Optimizing Green Space Landscape Patterns Based On High Resolution Image

Using the high spatial resolution Quickbird imagery acquired in2005, covering Futiandistrict of Shenzhen city as the major data source, based on the outcomes produced from theobject-oriented multi-scale segmentation algorithms, stepwise discriminant analysis (SDA) inpartnership with classification and regression trees (CART) was applied to construct a multi-scaleand multi-variant classification model to extract information on urban green lands. Landscapeecology methodologies coupled with GIS technologies was utilized to address issues related tourban green land classification algorithms, the selection of landscape pattern indices and thelayout of urban green lands, and further to analyze and optimize urban green land patterns andconnectivity. The findings from the current work were summarized as follows:(1) The classification model created by applying object-oriented classification algorithms,SDA in conjunction with CART was able to improve the accuracies of classification, or to reducethe dimensionalities of features while maintaining the accuracies without significant reduction.The training and validation accuracies of CART16model were estimated at94.44%and83.37%respectively, with kappa coefficients were0.93and0.81. The amount of features fed to the modelwas minimal and dropped to a half of the original features. After post-classification process of theoriginal classification result, the last overall accuracy of the land use classification was90.29%,with a Kappa coefficient of0.88. In which, the producer accuracy and the user accuracy of theurban green lands class were98.39%and97.60%respectively.(2) The area of the green lands in the study area totaled30.67km2, with a coverage rate of41.46%. In terms of functionalities of the green lands, the other type of green lands was thelargest one, with a area share of50%, followed by the green parks, with a ratio of23.73%. Theresidential green space, road green space and the other affiliated green space were smallest inarea, and they were widely distributed and highly fragmented. There was a heavy deficiency inprotective green space, mainly distributing the regions of water shores, river banks, futian blockand meilin block. In terms of patch size, the amount of large green patches was small but theyhad a areal ratio of79.26%of the total green space. There were a large number of green patcheswith a small size, with a wide distribution in the study area.(3) In general, the shape of green land patches was simple, regular, and the landscapediversity was low, with an uneven distribution. The green parks, road green space and the othergreen space were characterized by big area, centralized distribution, simplicity and regularity,high aggregation and dominance, and low fractal dimensions. There was a high fragmentation forthe green space, and the small-size green patches had the highest fragmentation, which wasbecause that the small-size green space constituted by the residential green space、road greenspace and the other affiliated green space, was extremely sparsely scattered.(4) The integrated connectivity of green space in the study area was low, the large and medium-size patches that have relative high connectivity significance were unevenly distributed,which caused a low connection among green patches, the termination of naturally ecologicalprocesses and a low stability of landscape. The green parks and other green space were the majorcomponents, with a large size and high connectivity. Large and medium-size green patches play acritical role in urban planning and urban construction, which should be protected and developedwith caution.(5) Taking those green patches greater than5000m2as our new target, the top10greenpatches in area were selected to implement a grading buffer analysis based on the connectivitydistance thresholds to determine those key regions that need to increase green patches to improvethe connectivity, or to recover the existing green space. Through the recovery of renovation ofgreen space in those key regions, the connectivity of green space in Futian district has beensignificantly improved.