Available via license: CC BY 4.0
Content may be subject to copyright.
Spatial-Temporal Analysis of Urban Heat Islands in
Lilongwe, Zomba, Blantyre, and Mzuzu Cities:
Examining the role of Urban Green Spaces
Japhet N Khendlo
Mzuzu University https://orcid.org/0009-0008-0970-0237
Research Article
Keywords: Urban Heat Island, Land surface Temperature, LULC, Urban Green spaces
Posted Date: August 5th, 2024
DOI: https://doi.org/10.21203/rs.3.rs-4851778/v1
License: This work is licensed under a Creative Commons Attribution 4.0 International License.
Read Full License
Additional Declarations: The authors declare no competing interests.
[1]
Spatial-Temporal Analysis of Urban Heat Islands in Lilongwe, Zomba, Blantyre, 1
and Mzuzu Cities: Examining the role of Urban Green Spaces 2
3
Abstract 4
This study examines significant environmental transformations in Mzuzu, Lilongwe, Blantyre, and Zomba cities in Malawi 5
over the past 23 years, focusing on changes in land cover and land surface temperatures (LSTs). Our findings reveal substantial 6
decreases in forest cover and vegetation, alongside notable increases in bare land and built-up areas across all cities. These 7
changes are strongly correlated with rising LSTs, as evidenced by a highly significant negative correlation between LSTs and 8
forest/vegetation cover (R = -1.00) and a positive correlation with bare land and built-up areas. For instance, from 2000 to 9
2023, Mzuzu's minimum LST increased from 14.6°C to 19.5°C, and the maximum LST rose from 26.4°C to 34.3°C. In 10
Lilongwe, the minimum LST increased from 18.4°C to 21.7°C, and the maximum LST from 35.4°C to 46.8°C. The observed 11
trends highlight the critical need for urban planning that integrates green infrastructure and sustainable land use practices. 12
Key Words: Urban Heat Island; Land surface Temperature, LULC, Urban Green spaces 13
14
Introduction 15
Rapid urban expansion typically converts open spaces, such as natural land cover with soil and vegetation, into artificial 16
surfaces like concrete, asphalt, and other impermeable materials. This transformation causes changes in the absorption and 17
reflection of solar radiation and the surface energy balance, leading to noticeable differences in surface radiance and air 18
temperature between urban and rural areas. This temperature difference creates a phenomenon known as Urban Heat Islands 19
(UHIs) (Abulibdeh, 2021; Philip, 2022; Jabbar, Hamoodi and Al-Hameedawi, 2023). The intensity and thermal balance of 20
Urban Heat Islands (UHIs) are influenced by various factors, including weather and climate conditions, the extent of 21
urbanization, population dynamics, urban size and density, topography, urban design, presence of water bodies, land use and 22
cover, green areas, human-made heat, and construction materials (Igun and Williams, 2018; Reisi et al, 2019). 23
These factors will influence the intensity and spatial patterns of Urban Heat Islands (UHIs). Conversely, UHIs will impact 24
surface temperatures, evaporation rates, energy consumption, the degree of solar radiation absorption, water quality due to 25
heat transfer to the soil, greenhouse gas levels, air pollutant emissions, the area's albedo, and heat storage capacity. Ultimately, 26
these changes can compromise human health and comfort. Additionally, UHIs can affect wind turbulence, directly altering the 27
near-surface atmospheric environments over urban areas (Batty, Xie and Sun, 1999; Gee and Sarker, 2013; Gill et al., 2018). 28
The relationship between urban features and Urban Heat Islands (UHIs) can be summarized in four key points: (1) Green 29
areas, pavements, and buildings are crucial in local temperature variations; (2) The arrangement of buildings and urban 30
structures affects UHI development; (3) Tall buildings and narrow streets intensify UHI effects by trapping warm air and 31
reducing airflow; (4) Water and green spaces play a significant role in mitigating UHI effects (Gago et al., 2013; You et al, 32
2023). 33
Several studies (Tiangco, Lagmay and Argete, 2008; Sharifi and Lehmann, 2014; Oluseyi, Fanan and Magaji, 2015; Zhao et 34
