Assessment of suffosion-related hazards in karst areas using GIS technology

This paper describes an application of the geographic information system (GIS) technology to a ground stability assessment in the karst area of Dzerzhinsk, Russia. In the stability analysis, the groundwater level changes in the karst aquifer could cause suffosion sinkholes when the gravitational force was greater than the soil strength. The GIS technology was used to combine various data and to delineate the zones of potential gravitational collapse and suffosion collapse in the area.

Landslide susceptibility mapping using hybrid random forest with GeoDetector and RFE for factor optimization

The present study aims to develop two hybrid models to optimize the factors and enhance the predictive ability of the landslide susceptibility models. For this,...     

Recognition of large scale deep-seated landslides in forest areas of Taiwan using high resolution topography

Large deep-seated landslides can be reactivated during intense events, and they can evolve into destructive failures. They are generally difficult to recognize in the field, especially when they develop in densely forested areas. A detailed and constantly updated inventory map of such phenomena, and the recognition of their topographic signatures is absolutely a key tool for landslide risk mitigation.The aim of this work is to test in forested areas, the performance of the new automatic and objective methodology developed by Tarolli et al. (2012) for geomorphic features extraction (landslide crowns) from high resolution topography (LiDAR derived Digital Terrain Models – DTMs). The methodology is based on the detection of landslides through the use of thresholds obtained by the statistical analysis of variability of landform curvature. The study was conducted in a high-risk area located in the central-south Taiwan, where an accurate field survey on landsliding processes and a high-quality set of airborne laser scanner elevation data are available. The area has been chosen because some of the deep-seated landslides are located near human infrastructures and their reactivation is highly dangerous. Thanks to LiDAR’s capability to detect the bare ground elevation data in forested areas, it was possible to recognize in detail landslide features also in remote regions difficult to access. The results, if compared with the previous work of Tarolli et al. (2012), mainly focused on shallow landslides, and in a not forested area, indicate that for deep-seated landslides, where the crowns are more evident, and they are present at large scale, the tested methodology performs better (higher quality index). The method can be used to interactively assist the interpreter/user on the task of deep-seated landslide hazard mapping, and risk assessment planning of such regions.     

Assessment of rockburst risk using multivariate adaptive regression splines and deep forest model

Acta Geotechnica - Rockburst is a major instability issue faced by underground excavation projects, which is induced by the instantaneous release of a large amount of strain energy stored in rock...     

Prediction of undrained shear strength using extreme gradient boosting and random forest based on Bayesian optimization

Accurate assessment of undrained shear strength(USS)for soft sensitive clays is a great concern in geotechnical engineering practice.This study applies novel data-driven extreme gradient boosting(XGBoost)and random forest(RF)ensemble learning methods for capturing the relationships between the USS and various basic soil parameters.Based on the soil data sets from TC304 database,a general approach is developed to predict the USS of soft clays using the two machine learning methods above,where five feature variables including the preconsolidation stress(PS),vertical effective stress(VES),liquid limit(LL),plastic limit(PL)and natural water content(W)are adopted.To reduce the dependence on the rule of thumb and inefficient brute-force search,the Bayesian optimization method is applied to determine the appropriate model hyper-parameters of both XGBoost and RF.The developed models are comprehensively compared with three comparison machine learning methods and two transformation models with respect to predictive accuracy and robustness under 5-fold cross-validation(CV).It is shown that XGBoost-based and RF-based methods outperform these approaches.Besides,the XGBoostbased model provides feature importance ranks,which makes it a promising tool in the prediction of geotechnical parameters and enhances the interpretability of model.     

