Next Article in Journal
Spatiotemporal Prediction of Monthly Sea Subsurface Temperature Fields Using a 3D U-Net-Based Model
Next Article in Special Issue
Using UAV Time Series to Estimate Landslides’ Kinematics Uncertainties, Case Study: Chirlești Earthflow, Romania
Previous Article in Journal
Winter Wheat Lodging Area Extraction Using Deep Learning with GaoFen-2 Satellite Imagery
Previous Article in Special Issue
Radiant Power Patterns Inferred from Remote Sensing Using a Cloud Computing Platform, during the 2021 Fagradalsfjall Eruption, Iceland
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
Remote Sensing
Volume 14
Issue 19
10.3390/rs14194888
remotesensing-logo
Submit to this Journal Review for this Journal Propose a Special Issue
► Article Menu

Article Menu

share Share announcement Help format_quote Cite question_answer Discuss in SciProfiles thumb_up
...
Endorse textsms
...
Comment
first_page
settings
Order Article Reprints
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Using GEOBIA and Vegetation Indices to Assess Small Urban Green Areas in Two Climatic Regions

by
Ana Maria Popa
1,2,
Diana Andreea Onose
1,2,*,
Ionut Cosmin Sandric
2,
Evangelos A. Dosiadis
3,
George P. Petropoulos
3,
Athanasios Alexandru Gavrilidis
1,2 and
Antigoni Faka
3
1
Centre for Environmental Research and Impact Studies, University of Bucharest, Bd. N. Balcescu 1, 010041 Bucharest, Romania
2
Faculty of Geography, University of Bucharest, Bd. N. Balcescu, 1, 010041 Bucharest, Romania
3
Department of Geography, Harokopio University of Athens, El. Venizelou 70, Kallithea, 17671 Athens, Greece
*
Author to whom correspondence should be addressed.
Remote Sens. 2022, 14(19), 4888; https://doi.org/10.3390/rs14194888
Submission received: 7 September 2022 / Revised: 26 September 2022 / Accepted: 27 September 2022 / Published: 30 September 2022
(This article belongs to the Special Issue Novel Methods and Applications in Satellite and Aerial Imagery Time Series Analysis)
Browse Figures
Versions Notes

Abstract

:
The importance of small urban green areas has increased in the context of rapid urbanization and the densification of the urban tissue. The analysis of these areas through remote sensing has been limited due to the low spatial resolution of freely available satellite images. We propose a timeseries analysis on 3 m resolution Planet images, using GEOBIA and vegetation indices, with the aim of extracting and assessing the quality of small urban green areas in two different climatic and biogeographical regions: temperate (Bucharest, Romania) and mediterranean (Athens, Greece). Our results have shown high accuracy (over 91%) regarding the extraction of small urban green areas in both cities across all the analyzed images. The timeseries analysis showed consistency with respect to location for around 55% of the identified surfaces throughout the entire period. The vegetation indices registered higher values in the temperate region due to the vegetation characteristics and city plan of the two cities. For the same reasons, the increase in the vegetation density and quality, as a result of the distance from the city center, and the decrease in the density of built-up areas, is more obvious in Athens. The proposed method provides valuable insights into the distribution and quality of small urban green areas at the city level and can represent the basis for many analyses, which is currently limited by poor spatial resolution.

Graphical Abstract

1. Introduction

The rapid urbanization that characterizes society today comes with two main trends in urban development: the densification of the urban area with the aim of creating compact cities and expansion into the peripheral areas, often through urban sprawl. Both models are currently being critically analyzed from the perspective of sustainable urban development, one of the main goals of the HABITAT Agenda [1] and The United Nations’s Sustainable Development Goals [2]. Cities are facing a wide range of challenges, which have an effect on environmental quality and the wellbeing of residents [3].
One of the main challenges to achieving sustainable, resilient, and inclusive cities [4] is ensuring the optimum amount of available and accessible high-quality urban green areas. The expansion of cities has increased the difficulty of reaching the existing green spaces from the new residential areas developed on the urban outskirts [5], while the increase in density has multiplied the pressures to which they are subjected [6]. Due to the lack of undeveloped public open spaces inside cities, many of the new larger green areas are located on the outskirts [7,8]. Therefore, small urban green spaces have become a viable solution for increasing the presence of vegetation in cities [9,10] since they are easier to plan than the larger spaces due to their surfaces and the diversity of existing approaches, both traditional (street alignments, residential gardens, and pocket parks) and modern (green roofs, green walls, and rain gardens).
Small urban green areas are patches covered by vegetation within cities. The accepted size for a small green area is debatable, with many studies focusing on the category rather than on the surface. Among the proposed threshold values, one of the most used is 0.5 ha [5,11], but depending on the object of the analysis or the legislative context, researchers have also opted for other limits. In Romania, for example, the Green Areas Law enforces a surface of a minimum of 1 ha in order for a green area to be acknowledged as a park and a maximum of 1 ha for an area to be regarded as a green square [12].
Small urban green spaces represent stepping stones in the green infrastructure [13] and provide cumulative benefits correlated with the network’s density. As with all urban green spaces, the small ones contribute to increasing environmental quality through aspects such as climate change mitigation [14], pollution and noise control, the regulation of the hydrological circuit, or the creation of habitats [15]. Moreover, they positively impact both physical and mental health [16] and help increase human wellbeing and the quality of life [17].
Research has mainly concentrated on large urban green areas [18], especially urban parks and forests, since they are very important elements at the urban level and provide a wide range of benefits. Research in the field of small urban green areas is still scarce and its distribution and quality have generally been neglected, even if as a whole, they represent a high percentage of the vegetation in cities. Studies focusing on small urban green areas usually focus on their use [5], their importance in terms of population health and the perception the population has in relation to their design [10], or associated benefits [19]. There is also a category of technical studies that focuses, in particular, on exploring the characteristics of modern, small urban green areas, such as green roofs [11], green walls, or rain gardens, and quantifying their input in terms of improving environmental quality.
Remote sensing is widely used in the assessment of urban green areas but it especially targets large areas [20], as they are easier to assess by means of automatic and semi-automatic methods. Thus far, research studies have mainly focused on the distribution and dynamics of urban green areas [21,22] and their characteristics, such as species composition [23] and the quality and health of vegetation in cities [24,25].
Remote-sensing analysis is conditioned by the resolution and characteristics of the available images, their spatial and temporal coverage, and access to data [18]. The most used images for the extraction of urban green areas are NASA’s Landsat [22,26] and ESA’s Sentinel [27,28] due to their long-term coverage. Yet, their spatial resolutions of 30 m and 10 m, respectively, do not support the accurate extraction of small patches of vegetation. New imagery developed over the last few years, such as Pleiades [28,29] and Planet, have allowed researchers to dive into the analysis of the distribution and quality of small urban green areas. Furthermore, high resolution images (under 1 m spatial resolution) such as IKONOS, GeoEye, and World View 3 and 4 are difficult to use on a large scale in research due to their costs.
The extraction and analysis of small green areas through remote sensing also faces methodological challenges. OBIA, one of the most widely used methods in the field, can create irregular shapes from the analyzed objects through the segmentation process [30]. To solve this problem, it has been suggested that an appropriate segmentation scale and a multi-temporal analysis of object-based classification should be used [31]. Another issue in the classification of small urban green areas is the presence of shadows in the imagery [18], since they very often occupy the areas near buildings, which cannot be properly seen. Hyperspectral images are recommended in this situation since different combinations of bands can help better distinguish between elements [32].
In a recent study, Shahtahmassebia et al. [18] performed a review regarding the remote sensing of urban green spaces and highlighted the need for a more detailed investigation of small urban green areas. They recommended developing a timeseries analysis and thematic applications, among others. Our study contributes to filling this gap by trying to respond to three research questions: (1) is the application of GEOBIA (Geographic Object-Based Image Analysis) to Planet images suitable for extracting small urban green areas, (2) is the method suitable irrespective of the biogeographical region to which it is applied, and (3) does the method provide valuable results in terms of the quality of small urban green areas regardless of the biogeographical region?
In order to respond to these questions, we tested the method on two cities located in different climatic and biogeographical regions: Bucharest (Romania) and Athens (Greece). In line with the proposed research questions, the study proposed two objectives:
(1)
to determine whether small urban green areas can be extracted by applying GEOBIA to Planet timeseries in two different biogeographical and climatic regions;
(2)
to compare the quality of small urban green areas in two different biogeographical and climatic regions.