al., 2016; Gill et al., 2018; Lee et al., 2021; Li, Stringer and Dallimer, 2021; Acosta et al., 2024) have explored the relationship 35
between. Land Surface Temperature (LST) and the distribution of impervious urban surfaces and green areas. However, there 36
is limited research focusing on cities in Malawi. Therefore, this study focused on the four major cities of Malawi—Blantyre, 37
Mzuzu, Zomba, and Lilongwe. The primary aim was to evaluate the effects of reduced urban green spaces on Urban Heat 38
Islands in these cities, covering the period from 2000 to 2023. 39
40
Study area41
The cities of Mzuzu, Lilongwe, Blantyre, and Zomba are located in the northern, central, southern, and eastern regions of 42
Malawi, respectively. They serve as the main urban and commercial canters of the country. Lilongwe is the capital city, Zomba 43
is the former capital, Blantyre is the business hub, and Mzuzu is the primary city in the northern region. The geographical 44
coordinates for these cities are as follows: Mzuzu is situated at 11.4390° S, 34.0084° E; Lilongwe at 13.9865° S, 33.7681° E; 45
Blantyre at 15.7667° S, 35.0168° E; and Zomba at 15.3766° S, 35.3357° E 46
2.0 Research design and methods 47
2.1 Land cover land use classification 48
The study involved land use and land cover classification using Random Forest supervised classification to quantify the land 49
changes from the year 2000 through 2023. Accuracy assessment of the classified images was assessed using Kappa coefficient 50
and Overall accuracy. The results of the accuracy assessment are shown Table1 below.51
52
53
54
Zomba
Lilongwe
Blantyre
Mzuzu
Kappa
Overall
Kappa
Overall
Kappa
Overall
Kappa
Overall
2000
0.87
0.79
0.85
0.82
0.87
0.83
0.86
0.81
2010
0.83
0.81
0.83
0.89
0.88
0.79
0.92
0.79
2023
0.94
0.95
0.91
0.90
0.93
0.90
0.87
0.90
[2]
2.2 Land surface Temperature 55
LST is largely affected by land surface emissivity that is related to NDVI. Therefore, NDVI-based emissivity method was 56
applied in this study for extracting LST from Band 6 of Landsat TM 5 and Band 10 of Landsat 8 OLI, using the following 57
process: 58
Equation1. Calculation of top of the atmosphere (TOA) radiance: For TM 5 thermal band, the digital number (DN) of each 59
pixel was converted to spectral radiance using Eq. 1 (NASA, 2010). 60
61
Where L is TOA spectral radiance (W/m2 sr m), Lmax is spectral radiance scales to QCalmax; Lmin is spectral radiance scales 62
to QCalmin; QCalmax is the maximum quantized calibrated pixel value (typically=255); QCalmin is the minimum quantized 63
calibrated pixel value (typically=1); and QCal is quantized calibrated DN. 64
Equation 2 was employed to obtain TOA radiance for Landsat 8 OLI based on the radiance rescaling factors in Metadata file. 65 66
Where L is TOA spectral radiance (W/m2 sr m), ML is band-specified multiplicative rescaling factor; QCal is quantized 67
calibrated DN; and AL is band-specified additive rescaling factors(Zhao et al., 2016). 68
Equation 3. Calculation of brightness temperature (TB ): TOA radiance was transformed to brightness temperature with 69
Equation. 70
71
Where, TB is brightness temperature (K); L is top of atmospheric radiance (W/m2 sr/m); K1 and K2 are Calibration 72
constants (W/m2sr m) which can be identified using metadata file associated with satellite images; and -273.15 is used for 73
converting brightness temperature from Kelvin to Celsius. 74
Equation 3. Calculation of NDVI: NDVI was calculated based on the reflectance values of visible red (ρRed and near infrared 75
(ρNIR ) bands obtained by atmospheric correction according to Eq. 4 (Sharifi and Lehmann, 2014; Reisi et al., 2019) 76
77
Calculation of NDVI is necessary for calculating fractional vegetation cover (Pv) and emissivity (). 78
Equation 4. Calculation of fractional vegetative cover (Pv); Pv value was calculated using NDVImax and NDVImin. 79
80
Equation6. Calculation of Land surface emissivity (LSE or ): An NDVI thresholds method was applied for LSE estimation 81
from Landsat 8 OL Image using equation 6 (Oluseyi et al., 2015; Ali, Patnaik and Madguni, 2017); 82
