A conceptual model for identifying the risk susceptibility of urban green spaces using geo-spatial techniques
- First Online:
- Cite this article as:
- Bardhan, R., Debnath, R. & Bandopadhyay, S. Model. Earth Syst. Environ. (2016) 2: 144. doi:10.1007/s40808-016-0202-y
- 654 Downloads
Urban green spaces are often regarded as the harbinger of sustainability in the rapidly urbanizing world. This study forwards a conceptual framework towards urban green space (UGS) management by the quantification of risk susceptibility of the UGS at a neighborhood level by using remote sensing data and geo-spatial techniques. Objective measure of the UGS was performed using weighted evaluation of NDVI data at a 20 m × 20 m grid over the city of Kolkata. The normalized green index (NGI) was developed to quantify the implication of the built-up spaces in UGS risk susceptibility in the urban fabric. Both the satellite image data and the NGI values were spatially auto-correlated using bivariate Moran’s I to derive the intricate relationship between built-up area and green spaces using LISA. It was observed that the low green-spaces were greatly influenced by the high built-up area around it. This inference was extended on the Kolkata’s grid, the results revealed that the older part of the city had the most risk susceptible zones, whereas the area surrounded by wetlands were the most stable region bearing high resilience to urbanization. Hence, by determining the most risky zones for UGS degradation, planners and policy makers can efficiently allocate resources towards the sustainable development of the city by conserving and promoting UGS.
KeywordsUrban green spaces Geo-spatial Concept-model Environment Spatial-autocorrelation
Sustainable development goals (SDG) no. 15, aims at protecting, restoring and promoting sustainable use of terrestrial ecosystems, sustainably manage forest and combat desertification by conserving land degradation and preventing bio-diversity loss by 2030 (United Nations 2014). Rapid urbanization and population explosion is the biggest force that is irreversibly transforming terrestrial ecosystems. 11 out of 19 megacities of the world, with more than 10 million inhabitants are in Asia, with India and China alone accommodates more than 1 billion population (Zhao et al. 2006). Here, more than 50 % of the population lives in less than 3 % of the urbanized terrestrial surface, creating tremendous pressure on the land, triggering ecological consequences (Singh et al. 2010). Rapid urbanization in India is changing the dynamics between ecology, economy and the society. In the last 50 years, the urban population in India has grown fivefolds, where 60 % of this growth is attributable to natural growth, and the remaining 40 % is due to migration and spatial expansion (Sivaramakrishnan et al. 2005; Taubenböck et al. 2009). The urban population in India is likely to grow to 500 million in the next 50 years, and will continue to have substantial impact on the ecology, economy and society at local, regional, and global scales (Singh et al. 2010). Hence, redesigning and optimizing urban systems to mitigate the impact of rapid urbanization and population expansion, while being carbon neutral, is an urgent necessity to combat climate change, and is a serious public policy and adaptation management challenge for India (Revi 2008).
In this purview, one of the effective policy instrument is the protection of green-areas within urban social-ecological systems to amend several problems of city-living. Green areas or the ‘Urban Green Spaces (UGS)’ in the city comprises all urban parks, forests and related vegetation that add value to the urbanites (Gupta et al. 2012). The benefits of UGS are wide ranging that include enhancing of public health, social solidity, climate change mitigation, pollution reduction, biodiversity conservation and uplift the quality of ecosystem services (Singh et al. 2010). Apart from maintaining the quality and sustainability, UGS functions as a visual screen and act as noise barriers and avoid too much spatial uniformity (Gupta et al. 2012). Hence in one word green spaces contribute in reducing urban vulnerabilities.
