Spatial modeling of flood probability using geo-environmental variables and machine learning models,case study: Tajan watershed,Iran

The main objective of this study was to produce flood susceptibility maps for Tajan watershed, Sari, Iran using three machine learning (ML) models including Self-Organization Map (SOM), Radial Basis Function Neural Network (RBFNN), and Multi-layers Perceptron (MLP). To reach such a goal, different physical-geographical factors (criteria) were integrated and mapped. 212 flood inventory map was randomly divided into training and testing datasets, where 148 flood locations (70%) were used for training and the remaining 64 locations (30%) were employed for testing. Model validation was performed using several statistical indices and the area under the curve (AUC). The results of the correlation matrix showed, three factors slope (0.277), distance from river (0.263), and altitude (0.223) were the most important factors affecting flood. The accuracy evaluation of the flood susceptibility maps through the AUC method and K-index shows that in the validation phase RBFNN (AUC = 0.90) outperform the MLP (AUC = 0.839) and SOM (AUC = 0.882) models. The highest percentage flood susceptibility of the area in MLP, SOM and RBFNN models is related to moderate (28.7%), very low (40%) and low (37%), respectively. Also, the validation results of the models using the Relative Flood Density (RFD) approach showed that very high class had the highest RFD value.     

Flood susceptibility mapping by ensemble evidential belief function and binomial logistic regression model on river basin of eastern India

Flooding is the overflow of water from stream, river, lake and sea that occurs all over the world and has disastrous effects on human society and environment. Frequent severe flood event in eastern India cause of death and damages every year so, the development of flood susceptibility method is needed for identifying the flood vulnerability areas to reduce the damages. Techniques of Remote Sensing (RS) and Geographical Information System (GIS) can help to flood susceptibility modeling by the accrued and analyzing huge amount of data in short time. The main objectives of this study are to determine the effectiveness of Evidence Belief Function (EBF), binomial Logistic Regression (LR) and ensemble of EBF and LR (EBF-LR) model with RS and GIS techniques for flood susceptibility mapping and spatial prediction of flood-susceptible areas in the Koiya river basin of West Bengal, India. Eight flood conditioning factors; Land use and land cover (LULC), soil, rainfall, normalized differences vegetation index (NDVI), distance to river, elevation, topographic wetness index (TWI) and stream power index (SPI) have been used, and total 264 historical flooding points were mapped, and randomly divided in to training (70%) and validating (30%) dataset. Flood susceptibility map has been generated by applying EBF, LR and ensemble EBF-LR method with the help of training and eight causative factors dataset. The maps have been divided in to six classes; extremely low, very low, low, moderate, high and very high. The receiver operating characteristic (ROC) curve has been used to accuracy assessment of the susceptibility map, and the area under curve (AUC) disclosures 87.9%, 85.2% and 84.1% prediction rate for the EBF-LR, EBF and LR model, respectively. This study is helpful to flood management program, dissection makers and planning in local administrative level.     

冰雪天气下基于MFOA-KELM残差修正的跑道温度混合预测

道面温度短时精准预测是跑道积冰预警的关键因素之一, 为了解决单一机理预测模型随预测时间延长而造成误差累积的问题, 提出了一种冰雪天气下跑道温度混合预测方法。将跑道温度机理预测模型与核极限学习机(KELM)相结合, 建立一种数据驱动修正残差的跑道温度机理预测模型。针对果蝇优化算法(FOA)收敛速度慢、易陷入局部最小值的问题, 引入权值更新函数和距离扩充因子, 调整果蝇的全局寻优效果, 避免陷入局部极小值。利用改进的果蝇优化算法(MFOA)对KELM的正则化参数与核参数联合优化, 以冰雪天气下跑道温度实际数据为例, 建立基于改进果蝇优化核极限学习机(MFOA-KELM)的跑道温度混合预测模型, 并在不同时间尺度下对该混合预测模型进行仿真测试。实验结果表明:与单一机理预测模型相比, 当预测时长为120 min时, MFOA-KELM混合预测模型的平均绝对误差至少减小了61.43%, 在残差阈值为±0.5℃时, 平均预测准确率为91.25%。可见, MFOA-KELM混合预测模型具有更高的预测准确性, 研究结论显示该混合预测方法能够为机场跑道温度短时精准预测提供新思路。      

Deep learning-based framework for monitoring of debris-covered glacier from remotely sensed images

