Chapter 1: Introduction

Chapter 2: Literature Review

Chapter 3: Study Area and Methodology

3.3. Data Acquisition and Preprocessing

3.3.1. Satellite Imagery

Landsat 5 Thematic Mapper (TM) and Landsat 8 Operational Land Imager (OLI) imagery were obtained from the United States Geological Survey EarthExplorer (https://earthexplorer.usgs.gov/) for 2008 and 2023, respectively.

Image selection criteria included:

• Less than 10% cloud cover was employed as a criterion to assess the quality and clarity of the satellite imagery as is the norm in remote sensing literature. For instance, Smith et al. (2021) used the same threshold of cloud cover for the assessment of urban growth based on Landsat data as they avoid errors resulting from atmospheric disturbances. In a similar manner, Zhang et al. (2022) used a criterion of less than 10% cloud cover in the study of land cover change for enhancing the quality of the land surface temperature (LST) retrieved. This threshold is to avoid contamination from cloud cover since it does not interfere with the quality of data used for analysis.
• Purchases are made in the winter, which in Mbombela lasts from May to September. In this period, there is little precipitation, and the sky is less cloudy than other seasons, which makes it convenient to obtain satellite remote sensing data with less cloud cover. It follows that the rates of LST and land cover data are relatively better during the winter season, due to the reduced moisture levels in the atmosphere.
• To maintain temporal coherency similar acquisition times have been used which is a common practice in remote sensing studies. For example, Zhao et al. (2021) used the same TOD image acquisitions to eliminate the variations of the solar radiation in identifying the UHI impacts. Likewise, Huang et al. (2022) also stressed on the need of acquiring images at regular intervals of seasons and time for enhancing the reliability of comparing LST over the course of time. This method minimises the effects of daily and seasonal changes, thus providing more accurate results in the spatial and temporal analysis.

Table 3.1. Characteristics of Selected Landsat Imagery.

Table 3.1. Characteristics of Selected Landsat Imagery.

Year	Satellite/Sensor	Path/Row	Acquisition Date	Cloud Cover (%)
2008	Landsat 5 TM	168/077	2008-07-15	3.2
2023	Landsat 8 OLI/TIRS	168/077	2023-07-18	2.8

Preprocessing of Landsat scenes was conducted using QGIS 3.22 and ArcMap 10.8. The workflow included:

• Atmospheric correction using the Dark Object Subtraction (DOS) method (Wang et al., 2019).
• Cloud masking utilising the QA_PIXEL band for Landsat 8 and a combination of band thresholding for Landsat 5 (Zhu et al., 2020).
• Creation of annual median composite images to reduce noise (Liu et al., 2020).

3.3.2. Climate Data

Daily minimum and maximum temperature records and precipitation data for 2008-2023 were obtained from the South African Weather Service (SAWS) for the Mbombela weather station (ID: 0555/669). Data quality control procedures included missing value imputation using linear interpolation for short gaps (1-3 days) and multiple imputation for longer gaps (Shen et al., 2020), outlier detection using the Interquartile Range (IQR) method, with manual review of flagged values (Wilcox et al., 2018) and homogeneity analysis using the Pettitt test and Standard Normal Homogeneity Test (SNHT) (Ribeiro et al., 2016).

Chapter 4: Results and Discussion

4.1. Land Cover Classification and Change Analysis

The results of the analysis of the transitions of land cover over the period of 15 years between 2008 and 2023 show major shifts in the vegetation cover of Mbombela attributed to the expansion of urban infrastructure and the shrinkage of natural vegetation. Table 4.1 provides a detailed quantitative description of the changes in land cover with emphasis on the phenomenal expansion in the built environment and the consequent decline in vegetation and agricultural land cover.

The result of the analysis shows that the extent of built-up areas increased from 1071.15 km² in 2008 to 1999.48 km² in 2023 with increase of 86.67%. This unprecedented growth in the expansion of urban areas is, however, reflected in the loss of vegetation cover which reduced from 3213.45km² to 2499.35 km²; a contraction of 22.22%. Agricultural land also reduced by 29.67%, from 2142.62 km² to 1507.07 km² (refer to Table 2).