Global UGS estimation suggests that developed countries have more green spaces compared to the cities in developing countries, which often fall below the minimum standards of 9 m2 green open space per city dweller by the World Health Organization (WHO) (Singh et al. 2010). The UGS coverage in India are not well-studied, some studies on Bangalore (Nagendra and Gopal 2011; Sudha and Ravindranath 2000), Delhi (FSI 2009) and Chandigarh (Chaudhry and Tewari 2010), focus more on the subjective methods of perceived, self-reported hedonism about green spaces involve the method of survey question audit by trained raters who apply specific criteria (e.g., presence/absence of various features) to assess natural environments. Such subjective measures are time consuming and expensive (Gupta et al. 2012). With the advent of remote sensing (RS) and geographical information system (GIS), objective measures of UGS have become more communicative, by virtue of its receptivity, synoptic view and larger area coverage. Forster (1983) showed the applicability of Landsat imagery for producing quality index for urban areas which included rating of green spaces (Gupta et al. 2012; Sivaramakrishnan et al. 2005; Sudha and Ravindranath 2000). Liu and Liu (2008) estimated green spaces using the ecological niche modeling technique, but concluded that it should not be used for UGS construction, as it does not address spaces and human barrier conflicts. Image processing allows extraction of features of interest, based on reflectance characteristics, from satellite data in the form of indices through different algorithms and mathematical indices. The Normalized Difference Vegetation Index (NDVI) is one such index that is used to highlight vegetation areas on a remote sensing data and is widely applied for studying global environmental and climate change(Gao 1996). It is calculated as a ratio difference between measured canopy reflectance in the red and near infrared bands respectively (Bhandari et al. 2012). The use of GIS has proved useful in identifying vegetation cover, preparing urban green data inventory and mapping potential value of a green areas using various attribute and location-based queries and, data integration techniques. Assessing multi-attribute values like risk or vulnerability of such urban assets can also be performed in GIS using advanced spatial algorithms, like multi-criteria mapping, spatial overlay, spatial auto-correlation and spatial regression. Such techniques enable in identifying spatial hot and cold spots of risks and vulnerability.
In city planning, UGS are considered as a key element for reducing urban vulnerability from external stresses of urbanization, hence are often treated as homogeneous goods, whose loss can be quantified as a uniform value reduction in a continuum scale (Lee et al. 2015; McConnells and Walls 2005). However, in reality it is more complex, they are actually heterogeneous goods having hierarchy of distinct values based on the functionality and services they provide. Thus create varied resistance towards urban vulnerability reduction.
The efficiency of UGS are dependent on their own risk liability. Urban green assets has an inherent resiliency, which is a function of the intrinsic risk susceptibility of the green space, based on its regeneration and rejuvenation capability and contextual association to built-spaces. Therefore, to internalize the beneficiary role of UGS, not all green spaces need to be statutorily preserved at the same time. This innate risk of the UGS can be utilized to derive a hierarchical UGS risk function and only those greens that are at immediate threat to extinction or will collapse can be preserved. Currently there exists no standard methodology to identify the risk susceptibility of the UGS. Studies that deal with the quality of green spaces mostly focus on characterizing them based on multiple environmental attributes and rarely address the issue of the intrinsic resilience. (Gupta et al. 2012; Sati et al. 2016). Remote sensing and GIS techniques can be aptly used for this function.
In this study we forward a conceptual framework for identifying the risk susceptibility of UGS as an effect of urbanization. Here, the risk susceptibility of UGS is defined as a function of its greenness or the vegetation health and its exposure to stress from urbanization. We use satellite imagery to capture the status of UGS in a rapidly urbanizing area, and generate a green space risk susceptibility map using Anselin’s (1995) spatial dependence model of bivariate-Local Index of Spatial Association (LISA).
Urban green spaces are an intrinsic part of a sustainable built environment. The importance of green spaces in the neighborhood transcends environmental, social and economic benefits. However, there are complex risks associated with the development and conservation of UGS. In this study, we forward a methodology that would not only help planners or policy makers to decide how much emphasis should be given to a particular spatial range of UGS, but also optimize resource allocation, in the preservation, protection and conservation of it. This work picks from the concluding remarks of Gupta et al. (2012)where they derived a tool/technique to objectively measure the quality of green spaces in an urban neighborhood using RS and GIS technology. They concluded by converging their study towards the importance of such RS enabled tool at the household level for urban-environment sustainability. Here, we pick this end and put forward a conceptual framework for studying the risk susceptibility decision making of UGS from built–space interaction.
