Variations in urban land surface temperature intensity over four cities in different ecological zones

This study aims at assessing variations and changes in the intensity of urban land surface temperature (LST) over four major cities in different ecological zone. The study intends to examine the contributions of different land cover types and variation in ecological locations on the intensity of urban LST. Remote Sensing and GIS techniques were used to measure the extent of the LST intensity over different cities and implications of land use/land cover (LULC) changes, using the Landsat TM/ ETM from 1984 to 2012, and Landsat OLI/TIRS from 2015 to 2019. The contributions of different landscape types to urban LST intensity were examined, using contribution index (CI) and Landscape index (LI) methods while the relationship between urban LST, and changes in LULC was examined using zonal statistics. The results revealed that the spatial and temporal changes in the LULC have greatly influenced the LST in the cities, though this varies from identified LULC. Changes in estimated LST vary from 0.12 to 1 °C yearly, while the changes are much intensified in the core section of the cities. The contribution of each landscapes varies, − 0.25 < CI > − 1.17 for sink landscape and 0.24 < CI > 1.05 for source landscape. The results further reveal that as LI ≥ 1, the contribution of source landscape to intensity of LST is lesser than that of sink landscape, but LI ≤ 1 shows that source landscapes contribute more to intensity of LST than sink landscapes. This might be as a result of changes in the vegetation cover between 1984 and 2019 as revealed in LULC change. Loss in the vegetal cover is anthropogenically induced leading to an increase in built-up and impervious surfaces resulted in mean monthly and yearly temperature changes. It is observed that the core and densities areas of cities witnessed higher LST compared with the rural area. The study concludes that different types of land cover within an urban area can affect the spatial pattern of urban LST, though this varies from one ecological zone to another and distribution of LST intensity in the urban area depends on its changes LULC. Thus, as cities’ population is expected to keep expanding there is a need to establish more viable linkages between the ever-growing population and land use patterns. The major findings from this study are useful in informing policymakers of the need to promote more sustainable urban development in the cities.


Methodology
The study area. This study involves different cities from ecological zones in Nigeria: Akure, Ibadan, Lagos, and Saki (Fig. 1). The study cities were purposively selected to cover different ecological zones: Guinean Savannah, derived Savannah, tropical rainforest and mangrove zones ( Table 1). The four cities being considered in this study have experienced some level of urban growth over the years. In recent years, there has been a major increase in the built-up of the area in Akure but the thick forest vegetation has been massively depleted to give room for farming and other agricultural activities, a similar change in the land use from thick vegetal cover to agricultural farmlands could also be observed in Saki. Lagos is a large port city in southwestern Nigeria and one of the most rapidly developing cities in Africa with a landmass of about 2341.72 km 2 . Lagos is also a key cultural centre of the whole region, with very interesting and old local traditions, cuisine, music, and lifestyle. Tourism plays an important role in the city's economy, and most of the guests come to the city when it is time for numerous music festivals and cultural events. Ibadan city, the administrative capital of Oyo state, lies completely within the boundary between the tropical forest zone and derived savannah. According to the Koppen climate classification, it has a tropical wet and dry climate. Its mean total rainfall is 1420.06 mm. Its mean maximum temperature is 26.46 °C while its mean minimum temperature is 21.42 °C.
Multi-year dataset acquisition. A multi-year Landsat dataset, from 1984 to 2019 were used in the study, to examine long period LST. The dataset cover both dry and wet seasons, at least a Landsat scene for a month, depending on low cloud coverage. The Landsat Thematic Mapper (TM), Enhanced Thematic Mapper Plus (ETM+) and an Operational Land Imager-Thermal Infrared Sensor (OLI-TIRS) dataset from 1984 to 2019 were used in this study to assess change in urban LST. The complete multi-year Landsat dataset was used to develop long term LST database for the four cities, results of 1984,1994,2004 and 2019 were presented to show the decadal variation. To examine the changes in Land use/Land cover (LULC) within and around the cities, as it varies over years, Landsat datasets for years 1984, 2004 and 2019 were used for classification.
