Commonly used data-driven models for landslide susceptibility prediction (LSP) can be mainly classified as heuristic, general statistical or machine learning models. This study plans to compare the prediction performance of these data-driven models on the landslide susceptibility mapping, thus further to explore the inherently features of these data-driven models. As a result, a more accurate and reliable LSP can be realized through choosing an optimal data-based model. A heuristic model represented by the analytic hierarchy process (AHP), a general statistical model represented by the general linear model (GLM) and information value (IV) model, and machine learning models represented by binary logistic regression (BLR), Multilayer Perceptron (MLP), back-propagation neural network (BPNN), support vector machine (SVM) and C5.0 decision tree (C5.0 DT) are adopted in this study. Shicheng County in China is used as the study area. In total, 369 landslides identified through field investigation are classified as training (70%) and testing datasets (30%). Next, 13 landslide conditioning factors (elevation, slope, aspect, plan curvature, profile curvature, relief amplitude, total surface radiation, population density, Normalized difference vegetation index, distance to river, topographic wetness index and rock types) are acquired from data sources of the free remote sensing images, Digital Elevation Model, field investigation and government reports. The correlations between these conditioning factors and the landslide locations are determined by frequency ratio analysis. Then, the landslide susceptibility indexes (LSIs) calculated by the eight trained models are imported into GIS software to produce landslide susceptibility maps of Shicheng County. Finally, the area under receiver operating characteristic curve (AUC), the calculated LSIs are applied to assess the LSP performance of the present eight models. The testing results show that these eight models generate reasonable LSP results as a whole, further showing that the C5.0 DT is of the highest prediction accuracy with an AUC value of 0.868, followed by the SVM (0.813), BPNN (0.803), MLP (0.792), BLR (0.784), GLM (0.779), IV (0.774) and AHP (0.773). It can be inferred that the machine learning models have higher LSP performance than general statistical and heuristic models due to its high AUC accuracy and reasonable LSIs distribution features, while general statistical model is limited by its linear analysis and heuristic model is limited by subjective weighting process.

Comparisons of heuristic, general statistical and machine learning models for landslide susceptibility prediction and mapping

Artificial intelligence has been applied in wildfire science and management since the 1990s, with early applications including neural networks and expert systems. Since then, the field has rapidly ...

/pdf/a-review-of-machine-learning-applications-in-wildfire-3yit0f75yl.pdf

A review of machine learning applications in wildfire science and management

Artificial intelligence has been applied in wildfire science and management since the 1990s, with early applications including neural networks and expert systems. Since then the field has rapidly progressed congruently with the wide adoption of machine learning (ML) in the environmental sciences. Here, we present a scoping review of ML in wildfire science and management. Our objective is to improve awareness of ML among wildfire scientists and managers, as well as illustrate the challenging range of problems in wildfire science available to data scientists. We first present an overview of popular ML approaches used in wildfire science to date, and then review their use in wildfire science within six problem domains: 1) fuels characterization, fire detection, and mapping; 2) fire weather and climate change; 3) fire occurrence, susceptibility, and risk; 4) fire behavior prediction; 5) fire effects; and 6) fire management. We also discuss the advantages and limitations of various ML approaches and identify opportunities for future advances in wildfire science and management within a data science context. We identified 298 relevant publications, where the most frequently used ML methods included random forests, MaxEnt, artificial neural networks, decision trees, support vector machines, and genetic algorithms. There exists opportunities to apply more current ML methods (e.g., deep learning and agent based learning) in wildfire science. However, despite the ability of ML models to learn on their own, expertise in wildfire science is necessary to ensure realistic modelling of fire processes across multiple scales, while the complexity of some ML methods requires sophisticated knowledge for their application. Finally, we stress that the wildfire research and management community plays an active role in providing relevant, high quality data for use by practitioners of ML methods.

River flooding can be a highly destructive natural hazard. Numerous approaches have been used to study the phenomenon; however, insufficient knowledge regarding flood conditioning factors continues to hinder prevention and control measures. This research examines the hypothesis that by adding further conditioning factors to a dataset used in river flood modeling, increases the accuracy of the final susceptibility mapping result. Additionally, this study assesses the impact of individual conditioning factors on flood susceptibility mapping and their importance in the construction of precise mapping of potential flood regions. Two robust machine learning approaches, Decision Tree (DT) and Support Vector Machine (SVM), were utilized to evaluate spatial correlations between flood conditioning factors and rate their level of importance for mapping the flood prone areas. For this purpose, two datasets were used; dataset 1 (DS1): Light Detection and Ranging (LiDAR) derived factors of altitude, slope, aspect, curvature, Stream Power Index (SPI), Topographic Wetness Index (TWI), Topographic Roughness Index (TRI), and Sediment Transport Index (STI) and dataset 2 (DS2): a combination of LiDAR derived factors supplemented by geology, soil, landuse/cover (LULC), distance from roads and distance from rivers parameters. An extreme flood event in 2011 in Brisbane, Australia was used as a case study, in which DT and SVM techniques were both applied, using both datasets. In addition, multi-collinearity, variance inflation factors (VIF), Pearson's correlation coefficients and Cohen's kappa analysis provided useful information regarding the inter-relationships of factors, as well as the influence of each factor on the precision of the final map. The area under curve (AUC) method was used for accuracy assessment. SVM and DT produced the highest accuracies of prediction, with rates of 85.52% and 88.47% respectively, using DS1 (the LiDAR dataset). Altitude, SPI and TRI were found to have a significant impact on the precision of the outcomes. It was concluded that the inclusion of additional factors in the modeling, does not necessarily guarantee the achievement of greater accuracy. However, the modeling method, can significantly alter outcomes.

