Pretoria, City of Tshwane (COT), Gauteng Province, South Africa is one of the cities that continues to experience rapid urban sprawl as a result of population growth and various land use, leading to the change of natural vegetation lands into impervious surface area (ISA). These are associated with transportation (paved roads, streets, highways, parking lots and sidewalks) and cemented buildings and rooftops, made of completely or partly impermeable artificial materials (e.g., asphalt, concrete, and brick). These landscapes influence the micro-climate (e.g., land surface temperature, LST) of Pretoria City as evidenced by the recent heat waves characterized by high temperature. Therefore, understanding ISA changes will provide information for city planning and environmental management. Conventionally, deriving ISA information has been dependent on field surveys and manual digitizing from hard copy maps, which is laborious and time-consuming. Remote sensing provides an avenue for deriving spatially explicit and timely ISA information. Numerous methods have been developed to estimate and retrieve ISA and LST from satellite imagery. There are limited studies focusing on the extraction of ISA and its relationship with LST variability across major cities in Africa. The objectives of the study were: (i) to explore suitable spectral indices to improve the delineation of built-up impervious surface areas from very high resolution multispectral data (e.g., WorldView-2), (ii) to examine exposed rooftop impervious surface area based on different colours, and their interplay with surface temperature variability, (iii) to determine if the spatio-temporal built-up ISA distribution pattern in relation to elevation influences urban heat island (UHI) extent using an optimal analytical scale and (iv) to assess the spatio-temporal change characteristics of ISA expansion using the corresponding surface temperature (LST) at selected administrative subplace units (i.e., local region scale). The study objectives were investigated using remote sensing data such as WorldView-2 (a very high-resolution multispectral sensor), medium resolution Landsat-5 Thematic Mapper (TM) and Landsat-8 OLI (Operational Land Imager) and TIRS (Thermal Infrared Sensor) at multiple scales. The ISA mapping methods used in this study can be grouped into two major categories: (i) the classification-based approach consisting of an object-based multi-class classification with overall accuracy ~90.4% and a multitemporal pixel-based binary classification. The latter yielded an area under the receiver operating characteristic curve (AUROC) = 0.8572 for 1995, AUROC = 0.8709 for 2005, AUROC = 0.8949 for 2015. (ii) the spectral index-based approach such as a new built-up extraction index (NBEI) derived in this study which yielded a high AUROC = ~0.82 compared to Built-up Area Index (BAI) (AUROC = ~0.73), Built-up spectral index (BSI) (AUROC = ~0.78), Red edge / Green Index
(RGI) (AUROC = ~0.71) and WorldView-Built-up Index (WV-BI) (AUROC = ~0.67). The multitemporal built-up Index (BUI) also estimated with AUROC = 0.8487 for 1993, AUROC = 0.8302 for 2003, AUROC = 0.8790 for 2013. This indicates that all these methods employed, mapped ISA with high predictive accuracy from remote sensing data. Furthermore, the single-channel algorithm (SCA) was employed to retrieve LST from the thermal infrared (TIR) band of the Landsat images. The LST overall retrieval error for the entire study generally was quite low (overall root mean square RMSE ≤ ~1.48OC), which signifies that the Landsat TIR used provided good results for further analysis. In conclusion, the study showed the potential of multispectral remote sensing data to quantify ISA and evaluate its interaction with surface temperature variability despite the complex urban landscape in Pretoria. Also, using impervious surface LST as a complementary metric in this research helped to reveal urban heat island distribution and improve understanding of the spatio-temporal developing trend of urban expansion at a local spatial scale.
Rapid urbanization because of population growth has led to the conversion of natural lands into large man-made landscapes which affects the micro-climate. Rooftop reflectivity, material, colour, slope, height, aspect, elevation are factors that potentially contribute to temperature variability. Therefore, strategically designed rooftop impervious surfaces have the potential to translate into significant energy, long-term cost savings, and health benefits. In this experimental study, we used the semi-automated Environment for Visualizing Images (ENVI) Feature Extraction that uses an object-based image analysis approach to classify rooftop based on colours from WorldView-2 (WV-2) image with overall accuracy ~90.4% and kappa coefficient ~0.87 respectively. The daytime retrieved surface temperatures were derived from 15m pan-sharpened Landsat 8 TIRS with a range of ~14.6OC to ~65OC (retrieval error = 0.38OC) for the same month covering Lynwood Ridge a residential area in Pretoria. Thereafter, the relationship between the rooftops and surface temperature (LST) were examined using multivariate statistical analysis. The results of this research reveal that the interaction between the applicable rooftop explanatory features (i.e., reflectance, texture measures and topographical properties) can explain over 22.10% of the variation in daytime rooftop surface temperatures. Furthermore, analysis of spatial distribution between mean daytime surface temperature and the residential rooftop indicated that the red, brown and green roof surfaces show lower LST values due to high reflectivity, high emissivity and low heat capacity during the daytime. The study concludes that in any study related to the spatial distribution of rooftop impervious surface area surface temperature, effect of various explanatory variables must be considered. The results of this experimental study serve as a useful approach for further application in urban planning and sustainable development.
