = number of years; ab = absolute change ± ). 871 R.R. Gyimah et al. T h e E g y p t i a n J o u r n a l o f R e m o t e S e n s i n g a n d S p a c e S c i e n c e s 26 (2023) 861–880 Fig. 8. (continued). 872 R.R. Gyimah et al. T h e E g y p t i a n J o u r n a l o f R e m o t e S e n s i n g a n d S p a c e S c i e n c e s 26 (2023) 861–880 Fig. 8. (continued). designed and managed to provide a variety of ecosystem services. The surfaces within the built environment. Digital agriculture as suggested grey infrastructure is man-made features such as pipes, ditches, swales, by Jiang et al. (2022) is another way to improve urban microclimate. culverts, and retention ponds that provide the same services (Li et al., According to Jiang et al. (2022), use of information technologies like 2020). The blue infrastructure relates to urban water systems such as artificial intelligence, robotics, phones, internet and e-commerce in ponds, lakes, streams, rivers and stormwater management systems to digital agriculture can improve the efficiencies of agricultural activities provide ecosystem services. Li et al. (2020), proposed that while each of to reduce greenhouse gas emissions. the “colour” strategies is good individually, an integrated green, blue and grey infrastructure is also key. For instance, the combination of trees and pavilions has a much stronger cooling effect than either trees or 4.3. The point of inflexion in the inverse relationship of LST, vegetation pavilions alone (Xu et al., 2017). Li et al (2020) also suggested that and built-up treated wastewater could be suitable for parks, gardens and squares while green infrastructure should be applied to residential areas. The Another finding of the study is the 30.0⁰C LST point of inflexion integrated strategy includes high raising buildings to reduce impervious (tipping point) that pivots the inverse relationship between vegetation loss and rising built-up environment, spurring UHIs in GAMA. This 30⁰C 873 R.R. Gyimah et al. T h e E g y p t i a n J o u r n a l o f R e m o t e S e n s i n g a n d S p a c e S c i e n c e s 26 (2023) 861–880 is the tipping point beyond which the built-up area overtakes the Sun et. al, (2012) urban LST increases with an increasing density of vegetation area in size and vice-versa in GAMA. This inverse relation- built-up areas and decreases with vegetation cover growth. ship is consistent with many authors (Fu and Weng, 2016; Ning et al., 2018; Tuffour-Mills et al., 2020; Zhang et al., 2009). For instance, Tuf- four-Mills et al. (2020), reported a similar relationship in Ghana. This 4.4. Spatial differences in decreasing and increasing temperature at the finding implies that at a mean LST of 30⁰C there is a zero correlation community level between vegetation and built-up cover. Above 30⁰C, there is a negative correlation between vegetation and built-up LULCC. Many authors Characterizing the observed communities into five LST quintiles including Ning et al. (2018) noted this point of inflexion at different (Fig. 7&14), it was found that communities that harbour the central thresholds of LST and LULCC in their studies. Essentially, this finding business district and manufacturing industries belong to the highest provides a statistical guide for checking the imbalance (inequality) of (fifth) LSTq (28.5⁰C ± 0.4). The highest LSTq5 communities had more vegetation cover and built-up area cover in GAMA. buildings, tarred, paved or tiled surfaces and few vegetation areas than The drivers of vegetation loss and rising LST can be associated with other communities. This corroborates Fu and Weng (2016) and Meyers the expansionary trend of built-up, particularly impervious surfaces as et al. (2020) that built infrastructure generates higher heat-holding ca- depicted by the inverse correlation explained below and illustrated in pacity during the day and releases it gradually at night and this causes land surface quintiles (LSTq) of Fig. 7 and Fig. 14. This observed phe- the UHIs. The findings also agree with Das and Das (2022), who indi- nomenon confirmed the report of Ahmed (2018), Pelta and Chudnovsky, cated higher LST within built-up environments. (2017) and Puplampu and Boafo (2021). The UHI effect is experienced The fourth quintile (LSTq4) communities are known for their high- in most populated and high-density built-up environments and rarely in density human populations, transport activities and housing facilities. vegetative zones in Egypt, Tel-Aviv, Accra Metropolitan Areas (Ahmed, They also had some bare land (football parks) and little vegetation. 