Is ward-level calculation of urban green space availability important?—A case study on Vellore city, India, using the histogram-based spectral discrimination approach

Abstract

How much green space is available for individuals is a major question that city planners are generally interested in, and the present study aimed to address this issue in the context of Vellore, India, through two approaches, namely, the per capita and the geographical area approach. In existing studies, urban green space (UGS) was only calculated at the macro level, i.e., for the city as a whole. Micro-or ward-level analysis was not attempted before, and the present study carried out the same to get a clear picture of the amount of greenery available in each ward of a city. For this purpose, a two-step approach was proposed where the histograms of Google Earth (GE) images were analyzed first to check whether the green cover types such as trees, shrubs/grassland, and cropland were spectrally different. Then, classification techniques such as ISODATA, maximum likelihood, support vector machine (SVM), and object-based methods were applied to the GE images. It was found that SVM performed well in extracting different green cover types with the highest overall accuracy of 93% and Kappa coefficient of 0.881. It was found that when considering the city as a whole, the amount of UGS available is 42% of the total area, which is more than the recommended range of 20–40%. Similarly, the available UGS per person is 97.84 m2, which is far above the recommended 12 m2/person. However, the micro-level analysis revealed that some of the wards have not satisfied the criteria of per capita and percentage area, though the city as a whole has satisfied both the criteria. Thus, the results indicate the importance of calculating the urban green space availability at the ward level rather than the city level as the former gives a closer look at the surplus and deficit areas. The results of terrestrial LiDAR survey at individual tree level revealed that if trees are located adjacent to buildings or roads, it results in fewer heat islands compared to the case where there are no trees.