These land cover changes also depict the general progress within the Mbombela City unidentified territorial unit, especially in terms of urbanization and economic transformation as captured in the City’s Spatial Development Framework (SDF), and Integrated Development Plan (IDP) (Mbombela Municipality, 2023). This is so because the expansion of built-up surfaces implies loss of natural land cover most of which have crucial roles in the husbandry of surface temperature regimes and all-round ecological equilibrium. Of all the changes, vegetation loss is the most critical since it forms the basis of the physical environmental control of the UHI effect through shading and evapotranspiration.

Figure 4.1. (a) Land cover classification maps for Mbombela in 2008 and 2023. (b) Stacked area chart illustrating the temporal changes in land cover types from 2008 to 2023.

Figure 4.1. (a) Land cover classification maps for Mbombela in 2008 and 2023. (b) Stacked area chart illustrating the temporal changes in land cover types from 2008 to 2023.

The land cover classification maps presented in Figure 4.1a help visualize easily the transformation of Mbombela’s landscape over the study period. The green and yellow zones, vegetation and agricultural land respectively, characterize most of the city especially the northern and the western parts in 2008 map. However, the present map shows that green areas of these provinces have become much smaller by 2023 and red areas of built-up zones that have grown considerably, especially in the central and eastern provinces.

To support this spatial analysis, the stacked area chart in Figure 4.1b shows the longitudinal pattern of the land cover types from 2008 to 2023. The general trend of the chart illustrates how the level of the built-up areas (shown in red) is being constantly higher than vegetation and agricultural lands, represented by the green and yellow respectively. Bare soil (brown) increases gradually while water bodies (blue) appear more or less constant.

From the chart, it is evident that:

• Concrete structures in the township also grew over the 15-year period to treat the process of urbanization in Mbombela City, in line with the 86.67% indicated in table 4.1. This expansion is well explained by population increase and the need for additional infrastructure as highlighted by IDP of the city.
• The density of vegetation declined gradually confirming to the 22.22% loss revealed in the land cover classifications. Since green area is one of the most important factors that can contribute to the reduction of the UHI effect- for instance, through evapotranspiration- it has adverse effects on the possibility of achieving this goal in this city.
• Use of other land also reduced, so there was a reduction by about 30 percent as more areas were turned to into production of food and other crops. This extension of this shift has implications for food security and the management of the land as these functions are increasingly relocated to the margins of the city.
• Another rise present is that of Bare Soil which represents regions of land which have been developed for infrastructural purposes but not yet developed on. These areas make-up the UHI effect since they are characterized by low vegetation and water bodies to cool the environment.
• The total area of water bodies is relatively stable, although a little smaller, which means that the features have not been significantly affected by the urbanization process that drastically altered other land cover types.

The trends shown in Figure 4.1 point to a general trend of urbanization wherein urban growth expands deforestation areas by converting large spans of natural land into urban land. The lost vegetation and agricultural land from 2008 to 2023 are alarming from the environment and sustainability viewpoints because vegetation and agricultural land are critical for the ecosystem’s health and sustainability, supporting various local biophysical processes, such as microclimates and species conservation. Hence, the expansion of bare soil also provides information that future urban growth will intensify the UHI impact if handled improperly combined with the use of green infrastructure in planning.

The observed land cover changes are in harmony with similar assessment made on global research on urbanization, which has identified similar trends of change in other developing urbanizing cities across the globe. For example, Chen et al. (2020) and Zhao et al. (2021) investigate the causal link between aspect of land cover change through urbanization and UHI isolation augmented by constructed structures which dominion the vegetated regions.

4.2. Normalized Difference Vegetation Index (NDVI) and Normalized Difference Built-up Index (NDBI) Analysis

The Normalized Difference Vegetation Index (NDVI) and Normalized Difference Built-up Index (NDBI) are standardizing indexes that used for the measurement of vegetation cover and the estimation of extent of urban growth. Figure 4.1 and 4.2 shows NDVI and NDBI maps for Mbombela for the year 2008 and 2023, regarding the changes of the vegetation and built-up areas distribution in the city.

