Study On The Grid Transformaiton Of Population Data At The Service Of Earthquake Emergency

China lies in two of the world’s most seismically active zones, thus is one of the countriessuffering most from major devastating. When a destructive earthquake occurs, seismicdepartments should response as soon as possible to obtain the disaster distribution, estimate thedamage and draw up emergency rescue plans, which cannot be achieved without the help of theearthquake emergency database. But the exiting database has already shown its limitation in recentreal applications:(1) some data items such as administrative division data, traffic data, especiallypopulation and building data which are closely related to disaster assessment need to be updated.(2) Most of the data is statistical data at the level of administrative districts, which is lack of spaceprecision and leads to high bias of earthquake disaster assessment. So it is needed to search amethod, using the data with higher resolution, more effective in precision to enrich the exitingearthquake emergency database. In order to solve the above-mentioned problems, this thesis tookpopulation data as research object, and studied the grid transformation technique of populationdata in Yunnan Province.Firstly, in this thesis, the research status quo and problems of grid transformation of populationdata were reviewed, and future trends were put forward. Aiming at the problems to be solved, theauthor proposed out some methods and strategies and then formed the basic structure and thewhole idea of this thesis as follows: Based on data of Yunnan province, in combination ofpopulation statistics, basic GIS data and remote sensing data with high resolution, using spatialanalysis and statistical analysis technique, on the basis of geomorphology zoning, the models ofpopulation density weight were built at spatial scale of1km2grid-cells for towns and rural areas,respectively. Then, according to population statistics at township level and population densityweight above-mentioned, the population density of each grid was calculated.The question of the grid transformation of population data is primarily an issue in the technicalsense, which needs some technical supports. By using GIS, this research solved the followingquestions: projection transformation, matching population statistics to administrative division data,urban and rural separation, calculating slope and rolling of terrain surface based on DEM,generating1km2grid vector data, spatial statistics analysis of population distribution factors based on the grid or administrative division ranges, and extracting the residential information fromremote sensing data.Based on three different methods, the classification of population agglomeration index, clusteranalysis of population distribution factors and geomorphology zoning, the three results ofpopulation distribution regionalization experiment were compared and the result based ongeomorphology zoning was chosen to be the best option. And then, this work attained to thefollowing conclusions about the macro-space population distribution of Yunnan: the effect ofaverage elevation on the vertical zonal distribution of population cannot be observed, but there is acorrelation between the population density and surface rolling. At a macro-level, populationdensity can be divided into three levels, and its space distribution presents that the middle east ofYunnan Province has the highest population density level, the southern has the second and thenorthwest has the last.The question of model level conversion is the focused and difficult problem in the researchabout grid transformation of population data and its objective existence has been verified in thisstudy. To solve this problem, by using a method based on the residential area to estimate gridpopulation density, a number of population data samples were acquired to support the populationdensity weight modeling, which can reveal the statistical regularities between population densityand its factors at1km2grid-cell level.In combination of the modeling scale and the limitation of fundamental data, this thesis chose asubjective way for modeling index selection. Some land-use data such as farmland and residentialland which is directly related to population distribution and some strong indicators of populationdistribution such as roads, residential spots and POI (point of interest) from navigable databasewere selected to help the modeling. The outcome statistical analysis shows that taking road andresidential point data as the supplement is an effective way to solve the information loss ofland-use data.For towns, a model based on road density and distances from town centers was used to calculatethe population density weight value. The result shows that the road forms the frame structure ofthe population distribution, and the population density decreases with the increase of the distancefrom the town center.For rural areas, based on geomorphology zoning and support by the modeling samples which were obtained from grid population density estimation, the rural population density weight modelwas built in each geomorphology zone by using multiple regression analysis. Moreover, the modelfactors were sorted by qualitative principle of their influences. And the sort result was used toadjust model coefficients, and made the final model more reasonable from the view of geography.Finally, a group of field investigated population data was adopted to testify the precision of themodels. The correlation and regression analysis shows that the linear correlation coefficientbetween simulate and investigate results is0.919and the goodness of fit index is0.948. The errorstatistical result shows that there are about70percent of the test samples within the error rangefrom-30%to30%. The results above demonstrate that the population density distribution mapgenerated in this study has a high precision, which can be validated by the precision test.