In recent years, deep learning (DL) methods have proven their efficiency for various computer vision (CV) tasks such as image classification, natural language processing, and object detection. However, training a DL model is expensive in terms of both complexities of the network structure and the amount of labeled data needed. In addition, the imbalance among available labeled data for different classes of interest may also adversely affect the model accuracy. This paper addresses these issues using a new convolutional neural network (CNN) based architecture. The proposed network incorporates both spatial and spectral information that combines two sub-networks: spatial-CNN and spectral-CNN. The spectral-CNN extracts spectral information, while spatial-CNN captures spatial information. Moreover, to make the features more robust, a multiscale spatial CNN architecture is introduced using different kernels. The final feature vector is formed by concatenating the outputs obtained from both spatial-CNN and spectral-CNN. To address the data imbalance problem, a generative adversarial network (GAN) was used to generate data for the underrepresented class. Finally, relatively a shallower network architecture was used to reduce the number of parameters in the network and improve the processing speed. The proposed model was trained and tested on Senitel-2 images for the classification of the debris-covered glacier. The results showed that the proposed method is well-suited for mapping and monitoring debris-covered glaciers at a large scale with high classification accuracy. In addition, we compared the proposed method with conventional machine learning approaches, support vector machine (SVM), random forest (RF) and multilayer perceptron (MLP).     

Advanced ensemble modeling method for space object state prediction accounting for uncertainty in atmospheric density

For objects in the low Earth orbit region, uncertainty in atmospheric density estimation is an important source of orbit prediction error, which is critical for space traffic management activities such as the satellite conjunction analysis. This paper investigates the evolution of orbit error distribution in the presence of atmospheric density uncertainties, which are modeled using probabilistic machine learning techniques. The recently proposed “HASDM-ML,” “CHAMP-ML,” and “MSIS-UQ” machine learning models for density estimation (Licata and Mehta, 2022b; Licata et al., 2022b) are used in this work. The investigation is convoluted because of the spatial and temporal correlation of the atmospheric density values. We develop several Monte Carlo methods, each capturing a different spatiotemporal density correlation, to study the effects of density uncertainty on orbit uncertainty propagation. However, Monte Carlo analysis is computationally expensive, so a faster method based on the Kalman filtering technique for orbit uncertainty propagation is also explored. It is difficult to translate the uncertainty in atmospheric density to the uncertainty in orbital states under a standard extended Kalman filter or unscented Kalman filter framework. This work uses the so-called “consider covariance sigma point (CCSP)” filter that can account for the density uncertainties during orbit propagation. As a test-bed for validation purposes, a comparison between CCSP and Monte Carlo methods of orbit uncertainty propagation is carried out. Finally, using the HASDM-ML, CHAMP-ML, and MSIS-UQ density models, we propose an ensemble approach for orbit uncertainty quantification for four different space weather conditions.     

A multilayer perceptron and multiclass support vector machine based high accuracy technique for daily rainfall estimation from MSG SEVIRI data

The current paper introduces a new multilayer perceptron (MLP) and support vector machine (SVM) based approach to improve daily rainfall estimation from the Meteosat Second Generation (MSG) data. In this study, the precipitation is first detected and classified into convective and stratiform rain by two MLP models, and then four multi-class SVM algorithms were used for daily rainfall estimation. Relevant spectral and textural input features of the developed algorithms were derived from the spectral MSG SEVIRI radiometer channels. The models were trained using radar rainfall data set colected over north Algeria. Validation of the proposed daily rainfall estimation technique was performed by rain gauge network data set recorded over north Algeria. Thus, several statistical scores were calculated, such as correlation coefficient (r), root mean square error (RMSE), mean error (Bias), and mean absolute error (MAE). The findings given by: (r = 0.97, bias = 0.31 mm, RMSE = 2.20 mm and MAE = 1.07 mm), showed a quite satisfactory relationship between the estimation and the respective observed daily precipitation. Moreover, the comparison of the results with those of two advanced techniques based on random forests (RF) and weighted ‘k’ nearest neighbor (WkNN) showed higher accuracy obtained by the proposed model.     

高精度多维限制器的性能分析

目前常用的限制器大都是基于一维构造,无法在多维情况下保证物理量的单调特性进而导致非物理振荡.为弥补传统方法的这一构造缺陷,多维限制器(MLP)通过多维修正使单元通量值介于周围相邻单元通量的最大值和最小值之间,在保证求解精度的情况下有效避免了多维振荡.基于一维激波管、无黏涡及激波边界层干扰等算例,对高精度MLP的特性进行了研究分析.结果显示:3阶MLP在连续和间断区域均可有效地避免多维振荡;与高阶WENO(Weighted Essentially Non-Oscillatory)方法相比,3阶MLP不仅算法简单、易于实现,还可显著提高求解的精度、保单调性及收敛性.因此可用于工程及科学研究的复杂流动,具有较好的应用前景.      