Evaluating changes in built-up impervious surface area (ISA) to understand the urban heat island (UHI) extent is valuable for governments in major cities in developing countries experiencing rapid urbanization and industrialization. This work aims at assessing built-up ISA spatio-temporal and influence on land surface temperature (LST) variability in the context of urban sprawl. Landsat-5 Thematic Mapper (TM) and Landsat-8 OLI (Operational Land Imager) and TIRS (Thermal Infrared Sensor) were used to quantify ISA using built-up Index (BUI) and spatio-temporal dynamics from 1993-2013. Thereafter using a suitable analytical sampling scale that represents the estimated ISA-LST, we examined its distribution in relation to elevation using the Shuttle Radar Topography Mission (SRTM) and also create Getis-Ord Gi* statistics hotspot maps to display the UHI extent. The BUI ISA extraction results show a high predictive accuracy with area under the receiver operating characteristic curve, AUROC = 0.8487 for 1993, AUROC = 0.8302 for 2003, AUROC = 0.8790 for 2013. The ISA spatio-temporal changes within ten years interval time frame results revealed a 14% total growth rate during the study year. Based on a suitable analytical scale (90x90) for the hexagon polygon grid, the majority of ISA distribution across the years was at an elevation range of between >1200m – 1600m. Also, Getis-Ord Gi* statistics hotspot maps revealed that hotspot regions expanded through time with a total growth rate of 19% and coldspot regions decreased by 3%. Our findings can represent useful information for policymakers by providing a scientific basis for sustainable urban planning and management.
Over the years, rapid urban growth has led to the conversion of natural lands into large man-made landscapes due to enhanced political and economic growth. This study assessed the spatio-temporal change characteristics of impervious surface area (ISA) expansion using its surface temperature (LST) at selected administrative subplace units (i.e., local region scale). ISA was estimated for 1995, 2005 and 2015 from Landsat-5 Thematic Mapper (TM) and Landsat-8 OLI (Operational Land Imager) and TIRS (Thermal Infrared Sensor) images using a Random Forest (RF) algorithm. The spatio-temporal trends of ISA were assessed using an optimal analytical scale to aggregate ISA LST coupled with weighted standard deviational ellipse (SDE) method. The ISA was quantified with high predictive accuracy (i.e., AUROC = 0.8572 for 1995, AUROC = 0.8709 for 2005, AUROC = 0.8949 for 2015) using RF classifier. More than 70% of the selected administrative subplaces in Pretoria experienced an increase in growth rate (415.59%) between 1995 and 2015. LST computations from the Landsat TIRS bands yielded good results (RMSE = ~1.44OC, 1.40OC, ~0.86OC) for 1995, 2005 and 2015 respectively. Based on the hexagon polygon grid (90x90), the aggregated ISA surface temperature weighted SDE analysis results indicated ISA expansion in different directions at the selected administrative subplace units. Our findings can represent useful information for policymakers in evaluating urban development trends in Pretoria, City of Tshwane (COT).
Globally, the unprecedented increase in population in many cities has led to rapid changes in urban landscape, which requires timely assessments and monitoring. Accurate determination of built-up information is vital for urban planning and environmental management. Often, the determination of the built-up area information has been dependent on field surveys, which is laborious and time-consuming. Remote sensing data is the only option for deriving spatially explicit and timely built-up area information. There are few spectral indices for built-up areas and often not accurate as they are specific to impervious material, age, colour, and thickness, especially using higher resolution images. The objective of this study is to test the utility of a new built-up extraction index (NBEI) using WorldView-2 to improve built-up material mapping irrespective of material type, age and colour. The new index was derived from spectral bands such as Green, Red edge, NIR1 and NIR2 bands that profoundly explain the variation in built-up areas on WorldView-2 image (WV-2). The result showed that NBEI improves the extraction of built-up areas with high accuracy (area under the receiver operating characteristic curve, AUROC = ~0.82) compared to the existing indices such as Built-up Area Index (BAI) (AUROC = ~0.73), Built-up spectral index (BSI) (AUROC = ~0.78 ), Red edge / Green Index (RGI) (AUROC = ~0.71) and WorldView-Built-up Index (WV-BI) (AUROC = ~0.67). The study demonstrated that the new built-up index could extract built-up areas using high-resolution images. The performance of NBEI could be attributed to the fact that it is not material specific, and would be necessary for urban area mapping.
