Morphology-Based Modeling Of Aggregation Effect On The Categorical Raster Data

Scale is the window for humans to observe the geographical objects and phenomena in reality.Scale effects,defined as the differences among geographical space entities,events and processes on different time and space scales,is reflected in geography and other disciplines.The "natural principle" proposed in geography defines the general rule for people to observe "reality" through the scale window.Revealing and modeling the scale effect of geospatial phenomena is the basis for understanding the multi-scale representation of spatial objects and is also a key for searching optimal expression scales and optimizing scaling approaches.At present,the traditional scale effect analysis method only quantifies the scale effect of metrics from multi-scale data or describes scale effect from different scale transform results.These strategies only stay in the revealing and description of the scale effect and does not consider the influence of the characteristics of the landscape itself to multi-scale data representation.It is difficult to elucidate the relevant factors behind the scale effect in a complex land surface environment.Therefore,this study aims at the problem of aggregation scale effect of land cover data,and proposes the idea of modeling scale effect from the three levels of landscape,(i.e.,patch level,class level and scene level).Based on the hypothesis that patch morphology has a decisive influence on the evolution of patches in the mode aggregation scale,and the method of constructing a scale model of area aggregation based on patch morphology is developed and expected to describe the scale effect of the patch from the mechanism.According to this model,the behavior of the landscape category area scale is predicted.Firstly,the review and systematic analysis of the up-scaling methods and its scale effects of the existing categorical raster data are conducted,and the main problems of existing scale effect analysis methods are pointed out.On this basis,the feasibility and preliminary thinking of scale effect modeling are proposed.The upscaling method is one of the important ways to implement multi-scale spatial data representation.The common point of the upward evaluation is to use metrics to quantify the amount of information loss at different scales and come up with a series of regular scaling rules.The analysis of the scale effect of this kind of landscape metrics analysis method is still “one scene,one discussion” and not universal.Landscape ecology defines three levels of landscape metrics,i.e.,patch level,class level,and scene level.The aggregate effect of land cover data also exists in these three levels,but,we have found that scale effects at the category and scene levels can be synthesized by using patches and their spatial relationships.Therefore,it is an effective way to use the patch-level aggregation scale effect model for building category raster data analyze phenomena and process scale behavior.Furthermore,a method of constructing scale effect model of area aggregation based on patch morphology is developed.Based on the observation that the patch geometry has a decisive influence on the evolution of the patch in scale,the area-morphology-scale statistical model of landscape patch aggregation effect is established.In the modeling,it is assumed that the aggregation effect of patch size is positively correlated with the patch morphology.The patch area is an important variable of the landscape and selected as the model dependent variable.Mathematical models are given based on modeling ideas.Due to the scale effect models of different patch types are significantly different,patches are first divided into patches of “Point”,“Line” and “Area” type,and then calculated their patch morphology metrics.The scale is an independent variable,and the area ratio is a statistical model for the dependent variable.Insufficient data points in the model are obtained by interpolation.Next,the model independent variable-patch morphology metrics is analyzed,and the appropriate metrics is selected as the modeling parameter.This paper analyzes the applicability of patch morphological metrics.It is difficult to describe the life cycle of the relative target expression scales of patches in terms of existing metrics,and a support radius metric is proposed.In view of the existence of many types of patch metrics and the duplication of information description,the existing dimensional reduction research methods for landscape metrics are used for reference.This paper designs dimension reduction experiments for patch level morphological description metrics and expects to characterize the maximum degree of features using the minimum index variables.information.Furthermore,the correlation analysis and principal component analysis method are used to calculate a number of sample data.Through analysis of a large number of sample points,the support radius metric and filling degree metric are selected as the model patch morphological variables.Finally,the accuracy of the model is verified and the application of category area prediction is carried out.In order to verify the validity of the model,this model is used to compare the existence and area of different types of patches.The accuracy of the model is verified by the ratio of patch area to the original area of patch(area ratio of patches)after polymerization.The model is applied to the area estimation of a single category.The support radius metric proposed in the modeling can measure the survival period of the relative expression scale of the patch.The length of this life cycle can lead to changes in the number of patches,density,and other information,and this study analyzes the scale of the landscape pattern under the support radius.effect.The experimental results show that different types of patch scale effect modeling can reflect the scale evolution behavior of patch area and accurately predict the life cycle of the patch.The modeling ideas and methods can not only reveal the multi-scale characteristics of complex geographical phenomena,but also effectively reveal the relevant influencing factors behind the change of geographical phenomena.Meanwhile,this method can also be used for modeling and analysis of other metrics.In this paper,we proposed the scale effect of category raster data aggregation to establish the overall idea,and then we developed the clustering effect model of raster data based on patch morphology and verified and applied the model.The models established for different types of patches reflect the scale-effects of the plausibility of this type of patch,such as the gradual decay of the area of "line" patches,and the "area" patches decay abruptly at a certain scale.This model can predict the life cycle of the patch,the area of the composite category,and reflect the internal factors of the fluctuation of the category area.The accuracy of the model is related to the patch sample.In the future,the model accuracy can be further improved by increasing the number of patch samples and ensuring the complete coverage of the sample in the form factor space.The method used to establish the scale evolution model of aggregated raster data in this paper can also be used to construct the scale effect of other metrics in future research.