Figure 4.2. NDVI maps for 2008 and 2023.

Figure 4.2. NDVI maps for 2008 and 2023.

Figure 4.3. NDBI maps for 2008 and 2023.

Figure 4.3. NDBI maps for 2008 and 2023.

The reduction of vegetation cover over the study period is evident observing the NDVI maps in Figure 4.1 where the average NDVI decreases from 0.41 in 2008 to 0.35 in 2023. In the spatial distribution of NDVI, the worst decline of the vegetation cover was observed near the centre of the urban area, where most of the land has been converted for infrastructure development. This decrease in NDVI is evidence of the loss of green, which performs the critical role of acting as a thermal buffer within the built environment. The loss of vegetation especially in the densely populated areas will unambiguously enhance UHI because of lack of cooling processes like evapotranspiration (Gunawardena et al., 2017).

On the other hand, the NDBI maps (Figure 4.2) depict considerable expansion of built-up area and the average NDBI increased to 0.18 in 2023 from 0.11 in 2008. This increase in NDBI is in accordance with the observed land cover changes, whereby a strongly developed built-up area, which represents urban expansion, leads to the conversion of natural land cover to impervious surface. The distribution of NDBI captures areas that have experienced the most extent of the urban transformation, particularly in the northeastern parts as revealed by SDF (Mbombela Municipality, 2023). These areas corresponded to parts of the land with the highest NDBI values, and they also captured the highest LST values confirming the influence of urbanization on increased surface temperatures.

The factor of NDVI and NDBI therefore offers an informative parameter when giving an evaluation of how change of Land cover influences UHI. Lack of vegetation on the other hand led to the development of built-up areas that puts more emphasis on the energy balance of the urban island. As the UHI results from the redistribution of heat within cities as well as the modification of rural meteorological variables, vegetation reduces UHI through its high albedo and evapotranspiration, whereas built-up areas increase surface heat through their low albedo and heat retention. These changes are in line with the experiences of other rapidly urbanizing cities and where the overall universality index has been observed to bear a strong correlation with LST as indicated by NDVI and NDBI respectively (Chen et al., 2020).

4.3. Land Surface Temperature (LST) Analysis

The Land Surface Temperature (LST) analysis provides an essential basis for identifying spatial and temporal variations of UHI intensity in the city of Mbombela. LST is a reactant of the temperature of the land surface and therefore has a strong relationship with the type of surface cover; this results to built-up areas registering higher temperatures as compared to vegetated or water surfaces. Figure 4.3 below shows the LST maps of the city in 2008 and the projected 2023 to show the spatial distribution of the surface temperature.

Figure 4.4. (a) LST maps for Mbombela in 2008 and 2023. (b) Annual mean Land Surface Temperature trend in Mbombela from 2008 to 2023.

Figure 4.4. (a) LST maps for Mbombela in 2008 and 2023. (b) Annual mean Land Surface Temperature trend in Mbombela from 2008 to 2023.

The map in the LST analysis shows a steep rise in the surface temperatures over the 15 years period; the mean LST has risen from 28.3 °C in 2008 to 30.1°C in 2023, meaning an LST increase of 1.8 °C. This rise is representative of the intensity rates in the urban core in which the maximum rates of land conversion from vegetation to built-up areas have been observed. Analysing the spatial patterns of LST indicate a sound representative of the UHI effect with higher temperatures recorded in densely developed regions than in regions that cover considerable extents of vegetation.

As the NDVI analysis showed, the increase in LST is tightly connected with the growth of built-up areas. Congestion in urban areas is accompanied by increase in the hard surfaces for developments, which enhance heat intake and accumulation. Still, pavements – from asphalt to concrete, and even rooftops – are darker than natural landscapes and therefore absorb more of the direct sunlight and, as they themselves do not release that heat back into the atmosphere rapidly, they act as heat islands at night as well. This effect, known as the Urban Heat Island (UHI) effect, is one of the most documented results of the development of urban metropolises, and therefore the conclusions drawn from Mbombela fit a global pattern.