2018; Pelta and Chudnovsky, 2017; Puplampu and Boafo, 2021). This Although some hot surface temperatures were found in places that finding also buttresses the point of Sun et. al, (2012), Songsore and contained extensive bare land and sparse vegetation, Meyers et al. McGranahan (2012) and Xu et al. (2021), that an increasingly imper- (2020) explained that the sparse vegetation and bare land cover cool vious surface changes the properties of the ground surface and directly down more quickly at night. They are also relatively cooler during wet impacts the LST of urban ecology. According to Imhoff et. al, (2010) and periods than in the built environment (Meyers et al., 2020). The second LSTq2 and first LSTq1 communities are known for having fewer built Fig. 9. Trends of GAMA community-level land surface temperature (LST) from 1991 to 2021 in (a) Very high, (b) High, (c) Moderate, (d) Low and (e) Very low LST quintiles. 874 R.R. Gyimah et al. T h e E g y p t i a n J o u r n a l o f R e m o t e S e n s i n g a n d S p a c e S c i e n c e s 26 (2023) 861–880 Fig. 9. (continued). environments, bare lands and extensive vegetative cover encompassing communities had a lesser anthropogenic impact on the natural land the Legon West and the Achimota Forest Reserve near Kisseman. surface than the higher LSTq communities and vice versa. Source: Google Earth, 2022; Assessed on 20/01/2022; 13:14. Furthermore, the trends of communities LST soared in the very high, high and moderate LSTq communities but decreased in the low and very 4.5. Land surface temperature implications for economic status and low LSTq communities. The rising LST in the former may be attributed to health risks the building of more impervious surfaces, corroborating (Chen et al., 2022; Quan et al., 2022). On the contrary, the trend of decreasing LST in The relation between the community LST and income status of the low and very low LSTq may be due to low-density residential GAMA communities was found to be statistically insignificant. In China, development, frequency of rainfall due to their proximity to highlands the expansion of heated surfaces, radiated from high-income urban core and land litigation which has delayed urban development in those areas areas to replace vegetative cool areas outside the core business districts as portrayed Zhao et al., (2021). In other words, the lower LSTq of Hangzhou (Lin et al., 2018). Therefore, the insignificance of the community LST and income-status nexus in GAMA is incongruent with 875 R.R. Gyimah et al. T h e E g y p t i a n J o u r n a l o f R e m o t e S e n s i n g a n d S p a c e S c i e n c e s 26 (2023) 861–880 Fig. 10. Spatial correlation of community income status, vegetation and LST quintiles in GAMA as of 2021. Fig. 11. (a) Contribution of impervious surfaces to LST and (b) common impervious surfaces in the observed communities. many others (Lin et al., 2018; Sun et al., 2021). It contradicts Sun et al., GAMA was statistically significant. This finding corroborates Jenerette (2021) who reported that the status of socioeconomic activities drives et al. (2016) who reported that urban LST driven by land cover features vegetative cover loss and UHIs in Hangzhou. Hangzhou harbours high- impacts residential heat-related health risks, in Phoenix (Arizona) USA. technological industrial activities and impervious surfaces. The insig- In GAMA. The finding is also consistent with Das and Das (2016) who nificance of the current observation might be due to less industrial ac- reported that outdoor LST thermal conform within built-up areas was tivities in most of the observed high-income communities like Ridge in more uncomfortable than other land cover types. The perceived com- the current study. mon health risks of high LST include heavy sweating, headache, dehy- However, the relation between the community LST and health risk in dration, intense thirst and skin rashes. These findings affirm WHO 876 R.R. Gyimah et al. T h e E g y p t i a n J o u r n a l o f R e m