Land surface temperature retrieve and assessment. The LST over the cities were retrieved from the thermal bands of multi-year Landsat datasets. Thermal bands 6 (10.40-12.50 μm), 6H (10.4-12.5 μm) and 10 (10.60-11.19 μm) of TM, ETM + and OLI-TIRS respectively, with spatial resolutions of 120 m, 60 m and 100 m respectively were used in the study to distinguish nature urban surface temperature in different cities. Also, multi-year Landsat datasets as it relates to changes in land use and land cover (LULC) and NDVI, over four cities in southwestern Nigeria (Fig. 1) were acquired from the United States Geological Survey (USGS), using a path of 191 and a row of 54 for Lagos, path of 191and a row of 55 for Ibadan, the path of 190 and a row of 55 for Akure and path of 191 and a row of 55 for Saki (Table 2).
Subsequently resampling the spatial resolutions to a uniform resolution of 30 m, these thermal bands were used to retrieve LST over the study area for three different periods were done following five major steps: (1) converting Digital Numbers (DNs) to spectral radiance, (2) obtaining effective at-sensor brightness temperature also known as black body temperature from the spectral radiance using Plank's inverse function, (3) estimating land surface emissivity, (4) extracting urban land surface temperature, and (5)  where L λ is the spectral radiance at the sensor's aperture in Wm −2 sr −1 μm −1 ; Q Cal is quantized calibrated pixel values in DNs, and M L and A L are band-specific rescaling factors given in the metadata file as radiance multiband x and radiance add band x respectively. The effective at-sensor brightness temperature also known as black body temperature (T B ) was obtained from the spectral radiance using Plank's inverse function: where T k is at-satellite brightness temperature (in Kelvin); K 2 is a calibration constant in Wm −2 sr −1 μm −1 ; K 1 is calibration constant 1 in Wm −2 sr −1 μm −1 ; and L λ is the spectral radiance at the sensor in Wm −2 sr −1 μm −1 . However, the land surface emissivity estimation was calculated using the temperature values obtained using Eq. (2), which referenced to a blackbody. Therefore, corrections for spectral emissivity (ε) are necessary according to the nature of the land cover.
In ∈K 1 L + 1  Dt represents the difference in the temperature between the sink or source landscape and the entire region while the proportion of the area that is source landscape or sinks landscape in the entire area are present as S. Thus for sink landscape, the contribution to LST intensity is the fraction of the area (S) that is part of the sink landscape multiplied by the difference in temperature (Dt) between the sink landscape and the entire region. For a sink landscape, the Dt values are negative but the contribution index is also positive for the source landscape. The landscape index approach was used to determine the contribution of both landscapes to the urban LST intensity as in Eq. (4): where landscape index (LI) is the function of the contribution index (CI) of both source and sink landscapes. This evaluated how the changes in land use/land cover within and around the cities affect the intensity of urban LST. Also, zonal statistics were used to calculate the relationship of the statistics and to show if a landscape zone feature contains overlapping the LST zones. This was also used to compare the intensity of the LST in a different zone.

