Analysis of the cooling effects of urban green spaces in mitigating micro-climate change using geospatial techniques in Adama City, Ethiopia

ABSTRACT

Greening the urban environment serves as an effective strategy to counteract the impacts of localized climate variations, such as temperature fluctuations and altered precipitation patterns. The main objective of this study is to examine how urban green spaces (UGS) in Adama City contribute to cooling the surrounding areas by using geospatial methods through considering land surface temperature and vegetation cover, thereby addressing micro-climate changes. Three different remotely sensed data of Landsat7 enhanced thematic mapper plus (ETM+) of the year 2000 and 2013 as well as Landsat8 operational land imagery/thermal infrared sensor (OLI/TIRS) (2023) were used in the study. The consistent land surface temperature data were retrieved from Landsat7 ETM+ and Landsat8 OLI/TIRS using mono window and split window algorithms, respectively. Results showed that the proportion of urban green spaces to other land use/land cover was reduced from 29.3 km² (21.20%) in 2000 to 18.17 km² (13.15%) in 2023. Due to the increment of built-up areas and reduction of vegetation cover, the land surface temperature of the city is increasing. The analysis determined that an optimal green space area of 5.5 ± 0.5 hectares in Adama City can effectively reduce surface temperatures by approximately 2.85 degrees Celsius. This study will contribute to understanding the role of vegetation cover in reducing urban heat effect, and also assist policymakers in regard to urban land use planning.

KEYWORDS:

• Cooling efficiency
• micro-climate
• urban green spaces

1. Introduction

Earth’s surface temperature is a product of the balance between incoming solar energy and outgoing radiation energy (Roza et al., 2017). The warmer the earth gets, the more the energy it radiates out. Earth is being experiencing a warmer atmosphere since the pre-industrial era and contributed to a significant increase in the global mean temperature (IPCC, 2014). Urban heat island (UHI) is a result of rapid urbanization which is described as urban areas with significantly warmer temperature than its nearby rural areas (Kong et al., 2016). There are a number of contributing factors which play significant role in the formation of UHI; for instance, low albedo materials, air pollutant, wind blocking, clearance of trees and increased use of air conditioner (Nuruzzaman, 2015).

Fast urbanization leads to reduction of vegetation areas; and increases land surface temperature and consequently changes urban micro-climate (Nor, 2013). Increased replacement of natural green areas to urbanized areas, diminishing of agricultural lands, expansion of impervious surfaces, extensions of barren land because of the built-up areas have led to significant changes in the local climate conditions. Due to all these factors temperature distribution in urban areas is expressively warmer than its surrounding sub-urban areas (Effat et al., 2014; Senanayake et al., 2013).

As one of the basic elements of the urban environment, urban green space is the only type of land use with natural or semi-natural conditions inside a city; and plays a significant role in protection of the ecological environment of cities (Ngom et al., 2016; Zhong et al., 2011). Vegetation is a vital element of global environment. It modifies the ecosystem through water preservation, terrestrial soil constancy and atmospheric circulation. It also helps to sustain a balance of ecosystem prominently. Urban greenery also acts as a natural agent against air pollution in the urban environment (Buyadi et al., 2015).

Urban green infrastructure offers numerous advantages beyond just decreasing land surface temperature. Urban green infrastructure plays a crucial role in climate change adaptation by mitigating the urban heat island effect, reducing energy consumption for cooling, and providing shade and cooling effects in built-up areas (Ngom et al., 2016). These adaptations help cities become more resilient to extreme heat events associated with climate change. Green infrastructures also help to filter pollutants from the air, resulting in improved air quality. Trees absorb carbon dioxide and other harmful gases while releasing oxygen, contributing to a healthier urban environment (Teferi & Abraha, 2017).

Trees and green spaces contribute considerably to the improvement of the urban climate and to UHI mitigation. Decrease of the temperature is achieved through trees that provide solar protection, affect air movements and heat exchange, absorb solar radiation and cool the air through evapotranspiration processes. It should be noted that urban parks may extend their cooling potential and decrease ambient temperatures in adjacent urban zones depending on the thermal balance of the overall area under study. Urban parks provide thermal comfort and a high mitigation potential (Cohen et al., 2013). Because of transpiration, greenery plays a significant role in alleviating UHIs by dropping temperature and increasing humidity. Their cooling effects are especially important; and they have been regarded as natural resources for city planning (Sandra et al., 2011).

Geospatial techniques offer a comprehensive and efficient means of assessing the cooling effects of green spaces by providing precise spatial data on land surface temperature and vegetation cover. By integrating remote sensing and geographic information system (GIS) technologies, these methods enable large-scale analysis, temporal monitoring, and visualization of the impact of green spaces on urban micro-climates, facilitating evidence-based decision-making for sustainable urban planning and climate resilience strategies (Feyisa et al., 2014; Merga et al., 2022).