o t e S e n s i n g a n d S p a c e S c i e n c e s 26 (2023) 861–880 Fig. 12. Perceived probability (Likert scale of 1 to 5) of exposure to heat conditions by participants. Fig. 13. (a) Perception of common health risk (a) associated with heat exposure and (b community-level land surface temperature quintile. (2013), Mantey et al. (2014) and Krehbiel et al. (2016). According to on a cold cement floor and leaving the home are also common. There- WHO (2013) and Jenerette et al., (2016), there are many of these heat fore, green, blue and grey spatial planning of city cooling systems re- related-health effects of rising LST to be avoided. Elsewhere, high LST mains crucial for sustainable city growth in the face of LULC and LST exacerbates city heat and air pollution by causing smog. Smog is a changes and health risks (Jenerette et al., 2016; Chen et al., 2022). hazardous pollutant that makes it harder to breathe leading to asthma attacks most frequently among children (WHO, 2013). Therefore, strengthening key health systems and improving the management of the 4.6. Data limitations health risks of UHI would be required in GAMA. Some of the health adaptation strategies of the urban community dwellers in GAMA Data used for this study included; (1) the Landsat satellite images corroborate Mantey et al. (2014) and Puplampu et al. (2021). The downloaded from the USGS website at a spatial resolution of 30 m and in observed high LST adaptation strategies encompass planting shade trees, the month of December which is the dry season and therefore measuring use of air conditioning (AC) and electric fans, opening doors and win- peak period temperature and not the average for the year. Again, using a dows and folding curtains for more ventilation. Some of these adapta- single scene image may not reflect the annual temperature, however, tion strategies like ACs may create environmental liabilities. Opening this was necessitated by the lack of cloud-free data. In making up for the doors may increase mosquitoes in rooms, whiles the use of air condi- shortfall, we used google earth engine archive data to provide time se- tioners increases household energy bills. According to Kayaga et al. ries on temperature, which to a large extent reflects annual temperature (2021), extreme heat events create high demand for water and elec- trends. It’s equally true that the google earth engine may not have tricity and this demand exceeds supply, leading to shortages and severe provided the entire scene at a time, it helps in generating the annual pressure on water and energy resources in Ghana. Other adaptation average temperature. In this study, using google earth engine provided strategies like taking a freshwater bath, using shade/shade trees, 341 Landsat images in estimating the annual for the study period of 25 partially undressed, wearing light clothes, drinking more water, laying years. (2) The primary data used in assessing local perception was based on 877 R.R. Gyimah et al. T h e E g y p t i a n J o u r n a l o f R e m o t e S e n s i n g a n d S p a c e S c i e n c e s 26 (2023) 861–880 Fig. 14. Rising land surface temperature quintiles. a semi-structured questionnaire we designed using google forms and risks implications. The common health risks of extreme heat/urban heat administered online through social media platforms to conduct a cross- islands include heavy sweating that causes discomfort, headaches, sectional assessment of socioeconomic and health risks of urban LST in dehydration and intense thirst. The differences in community LST and GAMA. The biggest limitation is the fact that the study has no control income status were not statistically significant. This implies that relying over the respondents’ selection. Such a method of data collection often on community-level or local microclimate estimates to inform LST pol- favors the educated population and their understanding of the questions. icy can misguide decisions and the urgency of taking climate action. The However, we find the data useful given that 103 respondents from 56 policy implication of this study points to a need for prioritizing efforts to communities actually responded to the survey. Perhaps the 103, may curb trade-offs of green spaces for heated surfaces and its health and constitute