we retrieved the historical patterns of LULC and LST changes and their relationship with human activities in the research area after using the methods presented in the preceding sections.
Changes in LULC Class Pattern in the Past
The past trends of LULC and LST changes for the period (1988 to 2016) were derived using Landsat data, and the MLC method is shown in Fig. 2. There were two distinct patterns (Fig. 2): the first, gradually increasing the area under built-up class, and the second, decreased vegetation, agriculture, and bare soil areas over the study period. The overall accuracy of the MLC classifications for 1988, 2002 and 2016 was 89.15%, 95.08% and 96.16%, respectively (Table 1 and Table 2).
Table 1 Landsat satellite images details used in the study
Acquired Date (M/D/Y)
|
Scenes ID
|
Sensors
|
Cloud cover (%)
|
Path/ Row
|
5/4/1988
|
LT51500371988125RSA00
|
Landsat 4–5 (TM)
|
<1%
|
150/37
|
5/19/2002
|
LE71500372002139SGS00
|
Landsat-5 (TM)
|
<1%
|
150/37
|
5/17/2016
|
LC81500372016138LGN01
|
Landsat 8 (OLI)
|
<1%
|
150/37
|
The findings showed that between 1988 and 2016, there were an increase of +9.94% in built-up area due to conversion of vegetation, bare soil, and agricultural areas into built-up areas (Table 3). This results from the study area's recent thirty years of increasing urbanization. Since 1988, several reasons, including rapid population increase, migration from neighboring provinces, especially after the 2005 earthquake, and instability in the eastern and western allied areas, have led to urbanization in the studied area.
Table 2 Accuracy assessment of Maximum Likelihood Classification method to classify the land use/cover types
Years
|
Users Accuracy (%)
|
Producers Accuracy (%)
|
Total accuracy (%)
|
Kappa coefficient
|
1988
|
81.29
|
82.68
|
89.15
|
0.85
|
2002
|
93.18
|
96.18
|
95.08
|
0.93
|
2016
|
90.75
|
97.89
|
96.14
|
0.94
|
The current study results support the findings of Srivanit and Hokao, (2012), who documented that social, economic, and political factors caused urban expansion. The vegetation of the study area indicated a decreasing trend from 1988 to 2016 (Table 3), attributed to rapid urbanization since 1988. (Akbar et al. 2019), reported that vegetation in Islamabad Capital Territory decreased due to rapid expansion and replacing the vegetation cover with impervious surfaces. According to (Ullah et al. 2019a), massive deforestation and population influx significantly affected LULC dynamics in the past 30 years. Similar findings have been reported by (Pham et al. 2015), who found a negative impact of urbanization on vegetation cover and agricultural land in Vietnam.
Table 3 - LULC changes for 1988, 2002 and 2016
Class name
|
1988
|
2002
|
2016
|
Change (km2)
|
Net change (%)
|
1988-2002
|
2002-2017
|
(1988‒2016)
|
Built-up Area
|
148.03
|
188.73
|
237.61
|
+40.70
|
+48.88
|
+9.94
|
Bare soil
|
410.02
|
397.88
|
375.27
|
-12.14
|
22.21
|
-3.81
|
Vegetation
|
174.42
|
171.16
|
156.42
|
-3.26
|
-14.74
|
-1.99
|
Agriculture
|
162.44
|
138.54
|
126.92
|
-23.90
|
-11.62
|
-3.94
|
Water Bodies
|
6.47
|
5.39
|
4.84
|
-1.08
|
-0.55
|
-0.18
|
The past pattern of LST Changes for the period 1988-2016
Past patterns of LST changes were obtained using thermal bands of Landsat by utilizing equations described in section 2, shown in Fig. 3. The results indicated that the first three LST classes (low, secondary moderate and moderate) declined from 1988 to 2016 (Fig. 4). The region under the low LST class showed a declining trend and decreased from 1% to 0.06% from 1988 to 2016 (Fig. 4). Similarly, areas for the secondary moderate and moderate LST classes decreased from 9.44% to 1.79% and 15.32% to 15.02%, respectively. The area within the range of secondary high class also decreased from 61.97% to 34.64% during the study period. The region within the range of the high LST class showed an opposite trend and increased from 12.27% to 48.48% (Fig. 4). The reasons for the LST variations may be both urbanization and climate change. Rapid urbanization contributed to a significant construction that increased LST, further exacerbated by climate change. Both the warming of cities and global climate change have an impact on the LST of the study area. The majority of LST regions are located in populated areas (Fig. 3), followed by vegetation, agricultural, bare soil, and water bodies. These findings support (Ahmed and Ahmed 2012), which indicated lower LST for flora, agriculture, and water bodies and greater LST for built-up areas and bare soil.