2. Data and Methodology

2.1. Study Area

The proposed method was developed and tested using two European capitals as case studies: Bucharest, the capital of Romania, which is situated in a plain area, and Athens, the capital of Greece, located in a hollow area and surrounded by the Saronic Gulf (Aegean Sea) (Figure 1). Both cities are the largest in their respective countries, with Bucharest having 2 million inhabitants and an area of 240 km2 [33], while the number of inhabitants in the Greater Athens Area has reached 3.1 million in an area of 360 km2 [34].
The two study areas are characterized by different climatic conditions that influence the quality and quantity of urban green spaces. Bucharest has a temperate-continental climate with a transition effect [36], while Athens enjoys a typical Mediterranean climate, with hot, dry summers and mild, rainy winters (Table 1). Over the last few years, both areas have been affected by heatwaves. Both cities contain large, dense, built-up areas and are affected by urban sprawl and intense air pollution, which, combined with the destruction of peri-urban forests by wildfires or changes in land cover, have generated an urban heat island effect with an intensity as high as 10 °C in Athens [37], compared with an average of 5 °C in Bucharest [38].
The planning systems in which the cities developed have considerably influenced the surface and morphology of green areas. Bucharest is a mix of socialist neighborhoods, containing all-important public services (including green areas, both large and small); historical areas, usually represented by dense single-family residences; and modern projects, mainly represented by office spaces [41,42]. Over the past three decades, the city has experienced a chaotic development as a result of two key processes: urban sprawl and densification [43,44]. In contrast, Athens began to experience spontaneous, undesigned urban development in rural areas around its historical center starting in the 1920s, with the uncontrolled and unplanned outward expansion of the urban tissue continuing to this day [45,46]. Consequently, both cities have limited public open spaces.
According to the European Environment Agency [47], Bucharest and Athens possess a low percentage of urban green and blue spaces compared with the other European capitals and are ranked 31 and 36, respectively, out of the 37 analyzed cities. The green space network in the two cities is represented by both large green areas (such as parks, cemeteries, sports areas, and forests) and small urban green areas, such as pocket parks, residential and institution gardens, street alignments, and green roofs, which represent a significant part of the total. National statistics may differ slightly from the European figures, since each country regulates their green spaces differently. Romania, for example, does not include green roofs or playgrounds but does include cemeteries and sports areas [12], which may contain extensive, impervious surfaces. Greece regulates green spaces along with other public areas—such as sidewalks, bike paths, or playgrounds, which may not include green areas—and suburban green areas [48].

2.2. Extracting Small Urban Green Areas Using GEOBIA

For the present study, we used four-band Planet imagery with a 3 m spatial resolution [49] for the period from 2018–2020. We selected summer images retrieved between 1st June and 15th July when the vegetation season was at its height in the temperate region in the northern hemisphere where Bucharest is located. The evergreen character of the Mediterranean vegetation also made the timeframe suitable for Athens. We extracted small urban green areas using Geographic Object-Based Image Analysis (GEOBIA) [30,50,51]. In the segmentation step, the scale factor was chosen based on several trials with values between 18 and 20 [52,53]. Random forest classification was used to classify the objects obtained in the segmentation step into three main land use types (green, developed, and water) [29,54,55]. The validation of the obtained classification was performed using confusion matrix to calculate the overall accuracy and Kappa index [28,38]. We classified four images for each city in order to validate the method, with each image covering the whole urban area. All classified images had an overall accuracy higher than 91%, which is considered very good [27], and a Kappa index higher than 0.84 (Table 2).
As it was only required to maintain the small urban green areas in our database, we used the OSM 2021 data [56] to erase the land uses associated with large green areas, such as cemeteries, farmland, farmyards, forests, meadows, allotments, nature reserves, parks, recreation grounds, and scrubs from the obtained classifications. Moreover, we also deleted the areas classified as green outside the built-up limit, which were not identified as green by OSM, but also did not represent urban green areas. Afterwards, we used the roads from the same database to split the remaining green areas into parcels and deleted those with an area over 2 ha. The resulting dataset was used in the next steps of the analysis, which focused only on the small urban green spaces.

2.3. Vegetation Indices Used for Assessing the Quality of Small Urban Green Areas

To analyze the small urban green areas, we selected two widely used vegetation indices: the Normalized Difference Vegetation Index (NDVI) [57] and the Modified Soil Adjusted Vegetation Index 2 (MSAVI2) [58]. Considering the visible spectrum and near infrared provided by the Planet images [59], we calculated NDVI and MSAVI2 using the formula in Table 3. NDVI is an indicator of vegetation used for evaluating the abundance of vegetation and its characteristics, based on a scale from −1 to +1, where 0.2 is considered a threshold for impervious areas [60,61]. MSAVI2 is considered to provide better results than NDVI in terms of distinguishing different canopy structures, as well as in the case of early stages of vegetation or types of vegetation, which do not entirely cover the soil, even when they are fully developed [62]. Usually, these two indicators are used together to obtain better information [29]. In this study, they were used together to assess and understand whether the areas with grass or low height vegetation can be equally assessed by both indices. MSAVI2 is less influenced by the soil, as it is able to detect short grass present in different green areas, in contrast to NDVI, for which the size of the leaves has an important impact on vegetation detection. Due to objective considerations, such as the wavelength of the images, it was not possible to calculate indices referring to water, nitrogen, or carbon [63,64,65].
In the context of the current study, we refer to quality of small urban green areas from the point of view of the information provided by the two indicators: NDVI and MSAVI2. Therefore, through quality we understand the density and health of vegetation, since the assessed indicators are strongly correlated with photosynthetic activity, biomass, plant and soil moisture, and plant stress [66,67].
Table
Table 3. Vegetation indices used in the analysis [68].
Table 3. Vegetation indices used in the analysis [68].
Index CodeIndexFormula
MSAVI2Modified Soil Adjusted Vegetation Index 2 2 × N I R + 1 − ( 2 × ( N I R + 1 ) 2 − 8 ( N I R − R e d ) 2
NDVINormalized Difference Vegetation Index ( N I R − R e d ) ( N I R + R e d )
An optimized hot spot analysis was performed based on the Getis-Ord GI z-scores [69,70] for NDVI and MSAVI2 to identify statistically significant spatial clusters of high values (hot spots) and low values (cold spots) [71].

2.4. Factors Influencing the Quality of Small Urban Green Spaces

In order to understand the spatial variations in the quality of small urban green areas, as indicated by the two calculated indicators (NDVI and MSAVI2), we performed statistical analysis, which helped in quantifying the variations. Based on scientific literature, we selected three indicators that proved useful when assessing the characteristics of urban green spaces in order to assess the spatial distribution of the quality of small urban green areas. The three selected indicators were the distance from the main roads [65] (as defined in the OSM database), the distance from the city center [62], and built-up density [72] (calculated using OSM data for a fishnet of 100 × 100 m). The small urban green areas extracted in the previous step were clipped using the same fishnet and for each resulting patch (with a surface of maximum 1 ha) we calculated the distance from the closest main road, the city center, and the built-up density. For each patch, we also applied spatial statistics and calculated the z-score for NDVI and MSAVI2 [73].
After calculating the three indicators for each patch of small urban green area in the two cities, we identified the minimum and maximum values. For each indicator and each city, we created five classes using the equal interval method. For the built-up indicator, the values ranged from 0 (no built-up areas in the analyzed cell) to 1 (the cell was entirely occupied by the built-up area).
One-way ANOVA was performed using SPSS [74] in order to compare the effect of the three indicators on the quality of small urban green areas. The average values of NDVI and MSAVI2 per patch of small urban green area were used as dependent variables. As independent variables, we used the three indicators: distance from the green patch to the main roads—Class_roads, distance from the city center—Class_center, and built-up density—Class_built.
Furthermore, we tested the differences in average NDVI and MSAVI2 values between the five classes established for each indicator. We performed post hoc tests (Tukey HSD test) [75] in order to identify the different classes between which differences were registered. The calculation of the vegetation indexes, the zonal statistics, and the hot spot analysis were optimized by developing Python scripts.