Mapping of groundwater potential zones in the drought-prone areas of south Madagascar using geospatial techniques

The southern regions of Madagascar have the country’s lowest water supply coverage and are highly vulnerable to drought. Access to potable drinking water is a major challenge for the local population. Chronic droughts lead to annual emergency appeals to save the lives of acute malnourished children. UNICEF’s response consisting in providing potable drinking water through the drilling of boreholes has been challenged by the complex hydrogeology, the low yield of boreholes and high-level salinity of water, the lack of reliable groundwater data and the weak capacity of the drilling sector. These constraints result in a high rate of drilling failure. To improve drilling success and provide more potable drinking water to local communities, it is vital to undertake reliable groundwater investigation.UNICEF Madagascar and the European Union delegation in Madagascar collaborated on the use of satellite imagery to improve sector knowledge and access to safe and clean water for local communities in southern Madagascar. The methodology relies on produce thematic layers of groundwater potential areas. Later, these thematic layers were overlaid with ground-based hydrogeological data to map the groundwater potential zones (GWP) and identify the most suitable sites for borehole siting and drilling. Findings of this study are very encouraging, and the integrated approach used has proven its applicability in mapping groundwater potential areas in the eight drought-affected areas of south Madagascar. The groundwater potential zone map is being used by UNICEF and partners to plan water supply projects and identify the best sites for positioning new boreholes and reduce the likelihood of drilling failure. Additionally, the project developed a database of groundwater resources, which will improve knowledge of the regional hydrogeological context and strengthen the capacity of the water sector. Lessons learnt from this study show that an integration of the groundwater potential zone map with demographics and water demand information will help identifying priority areas for detailed studies. Moreover, a capacity building activity is required for knowledge/technology transfer to the Ministry of Energy, Water and Hydrocarbons (MEEH), allowing the possibility of scaling-up this integrated approach to the rest of Madagascar. Finally, strengthening the capacity of the MEEH and refining this approach as suggested above will certainly help in the pursuit to improve equitable access to safe and clean water for households located in the drought-affected areas of southern Madagascar, allowing them to be more resilient to the effects of climate change.     

Modeling forest fire risk in the northeast of Iran using remote sensing and GIS techniques

Fire in forested areas can be regarded as an environmental disaster which is triggered by either natural forces or anthropogenic activities. Fires are one of the major hazards in forested and grassland areas in the north of Iran. Control of fire is difficult, but it is feasible to map fire risk by geospatial technologies and thereby minimize the frequency of fire occurrences and damages caused by fire. The fire risk models provide a suitable concept to understand characterization of fire risk. Some models are map based, and they combine effectively different forest fire–causing variables with remote sensing data in a GIS environment for identifying and mapping forest fire risk. In this study, Structural Fire Index, Fire Risk Index, and a new index called Hybrid Fire Index were used to delineate fire risk in northeastern Iran that is subjected to frequent forest fire. Vegetation moisture, slope, aspect, elevation, distance from roads, and vicinity to settlements were used as the factors influencing accidental fire starts. These indices were set up by assigning subjective weight values to the classes of the layers based on their sensitivity ratio to fire. Hot spots data derived from MODIS satellite sensor were used to validate the indices. Assessment of the indices with receiver operating characteristic (ROC) curves shows that 76.7 % accuracy of the HFI outperformed the other two indices. According to the Hybrid Fire Index, 57.5 % of the study area is located under high-risk zone, 33 % in medium-risk zone, and the remaining 9.5 % area is located in low-risk zone.     

Assessment of groundwater quality using multivariate statistical techniques in Hashtgerd Plain, Iran

Multivariate statistical techniques, such as cluster analysis (CA), factor analysis (FA), principal component analysis (PCA), and discriminant analysis (DA), were applied for the evaluation of variations and the interpretation of a large complex groundwater quality data set of the Hashtgerd Plain. In view of this, 13 parameters were measured in groundwater of 26 different wells for two periods. Hierarchical CA grouped the 26 sampling sites into two clusters based on the similarity of groundwater quality characteristics. FA based on PCA, was applied to the data sets of the two different groups obtained from CA, and resulted in three and five effective factors explaining 79.56 and 81.57% of the total variance in groundwater quality data sets of the two clusters, respectively. The main factors obtained from FA indicate that the parameters influencing groundwater quality are mainly related to natural (dissolution of soil and rock), point source (domestic wastewater) and non-point source pollution (agriculture and orchard practices) in the sampling sites of Hashtgerd Plain. DA provided an important data reduction as it uses only three parameters, i.e., electrical conductivity (EC), magnesium (Mg²⁺) and pH, affording more than 98% correct assignations, to discriminate between the two clusters of groundwater wells in the plain. Overall, the results of this study present the effectiveness of the combined use of multivariate statistical techniques for interpretation and reduction of a large data set and for identification of sources for effective groundwater quality management.     