a small response, but the relevance is the point that we did not climate action implications. It calls for urban forestry and sustainable make inferences from the study, rather it is an opinion study and the climate-resilient city development of GAMA communities. These include respondents were fairly distributed across the entire study area. building multi-purpose green cooling housing, roads, health and in- We concluded that the study data may have some limitations, it dustrial infrastructure. useful and the findings were equally relevant as they concur with many other recent studies including Puplampu, and Boafo, 2021; and Owusu, Declaration of Competing Interest 2018. The authors declare that they have no known competing financial 5. Conclusions interests or personal relationships that could have appeared to influence the work reported in this paper. In conclusion, the current study has adduced new evidence that advances the frontiers of knowledge concerning the relationship of LST, Appendix A. Supplementary material NDVI, NBVI and their health risks implications to urban community dwellers in GAMA. The built environment grew by 258.7% (849.6 km2) Supplementary data to this article can be found online at https://doi. at an annual rate of 8.6% while the vegetation declined by 74.6% (846.3 org/10.1016/j.ejrs.2023.09.004. km2) at 2.5% per annum. The mean LST of GAMA rose by 0.8⁰C in 30 years at 0.03⁰C per annum. A mean LST of 30⁰C corresponds to the References tipping point in the inverse relationship between vegetation and the built environment. The trends of community LST increased in the very Acheampong, R.A., 2019. Spatial planning and the urban informal economy. Spatial high, high and moderate LST quintile communities, but the LST Planning Ghana. https://doi.org/10.1007/978-3-030-02011-8. Addae, B., Oppelt, N., 2019. Land-use/land-cover change analysis and urban growth decreased in the low and very low LST quintile communities. The in- modelling in the Greater Accra Metropolitan Area (GAMA). Ghana. Urban Sci. 3, 26. come status of communities did not reflect their LST quintiles. While the https://doi.org/10.3390/urbansci3010026. general LST of GAMA increased and study participants express Adeyeri, O.E., Akinsanola, A.A., Ishola, K.A., 2017. Investigating surface urban heat island characteristics over Abuja, Nigeria: Relationship between land surface discomfort with heated rooms and outdoor temperatures had health temperature and multiple vegetation indices. Remote Sens. Appl.: Soc. Environ. 7, 57–68. https://doi.org/10.1016/j.rsase.2017.06.005. 878 R.R. Gyimah et al. T h e E g y p t i a n J o u r n a l o f R e m o t e S e n s i n g a n d S p a c e S c i e n c e s 26 (2023) 861–880 Adjei-Boadi, D., Agyei-Mensah, S., Adamkiewicz, G., Rodriguez, J.I., Gemmell, E., Jiang, S., Zhou, J., Qiu, S., 2022. Digital agriculture and urbanization: mechanism and Ezzati, M., Baumgartner, J., Owusu, G., 2022. Neighbourhood, built environment empirical research. Technol. Forecasting Social Change 180 (March 2021), 121724. and children’s outdoor play spaces in urban Ghana: Review of policies and https://doi.org/10.1016/j.techfore.2022.121724. challenges. Landsc. Urban Plan. 218, 104288 https://doi.org/10.1016/j. Johnson, D.P., Stanforth, A., Lulla, V., Luber, G., 2012. Developing an applied extreme landurbplan.2021.104288. heat vulnerability index utilizing socioeconomic and environmental data. Appl. Afaq, Y., Manocha, A., 2021. Ecological Informatics Analysis on change detection Geogr. 35, 23–31. https://doi.org/10.1016/j.apgeog.2012.04.006. techniques for remote sensing applications : a review. Ecol. Inform. 63, 101310 Jones, H.G., Vaughan, R.A., 2010. Remote Sensing of Vegetation: Principles, Techniques, https://doi.org/10.1016/j.ecoinf.2021.101310. and Applications. Oxford University Press, New York, p. 353. Ahmed, S., 2018. Assessment of urban heat islands and impact of climate change on Kafy, A.-A., Faisal, A.