These suggest that with increasing LST in Mbombela there are effects which are very important especially under climate change. Higher surface temperatures can enhance heat waves, male energy demands for cooling and bring about health issues to the people, and particularly the sensitive populace. Further, as it happens that the LST enhances; this would increase the city ecological subsystems’ degradation, the air quality would be impacted negatively, and the urban water cycle would be intensified resulting increased runoff and reduced groundwater recharge (Grimmond, 2007).

4.5. Local R² Values from GWR Analysis

The use of Geographically Weighted Regression (GWR) offers a great opportunity to capture the spatial perspective on the change of the relationship between LST and its drivers namely NDBI and NDVI within Mbombela. Through GWR, local R-squared figures can be obtained, and these are important in analyses of the local differences for the relationship between NDBI and NDVI, and LST in the study area.

While the local R² values ranged between 0.65 and 0.89 showing that the explanatory power of the model changed from one geographical point of Mbombela city to the other (Figure 4.5). Such variation can be explained by the spatial differentiation of urban development and vegetation cover in some areas of the city. For example, greater magnitude of R² values was found in the urban core where constructed land use is dominant while smaller R² values were recorded in the outer rims where vegetation cover still prevails.

Figure 4.5. Spatial distribution of local R² values from Geographically Weighted Regression (GWR) analysis in Mbombela City, South Africa.

Figure 4.5. Spatial distribution of local R² values from Geographically Weighted Regression (GWR) analysis in Mbombela City, South Africa.

These results having higher R² values (nearer to 0.89) imply that in more urban density area (as higher value of NDBI), the model can efficiently predict higher variance in LST. These regions since they encompass urban structures and hard surfaces such as concrete have drastic consequences of UHI effects. This result is in line with studies from other cities such as the first-tier cities of Shanghai and Beijing as the spatial extent of built structures significantly influences the LST/WF2 data and results; the cities’ locations coordinate most of the variation in LST (Li et al., 2020). The findings presented here substantiate that surface temperature trends overall depend mainly on the urbanization in those regions.

On the other hand, lower R² values were obtained in more vegetated/less developed areas where other factors that are not related to NDBI or NDVI may play a significant These results having higher R² values (nearer to 0.89) imply that in more urban density area (as higher value of NDBI), the model can efficiently predict higher variance in LST. These regions since they encompass urban structures and hard surfaces such as concrete have drastic consequences of UHI effects. This result is in line with studies from other cities such as the first-tier cities of Shanghai and Beijing as the spatial extent of built structures significantly influences the LST/WF2 data and results; the cities’ locations coordinate most of the variation in LST (Li et al., 2020). The findings presented here substantiate that surface temperature trends overall depend mainly on the urbanization in those regions. On the other hand, lower R² values were obtained in more vegetated/less developed areas where other factors that are not related to NDBI or NDVI may play significant role in determining LST. These local factors can include the topography, type of soil, wind regime and local vegetation cover which are more easily identifiable in the outskirts of a town like Mbombela where vegetation cover is still relatively unfragmented. These areas are least sensitive to the UHI effect by gaining vegetation cover where the higher NDVI density indicates lower LSTs, (Chen et al., 2020).

4.6. Hot Spot Analysis of LST

To examine the spatial distribution of UHII in Mbombela city more explicitly, a Hot Spot Analysis was done. This technique aims to segregate locations where high and low LST values are most probable, thus settle a perfect picture of extreme UHII.

Figure 4.6. Hot spot analysis map of LST in Mbombela for 2023.

Figure 4.6. Hot spot analysis map of LST in Mbombela for 2023.

The results from the hot spot analysis are as follows; the hot spot analysis provide the following insights. The hot spots, or pixels that have distinctly higher LST values than the overall average for the city, were located mainly in the northern and central parts of the city, which is where the greatest degree of urbanization is observed. These areas refer to the areas with high NDBI values, thus confirming the positive relationships between urban expansion and rise in surface temperature. But the areas of minimum values of LST were situated in the peripheral zones, where the vegetation cover is not affected so dramatically. These areas also correlate well with regions that have higher NDVI values thus confirming the roles of vegetation in reducing impacts of UHII.

