Analysis Of Landscape Dynamics In Urban Forest Parks Based On High-resolution Images

In this paper, Purple Mountain Forest Park in Nanjing was chosen as the case study area.First, high-resolution remote sensing images in2004and2007were collected as the major datasource. Second, five methods including object-oriented classification, dynamic analysis oflandscape patterns, cellular automaton, MARCOV chain model and the minimum cumulativeresistance model were applied to analyze the dynamic of land utilization and landscape pattern atPurple Mountain on various platforms such as ArcGIS, ENVI, EX, Maxent and IDRISI. Third,based on the analysis results mentioned above, the characteristics of landscape pattern in PurpleMountain in2011had been predicted.Prediction accuracy of CA-Marcov model was evaluatedby calculation of the confusion matrix and kappa coefficient. Finally, based on the predictionresults of future landscape pattern in Purple Mountain, sustainable management measures forprotection and utilization of Purple Mountain were put forward. The research results are showedas follows.(1)By comparing the classification accuracy of fusion images of four vegetation indexes andremote sensing images, it can be seen that the normalized differential vegetation index (NDVI)outperformed the other three indexes. In RS-based monitoring of vegetation dynamic, theperformance of NDVI is the best.(2)It can be seen by comparison of landscape pattern index from2004to2007, the numberof patch had risen to71840in2007, while the average area of patch had dropped to0.07km2. Thepatches increased in number,while the mean area of patches reduced and the landsacape patternbecame more fragmented in Purple Mountain from2004to2007. By comparing the landscapepattern index and their ranks at the scale of landscape classes, it can be seen that the patchnumber of broad-leaved forests significantly increased from2004to2007, the number of forestpatches in Purple Mountain increased, and the farmland and grassland patches obviouslydecreased. Thirdly, from2004to2007, landscape edge density (ED) was generally increased, andthe average patch area (MPS) was reduced, indicating that the degree of landscape fragmentationis continuously increasing. Finally, from2004to2007,the landscape pattern in Purple Mountainshowed a strengthed trend of heterogeneity and fragmentation.(3)From the transformation matrix of land-use types during the period of2004to2007,it canbe seen that the land used for broad-leaved forests and roads significantly increased,while thearea of coniferous forests and farmlands obviously decreased through conversion intobroad-leaved forests and roads. The area of grassland was slightly decreased. Parts of farmlandwas converted into the grassland and the area of water was not obvious changed. Generally, thearea of road and broad-leaved forests showed an increasing trend. the conversion of coniferousforests, broad-leaved forests and farmland into other land types was obvious from the viewpointof spatial pattern. (4) It can be seen from the prediction result of minimum cumulative resistance model thatthe resistance lines of landscape security pattern of Purple Mountain were mainly distributed onthe northern slope of the mountain. There were three sources which were suitable for thehabituation of mammals in the study area. These habitats located on the upper northern slope ofPurple Mountain had the following characteristics: high altitude; high vegetation coverage; farfrom the highway, railway and human settlements, but close to water source regions such as lakes,reservoirs; less disturbance from human activities.