-A., Al Rakib, A., Roy, S., Ferdousi, J., Raikwar, V., Kona, M.A., socioeconomic over Suez Governorate using remote sensing and GIS techniques. Fatin, S.M.A.A., 2021. Predicting changes in land use/land cover and seasonal land Egypt. J. Remote Sens. Sp. Sci. 21, 15–25. https://doi.org/10.1016/j. surface temperature using multi-temporal landsat images in the northwest region of ejrs.2017.08.001. Bangladesh. Heliyon 7 (7), e07623. Ampim, P.A.Y., Adiku, S.G.K., Sloan, J.J., 2015. Green roofs : A possible best Kane, S.N., Mishra, A., Dutta, A.K., 2016. Preface: International Conference on Recent management practice for enhancing the environmental quality of Ghanaian cities 9, Trends in Physics (ICRTP 2016). J. Phys. Conf. Ser. 755. doi: 10.1088/1742-6596/ 701–711. doi: 10.5897/AJEST2014. 755/1/011001. Athukorala, D., Murayama, Y., 2020. Spatial variation of land use/cover composition and Kavzoglu, T., Colkesen, I., 2009. A kernel functions analysis for support vector machines impact on surface urban heat island in a tropical sub-Saharan City of Accra Ghana. for land cover classification. Int. J. Appl. Earth Obs. Geoinf. 11, 352–359. https:// Sustain. 12 (19), 7953. https://doi.org/10.3390/SU12197953. doi.org/10.1016/j.jag.2009.06.002. Avashia, V., Garg, A., Dholakia, H., 2021. Understanding temperature related health risk Kayaga, S.M., Amankwaa, E.F., Gough, K.V., Wilby, R.L., Abarike, M.A., Codjoe, S.N.A., in context of urban land use changes. Landsc. Urban Plan. 212, 104107. https://doi. Kasei, R., Nabilse, C.K., Yankson, P.W.K., Mensah, P., Abdullah, K., Griffiths, P., org/10.1016/j.landurbplan.2021.104107. 2021. Cities and extreme weather events: impacts of flooding and extreme heat on Chen, S., Haase, D., Qureshi, S., Firozjaei, M.K., 2022. Integrated land use and urban water and electricity services in Ghana. Environ. Urban. 33, 131–150. https://doi. function impacts on land surface temperature: implications on urban heat mitigation org/10.1177/0956247820952030. in berlin with eight-type spaces. Sustain. Cities Soc. 83, 103944. Kim, S.W., Brown, R.D., 2021. Urban heat island (UHI) variations within a city boundary: Codjoe, S.N.A., Gough, K.V., Wilby, R.L., Kasei, R., Yankson, P.W.K., Amankwaa, E.F., a systematic literature review. Renew. Sustain. Energy Rev. 148, 111256 https://doi. Abarike, M.A., Atiglo, D.Y., Kayaga, S., Mensah, P., Nabilse, C.K., Griffiths, P.L., org/10.1016/j.rser.2021.111256. 2020. Impact of extreme weather conditions on healthcare provision in urban Krehbiel, C.P., Jackson, T., Henebry, G.M., 2016. Web-enabled landsat data time series Ghana. Soc. Sci. Med. 258, 113072 https://doi.org/10.1016/j. for monitoring urban heat island impacts on land surface phenology. IEEE J. Sel. socscimed.2020.113072. Top. Appl. Earth Obs. Remote Sens. 9, 2043–2050. https://doi.org/10.1109/ Das, M., Das, A., 2020. Exploring the pattern of outdoor thermal comfort (OTC) in a JSTARS.2015.2496951. tropical planning region of eastern India during summer. Urban Clim. 34 (August), Kumari, B., Shahfahad, Tayyab, M., Ahmed, I.A., Baig, M.R.I., Ali, M.A., Asif, Usmani, T. 100708 https://doi.org/10.1016/j.uclim.2020.100708. M., Rahman, A., 2021. Land use/land cover (LU/LC) change dynamics using indices DESA-UN DESA-UN, 2018. World Urbanization Prospects: The 2018 Revision, overlay method in Gautam Buddha Nagar District-India. GeoJournal. doi: 10.1007/ Demographic Research. New York. s10708-021-10374-w. Dissanayake, D., Morimoto, T., Murayama, Y., Ranagalage, M., Handayani, H.H., 2019. Li, J., Wang, Y., Ni, Z., Chen, S., Xia, B., 2020. An integrated strategy to improve the Impact of urban surface characteristics and socio-economic variables on the spatial microclimate regulation of green-blue-grey infrastructures in specific urban forms. variation of land surface temperature in Lagos City, Nigeria. Sustain. 11 (1), 25. J. Clean. Prod. 271, 122555 https://doi.org/10.1016/j.jclepro.2020.122555. Doe, E.K., Aikins, B.E., Njomaba, E., Owusu, A.B., 2018. Land use land cover change Lin, Y., Jim, C.Y., Deng, J., Wang, Z., 2018. Urbanization effect on spatiotemporal within Kakum Conservation Area in the Assin South District from 1991–2015. West thermal patterns and changes in Hangzhou (China). Build. Environ. 145, 166–176. African J. Appl, Ecol. https://doi.org/10.1016/j.buildenv.2018.09.020. Easterling, D.R., Horton, B., Jones, P.D., Peterson, T.C., Karl, T.R., Parker, D.E., Liu, C.C., Lin, T.C., Yuan, K.Y., Chiueh, P.T., 2022. Spatio-temporal prediction and factor Salinger, M.J., Razuvayev, V., Plummer, N., Jamason, P., Folland, C.K., 1997. identification of urban air quality using support vector machine. Urban Clim. 41, Maximum and minimum temperature trends for the globe. Science 277 (5324), 101055 https://doi.org/10.1016/j.uclim.2021.101055. 