基于集成学习的太阳质子事件短期预报方法

太阳质子事件是一种由太阳活动爆发时喷射并传播到近地空间的高能粒子引起的空间天气现象。这些高能粒子会对航天器和宇航员产生严重危害,对太阳质子事件进行准确的短期预报是航天活动灾害预防的重要内容。针对当前主要预报模型中普遍存在的高虚报率问题,提出了一种基于集成学习的太阳质子事件短期预报方法,利用第23个太阳活动周数据,建立了一种集成8种机器学习模型的太阳质子事件短期预报系统。实验结果表明,本文方法在取得了80.95%的报准率的同时,将虚报率降低至19.05%,相比现有的预报系统具有较为明显的优势。      

基于GA-SVM的GNSS-IR土壤湿度反演方法

针对提高大范围土壤湿度测量精度的问题,研究了土壤湿度的全球卫星导航系统干涉测量法（GNSS-IR）,提出了一种基于支持向量机（SVM）的土壤湿度反演模型,利用遗传算法（GA）的自动寻优功能寻找SVM的最佳参数。结果表明,GA-SVM模型在测试集上得到的土壤湿度反演值与实测值的平均绝对百分比误差（MAPE）仅为0.69%,最大相对误差（MRE）为1.22%,线性回归方程决定系数达到了0.956 9。进一步与统计回归、粒子群优化的SVM模型（PSO-SVM）及反向传播（BP）神经网络方法进行对比,结果说明:在样本数目有限的情况下,GA-SVM方法更适用于土壤湿度的GNSS-IR技术反演,且反演精度较高,泛化性能良好。      

Prediction of landslides using ASTER imagery and data mining models

The aim of this study was to identify landslide-related factors using only remotely sensed data and to present landslide susceptibility maps using a geographic information system, data-mining models, an artificial neural network (ANN), and an adaptive neuro-fuzzy interface system (ANFIS). Landslide-related factors were identified in Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) satellite imagery. The slope, aspect, and curvature of topographic features were calculated from a digital elevation model that was made using the ASTER imagery. Lineaments, land-cover, and normalized difference vegetative index layers were also extracted from the imagery. Landslide-susceptible areas were analyzed and mapped based on occurrence factors using the ANN and ANFIS. The generalized bell-shaped built-in membership function of the ANFIS was applied to landslide susceptibility mapping. Analytical results were validated using landslide test location data. In the validation results, the ANN model showed 80.42% prediction accuracy and the ANFIS model showed 86.55% prediction accuracy. These results suggest that the ANFIS model has a better performance than does the ANN in predicting landslide susceptibility.     

基于代理模型的制导火箭炮发射诸元计算方法

针对制导火箭炮发射诸元的快速计算问题,提出了一种结合大样本数据和代理模型计算发射诸元的新方法。运用代理模型建立射角、无控弹道侧偏与炮位纬度、炮位高程、射向、射程、目标点高程及药温之间的函数关系,并根据射程和无控弹道侧偏的预测值对射向进行修正。仿真结果表明,高阶多项式响应面、相关函数为高斯函数的Kriging、高阶单项式径向基函数、核函数为高斯函数的最小二乘支持向量机、激活函数为正弦函数的超限学习机以及由上述单一代理模型构建的组合代理模型均具有较高的预测精度,各种单一代理模型对射角和无控弹道侧偏的预测时间均小于1 ms,证明了基于代理模型的射角和无控弹道侧偏预测方法切实可行,且通过对射向进行修正有效减小了由于地球自转引起的无控弹道侧偏。      

Evaluating the predictability of eight Atmospheric-Oceanic signals affecting Iran’s Droughts,employing intelligence based and stochastic methods