To identify the various parameters for UGS vulnerability, indices based classification of remotely sensed data from Indian Remote Sensing Satellite (IRS), Resourcesat-1 IRS P6 Linear Imaging Self Scanner (LISS-IV) (May, 2011) was used. It was then used to classify the UGS based on their greenness (vegetation health) into Very High (VH), High (H), and Low (L) values. LISS-IV is a multispectral high resolution sensor with a spatial resolution of 5.8 m at nadir and operating in three spectral bands 0.52–0.59, 0.62–0.68 and 0.76–0.86 μm of green, red and near infrared regions (NIR) of the electromagnetic field. NIR is extensively used for the characterization and identification of vegetation using the Normalized Difference Vegetation Index (NDVI) (Gupta and Jain 2005).
Characterization of NDVI and extraction of built-up area
It should be noted that NDVI is a representation of the degree of greenness of a particular remote sensing scene and is equal to the chlorophyll concentration i.e., all observed areas having chlorophyll density. It varies with the degree of absorption of red light by plant chlorophyll and the subsequent reflection of infrared radiation by moisture in the leaf (Bhandari et al. 2012). Therefore, NDVI of an urban area depicts both accessible and non-accessible green spaces, however, the city green area in terms of land use only reports the accessible green spaces like parks, playgrounds, gardens, etc. Consequently, the total greenness reported by NDVI image will be much higher than the green cover of land use maps. NDVI values range from +1 to −1, where values near to +1 denotes healthy vegetation, whereas values near to −1 denotes barren land. This range of +1 to −1 is dependent on the sensitivity of the NIR band of a specific sensor of the satellite (Carlson et al. 1994).
Classification of NDVI data and its interpretation
NDVI threshold values
Very high (VH)
Very dense and highly vigorous vegetation
Abundant and vigorous vegetation
Sparse and apathetic vegetation
The next step involved the extraction of the built-up area from the LISS IV data, by using the parametric maximum likelihood approach based supervised classification method. Supervised classification is based on the idea that the user can select maximum training samples from the image that are the representative of a specific class and then process the training sets (signatures). These training sets were used as references for the classification of all other pixels in the image. However, the accuracy of this classification is highly dependent on the number of the training sets that were used.
In our classification, we have taken 50 signatures for both built-up and non-built up classes. Finally, the LISS-IV imagery was classified into two classes (1) ‘Built up’ and (2) ‘Non-built up’ with an overall kappa accuracy of 96 %. The raster output from NDVI and built-up areas were co-registered and subtracted from each other (i.e., built-up areas -NDVI). This was done to remove overlapping pixels, such that built-up areas do not have pixels that also represent green vegetation. The threshold for retaining the pixels in the NDVI image, in spite of overlap with built-up area image was set at 0.2.
Development of the normalized green index (NGI) and normalized built-up index (NBI)
In order to quantify the status of UGS based on the NDVI characterization (see Table 1), weights were allotted to each class of the NDVI image, within a range of 1–5. The weights provided are 5, 3 and 1 for Low (L), High (H) and Very High (VH) NDVI classes, respectively. These weights were provided based on the inherent reliance of urban greens i.e., more NDVI means healthy vegetation and hence more inherent resilience. In other words less is the risk susceptibility. However, no weights were allotted to built-up pixels.
Spatial relationship between NGI and NBI
Moran’s I value ranges from −1 to +1, where +1.0 indicates clustering while a value near −1.0 indicates dispersion. A high positive local Moran’s I value infers that the location under study has similar high or low values as its neighbors, thus the locations are spatial clusters. The high–high clusters where, high values are surrounded by high value neighborhoods, and low–low clusters have low values surrounded by low value neighborhoods. In this study, high–high NGI and NBI values depict hotspots of healthy UGS and high urbanization respectively.
Calculating risk susceptibility of USG
Risk susceptibility classification
Risk susceptibility of UGS
Status of NGI
This table should be read as following: if the NGI value of a particular patch of USG is ‘HIGH’ and its surrounding values of NBI is also ‘HIGH’, then the greenness status of that particular green patch is “fatigued’ with a ‘MODERATE’ risk susceptibility. That means the green patch has higher resilience, to cope with the high exposure from NBI, however, the green patch is fatigued i.e., with prolonged exposure it might collapse. Similarly, when NGI value is ‘LOW’ and the neighboring NBI value is ‘HIGH’, then the risk susceptibility is very high, i.e., exposure from urbanization exceeds inherent resilience and hence will evidently lead to collapse of the green space.
