GEOSPATIAL DIMENSIONS OF LAND COVER TRANSITIONS AND LAND SURFACE TEMPERATURE IN ABUJA CITY, NIGERIA

Urbanization is often accompanied by succession of underlying land cover with impervious surfaces. Built intensification significantly alters the surface energy budget making cities warmer than their outlying suburbs, which signifies an ecological deterioration. Landsat imageries with scene covering Abuja city is processed using Google Earth Engine platform to estimate land cover and land surface temperature over the span of 30 years (1990-2020). Dimensions of land cover transitions were examined in-terms losses, gains, swaps, net-change and persistency. Thermal signature of each land cover type was estimated using land surface temperature. Urban thermal field variance index is computed from land surface temperature to evaluate the thermal conditions in the city. Results indicate that netchanges for built-up exhibited gains of 40% while agricultural land, bare-land and vegetation exhibited loss of 27%, 7% and 8% respectively. Built-up also showed the highest proportion of persistence (12%). Results shows that land surface temperature has increased by 2.01°C from 1990 to 2020. Agricultural land, bare-land and built-up were found with the highest temperature. Lowest temperature was found in waterbody and vegetation. The ecological evaluation showed that 47% of the city is experiencing bad to worst thermal condition. These findings provide further information that can contribute towards an informed spatial planning in cities.


INTRODUCTION
The worlds urban population is above 50% and projected to reach 70% by 2050 (UN-Habitat, 2019). Cognizant of the fact that 35% of the projection will come from China, India and Nigeria (UN-DESA, 2019). Urbanization is often accompanied by profound alteration of underlying natural surfaces such as vegetation (P W . The natural surfaces are transformed into impervious structures such as roads and buildings (Filho et al., 2021). The prevalence of impervious surfaces modifies biophysical properties such as albedo, latent heat and conductivity, thereby creating Urban Heat Island (UHI) phenomena (Oke, 1973). Extreme heat has the potential of effecting the thermal comfort and health of urban dwellers (Cavan et al., 2014;EPA, 2016). This raises concerns for prior planning towards achieving the United-Nation's Sustainable Development Goals (SDGs) (UN-Habitat, 2020).
Urbanization have occurred on relatively small-fraction of Earth's land surface, but yet it contributes significantly to the loss of natural ecosystems (UNCCD, 2017). An informed land management strategy have great potential of ensuring a more secure and sustainable urban future (Hishe, 2021). There is need for a spatially resolved analytics of land cover and its ecological footprints. Remote Sensing and geographic information system (GIS) have been recognized as effective geospatial technologies for observing earth's surface features (Manakos et al., 2021). Geodata analytics provides adequate spatial detail needed for a sustainable spatial planning. Land surface temperature (LST) is a key indicator of Earth's surface energy balance (Dissanayake, 2019).
In-depth analysis of cross-tabulation matrix of land cover change provides insights on the components of transitions (Adugna et al., 2017;Pontius et al., 2004). UHI intensity is extricable linked to the anomalous changes in LST and can be ecologically evaluated using the Urban Thermal Field Variance Index (UTFVI) (Alcantara et al., 2019). Abuja City has significantly urbanized since its inauguration as the capital of Nigeria (Gumel et al., 2020). This paper aims to explore the geospatial dimensions of land cover transitions and Land Surface Temperature (LST) in Abuja city, Nigeria. The specific objectives of this study are: (I) to analyse the dynamics of land cover in Abuja City, Nigeria; (II) to analyse the pattern of LST in relation to land cover changes; and (III) to evaluate the thermal comfort of Abuja City, Nigeria.

Study Area
The study focuses on Abuja City, Federal Capital Territory (FCT) of Nigeria (Figure 1). The area covers 256 km 2 and extends between latitude 7° 25' & 9° 20' Northing of the Equator and longitude 5° 45' & 7° 39' Easting of the Greenwich Meridians. Abuja City serves as the administrative and political headquarters for the Federal Government of Nigeria (FGN) (Abubakar, 2014). The area has two distinct seasons, rainy and dry. The rainy season starts in April and retreats by September, while the dry season starts in November and retreats by March (Adeyeri et al., 2017 The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XLIII-B3-2022 XXIV ISPRS Congress (2022 edition), 6-11 June 2022, Nice, France

Data and Pre-processing
Land cover and land surface temperature data (path:189, row:54) are excavated from Landsat collections in the GEE platform (https://earthengine.google.org/). The acquired scenes are cloud free level_1 tier (L_1T) products generated at medium resolution for 1990, 2001, 2014 and 2020 (Table 1). The images were selected within dry season to avoid phenological variability between time points (Verbesselt et al., 2010). The boundary shapefile for Abuja City was obtained from Abuja Geographical Information System (AGIS). The reference layer for training and validation is available in Google Earth. Landsat imageries were projected to Universal Transverse Mercator (UTM) Zone 32N and georeferenced to the World Geodetic System (WGS) 1984.  Table 1: Landsat data used in the study GEE cloud computing infrastructure facilitates advanced processing and analysis of images (Ermida et al., 2020). The L_1T Landsat imageries are calibrated to correct radiometric and geometric distortion. Prior to image analysis an atmospheric correction is performed. The acquired raw bands were converted to top of atmosphere (TOA) reflectance and radiance from digital numbers (DN) (Cetin et al., 2008). The multispectral bands were used to obtain land cover classes and Normalized Difference Vegetation Index (NDVI). While the TIR bands were used to retrieve LST. The methodology is summarized in (Figure 2).