Previous studies on the role of green spaces in mitigating urban heat islands have been limited by inadequate methodologies, making it challenging to accurately assess the specific contributions of individual green space patches. While some research has highlighted the significant cooling effects of urban parks, there is difficulty in generalizing findings due to varying methodologies and limited consideration of other types of green spaces, such as roadside vegetation and greenery around non-religious institutions. For example, studies by Feyisa et al. (2014), Teferi and Abraha (2017), and Samson et al. (2018) utilized quadrant division methods focusing primarily on selected parks, neglecting the broader spectrum of green spaces and their respective impacts.

Little is known about the quantitative role of green spaces is in mitigating micro-climate change in the study area which needs detail investigation to make specific recommendations for urban land use optimization, landscape planning and design to mitigate the heating effects. This matter necessitates assessing the cooling effect of green space on land surface temperature (LST) using geospatial techniques. Findings from the study will provide an understanding of the LST, urban heat island (UHI) and LU/LC status of the area as an input for planning and decision-making.

2. Materials and methods

2.1. The study area

This study was conducted in Adama City, which is located in the Great Ethiopian Rift Valley; and found at about 100 Km Southeast of Addis Ababa, Ethiopia. Its absolute location stretches between 8° 27′00″to 8°36′00″North latitude and 39° 12′30″ to 39° 20′ 30″ East longitude and covering a total area of 138.2 Km² (Figure 1).

Figure 1. Location map of Adama City.

Adama city has hot and dry weather for the greater part of winter and warm and sunny in summer. It lies in Great Rift Valley of East Africa; and the altitude of the central part of the city constitutes the lowest area.

The climate of Adama City comes under a sub-tropical agro-climatic zone. The city has hot and dry weather for the greater part of winter and warm and sunny in summer. There are four climatic seasons; Kremt (rainy period) Bega (dry period) Belg (small rains) and Meher (a spell between the long and small rain periods). The city experiences the hottest and coldest temperature during May and December respectively. On the average, rainy season is June, July, August and September; whereas, the dry month is in January, October, November and December. Among all months, the driest month is December. Adama City experiences the hottest and coldest temperature during May and December respectively (Central Statistical Authority CSA Adama Branch, 2009).

2.2. Data types and sources

Data for the study was collected from both primary and secondary sources (Table 1). The primary data source includes observations and fieldwork. The observation was made to collect ground truth points for validation of LU/LC types as well and different photographs were taken for further identification of each land use/land cover of the study area. These were carried out in order to identify the land use land cover types such as built-up, open spaces, water bodies, and the spatial distribution of green spaces. Secondary data sources include reviewing different relevant literature of the specific study area and related studies.

Table 1. Data types and sources

No.	Data category	Source	Purpose
1	Landsat images	USGS	For LU/LCC, LST, NDVI and, NDBI analysis
2	Temperature data	National meteorological agency (NMA), Ethiopia	To validate the value of temperature retrieved from satellite image
3	Adama city shape file	Adama city municipality	For preparation of result map

Two sets of remotely sensed data such as Landsat7 ETM+ of the year 2000 and 2013 and Landsat8 OLI/TIRS of the year 2023 were used for this study (Table 2).

Table 2. Remote sensing images

Image	Path	Raw	Pixel size	Acquisition date	Sources
Landsat ETM+ of 2000	167	54	30 × 30	11, Feb, 2000	U.S. Geological Survey
Landsat OLI/TIRS of 2013 & 2023	167	54	30 × 30	05, Jan, 2023	U.S. Geological Survey

2.3. Data processing and analysis methods

2.3.1. Land use/land cover classification

Image pre-processing which encompasses layer stacking; image resampling and false color combination was employed using ERDAS IMAGINE V15 software. A supervised classification approach was applied to classify the years 2000, 2013, and 2023 images with a maximum likelihood classifier algorithm (MLC). MLC is one of the eminent parametric classifiers used for supervised image classification (Foody et al., 1992).

Post-classification smoothing has also been done on the land use/land cover (LU/LC) to remove noisy pixels. Sample data from the field using hand-held global positioning system (GPS) and visual interpretation of the selected land use/land cover type were prepared, and then land cover polygons were made using ArcGIS 10.8 to extract and reclassify the urban green spaces. Finally, urban green land cover classes were identified and mapped for further analysis.

Finally, an accuracy assessment was performed to evaluate the quality of classification output. Error matrix based on assessment of the overall accuracy; producer’s accuracy, user’s accuracy, and kappa coefficient were utilized to evaluate the pixel-based classification output for LU/LC classification. The purpose of accuracy assessment is to quantitatively assess how efficiently the pixels were tested into the correct land cover classes (Bhatta, 2008).

2.3.2. Retrieval of Land Surface Temperature (LST)

Near-surface air temperature is a consequence of the complex effects of the turbulent heat transports produced by nearby heated surfaces (Avdan & Jovanovska, 2016). The Mono window algorithm for Landsat 7 and the split window algorithm for Landsat 8 were employed.