83
Where, w, v and s are water, vegetation C is surface roughness considered as 0 for a flat surface. For calculating emissivity 84
in Landsat 7 four cases were considered, 85
86
Equation8. Calculation of LST: equation (9) was used to convert brightness temperature to land surface temperature. 87
88
[3]
Where, is emitted radiance wavelength (11.45 m for Landsat 5 and 10.895 m for Landsat 8 OLI); is equal to 0.01438 89
mK and is produced using , in which “h” is the Planck’s Constant , “c” is the velocity of light 90 , and “b” is Boltzmann constant , and is surface emissivity. And UHI is the urban 91
heat islands, SD is the standard deviation and LSTm is the mean of the Land surface Temperature (Ouali et al., 2018).92
3.0 Results93
3.1 Land use and land cover classification 94
Over the past 23 years, Mzuzu City has experienced a substantial 72% decrease in forest cover and a 53% decline in vegetation. 95
Concurrently, there has been a 195% increase in bare land and a 135% rise in built-up areas. Spearman's rank correlation 96
indicates a highly significant negative correlation for both forest cover and vegetation (R = -1.00). Similarly, in the same 97
period, Lilongwe City has witnessed a 25% reduction in forest cover and a 33% decrease in vegetation. Bare land and built -98
up areas have expanded by 37% and 25%, respectively, with Spearman's rank correlation also demonstrating a highly 99
significant negative association for both forest cover and vegetation (R = -1.00).100
101
102
103
104
105
106
. 107
108
109
110
111
112
113
114
115
116
117
Figure 1. Showing land cover land use classification for Mzuzu City (a), Lilongwe City (b), Blantyre City (c) 118
and Zomba City (d) for the years 200 0 and 2023119
Over the past 23 years, Lilongwe City has undergone a 35% reduction in forest cover and a 29% decrease in vegetation. During 120
the same period, bare land and built-up areas have increased by 37% and 30%, respectively. Spearman's rank correlation 121
demonstrates a highly significant negative relationship for both forest cover and vegetation (R = -1.00). Similarly, over the 122
last 23 years, Zomba City has experienced a 57% reduction in forest cover and an 11% decrease in vegetation. Concurrently, 123
bare land and built-up areas have expanded by 111% and 212%, respectively. Spearman's rank correlation also indicates a 124
highly significant negative relationship for both forest cover and vegetation (R = -1.00).125
3.2 Land surface Temperature 126
The results indicate that minimum temperatures in Blantyre, Lilongwe, Zomba, and Mzuzu have increased by 24.1%, 17.9%, 127
28.5%, and 33.6%, respectively, from the year 2000 to 2023. Similarly, maximum temperatures have risen by 23.1%, 32.2%, 128
31.8%, and 29.9% in these cities over the same period, as detailed in Table 2 below 129
130
131
Yr:2023
(a)
(b)
(c)
(d)
Yr:2000
Yr:2000
Yr:2000
Yr:2000
Yr:2023
Yr:2023
Yr:2023
[4]
Table2. Minimum and maximum Land surface temperatures 132
133
134
. 135
136
137
138
139
140
141
142
143
144
145
146
147
148
Figure2: Showing minim and maximum Land surface temperatures for the years 2000 and 2023 for Blantyre 149
(a), Lilongwe (b), Mzuzu (c) and Zomba (d) cities. 150
3.3 Regression analysis for predicted vs observed LST 151
The correlation matrix reveals significant negative correlations between land surface temperatures (LSTs) and 152
forest/vegetation cover, suggesting that reductions in these covers correspond with increasing LSTs. In contrast, there are 153
strong positive correlations between LSTs and both bare land and built-up areas, indicating that expansions in these land cover 154
types are associated with rising LSTs. 155
156
157
Figure3. Showing correlation and regression analysis results for the City of Blantyre 158
Blantyre City
Lilongwe City
Zomba City
Mzuzu City
Year
Min
LST
(°C)
Max LST
(°C)
Min
LST
(°C)
Max LST
(°C)
Min
LST
(°C)
Max LST
(°C)
Min
LST
(°C)
Max LST
(°C)
2000
17.4
31.2
18.4
35.4
15.4
26.4
14.6
26.4
2010
18.4
34.2
19.9
38.9
16.9
28.9
17.4
30.9
2023
21.6
38.4
21.7
46.8
19.8
34.8
19.5
34.3
LST increase
4.2
7.2
3.3
