# Cloud Computation Using High-Resolution Images for Improving the SDG Indicator on Open Spaces

^{*}

## Abstract

**:**

^{2}within eight years.

## 1. Introduction

^{2}per capita and Tijuana (Mexico) with 2 m

^{2}per capita [9]. In general, the absence of green spaces accounts according to the World Health Organization (WHO) for 3.3% of global death [10].

_{Open}) compares the total built-up area of an urban agglomeration with the area that is open space (gray and green spaces). To operationalize this indicator, the continuously built-up area of an urban agglomeration (thus not the administrative city) is mapped with publicly available satellite images (e.g., Landsat, Sentinel) and the amount of open spaces and the street area (which are also important spaces for interaction) are mapped. The indicator (Built

_{Open}) is computed following [8]:

^{2}. Taking the example of Kampala, and the urban area mapped by the GHSL (528 km

^{2}) or the Atlas of Urban Expansion (513 km

^{2}), this would arrive at an image data cost of more than 5000 Euros for one city. To upscale this to all cities (e.g., as included in the JRC database of urban centers), which is approximately 13,000 cities, and for different years, this would generate enormous image data costs (for all cities of the database this would be close to 10 million Euros for one year). Furthermore, the processing of VHR image data requires enormous computational resources. Both problems (image and computational costs) can be at least partially solved for research purposes by high-resolution image repositories, i.e., the Planet image data repository (https://www.planet.com/markets/education-and-research/), which enables access to high and very-high-resolution images (ranging from 1 to 5 m spatial resolution) and Google Earth Engine (GEE) [16], providing an efficient computational environment. The study addresses the above-identified bottlenecks (access to high-resolution satellite imagery and the computational costs) by proposing a framework that is built on cloud computation, open imagery and open reference data (for research), with minimal manual interaction and tuning to support future scalability. Therefore, the aim of this study is to analyze whether high-resolution (HR) images (in particular, RadipEye images) provide a better estimation of the presence of built-up and open spaces in support of SDG indicator 11.7.1 at the urban scale. The processing of the images and mapping of built-up areas is carried out within GEE, a very efficient cloud computation environment. The paper is divided into the following sections. Section 2 provides an overview of the study area, the city of Kampala, and describes the input data. Section 3 elaborates on the methodology to assess the current GHSL for the study area and the machine learning workflow used to map built-up and non-built-up areas. Section 4 provides the results and shows the potential to ease the calculation of SDG indicator 11.7.1. Section 5 discusses the role of EO to provide consistent data for SDG indicators and the new avenues opened by cloud computation, and Section 6 details the main conclusions.

## 2. Study Area and Input Data

_{Open}for 2010, 2016 and 2018, respectively, and b) changes in Built

_{Open}.

#### 2.1. Overview of the Study Area

#### 2.2. Input Data

^{2}, which allowed excluding small isolated structures. For the gray open space class, samples were selected along major and secondary roads, ensuring that samples did not fall on roofs. Water and vegetation samples were selected in the zones of natural areas. For 2016, we extracted reference data from the ESA Africa land cover map (http://2016africalandcover20m.esrin.esa.int/). As the 2016 data were of the coarser resolution, the following rules were implemented. Samples of buildings and roads from 2010 were used and it was determined whether the areas changed between 2010 and 2016, as well as whether the ESA land cover map highlighted them as built-up areas. Water points from 2010 were also used and changes in the waterline (Lake Victoria) were updated. Vegetation points were extracted from the ESA Africa land cover map (restricting the selection to the classes trees, shrubs and grassland). For 2018, we used the OSM dataset [23] for the city of Kampala; we only extracted buildings with an area larger than 100 m

^{2}. For gray open spaces, samples were selected along major and secondary roads, ensuring that samples did not fall on roofs. Water and vegetation samples were selected in the zones of natural areas.

## 3. Methodology

#### 3.1. Assessment of the GHSL for the Study Area

- Taking 50% built-up area (building footprints) as the threshold for classifying a cell as built-up area.
- Taking 10% built-up area (building footprints) as the threshold for classifying a cell as built-up area.
- Taking more than 0% built-up area (building footprints) as the threshold for classifying a cell as built-up area.

#### 3.2. Calculation of the Built_{Open} Indicator

_{Open}indicator based on cloud computation and machine learning classification of HR images. To do so, we used the city delimitation given by World Bank [20] as a boundary for the study area. The ground truth data required to train the machine learning algorithm was extracted from different sources, including municipal data and OSM data [23]. We also compare the Built

_{Open}indicator for different years, namely 2010, 2016, and 2018.

_{Open}index following the formula presented in (1).