Mono window algorithm is an algorithm that possesses the benefits of easy determination of parameters, extensive applications (Yu et al., 2014), and high precision reversal which could accurately reflect regional land surface heat distribution. The specific formulae are:

(1) $\begin{array}{l}\text{Ts=}\{\text{a6}\left(\text{1-C6-D6}\right)\text{+(b6(1-C6-}\\ \text{D6)+C6+D6)T6-D6Ta}\}\text{/C6}\end{array}$(1)

$\mathrm{C}6\mathit{\hspace{0.17em}}=6\mathit{\tau }6$

$\mathrm{D}6\mathit{\hspace{0.17em}}=\mathit{\hspace{0.17em}}\left(1-\mathit{\hspace{0.17em}}\mathit{\epsilon }6\right)\mathit{\hspace{0.17em}}\left[1+\mathit{\hspace{0.17em}}\left(1-\mathit{\hspace{0.17em}}\mathit{\epsilon }6\right)\mathit{\hspace{0.17em}}\mathit{\tau }6\right]$

Where; T6 is brightness temperature, ε6 is land surface emissivity, τ is atmospheric tan-emissivity in thermal infrared and Ta is an average atmospheric temperature which can be calculated by the parameter estimation method of the mono-window algorithm. Also, A6 = −67.355351, b6 = 0.458606, C6 and D6 are intermediate variables and Ts is the Land Surface Temperature (LST) to be calculated.

Split-window algorithm is used to retrieve land surface temperature from Landsat 8 data that has two bands (Band 10: 10.6–11.2 μm, Band 11: 11.5–12.5 μm). Split-window algorithm uses the brightness temperature of the two bands of Thermal Infrared (TIR), mean, and difference in land surface emissivity for estimating LST. The process of acquiring LST values follows the conversion of thermal infrared Digital Numbers (DNs) (Bands 10 and 11) to radiance Top of Atmosphere (TOA) and at-satellite brightness temperature. The effective at-sensor Brightness Temperature (BT) is also known as black body temperature which was obtained from the spectral radiance using Plank’s inverse function. Spectral radiance values for bands 10 and 11 were converted to radiant surface temperature under an assumption of uniform emissivity using pre-launch calibration constants for the Landsat 8 OLI sensor. After spectral radiance is converted to radiance, the raw digital numbers of the thermal bands are converted to Top of Atmosphere (TOA) brightness temperatures, which are the effective temperature viewed by the satellite under an assumption of emissivity using Planck’s equation (Sobrino et al., 2012).

(2) $\mathrm{LST}=\mathrm{TB}10+\mathrm{C}1(\mathrm{TB}10-\mathrm{TB}11)+\mathrm{C}2(\mathrm{TB}10-\mathrm{TB}11)\mathit{\hspace{0.17em}}+\mathrm{Co}\mathit{\hspace{0.17em}}+\mathit{\hspace{0.17em}}\left(\mathrm{C}3\mathit{\hspace{0.17em}}+\mathit{\hspace{0.17em}}\mathrm{C}4\mathrm{W}\right)\mathit{\hspace{0.17em}}\left(1-\mathrm{\Delta \epsilon }\right)\mathit{\hspace{0.17em}}+\mathit{\hspace{0.17em}}\left(\mathrm{C}5\mathit{\hspace{0.17em}}+\mathit{\hspace{0.17em}}\mathrm{C}6\mathrm{W}\right)\mathit{\hspace{0.17em}}\mathrm{\Delta \epsilon }$(2)

Where; C0, C1, C2, C3, C4, C5 and C6 is the split-window coefficients; TB10 = brightness temperature of band 10 (Kelvin K); TB11 = brightness temperature of band 11 (Kelvin K); $\mathit{\epsilon }$ is mean value of land surface emissivity (LSE) of TIR bands; W content of water vapors in the atmosphere; ∆$\mathit{\epsilon }$ = difference between LSE of bands 10 and 11 emissivity, difference of land surface emissivity, and then to estimate LST;

(3) $\mathrm{BT}=\frac{\mathit{K}2}{\mathit{ln}\left(\frac{\mathit{K}1}{\mathit{L\lambda }}+1\right)}$(3)

Where; BT is effective at-sensor brightness temperature (K); K2 is calibration constant 2 (K); K1 is calibration constant 1 (W/(m² *sr * μm)); Lλ is spectral radiance at sensors aperture (W/(m2 * sr * μm)); and Ln is a natural logarithm.

To compute Land Surface Emissivity (LSE), it is essential to know the characteristics of the earth’s surface and change the thermal radiance energy during calculation of LST (Sobrino et al., 2012). The emissivity is calculated using;

(4) $\mathit{\epsilon }\mathit{\hspace{0.17em}}=\mathit{\hspace{0.17em}}0.004{\mathit{\hspace{0.17em}}}^{\ast }\mathrm{PV}\mathit{\hspace{0.17em}}+\mathit{\hspace{0.17em}}0.986$(4)

Where; PV is Vegetation Proportion which is obtained using the following formula;

(5) $\mathrm{PV}={\left[\frac{\mathrm{NDVI}-\mathrm{NDVImin}}{\mathrm{NDVImax}-\mathrm{NDVImin}}\right]}^2$(5)

Where; NDVI is normalized difference vegetation index; NDVImin is a minimum value of normalized difference vegetation index and NDVImax is a maximum value of normalized difference vegetation index.