11.4
4.4
8.4
4.9
7.9
(a)
(b)
(c)
(d)
Yr:2000
Yr:2000
Yr:2000
Yr:2000
Yr:2023
Yr:2023
Yr:2023
Yr:2023
[5]
Figure4. Showing correlation and regression analysis results for Mzuzu. Zomba and Lilongwe Cities159
4.0 Discussion 160
The analysis of land cover changes and land surface temperatures (LSTs) across Mzuzu, Lilongwe, Blantyre, and Zomba cities 161
over the past 23 years reveals significant environmental transformations. Notably, there has been a substantial decrease in 162
forest cover and vegetation, coupled with an increase in bare land and built-up areas. These changes are consistently reflected 163
in all four cities, although the extent varies. In Mzuzu City, forest cover has decreased by 72%, and vegetation has declined 164
by 53%, while bare land and built-up areas have increased by 195% and 135%, respectively. Similarly, Lilongwe City has 165
experienced a 25% reduction in forest cover and a 33% decrease in vegetation, with bare land and built-up areas expanding by 166
37% and 25%. In Blantyre City, there has been a 35% reduction in forest cover and a 29% decrease in vegetation, alongside a 167
37% increase in bare land and a 30% increase in built-up areas. Zomba City has seen a 57% reduction in forest cover and an 168
11% decrease in vegetation, with bare land and built-up areas increasing by 111% and 212%, respectively. 169
These land cover changes are accompanied by significant increases in both minimum and maximum LSTs. For example, from 170
2000 to 2023, the minimum LST in Mzuzu City increased by 33.6%, while the maximum LST rose by 29.9%. Similar trends 171
are observed in the other cities, with Blantyre, Lilongwe, and Zomba experiencing notable increases in both minimum and 172
maximum LSTs. The results are in agreement with a research done by Ouali et al., (2018) and Igun and Williams, (2018) who 173
found direct link between the deterioration of forest covers and increase in the urban heat, creating urban heat islands. The 174
correlation analysis supports these observations, revealing highly significant negative correlations between LSTs and 175
forest/vegetation cover (R = -1.00). Conversely, there are strong positive correlations between LSTs and both bare land and 176
built-up areas. These findings indicate that reductions in forest and vegetation cover are strongly associated with rising LSTs, 177
while increases in bare land and built-up areas correspond with higher LSTs. 178
5.0 Conclusion and recommendation 179
The results highlight a clear and concerning trend of urban expansion and deforestation in Malawi's major cities, 180
leading to increased bare land and built-up areas. This urbanization process has contributed significantly to rising 181
[6]
LSTs, indicating a trend toward urban heat island effects. The strong negative correlations between LSTs and 182
forest/vegetation cover underscore the critical role of these natural covers in moderating temperatures. 183
Limitations 184
The spatial resolution of the satellite images could affect the results and also the results provides a snapshot in 185
time and does not take into account any changes which might happen after the research 186
Conflict of interest 187
The author declares that there is not conflict of interest as all related work has been duly cited and acknowledged 188
Funding 189
The research did not receive any form of funding 190
References 191
Abulibdeh, A. (2021) ‘Analysis of urban heat island characteristics and mitigation strategies for eight arid and semi-arid gulf region cities’, 192
Environmental Earth Sciences, 80(7), pp. 1–26. Available at: https://doi.org/10.1007/s12665-021-09540-7. 193
Acosta, M.P., Vahdatikhaki, F., Santos, J., Jarro, S.P. and Dorée, A.G. (2024) ‘Data-driven analysis of Urban Heat Island phenomenon based 194
on street typology’, Sustainable Cities and Society, 101(December 2023). Available at: https://doi.org/10.1016/j.scs.2023.105170. 195
Ali, S.B., Patnaik, S. and Madguni, O. (2017) ‘Microclimate land surface temperatures across urban land use/ land cover forms’, Global 196