基于随机森林算法的青藏高原AMSR2被动微波雪深反演

青藏高原因其复杂的地形地势和和积雪分布使得多种雪深算法未达到理想的精度。基于新一代被动微波数据AMSR2（Advanced Microwave Scanning Radiometer 2）, 应用随机森林算法（Random Forest, RF）将亮温（Brightness Temperature, BT）和亮温差（Brightness Temperature Difference, BTD）作为参数输入, 并将高程和纬度参数引入雪深反演模型中, 经过模拟退火算法进行有效反演因子筛选, 构建了基于随机森林算法的青藏高原雪深反演模型。结果表明： 与AMSR2全球雪深产品相比, 随机森林算法的拟合优度（R²）由0.41提升至0.60, 均方根误差（Root Mean Square Error, RMSE）由7.36 cm降至4.88 cm, 偏差（BIAS）由3.24 cm减小至-0.16 cm, 随机森林雪深反演模型在青藏高原的精度更高; 青藏高原平均海拔超过4 000 m, 当海拔大于青藏高原平均海拔时, 随机森林算法的反演效果最差, 但RMSE仅为3.78 cm, BIAS仅为-0.09 cm; 高原南部（25° ~ 30° N）因其复杂的地势和相对较少的气象站点使得反演效果较差, RMSE为5.94 cm, BIAS为-0.39 cm; 青藏高原的主要土地覆盖类型为草地, 随机森林算法在草地的RMSE约为3 cm, BIAS接近0 cm。     

基于斜坡单元与随机森林模型的元阳县崩滑地质灾害易发性评价

针对基于栅格单元与定性定量方法模型在地质灾害易发性评价中存在模型预测精度低且使用较为频繁的不足与弊端,采用斜坡单元与机器学习方法之一的随机森林模型相结合开展元阳县崩滑地质灾害易发性评价。在ArcGIS中,利用曲率分水岭法划分出7851个斜坡单元。经过大量统计研究与地质环境条件分析,选取工程地质岩组、地貌类型、高程、坡度、坡向、曲率、起伏度、河流距离、断层距离等9个因子作为评价指标,并通过SPSS软件,将9个评价指标与灾点发育特征的关系进行数据分析,得出各评价指标权重。在SPSS中,采用随机森林模型,建立易发性评价模型,将元阳县崩滑地质灾害易发性划分为低、中、高、极高4类,所占面积分别为410.06 km2、470.21 km2、550.02 km2和776.87 km2,分别占元阳县面积的18.58%、21.30%、24.92%和35.20%。经与详查结果对比,评价结果与实际高度吻合。利用ROC曲线得出区划结果精度AUC值为92.7%,区划结果相当好。研究显示,元阳县中部和西南两个部分地质灾害集中,易发性极高。     

Assessment of surface water quality using multivariate statistical analysis techniques: a case study from Ghrib dam,Algeria

This study explores the water quality status and pollution sources in Ghrib Dam, Algeria. It allows us to obtain more accurate information on water quality by applying a series of multivariate statistical techniques, including principal component analysis (PCA)/factor analysis (FA), hierarchical cluster analysis (CA), and multiple regression analysis (MRA). On 19 physicochemical parameters dataset over 5 years and from 6 different sites located in and around the lake. One-way analysis of variance (ANOVA) was used to investigate the statistically considerable spatial and seasonal differences. The results of ANOVA suggest that there exist a statistically significant temporal variation in the water quality of the dam for all parameters. On the other hand, only organic matter has a statistically significant spatial variation. In the multiple linear models, an association between organic and inorganic parameters was found; their origin comes from the mechanical erosion process of agricultural lands in the watershed. The PCA/FA identifies five dominant factors as responsible of the data structure, explaining more than 94.96% of the total variance in the water quality dataset. This suggests that the variations in water compounds’ concentration are mainly related to the multiple anthropogenic activities, as well as natural processes. The results of cluster analysis demonstrate that the sampling stations were divided in two similar groups, which indicates spatial homogeneity. While seasonal grouping has showed that the source of pollution was related to the level of runoff in the seasons.     