The calculated radiant surface temperature is then corrected for emissivity using the the following equation;

(6) $\mathrm{LST}=\mathit{\hspace{0.17em}}\frac{\mathrm{TB}}{1\mathit{\hspace{0.17em}}\left(\mathit{\lambda }\mathrm{TB}/\mathrm{P}\right)\mathrm{In}\mathit{\epsilon }}$(6)

Where; LST is Land Surface Temperature, TB is radiant surface temperature in (K), λ is the wave-length of emitted radiance (11.5 μm), ρ is Planck’s constant (6.26 × 10⁻³⁴ J s); and ε is land surface emissivity.

2.3.3. Extraction of built-up index

The normalized difference built-up index (NDBI) is often mixed with plant noise, and its values range from −1 to 1. The greater the NDBI is, the higher the proportion of built-up land is (Varshney, 2013). NDBI is derived from Landsat ETM and Landsat 8, from reflectance measurements in the red and mid-infrared (MIR) portion of the spectrum. The NDBI value is obtained using the following equation and applied to identify urban built area. Extraction NDBI is calculated using the following equation.

(7) $\mathrm{NDBI}=\frac{\mathrm{MIR}-\mathrm{NIR}}{\mathrm{MIR}+\mathrm{NIR}}$(7)

Where; MIR is Md Infrared (Band 6 for Landsat 8) and NIR is Near Infrared (Band 5 for Landsat 8) (Zhou et al., 2014).

2.3.4. Extraction of normalized difference vegetation index (NDVI)

The spatial distribution of urban green spaces in Adama city was extracted and analyzed on the basis of the value of NDVI. The values of NDVI generally range between −1 to + 1. The −1 value depicts the absence of vegetation and + 1 value shows the presence and density of vegetation. Moreover, NDVI values are suitable for the calculation of change detection analysis. Very low value of NDVI (0.1 and below) correspond to barren areas of rock, sand, or snow (Pervaiz et al., 2018).

The values of NDVI can further categorized as non-vegetated for the value < 0, unhealthy vegetation for the value between 0.02–0.03 and moderate values represent bush and grasses (0.2 to 0.3), whereas, high NDVI value correspond to dense vegetation (0.6 to 0.8) (Gandhi et al., 2015). These activities was computed and analyzed in ArcGIS Software environment using the following formula.

(8) $\mathrm{NDVI}=\mathit{\hspace{0.17em}}\frac{\mathrm{NIR}-\mathrm{R}}{\mathrm{NIR}+\mathrm{R}}$(8)

Where; NIR is a Near Infrared (the pixel Digital Number (DN) of TM Band 4 and Band 5 for Landsat 7 and 8, respectively); and R is Red Band (Band 3 and Band 4 for Landsat7 and 8, respectively) (Gandhi et al., 2015).

2.3.5. Determination of green spaces’ cooling effect

To determine the contribution of green spaces in relation to other LULC types in mitigating UHI, each green space patches was extracted with corresponding LST values for the identified years. In consideration of the possible varying temperature conditions during LST acquisition, the temperature means of green spaces in relation to the entire landscape for the identified years was computed separately. With the mean temperature within the study area as baseline, the thermal influence of each of the green space patches to the entire urban landscape was computed by the contribution index (CI) (X. L. Chen et al., 2006) using function:

(9) $\mathrm{CI}\mathit{\hspace{0.17em}}=\mathit{\hspace{0.17em}}{\mathrm{D}}^{\ast }\mathrm{St}$(9)

Where: CI is Contribution Index i.e. the influence of each green space patch to the entire landscape, D is the difference in mean temperature between the green space patches and the entire urban landscape; and St is the proportion of the area to the entire landscape.

The Green Index (GI) percentage of green space patches was calculated based on binary classification (green and non-green classes) of NDVI measurements. The negative values of NDVI measurement is classified as built up area and positive values are classified as green class (Kshama, 2012).

Green space cooling effect is determined as the difference in temperature between inside the green space and the average land surface temperature of the entire landscape (Choi et al., 2012). To determine the cooling effects of green spaces, surface temperature of green spaces and the entire land scape should be computed using the following equation.

(10) $\mathrm{\Delta T}\mathit{\hspace{0.17em}}=\mathit{\hspace{0.17em}}\left({\mathrm{T}}_{\mathrm{u}}-\mathit{\hspace{0.17em}}{\mathrm{T}}_{\mathrm{p}}\right)$(10)

Where; ΔT is Change in Land Surface Temperature, T_p is the average LST of green space; and T_u is the average LST of the entire landscape.

2.3.6. Software used for the study

For the achievement of the study’s purpose, different software required is stated as table below (Table 3). General workflow for this study is presented in Figure 2.

