Monitoring of UN sustainable development goal SDG-9.1.1: study of Algerian “Belt and Road” expressways constructed by China

Research area and data

Research data

This paper uses WorldPop and HRSL population raster data, Landsat remote sensing imagery, Algerian urban area and administrative division map. The WorldPop project was launched in 2013 as part of a global population mapping project led by the Institute of Geographical Data, University of Southampton, UK (https://www.worldpop.org/). It aims to provide an open access archive of spatial demographic datasets for Central and South America, Africa and Asia to support development, disaster response and health applications. It uses raster prediction based on a random forest model to generate a 100 m spatial resolution population density, which is then used as a weighted surface and redistributed according to national population data (Stevens et al., 2015). LandScan uses the National Imagery and Mapping Agency’s (NIMA) Vector Map (VMAP) series data and Digital Terrain Elevation Data (DTED) for global coverage of roads and slopes, respectively (Bhaduri et al., 2002). HRSL population data has a resolution of 30 m and was obtained from the Center for International Earth Science Information Network (CIESIN) (Batran et al., 2018). Global Administrative Areas (GADM) is a database of the locations of the world’s administrative areas. Administrative areas in this database include countries, counties and departments (Hijmans, Garcia & Wieczorek, 2010). Landsat 5 and Landsat 8 data were obtained from the US Geological Survey (Table 2; El-Askary et al., 2014). These data were processed using several commands in ArcGIS 10.5 software.

Research methods

This paper aims to calculate the SDG 9.1.1 indicator in the proximity of five Algerian expressways built as part of China’s Belt and Road Initiative Project, from large to small areas. Figure 2 presents a flowchart of the three methods. (1) Obtaining large-scale statistics based on HRSL population data, using the nearest neighbor method and regional statistical method to evaluate and analyse the contribution of five Algerian expressways to monitor indicator 9.1.1 in the past five years; (2) For medium-sized areas, an evaluation method based on the NDBI building area index is proposed. (3) For small areas, an evaluation method based on high-resolution time series images is proposed. (4) analysis and comparison of the three population statistics methods.

(1) A large-scale evaluation method for indicator 9.1.1. based on HRSL population raster data.

In this paper, the population statistics of residents within two kilometers of five expressways are studied using HRSL population data and nearest neighbor and regional statistics methods. The nearest neighbor method locates the positions of pixels in the output image to the original image by the nearest neighbor method. It finds the nearest pixel from the original image and takes the value of the pixel as the value of the output image pixel. The nearest neighbor method is used to calculate the size of a single grid pixel. Then, by using the function of regional statistical analysis, all the grid pixel values are tabulated and, finally, the sum of all pixel values is calculated (Caraway, McCreight & Rajagopalan, 2014; Rossi, Dungan & Beck, 1994).

(2) Evaluation method of a medium area based on the NDBI building area index.

The Middle Area comprises the Bejaia-Haniff Link Expressway in Algeria and uses the NDBI building area population estimation method for demographic analysis. A two-kilometer NDBI distribution map of the Bejaia-Haniff Link Expressway road was calculated using Landsat remote sensing images from 2011, 2015, and 2019 and is presented in Fig. 3. The normalized difference building index (NDBI) indicates the distribution of land for regional construction (Zha, Ni & Yang, 2003; Li et al., 2017; Xu, 2008). The calculation method is shown in Eq. (1). The NDBI is mainly used to extract information on buildings in cities and towns but, according to Jin et al.’s (2020) research, it can also be used for villages and towns. There was a significant positive correlation between the remote sensing index of village buildings and surface temperature (P < 0.05). The Bejaia-Haniff Link Expressway has less urban area within 2 km. Most of the buildings are built in villages and towns. The article uses the city vector mask to remove the extracted urban buildings and retain the villages and towns.

Where NDBI represents the normalized building index, SMIR represents mid-infrared reflectance, and NIR represents near-infrared reflectance.

Supervised classification: The maximum likelihood method is used to extract a sample from the whole population to ensure that the same block with the same band value as the sample is assigned the same attributes belonging to the same type of feature (Cabral et al., 2018; Murthy, Raju & Badrinath, 2003; Otukei & Blaschke, 2010; Keuchel et al., 2003). Combining remote sensing images of the study area with data on its land types, we categorized the land types as residential, forest, cultivated, and bare by visual interpretation. Then, 100 uniformly distributed samples were selected from the entire image (Fig. 4). After many experiments, the separability between each sample type was greater than 1.8 and the maximum likelihood algorithm was selected for classification. Finally, combined with 0.3m high-resolution remote sensing image for accuracy verification, the accuracy of the verification results is above 90% (Li et al., 2014; Zhang et al., 2016; Pujiono et al., 2019).

NDBI was calculated using the Landsat-5 and Landsat-8 remote sensing image bands to extract the area of rural buildings, and the area of rural buildings within two kilometers of the expressway in 2011, 2015 and 2019 was estimated. In order to maintain the accuracy of extracting buildings and eliminate the effects of bare soil, cultivated land and water bodies, the bare soil and cultivated land were extracted by the supervised classification maximum likelihood method, while water bodies were extracted by visual interpretation. Finally, the NDBI was masked by areas of bare soil, cultivated land and water bodies to obtain a map of the rural building distribution. The total population is then calculated based on the total area of the building and the per capita living area.

(3) Small-scale evaluation method based on high-resolution time series images

The small area of the town of Tikhramtath in the Bejaia-Haniff buffer zone of Algeria was analysed in terms of demographics. A remote sensing method was used to monitor changes in buildings in 2011, 2015, 2017 and 2019 in small areas according to a multi-time remote sensing data map. A visual interpretation of the number of buildings and number of floors in the domains was carried out through Google Earth 0.24m high-resolution remote sensing images. First, the building area was calculated with the Calculate Geometry tool. Then, according to the size of the building, the population of each floor of each building was estimated to calculate the total population of the area.

(4) Accuracy evaluation and comparative analysis

Comparisons of the results of the three methods presented in the article were made. The root mean square error (RMSE) and relative RMSE (%RMSE) were used to measure the accuracy of WorldPop data, HRSL data, NDBI estimated building area data, and remote sensing visual interpretation data. The %RMSE values were obtained by dividing the RMSEs by the average of the number of censuses, which can reflect the accuracy of the model simulation. Finally, the most suitable population calculation methods for different regions were selected (Tan et al., 2017; Draper et al., 2013; Bhunia, Shit & Maiti, 2018).

(2)${\displaystyle \text{RMSE}=\sqrt{\frac{1}{N}\sum {\left(f_i-r_i\right)}^2}}$ (3)${\displaystyle \text{\% RMSE}=\frac{\text{RMSE}}{\frac{1}{N}\sum r_i}}$ Where fi is the estimated value of the ith group of data, that is, the estimated population density obtained after population spatialization; ri is the reference value of the ith data, that is, the population density value obtained from the census data; N represents group data.

Experimental results

Based on WorldPop population statistics

Algerian urban demographic results.

The statistics of the Algerian urban population are crucial to the statistics of the rural population. The rural population size is based on the total population minus the urban population. The distribution of the Algerian urban population from 2009 to 2019 is shown in Fig. 5. The sizes of the urban population in 2009, 2011, 2013, 2015, 2017 and 2019 were 11,396,261, 11,857,311, 12,632,705, 13,302,082, 14,019,482, and 14,820,436, respectively. In these years, the total rural population was 23,247,534, 24,126,163, 24,821,053, 25,788,594, 26,896,482, and 28,158,778 people.

According to the comparison of urban population changes, Fig. 5 shows that there are 84 cities in Algeria, five of which with an area of 200 km² or more, all of which are located in the northern part of Algeria. The population of urban areas has grown rapidly over the last 10 years, with a total growth of 3,424,085 people. The urban areas along the new expressways are particularly obvious, including Oran, Alger, Blida, Bouira, Sidi-Bel-Abbes and other larger cities.

Demographic results of the provinces of Algeria

Population data from the provinces of Algeria play an important role in calculating the proportion of the rural population that lives within 2 km of an expressway. Figure 6 shows that the top three most populous provinces in Algeria are Alger, Setif, and Oran, with populations of 3,037,455, 1,510,132 and 1,483,869, respectively. The three provinces with the smallest populations are Tindouf, Illizi, and Tamanghasset, with populations of 52,170, 54,636 and 181,217, respectively. The population growth rate of each province is low. In 2009-2011, the growth rate was above 10%, being 12.678% in Tindouf Province. In 2011–2013, the growth rate was above 10%.

The provinces with growth rates >10% during 2015–2017 were Naama Province, Tindouf Province, and Tiaret Province, with growth rates of 11.132%, 12.678% and 17.411%, respectively. Tizi Ouzou Province had the lowest growth rate at 0.516%. The provinces with growth rates >10% in 2017–2019 were the same, with growth rates of 11.694%, 12.678% and 22.306%, respectively, while Tizi Ouzou Province had the lowest growth rate at 0.545% (Fig. 7).

China’s construction of five expressways in Algeria: SDG-9.1.1 assessment

The Worldpop population raster data was counted as follows. (1) A expressway 2-km buffer surface was created using tool “Feature to polygon” (2) Natural Earth urban area data was based on the tool “Extract by mask” to obtain the urban population raster data, then the spatial analyze tool “Raster calculator” was used to obtain the rural population raster map. (3) The generated buffer surface was obtained using the “Extract by mask” tool to obtain 2-km rural population raster data. (4) The Worldpop cell raster layer was converted to a vector point layer through the “From raster to point tool” and the grid cell values were assigned to the point layer. (5) The required information was extracted from the point layer to a data table, then the population was counted (Tewari & Manning, 2017; Holt et al., 2018).

The distribution of the rural population after China’s construction of the Algeria expressway is shown in Fig. 8. The population of the Algeria Cherchell Ring Expressway is gradually increasing as shown in Table 3. Before the construction in 2011-2013, the growth rate was 1.77%. The population increased rapidly after the start of construction and was 6.66% in 2015. Population growth during the construction period was slower, dropping to 3.03%, and then increased afterward. From the start of construction to two years later, there was an increase of 5,733 people. The population growth rate within 2 km of the 53 km North-South Expressway decreased by 0.22% in 2013. After the start of construction, it gradually increased in 2017. After construction, the growth will be slower. The 53 km north-south expressway in Algeria will increase by 1,715 people from the start of construction to the second year after construction. The growth rate of the Bejaia-Haniff connection line of Algeria from pre-construction in 2011 to construction in 2015 gradually increased to 3.29%, the growth rate in 2015-2017 fell to 2.48%, and the growth rate after construction was 2.65%. The population increased by 15,825 during the Bejaia-Haniff Expressway construction period.

Table 3:

Statistics of highway populations.

Year	Cherchell Ring Expressway population	Growth rate	North-South Expressway population	Growth rate	Bejaia Expressway population	Growth rate
2011 (before construction)	39,581	4.82%	35,168	10.45%	136,392	2.18%
2013 (before construction)	40,282	1.77%	35,246	0.22%	140,084	2.71%
2015 (under construction)	42,964	6.66%	35,681	1.23%	144,698	3.29%
2017 (under construction)	44,265	3.03%	36,883	3.37%	148,286	2.48%
2019 (after construction)	46,015	3.95%	37,950	2.89%	152,217	2.65%

DOI: 10.7717/peerj.8953/table-3

As shown in Table 4, the population near the Tipaza expressway decreased. After the project started, the growth rate increased rapidly to 5.84%. During the construction process, the population growth rate dropped to 5.79%. After construction, the growth rate reached 7.09% two years later in 2013. The growth rate dropped to 2.66% from 2013 to 2015, and the population increased by 11,988 from the start to the end of construction. The growth rate of the Algeria East-West Expressway gradually increased from 2005 to 2009 to 6.36%. During the construction process, the population growth rate dropped to 3.68%. After completion of construction, the population growth rate was 4.85%. The population increased by 156,755 people during the East-West Expressway construction period. China’s construction of five expressways in Algeria has increased the rural population within the 2 km buffer zone by 192,016 in the eight years of construction. By 2019, the population increased by 329,291, accounting for 1.17% of the rural population. In 2009, 2011, 2013, 2015, 2017, 2019, the rural population within the 2 km buffer zone accounted for 4.79%, 4.80%, 4.87%, 4.89%, 4.90%, and 4.86% of the total rural population, respectively.

Table 4:

Statistics of the highway populations.

Year	Tipaza 48 km highway population	Growth rate	East-West Expressway population	Growth rate
2005 (before construction)	60,320	3.70%	764,851	3.61%
2007 (before construction)	60,212	−0.18%	797,051	4.21%
2009 (under construction)	63,729	5.84%	847,763	6.36%
2011 (under construction)	67,419	5.79%	878,946	3.68%
2013 (after construction)	72,200	7.09%	921,606	4.85%

DOI: 10.7717/peerj.8953/table-4

Large area range HRSL population raster data statistics results

The large-area HRSL population raster data analysis method is used the nearest neighbor method to calculate the size of a single grid pixel. Then, by using the ArcGIS function of regional statistical analysis, all the HRSL population raster grid pixel values are tabulated and, finally, the sum of all pixel values is calculated.

The distribution of HRSL raster data for the rural population living within 2 km of an expressway in 2015 is shown in Fig. 9. The populations of the Algeria East-West Expressway, Bejaia Expressway, Tipaza Expressway, North-South Expressway, and Cherchell Ring Expressway were 911,549, 127,471, 71,411, 30,583 and 41,657, respectively, accounting for approximately 4.12% of the 2015 rural population.

The population living within 2 km of the East-West Expressway was the largest of the five expressways, with a total population of 1,208,199, as shown in Table 5. The population distribution was closely related to the geography of Algeria. The northern part of Algeria comprises the coastal plains and hills of the Mediterranean coast, the central part is the Tell Atlas and Sahara Atlas Mountains, and the southern part of the Sahara desert is largely uninhabited.

Monitoring results of population estimation method for NDBI building area in the middle area

Figure 10 shows that the building areas within 2 km of the Bejaia-Haniff Expressway in 2011, 2015 and 2019 were 15.84 km², 17.77 km², and 21.77km², respectively in Algeria. By sampling 100 buildings to determine that the mean single building area is 290 m², the rural populations in 2011, 2015 and 2019 are estimated to be 273,118, 306,430 and 375,408, respectively. They accounted for 15.23%, 16.65%, and 19.64% of the rural populations of Bejaia, Bouira, and Boumerdes provinces, respectively.

The rural population of the Bejaia-Haniff Expressway in Algeria had the largest population growth rate in 2019 of 22.51%. Among them, the Tahalacht Industrial Zone in the northern part of the expressway, the Beni Mansour and the Tikhramtath town in the southern region, the Tala lbir town in the central region, Sidi Elash and Sidi Ayad have increased significantly.

Estimated population by remote sensing monitoring of a small area

Tikhramtath town is located at the beginning of the Bejaia Expressway, the closest to the Bejaia Expressway, and the towns and buildings are concentrated. The buildings increased significantly in 2011, 2015, 2017 and 2019, making it particularly suitable for small-area case selection.

The size of the building was vectored as a polygon feature by ArcGIS, then the ArcGIS tool “Calculate geometry” was used to calculate the area of the building. According to the 0.24-m high-resolution multi-time-series remote sensing image map, the number of building layers was compared with the image data of different periods. The number of layers of a single building was determined from the remote sensing image map. Finally, the layers of all buildings in Tikhramtath town were counted (Saadaoui et al., 2019).

A small area containing the town of Tikhramtath, within 2 km of the Bejaia-Hanif Expressway in Algeria (before, during and after construction) was selected for demographic analysis. A map of the building distribution is shown in Fig. 11. The population of Tikhramtath town in 2011, 2015, 2017 and 2019 was 1,790, 2,785, 3,365 and 3,870, respectively, accounting for 0.29%, 0.35%, 0.52%, and 0.60% of the total rural population in Bouira.

From the statistical chart, the size of buildings in the town of Tikhramtath within 2 km of the Bejaia-Hanif Expressway, the floor height, and the population of each floor of each building were estimated. The population of the Tikhramtath town in 2015, 2017 and 2019 is shown in Table 6. The growth rates in 2015, 2017 and 2019 were 55.59%, 20.83%, and 15.01%, respectively. The increase in population was mainly due to the construction of the Bejaia-Haniff Expressway.

Comparative analysis

(1) WorldPop and HRSL population statistics were used to compare regions around five expressways built by China’s Belt and Road Initiative Project in 2015. The results are shown in Table 7.

The table shows that there is a small difference between the WorldPop demographic data and the HRSL population data. The rural populations of the five expressways buffers differed by 1307, 5098, 17,227, 2706, and 51,172.

(2) Comparison of the World Bank’s WorldPop population statistics method, the NDBI building area estimation method, and the HRSL population statistics method for the 2 km population of the Bejaia-Hanif Expressway.

This paper uses the World Bank’s WorldPop population statistics, the NDBI building area estimation method, and HRSL population statistics to analyse the demographics of the Begaia-Haniff Expressway in Algeria. The results are shown in Table 8.

According to the statistics, the WorldPop population data and NDBI building area population estimates for 2011, 2015, and 2019 are quite different. The differences are 136,726, 161,732 and 223,191, respectively. The difference between the WorldPop and HRSL population statistics is small, at 17,227. The NDBI building area population estimation method is subjective, and the total population can only be judged roughly, resulting in large errors. Worldpop demographics and HRSL demographics are also due to the small difference in statistical results due to resolution issues.

(3) Table 9 compares the WorldPop population data and the small area remote sensing monitoring method for Tikhramtath town.

Table 9 shows that the results of the three methods for Tikhramtath town are quite different. The remote sensing monitoring visual interpretation method and WorldPop population statistics differed by 1782, 2766, 3349 and 3853, respectively. The remote sensing monitoring visual interpretation method differed from the HRSL population statistics by 2,690 people. The WorldPop population statistics are obviously inconsistent with the actual situation, while the remote sensing monitoring visual interpretation method locates the added buildings.

This paper aims to calculate SDG indicator 9.1.1 in the proximity of five Algerian expressways. The SDG-9.1.1 indicator represents the proportion of the rural population who live within 2 km of an all-season road. At present, researchers in China and elsewhere have used many methods to evaluate the SDG-9.1.1 indicator. Qiu et al. (2019), Mariathasan, Bezuidenhoudt & Olympio (2019) and Xu, Bai & Chen (2019) have proposed different methods. However, most of the above methods use existing population data that is updated slowly, such that the new population cannot be found in time. This paper proposes the use of remote sensing to monitor newly added buildings to estimate the new rural population. This method makes up for the problem of the slow update of existing population data.

Precision verification

We used GHS Population Grid data for accuracy verification. The GHS Population Grid is made up of residential population estimates for target year 2015 provided by CIESIN GPWv4 (Gridded Population of the World, now in its fourth version). The estimates were disaggregated from census or administrative units into grid cells, according to the distribution and density of built-up areas as mapped in the GHSL global layer for the corresponding epoch. An accuracy assessment table is shown in Table 10.

In the accuracy comparison, lower %RMSE values indicate higher accuracy. The accuracy of the WorldPop population data in the large area was 89.99%, which is slightly higher than the 81.02% accuracy of the HRSL population data. WorldPop population data was also most accurate in the middle region. The NDBI building area estimates are subjectively influenced by people and had the lowest accuracy of 2.58%. The accuracy of the population data for Tikhramtath town (in the small area) was generally low and that of the HRSL population data was higher than that of the other two groups, at 12.34%.