Dynamic Study Of Forest Landscape Pattern In Zengcheng District Of Guangzhou City Based On RS And GIS

With social and economic development,the requirements for forest resources by human are keeping increasing strongly.The rapid expansion of cities and the destruction of forest ecosystems have affected the balance of natural ecology and the long-term interests of mankind itself.Meanwhile,ecological civilization have been raised to an unprecedented level in China,the Guangdong Provincial Party Committee and the provincial government proposed to take action afforesting the province in depth,and strive to build a new ecological situation in the southern Guangdong Province and become No.1green ecological province in China.Fragmentation of forests leads to poor forest quality and weakens function of ecosystem.This paper studies the forest landscape the heterogeneity of the forest landscape in Zengcheng District of Guangzhou,and discusses the landscape change patterns and development trends in the past 10 years.The Landsat satellite imagery of the 2008,2011,and 2016 research areas was used as the data source.In the eCognition 9.0,the object-oriented classification method was used to classify the image data to obtain a high accuracy result.The results were input into Fragstats 4.2 to calculate the landscape indicators to help analyzing the landscape patterns.After calculating spatial and temporal interaction of the landscape from 2011 to 2016,CA-Markov model was used to predict the landscape composition in 2021,which will help the action of construction and urban forest management in the next five years.The main results of this article are as follows:(1)Forest is the dominant landscape type in Zengcheng District,its area had achieved considerable growth ca.3528.45 hm~2 within the study time interval,but a certain number of the previous forest lands degenerated into grassland and other landscape types landscape.The number and density of patches of all landscapes types increased.At the same time,the shape of all patches became more complex and the distribution became more discrete,indicating that the momentum of landscape in the study area was developing in fragmented.At the landscape level,the degree of landscape separation in the study area was relatively stable,but its value was relatively high.While the landscape became more fragmented,there was a unconspicuous degree of aggregation.Across the study period,all types of landscapes were developing in unbalanced.(2)In the analysis of landscape types shift,the most conspicuous landscape interaction change occurred between urban forest land,grassland and farmland,and the dynamic change of grassland was the most significant.With the analysis of the landscape pattern,it shows that the forest management and construction action in the study area had been well implemented and the area of forest was increased.However,the fragmentation of the forest landscape and the ever-present degradation of the forest land also indicate that the forest quality maintenance and upgrading work needs strengthen further.(3)Based on the pattern of land-based changes and landscape disturbance conditions during the period of 2011-2016,using the CA-Markov model to predict the landscape composition of in 2021,the results show that the urban forest area will continue to increase ca.3630 hm~2,and the construction land area will keep increasing.In future urban construction and forest management,the objective of the adjustment of forest landscape fragmentation should always be taken into account.(4)This article uses the object-oriented classification method to classify the landscape in the study area.According to the experimental results,the optimal segmentation parameters are set,and the effective image band,vegetation index,and topography are selected as the classification features.Then the CART decision tree classification produced a good classification result.In 2008 Landsat 5 image classification accuracy is 90.83%,Kappa value is 0.89;in 2011 Landsat 5 image classification accuracy is 89.17%,Kappa value is divided into 0.87;2016 Landsat 8 image classification accuracy is 90.83%,Kappa value Divided into 0.89.(5)eCognition is the most professional object-oriented classification software in the world.The software provides a variety of segmentation and classification algorithms,and a great deal of freedom and flexibility in the selection of classification features to achieve high-precision remote sensing image classification.With more and more effective classification features are created based on image information such as spectrum,texture,and shape,the image classification accuracy will be improved to a greater extent.Based on the long-term predictability of Markov model,CA-Markov model is combined with CA model to strengthen the ability to simulate the spatial variation of complex systems.The model is a functional module in IDRISI.In addition,the software also contains a variety of algorithms with great ability in simulating and predicting land cover.IDRISI has always paid attention to related theories and the development trend of technological frontiers,new modules are constantly added into it.Throughout the processing of the entire research data,ARCGIS has always been involved,so that the data smoothly computed in various software platforms.The landscape pattern analysis is generally considered to be guided by the landscape ecology theories and relied on the GIS,RS,GPS technology platform.With the advancement of satellite sensor information capture capabilities and the innovation of the software,their relationship will become more compact,and more advanced and complex algorithms can be easily implemented to obtain great accurate results.The guidance of landscape spatial analysis for related work will be more practical.