#### 3.2.1. Ensemble Classification

**Data pre-processing**

**Ensemble classifier**

#### 3.2.2. Consistency of Changes

- The classified mosaic of 2016 was taken as a reference because this dataset obtained a higher accuracy in the classification.
- Pixel values in the 2016 mosaic were replaced by the pixel value in the 2010 mosaic if those belong to built-up areas or gray spaces.
- Pixel values in 2010 were replaced by gray spaces as visual inspection confirmed that these pixels were wrongly classified as water, mainly being areas of shadows (e.g., along roads).
- Pixel values in the 2018 mosaic were substituted by the pixel value of the classified mosaic in the previous reference year, i.e., the classified mosaic for 2016.

_{Open}index, we created a common mask to achieve the same area coverage because it was not possible to find less than 10% cloud coverage scenes for the entire study area. These consistency rules also enabled the use of areas of cloud coverage without masking. This workflow allowed for minimal manual interaction, which is important for the development of scalable methods.

## 4. Results

#### 4.1. The Strength and Weaknesses of the GHSL for Built-Up and Open Space Mapping

_{Open}for SDG indicator 11.7.1.

#### 4.2. Image Classification and Built_{Open} Indicator

_{Red}) also contributes to the identification of classes as it was selected in two datasets, 2010 and 2018.

_{Open}is reported there as well. These results show that the amount of open space continuously declines and when calculating the amount of open space per capita (using the WorldPop population data [45] of the respective years), the share reduces from approximately 300 to 170 m

^{2}between 2010 and 2018.

## 5. Discussion

_{Open}index based on HR images and machine learning classification methods. The proposed method uses free-for-research available HR images and a free-for-research cloud platform such as GEE. Building on computational efficiency, HR images and harvesting of reference data from available databases open opportunities to scale up the method beyond this pilot case study. Based on the case study, we could demonstrate that the combination of data within GEE made it feasible to monitor the temporal dynamics of the Built

_{Open}index at a relatively high temporal granularity. At the metropolitan scale, such data enable monitoring the progress towards achieving SDG 11.7.1, in particular when adding local information about public access and population data split into gender, age groups and disabilities. Using a publicly accessibly population dataset (WorldPop [45]), our results showed that the share of open space per capita almost halved between 2010 and 2018, with a loss of 125 m

^{2}per capita (Table 7). Having such spatially detailed data allows specifying the shares of gray and green open spaces in different parts of the metropolitan area. For example, Kampala city is very densely built up with limited access to gray and green open spaces in the immediate surroundings per person. Furthermore, there is a major difference in open space availability between planned and well-serviced areas as compared to informal and deprived areas. We used the available municipal data for 2010 to analyze this difference. Results show that all informal areas within the city had on average of 20% less open space as compared to planned areas (example area is shown in Figure 9).

_{Open}indicator, is a ratio between the open spaces and street space divided by the amount of built-up area [8]. As such, the indicator is difficult to interpret for policy making and the general public. The indicator Built

_{Open}per capita in m

^{2}, which would already include the aspect of the population into the equation, would be much easier to communicate to the general public. This indicator can be more easily quantified for cities across the globe, combining existing population data (e.g., WorldPop data [46]) and data on gray and green open spaces extracted by the proposed method in this paper.

_{Open}indicator also expressed in a meaningful sense for laypersons, i.e., Built

_{Open}per capita.

## 6. Conclusions

_{Open}index using a machine learning algorithm applied to HR images in a cloud computing environment. We demonstrated that the proposed method has the potential to support the monitoring of urban dynamics and to inform policy making in addressing the SDG 11 targets, currently limited by coarse resolution data that do not sufficiently allow quantifying open spaces. Thus, using EO-based information to calculate Built

_{Open}per capita would allow supporting strategic urban planning aimed at urban sustainability, by providing an easily quantifiable indicator that shows one aspect of sustainability. We discussed the limitations of this study concerning the availability of ground truth data required to train machine learning models and showed that open data and volunteered geographic information (VGI) data support training and validation of such models. The developed method allows mapping the dynamics of open spaces, both gray and green spaces, and provides relevant base data for urban planning and management of cities, addressing a crucial aspect of the environmental quality of livability in cities. Future work includes extending our study to cover other cities where the GHSL presents problems in capturing the difference between built-up and open spaces, such as cities located in very arid regions.