3. Results

3.1. Distribution and Dynamics of Small Urban Green Areas

The analysis highlighted that a considerable share of the green areas in the cities are represented by small patches. The extraction of green features through GEOBIA revealed that, on average, 49% of Bucharest and 68% of Athens are covered by vegetation (including tree canopies overlapping built areas and infrastructure, agricultural land, protected areas, and abandoned surfaces covered by vegetation in the peripheral area of the cities). Between 36% (in the case of Athens) and 47% (in the case of Bucharest) of the surface identified as vegetation is represented by small urban green areas, which mainly include residential gardens (both public and private), pocket parks, street trees, and the gardens of institutions. Analyzing the dynamics of the small urban green areas between 2018 and 2020, we observed that, in Bucharest, these showed variations between 21% and 28% of the city’s surface with a slightly increasing trend (Figure 2a), while in Athens they registered a gradual decrease from 29% to 23% (Figure 2b).
The dynamics of the small urban green areas were analyzed by overlapping the four images for each city. In Bucharest, 37.11% of the surface of small urban green areas was identified as green on all four images (100% overlap), 18.57% overlaps on three images (>50% overlap), and 18.87% on two images (<50% overlap), while 25.44% of the green surface only appears on one of the classified images (Figure 3a). In Athens, 36.5% of the surface of the identified small urban green areas were distributed in this category in all four analyzed moments, while 15.67% are classified as green on three images (>50% overlap), 22.41% on two images (<50% overlap), and 25.41% appear on only one image (Figure 3b).
The percentage of overlaps highlights the certainty of the presence of small urban green areas in those locations. In the case of both cities, around 55% of the identified small green areas remain in this category almost across the entire analyzed timeframe. The areas with a high percentage of overlaps have different distributions across the two cities. Meanwhile, Bucharest has high overlaps mainly in the peripheral socialist neighborhoods, while in Athens, those areas are mostly located in the hills in the north-eastern part of the city where neighborhoods consisting of villas belonging to the higher class are located.

3.2. Quality of Small Urban Green Areas

The values of the NDVI and MSAVI2 highlight the characteristics and quality of small urban green spaces in the two cities. At the city level, Bucharest has higher values than Athens in the analyzed timeframe, even if the latter contains a larger surface of green areas; the same tendency exists when analyzing only the small urban green areas. The average values for the NDVI of the identified small urban green areas, calculated for the analyzed timeseries, are 0.53 for Bucharest and 0.29 for Athens; meanwhile, the average value of the MSAVI2 for the same surfaces is 0.69 for Bucharest and 0.44 for Athens. The highlighted tendency is explained by the different types of vegetation characterizing the two cities.
In the case of Bucharest, the average values of the NDVI, calculated for the small urban green areas, vary between 0.48 in 2018 and 0.59 in 2019. Meanwhile, Athens is characterized by smaller variations, with a minimum average of 0.26 in 2018 and a maximum average of 0.30 for the other three images. According to the registered values, the small urban green areas in Bucharest have denser and healthier vegetation than those in Athens. The values above 0.66 show that these small patches contain mature trees with a dense canopy, a situation that is characteristic of Bucharest, whereas those between 0.33 and 0.66 suggest the presence of bushes or scarcer vegetation, as may be found in Athens.
The highest values of the NDVI within small green patches in Bucharest are associated with the wealthier neighborhoods in the north, followed by some of the largest socialist neighborhoods in the city, Titan (in the east), Berceni (in the south), and Militari and Drumul Taberei (in the west) (Figure 4a). In Athens, the highest values are around 0.66 and characterize the north-eastern region of the city (Kifisia, Ekali), an area inhabited by the wealthier class (Figure 4b).
In Bucharest, large urban green areas (such as forests and parks) are characterized by higher values of the NDVI than small green patches throughout the entire analyzed timeframe. In 2019, we registered the highest values of the NDVI with 67% of the large urban green areas and 45% of the small urban green areas having values over 0.66. In Athens, few areas have NDVI values above 0.66, regardless of whether they are large or small.
The MSAVI2 average values, calculated for small urban green areas, varied between 0.64 in 2018 and 0.74 in 2019 in Bucharest and between 0.4 in 2018 and 0.46 in 2019 in Athens, with very similar values for 2020. Both the spatial distribution and the temporal dynamics of the MSAVI2 are similar to those of the NDVI. In Bucharest, almost all the surfaces identified as small green areas have average values over 0.6 (Figure 5a), indicating that the vegetation is sufficiently dense to cover the soil. However, in Athens, the registered average values for the MSAVI2 above 0.6 were between 6% in 2019 and 31% in July 2020. The majority of the small urban green areas in the city have average values between 0.4 and 0.6 (Figure 5b), highlighting the scarcity of Mediterranean vegetation. The spatial distribution of the values of the MSAVI2 reveal that both city centers lack appropriate green coverage.

3.3. Spatial Distribution of the Quality of Small Urban Green Areas

The Optimized Hot Spot Analysis showed the same spatial characteristics for both indicators and all four images analyzed for each city. The average values of the NDVI and MSAVI2 for small urban green areas showed that strong spatial clustering—marked as hot spots in the northern and western part of Bucharest—exists (Figure 6a), whereas the center and the eastern regions are identified as cold spots. In Athens, the clustering shows hot spots in the northern and eastern parts of the city (Figure 6b). In both cities, there is a tendency to create hot spots in areas near the forests (in the north in Bucharest and the north-east in Athens) and cold spots in the central areas. Small urban green areas tend to have higher average values when they are in the proximity of large green areas.
Testing the spatial distribution of the quality of small urban green areas highlighted similar tendencies for the NDVI and MSAVI2 in both cities. The ANOVA tests revealed significant statistical differences between the quality of small urban green areas (measured through the NDVI and MSAVI2) for all three tested indicators—distance from main roads, distance from the city center, and built-up density (Table 4)—in both cities. The only exception was the image of Bucharest from 2018, in which we were unable to identify significant differences between the NDVI classes in relation to the built-up surface.
The most relevant indicator was the distance of the small green areas from the center of the city. The Tukey HSD test showed significant differences between all the analyzed classes in seven of the eight images. The image for Bucharest from 2018 only registered significant differences between some of the five classes. In the case of Athens, the quality of the small green areas exhibited a significantly clear increase from the city center to the peripheral areas, with differences between 0.08 (2019) and 0.13 (July 2020) for the mean values of the NDVI and between 0.09 (2019) and 0.14 (July 2020) for those of the MSAVI2. Regarding Bucharest, the values are much closer together, with only class 2 (located between 2.5 and 4.7 km from the city center) registering significantly higher values. This class is essentially located outside the inner-city ring of Bucharest and overlaps the large socialist neighborhoods in the city.
As in the case of the first indicator, the built-up density is better illustrated in Athens than in Bucharest. The quality of the small urban green areas decreases as the built-up density increases; the differences are registered by the mean values for the five classes, varying between 0.11 (2018) and 0.17 (June 2020) for the NDVI and between 0.15 (2018) and 0.22 (June 2020) for the MSAVI2. The Tukey HSD test highlights that in Athens there are significant differences between all the classes except classes 4 (61–80% built-up) and 5 (81–100% built-up) with respect to the two images in 2020. For Bucharest, the results are similar with those registered for the distance from the city center, with small variations between the means of the five classes and no consistent spatial pattern.
In the case of the distance from the main roads, the results are similar for both cities. The post-hoc tests showed several classes with no significant differences between them, especially those closer to the main roads compared with those farthest away. In both cities, the highest-class averages for the NDVI and MSAVI2 were registered for the medium distances. Multiple comparisons showed that there is a correlation between the results obtained with the Optimized Hot Spot Analysis and the differences between groups.

