Distribution Pattern And Dynamics Study Of Vegetation Covers At Yellow River Delta

Important roles of the vegetation cover changes for the ecological security pattern at the regional scale has been largely documented. Yellow River Delta, with the most fast depositing speed for land of the world, is characteristic of rapid and complex changes for its LUCC. Now most of the study for the vegetation study focuses on either the classical relationship between vegetation community and environmental factors, or the land cover changes integrating more social factors than the ecological factors, lacking of the study of vegetation covers at large scale.A SPOT imagine was employed here for the interpretation of vegetation covers at Yellow River Delta. Different environmental layers and geographic analysis methods, such as buffer analysis, average nearest neighbor distance analysis and mean center analysis, was integrated into the distribution pattern study of the vegetation covers. Yellow River, aced as the most important fresh water provider, is the macro gradient of water, but with the existence of human disturbance, the law is not distinct to conclude. The coastline is the macro gradient of soil salinity, but the law is not distinct to conclude with the existence of river mouth and the heterogeneity of the tideland. For the study of the average elevation and average soil salinity, the law is distinct for the major natural vegetation covers with great correspondence with the conclusion of ground survey. The result of the average nearest neighbor distance analysis shows that the distribution of the major natural vegetation covers is characteristic of clamp, and the spatial clamp distribution could perfectly be expressed by the succession theory of the Yellow River Delta after the analysis of mean center method. At the perpendicular direction of the coastline is the distribution of the salt-oriented succession, while at the perpendicular direction of Yellow River is the distribution of the water- oriented succession.Using the Pre-classification comparison method, the dynamic change of NDVI, which is most popular used vegetation index, is studied in Yellow River Delta. The wavelet method is employed here, for the detecting of vegetation cover change, the performance of which is extremely well. In light of the NDVI imagine extracted from three TM images, the high value area of NDVI is distributed among the adjacent area of Yellow River, and the former Yellow River, and the low value area of the NDVI is distributed among the coastline area, the result of which is similar to the existed study. With the result of nonlinear regression, the distribution of NDVI is related to the distribution of soil salinity. For a whole, the dynamic change of NDVI from 1996 to 2005 is relatively calm. Distribution of significant change area is characterized of a ring-shaped. The most significant changed area is located at the adjacent area of coastline, and the most significant changed area is located at the buffer area of coastline, which is on the neighbor of the sea. The patches of changed area from 1996 to 2001, in area, are more than that from 2001 to 2005. The wavelet disposed images at three levels show the similar distribution of initial image. There are no significant difference between initial image and level 1 image, level 2 image. The image of level 3 deprived of a sounding portion of noise and odd value, hence, the wavelet dispose is available for the interpretation of change rules.We employed the mathematic method for the qualification of the driving force for the vegetation cover change. The analysis for the EDA reveals that the 25.64% of the total changes occurred at the 0～5 Km area of the coastline buffer area, while 47.20% at the 0～10 Km. As the increase of the distance from the city, the number of changed pixels is characteristic of a ripple-shaped curve, and the major area for changed is the 10Km～25Km buffer area of the city. As the increase of the distance to Yellow River, the number of changed pixels is on the decreased trend. About 44.79% of the changed pixels locate at the 0～10 Km buffer area of the Yellow River, and almost of the changed pixels locate at the 0～5 Km buffer area of the Yellow River. Almost the changed pixels occurs at the -0.20~0.60m range of elevation, and, also the number of changed pixels is characteristic of a ripple-shaped curve, as the increase of the elevation. As the increase of the distance to the roads, the number of changed pixels decreased sharply. The achievement of the roads could be understood as a indicator of the human activity, hence, the changes occurred attributed to the human activity. Using the 300m×300m sliding window, we scanned the binary imagine from 1996 to 2001, calculating the changed pixels of the 100 pixels, which could be used as the probability of the land-use change. At the same time, we calculate the mean average Lc( the distance from the pixel to the coastline), Ly(the distance from the pixel to the coastline), Lxg (the distance from the pixel to the local city - Xianhe and Gudao), Lr( the distance form the pixel to the local major roads) and De( the elevation of each pixel). Using the multiple stepwise multiple regression procedure, we got the weights of the five variables for each land-use type and the expression of the calculation of suitability maps for each maps:p = (65.892 - 0.0049* Lr - 0.044* De + 0.00062* Lc - 0.000065* Lxg)/100(F = 1883.9, p = 0.000)Due to the stepwise regression, the expression of the regression could be used for interpretation of land-use change. According to expression, the dominate factor determining the land-use change is the elevation, and the second is the distance to roads. However, the distance to Yellow River is not included into the expression, which may relate to that the influence of the Yellow River for the land-use change may exist in a limited range.The Detrend Canonical Correspondence Analysis method is used here for the interpretation of relationship between the landscape pattern and human activity. The first two axes of the landscape index matrix could interpret 81.6% and 8.3% of the total variance of the matrix respectively, which is able to represent the landscape character. The most related variable of the human activity is the PF, and POD, PB, and PW in turn for the first axis, while, for the second axis, the most related variable of the human activity is the AA and RF. Due to the 81.6% of the total variance interpreted by the first axis, the most determined variable of the landscape pattern is PF. With no constrained matrix, the Eigen value is 0.00631. And with the landscape index matrix as the constrained matrix, the Eigen value is 0.006. Hence, the percentage interpreted by human activity is 95.08%, from which a conclusion could be drawn that the landscape pattern is the result of human activity.Assisted by the Cellular automate model, we accomplished the simulation of Land-use change of multiple category land-use type and the human disturbance. Using the 5 by 5 sliding window , the transition area matrix and suitability maps, we specify the transition state and simulate the prediction map of 2005, with which the actual 2005 land-use map would be compared. Using the Kappa coefficient equation above, we got the Kappa coefficient and overall correct percent of the model MCA and AACA. The Kappa coefficient of and overall correct percent of MCA are 0.379 and 46.94%, while, for the AACA model, are 0.447 and 54.92%. GEOMOD model is a binary model used for simulating the land-use change. Here, we re-classed the vegetation maps, and set the 2001 map as the initial simulating map, constrained with the actual 2005 vegetation map and the suitability map generate by stepwise regression. Compared with the actual reference map of 2005, the result of simulation is good for its performance, with kappa coefficient 0.398 and overall accuracy 69.85%.