Changes in land use/land cover as relate to urban LST intensity. Multispectral image classification
was used to extract thematic information from satellite images. It aids in having a good understanding of the different land cover classes found within an area of interest. Classification in this study was performed using the Maximum likelihood classification form. The maximum likelihood method of classification algorithm is one of the common parametric classifiers used for supervised classification 36,37 . The algorithm is used for Computing the weighted distance or likelihood (D) of unknown measurement vector (X) belonging to one of the known classes (Mc) which is based on the Bayesian equation: A supervised classification was performed by creating a training sample, and based on a spectral signature curve, different land-use classes were created including Built-up area, Waterbody, Cultivation, Natural vegetation and Rock outcrop. While the land surface temperature was extracted using the equation: Since the LST values were obtained in degree Kelvin, LST was converted into degrees Celsius. The LST is the land surface temperature is the function of T B, λ, ρ and ε, such as λ is the central wavelength of emitted radiance (11.45 μm for TM and ETM+; 10.90 μm for OLI-TIRS); T B is at-satellite brightness temperature; ρ is h × c/σ (1.438 × 10 −2 mK or 14,380 µmK); h is Planck's constant (6.26 × 10 −34 Js); c is the velocity of light (2.998 × 10 8 m/s); www.nature.com/scientificreports/ σ is Stefan Boltzmann's constant (1.38 × 10 −23 J K −1 ); and ε is the land surface emissivity. In this study, the relationship between LST and vegetation availability was calculated using zonal statistics. Likewise, each of the LULC categories is assigned an emissivity value by reference to the emissivity classification scheme by Snyder et al. 38 . A heterogeneous surface (that is, a mixture of bare soil, exposed water surface and vegetation) was considered in this study. The emissivity of a heterogeneous surface (mixed pixels) was computed using Eq. (3), taking into account the proportion of vegetation in each pixel (f v ) and the cavity effect due to surface roughness (C λ ).f v was estimated using Eq. (4) 33 , while C λ was calculated using Eq. (5), with the geometrical (shape) factor 'F' having the mean value 0.55, according to Sobrino et al. 39 : NDVI is the Normalized Difference Vegetation Index as computed in Eq. (10) for the respective years considered in this study; NDVI s and NDVI v are Normalized Difference Vegetation Index Threshold values for soil pixels (NDVI s = 0.2) and pixels of full vegetation (NDVI v = 0.5) respectively (Eq. 4), as proposed by Sobrino et al. 39 and Sobrino et al. 33 ; and ε s and ε v are the emissivity (ε) of soil pixels and full vegetation pixels with the mean values 0.97 and 0.99, respectively. The final expression for land surface emissivity is given by: Detail flow of the methodology is presented in Fig. 2.

Results and discussion
In this study, the results revealed three major outcomes:  (Tables 3, 4, 5, 6) over the years, as source landscapes increases, the sink landscapes decreases from 1984 to 2019. The S values for sink and source landscapes, in Akure for example, changes from were 67% and 33% in 1984 to 49% and 51% in 2019, respectively. Similar increases were observed in all other cities but much more high in Ibadan and Lagos. In Ibadan, the S value for sink landscape was 40% in 1984 but changed to 29% in 2019 while source landscape change changed from 60% in 1984 to 71% in 2019 ( Table 5). Much of these changes are observed in Lagos where S value for sink landscape was 47% in 1984 but changed to 25% in 2019 while the source landscape change changed from 53% in 1984 to 75% in 2019. Both spatial and temporal changes in the landscapes types affect the CI of each landscape and the intensity of the LST in all the cities (Tables 3, 4, 5, 6). In terms of the individual contribution of each landscape type in the cities of study, source landscape contributes positively to the intensity of LST while sink landscape has a negative contribution as presented in CI in Tables 3, 4 Table 6). The results imply that as LI ≥ 1, the contribution of source landscape to intensity of LST is lesser than that of sink landscape, but LI ≤ 1 shows that source landscapes contribute more to intensity of LST than sink landscapes (Tables 3, 4, 5, 6). In all the cities examined in this study, the core part of each city was much warmer compared to the neighbouring areas (Figs. 3, 4, 5, 6). The intensity of the LST increased over the years but much more in the centre of the cities.
The general observation in Akure is that only the central part of the city was much warmer in 1984 compared to other past of the city but a significant increase in LST was observed towards the northeastern part of Akure as of 2004 and this scenario appears much more towards the northwestern and southern part of the city in 2019 (Fig. 4).