364–367. Malley, C.O., Piroozfar, P., Farr, E.R.P., Pomponi, F., 2015. Urban Heat Island (UHI) Ermida, S.L., Soares, P., Mantas, V., Göttsche, F., Trigo, I.F., 2020. Google earth engine mitigating strategies: a case-based comparative analysis. Sustain. Cities Soc. 19, open-source code for land surface temperature estimation from the landsat series. 222–235. https://doi.org/10.1016/j.scs.2015.05.009. Remote Sens. 12, 1471. https://doi.org/10.3390/rs12091471. Mantey, S., Tagoe, N., Abaidoo, C., 2014. Estimation of land surface temperature and Foody, G.M., 2020. Explaining the unsuitability of the kappa coefficient in the vegetation abundance relationship – a case study estimation of land surface assessment and comparison of the accuracy of thematic maps obtained by image temperature and vegetation. In: 3rd UMaT Bienn. Int. Min. Miner. Conf. 0–10. classification. Remote Sens. Environ. 239, 111630 https://doi.org/10.1016/j. Mantey, S., Tagoe, N.D., Abaidoo, C.A., 2014. Estimation of land surface temperature rse.2019.111630. and vegetation abundance relationship – a case study. In: 3rd UMaT Biennial Fu, P., Weng, Q., 2016. A time series analysis of urbanization induced land use and land International Mining & Mineral Conference. cover change and its impact on land surface temperature with Landsat imagery. Meyers, J., Langston, A., Devereux, D., Lin, B., 2020. Mapping land surface temperatures Remote Sens. Environ. 175, 205–214. and heat vulnerability in Darwin. CSIRO, Australia, City of Darwin. Ghana Statistical Service, 2021. Ghana 2021 Population and Housing Census, Population Minu, S., Shetty, A., 2015. A comparative study of image change detection algorithms in of Regions and Districts, Volume 3A. Accra, Ghana. MATLAB. Aquat. Procedia 4, 1366–1373. https://doi.org/10.1016/j. Gonzalez-Trevizo, M.E., Martinez-Torres, K.E., Armendariz-Lopez, J.F., Santamouris, M., aqpro.2015.02.177. Bojorquez-Morales, G., Luna-Leon, A., 2021. Research trends on environmental, Nangombe, S.S., Zhou, T., Zhang, W., Zou, L., Li, D., 2019. High-temperature extreme energy and vulnerability impacts of Urban Heat Islands: an overview. Energ. events over Africa under 1.5 and 2 ◦C of global warming. J. Geophys. Res. Atmos. Buildings 246, 111051. https://doi.org/10.1016/j.enbuild.2021.111051. 124 (8), 4413–4428. Guha, S., Govil, H., Gill, N., Dey, A., 2021. A long-term seasonal analysis on the Ning, J., Gao, Z., Meng, R., Xu, F., Gao, M., 2018. Analysis of relationships between land relationship between LST and NDBI using Landsat data. Quat. Int. 575–576, surface temperature and land use changes in the Yellow River Delta. Front. Earth Sci. 249–258. https://doi.org/10.1016/j.quaint.2020.06.041. 12 (2), 444–456. Halder, B., Bandyopadhyay, J., Banik, P., 2021. Monitoring the effect of urban Ningrum, W., 2018. Urban Heat Island towards Urban Climate. IOP Conf. Ser. Earth development on urban heat island based on remote sensing and geo-spatial approach Environ. Sci. 118, 012048. in Kolkata and adjacent areas, India. Sustain. Cities Soc. 74, 103186 https://doi.org/ Nyamekye, C., Kwofie, S., Agyapong, E., Ofosu, S.A., Arthur, R., Appiah, L.B., 2021. 10.1016/j.scs.2021.103186. Integrating support vector machine and cellular automata for modelling land cover Huang, C., Davis, L.S., Townshend, J.R.G., 2002. An assessment of support vector change in the tropical rainforest under equatorial climate in Ghana. Curr. Res. machines for land cover classification. Int. J. Remote Sens. 23, 725–749. https://doi. Environ. Sustain. 3, 100052 https://doi.org/10.1016/j.crsust.2021.100052. org/10.1080/01431160110040323. Oduro, C.Y., Adamtey, R., Ocloo, K., 2015. Urban growth and livelihood transformations Huang, S., Tang, L., Hupy, J.P., Wang, Y., Shao, G., 2021. A commentary review on the on the fringes of African cities: a case study of changing livelihoods in Peri-Urban use of normalized difference vegetation index (NDVI) in the era of popular remote Accra. Environ. Nat. Resour. Res. 5, 81–98. https://doi.org/10.5539/enrr.v5n2p81. sensing. J. For. Res. 32, 1–6. https://doi.org/10.1007/s11676-020-01155-1. Ojeh, V.N., Balogun, A.A., Okhimamhe, A.A., 2016. Urban-rural temperature differences Imhoff, M.L., Zhang, P., Wolfe, R.E., Bounoua, L., 2010. Remote sensing of the urban heat in Lagos. Climate 4, 1–18. https://doi.org/10.3390/cli4020029. island effect across biomes in the continental USA. Remote Sens. Environ. 