A practical trial of landslide detection from single-temporal Landsat8 images using contour-based proposals and random forest: a case study of national Nepal

Landslides are frequent all around the world, causing tremendous loss to human beings. Rapid access to the locations where landslides occur is crucial for emergency response. Most researches in landslide detection from remotely sensed images focus on small regions, which are handpicked. That makes it easy to distinguish landslides from background objects, but hard to apply in practical cases. The complicated non-landslide background pixels increase the difficulty to accurately detect landslides. In this study, we propose a technique framework to remove non-landslide background pixels for national Nepal using 12 Landsat8 images and digital elevation model (DEM). DEM is useful in removing flat areas, where landslides are less likely to occur. The framework consists of three sections: image enhancement, landslide proposal extraction, and detection model setup. Bare land, including landslides, is enhanced using vegetation index after haze/cloud re-movement. Later, calculate connective contours and propose them as potential regions that may contain landslides. For each proposal, calculate texture feature and build detection model using one of the Landsat8 images, which is further applied on other images to check its applicability and robustness. The assessment shows that the method is able to remove 99% of the background pixels in the scale of national Nepal, taking over billions of pixels. Even there is still much to do to achieve high accurate landslide detection results from large-scale images, the experiment validates a strong potential applicability for the proposed method in large-scale landslide-related analysis.     

随机森林回归模型在地下水水质评价的新应用

地下水水质评价是地下水研究的一个重要课题,研究更精确、更适用的地下水水质评价模型具有重要意义。近年来应用较多的评价模型存在适用性小、精确度低的问题,其中主要原因在于现有的水质评价标准中,不同因子的等级分类标准不一,同一因子不同等级取值区间不一致等问题,而现有模型无法克服不同因子自身评价等级不匹配以及无法有效处理多重因子之间的相互影响。针对水质评价存在的问题,本文将研究一个能有效克服上述问题,且精确度较高、适用性更广的水质评价模型。基于北京市地下水监测网中污染源监测网的数据,选取大兴区作为研究区域,以大兴区2018年至2020年持续进行监测的39个点位作为研究点,根据监测点历年监测水样,选取化验数据中影响水质环境较为明显的pH值、Cl^-、NO^-₃、SO₄^2-、Na⁺、NH⁺₄、Mn^-、耗氧量(COD_Mn)、总硬度、溶解性总固体这10项指标作为水质评价因子,利用随机森林回归...     

基于随机森林算法的砂土液化预测方法

砂土液化的影响因素较多且复杂。以唐山大地震的72个场地的实测液化样本数据为例,在不丢失任何信息的前提下,选取了8个砂土液化的判别指标,通过计算样本数据的Gini系数,采用CART算法的决策树对数据的特征属性进行划分。在此基础之上,通过增加多个决策树构造随机森林的方式,在一定程度上降低了单个决策树学习过度造成的过拟合风险,同时,通过10轮交叉验证的方式确定了决策树的最大高度为5,随机森林中决策树的个数为20时,模型的效果达到最佳。研究结果表明,与抗震设计规范中的标贯试验法判别公式相比,决策树模型和随机森林模型的训练结果和预测结果有显著提高,尤其是随机森林模型在训练样本和预测样本上均没有出现误判,稳定性更高。     

基于随机森林回归分析的岩体结构面粗糙度研究

岩体结构面粗糙度系数是快速估算结构面峰值抗剪强度的重要参数。但是结构面轮廓曲线复杂,单一统计参数无法量化表征粗糙度。为解决这一问题,收集了112条结构面轮廓曲线起伏角、起伏度、迹线长度3方面的8项统计参数,利用随机森林回归模型交叉验证的方法评估统计参数的重要性。结果表明：最大起伏度、起伏高度标准偏差、平均起伏角、起伏角标准差、平均相对起伏度及粗糙度剖面指数等6项统计参数重要性占比达到93.2%,且回归拟合系数趋于平稳,基于重要性评估结果建立最优超参数决策树数目(n_tree)为400、参与节点分割的数目(m_try)为2的随机森林回归模型,模型预测结果拟合优度高达98.1%。与基于坡度均方根、结构函数及粗糙度剖面指数等传统线性回归结果对比,随机森林回归模型结果精度更高,误差更小,拟合优度提高6%以上,表明随机森林回归模型更适用于结构面粗糙度反演。     