Since the risk susceptibility values were subjectively measure, for ease of calculation we convert them to objective values by assigning weights in accordance to their hierarchy of risk susceptibility.
Identification of exposure quotient of USG and specific green assets under maximum risk
Inherent resilience of USG (NDVI weights) ↓
Risk susceptibility weights →
Land-use status (2008)
Classification of UGS based on its vegetation health
Very high (VH)
Descriptive of different classes of UGS
Normalized Green Index (NGI)
Very high (VH)
The bivariate Moran’s I of NGI(x) and NBI (lag of y) was −0.393. The negative value implied an inverse relationship between NGI and NBI. This infers that the low green-spaces are surrounded by high built-up areas.
In this study, a conceptual framework was forwarded to identify urban green spaces that are at high risks due to urbanization. This framework will be an essential tool to urban planners and policy makers to facilitate sustainable urban built environment bye-laws and guidelines. The risk susceptibility of the UGS was calculated using NDVI values, classified into three categories: Low (L), High (H) and Very High (VH). In order, to convert this subjective measure into an objective measure, weights were assigned. This objective measure was termed as NGI. This NDVI of Kolkata is illustrated in Fig. 6. Following steps included, deriving the spatial relationships between NGI and NBI, and finally converging the spatial correlations, into USG’s risk susceptibility index by using bivariate Moran’s I.
If we look at the spatial distribution of the green spaces at moderate to high risks, it can be seen that more risk inclined green spaces are concentrated in the North, West and the central part of the city, where the consolidation of built-up spaces are high, as illustrated in Fig. 7. The Southern and the Eastern part of the city pose less risk on account of the wetlands. The highest susceptibility of UGS were found to cluster in the Northern part of the city, which are the older parts of the Kolkata city. These areas have high sprawling due to heavy urbanization, thus, consequently manifesting in high risk susceptibility for the green spaces.
Uncontrolled urbanization in India is the biggest driver of rapid environmental and green space degradation. (Maiti and Agrawal 2005). The urban heat island effect is also another outcome of rapid green asset degradation, which is shrinking the scope and boundaries for energy and health sustainability. Green spaces are the most essential component of the built-environment. The presence of green spaces not only improves health, social well-being, but mitigates the ill-effects of climate change (Tu et al. 2016). Thus the green spaces cater to every normative layer of sustainable urban development. It is an utmost necessity to preserve and protect the green spaces in the city to foster inclusive growth. However, urban planners and policy makers, lacks appropriate tools to identify and solve the resource allocation problems associated with the UGS design. This study demonstrated that remote sensing technology when coupled with GIS tools, can be a valuable technique towards identification and monitoring of green spaces that are at high risks. Future work lies in the refinement of the spatial data using LIDAR or other advanced high resolution satellite imagery systems, so that more realistic models can be analyzed at the neighborhood and individual level.
In diverse megacity like Kolkata, it is extremely difficult to find spaces for the development of UGS. Moreover, the current green spaces are at the brink of extinction due to rapid urbanization and population explosion. The methodology that is forwarded in this study, not only assess the state of UGS in the built environment of the city, but also pin-points high risk zones. This quarantine of high risk UGSs, can help in efficient resource allocation by the policy makers. This tool enables the planner to focus on green assets at a micro- level, thus, enabling a bottom up planning strategies. The conceptual framework which is devised in this study for the quantification of the green space can be applied to other Indian cities, where the process of urbanization is fast and urban greens is stressed. Due to its low resource intensity and ease of computation, this method can be easily adopted by planners and policy makers for sustainable development and rapid planning process for any developing nations. Therefore, fostering more community engagement and nurturing pro-environmental behavior among the habitants. Strengthening the green cover right from the household to the neighborhood to city level, which will not only stabilize the environmental degradation and climate change, but will also push the community and the country towards fulfilling the UN-Sustainable Development Goals.
This work was supported by the Ministry of Human Resource and Development (MHRD), Govt. of India for funding the project titled ‘Urban Form and Extreme Precipitation Events: Are compact cities more vulnerable to flooding?’ (14IMHPCU001) conducted by Center for Urban Science and Engineering (C-USE), Indian Institute of Technology Bombay. The authors acknowledge the valuable suggestions and comments of Prof. Arnab Jana, C-USE, IIT Bombay, for improving the quality of this paper.