Algorithm for Urban Spontaneous Green Space Detection based on Optical Satellite Remote Sensing

In this study we developed an algorithm for spontaneous green space identification in the cities. The algorithm was developed based on NDVI analyses (calculated from Sentinel-2 data) in Warsaw's green spaces, Poland. We analysed NDVI values in nearly 2000 points which allowed us to develop an algorithm for a distinction between managed and spontaneous vegetation, the algorithm results show that in all vegetation classes NDVI values obtained for managed vegetation are lower than for spontaneously occurring one. Application of algorithm requires a time series of satellite images, for all each of them, the NDVI threshold value is calculated. Finally, based on the number of exceedances of the threshold value each green space pixel is assigned to managed or spontaneous vegetation class. The validation shows that the algorithm achieves 70% accuracy for both classes.