Parameterization,sensitivity analysis,and inversion: an investigation using groundwater modeling of the surface-mined Tivoli-Guidonia basin (Metropolitan City of Rome,Italy)

With respect to model parameterization and sensitivity analysis, this work uses a practical example to suggest that methods that start with simple models and use computationally frugal model analysis methods remain valuable in any toolbox of model development methods. In this work, groundwater model calibration starts with a simple parameterization that evolves into a moderately complex model. The model is developed for a water management study of the Tivoli-Guidonia basin (Rome, Italy) where surface mining has been conducted in conjunction with substantial dewatering. The approach to model development used in this work employs repeated analysis using sensitivity and inverse methods, including use of a new observation-stacked parameter importance graph. The methods are highly parallelizable and require few model runs, which make the repeated analyses and attendant insights possible. The success of a model development design can be measured by insights attained and demonstrated model accuracy relevant to predictions. Example insights were obtained: (1) A long-held belief that, except for a few distinct fractures, the travertine is homogeneous was found to be inadequate, and (2) The dewatering pumping rate is more critical to model accuracy than expected. The latter insight motivated additional data collection and improved pumpage estimates. Validation tests using three other recharge and pumpage conditions suggest good accuracy for the predictions considered. The model was used to evaluate management scenarios and showed that similar dewatering results could be achieved using 20 % less pumped water, but would require installing newly positioned wells and cooperation between mine owners.     

Assessment of groundwater vulnerability in Mettur region, Tamilnadu, India using drastic and GIS techniques

A study was carried in Mettur taluk, Salem district of Tamilnadu, India to develop a DRASTIC vulnerability index in GIS environment owing to groundwater pollution with increasing population, industries, and agricultural activities. Seven DRASTIC layers were created from available data (depth to water table, net recharge, aquifer media, soil media, topography, impact of vadose zone, and hydraulic conductivity) and incorporated into DRASTIC model to create a groundwater vulnerability map by overlaying the hydrogeological parameters. The output map indicates southwestern part of the study area with high pollution potential, northern and northwestern parts as moderate pollution potential and northeastern parts as low and no risk of pollution potential. For validating the vulnerability assessment, a total of 46 groundwater samples were collected from different vulnerability zones of the study area for two different seasons (pre- and post-monsoon) and analyzed for major anions and cations. Higher ionic concentrations were noted in wells located near highly industrialized, urbanized, and agricultural active zones. The water types represent Na–Mg–HCO₃ and Na–Cl–HCO₃ type indicating dominance of anthropogenic-related activities. Nitrate and chloride were demarcated as pollution indicators and correlated with DRASTIC vulnerability map. The results show that southwestern, northwestern, and northern parts of the study area recorded with high and moderate vulnerable zones, record higher nitrate values. In contrast to DRASTIC method predicted, low vulnerable zones show higher chloride concentration may be due to agricultural and urban development.     

Assessment of groundwater potential zone using remote sensing,GIS and multi criteria decision analysis techniques

The sustainable development and management of groundwater resource needs quantitative assessment, based on scientific principle and recent techniques. In the present study, groundwater potential zone is being determined using remote sensing, Geographical Information System (GIS) and Multi-Criteria Decision Analysis (MCDA) techniques using various thematic layers viz. geomorphology, geology, drainage density, slope, rainfall, soil texture, groundwater depth, soil depth, lineament and land use/ land cover. The Analytic Hierarchy Approach (AHP) is used to determine the weights of various themes for identifying the groundwater potential zone based on weights assignment and normalization with respect to the relative contribution of the different themes to groundwater occurrence. Finally, obtained groundwater potential zones were classified into five categories, viz. low, medium, medium-high, high and very high potential zone. The result depicts the groundwater potential zone in the study area and found to be helpful in better development and management planning of groundwater resource.