The intensity of LST clusters in high-density built-up areas in the different cities. It implies thus that the expansion of built-up areas further influences the intensity of LST in the urban area. The intensity of LST increases along the region of the cities with the high rate of high-density and expansion of built-up areas. These results further indicate more of a distinction between the built-up and the vegetal cover in the area as they contribute to the intensity of LST in the study cities. However, the observation of temperature that show a higher In Lagos for example, it could be visually seen (Fig. 5) high cluster of LST in the southern part of the city in 1984 but there appears a significant spread with much more intensity of LST across every part of the city as observed in 2019. In other cities, the high cluster of LST is obvious in the later years from 2004 to 2019. The reasons for these outcomes are very obvious as the majority of source landscapes in the cities centres are usually built-up surfaces distinguished with the concrete surface while sink landscapes are mainly vegetation and water surfaces. This is much more apparent in Lagos with a strong heterogeneity of land surface characteristics, increasing built-up and soil between the coast and mainland, but the vegetation within the city is exceedingly degraded, leading to high intensity of LST 40,41 . Thus, source landscapes are noted for large exposed concrete surfaces with high heat absorption which eventually increases the temperature in the cities while much evapotranspiration and reduction of temperature takes place in the sink landscapes.         Table 7). The results reveal that as the built-up areas increases in the cities (Fig. 8), the intensity of LST is much more upsurged (Figs. 3, 4, 5, 6). Change in LULC is noticeable in all the cities examined in this study. For example, the total land area of Akure, used in this study, is approximately 1638 km 2 . In 1984, the built-up was about 33 km 2 which increase to nearly 100 km 2 in 2019. Cultivation also increased from 217.6  (Table 7). What is obvious from these results is that as both built-up within the city and the cultivation land around the city expanding, vegetation decreases (Fig. 7a), leading to much of the LST intensification in the city (Fig. 3). A similar scenario is obvious in Ibadan in 1984 where for instance, the built-up and cultivation increase from 101 km 2 and 242 km 2 to 343 km 2 and 269 km 2 in 2019, respectively. These expansions decrease in vegetation land from 702 km 2 in 1986 to 434 km 2 in 2019 (Table 7). In Lagos in 1984, the built-up was about 296 km 2 but increased each year to 1053 km 2 in 2019 Cultivation, however, decreased from 193 km 2 in 1984 to nearly 165 km 2 in 2019. The reason for the reduction in cultivation land in Lagos is obvious, since the city is a major industrial location in the country, only a few lands are used for  Table 7). Much of the changes in the LST in the cities (Figs. 3, 4, 5, 6) are the results of changes in the built-up area (Fig. 8). The relationship between LST and LULC was further examined by calculating the zonal statistics of the images for the different years across all four cities (Fig. 6). The result reveals that the mean surface temperature variations over different land cover types. For Akure, the mean temperature over built-up ranges between 24 and 29 °C from 1984 and 2019, this is a high temperature. For the Ibadan area, the mean surface temperature variation over different land cover types revealed that the majority of the bare surface and impervious surface of Ibadan metropolis had temperatures between 22 and 38 °C, while the mean temperature for built-up ranges between 26 and 29 °C from 1984 and 2019 in which by implication of the interpretation of the zonal statistic corresponding to high temperature. For Lagos metropolis, the mean surface temperature variation over different land cover types revealed that the majority of the bare surface and impervious (built-up) surface of Lagos metropolis had temperatures between 20 and 39 °C, while the mean temperature for built-up ranges between 25 and 30 °C from 1984 and 2019. For the Saki area, the mean surface temperature variation over different land cover types revealed that the mean temperature for built-up ranges between 29 and 36 °C from 1984 and 2019 in which by implication of the interpretation of the zonal statistics corresponding to high temperature. The highest mean temperature was observed in Saki and this again is as a result of the ecological location of Saki in the Guinean Savannah.
The results are indicated that the LST is strongly correlated with change in LULC. Hence areas with the least vegetation are experiencing higher land surface temperatures. Again in this study, LST and LULC were found to be closely correlated in all land cover categories, especially in urban and vegetated/rural areas across the four cities that are being considered. Putting the side of the image by side one another could help to visually observe the relationship amongst these variables, seeing the correlation between LST and LULC in a specific location as well as their disparities as they are located in different ecological zones. Thus the results from this study reveal that anthropogenic factors such as urban encroachment into agricultural or rural land due to urban expansion and farming activities have contributed immensely to the removal of vegetation cover. As a result, this alteration in land cover is most responsible for the variation in land surface temperatures over time and space. Besides, the impervious surface is one of the most important land cover types and a feature of urban/suburban environments which affects the LST in the urban area. It is known to affect urban surface temperatures by altering the   www.nature.com/scientificreports/ sensible and latent fluxes. Impervious surfaces represent materials that do not absorb water or dampness [40][41][42] . Most materials used in urban constructions, for example, housetops, roads, interstates, parking areas and walkways are impervious.