## Author Contributions

## Funding

## Acknowledgments

## Conflicts of Interest

## Appendix A. Textural Feature Formulas

- $p\left(i,j\right)$ is the (i,j)
^{th}entry in a normalized gray tone matrix, - ${p}_{x}\left(i\right)={\sum}_{i=1}^{{N}_{g}}P\left(i,j\right)$, is the i
^{th}entry in the marginal probability matrix computed by summing the rows of $p\left(i,j\right)$ for fixed i, - ${p}_{y}\left(j\right)={\sum}_{i=1}^{{N}_{g}}P\left(i,j\right)$, is the j
^{th}entry in the marginal probability matrix computed by summing the columns of $p\left(i,j\right)$ for fixed j, - ${N}_{g}$ is the number of distinct gray levels in the quantized image,
- ${p}_{x+y}\left(k\right)={\sum}_{i=1}^{{N}_{g}}{\sum}_{j=1}^{{N}_{g}}p{\left(i,j\right)}_{i+j=k}$, and ${p}_{x-y}\left(k\right)={\sum}_{i=1}^{{N}_{g}}{\sum}_{j=1}^{{N}_{g}}p{\left(i,j\right)}_{\left|i-j\right|=k}$

Name/Formula | Name/Formula |
---|---|

Angular Second Moment ${f}_{1}={\displaystyle {\displaystyle \sum}_{i=1}^{{N}_{g}}}{\displaystyle {\displaystyle \sum}_{j=1}^{{N}_{g}}}{\left\{p\left(i,j\right)\right\}}^{2}$ | Contrast ${f}_{2}={\displaystyle {\displaystyle \sum}_{n=0}^{{N}_{g-1}}}{n}^{2}\left\{{{\displaystyle {\displaystyle \sum}_{i=1}^{{N}_{g}}}{\displaystyle {\displaystyle \sum}_{j=1}^{{N}_{g}}}p\left(i,j\right)}_{\left|i-j\right|=n}\right\}$ |

Correlation ${f}_{3}={\displaystyle {\displaystyle \sum}_{i=1}^{{N}_{g}}}{\displaystyle {\displaystyle \sum}_{j=1}^{{N}_{g}}}\frac{\left(i,j\right)p\left(i,j\right)-{\mu}_{x}{\mu}_{y}}{{\sigma}_{x}{\sigma}_{y}}$ | Variance ${f}_{4}={\displaystyle {\displaystyle \sum}_{i=1}^{{N}_{g}}}{\displaystyle {\displaystyle \sum}_{j=1}^{{N}_{g}}}{\left(i-\mu \right)}^{2}p\left(i,j\right)$ |

Inverse Difference Moment ${f}_{5}={\displaystyle {\displaystyle \sum}_{i=1}^{{N}_{g}}}{\displaystyle {\displaystyle \sum}_{j=1}^{{N}_{g}}}\frac{1}{1+{\left(i-j\right)}^{2}}p\left(i,j\right)$ | Sum Average ${f}_{6}={\displaystyle {\displaystyle \sum}_{i=2}^{2{N}_{g}}}i{p}_{x+y}\left(i\right)$ |

Sum Variance ${f}_{7}={\displaystyle {\displaystyle \sum}_{i=2}^{2{N}_{g}}}{\left(i-{f}_{8}\right)}^{2}{P}_{x+y}\left(i\right)$ | Sum Entropy ${f}_{8}=-{\displaystyle {\displaystyle \sum}_{i=2}^{2{N}_{g}}}{p}_{x+y}\left(i\right)log\left\{{p}_{x+y}\left(i\right)\right\}$ |

Entropy ${f}_{8}=-{\displaystyle {\displaystyle \sum}_{i=2}^{2{N}_{g}}}{p}_{x+y}\left(i\right)log\left\{{p}_{x+y}\left(i\right)\right\}$ | Difference Variance ${f}_{10}=$ variance of ${p}_{x-y}$ |

Difference Entropy ${f}_{11}={\displaystyle {\displaystyle \sum}_{i=0}^{{N}_{g-1}}}{p}_{x-y}\left(i\right)log\left\{{p}_{x-y}\left(i\right)\right\}$ | Information Measures of Correlation 1 ${f}_{12}=\frac{HXY-HXY1}{\mathrm{max}\left\{HX,HY\right\}}$where, $HXY=-{\displaystyle {\displaystyle \sum}_{i=1}^{{N}_{g}}}{\displaystyle {\displaystyle \sum}_{j=1}^{{N}_{g}}}p\left(i,j\right)\mathrm{log}\left(p\left(i,j\right)\right)$ $HXY1=-{\displaystyle {\displaystyle \sum}_{i=1}^{{N}_{g}}}{\displaystyle {\displaystyle \sum}_{j=1}^{{N}_{g}}}p\left(i,j\right)\mathrm{log}\left\{{p}_{x}\left(i\right){p}_{y}\left(j\right)\right\}$ $HX$ and $HY$ are entropies of ${\mathrm{p}}_{x}$ and ${p}_{y}$ |

