Since the 1950s, Saudi Arabia has experienced rapid urbanization [1
]. The percentage of the population that lives in urban areas in Saudi Arabia has increased from 21% in 1950, through 58% in 1975, to 83% in 2015, and is estimated to reach 86% by 2030 and 90% by 2050 [2
]. These large increases in urban population over a short time period have led to the creation of large urban settlements to meet the need for new housing, businesses, industrial areas and transport infrastructure [3
]. Therefore, accurate, consistent and timely information on national trends in urbanization and the growth of cities in space and time is required to enable urban and regional planners and policy-makers to understand the driving forces underpinning the processes of urbanization, to anticipate the consequences of rapid urbanization, and ultimately to make more reliable predictions to achieve sustainable urban development [6
The space–time pattern of urban growth can be measured from traditional sources including national censuses [7
]. However, in Saudi Arabia, such sources lack adequate spatial resolution, may be affected by abrupt, artificial administrative boundaries, and have low temporal frequency, limiting their usefulness for dynamic analysis. Remotely sensed imagery offers an alternative source of information for measuring the space–time growth of cities and urban dynamics with the benefits of providing a synoptic view, a fixed spatial resolution, and frequent revisit period. Fine and moderate spatial resolution images, such as from WorldView, GeoEye, Systeme Probatoire d’Observation de la Terre (SPOT) and Landsat, have been used widely for mapping urban areas [8
]. However, for a vast country like Saudi Arabia, which covers an area of more than 2 million km2
, processing such fine spatial resolution images can be prohibitively time consuming and expensive. For national coverage, it is more appropriate to consider coarse spatial resolution imagery. One source of imagery that has particular relevance to mapping urban areas is the Defense Meteorological Satellite Program (DMSP) Operational Line-Scan System (OLS) of stable night-time (SNT) light images. Such imagery has been used as a valuable source for assessing urban growth extent across very large areas, i.e., the national, regional and global scales.
DMSP-OLS has become a uniform and valuable data source for observing emission sources, such as city and industrial lights, gas flares and large natural fires [6
]. DMSP-OLS data have also been used widely to investigate urban expansion and density [7
]. However, there are several limitations associated with SNT data, including the absence of on-board calibration of the DMSP-OLS sensor, which makes it non-comparable with time-series data derived from other satellites. This can be improved by applying a calibration process to SNT data [24
]. Another issue is the blooming effect in the SNT data, where the lit area exceeds the urban area. This causes difficulty in separating the urban area from its surrounding non-urban area. This issue can be addressed by applying methods such as the empirical threshold technique [18
], classification-based methods [25
] and index-based methods [28
Due to its simplicity and comparatively reasonable accuracy and reliability, the threshold-based approach has been utilized commonly for deriving urban areas from night-time images [15
], although this method often overestimates urban areas in larger cities and underestimates urban areas in small cities and towns. Using a single threshold may be inappropriate for regional and national levels where the desired threshold may vary across space and time [15
]. For example, the appropriate thresholds for extracting urban areas from the DMSP-OLS images ranged between seven and 63 in different regions of the world [15
]. These discrepancies in the threshold values may be attributed to spatial heterogeneity in terms of urban structure or socio-economic status [15
]. Table 1
summarizes the research on urban areas measured by night-time lights in terms of study area, methodology and accuracy.
According to the authors’ knowledge, no consistent or comparable information exists on urban growth in space and time in Saudi Arabia at the national and provincial levels. A variety of studies have been undertaken to detect urban growth using mainly Landsat satellite sensor imagery for certain major cities in Saudi Arabia, such as Riyadh city (in central Saudi Arabia) between 1972 and 2005 [34
] and between 1990 and 2014 [35
], Makkah and Taif (in western Saudi Arabia) between 1986 and 2013 [36
], Jeddah (in western Saudi Arabia) between 1964 and 2007 [37
], and Alkhobar (eastern Saudi Arabia) between 1990 and 2013 [8
]. However, there has been a paucity of studies addressing urban growth across the whole of Saudi Arabia over recent years, leaving a wide gap in knowledge on this important issue.
This research aimed to quantify the spatiotemporal dynamics of urban growth in Saudi Arabia between 1992 and 2013 using DMSP-OLS SNT data for the first time. The analysis was carried out over seven-year intervals, resulting in three time periods (1992–1999, 1999–2006 and 2006–2013). Moreover, due to their political, economic and religious importance, the spatial patterns of growth in four highly urbanized cities (Makkah, Riyadh, Jeddah and Dammam) situated in different parts of the country were considered in greater detail.
Timely and accurate information on urban growth across large areas is required for understanding the drivers of that growth and predicting future growth, as well as responding to a range of environmental and socio-economic problems. Remotely sensed satellite sensor data have been demonstrated to be suitable for urbanization studies. A variety of remotely sensed satellite sensor data are available. However, DMSP-OLS SNT lights data have advantages for urbanization studies because: (1) the coarse spatial resolution (1 km) of the DMSP-OLS SNT data allows large-scale coverage and speedy data processing and requires less labor for data analysis and interpretation, (2) urban phenomena are often clearer and more readily identified in night-time lights satellite sensor data compared then with daytime satellite sensor data, and (3) the high-temporal DMSP-OLS data (1992 to 2013) provide a suitable source of time series imagery for monitoring urban dynamics.
This research describes the application of DMSP-OLS SNT data to analyze the spatial growth patterns in Saudi Arabia between 1992 and 2013. In this research, the lack of on-board sensor calibration was corrected by adopting robust stepwise calibration [41
], intra-annual composition and inter-annual series correction methods [15
]. The hybrid calibration approach greatly improved the continuity and comparability of the DMSP-OLS SNT data. The corrected DMSP-OLS SNT data were assessed using socio-economic variables at the national level for the period from 1992 to 2013 and large correlations were observed, with a mean
value of 0.98. The urban pattern across Saudi Arabia is spatially inhomogeneous and, thus, different thresholds, based on demographic information, were used to map urban areas. The extracted urban areas were assessed at the national level using socio-economic variables and at the city level (four representative cities) using Landsat data. Assessment at the national level produced a large correlation, with an average
value of 0.98. Accuracy assessment at the city level revealed that most of the urban areas were identified accurately from the DMSP-OLS SNT data compared with those derived from the Landsat data, with an average overall accuracy and Kappa statistic of 86% and 0.71, respectively.
Most Saudi cities continued to experience rapid urbanization growth during the period from 1992 to 2013. The robust, calibrated estimates from this analysis reveal that the urban area increased nearly three-fold from 8184 km2 in 1992 to 24,699 km2 in 2013. The large increase in urban area across Saudi Arabia coincided with a significant increase in the national economy that supported investments in services, transportation and industry. The largest urban areas were located in Makkah, Riyadh and Eastern provinces because of their importance. In contrast, the largest urban growth rates occurred in the remaining provinces coinciding with geographically varying development policies.
DMSP-OLS SNT data can be used by urban and regional planners and policy makers to better quantify urban growth spatially and temporally, understand the current driving forces underpinning the urbanization process, make more reliable predictions of future growth, and help to achieve sustainable urban development through careful planning. Despite the calibration approach taken, which helped to ensure that a range of biases were corrected, some uncertainties remain in the DMSP-OLS SNT data (blooming and saturation effects). Therefore, in future research, land surface temperature as well as ancillary spatial data, such as vegetation, water and bare land layers, should be used to reveal how and where the accuracy of urban area mapping might be increased further. In addition, finer spatial resolution night-time data, such as from the Visible Infrared Imager Radiometer Suite (VIIRS), should be investigated as an alternative validation dataset.