The mean Land surface temperature Variations in different Land use land cover types
The present research examined the relationship between LST and LULC changes because LULC changes reduce vegetative cover, increasing LST and generate UHI in urban areas, which has various negative effects on human health and the ecosystem, including heat stress and the loss of biodiversity. Our findings show that the built-up areas, bare soil, agricultural, and vegetation have the greatest mean LST (Fig. 5). These results corroborate those made by Kafy et al. (2020), who evaluated the effect of LULC courses on LST in the Rajshahi area of Bangladesh to analyze the connection between thermal signatures and LULC classes. The current data demonstrate that vegetation has been replaced by built-up areas, increasing the LST of the study region. LST also increased for all classes of LULC and expanded beyond the built-up area (Fig. 5). This is in line with (Terfa et al. 2020), who found that LST increased beyond the built-up area in several cities of Ethiopia. Our results were similar to (Ren et al. 2008) in Northern China. However, the LST increase is relatively higher in our study area, possibly due to the rapid urban expansion in the past three decades. The findings of our study suggested the presence of urban warming effects in Islamabad Capital Territory.
The Relationship of Human Modification with the LST Variations
The human modification map for the year 2016 is presented in the Fig. 6. It showed that the highest human modification occurred in the built-up area, which generally has a high human footprint. The relationship between human modification and LST was evaluated by assessing the relationship between HMI, LULC, and LST explained in the following sections.
Human modification index in different temperature zones
The findings of HMI for different LST zones showed that the highest human modification was observed in the highest LST class, followed by the secondary high, moderate, secondary moderate, and lower LST range (Fig. 7). The GWR result is presented in Fig 8. The GWR analysis between HMI and LST indicated a strong spatial relationship, especially in built-up areas (Fig. 8). Our results were similar to those reported by (Chu et al. 2020), who assessed the relationship between HMI and LST on Hainan Island South China Sea. High human modifications in higher LST classes indicated that the anthropogenic activities caused the replacement of the vegetation with built-up areas that increased LST in the study region. High human modifications are usually associated with urban expansion caused by high natural population growth influx of populations into urban areas. Over the past 30 years, rapid urbanization occurred in Pakistan due to a population explosion which has grown 2% each year. By 2030, it will be close to 300 million, and by 2050, 450 million ([CSL STYLE ERROR: reference with no printed form.]). Political and economic issues Weng, (2001) and a rapid population growth rate are among the main causes of urban expansion in our study region, and they are not an exception (Boukorin and Al-shihri, 2015; Aboukorin and Al-shihri, 2015). contribute to urban expansion. These factors resulted in rapid human modification of the study area, which increased the LST.
Variation of human modification index and mean LST in LULC types
The built-up area has the highest HMI, followed by agricultural land, bare soil, vegetation, and water bodies (Fig. 8). The mean LST showed a similar pattern for the study period except for bare soil. The highest mean LST was found in the built-up area, while the lowest was in water bodies. The effect of human activities of LST is widely reported in the literature (Akbar et al., 2019; Sajjad and Iqbal, 2012; Sarmah et al., 2018; Yu et al., 2018). A study carried out in China by Weng et al. (2004) found that LST is affected by human activities and increased mean LST in urban impervious surfaces (Iqbal M et al. 2012) (Ahmed and Ahmed 2012). Our results were also consistent with the findings of (Isa et al. 2013), who found similar results in Kuala Lumpur, Malaysia. It is evident from Fig. 8 those human activities affected LST in all LULC types. High human activities are generally associated with high population density, affecting LST in various parts of the world. There is a very high possibility that human effects might explain UHI and heatwave incidences (Joughin et al. 2002).