The large-scale atmospheric-oceanic phenomena are among the main effective factors in the droughts in the Middle East, especially in Iran. Since these effects are usually delayed, their relevant signals can be useful for predicting droughts. As a result, the provision of a precise prediction of these signals can be efficient in increasing the drought prediction prospect. The current study predicts 8 cases of the most effective oceanic signals on the droughts which have been investigated in Iran. To do so, the problem-solving method with the time series prediction approach is based on the two model types intelligence-based (including multilayer perceptron [MLP] and support vector machine [SVM]) and stochastic (including Autoregressive Integrated Moving Average [ARIMA]) has been used. The model's input for each index included the time lags of the same index itself, which was determined by the autocorrelation function. Based on the evaluation criteria, the results were indicative of the weak predictability of the North Atlantic Oscillation (NAO) and Arctic Oscillation (AO), while the Extreme Eastern Tropical Pacific sea surface temperature (Niño [1 + 2]), East Central Tropical Pacific sea surface temperature (Niño [3 + 4]), and Oceanic Niño Index (ONI) were predicted with very good accuracy, and there is a high overlap between their predictions and observations (95.9 % < R² < 99.3 %). In the extreme events also, the rate of normalized forecasting error for Niño (1 + 2), Niño (3 + 4), and ONI were in the medium (20–30 %), good (10–20 %), and excellent (0–10 %) ranges, respectively. The comparison between the models also indicates a partial superiority of the ARIMA stochastic model over the SVM and MLP models. The overall results of the study are indicative of the applicability of the predictions of the three mentioned indices as the inputs to increase precipitation and drought forecasting prospects in Iran (as well as all regions affected by them); which have the research value for further studies in terms of drought forecasting.     

Environmental degradation analysis using NOAA/AVHRR data

This paper proposes a particular approach to assess information about soil degradation, based on a methodology to calculate soil color from NOAA/AVHRR data. As erosive processes change physical and chemical properties of the soil, altering, consequently, the superficial color, monitoring the change in color over time can help to identify and analyze those processes. A relationship among the soil color (described in the Munsell Color System, i.e., in terms of Hue, Value and Chroma), vegetation indices, surface temperature and emissivity has been established, which is based on the theoretical model. The methodology has three main phases: determination of the regression models among soil color and vegetation indices, emissivity and surface temperature; generation of digital soil color models; and statistical evaluation of the estimated color. The tests showed that the methodology is efficient in determining soil color using the various vegetation indices (i.e., Normalized vegetation index NDVI, Modified soil adjusted vegetation index MSAVI). One vegetation index, i.e., Purified adjusted vegetation index (PAVI) is proposed to subsidies the effect of vegetation over the soil. Best results were obtained for the Hue color component. To further test the methodology, the estimated digital color models were compared with the characteristic color of soil classes in the test area. The results of this application confirmed the methodology’s capacity to determine the soil color from NOAA/AVHRR data. This type of study is quite helpful to know the erosion of soil as well as some abrupt change in soil due to natural hazards by space borne or air-borne sensors.     

基于树模型机器学习方法的GNSS-R海面风速反演

GNSS-R是基于GNSS卫星反射信号的一种新技术.GNSS-R技术可以运用到海面风场反演中,传统的GNSS-R技术反演海面风场主要有波形匹配和经验函数两种方法,风速反演精度约为2m·s-1.波形匹配方法耗时多,计算量大;经验函数方法通常只使用少量物理观测量,会造成信息浪费,损失一定的反演精度.为了提高海面风速的反演精度,引入机器学习领域常用的树模型算法决策树、随机森林、GBDT等对海面风速进行预测.利用GNSS-R与ECMWF数据构成训练集和验证集,训练集用于模型学习,验证集用于检验模型的反演效果.实验结果显示,决策树和随机森林预测误差约为0.6m·s-1,GBDT等算法的预测误差约为2m·s-1,满足风速反演要求.与GNSS-R传统反演方法相比,机器学习树模型算法效果更好,在验证集上表现稳定且误差较小.因此,可以将机器学习树模型算法运用到海面风速反演中.      

基于机器学习方法对23太阳活动周质子事件的研究

在耀斑伴随日冕物质抛射(CME)事件编目数据的基础上,进行太阳质子事件(SPE)匹配,构建研究数据集.利用Apriori算法挖掘SPE与耀斑级别、耀斑发生日面位置以及CME角宽度和速度的关联关系.结果 表明:X级耀斑、全晕CME、高速(＞1000 km.s-1) CME和日面西半球耀斑是最可能伴随质子事件的4种特征,其...     