4. Discussions

4.1. Method’s Efficiency for Extracting Small Urban Green Areas

GEOBIA is an efficient method that is widely used for extracting different types of land use at the city level, including urban green areas [76,77]. We further advanced this tendency and calibrated this method to efficiently extract the small urban green areas using high-resolution images. Our data showed a high level of accuracy when identifying green areas (over 91% across all analyzed images) in both the analyzed climatic and biogeographic regions, namely, mediterranean and temperate.
The small urban green areas we extracted from the images do not exactly represent the physical surface of the green areas, but the canopies for those that contain trees [18]. For this reason, combined with the resolution of the images and the fact that the selected method allows for the exclusion of artificial areas, which are included in green spaces [78] (e.g., large alleys in parks and roofs of sports areas), but also the inclusion of green areas that may not be considered as such in the legislation (e.g., informal green areas that may include, among others, green roofs or private residential gardens [7]), the resultant green areas cannot be compared with the official statistics. Instead, these should be comparable with the data provided by the Copernicus Land Monitoring Service and other similar databases since they use the same range of methods [38].
Image resolution is a positive aspect of our analysis. Sun et al. [79] demonstrated that urban green spaces can be successfully identified using images with a spatial resolution between 2 m and 16 m. With a 3 m resolution, Planet images are the best available satellite images and are freely provided for research by Planet. They offer the opportunity to study small urban green areas, which was not possible with other widely used types of satellite imagery, such as Landsat (30 m resolution) and Sentinel (10 m resolution). The research into urban green areas using Planet images is still in its infancy; however, the application of the latter is wide-ranging. For example, Pascual et al. [80] used these images in order to predict the risk of tree mortality in a tropical eucalypt forest in Brazil.
Our analysis showed that very small green areas (such as the green patches on roundabouts) and linear elements (such as street alignments) have lower validation scores than pocket parks or residential gardens. To analyze these accurately, there is a need for images with a better resolution, technologies that may not be publicly available, and a computational capacity that increases exponentially with the degree of detail of the analysis [81].
The use of timeseries ensures validation [82], especially when analyzing an element with high spatial dynamics such as small urban green areas. In both the above case studies, in more than 50% of the cases, small urban green areas are identified as such on at least three of the four images. The rest of the surface may be identified as small green areas only on some images, due to the technical factors or changes in land cover. Among the relevant technical factors, we highlight the quality of the images and the shadow effect, which especially affects the areas near high buildings where small green areas may be located [18]. Researchers have explored several methods in order to minimize this problem, such as the use of four masks (vegetation, height, shadow, and distance) [83].
Land cover change is a significant aspect when analyzing small urban green areas [82,84] since their surfaces and often unclear legal status make them vulnerable to transformations [44]. For example, in Bucharest, in the context of a volatile legislative framework, many small green areas have been transferred into private property and transformed in accordance with other land uses [85,86]. Moreover, there are areas that are not actively managed and, therefore, depending on the climatic conditions and the works that take place, these may be covered by ruderal vegetation [87], bare soil, or even urban waste. This might be the case in abandoned industrial or agricultural areas, brownfields, or even different types of gardens depending on their management [26]. On a smaller scale, the differences between the images may be attributable to various reasons, such as using artificial or natural grass on football fields, the construction or demolition of buildings, the clearing of grass on empty plots, and the pruning of trees and bushes, which are not performed to the same extent every year.
A particular case is the difference in the surface of small green areas registered in Bucharest between June and July 2020. Apart from the factors discussed above, which partially explain the difference, in this case, the climatic conditions are also relevant. Drought changes the shape and morphology of leaves and makes them interact differently with the electromagnetic spectrum, thereby diminishing the near infrared reflectance and, therefore, the surfaces identified as green. In order to obtain the best possible results when analyzing the dynamics of green areas in a certain region, one should use images acquired in similar climatic conditions (e.g., similar rainfall in the period before acquisition).

4.2. Insights into the Quality of Small Urban Green Areas

In line with prior research [88], our findings highlighted higher values of the NDVI in Bucharest, which has a temperate climate, than in Athens, which enjoys a Mediterranean climate. This aspect is related to both the biogeographical characteristics of the two regions and the planning decisions implemented in the cities. The vegetation in Athens is mainly represented by evergreen species, such as oak and cypress, and in the peripheral areas there are olive groves [89]. Many urban green spaces in the city contain oleander, olive, lemon, or orange trees. The vegetation in Bucharest mainly consists of deciduous species, such as linden, hornbeam, American maple, and platanus.
Mediterranean vegetation is adapted to cope with drought stress and has a low water content [89], which explains the lower levels of the NDVI in comparison with the vegetation in the temperate climate during summer. Meanwhile, evergreen vegetation has similar NDVI values over the year [90]; deciduous species in the temperate zone register maximum values during late spring and early summer when the biomass and photosynthesis are at their peaks [67].
Our results showed that both cities have neighborhoods with very well-developed small urban green areas (Figure 7), but their share within the city is very different. In Bucharest, the socialist neighborhoods (e.g., Drumul Taberei), which comprise the majority of the multi-family residential spaces, have very well-developed small green areas, mainly represented by residential gardens, pocket parks, and street alignments. These areas, which were planned during the socialist regime, along with the residential buildings they serve, have dense and well-developed vegetation (usually including mature trees that are 40–50 years old) rendering them NDVI hot spots at the city level. In contrast, multi-family residential areas in Athens (e.g., Kalithea) have low values with respect to the NDVI due to the scattered character of the vegetation that generates very low values of the NDVI (around 0.3). The single-family residential areas have comparable NDVI values in Bucharest and Athens, but in the case of the latter, these cover small areas, usually in the eastern and northeastern peripheries.
When assessing the quality of small green areas in relation to the distance from the city center, the distance from roads, and the built-up density, there were smaller differences between the NDVI and MSAVI2 classes’ averages in Bucharest compared to Athens. This is explained by the fact that in the urban core of Athens vegetation is very scarce and is dominated by individual trees; meanwhile, on the periphery, there are residential gardens (Figure 7) and pocket parks. Even if the spatial resolution of the utilized images is very good, the 9 sqm pixels cannot optimally represent individual trees, especially those with small crowns, as in the case of many species encountered in the small urban green areas in the Mediterranean region. Even if the pixel is identified as green, the intrusion of the artificial areas surrounding the trees will lower the registered NDVI value [91]. On the other hand, in Bucharest, the differences are smaller between the center and the periphery in terms of small green areas, their surfaces, and their characteristics being less contrasting.
In both Bucharest and Athens, the highest values of the NDVI and MSAVI2 were usually registered at a medium distance from the main roads. This phenomenon might be related to the poor air quality in the proximity of main roads and the lack of maintenance works in the peripheral areas, which are located a great distance from them.
Our results regarding the variations in the NDVI and MSAVI2 with respect to the built-up density in Athens are in line with those of Yang et al. [92], who found that in China, a more concentrated building density generates a poorer quality of urban green areas in similar natural conditions. In Bucharest, this relationship is much less significant, with fewer differences between the classes and no clear tendency. This might be explained by the characteristics of the temperate deciduous vegetation, which is often higher than the buildings—especially in single-family residential areas—and, therefore, visible. Nawar et al. [24] argue that the decrease in the surface of green areas is directly related to the decrease in their quality. Our analysis did not focus on the evolution of the quality of individual green patches. Instead, it highlighted that small urban green areas are characterized by lower values of the NDVI and MSAVI2 than large areas (associated mainly with parks and forests).

4.3. Influence of Climatic and Biogeographical Characteristics

We tested the method for extracting the small urban green areas in two cities from different climatic and biogeographical areas to ensure its higher potential of application and generalization. Considering the different climate zones of the analyzed cities, we focused on creating training samples specifically for each city. Even though the acquired images were from the same period, the texture and the spectral response were different across the two cities. In the case of Bucharest, vegetation in small patches was easy to observe, while for Athens this was more difficult, due to the low water content of the vegetation [93]. In Athens, there are areas covered by low vegetation, which may become drier during certain periods; therefore, their spectral response may lead to different results—the drier the vegetation, the more likely it is that these areas are classified as “developed”.
The identified surfaces of small green areas are related to the climatic conditions in the period prior to the analyzed images. In Bucharest, for example, the smallest surfaces of the small green areas, registered in June 2018 and July 2020, are associated with the lowest rainfall [92] and the highest average temperature at the time of analysis (Table 5). The variation in the NDVI and MSAVI2 values over the years is explained similarly, with the two indicators registering higher values related to denser and healthier vegetation in the years with higher rainfall. The values of the NDVI and MSAVI2, calculated at the patch level, register higher differences in Bucharest than in Athens along the analyzed timeseries. This may also be explained by the significant climatic variations [94] in Bucharest compared to those registered in Athens during the period before the images were acquired. The fact that the spatial tendency of the NDVI and MSAVI2 values remains stable in relation to the distance from the city center and the built-up density show the validity of the method no matter the yearly climatic variations.
The two indicators we selected for the analysis complement one another, especially in the case of the Mediterranean climate. In the case of Bucharest, the large surfaces with high values of the NDVI reduce the contribution of the MSAVI2. Virtually, all the surfaces identified as small green areas on all the images have the potential of developing dense vegetation. However, in Athens, the MSAVI2 better highlights the areas where there is a high potential for the development of dense vegetation and identifies the areas with herbaceous vegetation, which cover larger surfaces in this Mediterranean city.

5. Conclusions

The novelty of our research relates, in the first place, to the resolution of the utilized satellite images, which allows for the analyses of small urban green areas at the city level. Past research only focused on large natural or urban green areas, or on small scale case studies. Our study is one of the first to analyze small green areas at the urban level and provide a complex image of their distribution and quality. GEOBIA proved reliable in the analysis of small urban green areas and the use of timeseries improves its results. The method is easily replicable and an increased number of analyses on the subject would support the elaboration of a guideline to establish the suitable parameters for GEOBIA. The method rendered good results with respect to both the temperate and Mediterranean climatic regions, with the main uncertainties being related to the individual trees and street alignments, which require an even better spatial resolution, and to the intense dynamics of these areas. An advantage of the presented method is the possibility of including private green spaces in assessments, which are difficult to analyze through field methods.
Planet images support the calculation of some vegetation indices that can provide a general image of the state and quality of small urban green areas. Both the NDVI and MSAVI2 registered higher values in the temperate region due to the climatic and biogeographical characteristics supporting a greater vegetation density and water content. Testing the method in two different climatic regions proved its potential for generalization and revealed valuable insights in relation to the characteristics of small urban green areas. Future studies may target other climatic and biogeographical regions to ensure the validation of these results.
The proposed methodological framework represents the basis for a large number of applications that require an accurate, easy-to-implement method for extracting urban green areas. Such studies may relate to the assessment of cities’ sustainability, the quality of life in urban areas, and health and epidemiological studies.