114, Owusu, A.B., 2018. An assessment of urban vegetation abundance in Accra Metropolitan 504–513. https://doi.org/10.1016/j.rse.2009.10.008. Area, Ghana: a geospatial approach. J. Environ. Geogr. 11, 37–44. https://doi.org/ IPCC, 2021. Climate Change 2021: The Physical Science Basis. Contribution of Working 10.2478/jengeo-2018-0005. Group I to the Sixth Assessment Report of the Intergovernmental Panel on Climate Park, Y., Guldmann, J.M., Liu, D., 2021. Impacts of tree and building shades on the urban Change [Masson-Delmotte, V., P. Zhai, A. Pirani, S. L. Connors, C. Péan, S. Berger, N. heat island: combining remote sensing, 3D digital city and spatial regression Caud, Y. Chen,. Cambridge Univ. Press In Press. approaches. Comput. Environ. Urban Syst. 88, 15. https://doi.org/10.1016/j. Jenerette, G.D., Harlan, S.L., Buyantuev, A., Stefanov, W.L., Declet-Barreto, J., compenvurbsys.2021.101655. Ruddell, B.L., Myint, S.W., Kaplan, S., Li, X., 2016. Micro-scale urban surface Pelta, R., Chudnovsky, A.A., 2017. Spatiotemporal estimation of air temperature patterns temperatures are related to land-cover features and residential heat related health at the street level using high resolution satellite imagery. Sci. Total Environ. 579, impacts in Phoenix, AZ USA. Landscape Ecol. 31 (4), 745–760. 675–684. https://doi.org/10.1016/j.scitotenv.2016.11.042. 879 R.R. Gyimah et al. T h e E g y p t i a n J o u r n a l o f R e m o t e S e n s i n g a n d S p a c e S c i e n c e s 26 (2023) 861–880 Puplampu, D.A., Boafo, Y.A., 2021. Exploring the impacts of urban expansion on green IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens. 14, 8316–8325. https://doi.org/ spaces availability and delivery of ecosystem services in the Accra metropolis. 10.1109/JSTARS.2021.3105586. Environ. Challenges 5, 100283. https://doi.org/10.1016/j.envc.2021.100283. Tian, Y., Chen, C., Chen, X., Zhang, Q., Sun, R., 2020. Research on real-time analysis Quan, Q., Liang, W., Yan, D., Lei, J., 2022. Influences of joint action of natural and social technology of urban land use based on support vector machine. Pattern Recogn. Lett. factors on atmospheric process of hydrological cycle in Inner Mongolia, China. 133, 320–326. https://doi.org/10.1016/j.patrec.2020.03.022. Urban Clim. 41, 101043 https://doi.org/10.1016/j.uclim.2021.101043. Tuffour-Mills, D., Antwi-Agyei, P., Addo-Fordjour, P., 2020. Trends and drivers of land Ranagalage, M., Wang, R., Gunarathna, M.H.J.P., Dissanayake, D.M.S.L.B., cover changes in a tropical urban forest in Ghana. Trees, For. People 2, 100040. Murayama, Y., Simwanda, M., 2019. Spatial forecasting of the landscape in rapidly https://doi.org/10.1016/j.tfp.2020.100040. urbanizing hill stations of South Asia: a case study of Nuwara Eliya, Sri Lanka UN-HABITAT, 2020. World Cities Report 2020; The Value of Sustainable Urbanization. (1996–2037). Remote Sens. 11 (15), 1743. doi: 10.18356/c41ab67e-en. Ringard, J., Dieppois, B., Rome, S., Diedhiou, A., Pellarin, T., Konaré, A., Diawara, A., Vision, M., 1993. Image pre-processing 56–57. Konaté, D., Dje, B.K., Katiellou, G.L., Seidou Sanda, I., Hassane, B., Vischel, T., Völker, S., Baumeister, H., Classen, T., Hornberg, C., Kistemann, T., 2013. Evidence for Garuma, G.F., Mengistu, G., Camara, M., Diongue, A., Gaye, A.T., Descroix, L., 2016. the temperature-mitigating capacity of urban blue space - a health geographic The intensification of thermal extremes in west Africa. Glob. Planet. Change 139, perspective. Erdkunde 67, 355–371. https://doi.org/10.3112/erdkunde.2013.04.05. 