Figure 2. Methodological flow chart.

Table 3. Software used for the study

No	Software	Purpose
1	ArcGIS 10.3	Analyzing and visualization of spatial data
2	ERDAS 2015	For Image classification and accuracy assessment
3	Google Earth Pro & GPS	For LULC accuracy assessment

3. Results and discussion

3.1. Accuracy assessment

Accuracy assessment was done from land use land cover types of 2000, 2013 and 2023 for the analysis of cooling effects of green spaces in mitigating micro-climate change in Adama City (Table 4).

Table 4. Accuracy assessment of land use and land cover for 2000, 2013 and 2023

LU/LC class	2000		2013		2023
	Producers accuracy	Users accuracy	Producers accuracy	Users accuracy	Producers accuracy	Users accuracy
	(%)	(%)	(%)	(%)	(%)	(%)
Dense veg.	94.5	98.3	99.3	99.8	98.6	99.8
Built-up	98.8	95.9	98.9	97.4	97.3	98.22
Open space	100	99.4	100	99.4	100	98.4
Sparse veg.	99.5	99.5	43.2	97.2	92.6	98
Water bodies	98.6	99.8	99.3	98.3	94.5	99.8
Overall accuracy	98.4%		97.6%		98.7%
Kappa coefficient	0.96		0.97		0.97

3.2. Land use/land cover (LU/LC) map of the period 2000 to 2023

The LU/LC of the year 2000 shows that the city had a built-up area with a total area of 31.75 Km² (22.97%) and dense and sparse vegetation cover with a total area of 102.26 Km² (74%) respectively. However, these green spaces were rapidly eradicated and replaced by built-up areas. It is rapidly increasing which corresponds to the context of rapid urbanization in Adama City.

As part of developing countries’ cities, Adama City is also facing a high rate of land use/land cover change due to urban development activities. Due to the increase in population, industrialization and other natural and human activities, land use/land cover of the city is changing. The transformation of agricultural lands into urban areas has greatly affected the land components and the environment. Consequently, the natural vegetation in and around the city is also converted into impervious surfaces. Expansion led by government and private sectors mainly mass housing programs like condominiums, single residential and real state constructions, and other developmental activities such as the construction of roads and factories are visible evidence for the eradication of green spaces and rise in land surface temperature in the city. The proportion of urban green spaces (UGS) particularly, dense vegetation which includes indigenous forest and thicket were reduced from 29.3 Km²(21.20%) in 2000 to 18.17 Km² (13.15%) in 2023, whereas built-up area increased from 31.75 Km²(22.97%) to 86.0 Km² (62.24%); whereas, built-up areas occupy the dominant LU/LC classes which is the most significant change in land cover type proportion (Table 5 and Figure 3).

Figure 3. LU/LC maps of the year 2000 (A), 2013 (B) and 2023(C).

Table 5. Land Use/land cover (LU/LC) of the Year 2000, 2013 and 2023

Class name	2000		2013		2023
	Area (lkm^2)	(%)	Area (km^2)	(%)	Area (km^2)	(%)
Dense veg.	29.38	21.26	24.14	17.47	18.17	13.15
Built-up	31.75	22.97	67.13	48.57	86.0	62.24
Open Space	2.26	1.64	3.17	2.29	0.36	0.26
Sparse veg.	72.88	52.74	41.34	29.92	32.27	23.35
Water body	1.93	1.40	2.42	1.74	1.38	1.00
Total	138.2	100.00	138.2	100.00	138.2	100.00

Studies conducted by; Teferi and Abraha (2017) and Dagnachew (2018) also indicate that built-up areas consume a considerable amount of land from vegetation cover during urban development. These reveal that rapid population growth and expanding built-up areas in cities reduce the cooling effects of vegetation cover. Of all land use land cover types, bare land (rocky areas) and built-up areas are the most land cover types that experience high land surface temperature. Green spaces such as parks, trees along streets, and gardens absorb solar radiation, provide shade, and release moisture through transpiration, effectively cooling the surrounding air. This helps alleviate the urban heat island effect, where cities experience higher temperatures compared to rural areas due to human activities and built-up infrastructure. Overall, the presence of green spaces in urban areas plays a crucial role in creating more comfortable and sustainable living environments by reducing heat stress and enhancing the overall well-being (Merga et al., 2022; Moisa & Gemeda, 2021; Moisa et al., 2022).

3.2.1. Land surface temperature analysis

Land surface temperature (LST) distribution was classified into appropriate ranges and color-coded to generate a thermal pattern distribution map of the study area. The result of this study shows that different LU/LC classes have different LST values. The average surface temperature of Adama City has increased from 28.25°C in 2013 to 31.78°C in 2023 with differences of 3.53°C (Table 6). High LST values were found in bare lands (vacant spaces, uncovered areas) and built-up areas. This finding clearly shows that both bare land and built-up areas are the two major possible causes for the rise of surface temperature in the study area. LST value ranges from 30–34°C in 2000 was grew from 8.23 km² to 38.97 Km² in 2023. The high rate of activities in the manufacturing and transport sector along with power generation accelerates the mean LST.