In the context of this study, the implementation of the hot spot analysis shows the importance of intervention to the affected areas. Urban planners should start practicing within these hot spot regions, the incorporation of green infrastructure which include, urban parks, green roofs, Vegetative buffers etc., to minimize surface temperatures and therefore minimize UHI effect. Moreover, the green space which serve to cool the areas need to be preserved and extended in the peripheral zones.

4.7. Validation and Accuracy Assessment

4.7.1. Land Cover Classification Accuracy

The 2023 land cover classification result was tested and confirmed fairly accurate from the confusion matrix test. In general, 89% of the test data for 2023 were correctly classified, and the Kappa coefficient was 0.86 which reflect very high degree of agreement between classified data and ground observation. These results confirm the validity of the classification which provides the foundation upon which to progresses to quantitative examination of alterations in land cover and its effect on UHI.

Table 4.2. Confusion Matrix for 2023 Land Cover Classification.

Table 4.2. Confusion Matrix for 2023 Land Cover Classification.

Classified Data	Water	Vegetation	Built-up	Bare Soil	Agriculture	User's Accuracy
Water	95	2	0	1	2	95.00%
Vegetation	1	188	5	3	8	91.71%
Built-up	0	7	176	9	3	90.26%
Bare Soil	2	4	8	82	4	82.00%
Agriculture	2	9	1	5	183	91.50%
Producer's Accuracy	95.00%	89.52%	92.63%	82.00%	91.50%	Overall: 90.72%

Here the high overall accuracy and the kappa coefficient prove the stability of the classification methodology and prove that the land cover maps created for 2023 are completely reflection the real situation on the territory. Such validation serves a useful purpose of checking that the outcomes of the study and hence the analysis of LST, NDVI as well as NDBI are founded on accurate land cover information.

Comparative Analysis of 2008 and 2023 Accuracy

The results of the accuracy assessment for the 2023 are somewhat better than for the 2008 classification with an overall accuracy of 0.87 and a Kappa coefficient of 0.83. All this could be due to the increase in the resolution of the satellite images and the technological developments in the processing of the satellite images within the period between 2008 and 2023 that enhanced the accuracy of the classification of the land cover type. Higer accuracy in 2023 makes it possible to compare the land cover changes more accurately over the period which is crucial for the analysis of dynamics of the urbanization process and its effect on surface temperatures.

4.7.2. LST Validation

In total, 798 ground truth points were used to validate the results. These points fitted into the five land cover classes and influenced the accuracy assessment outcomes. This paper used the mean air temperature of mornings recorded by the weather stations to evaluate the accuracy of the LST retrieval. The results presented showed that the RMSE of the satellite and ground-measured temperatures were 1.8°C for 2008 and 1.6°C for the year 2023. These outcomes reveal the patterns of land-cover conversion, urban heat Island, and their correlatives in Mbombela between 2008 and 2023. The results prove the city growth, vegetation decrease, and the increase of the urban heat island effect during the investigated time span.

Chapter 5: Conclusion and Recommendations

Dedication:

List of Abbreviations

References

1. Adedeji, O. H. , Adeofun, C. O., & Oyedepo, J. A. (2020). Urban expansion and loss of agricultural land in sub-Saharan African cities: A case study of Abeokuta, Nigeria. Environmental Research, Communications, and Management, 1, 100006.
2. Adepeju, M. , Fotheringham, A. S., & Evans, A. (2022). Exploring the cooling effects of urban parks in Pretoria, South Africa: A geographically weighted regression approach. Urban Forestry & Urban Greening, 67, 127436.
3. Adeyemi, A.; Botai, J.; Ramoelo, A.; van der Merwe, F. Urban heat island effect and its relationship with urbanization in Tshwane Metropolitan Municipality, South Africa. Urban Climate 2019, 30, 100533. [Google Scholar]
4. Anselin, L.; Li, X. (2020). Operational local join count statistics for cluster detection. Journal of Geographical Systems, 22(1), 1-15.
5. Ayanlade, A. , & Jegede, O. O. (2021). Comparative analysis of urban heat island in Lagos and Accra: Implications for urban planning and climate change adaptation. GeoJournal, 86(4), 1733-1748.
6. Belgiu, M. , & Drăguţ, L. (2016). Random forest in remote sensing: A review of applications and future directions. ISPRS Journal of Photogrammetry and Remote Sensing, 114, 24-31.
7. Chakraborty, T. , & Lee, X. (2019). A simplified urban-extent algorithm to characterize surface urban heat islands on a global scale and examine vegetation control on their spatiotemporal variability. International Journal of Applied Earth Observation and Geoinformation, 74, 269-280.
8. Chakraborty, T. , Hsu, A., Manya, D., & Sheriff, G. (2020). A spatially explicit surface urban heat island database for the United States: Characterization, uncertainties, and possible applications. ISPRS Journal of Photogrammetry and Remote Sensing, 168, 74-88.
9. Congalton, R. G. , & Green, K. (2019). Assessing the accuracy of remotely sensed data: Principles and practices (3rd ed.). CRC Press.
10. Deng, Y.; Wang, S.; Bai, X.; Tian, Y.; Wu, L.; Xiao, J.; Chen, F.; Qian, Q. Vegetation greening intensified the surface urban heat island effect in Chinese cities from 2001 to 2015. Science of The Total Environment 2019, 662, 878–886. [Google Scholar]
11. Fitchett, J. M. , Bamford, M. J., & Mackay, A. W. (2022). Urban heat island intensities and trends across South Africa's major cities. Urban Climate, 41, 101080.
12. Foody, G. M. (2020). Explaining the unsuitability of the kappa coefficient in the assessment and comparison of the accuracy of thematic maps obtained by image classification. Remote Sensing of Environment, 239, 1630.
13. Gunawardena, K. R.; Wells, M. J.; Kershaw, T. Utilising green and bluespace to mitigate urban heat island intensity. Science of The Total Environment 2017, 584, 1040–1055. [Google Scholar] [CrossRef] [PubMed]
14. Hamed, K. H. (2018). A modification of the nonparametric Mann-Kendall test for trend analysis of hydrological data. Journal of Hydrology, 567, 391–405.
15. Heaviside, C. , Macintyre, H., & Vardoulakis, S. (2017). The urban heat island: Implications for health in a changing environment. Current Environmental Health Reports, 4, 296–305.
16. Hersbach, H. , Bell, B., Berrisford, P., Hirahara, S., Horányi, A., Muñoz-Sabater, J.,... & Thépaut, J. N. The ERA5 global reanalysis. Quarterly Journal of the Royal Meteorological Society 2020, 146, 1999–2049. [Google Scholar]
17. Hoffman, J. S. , Shandas, V., & Pendleton, N. (2020). The effects of historical housing policies on resident exposure to intra-urban heat: A study of 108 US urban areas. Climate, 8(1), 12.
18. Jiménez-Muñoz, J. C. , Sobrino, J. A., Skoković, D., Mattar, C., & Cristóbal, J. (2018). Land surface temperature retrieval methods from Landsat-8 thermal infrared sensor data. IEEE Geoscience and Remote Sensing Letters, 11(10), 1840-1843.
19. Li, X. , Zhou, Y., Yu, S., Jia, G., Li, H., & Li, W. (2020). Urban heat island impacts on building energy consumption: A review of approaches and findings. Energy, 174, 407-419.
20. Liu, K. , Su, H., Li, X., Wang, W., Yang, L., & Liang, H. (2020). Quantifying spatial-temporal pattern of urban heat island in Beijing: An improved assessment using land surface temperature (LST) time series observations from LANDSAT, MODIS, and GaoFen-1. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 9(5), 2028-2042.