Land Cover Classification
Random Forest (RF) algorithm was used to classify the land cover for the year 1990, 2001, 2014 and 2020. RF non-parametric image classifier yields higher accuracy amongst the other types machine leaning algorithm (Talukdar et al., 2020). The land cover classification scheme consisted of five (5) classes; agricultural land, bare-land, built-up area, vegetation and waterbody. The validation of RF classified land cover produced an overall accuracy (OA) of 82.83%, 88.64%, 90.40% and 93.76% for the years 1990, 2001, 2014 and 2020. The kappa coefficient of the reference epochs ranged between 0.80 to 0.9.

Post Classification Comparison
The classified images of 1990, 2001, 2014 and 2020 were overlaid to analyze the spatial dynamics of land cover. An extended two-dimensional comparison matrix of 1990 and 2020 was produced to dictate the declining, increasing and static patterns of land cover categories. The components of land cover transitions were analyzed based on losses, gains, swaps, netchange and persistency (Braimoh, 2006;Pontius et al., 2004).

Urban Thermal Field Variance Index
Urban thermal field variance index (UTFVI) was used for the ecological evaluation of UHI in Abuja City. UTFVI is given by Eq. (6) (Guha et al., 2018).
Where, UTFVI = Urban Thermal Field Variance Index Ts = Land Surface Temperature (Kelvin) Tm = Mean of Land Surface Temperature (Kelvin) The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XLIII-B3-2022 XXIV ISPRS Congress (2022 edition), 6-11 June 2022, Nice, France

Urban Heat Island
The UHI assessment of heat stress is given by Eq. (7).
Where, UHI = Urban Heat Island LST = Land Surface Temperature SD = Standard Deviation 3. RESULTS

Land Cover Dynamics
The classified Landsat images of Abuja City for the year 1990, 2001, 2014 and 2020 are shown in (Figure 3). The images are ordered into five (5) land cover classes; agricultural land, bareland, built-up, waterbody and vegetation. The land cover has shown progressions between 1990 and 2020 due urbanization.   Table 2: Percentage of land cover change from 1990 to 2020

Land Cover Transitions
The components of land cover transitions between 1990 to 2020 are presented in the order of loss, gain, total change, swap and net-change (Table 3). The result indicate that built-up area experienced the highest gain 44% and also the least loss at 4%. While, agricultural land experienced the highest loss at 34%. Then followed by vegetation that lost 21% and gained 13%. The highest change attributable to the absolute net gain was found in built-up areas (40%) and the least was found in bare-land (7%). Swap type of change was highest for vegetation, which has experienced nearly a pure swap of its total transition pattern.  Table 3: Components of land cover transitions  The proportion of each land cover category that remained static between 1990 and 2020 is presented as the diagonal values (Table  4). The results indicate that built-up has the highest persistence at 11.97%. Followed by agricultural land at 10.10%. While waterbody experienced the lowest persistence at 0.19%. An estimated 6.19% of vegetation persisted.

100
The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XLIII-B3-2022 XXIV ISPRS Congress (2022 edition), 6-11 June 2022, Nice, France The minimum, maximum, standard Deviation and mean LST has shown a gradual increase between 1990 and 2020 ( Table 5). The minimum LST ranged from 19°C to 23°C, while maximum LST ranged from 36°C to 43°C and mean LST ranged between 30°C to 34°C. The mean LST increased by 2.56 °C from 1990 to 2020. The result concur with Oke, 1973). The mean thermal signature for each land cover type is shown in (Figure 5). Vegetation and water bodies exhibited the lowest mean LST, which can be attributed to their higher latent heat transfer and evapotranspiration (Ahmed et al., 2002). Highest mean LST was observed in agricultural land, bare-land and builtup areas. Bare-land reflects incoming radiation thereby warming up faster than other land cover types (Huang et al., 2015). These findings justify the dependence of LST on land cover types.

Evaluation of Thermal Comfort in Abuja city
The Urban Thermal Field Variance Index (UTFVI) was used to quantitatively describe the thresholds of thermal and ecological comfort in relation to Urban Heat Island (UHI) intensity in Abuja City ( Figure 6). UHI intensity is extricable linked to anomalous changes in Land Surface Temperature (LST) (Alcantara et al., 2019). Built intensification alters surface energy budget (EPA, 2016), thereby decreasing the thermal comfort within cities. Figure 6: UTFVI for Abuja City

CONCLUSION
This paper analysed the spatial dimensions of land cover transitions and land surface temperature in Abuja city, Nigeria. Google Earth Engine (GEE) platform offers great potential in processing Landsat imagery to retrieve Land Surface Temperature (LST) and to classify land cover using Random Forest (RF) at high accuracy. The Landsat imageries were obtained during the dry season to avoid phenological variability between time points (Verbesselt et al., 2010). The analysis of transition matrix reveals that Abuja City has undergone significant change in land cover since its inauguration in 1990. Built-up is the dominant land cover that has significantly gained from the loss of other categories like vegetation, agricultural and bare-land. Built-up has also shown the highest level of persistence of changing to other land cover. The descriptive statistics of LST showed a significant increase in temperature from 1990 to 2020. Vegetation and water bodies recorded the lowest mean LST, which can be attributed to their higher latent heat transfer and evapotranspiration (Ahmed et al., 2002). Highest mean LST was observed in agricultural land, bare-land and built-up areas. Bare-land reflects incoming radiation thereby warming up faster than other land cover types (Huang et al., 2015). These findings justify the dependence of LST on land cover types. The urban thermal field variance index of Abuja City shows that a sizable portion of Abuja City is experience bad to worst thermal comfort. There is need for a proactive land management strategy that will contribute towards the achievement the goals of SDGs and improving the thermal comfort of urban dwellers. The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XLIII-B3-2022 XXIV ISPRS Congress (2022 edition), 6-11 June 2022, Nice, France