Author Contributions

Conceptualization, D.A.O.; methodology, I.C.S., D.A.O. and G.P.P.; software, I.C.S. and A.M.P.; validation, A.M.P. and E.A.D.; formal analysis, A.M.P. and E.A.D.; data curation, A.M.P.; writing—original draft preparation, A.M.P., D.A.O. and E.A.D.; writing—review and editing, D.A.O., A.A.G., I.C.S., G.P.P. and A.F.; visualization, A.M.P. and A.A.G.; supervision, D.A.O.; project administration, D.A.O.; funding acquisition, D.A.O. All authors have read and agreed to the published version of the manuscript.

Funding

This research was funded by the Ministry of Research, Innovation and Digitization, CNCS-UEFISCDI (Romania), grant number PN-III-P1-1.1-TE-2019-1543, within PNCDI III, Contribution of small urban green infrastructure in achieving environmental justice (SmallGreen).

Acknowledgments

This work was supported by a grant of the Ministry of Research, Innovation and Digitization, CNCS/CCCDI-UEFISCDI, project number PN-III-P1-1.1-TE-2019-1543, within PNCDI III, Contribution of small urban green infrastructure in achieving environmental justice (SmallGreen).

Conflicts of Interest

The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript; or in the decision to publish the results.

References

  1. United Nations. Habitat III New Urban Agenda; United Nations: New York, NY, USA, 2017. [Google Scholar]
  2. United Nations Sustainable Development Goals. Available online: https://sustainabledevelopment.un.org/index.php?menu=1300 (accessed on 16 May 2019).
  3. Bush, J.; Doyon, A. Building Urban Resilience with Nature-Based Solutions: How Can Urban Planning Contribute? Cities 2019, 95, 102483. [Google Scholar] [CrossRef]
  4. United for Smart Sustainable Cities. Implementing Sustainable Development Goal 11 by Connecting Sustainability Policies and Urban-Planning Practices through ICTs; United Nations: New York, NY, USA, 2017. [Google Scholar]
  5. Peschardt, K.K.; Schipperijn, J.; Stigsdotter, U.K. Use of Small Public Urban Green Spaces (SPUGS). Urban For. Urban Green. 2012, 11, 235–244. [Google Scholar] [CrossRef]
  6. Popa, A.-M.; Ioja, I.-C.; Nita, M.-R.; Onose, D.-A. Urban Sustainability Assessment of Romanian Cities. Carpathian J. Earth Environ. Sci. 2022, 17, 373–385. [Google Scholar]
  7. Huang, Y.; Lin, T.; Zhang, G.; Jones, L.; Xue, X.; Ye, H.; Liu, Y. Spatiotemporal Patterns and Inequity of Urban Green Space Accessibility and Its Relationship with Urban Spatial Expansion in China during Rapid Urbanization Period. Sci. Total Environ. 2022, 809, 151123. [Google Scholar] [CrossRef] [PubMed]
  8. Siddique, S.; Uddin, M.M. Green Space Dynamics in Response to Rapid Urbanization: Patterns, Transformations and Topographic Influence in Chattogram City, Bangladesh. Land Use Policy 2022, 114, 105974. [Google Scholar] [CrossRef]
  9. Picard, P.M.; Tran, T.T.H. Small Urban Green Areas. J. Environ. Econ. Manag. 2021, 106, 102418. [Google Scholar] [CrossRef]
  10. Naghibi, M.; Faizi, M.; Ekhlassi, A. Design Possibilities of Leftover Spaces as a Pocket Park in Relation to Planting Enclosure. Urban For. Urban Green. 2021, 64, 127273. [Google Scholar] [CrossRef]
  11. Zhang, X.; Ni, Z.; Wang, Y.; Chen, S.; Xia, B. Public Perception and Preferences of Small Urban Green Infrastructures: A Case Study in Guangzhou, China. Urban For. Urban Green. 2020, 53, 126700. [Google Scholar] [CrossRef]
  12. Parlamentul Romaniei. Legea 24/2007; Parlamentul Romaniei: Bucharest, Romania, 2012. [Google Scholar]
  13. Nordh, H.; Østby, K. Pocket Parks for People—A Study of Park Design and Use. Urban For. Urban Green. 2013, 12, 12–17. [Google Scholar] [CrossRef]
  14. Park, J.; Kim, J.H.; Lee, D.K.; Park, C.Y.; Jeong, S.G. The Influence of Small Green Space Type and Structure at the Street Level on Urban Heat Island Mitigation. Urban For. Urban Green. 2017, 21, 203–212. [Google Scholar] [CrossRef]
  15. Strohbach, M.W.; Lerman, S.B.; Warren, P.S. Are Small Greening Areas Enhancing Bird Diversity? Insights from Community-Driven Greening Projects in Boston. Landsc. Urban Plan. 2013, 114, 69–79. [Google Scholar] [CrossRef]
  16. Lin, W.; Chen, Q.; Jiang, M.; Zhang, X.; Liu, Z.; Tao, J.; Wu, L.; Xu, S.; Kang, Y.; Zeng, Q. The Effect of Green Space Behaviour and per Capita Area in Small Urban Green Spaces on Psychophysiological Responses. Landsc. Urban Plan. 2019, 192, 103637. [Google Scholar] [CrossRef]
  17. Kolokotsa, D.; Lilli, A.; Lilli, M.A.; Nikolaidis, N.P. On the Impact of Nature-Based Solutions on Citizens’ Health & Well Being. Energy Build. 2020, 229, 110527. [Google Scholar] [CrossRef]
  18. Shahtahmassebia, A.R.; Chenlu, L.; Yifan, F.; Yani, W.; Yue, I.; Muye, G.; Ke, W.; Arunima, M.; George Alan, B. Remote Sensing of Urban Green Spaces: A Review. Urban For. Urban Green. 2021, 57, 126946. [Google Scholar] [CrossRef]
  19. Peschardt, K.K.; Stigsdotter, U.K. Associations between Park Characteristics and Perceived Restorativeness of Small Public Urban Green Spaces. Landsc. Urban Plan. 2013, 112, 26–39. [Google Scholar] [CrossRef]
  20. Petropoulos, G.P.; Kalivas, D.P.; Georgopoulou, I.A.; Srivastava, P.K. Urban Vegetation Cover Extraction from Hyperspectral Imagery and Geographic Information System Spatial Analysis Techniques: Case of Athens, Greece. J. Appl. Remote Sens. 2015, 9, 096088. [Google Scholar] [CrossRef]
  21. Kuklina, V.; Sizov, O.; Fedorov, R. Green Spaces as an Indicator of Urban Sustainability in the Arctic Cities: Case of Nadym. Polar Sci. 2021, 29, 100672. [Google Scholar] [CrossRef]
  22. Wellmann, T.; Schug, F.; Haase, D.; Pflugmacher, D.; van der Linden, S. Green Growth? On the Relation between Population Density, Land Use and Vegetation Cover Fractions in a City Using a 30-Years Landsat Time Series. Landsc. Urban Plan. 2020, 202, 103857. [Google Scholar] [CrossRef]
  23. Pu, R.; Landry, S. A Comparative Analysis of High Spatial Resolution IKONOS and WorldView-2 Imagery for Mapping Urban Tree Species. Remote Sens. Environ. 2012, 124, 516–533. [Google Scholar] [CrossRef]
  24. Nawar, N.; Sorker, R.; Chowdhury, F.J.; Mostafizur Rahman, M. Present Status and Historical Changes of Urban Green Space in Dhaka City, Bangladesh: A Remote Sensing Driven Approach. Environ. Chall. 2022, 6, 100425. [Google Scholar] [CrossRef]
  25. Abutaleb, K.; Mudede, M.F.; Nkongolo, N.; Newete, S.W. Estimating Urban Greenness Index Using Remote Sensing Data: A Case Study of an Affluent vs Poor Suburbs in the City of Johannesburg. Egypt. J. Remote Sens. Sp. Sci. 2021, 24, 343–351. [Google Scholar] [CrossRef]
  26. Grădinaru, S.R.; Kienast, F.; Psomas, A. Using Multi-Seasonal Landsat Imagery for Rapid Identification of Abandoned Land in Areas Affected by Urban Sprawl. Ecol. Indic. 2019, 96, 79–86. [Google Scholar] [CrossRef]
  27. Kranjčić, N.; Medak, D.; Župan, R.; Rezo, M. Support Vector Machine Accuracy Assessment for Extracting Green Urban Areas in Towns. Remote Sens. 2019, 11, 655. [Google Scholar] [CrossRef]
  28. Lebourgeois, V.; Dupuy, S.; Vintrou, É.; Ameline, M.; Butler, S.; Bégué, A. A Combined Random Forest and OBIA Classification Scheme for Mapping Smallholder Agriculture at Different Nomenclature Levels Using Multisource Data (Simulated Sentinel-2 Time Series, VHRS and DEM). Remote Sens. 2017, 9, 259. [Google Scholar] [CrossRef]
  29. Zylshal; Sulma, S.; Yulianto, F.; Nugroho, J.T.; Sofan, P. A Support Vector Machine Object Based Image Analysis Approach on Urban Green Space Extraction Using Pleiades-1A Imagery. Model. Earth Syst. Environ. 2016, 2, 54. [Google Scholar] [CrossRef]
  30. Ma, L.; Schmitt, M.; Zhu, X. Uncertainty analysis of object-based land-cover classification using Sentinel-2 time-series data. Remote Sens. 2020, 12, 3798. [Google Scholar] [CrossRef]
  31. Baker, F.; Smith, G.R.; Marsden, S.J.; Cavan, G. Mapping Regulating Ecosystem Service Deprivation in Urban Areas: A Transferable High-Spatial Resolution Uncertainty Aware Approach. Ecol. Indic. 2021, 121, 107058. [Google Scholar] [CrossRef]
  32. Yang, Z.; Willis, P.; Mueller, R. Impact of Band-Ratio Enhanced AWiFS Image to Crop Classification Accuracy. In Proceedings of the Pecora 17—The Future of Land Imaging… Going Operational, Denver, CO, USA, 18–20 November 2008; Volume 17, pp. 1–11. [Google Scholar]
  33. INS Baze de Date Statistice. Available online: http://statistici.insse.ro:8077/tempo-online/#/pages/tables/insse-table (accessed on 26 November 2020).
  34. Hellenic Statistical Authority. Statistical Database; Hellenic Statistical Authority: Piraeus, Greece, 2020. [Google Scholar]
  35. Atlas, U. Copernicus Land Monitoring Service; European Environment Agency (EEA): Copenhagen, Denmark, 2018. [Google Scholar]
  36. Dumitrescu, E. Clima Municipiului București; Editura Ars Docendi: Bucharest, Romania, 2007. [Google Scholar]