Journal of Environmental Science and Management, 3(3), pp. 231–242. Available at: https://doi.org/10.22034/gjesm.2017.03.03.001. 197
Batty, M., Xie, Y. and Sun, Z. (1999) ‘Modeling urban dynamics through GIS-based cellular automata’, Computers, Environment and Urban 198
Systems, 23(3), pp. 205–233. Available at: https://doi.org/10.1016/S0198-9715(99)00015-0. 199
Gago, E.J., Roldan, J., Pacheco-Torres, R. and Ordóñez, J. (2013) ‘The city and urban heat islands: A review of strategies to mitigate adverse 200
effects’, Renewable and Sustainable Energy Reviews, 25, pp. 749–758. Available at: https://doi.org/10.1016/j.rser.2013.05.057. 201
Gee, O.K. and Sarker, M.L.R. (2013) ‘Monitoring the effects of land use/landcover changes on urban heat island’, Earth Resources and 202
Environmental Remote Sensing/GIS Applications IV, 8893(May), p. 889304. Available at: https://doi.org/10.1117/12.2029035. 203
Gill, S.E., Handley, J.F., Ennos, A.R. and Pauleit, S. (2018) ‘Adapting cities for climate change: The role of the green infrastructure’, Planning 204
for Climate Change: A Reader in Green Infrastructure and Sustainable Design for Resilient Cities, pp. 195–205. 205
Igun, E. and Williams, M. (2018) ‘Impact of urban land cover change on land surface temperature’, Global Journal of Environmental Science 206
and Management, 4(1), pp. 47–58. Available at: https://doi.org/10.22034/gjesm.2018.04.01.005. 207
Jabbar, H.K., Hamoodi, M.N. and Al-Hameedawi, A.N. (2023) ‘Urban heat islands: a review of contributing factors, effects and data’, IOP 208
Conference Series: Earth and Environmental Science, 1129(1), pp. 0–9. Available at: https://doi.org/10.1088/1755-1315/1129/1/012038. 209
Lee, K., Kim, Y., Sung, H.C., Kim, S.H. and Jeon, S.W. (2021) ‘Changes in Surface Urban Heat Island Effect with the Development of New 210
Towns’. Available at: https://www.researchsquare.com/article/rs-515349/latest. 211
Li, X., Stringer, L.C. and Dallimer, M. (2021) ‘The spatial and temporal characteristics of urban heat island intensity: Implications for east 212
africa’s urban development’, Climate, 9(4). Available at: https://doi.org/10.3390/cli9040051. 213
Oluseyi, I.O., Fanan, U. and Magaji, J.Y. (2015) ‘An evaluation of the effect of land use / cover change on the surface temperature of Lokoja 214
town , Nigeria An evaluation of the effect of land use / cover change on the surface temperature of Lokoja town , Nigeria’, 3 (October), pp. 215
86–90. Available at: https://doi.org/10.5897/AJEST09.014. 216
Ouali, K., El Harrouni, K., Abidi, M.L. and Diab, Y. (2018) ‘The Urban Heat Island phenomenon modelling and analysis as an adaptation of 217
Maghreb cities to climate change’, MATEC Web of Conferences, 149, p. 02090. Available at: 218
https://doi.org/10.1051/matecconf/201814902090. 219
Philip, V.V. (2022) ‘MITIGATION PLANNING FOR URBAN HEAT ISLAND’, 7(2), pp. 339–342. 220
Reisi, M., Ahmadi Nadoushan, M. and Aye, L. (2019) ‘Remote sensing for urban heat and cool islands evaluation in semi-arid areas’, Global 221
Journal of Environmental Science and Management, 5(3), pp. 319–330. Available at: https://doi.org/10.22034/gjesm.2019.03.05. 222
Sharifi, E. and Lehmann, S. (2014) ‘Comparative analysis of surface urban heat island effect in central sydney’, Journal of Sustainable 223
Development, 7(3), pp. 23–34. Available at: https://doi.org/10.5539/jsd.v7n3p23. 224
Tiangco, M., Lagmay, A.M.F. and Argete, J. (2008) ‘ASTER-based study of the night-time urban heat island effect in Metro Manila’, 225
International Journal of Remote Sensing, 29(10), pp. 2799–2818. Available at: https://doi.org/10.1080/01431160701408360. 226
You, M., Huang, J. and Guan, C.H. (2023) ‘Are New Towns Prone to Urban Heat Island Effect? Implications for Planning Form and Function’, 227
Sustainable Cities and Society, 99(June), p. 104939. Available at: https://doi.org/10.1016/j.scs.2023.104939. 228
Zhao, M., Cai, H., Qiao, Z. and Xu, X. (2016) ‘Influence of urban expansion on the urban heat island effect in Shanghai’, International Journal 229
of Geographical Information Science, 30(12), pp. 2421–2441. Available at: https://doi.org/10.1080/13658816.2016.1178389. 230