Conclusion
The study attempted to compare spatiotemporal variations and intensity of land surface temperature (LST) in four cities located in different ecological zones. Based on the satellite dataset, remote sensing and Geographic Information System (GIS) techniques, the study assesses variations in urban LST in the cities over a period of 30 years. The drivers of LST and consequently urban heat islands, including changes in vegetation cover, bare surface (considering soil moisture content), water bodies and population (hence built density and impervious surfaces) are examined in the study. Compared to the traditional meteorological observation method, remote sensing technology has the advantage of wide coverage, which makes the large-scale assessment of LST possible 43 .
The major findings from this study reveal: (1) spatial and temporal variations in the LST in the cities; (2) change in the land use and land (LULC) are responsible for the change in the LST noted in the cities; (3) significant relationship between change LST, LULC and Ecological location of the cities. Generally, most locations in the cities with a relatively low temperature in 1984 were converted to hot spots in 2004 and hotter in 2019. What is obvious from this study is that maximum air temperature was in built-up areas of the city while the minimum temperature observed was in areas with dense vegetation. This is noticeable to be a result of an increase in built-up areas over the years and a corresponding decrease in vegetation 5,44 . This implies that the spatial and temporal changes in the land uses have greatly influenced the increase in the land surface temperature of each of the identified land uses. The study further reveals increasing variations in LST in all land uses which are results of rapid change LULC and modification of urban landscapes. Cities demonstrate greater temperature in their centre than the surrounding rural areas, which is known as Urban Heat Island effect. These forms a temperature difference between the cities and the surrounding suburbs because of the effect which causes discomfort to the city dwellers 45 . In recent years, the results from the present study demonstrate that a huge amount of vegetation cover is replaced by an artificial built surface that absorbs incoming solar radiation or heat and makes the cities warmer than ever before. The reasons for this outcome is very obvious as the majority of source landscapes in the cities centres are usually built-up surfaces distinguished from the concrete surface. According to the physical properties of the materials, thermal conductivity, and heat capacity, the materials with higher thermal conductivity conduct heat to their interior easier, whereas materials with high heat capacity store more heat and, as more heat is retained, the temperature of the material increases 40,46,47 .
It has been reported in the literature that the phenomenon exists in almost every big city. Numerous factors are held accountable for this effect, including land use and land cover changes in urban areas 48 , anthropogenic heat release, climatic conditions and air pollutants 42,45 . Lagos and Ibadan for instance are some of the fastest-growing cities in Africa today due to the rapid physical expansion of the cities. It is also reported that the population of the people living in Lagos has increased up to five times in the past 30 years which invariably means that there will be massive crowding in terms of establishing structures increased competition for amenities and increased thermal discomfort due to overcrowding of people as well as changes in land use and land cover. These findings are coherent with the results of the studies by Dewan and Corner 48 and Ayanlade 6 which revealed the impacts of land use and land cover changes on urban LST. Moreover, the results from the present study disclose a strong relation between LULC and urban temperature, especially regarding vegetation and built density. The major findings and innovation of this study are that different types of land cover within an urban area can affect the spatial pattern of the LST, though this varies from one ecological zone to another. Normally, the distribution of temperature in an urban area depends on its urban LULC. Based on findings from the present study, it can be established that vegetation availability plays a very low significant role in the temperature of the metropolis. The results from this study are also useful for urban heat mitigation efforts, which will aid policy maker in dealing with city challenges. Essentially as these are the factors directly related to the cities' land surface temperature regime. The major findings from this study are useful in informing policy that will promote more sustainable urban development in the cities. What can be learned from this study is that cities' population is expected to keep expanding and therefore there is a need to establish more viable linkages between the growing population and land use patterns, essentially because these are the factors directly related to the cities' land surface temperature regime.