Identifying the essential flood conditioning factors for flood prone area mapping using machine learning techniques

Forest fires have caused considerable losses to ecologies, societies, and economies worldwide. To minimize these losses and reduce forest fires, modeling and predicting the occurrence of forest fires are meaningful because they can support forest fire prevention and management. In recent years, the convolutional neural network (CNN) has become an important state-of-the-art deep learning algorithm, and its implementation has enriched many fields. Therefore, we proposed a spatial prediction model for forest fire susceptibility using a CNN. Past forest fire locations in Yunnan Province, China, from 2002 to 2010, and a set of 14 forest fire influencing factors were mapped using a geographic information system. Oversampling was applied to eliminate the class imbalance, and proportional stratified sampling was used to construct the training/validation sample libraries. A CNN architecture that is suitable for the prediction of forest fire susceptibility was designed and hyperparameters were optimized to improve the prediction accuracy. Then, the test dataset was fed into the trained model to construct the spatial prediction map of forest fire susceptibility in Yunnan Province. Finally, the prediction performance of the proposed model was assessed using several statistical measures—Wilcoxon signed-rank test, receiver operating characteristic curve, and area under the curve (AUC). The results confirmed the higher accuracy of the proposed CNN model (AUC 0.86) than those of the random forests, support vector machine, multilayer perceptron neural network, and kernel logistic regression benchmark classifiers. The CNN has stronger fitting and classification abilities and can make full use of neighborhood information, which is a promising alternative for the spatial prediction of forest fire susceptibility. This research extends the application of CNN to the prediction of forest fire susceptibility.

https://link.springer.com/content/pdf/10.1007%2Fs13753-019-00233-1.pdf

Forest Fire Susceptibility Modeling Using a Convolutional Neural Network for Yunnan Province of China

Rampant pasture burning has lead to various forest fires taking their toll over the health of many forests. Nanda Devi Biosphere Reserve, located in the northern part of India, witnessed a majority of these incidents in the recent past, though, it remains comprehensively untouched from research studies. The scale of these wildfires has led to an immense requirement of preventive measures to be taken for recuperating from such events. This requires for an in-depth analysis of the study area, its history of wildfires and their causes. These efforts would assist in laying a blueprint for a contingency plan in the event of a wildfire. This work proposes an evolutionary optimized gradient boosted decision trees for preparing wildfire susceptibility maps for the study area that would aid in the government’s forest preservation and disaster management activities. The study took 18 ignition factors of elevation, slope, aspect, plan curvature, topographic position index, topographic water index, normalized difference vegetation index, soil texture, temperature, rainfall, aridity index, potential evapotranspiration, relative humidity, wind speed, land cover and distance from roads, rivers and habitations into consideration. The study revealed that approximately 1432.025 km2 of area was very highly susceptible to forest fires while 1202.356 km2 was highly susceptible to forest fires. The proposed model was compared against various machine learning models such as random forest, neural networks and support vector machines, and it outperformed them by achieving an overall accuracy of 95.5%. The proposed model demonstrated good prospects for application in the field of hazard susceptibility mappings.

GIS-based evolutionary optimized Gradient Boosted Decision Trees for forest fire susceptibility mapping

July 2013 unleashed one of the worst floods in the history of mankind for the North Indian state of Uttarakhand. Four years later and still little can be said about the state of affairs in the region that is still recuperating from the loss of lives, property and biodiversity. This paper attempts to utilize the state-of-the-art machine learning techniques for flood susceptibility assessment in the flood ravaged district of Chamoli. Remote Sensing and Geographic Information Systems hand-in-glove with procured satellite images unravel the mystery behind the district's traits that conditioned it for the final encounter with the unfortunate disaster. The predictive accuracy of the proposed particle swarm optimized-support vector machine was compared against various conventional models like Random Forest, Neural Networks. Outperforming them, it achieved an accuracy of 96.55%. Flood susceptibility map for the study area was generated indicating that approximately 1,832.3 sq. km area is highly susceptible to floods.

Flood susceptibility mapping using GIS-based support vector machine and particle swarm optimization: A case study in Uttarakhand (India)

Inherent complex topography and drastic weather patterns together have concocted various natural disasters worldwide. In difficult terrains such as those prevalent in the North-Eastern regions of I...

Shruti Sachdeva

Papers

GIS-based evolutionary optimized Gradient Boosted Decision Trees for forest fire susceptibility mapping

Flood susceptibility mapping using GIS-based support vector machine and particle swarm optimization: A case study in Uttarakhand (India)

A novel voting ensemble model for spatial prediction of landslides using GIS