66–77. Wemegah, C.S., Yamba, E.I., Aryee, J.N.A., Sam, F., Amekudzi, L.K., 2020. Assessment of Rockström, J., Steffen, W., Noone, K., Persson, Å., Chapin, F.S., Lambin, E., Lenton, T.M., urban heat island warming in the greater accra region. Sci. African 8, e00426. Scheffer, M., Folke, C., Schellnhuber, H.J., Nykvist, B., de Wit, C.A., Hughes, T., van WHO, 2013. Protecting health from climate change: vulnerability and adaptation der Leeuw, S., Rodhe, H., Sörlin, S., Snyder, P.K., Costanza, R., Svedin, U., assessment. WHO Libr. Cat. Data. Falkenmark, M., Karlberg, L., Corell, R.W., Fabry, V.J., Hansen, J., Walker, B., Wilby, R.L., Kasei, R., Gough, K.V., Amankwaa, E.F., Abarike, M., Anderson, N.J., Liverman, D., Richardson, K., Crutzen, P., Foley, J., 2009. Planetary boundaries: Codjoe, S.N.A., Griffiths, P., Kaba, C., Abdullah, K., Kayaga, S., Matthews, T., exploring the safe operating space for humanity. Ecol. Soc. https://doi.org/10.5751/ Mensah, P., Murphy, C., Yankson, P.W.K., 2021. Monitoring and moderating ES-03180-140232. extreme indoor temperatures in low-income urban communities. Environ. Res. Lett. Roy, S., Pandit, S., Eva, E.A., Bagmar, M.S.H., Papia, M., Banik, L., Dube, T., Rahman, F., 16 (2), 024033. Razi, M.A., 2020. Examining the nexus between land surface temperature and urban Xu, L., Liu, X., Tong, D., Liu, Z., Yin, L., Zheng, W., 2022. Forecasting Urban Land Use growth in Chattogram Metropolitan Area of Bangladesh using long term Landsat Change Based on Cellular. series data. Urban Clim. 32, 100593 https://doi.org/10.1016/j.uclim.2020.100593. Xu, L., Liu, X., Tong, D., Liu, Z., Yin, L., Zheng, W., 2022b. Forecasting urban land use Saha, S., Saha, A., Das, M., Saha, A., Sarkar, R., Das, A., 2021. Analyzing spatial change based on cellular automata and the PLUS model. Land 11, 652. https://doi. relationship between land use/land cover (LULC) and land surface temperature org/10.3390/land11050652. (LST) of three urban agglomerations (UAs) of Eastern India. Remote Sens. Appl.: Soc. Yao, R., Wang, L., Huang, X., Sun, L., Chen, R., Wu, X., Zhang, W., Niu, Z., 2021. A robust Environ. 22, 100507 https://doi.org/10.1016/j.rsase.2021.100507. method for filling the gaps in MODIS and VIIRS land surface temperature data. IEEE Shao, Y., Lunetta, R.S., 2012. Comparison of support vector machine, neural network, Trans. Geosci. Remote Sens. 59, 10738–10752. https://doi.org/10.1109/ and CART algorithms for the land-cover classification using limited training data TGRS.2021.3053284. points. ISPRS J. Photogrammetry Remote Sensing 70, 78–87. https://doi.org/ Zhang, N., Chen, Y., Luo, L., Wang, Y., 2017. Effectiveness of different urban heat island 10.1016/j.isprsjprs.2012.04.001. mitigation methods and their regional impacts. Am. Meteorol. Soc. 18, 2991–3012. Simwanda, M., Ranagalage, M., Estoque, R.C., Murayama, Y., 2019. Spatial analysis of https://doi.org/10.1175/JHM-D-17-0049.1. surface urban heat Islands in four rapidly growing african cities. Remote Sens. 11, Zhang, Y., Odeh, I.O.A., Han, C., 2009. Bi-temporal characterization of land surface 1–20. https://doi.org/10.3390/rs11141645. temperature in relation to impervious surface area, NDVI and NDBI, using a sub- Songsore, J., 2020. The urban transition in ghana: urbanization, national development pixel image analysis. Int. J. Appl. Earth Obs. Geoinf. 11 (4), 256–264. and poverty reduction. Ghana Soc. Sci. J. 17, 120–172. Zhao, T., Shi, J., Entekhabi, D., Jackson, T.J., Hu, L., Peng, Z., Yao, P., Li, S., Kang, C.S., Songsore, J., McGranahan, G., 2012. Scaling Urban Environmental Challenges, Scaling 2021. Retrievals of soil moisture and vegetation optical depth using a multi-channel Urban Environmental Challenges: From Local to Global and Back. Routledge. doi: collaborative algorithm. Remote Sens. Environ. 257, 112321 https://doi.org/ 10.4324/9781849772471. 10.1016/j.rse.2021.112321. Songsore, J., Nabila, J.S., Yvon, Y., Avle, S., Bosque-Hamilton, E.K., Amponsah, P.E., Zheng, Y., Tang, L., Wang, H., 2021. An improved approach for monitoring urban built- OAlhassan, O., Satterthwaite, D., 2006. Environmental Health Watch and Disaster up areas by combining. J. Clean. Prod. 328, 129488 https://doi.org/10.1016/j. Monitoring in the Greater Accra Metropolitan Area (Gama), 2005. University of jclepro.2021.129488. Ghana, Legon-Accra, Ghana. Zhou, D., Bonafoni, S., Zhang, L., Wang, R., 2018. Remote sensing of the urban heat Sun, D., Hu, C., Wang, Y., Wang, Z., Zhang, J., 2021. Examining spatio-temporal island effect in a highly populated urban agglomeration area in East China. Sci. Total characteristics of urban heat islands and factors driving them in Hangzhou, China. Environ. 628–629, 415–429. https://doi.org/10.1016/j.scitotenv.2018.02.074. 880