Information Measures of Correlation 2 ${f}_{13}={\left(1-{e}^{\left[-2.0\left(HXY2-HXY\right)\right]}\right)}^{1/2}$, where $HXY2=-{\displaystyle {\displaystyle \sum}_{i=1}^{{N}_{g}}}{\displaystyle {\displaystyle \sum}_{j=1}^{{N}_{g}}}{p}_{x}\left(i\right){p}_{y}\left(j\right)\mathrm{log}\left\{{p}_{x}\left(i\right){p}_{y}\left(j\right)\right\}$ | Maximal Correlation Coefficient ${f}_{14}={\left(\mathrm{sec}\mathrm{ond}\mathrm{largest}\mathrm{eigen}\mathrm{value}\mathrm{of}\mathrm{Q}\right)}^{\frac{1}{2}}$ where ${Q}_{\left(i,j\right)}={\displaystyle {\displaystyle \sum}_{k=0}^{{N}_{g-1}}}\frac{p\left(i,k\right)p\left(j,k\right)}{{p}_{x}\left(i\right){p}_{y}\left(k\right)}$ |

Dissimilarity ${f}_{15}={\displaystyle {\displaystyle \sum}_{i=1}^{{N}_{g}}}{\displaystyle {\displaystyle \sum}_{j=1}^{{N}_{g}}}{\left|i-j\right|}^{2}p\left(i,j\right)$ |

- $s\left(i,j,\delta ,T\right)$ is the (i,j)
^{th}entry in a normalized Gray Level Co-occurrence Matrix, equivalent to$p\left(i,j\right)$, - $T$ represents the region and shape used to estimate the second order probabilities, and
- $\delta =\left(\mathsf{\u2206}x,\mathsf{\u2206}y\right)$ is the displacement vector.

Description | Formula |
---|---|

Inertia | $I\left(\delta ,T\right)={\displaystyle {\displaystyle \sum}_{i=0}^{L-1}}{\displaystyle {\displaystyle \sum}_{j=0}^{L-1}}{\left(i-j\right)}^{2}s\left(i,j,\delta ,T\right)$. |

Cluster shade | $A\left(\delta ,T\right)={\displaystyle {\displaystyle \sum}_{i=0}^{L-1}}{\displaystyle {\displaystyle \sum}_{j=0}^{L-1}}{\left(i+j-{\mu}_{i}-{\mu}_{j}\right)}^{3}s\left(i,j,\delta ,T\right)$ |

Cluster prominence | $B\left(\delta ,T\right)={\displaystyle {\displaystyle \sum}_{i=0}^{L-1}}{\displaystyle {\displaystyle \sum}_{j=0}^{L-1}}{\left(i+j-{\mu}_{i}-{\mu}_{j}\right)}^{4}s\left(i,j,\delta ,T\right)$ |

## References

- Douglas, O.; Lennon, M.; Scott, M. Green space benefits for health and well-being: A life-course approach for urban planning, design and management. Cities
**2017**, 66, 53–62. [Google Scholar] [CrossRef][Green Version] - UN-Habitat. Streets as Public Spaces and Drivers of Urban Prosperity; UN-Habitat: Nairobi, Kenya, 2013. [Google Scholar]
- Wicht, M.; Kuffer, M. The continuous built-up area extracted from ISS night-time lights to compare the amount of urban green areas across European cities. Eur. J. Remote Sens.
**2019**, 52, 58–73. [Google Scholar] [CrossRef] - Wang, J.; Kuffer, M.; Sliuzas, R.; Kohli, D. The exposure of slums to high temperature: Morphology-based local scale thermal patterns. Sci. Total Environ.
**2019**, 650, 1805–1817. [Google Scholar] [CrossRef] [PubMed] - Kim, J.-H.; Lee, C.; Sohn, W. Urban Natural Environments, Obesity, and Health-Related Quality of Life among Hispanic Children Living in Inner-City Neighborhoods. Int. J. Environ. Res. Public Health
**2016**, 13, 121. [Google Scholar] [CrossRef] - Dayo-Babatunde, B.; Martinez, J.; Kuffer, M.; Kyessi, A.G. The Street as a Binding Factor: Measuring the Quality of Streets as Public Space within a Fragmenting City: The Case of Msasani Bonde la Mpunga, Dar es Salaam, Tanzania. In GIS in Sustainable Urban Planning and Management; van Maarsevee, M.F.A.M., Martinez, M., Flacke, J., Eds.; CRC Press: Boca Raton, FL, USA, 2018; pp. 183–201. [Google Scholar]
- Shrestha, S.R.; Sliuzas, R.; Kuffer, M. Open spaces and risk perception in post-earthquake Kathmandu city. Appl. Geogr.
**2018**, 93, 81–91. [Google Scholar] [CrossRef] - UN-Habitat. Metadata Indicator 11.7.1: Average Share of the Built-up Area of Cities that is Open Space for Public Use for All, by Sex, Age and Persons with Disabilities; United Nations Statistics Division, Development Data and Outreach Branch: New York, NY, USA, 2018. [Google Scholar]
- OECD. Green Area per Capita; UN-Habitat: Nairobi, Kenya, 2018. [Google Scholar]
- WHO. Urban Green Spaces. Available online: https://www.who.int/sustainable-development/cities/health-risks/urban-green-space/en/ (accessed on 1 March 2020).
- Esch, T.; Thiel, M.; Schenk, A.; Roth, A.; Müller, A.; Dech, S. Delineation of urban footprints from TerraSAR-X data by analyzing speckle characteristics and intensity information. IEEE Trans. Geosci. Remote Sens.
**2010**, 48, 905–916. [Google Scholar] [CrossRef] - Taubenböck, H.; Esch, T.; Felbier, A.; Wiesner, M.; Roth, A.; Dech, S. Monitoring urbanization in mega cities from space. Remote Sens. Environ.
**2012**, 117, 162–176. [Google Scholar] [CrossRef] - Pesaresi, M.; Guo, H.D.; Blaes, X.; Ehrlich, D.; Ferri, S.; Gueguen, L.; Halkia, M.; Kauffmann, M.; Kemper, T.; Lu, L.L.; et al. A global human settlement layer from optical HR/VHR RS data: Concept and first results. IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens.
**2013**, 6, 2102–2131. [Google Scholar] [CrossRef] - Florczyk, A.J.; Melchiorri, M.; Corbane, C.; Schiavina, M.; Maffenini, M.; Pesaresi, M.; Politis, P.; Sabo, S.; Freire, S.; Ehrlich, D.; et al. Description of the GHS Urban Centre Database 2015; Public Release 2019; Version 1.0; Office of the European Union: Luxembourg, 2019. [Google Scholar]
- Angel, S.; Blei, A.M.; Parent, J.; Lamson-Hall, P.; Galarza-Sanchez, N.; Civco, D.L.; Thom, K. Atlas of Urban Expansion—2016 Edition; The NYU Urbanization Project: New York, NY, USA, 2016. [Google Scholar]
- Gorelick, N.; Hancher, M.; Dixon, M.; Ilyushchenko, S.; Thau, D.; Moore, R. Google Earth Engine: Planetary-scale geospatial analysis for everyone. Remote Sens. Environ.
**2017**, 202, 18–27. [Google Scholar] [CrossRef] - Uganda Bureau of Statistics. Statistical Abstract; Uganda Bureau of Statistics: Kampala, Uganda, 2019. [Google Scholar]
- UN-Habitat. The State of African cities 2014. Re-imagining Sustainable Urban Transitions; UN-Habitat: Nairobi, Kenya, 2014. [Google Scholar]
- Pérez-Molina, E.; Sliuzas, R.; Flacke, J.; Jetten, V. Developing a cellular automata model of urban growth to inform spatial policy for flood mitigation: A case study in Kampala, Uganda. Comput. Environ. Urban Syst.
**2017**, 65, 53–65. [Google Scholar] [CrossRef] - World Bank. Kampala Metropolitan Area Spatial Extent; World Bank: Washington, DC, USA, 2017. [Google Scholar]
- Planet Team. Planet Application Program Interface: In Space for Life on Earth. Available online: https://api.planet.com (accessed on 1 January 2019).
- Planet Team. Rapideye™ Imagery Product Specifications; Planet Labs Inc.: San Francisco, CA, USA, 2016. [Google Scholar]
- OpenStreetMap contributions. Planet Dump. 2017. Available online: https://planet.osm.org (accessed on 9 December 2018).
- Duque, J.C.; Patino, J.E.; Betancourt, A. Exploring the Potential of Machine Learning for Automatic Slum Identification from VHR Imagery. Remote Sens.
**2017**, 9, 895. [Google Scholar] [CrossRef][Green Version] - Pesaresi, M.; Ehrlich, D.; Ferri, S.; Florczyk, A.J.; Freire, S.; Halkia, M.; Julea, A.; Kemper, T.; Soille, P.; Syrris, V. Operating Procedure for the Production of the Global Human Settlement Layer from Landsat Data of the Epochs 1975, 1990, 2000, and 2014; Publications Office of the European Union: Luxembourg, 2016. [Google Scholar]
- Aguilar, R.; Zurita-Milla, R.; Izquierdo-Verdiguier, E.; de By, R. A Cloud-Based Multi-Temporal Ensemble Classifier to Map Smallholder Farming Systems. Remote Sens.
**2018**, 10, 729. [Google Scholar] [CrossRef][Green Version] - Louhaichi, M.; Borman, M.M.; Johnson, D.E. Spatially Located Platform and Aerial Photography for Documentation of Grazing Impacts on Wheat. Geocarto Int.
**2001**, 16, 65–70. [Google Scholar] [CrossRef] - Huete, A.; Didan, K.; Miura, T.; Rodriguez, E.P.; Gao, X.; Ferreira, L.G. Overview of the radiometric and biophysical performance of the MODIS vegetation indices. Remote Sens. Environ.
**2002**, 83, 195–213. [Google Scholar] [CrossRef] - Huete, A.R. A soil-adjusted vegetation index (SAVI). Remote Sens. Environ.
**1988**, 25, 295–309. [Google Scholar] [CrossRef] - Haboudane, D.; Miller, J.R.; Tremblay, N.; Zarco-Tejada, P.J.; Dextraze, L. Integrated narrow-band vegetation indices for prediction of crop chlorophyll content for application to precision agriculture. Remote Sens. Environ.
**2002**, 81, 416–426. [Google Scholar] [CrossRef] - Gitelson, A.A.; Kaufman, Y.J.; Stark, R.; Rundquist, D. Novel algorithms for remote estimation of vegetation fraction. Remote Sens. Environ.
**2002**, 80, 76–87. [Google Scholar] [CrossRef][Green Version] - McFeeters, S.K. The use of the Normalized Difference Water Index (NDWI) in the delineation of open water features. Int. J. Remote Sens.
**1996**, 17, 1425–1432. [Google Scholar] [CrossRef] - Space4water. Normalized Difference Water Index (NDWI). Available online: http://space4water.org/taxonomy/term/1315 (accessed on 3 April 2020).
- Haralick, R.M.; Shanmugam, K.; Dinstein, I. Texture features for image classification. IEEE Trans. Syst. Man. Cybern.
**1973**, 3, 610–621. [Google Scholar] [CrossRef][Green Version] - Conners, R.W.; Trivedi, M.M.; Harlow, C.A. Segmentation of a high-resolution urban scene using texture operators. Comput. Vis. Graph. Image Process.
**1984**, 25, 273–310. [Google Scholar] [CrossRef] - Shen, H.; Abuduwaili, J.; Ma, L.; Samat, A. Remote sensing-based land surface change identification and prediction in the Aral Sea bed, Central Asia. Int. J. Environ. Sci. Technol.
**2019**, 16, 2031–2046. [Google Scholar] [CrossRef] - Wurm, M.; Weigand, M.; Schmitt, A.; Geiß, C.; Taubenböck, H. Exploitation of textural and morphological image features in Sentinel-2A data for slum mapping. In Proceedings of the Joint Urban Remote Sensing Event (JURSE), Dubai, United Arab Emirates, 6–8 March 2017; pp. 1–4. [Google Scholar]
- Pesaresi, M.; Gerhardinger, A.; Kayitakire, F. A robust built-up area presence index by anisotropic rotation-invariant textural measure. IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens.
**2008**, 1, 180–192. [Google Scholar] [CrossRef] - Izquierdo-Verdiguier, E.; Zurita-Milla, R. An evaluation of Guided Regularized Random Forest for classification and regression tasks in remote sensing. Int. J. Appl. Earth Obs. Geoinf.
**2020**, 88, 102051. [Google Scholar] [CrossRef] - Rodriguez-Galiano, V.F.; Ghimire, B.; Rogan, J.; Chica-Olmo, M.; Rigol-Sanchez, J.P. An assessment of the effectiveness of a random forest classifier for land-cover classification. ISPRS J. Photogramm. Remote Sens.
**2012**, 67, 93–104. [Google Scholar] [CrossRef] - Mountrakis, G.; Im, J.; Ogole, C. Support vector machines in remote sensing: A review. ISPRS J. Photogramm. Remote Sens.
**2011**, 66, 247–259. [Google Scholar] [CrossRef] - Camps-Valls, G.; Gomez-Chova, L.; Calpe-Maravilla, J.; Martin-Guerrero, J.D.; Soria-Olivas, E.; Alonso-Chorda, L.; Moreno, J. Robust support vector method for hyperspectral data classification and knowledge discovery. IEEE Trans. Geosci. Remote Sens.
**2004**, 42, 1530–1542. [Google Scholar] [CrossRef] - Kavzoglu, T.; Colkesen, I. A kernel functions analysis for support vector machines for land cover classification. Int. J. Appl. Earth Obs. Geoinf.
**2009**, 11, 352–359. [Google Scholar] [CrossRef] - Leonita, G.; Kuffer, M.; Sliuzas, R.; Persello, C. Machine Learning-Based Slum Mapping in Support of Slum Upgrading Programs: The Case of Bandung City, Indonesia. Remote Sens.
**2018**, 10, 1522. [Google Scholar] [CrossRef][Green Version] - Linard, C.; Gilbert, M.; Snow, R.W.; Noor, A.M.; Tatem, A.J. Population Distribution, Settlement Patterns and Accessibility across Africa in 2010. PLoS ONE
**2012**, 7, 8. [Google Scholar] [CrossRef][Green Version] - Wardrop, N.A.; Jochem, W.C.; Bird, T.J.; Chamberlain, H.R.; Clarke, D.; Kerr, D.; Bengtsson, L.; Juran, S.; Seaman, V.; Tatem, A.J. Spatially disaggregated population estimates in the absence of national population and housing census data. Proc. Natl. Acad. Sci. USA
**2018**, 115, 3529. [Google Scholar] [CrossRef][Green Version] - Grippa, T.; Georganos, S.; Zarougui, S.; Bognounou, P.; Diboulo, E.; Forget, Y.; Lennert, M.; Vanhuysse, S.; Mboga, N.; Wolff, E. Mapping Urban Land Use at Street Block Level Using OpenStreetMap, Remote Sensing Data, and Spatial Metrics. ISPRS Int. J. Geo-Inf.
**2018**, 7, 246. [Google Scholar] [CrossRef][Green Version] - Mahabir, R.; Stefanidis, A.; Croitoru, A.; Crooks, A.; Agouris, P. Authoritative and volunteered geographical information in a developing country: A comparative case study of road datasets in Nairobi, Kenya. ISPRS Int. J. Geo-Inf.
**2017**, 6, 24. [Google Scholar] [CrossRef] - Huang, C.; Yang, J.; Lu, H.; Huang, H.; Yu, L. Green Spaces as an Indicator of Urban Health: Evaluating Its Changes in 28 Mega-Cities. Remote Sens.
**2017**, 9, 1266. [Google Scholar] [CrossRef][Green Version] - Giezen, M.; Balikci, S.; Arundel, R. Using Remote Sensing to Analyse Net Land-Use Change from Conflicting Sustainability Policies: The Case of Amsterdam. ISPRS Int. J. Geo-Inf.
**2018**, 7, 381. [Google Scholar] [CrossRef][Green Version]

**Figure 1.**(

**a**) The 2014 Global Human Settlement Layer (GHSL) (built-up areas in yellow) and building footprints (in red); (

**b**) building footprints (in red) on top of a Google Earth image.

**Figure 2.**(

**a**) Atlas of Urban Expansion (2015); (

**b**) building footprints (in red) on top of the built-up areas of the Atlas of Urban Expansion (gaps Google Earth image).

**Figure 3.**Greater Kampala Metropolitan area and its administrative subdivisions (the background GHSL and main streets from OpenStreetMap (OSM).

**Figure 4.**(

**a**) Mosaic of RapidEyes scenes in the study area for 2010, (

**b**) Mosaic of RapidEyes scenes in the study area for 2016, and (

**c**) Mosaic of RapidEyes scenes in the study area for 2018.

Year | Data Source | Criteria | Total Points |
---|---|---|---|

2010 | Municipal Map City of Kampala | Built-up: Buildings with an area >100 m^{2}Centre line of roads: 4m buffer around roads Vegetation: natural vegetation areas Water: Water bodies | 3996 |

2016 | ESA Africa Land cover map of 2016 | Built-up 2010 (check for changes): Buildings with an area >100 m^{2}Centre line of roads 2010 (check for changes): 4m buffer around roads ESA vegetation: grass, greenfield, forest, garden, brownfield, greenspace Water: Water bodies | 3931 |

2018 | OSM | Built-up: buildings with an area >100 m^{2}Centre line of roads: 4m buffer around roads Vegetation: grass, greenfield, forest, garden, brownfield, greenspace Water bodies | 3893 |

**Table 2.**Features extracted from a single multi-spectral RapidEye image [GLCM: Gray Level Co-occurrence Matrix].

Features | Features per Image |
---|---|

Spectral image bands | 5 |

Indices | 9 |

GLCM-based textural features | 180 |

Total | 194 |

**Table 3.**Indices and formulas. Rapid-eye band name abbreviations are: R = red, RE = red edge, G = green, B = blue and NIR = near infrared.