  37. Gaitani, N.; Spanou, A.; Saliari, M.; Synnefa, A.; Vassilakopoulou, K.; Papadopoulou, K.; Pavlou, K.; Santamouris, M.; Papaioannou, M.; Lagoudaki, A. Improving the Microclimate in Urban Areas: A Case Study in the Centre of Athens. Build. Serv. Eng. Res. Technol. 2011, 32, 53–71. [Google Scholar] [CrossRef]
  38. Cheval, S.; Popa, A.-M.; Șandric, I.; Iojă, I.-C. Exploratory Analysis of Cooling Effect of Urban Lakes on Land Surface Temperature in Bucharest (Romania) Using Landsat Imagery. Urban Clim. 2020, 34, 100696. [Google Scholar] [CrossRef]
  39. ANM. Geografie; ANM: Bucharest, Romania, 2018. [Google Scholar]
  40. Hellenic National Meteorological Service. Climatic Data for Nea Filadefia Station; Hellenic National Meteorological Service: Hellinikon, Greece, 2010. [Google Scholar]
  41. Nae, M.; Dumitrache, L.; Suditu, B.; Matei, E. Housing Activism Initiatives and Land-Use Conflicts: Pathways for Participatory Planning and Urban Sustainable Development in Bucharest City, Romania. Sustainability 2019, 11, 6211. [Google Scholar] [CrossRef]
  42. Ianoş, I.; Sorensen, A.; Merciu, C. Incoherence of Urban Planning Policy in Bucharest: Its Potential for Land Use Conflict. Land Use Policy 2017, 60, 101–112. [Google Scholar] [CrossRef]
  43. Iojă, I.C.; Osaci-Costache, G.; Breuste, J.; Hossu, C.A.; Grădinaru, S.R.; Onose, D.A.; Nită, M.R.; Skokanová, H. Integrating Urban Blue and Green Areas Based on Historical Evidence. Urban For. Urban Green. 2018, 34, 217–225. [Google Scholar] [CrossRef]
  44. Badiu, D.L.; Onose, D.A.; Niță, M.R.; Lafortezza, R. From “Red” to Green? A Look into the Evolution of Green Spaces in a Post-Socialist City. Landsc. Urban Plan. 2019, 187, 156–164. [Google Scholar] [CrossRef]
  45. Chorianopoulos, I.; Pagonis, T.; Koukoulas, S.; Drymoniti, S. Planning, Competitiveness and Sprawl in the Mediterranean City: The Case of Athens. Cities 2010, 27, 249–259. [Google Scholar] [CrossRef]
  46. Kassomenos, P.; Kissas, G.; Petrou, I.; Begou, P.; Khan, H.S.; Santamouris, M. The Influence of Daily Weather Types on the Development and Intensity of the Urban Heat Island in Two Mediterranean Coastal Metropolises. Sci. Total Environ. 2022, 819, 153071. [Google Scholar] [CrossRef]
  47. European Environment Agency. How Green Are European Cities? Green Space Key to Well-Being—But Access Varies; European Environment Agency: Copenhagen, Denmark, 2022. [Google Scholar]
  48. Greek Ministry of Environment and Energy. Law 4280/2014: Environmental Upgrade and Private Urban Planning—Sustainable Development of Settlements—Regulations of Forest Legislation and Other Provisions; Greek Ministry of Environment and Energy: Athens, Greece, 2014. [Google Scholar]
  49. Planet Team. Planet Application Program Interface: In Space for Life on Earth; Planet Team: San Francisco, CA, USA, 2021. [Google Scholar]
  50. Timilsina, S.; Aryal, J.; Kirkpatrick, J.B. Mapping Urban Tree Cover Changes Using Object-Based Convolution Neural Network (OB-CNN). Remote Sens. 2020, 12, 3017. [Google Scholar] [CrossRef]
  51. Degerickx, J.; Hermy, M.; Somers, B. Mapping Functional Urban Green Types Using High Resolution Remote Sensing Data. Sustainability 2020, 12, 2144. [Google Scholar] [CrossRef]
  52. Gulcin, D.; Akpınar, A. Mapping Urban Green Spaces Based on an Object-Oriented Approach. Bilge Int. J. Sci. Technol. Res. 2018, 2, 71–81. [Google Scholar] [CrossRef]
  53. Blaschke, T. Object Based Image Analysis for Remote Sensing. ISPRS J. Photogramm. Remote Sens. 2010, 65, 2–16. [Google Scholar] [CrossRef]
  54. Puissant, A.; Rougiera, S.; Stumpf, A. Object-Oriented Mapping of Urban Trees Using Random Forest Classifiers. Int. J. Appl. Earth Obs. Geoinf. 2014, 26, 235–245. [Google Scholar] [CrossRef]
  55. Belgiu, M.; Drăgu, L. Random Forest in Remote Sensing: A Review of Applications and Future Directions. ISPRS J. Photogramm. Remote Sens. 2016, 114, 24–31. [Google Scholar] [CrossRef]
  56. OpenStreetMap Contributors Roads, Buildings, Land Use. Available online: https://planet.openstreetmap.org (accessed on 15 October 2021).
  57. Rouse, J.W.; Haas, R.H.; Schell, J.A.; Deering, D.W. Monitoring Vegetation Systems in the Great Plains with ERTS. In Proceedings of the Third ERTS Symposium, Washington, DC, USA, 1–14 December 1973; NASA: Washington, DC, USA, 1974; pp. 309–317. [Google Scholar]
  58. Qi, J.; Chehbouni, A.; Huete, A.R.; Kerr, Y.H.; Sorooshian, S. A Modified Soil Adjusted Vegetation Index. Remote Sens. Environ. 1994, 48, 119–126. [Google Scholar] [CrossRef]
  59. Planet Team. Planet Imagery Product Specification; Planet Team: San Francisco, CA, USA, 2022. [Google Scholar]
  60. Feng, S.; Fan, F. A Hierarchical Extraction Method of Impervious Surface Based on NDVI Thresholding Integrated with Multispectral and High-Resolution Remote Sensing Imageries. IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens. 2019, 12, 1461–1470. [Google Scholar] [CrossRef]
  61. Hashim, H.; Latif, A.Z.; Adnan, N.A. Urban Vegetation Classification with NDVI Threshold Value Method with Very High Resolution (VHR) PLEIADES Imagery. Environ. Sci. 2019, 42, 237–240. [Google Scholar] [CrossRef]
  62. Fabijańczyk, P.; Zawadzki, J. Spatial Correlations of NDVI and MSAVI2 Indices of Green and Forested Areas of Urban Agglomeration, Case Study Warsaw, Poland. Remote Sens. Appl. Soc. Environ. 2022, 26, 100721. [Google Scholar] [CrossRef]
  63. Verrelst, J.; Rivera-Caicedo, J.P.; Reyes-Muñoz, P.; Morata, M.; Amin, E.; Tagliabue, G.; Panigada, C.; Hank, T.; Berger, K. Mapping Landscape Canopy Nitrogen Content from Space Using PRISMA Data. ISPRS J. Photogramm. Remote Sens. 2021, 178, 382–395. [Google Scholar] [CrossRef]
  64. Erener, A. Remote Sensing of Vegetation Health for Reclaimed Areas of Seyitömer Open Cast Coal Mine. Int. J. Coal Geol. 2011, 86, 20–26. [Google Scholar] [CrossRef]
  65. Cârlan, I.; Mihai, B.A.; Nistor, C.; Große-Stoltenberg, A. Identifying Urban Vegetation Stress Factors Based on Open Access Remote Sensing Imagery and Field Observations. Ecol. Inform. 2020, 55, 101032. [Google Scholar] [CrossRef]
  66. Calderón-Contreras, R.; Quiroz-Rosas, L.E. Analysing Scale, Quality and Diversity of Green Infrastructure and the Provision of Urban Ecosystem Services: A Case from Mexico City. Ecosyst. Serv. 2017, 23, 127–137. [Google Scholar] [CrossRef]
  67. Wong, C.Y.S.; D’Odorico, P.; Bhathena, Y.; Arain, M.A.; Ensminger, I. Carotenoid Based Vegetation Indices for Accurate Monitoring of the Phenology of Photosynthesis at the Leaf-Scale in Deciduous and Evergreen Trees. Remote Sens. Environ. 2019, 233, 111407. [Google Scholar] [CrossRef]
  68. L3Harris Geospatial Broadband Greenness Vegetation Indexes. Available online: https://www.l3harrisgeospatial.com/docs/broadbandgreenness.html#Green7 (accessed on 3 October 2021).
  69. Getis, A.; Ord, J.K. The Analysis of Spatial Association by Use of Distance Statistics. Geogr. Anal. 1992, 24, 189–206. [Google Scholar] [CrossRef]
  70. Ord, J.K.; Getis, A. Local Spatial Autocorrelation Statistics: Distributional Issues and an Application. Geogr. Anal. 1995, 27, 286–306. [Google Scholar] [CrossRef]
  71. Esri Inc. Optimized Hot Spot Analysis (Spatial Statistics)—ArcGIS Pro | Documentation. Available online: https://pro.arcgis.com/en/pro-app/latest/tool-reference/spatial-statistics/optimized-hot-spot-analysis.htm (accessed on 3 March 2022).
  72. Zheng, Y.; Tang, L.; Wang, H. An Improved Approach for Monitoring Urban Built-up Areas by Combining NPP-VIIRS Nighttime Light, NDVI, NDWI, and NDBI. J. Clean. Prod. 2021, 328, 129488. [Google Scholar] [CrossRef]
  73. Esri Inc. Zonal Statistics as Table (Spatial Analyst)—ArcGIS Pro. Documentation. Available online: https://pro.arcgis.com/en/pro-app/latest/tool-reference/spatial-analyst/zonal-statistics-as-table.htm (accessed on 3 March 2022).
  74. IBM Corporation. IBM SPSS Statistics for Windows; IBM Corporation: Armonk, NY, USA, 2013. [Google Scholar]
  75. Jackson, P.H.; Ferguson, G.A. Statistical Analysis in Psychology and Education. J. R. Stat. Soc. Ser. A 1972, 135, 153. [Google Scholar] [CrossRef]
  76. Hossain, M.D.; Chen, D. Segmentation for Object-Based Image Anaalysis (OBIA): A Review of Algorithms and Challenges from Remote Sensing Perspective. ISPRS J. Photogramm. Remote Sens. 2019, 150, 115–134. [Google Scholar] [CrossRef]
  77. Bartesaghi-Koc, C.; Osmond, P.; Peters, A. Mapping and Classifying Green Infrastructure Typologies for Climate-Related Studies Based on Remote Sensing Data. Urban For. Urban Green. 2019, 37, 154–167. [Google Scholar] [CrossRef]
  78. Liu, L.; Coops, N.C.; Aven, N.W.; Pang, Y. Mapping Urban Tree Species Using Integrated Airborne Hyperspectral and LiDAR Remote Sensing Data. Remote Sens. Environ. 2017, 200, 170–182. [Google Scholar] [CrossRef]