Table 6. Summary of LST values of the year 2000, 2013 and 2023

LST Values (^0C)	Landsat ETM+ (2000)	Landsat ETM+ (2013)	Landsat OLI (2023)
Minimum	19	17.8	18.8
Maximum	36.4	38.2	39.7
Mean	27.08	28.25	31.78

The result is in agreement with (Avdan & Jovanovska, 2016; Feyisa et al., 2014; Merga et al., 2023 Samson et al., 2018) that the major factors for the increase of LST both in rural and urban areas are land use/land cover changes and unplanned use of land resources (Table 7 and Figure 4).

Figure 4. LST map of the year 2000(A) and 2013(B) and 2023(C).

Table 7. Variations of LST values from the year 2000 to 2023

2000			2013		Difference (%)
LST (^0C)	Area (km^2)	Area (%)	Area (km^2)	Area (%)	Difference (km^2)	2000–2023
<22	24.57	17.78	36.72	26.57	12.15	+10.42
22–26	64.51	46.69	32.84	23.76	−31.67	− 22.93
26–30	29.33	21.22	14.04	10.16	−15.29	− 11.06
30–34	8.23	5.95	38.97	28.2	30.74	+20.62
>34	11.54	8.35	15.63	11.31	4.09	+2.96

Results from Figure 4(a–c) showed that the mean land surface temperature of the study area has increased from 27.08°C in 2000, 28.25°C in 2013, and to 31.78°C in 2023. A great variation in surface temperature was observed in areas having dense vegetation, impervious surfaces, uncovered soil, and lowest altitude. A Northeastern part of the City, particularly, Goba Hiddi Kebele, experiences higher surface temperature with a mean LST value of 34.26 °C due to its lower altitude and bare surface. However central parts of the study area showed a decrease in land surface temperature. These variations in LST could have resulted from the existence of dense vegetation cover and its topographic setup (higher altitude).

Other studies have also shown similar results. For instance, (Melkamu & Meseret, 2019; Samson et al., 2018) pointed out that the lowest mean LST was recorded by water body and green vegetation classes. This means that areas with lower vegetation cover are experiencing higher land surface temperatures and vice versa. From this, it is clear that vegetation has a cooling and regulating effect on the surface temperature of an area. Studies showed that vegetated surfaces can contribute significantly to human comfort and better health conditions by decreasing the land surface temperature (Gémes et al., 2016; Melkamu & Meseret, 2019; Merga et al., 2024).

3.2.2. Normalized difference built-up index (NDBI)

This study clearly shows that the vegetation cover of the study area where decreased due to rapid expansion of different buildings. Therefore, built-up areas are an inducing many surface temperature variations (Figure 5). Of the total area, built-up area alone accounted for about 18.87 Km² (13.67%) in 2000; and 76.4 Km² (55.29%). Currently, the built-up area occupies the largest portion of the city. The central and Northern parts of the city are packed with man-made features which made the temperature to be higher than the surrounding regions.

Figure 5. Built-up areas of Adama City in 2000 (A), 2013 (B) and 2023 (C).

Some other studies also reveal that the sharp decrease in green spaces and rapid increase of built-up areas decline the cooling effect of green spaces (Žuvela-Aloise et al., 2016). High-temperature irregularities are comprehensively related to built-up land, densely populated areas, and greatly industrialized zones (Moisa et al., 2023; Samson et al., 2018).

3.3. Correlation of land surface temperature with built-up and green areas

These two indices were used to examine temperature variations and also determine their relationship with LST. Strong correlations between LST and both urban parameters were found in the study area. A positive correlation was identified between LST and built-up areas with R2 = 0.9276. This shows that the higher the built-up areas, the higher the LST values. The positive correlation found between the Normalized Difference Built-up Index (NDBI) and LST indicates that the built-up area is producing many LST variations and is the main contributor to urban heat islands (Figure 6). This finding is in agreement with Moisa et al. (2022) that the effect of the increment of the built-up area will result in severe effects of urban heating.

Figure 6. Relationship between LST and NDBI.

3.2.3. Normalized difference vegetation index (NDVI)

The results of NDVI of both 2000 and 2023 showed that the northeast and southwest parts of the study area have higher NDVI values. The low values of NDVI were also observed in dense residential areas with less vegetation coverage. Vegetation cover has decreased and non-vegetated area has been increasing gradually over the study period (Figure 7). The average value of NDVI of the year 2000 was reduced by half percent by the year 2023. Relatively, a high value of NDVI in 2023 is observed in the south and southwestern parts particularly in Bole Kebele whereas the northeast part i.e. and the rest parts of the study areas have low values of NDVI. By comparing the NDVI of the two different periods (2000 and 2023), it is observed that maximum NDVI values were decreased over the study period.

Figure 7. Normalized difference vegetation index for the year 2000(A), 2013(B) and 2023(C).