Indices | Formula |
---|---|

Enhanced Vegetation Index (EVI) | 2.5 × (NIR − R)/(NIR +6 × R − 7.5 × B + 1) |

Transformed Chlorophyll Absorption in Reflectance Index (TCARI) | 3 × ((RE − R) − 0.2 × (RE − G) × (RE/R)) |

Soil Adjusted Vegetation Index (SAVI) | (1 + L) × (NIR − R)/(NIR + R + L), where L = 0.5 |

Modified Soil Adjusted Vegetation Index (MSAVI) | $0.5\ast (2\ast NIR+1-\sqrt{{\left(2\times NIR+1\right)}^{2}-8\times \left(NIR-R\right)})$ |

Visible Atmospherically Resistance Index (VARI) | (G − R)/(G + R − B) |

Green Leaf Index (GLI) | (2 × G − R − B)/(2 × G + R + B) |

Normalized Difference Vegetation Index (NDVI) | (NIR − R)/(NIR + R) |

Normalized Difference Water Index with green band (NDWI_{Green}) | (G − NIR)/(G + NIR) |

Normalized Difference Water Index with red band (NDWI_{Red}) | (R − NIR)/(R + NIR) |

**Table 4.**Comparing the built-up area classification of the GHSL with building footprints: (1) without aggregation and (2–4) with aggregation to the 38 m cell size.

(1) Building Level | (2) 50% Built up | (3) 10% Built up | (4) >0% Built up | |
---|---|---|---|---|

Overall Accuracy | 0.354 | 0.268 | 0.711 | 0.836 |

Kappa | 0.075 | 0.022 | 0.395 | 0.564 |

**Table 5.**Selected features per year [b1: band 1, b2: band 2, b3: band 3, b4: band 4, b5: band 5, asm: angular second moment, contrast: contrast, corr: correlation, var: variance, idm: inverse difference moment, savg: sum average, svar: sum variance, sent: sum entropy, ent: entropy, dvar: difference variance, dent: difference entropy, imcorr1: information measures of correlation 1, diss: dissimilarity, inertia: inertia, shade: cluster shade, and prom: cluster prominence].

Feature\Year | 2010 | 2016 | 2018 |
---|---|---|---|

Image bands | b2, b5 | b1, b3, b4, b5 | b1, b5 |

Indices | NDWI_{Red}, VARI | NDWI_{Green}, TCARI, VARI | GLI, NDWI_{Red}, TCARI, VARI |

Textural—3 × 3 | b3_inertia, b3_shade, b5_dvar, b5_imcorr1 | b1_sent, b2_prom, b3_savg, b4_savg, b5_var | |

Textural—5 × 5 | b2_dvar, b3_inertia, b3_savg, b3_var, b4_contrast, b4_dvar, b4_var b5_contrast, b5_savg, b5_dvar, b5_corr | b3_dvar, b3_savg, b1_savg, b2_idm, b4_dent, b4_dvar, b5_asm, b5_contrast, b5_corr, b5_dent, b5_dvar, b5_ent, b5_idm, b5_imcorr1, b5_prom, b5_savg, b5_sent, b5_svar | b1_asm, b1_shade, b2_dvar, b2_savg, b2_shade, b3_savg, b3_shade, b4_inertia, b4_shade, b5_asm, b5_corr, b5_dent, b5_diss, b5_idm, b5_imcorr1, b5_prom, b5_savg, b5_sent, b5_svar |

Total | 15 | 29 | 30 |

**Table 6.**Assessment of Overall Accuracy (OA) using four classes (built-up area, gray spaces, green spaces and water) and three classes (GHSL classes: built-up area, non-built-up area and water).

Year | 2010 | 2016 | 2018 |
---|---|---|---|

OA 4 classes | 0.78 | 0.83 | 0.73 |

OA 3 classes | 0.84 | 0.88 | 0.86 |

**Table 7.**Land cover statistics per year in hectares (ha), the indicator Built

_{Open}per year in ha, and the Built

_{Open}per capita/year in m

^{2}.

Year | 2010 | 2016 | 2018 |
---|---|---|---|

Built-up area | 9087.40 | 14,698.49 | 24,374.61 |

Green space | 63,608.26 | 54,998.31 | 48,984.30 |

Gray space | 18,263.67 | 21,230.23 | 17,600.43 |

Built_{Open} | 9.00 | 5.19 | 2.73 |

Built_{Open} per capita in m^{2} | 294.97 | 212.80 | 170.15 |

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## Share and Cite

**MDPI and ACS Style**

Aguilar, R.; Kuffer, M. Cloud Computation Using High-Resolution Images for Improving the SDG Indicator on Open Spaces. *Remote Sens.* **2020**, *12*, 1144.
https://doi.org/10.3390/rs12071144

**AMA Style**

Aguilar R, Kuffer M. Cloud Computation Using High-Resolution Images for Improving the SDG Indicator on Open Spaces. *Remote Sensing*. 2020; 12(7):1144.
https://doi.org/10.3390/rs12071144

**Chicago/Turabian Style**

Aguilar, Rosa, and Monika Kuffer. 2020. "Cloud Computation Using High-Resolution Images for Improving the SDG Indicator on Open Spaces" *Remote Sensing* 12, no. 7: 1144.
https://doi.org/10.3390/rs12071144