  79. Sun, Y.; Meng, Q.; Sun, Z.; Zhang, J.; Zhang, L. Assessing the Impacts of Grain Sizes on Landscape Pattern of Urban Green Space. In Proceedings of the AOPC 2017: Optical Sensing and Imaging Technology and Applications, Beijing, China, 4–6 June 2017; SPIE: Bellingham, WA, USA, 2017; Volume 10462. [Google Scholar] [CrossRef]
  80. Pascual, A.; Tupinambá-Simões, F.; Guerra-Hernández, J.; Bravo, F. High-Resolution Planet Satellite Imagery and Multi-Temporal Surveys to Predict Risk of Tree Mortality in Tropical Eucalypt Forestry. J. Environ. Manag. 2022, 310, 114804. [Google Scholar] [CrossRef]
  81. Bello, I.M.; Zhang, K.; Su, Y.; Wang, J.; Aslam, M.A. Densely Multiscale Framework for Segmentation of High Resolution Remote Sensing Imagery. Comput. Geosci. 2022, 167, 105196. [Google Scholar] [CrossRef]
  82. Wu, W.B.; Ma, J.; Meadows, M.E.; Banzhaf, E.; Huang, T.Y.; Liu, Y.F.; Zhao, B. Spatio-Temporal Changes in Urban Green Space in 107 Chinese Cities (1990–2019): The Role of Economic Drivers and Policy. Int. J. Appl. Earth Obs. Geoinf. 2021, 103, 102525. [Google Scholar] [CrossRef]
  83. Aval, J.; Demuynck, J.; Zenou, E.; Fabre, S.; Sheeren, D.; Fauvel, M.; Adeline, K.; Briottet, X. Detection of Individual Trees in Urban Alignment from Airborne Data and Contextual Information: A Marked Point Process Approach. ISPRS J. Photogramm. Remote Sens. 2018, 146, 197–210. [Google Scholar] [CrossRef]
  84. Morar, C.; Lukić, T.; Valjarević, A.; Niemets, L.; Kostrikov, S.; Sehida, K.; Telebienieva, I.; Kliuchko, L.; Kobylin, P.; Kravchenko, K. Spatiotemporal Analysis of Urban Green Areas Using Change Detection: A Case Study of Kharkiv, Ukraine. Front. Environ. Sci. 2022, 10, 30. [Google Scholar] [CrossRef]
  85. Kronenberg, J.; Haase, A.; Łaszkiewicz, E.; Antal, A.; Baravikova, A.; Biernacka, M.; Dushkova, D.; Filčak, R.; Haase, D.; Ignatieva, M.; et al. Environmental Justice in the Context of Urban Green Space Availability, Accessibility, and Attractiveness in Postsocialist Cities. Cities 2020, 106, 102862. [Google Scholar] [CrossRef]
  86. Onose, D.A.; Iojă, I.C.; Niță, M.R.; Badiu, D.L.; Hossu, C.A. Green Struggle—Environmental Conflicts Involving Urban Green Areas in Bucharest City. In Making Green Cities—Concepts, Challenges and Practice; Breuste, J., Artmann, M., Ioja, I.C., Qureshi, S., Eds.; Springer: Berlin/Heidelberg, Germany, 2020. [Google Scholar]
  87. Włodarczyk-Marciniak, R.; Sikorska, D.; Krauze, K. Residents’ Awareness of the Role of Informal Green Spaces in a Post-Industrial City, with a Focus on Regulating Services and Urban Adaptation Potential. Sustain. Cities Soc. 2020, 59, 102236. [Google Scholar] [CrossRef] [PubMed]
  88. Wang, C.; Ren, Z.; Dong, Y.; Zhang, P.; Guo, Y.; Wang, W.; Bao, G. Efficient Cooling of Cities at Global Scale Using Urban Green Space to Mitigate Urban Heat Island Effects in Different Climatic Regions. Urban For. Urban Green. 2022, 74, 127635. [Google Scholar] [CrossRef]
  89. Khan, A.; Papazoglou, E.G.; Cartalis, C.; Philippopoulos, K.; Vasilakopoulou, K.; Santamouris, M. On the Mitigation Potential and Urban Climate Impact of Increased Green Infrastructures in a Coastal Mediterranean City. Build. Environ. 2022, 221, 109264. [Google Scholar] [CrossRef]
  90. Volpi, I.; Marchi, S.; Petacchi, R.; Hoxha, K.; Guidotti, D. Detecting Olive Grove Abandonment with Sentinel-2 and Machine Learning: The Development of a Web-Based Tool for Land Management. Smart Agric. Technol. 2023, 3, 100068. [Google Scholar] [CrossRef]
  91. Stateras, D.; Kalivas, D. Assessment of Olive Tree Canopy Characteristics and Yield Forecast Model Using High Resolution Uav Imagery. Agriculture 2020, 10, 385. [Google Scholar] [CrossRef]
  92. Yang, Z.; Fang, C.; Mu, X.; Li, G.; Xu, G. Urban Green Space Quality in China: Quality Measurement, Spatial Heterogeneity Pattern and Influencing Factor. Urban For. Urban Green. 2021, 66, 127381. [Google Scholar] [CrossRef]
  93. Neinavaz, E.; Skidmore, A.K.; Darvishzadeh, R.; Groen, T.A. Retrieving Vegetation Canopy Water Content from Hyperspectral Thermal Measurements. Agric. For. Meteorol. 2017, 247, 365–375. [Google Scholar] [CrossRef]
  94. Maselli, F.; Cherubini, P.; Chiesi, M.; Gilabert, M.A.; Lombardi, F.; Moreno, A.; Teobaldelli, M.; Tognetti, R. Start of the Dry Season as a Main Determinant of Inter-Annual Mediterranean Forest Production Variations. Agric. For. Meteorol. 2014, 194, 197–206. [Google Scholar] [CrossRef]
  95. Raspisaniye Pogodi Weather. Available online: https://rp5.ru (accessed on 1 September 2022).
Remotesensing 14 04888 g001 550
Figure 1. Location of the case studies and main land use categories, based on Urban Atlas data (Data from [35]).
Figure 1. Location of the case studies and main land use categories, based on Urban Atlas data (Data from [35]).
Remotesensing 14 04888 g001
Remotesensing 14 04888 g002 550
Figure 2. Variations of small urban green areas per image for Bucharest (a) and Athens (b).
Figure 2. Variations of small urban green areas per image for Bucharest (a) and Athens (b).
Remotesensing 14 04888 g002
Remotesensing 14 04888 g003 550
Figure 3. Small urban green area distribution for Bucharest (a) and Athens (b), from 2018–2020.
Figure 3. Small urban green area distribution for Bucharest (a) and Athens (b), from 2018–2020.
Remotesensing 14 04888 g003
Remotesensing 14 04888 g004 550
Figure 4. NDVI distribution of small urban green areas for July 2020 in Bucharest (a) and Athens (b).
Figure 4. NDVI distribution of small urban green areas for July 2020 in Bucharest (a) and Athens (b).
Remotesensing 14 04888 g004
Remotesensing 14 04888 g005 550
Figure 5. MSAVI2 distribution of small urban green areas for July 2020 in Bucharest (a) and Athens (b).
Figure 5. MSAVI2 distribution of small urban green areas for July 2020 in Bucharest (a) and Athens (b).
Remotesensing 14 04888 g005
Remotesensing 14 04888 g006 550
Figure 6. Optimized hot spot distribution of NDVI for July 2020 in Bucharest (a) and Athens (b).
Figure 6. Optimized hot spot distribution of NDVI for July 2020 in Bucharest (a) and Athens (b).
Remotesensing 14 04888 g006
Remotesensing 14 04888 g007 550
Figure 7. Sample of small green area with high NDVI values in Bucharest and Athens.
Figure 7. Sample of small green area with high NDVI values in Bucharest and Athens.
Remotesensing 14 04888 g007
Table
Table 1. Climatic characteristics of Bucharest and Athens.
Table 1. Climatic characteristics of Bucharest and Athens.
CityAnnual Mean TemperatureTemperature AmplitudeMinimum Monthly Average TemperatureMaximum Monthly Average TemperatureAnnual Average Amount of PrecipitationReference
Bucharest10.5 °C26 °C−3 °C in January23 °C in July585 mmANM [39]
Athens17.8 °C19.5 °C8.8 °C in January28.3 °C in July411.8 mmHellenic National Meteorological Service [40]
Table
Table 2. Accuracy results for GEOBIA classification at city level.
Table 2. Accuracy results for GEOBIA classification at city level.
CityAcquisition DateOverall Accuracy (%)Kappa
Athens13 June 201894.800.9020
08 July 201991.800.8470
21 June 202093.400.8783
22 July 202092.870.8662
Bucharest10 June 201896.200.9267
13 June 201993.200.8739
26 June 202095.000.9080
15 July 202093.000.8700
Table
Table 4. ANOVA test results for mean values of NDVI and MSAVI2 for Bucharest and Athens.
Table 4. ANOVA test results for mean values of NDVI and MSAVI2 for Bucharest and Athens.
DateDf between GroupsDf within GroupsF-Statistic/
p-Value		Class_RoadsClass_CenterClass_Built
NDVIMSAVI2NDVIMSAVI2NDVIMSAVI2
Athens
13 June 20184223966F357.03348.714778.364580.194203.574347.13
p0.000.000.000.000.000.00
8 July 20194149844F428.16413.182975.232984.122841.543006.98
p0.000.000.000.000.000.00
21 June 20204181283F385.94369.835220.605085.264531.014743.90
p0.000.000.000.000.000.00
22 July 20204242108F180.10203.383548.113484.793789.403897.32
p0.000.000.000.000.000.00
Bucharest
10 June 2018491753F34.3263.795.83256.521.3442.52
p0.000.000.000.000.250.00
13 June 2019492611F32.7532.91340.22351.81144.91126.52
p0.000.000.000.000.000.00
26 June 20204107680F26.8726.24230.55217.16116.02107.57
p0.000.000.000.000.000.00
15 July 20204100240F518.35507.652274.742216.171206.331182.90
p0.000.000.000.000.000.00
Table
Table 5. Climatic conditions in Athens and Bucharest within the 30-day period before the analyzed images (Data from [95]).
Table 5. Climatic conditions in Athens and Bucharest within the 30-day period before the analyzed images (Data from [95]).
CityDateRainfall (mm)Rainy Days (No.)Average Temperature (°C)
Athens13 June 20180024.1
8 July 20190027.1
21 June 20200021.5
22 July 20200026.3
Bucharest10 June 201831921.2
13 June 20191571420.1
26 June 20201091419.8
15 July 202083923.7
Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Share and Cite