The study shows that 64.54 Km² (46.71%) of the study area in 2000 was covered by NDVI values greater than 0.2. But in the year 2023, only 47.84 Km² (34.62%) of the study area was covered by the same NDVI class. These confirmed that there has been a dynamic vegetation cover change from one class to the other in the study area. NDVI values between 0.06 and 0.4 represent agricultural fields in the surrounding periphery. Patches of dense vegetation cover in the southern parts of the City show relatively high NDVI values. A number of studies have also shown that NDVI values of river banks and around water bodies experience higher NDVI values than other classes, owing to the presence of agricultural land (Feyisa et al., 2014; Samson et al., 2018). The NDVI values of the different years showing that there has been marked vegetation cover change during the study period (Table 8).

Table 8. Comparisons of the NDVI values in the year 2000 and 2023

NDVI values	Landsat ETM+ (2000)	Landsat ETM+ (2013)	Landsat OLI (2023)
Min.	−0.31	−0.265	−0.128
Max.	0.62	0.516	0.456
Mean	0.35	0.28	0.124

The results indicated that the high value of NDVI was distributed in the outskirts and in southwestern parts of the city. For green areas, a negative correlation was identified with the LST of the study area, that is, R2 = −0.9286. It can be concluded that, if the area is densely vegetated, the LST is found to be lower. Healthy vegetative cover plays a key role in lowering of the surface temperature. The high value of LST was detected in built-up areas and bare faces; whereas low surface temperature was found in areas covered with vegetation. The decrease in mean patch size may increase LST because a larger, continuous green space produces stronger cool island effects than that of several small pieces of green spaces.

The analyzed Landsat images of 2000 and 2023 indicated that NDVI and LST have indirect relationships. A Low NDVI value has high LST and high NDVI values have low LST. The results of this study showed that NDVI was correlated with LST with statistical significance. This finding is consistent with studies conducted by X. L. Chen et al. (2006); and Buyadi et al. (2015) which confirmed that there is a negative correlation between LST and the richness of green spaces measured by NDVI (Figure 8).

Figure 8. Relationship between LST and NDVI.

Trees and other plants help cool the environment, making green space a simple and effective way to mitigate urban heat island effects. Therefore, the effects of the increase in patch density on LST can be explained by a decrease in the mean patch size of green spaces. Generally, a negative correlation was found between NDVI values with LST. This finding is in agreement with Zhou et al. (2014); and Feyisa et al. (2014) and Melkamu and Meseret (2019) that green spaces can lower surface and air temperatures by providing shade that prevents land surfaces from direct heating from sunlight.

3.3. Estimation of LST in the selected test spots of Adama city

From the total of fourteen (14) kebeles in Adama City, Gooba Hiddi kebele experiences extreme LST values. The mean LST of the kebele was found to be higher than other kebeles with a mean LST value of 34.26°C. The area is covered by man-made features (built-up area) and found at the lowest altitude with a large portion of bare lands. The increment of built-up area in the selected site will result in a severe effect of urban heating whereas the increment of green areas within the selected site is seen to be the most suitable measure to reduce the LST value by about 5.1°C. Estimating the land surface temperature (LST) values when both the built-up area and green spaces in the selected test spot are increased by different percentages (25%, 50%, 75%, and 100%) compared to their current levels (Table 9). Essentially, it’s exploring how changes in the extent of built-up areas and green spaces affect the temperature of the land surface in a particular test area, depicted in Figure 9.

Figure 9. Estimation of LST in the selected test spot in Adama City.

Table 9. Estimated LST value of selected test spot (Gooba Hiddii Kebele)

Increase of built-up area by	Estimated mean LST in (°C)	Increase green area by	Estimated mean LST (°C)
25%	35.02	25%	32.98
50%	35.78	50%	31.76
75%	36.54	75%	30.46
100%	37.31	100%	29.16

3.4. Cooling effects of urban green space

Variations in land surface temperature and cooling efficiency of green spaces in Adama City have been evaluated. Areas of green space patches are in a logarithmic relationship with the maximum temperature difference with the coefficient correlation of (R2 = 0.6922). This demonstrates that the areas of green spaces are highly associated with the cooling range and maximum surface temperature difference. Based on the statistics of the cooling range appropriate to the areas of the green spaces, a regression analysis of the area with the cooling extent and maximum surface temperature difference respectively was made. The range of 5.5 ± 0.5 ha can bring a change in surface temperature by about 2.85°C. This indicates that the bigger the size of green space, the higher the cooling efficiency of green space patches. The inclination of the model curve is very steep on the left-hand side which means that the cooling efficiency of green space increases with the size of green spaces. On the right-hand side, the gradient is low and the cooling efficiency is reduced and stable. A threshold value of the cooling Efficiency of green space is then calculated as (TVoE = 5.50 ha) (Figure 10).

Figure 10. Relationship between change in LST (°C) and green space area (ha).