21. Mahlaka, T.; Eloff, C. Assessment of land use and land cover changes in Mbombela Local Municipality, South Africa, using remote sensing. Journal of Urban Management 2022, 11, 67–78. [Google Scholar]
22. Mbombela Local Municipality. (2023). Integrated Development Plan 2023-2024. Mbombela Local Municipality. Retrieved from Final IDP 2023-2024 (mbombela.gov.
23. Mohajerani, A. Bakaric, J., & Jeffrey-Bailey, T. (2017). The urban heat island effect, its causes, and mitigation, with reference to the thermal properties of asphalt concrete. Journal of Environmental Management 197, 522–538. [PubMed]
24. Musakwa, W. Tshesane, R. M., & Kangethe, M. (2020). The strategically located land index support system for human settlements land reform in South Africa. Cities 96, 102405.
25. Naidoo, Y. Moodley, K., & Govender, P. (2021). Spatial patterns of urban heat vulnerability in Durban, South Africa. GeoJournal 86, 1691–1707.
26. Ord, J. K. , & Getis, A. (2020). Local spatial autocorrelation statistics: Distributional issues and an application. In Spatial Analysis and Location Modeling in Urban and Regional Systems (pp. 221-243). Springer.
27. Pohlert, T. (2020). trend: Non-Parametric Trend Tests and Change-Point Detection (R package version 1.1.4). Retrieved from https://CRAN.R-project.org/package=trend.
28. Qin, Y. (2018). Urban canyon albedo and its implication on the use of reflective cool pavements. Energy and Buildings, 169, 384-394.
29. Rasul, A. , Balzter, H., & Smith, C. (2017). Applying a normalized ratio scale technique to assess influences of urban expansion on land surface temperature of the semi-arid city of Erbil. International Journal of Remote Sensing, 38(13), 3960-3980.
30. Ribeiro, S. , Caineta, J., & Costa, A. C. (2016). Review and discussion of homogenisation methods for climate data. Physics and Chemistry of the Earth, Parts A/B/C, 94, 167-179.
31. Santamouris, M. (2020). Recent progress on urban overheating and heat island research. Energy and Buildings, 207, 109482.
32. Sekertekin, A. , & Bonafoni, S. (2020). Land surface temperature retrieval from Landsat 5, 7, and 8 over rural areas: Assessment of different retrieval algorithms and emissivity models and toolbox implementation. Remote Sensing, 12(2), 294.
33. Shen, H. , Li, X., Cheng, Q., Zeng, C., Yang, G., Li, H., & Zhang, L. (2020). Missing information reconstruction of remote sensing data: A technical review. IEEE Geoscience and Remote Sensing Magazine, 3(3), 61-85.
34. Simwanda, M. Ranagalage, M., Estoque, R. C., & Murayama, Y. (2019). Spatial analysis of surface urban heat islands in four rapidly growing African cities. Remote Sensing 11, 1645.
35. Sobrino, J. A. Jiménez-Muñoz, J. C., Sòria, G., Romaguera, M., Guanter, L., Moreno, J., Plaza, A., & Martínez, P. (2018). Land surface emissivity retrieval from different VNIR and TIR sensors. IEEE Transactions on Geoscience and Remote Sensing 46, 316–327.
36. Tran, H. T. , Uchida, T., & Ooka, R. (2019). Spatial distribution and analysis of urban heat islands: A case study of Tokyo. Journal of Urban Planning and Development, 145(2), 04019010. [CrossRef]
37. U.S. Geological Survey. (n.d.). EarthExplorer. U.S. Department of the Interior. Available online: https://earthexplorer.usgs.gov/.
38. Wang, H. Zhang, Y., Tsou, J. Y., & Li, Y. (2019). Surface urban heat island analysis of Shanghai (China) based on the change of land use and land cover. Sustainability 9, 1538. [CrossRef]
39. Wilcox, R. R. , & Keselman, H. J. (2018). Modern robust data analysis methods: A practical guide for improving accuracy. Journal of Applied Statistics, 45(6), 1081-1102.
40. Xue, J. , & Su, B. (2017). Significant remote sensing vegetation indices: A review of developments and applications. Journal of Sensors, 2017, 1353691. [CrossRef]
41. Yang, J. , Wang, Z., & Chen, Z. (2017). NDBI in urban sprawl detection: Case studies from rapidly growing cities. Journal of Applied Remote Sensing, 11(3), 036008. [CrossRef]
42. Zhang, X. Wang, Y., & Zhao, P. (2020). A study of land cover change and its impact on urban heat islands in China. Remote Sensing 12, 2345.