Comparative evaluation of geological disaster susceptibility using multi-regression methods and spatial accuracy validation

Geological disasters not only cause economic losses and ecological destruction, but also seriously threaten human survival. Selecting an appropriate method to evaluate susceptibility to geological disasters is an important part of geological disaster research. The aims of this study are to explore the accuracy and reliability of multi-regression methods for geological disaster susceptibility evaluation, including Logistic Regression(LR), Spatial Autoregression(SAR), Geographical Weighted Regression(GWR), and Support Vector Regression(SVR), all of which have been widely discussed in the literature. In this study, we selected Yunnan Province of China as the research site and collected data on typical geological disaster events and the associated hazards that occurred within the study area to construct a corresponding index system for geological disaster assessment. Four methods were used to model and evaluate geological disaster susceptibility. The predictive capabilities of the methods were verified using the receiver operating characteristic(ROC) curve and the success rate curve. Lastly, spatial accuracy validation was introduced to improve the results of the evaluation, which was demonstrated by the spatial receiver operating characteristic(SROC) curve and the spatial success rate(SSR) curve. The results suggest that: 1) these methods are all valid with respect to the SROC and SSR curves, and the spatial accuracy validation method improved their modelling results and accuracy, such that the area under the curve(AUC) values of the ROC curves increased by about 3%–13% and the AUC of the success rate curve values increased by 15%–20%; 2) the evaluation accuracies of LR, SAR, GWR, and SVR were 0.8325, 0.8393, 0.8370 and 0.8539, which proved the four statistical regression methods all have good evaluation capability for geological disaster susceptibility evaluation and the evaluation results of SVR are more reasonable than others; 3) according to the evaluation results of SVR, the central-southern Yunnan Province are the highest sus-ceptibility areas and the lowest susceptibility is mainly located in the central and northern parts of the study area.

A Vector-based Cellular Automata Model for Simulating Urban Land Use Change

Cellular Automata(CA) is widely used for the simulation of land use changes. This study applied a vector-based CA model to simulate land use change in order to minimize or eliminate the scale sensitivity in traditional raster-based CA model. The cells of vector-based CA model are presented according to the shapes and attributes of geographic entities, and the transition rules of vector-based CA model are improved by taking spatial variables of the study area into consideration. The vector-based CA model is applied to simulate land use changes in downtown of Qidong City, Jiangsu Province, China and its validation is confirmed by the methods of visual assessment and spatial accuracy. The simulation result of vector-based CA model reveals that nearly 75% of newly increased urban cells are located in the northwest and southwest parts of the study area from 2002 to 2007, which is in consistent with real land use map. In addition, the simulation results of the vector-based and raster-based CA models are compared to real land use data and their spatial accuracies are found to be 84.0% and 81.9%, respectively. In conclusion, results from this study indicate that the vector-based CA model is a practical and applicable method for the simulation of urbanization processes.

Progress on spatial prediction methods for soil particle-size fractions

Soil particle-size fractions(PSFs),including three components of sand,silt,and clay,are very improtant for the simulation of land-surface process and the evaluation of ecosystem services.Accurate spatial prediction of soil PSFs can help better understand the simulation processes of these models.Because soil PSFs are compositional data,there are some special demands such as the constant sum(1 or 100%) in the interpolation process.In addition,the performance of spatial prediction methods can mostly affect the accuracy of the spatial distributions.Here,we proposed a framework for the spatial prediction of soil PSFs.It included log-ratio transformation methods of soil PSFs(additive log-ratio,centered log-ratio,symmetry log-ratio,and isometric log-ratio methods),interpolation methods(geostatistical methods,regression models,and machine learning models),validation methods(probability sampling,data splitting,and cross-validation) and indices of accuracy assessments in soil PSF interpolation and soil texture classification(rank correlation coefficient,mean error,root mean square error,mean absolute error,coefficient of determination,Aitchison distance,standardized residual sum of squares,overall accuracy,Kappa coefficient,and Precision-Recall curve) and uncertainty analysis indices(prediction and confidence intervals,standard deviation,and confusion index).Moreover,we summarized several paths on improving the accuracy of soil PSF interpolation,such as improving data distribution through effective data transformation,choosing appropriate prediction methods according to the data distribution,combining auxiliary variables to improve mapping accuracy and distribution rationality,improving interpolation accuracy using hybrid models,and developing multi-component joint models.In the future,we should pay more attention to the principles and mechanisms of data transformation,joint simulation models and high accuracy surface modeling methods for multi-components,as well as the combination of soil particle size curves with stochastic simulations.We proposed a clear framework for improving the performance of the prediction methods for soil PSFs,which can be referenced by other researchers in digital soil sciences.