The formula of the cooling efficiency curve of green space is y = 0.6641ln(x) + 1.5997 (R² = 0.692). This turning point is the maximum ΔLST and is defined as the cooling intensity of urban green spaces. The cooling efficiency is expressed as a logarithmic curve between the area of each UGS and its maximum ΔLST.

In this study, to further validate the results, the size of green space patches was divided into three segments (<1 ha, 1–4 ha, 2–4 ha, and greater than 5 ha. The results indicate that in the 0–5 ha segments, correlations between ΔLST and the size of green space show a positive relationship, which changes to a negative relationship in the 5 ha segment. This study validates that the calculated threshold value of efficiency (TVoE) in this study is consistent (Figure 11).

Figure 11. LST change with respect to areas of green spaces.

The TVoE of the study clearly shows that smaller green spaces have a positive relation with change in LST, but green spaces with an area greater than 5.5 ± 0.50 ha have a negative relation with LST change. Some studies also revealed that the cooling effect of green spaces has a size-based threshold value. For instance, X. Chen et al. (2012) pointed out that 5 ha of green space area is a vital onset value for cooling. Study conducted by Zhaowu et al. (2018) also indicated that 4.55 ha of green space area is an important threshold value for cooling; and agreed that the maximum cooling extent of Urban Green Spaces (UGS) is stated as the distance between the edge of the vegetation cover and the first turning point of temperature drop compared with the urban green space temperature.

3.5. LU/LC and its contributions to urban surface heating

Among the five land use and land cover (LU/LC) classes examined, only built-up areas and vacant spaces (bare land) showed positive contribution indices in 2023, with values of + 1.75 and + 0.25, respectively. Built-up areas accounted for 12.6% of the landscape, while vacant spaces represented 8.7% (Figure 12). Notably, built-up and dense green space classes consistently exhibited significantly positive and negative deviations, respectively, compared to the overall landscape composition. Areas classified as built-up, such as urbanized zones, consistently contribute to local temperature increases. This suggests that urbanization is a significant factor in raising temperatures locally and may also have implications for global warming. On the other hand, areas categorized as green spaces, such as parks and vegetation-covered areas, show minimal contribution to heat buildup. The negative contribution index observed in vegetation-covered areas indicates that these spaces act as effective heat sinks, helping to mitigate the urban heat island effect. Overall, this underscores the importance of incorporating green spaces into urban environments to counteract rising temperatures associated with urbanization.

Figure 12. Heat contribution index of LU/LC of the years 2000 and 2023.

This finding follows the built-up heat source and vegetation heat sink that characterize most urban landscapes. In this regard, as suggested by a number of authors (X. L. Chen et al., 2006; Choi et al., 2012; Tong et al., 2005), consideration of urban heat sinks is critical in designing sustainable urban plans. Such initiatives could offset the local high urban temperature and mitigate global warming.

Conclusion

This study analyzed the cooling effects of urban vegetation cover in mitigating micro-climate change using geospatial techniques in Adama City central Ethiopia. The study revealed that the radical changes in land use/land cover dynamics. The study compared surface temperature with green and non-green areas to analyze the effects of green spaces on temperature and consequently UHI, and has analyzed the cooling effects of urban green spaces from the year 2000 to 2023. The rapid expansion of built-up areas and reduction of green spaces in the city are the major possible causes for the rise of land surface temperature. The study revealed that radical changes in land use/land cover (LU/LC) dynamics. NDVI of the year 2000 having a value > 0.2 was reduced from 64.54 Km² (46.71%) to 47.84 Km² (34.62%) in the year 2023. It clearly shows that the expansion of built-up land has caused significant land cover change (LCC) as well as changes in the LST. It is a very interesting fact that LST distribution in Adama City is very closely related to the distribution of vegetation cover (NDVI) and built-up areas (NDBI) with values of R² = 928; and R² = 927, respectively. Differences in temperature between the areas covered and uncovered with vegetation clearly reveal that vegetation can minimize surface temperature and can significantly mitigate UHI effects. The derivation of cooling efficiency that is significant for urban planning and decision makers was calculated as 5.5 ± 0.50 ha which means that when the Adama City municipality implements landscape planning, a green space size of 5.5 ± 0.50 ha is the most efficient to cool the heat effect. This study did not analyze the individual contributions of different types, species, shapes, and forms of green spaces to mitigating micro-climate changes separately. Therefore, further research is necessary to explore these factors in depth and gain a more comprehensive understanding of how various aspects of green spaces affect their cooling efficiency in urban areas. The study suggests strengthening the development of urban green spaces as an important strategy to mitigate the effects of micro-climate change.

Authors’ contributions

BBM is involved in data collection, literature work, data analysis, and manuscript writing. KWT and GA were also engaged in providing critical comments and approving the final manuscript.

Consent for publication

The authors agreed to publish in journal of urban ecosystem.

Acknowledgments

The authors would like to acknowledge Jimma University, College of Social Science and Humanities; and Oda Bultum University, Institute of Land Administration for providing facilities to conduct this research.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

No funding received for this research.