Improving orbit prediction accuracy through supervised machine learning

Due to the lack of information such as the space environment condition and resident space objects’ (RSOs’) body characteristics, current orbit predictions that are solely grounded on physics-based models may fail to achieve required accuracy for collision avoidance and have led to satellite collisions already. This paper presents a methodology to predict RSOs’ trajectories with higher accuracy than that of the current methods. Inspired by the machine learning (ML) theory through which the models are learned based on large amounts of observed data and the prediction is conducted without explicitly modeling space objects and space environment, the proposed ML approach integrates physics-based orbit prediction algorithms with a learning-based process that focuses on reducing the prediction errors. Using a simulation-based space catalog environment as the test bed, the paper demonstrates three types of generalization capability for the proposed ML approach: (1) the ML model can be used to improve the same RSO’s orbit information that is not available during the learning process but shares the same time interval as the training data; (2) the ML model can be used to improve predictions of the same RSO at future epochs; and (3) the ML model based on a RSO can be applied to other RSOs that share some common features.     

An ensemble deep learning based shoreline segmentation approach (WaterNet) from Landsat 8 OLI images

Shorelines constantly vary due to natural, urbanization and anthropogenic effects such as global warming, population growth, and environmental pollution. Sustainable monitoring of coastal changes is vital in terms of coastal resource management, environmental preservation and planning. Publicly available Landsat 8 OLI (Operational Land Manager) images provide accurate, reliable, temporal and up-to-date information about coastal areas. Recently, the use of machine learning and deep learning algorithms have become widespread. In this study, we used our public Landsat 8 OLI satellite image dataset to create a majority voting method which is an ensemble automatic shoreline segmentation system (WaterNet) to obtain shorelines automatically. For this purpose, different deep learning architectures have been utilized namely as Standard U-Net, Dilated U-Net, Fractal U-Net, FC-DenseNet, and Pix2Pix. Also, we have suggested a novel framework to create labeling data from OpenStreetMap service to create a unique dataset called YTU-WaterNet. According to the results, IoU and F1 scores have been calculated as 99.59% and 99.79% for the WaterNet. The results indicate that the WaterNet method outperforms other methods in terms of shoreline extraction from Landsat 8 OLI satellite images.     

基于误差最小的SVM最优分类面修正

针对C-支持向量机(C-SVM,C-Support Vector Machine)中惩罚系数C可能导致最优分类面不合理的问题,提出基于误差最小的SVM最优分类面修正方法.通过调整正负类分类间隔的约束条件,求解使训练样本总误差最小的偏置系数,并兼顾与正负类误差之差的绝对值的平衡,得到误差最小的更优分类面.实验证明该修正方法与C-SVM及其它修正方法相比,具有较高的分类精度和较强的抗噪声与野值数据干扰能力.     

舵面电动加载系统的自适应CMAC复合控制

针对无人机舵面电动加载系统具有非线性及多余力矩的特点,提出了一种自适应CMAC(Cerebellar Model Articulation Controller)神经网络与自适应神经元控制器并联构成复合控制结构.该控制策略以系统的指令输入和实际输出作为CMAC的激励信号,以系统的当前控制误差作为CMAC的训练信号.提出了利用误差在线自适应调整学习率的方法,消除了常规前馈型CMAC的过学习和不稳定现象.建立了无人机舵面电动加载系统的数学模型,给出了具体的控制结构和算法.仿真结果表明:该方法有效抑制了加载系统的多余力矩,增强了系统的稳定性,明显改善了舵面电动加载系统的动态性能.     

An improved cyclic multi model-eXtreme gradient boosting (CMM-XGBoost) forecasting algorithm on the GNSS vertical time series

The study of GNSS vertical coordinate time series forecasting is helpful for monitoring the crustal plate movement, dam or bridge deformation monitoring, and global or regional coordinate system maintenance. The eXtreme Gradient Boosting (XGBoost) algorithm is a machine learning algorithm that can evaluate features, and it has a great potential and stability for long-span time series forecasting. This study proposes a multi-model combined forecasting method based on the XGBoost algorithm. The method constitutes a new time series as features through the fitting and forecasting results of the forecasting model. The XGBoost model is then used for forecasting. In addition, this method can obtain higher precision forecasting results through circulation. To verify the performance of the forecasting method, 1095 epochs of data in the Up coordinate of 16 GNSS stations are selected for the forecasting test. Compared with the CNN-LSTM model, the experimental results of our forecasting method show that the mean absolute error (MAE) values are reduced by 30.23 %～52.50 % and the root mean square error (RMSE) values are reduced by 31.92 %～54.33 %. The forecasting results have higher accuracy and are highly correlated to the original time series, which can better forecast the vertical movement of the GNSS stations. Therefore, the forecasting method can be applied to the up component of the GNSS coordinate time series.