MDPI and ACS Style

Popa, A.M.; Onose, D.A.; Sandric, I.C.; Dosiadis, E.A.; Petropoulos, G.P.; Gavrilidis, A.A.; Faka, A. Using GEOBIA and Vegetation Indices to Assess Small Urban Green Areas in Two Climatic Regions. Remote Sens. 2022, 14, 4888. https://doi.org/10.3390/rs14194888

AMA Style

Popa AM, Onose DA, Sandric IC, Dosiadis EA, Petropoulos GP, Gavrilidis AA, Faka A. Using GEOBIA and Vegetation Indices to Assess Small Urban Green Areas in Two Climatic Regions. Remote Sensing. 2022; 14(19):4888. https://doi.org/10.3390/rs14194888

Chicago/Turabian Style

Popa, Ana Maria, Diana Andreea Onose, Ionut Cosmin Sandric, Evangelos A. Dosiadis, George P. Petropoulos, Athanasios Alexandru Gavrilidis, and Antigoni Faka. 2022. "Using GEOBIA and Vegetation Indices to Assess Small Urban Green Areas in Two Climatic Regions" Remote Sensing 14, no. 19: 4888. https://doi.org/10.3390/rs14194888

APA Style

Popa, A. M., Onose, D. A., Sandric, I. C., Dosiadis, E. A., Petropoulos, G. P., Gavrilidis, A. A., & Faka, A. (2022). Using GEOBIA and Vegetation Indices to Assess Small Urban Green Areas in Two Climatic Regions. Remote Sensing, 14(19), 4888. https://doi.org/10.3390/rs14194888

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop