Academia.eduAcademia.edu

Figure 13 – uploaded by Qi ZHOU

See full PDFdownloadDownload figure
Figure 13. Distribution of residuals for Waikato, when the geographic unit area was set larger than 0, 0.2, 0.4, and 0.6 km”

Figure 13 Distribution of residuals for Waikato, when the geographic unit area was set larger than 0, 0.2, 0.4, and 0.6 km”

Related Figures (21)

Figure 1. An explanation of the density and completeness of buildings. where, Densitypuiidings is the building density; }~ Building Area is the total area of build- ings in a geographic unit, and Aredgeographic unit is the land area of such a geographic unit. The Densityouitaings May vary from 0% (i.e. no building in a geographic unit) to 100% (i.e. filled with buildings). As an example, Figure 1(a) shows two schematic buildings (shown in black) in one geographic unit (shown by a dotted square). Suppose that the land area of this geographic unit is 1, the building density is then (0.1 + 0.2)/1 x 100% = 30%. In a this geographic unit is 1, the building density is then (0.1 + 0.2)/1 x 100% = 30%. In a
Figure 1. An explanation of the density and completeness of buildings. where, Densitypuiidings is the building density; }~ Building Area is the total area of build- ings in a geographic unit, and Aredgeographic unit is the land area of such a geographic unit. The Densityouitaings May vary from 0% (i.e. no building in a geographic unit) to 100% (i.e. filled with buildings). As an example, Figure 1(a) shows two schematic buildings (shown in black) in one geographic unit (shown by a dotted square). Suppose that the land area of this geographic unit is 1, the building density is then (0.1 + 0.2)/1 x 100% = 30%. In a this geographic unit is 1, the building density is then (0.1 + 0.2)/1 x 100% = 30%. In a
Figure 3. Study area of San Jose City in the United States.
Figure 3. Study area of San Jose City in the United States.
Figure 4. Study areas of Waikato and Hawke's Bay in New Zealand.
Figure 4. Study areas of Waikato and Hawke's Bay in New Zealand.
where, x denotes the OSM building density and y’ denotes the estimated completeness The distribution of residuals across grid cells was then plotted to show the percen: tage of grid cells in each residual range (e.g. 0-10%; Figure 6(b)). The residuals variec considerably with some estimated completeness values much lower than the reference completeness (less than —80%). However, the absolute residuals for 84% of the grid cell: were less than 10%, and for 90% of the grid cells the absolute residuals were less thar 20%. This indicates that, in most cases, the estimated completeness was close to the reference completeness. ee nen Ne en Figure 7 shows the residuals for all 636 of the 1 x 1 km grid cells across San Jose City. [he number of grid cells is larger than the 395 in the previous analysis because there are ‘wo additional types of grid cells. For the first type, there were no OSM buildings in the yeographic unit of a grid cell (i.e. green geographic units in Figure 7). In this case, both he OSM building density and OSM building completeness were 0, and thus the residual nas also equal to 0. For the second type, the reference building completeness was larger han 100% (i.e. gray geographic units in Figure 7), mainly due to the incompleteness of
where, x denotes the OSM building density and y’ denotes the estimated completeness The distribution of residuals across grid cells was then plotted to show the percen: tage of grid cells in each residual range (e.g. 0-10%; Figure 6(b)). The residuals variec considerably with some estimated completeness values much lower than the reference completeness (less than —80%). However, the absolute residuals for 84% of the grid cell: were less than 10%, and for 90% of the grid cells the absolute residuals were less thar 20%. This indicates that, in most cases, the estimated completeness was close to the reference completeness. ee nen Ne en Figure 7 shows the residuals for all 636 of the 1 x 1 km grid cells across San Jose City. [he number of grid cells is larger than the 395 in the previous analysis because there are ‘wo additional types of grid cells. For the first type, there were no OSM buildings in the yeographic unit of a grid cell (i.e. green geographic units in Figure 7). In this case, both he OSM building density and OSM building completeness were 0, and thus the residual nas also equal to 0. For the second type, the reference building completeness was larger han 100% (i.e. gray geographic units in Figure 7), mainly due to the incompleteness of
Figure 9. Results for both Waikato and Hawke’s Bay.
Figure 9. Results for both Waikato and Hawke’s Bay.
Table 1. Minimum and maximum values of OpenStreetMap (OSM) building density for different ranges of OSM building completeness, in three study areas.
Table 1. Minimum and maximum values of OpenStreetMap (OSM) building density for different ranges of OSM building completeness, in three study areas.
slope) usually varies among geographic units, and thus the slope may vary among the study cases. Figure 10 plots the relationships between geographic unit area and refer- ence building density in each grid cell for the three study areas, (a) San Jose, (b) Waikato and (c) Hawke's Bay. As an example of variation, Figure 11 shows the distribution of reference building densities for San Jose City. From this it is evident that the reference building densities for some commercial and industrial zones were higher than 30% or even 40%, but those for some residential zones located near urban fringes were lower
slope) usually varies among geographic units, and thus the slope may vary among the study cases. Figure 10 plots the relationships between geographic unit area and refer- ence building density in each grid cell for the three study areas, (a) San Jose, (b) Waikato and (c) Hawke's Bay. As an example of variation, Figure 11 shows the distribution of reference building densities for San Jose City. From this it is evident that the reference building densities for some commercial and industrial zones were higher than 30% or even 40%, but those for some residential zones located near urban fringes were lower
Figure 11. Distribution of reference building densities for San Jose City.
Figure 11. Distribution of reference building densities for San Jose City.
Figure 12. Distribution of residuals when applying the different slopes (ranging from 3.494 to 4.047) to San Jose City (a), Waikato (b), and Hawke's Bay (c).
Figure 12. Distribution of residuals when applying the different slopes (ranging from 3.494 to 4.047) to San Jose City (a), Waikato (b), and Hawke's Bay (c).
Table 2. Slopes for the different sizes of samples randomly chosen from the grid cells for San Jose City.
Table 2. Slopes for the different sizes of samples randomly chosen from the grid cells for San Jose City.
A Linear
A Linear
‘igure 18. Comparison between the linear and power fittings for San Jose City. MUTI sii J Mer aaa” dil Fourth, in this study, the urban areas were delineated based on the land use data in OSM, which may suffer from quality issues. For example, in some geographical units parks and high schools with a playground were classed as residential areas, for which the building densities were relatively lower. An urban area may also be delineated based on other source data (e.g. remote sensing or census data), although these data may also suffer from classification errors. As an example, the urban areas of Waikato and Hawke's Bay were also delineated with artificial surface data from Globe Land Cover Mapping (GLCM) at a resolution of 30m (http://ngcc.sbsm.gov.cn/article/en/ps/mp/201302/20130200001694. shtml). The results showed that the residuals for 70-76% of the grid cells were smaller
‘igure 18. Comparison between the linear and power fittings for San Jose City. MUTI sii J Mer aaa” dil Fourth, in this study, the urban areas were delineated based on the land use data in OSM, which may suffer from quality issues. For example, in some geographical units parks and high schools with a playground were classed as residential areas, for which the building densities were relatively lower. An urban area may also be delineated based on other source data (e.g. remote sensing or census data), although these data may also suffer from classification errors. As an example, the urban areas of Waikato and Hawke's Bay were also delineated with artificial surface data from Globe Land Cover Mapping (GLCM) at a resolution of 30m (http://ngcc.sbsm.gov.cn/article/en/ps/mp/201302/20130200001694. shtml). The results showed that the residuals for 70-76% of the grid cells were smaller
Figure 19. Distribution of residuals for Waikato and Hawke's Bay, respectively, in which the urban areas were delineated using the data in Globe Land Cover Mapping (GLCM) at a resolution of 30 m.
Figure 19. Distribution of residuals for Waikato and Hawke's Bay, respectively, in which the urban areas were delineated using the data in Globe Land Cover Mapping (GLCM) at a resolution of 30 m.
Related topics:
Data Quality (Computer Science)Quality EstimationGeographic Information Systems (GIS)Low-Rise High-Density Buildings
580 California St., Suite 400
San Francisco, CA, 94104
© 2025 Academia. All rights reserved