神经网络敏感性分析及其在遥感影像分类中的应用

针对基于连接权的神经网络敏感性分析方法中求取敏感性系数的不稳定性,提出一种优化连接权的神经网络敏感性分析方法。首先采用遗传算法根据误差最小化原则对神经网络进行优化,在优化的神经网络模型上进行基于连接权的敏感性分析。以1个数值模拟实例和华盛顿广场地区的遥感图像地物分类为例,验证所提方法的有效性。实验结果表明,所提方法求取输入变量的敏感性系数是稳定有效的,能有效筛选出遥感图像中对分类贡献较大的特征波段,达到降维的同时提高分类精度。     

Groundwater potential index in a crystalline terrain using remote sensing data

Demand for groundwater for drinking, agricultural and industrial purposes has increased due to uncertainty in the surface water supply. Agriculture is the main occupation of the rural people in Guntur district, Andhra Pradesh, India. Development of groundwater in the district is very less, indicating a lot of scope for further development of groundwater resources. However, assessment of groundwater conditions, particularly in a crystalline terrain, is a complex task because of variations in weathering and fracturing zones from place to place. Systematic studies for evaluation of groundwater potential zones have been carried out in a crystalline terrain of the district. Information on soils, geological formations and groundwater conditions is collected during the hydrogeological survey. Topographical and drainage conditions are derived from the Survey of India topographical maps. Geomorphological units and associated landform features inferred and delineated from the Indian remote sensing satellite imagery (IRS ID LISS III FCC) are moderately buried pediplain (BPM), shallow buried pediplain (BPS), valley fills (VF), structural hill (SH), residual hills (RH), lineaments and land use/land cover. A groundwater potential index (GPI) is computed for relative evaluation of groundwater potential zones in the study area by integrating all the related factors of occurrence and movement of groundwater resources. Accordingly, the landforms, BPM, BPS, VF, SH and RH, of the area are categorized as very good groundwater potential zone, good to moderate groundwater potential zone, moderate to poor groundwater potential zone, poor to very poor groundwater potential zone and very poor groundwater potential zone, respectively, for development and utilization of both groundwater and surface water resources for eliminating water scarcity. This study could help to improve the agrarian economy for better living conditions of the rural people. Taking the total weight-score of the GPI into account, a generalized classification of groundwater potential zones is evaluated for a quick assessment of the occurrence of groundwater resources on regional scale.     

Database extension for digital soil mapping using artificial neural networks

Cost and time are the two most important factors conditioning soil surveys. Since these surveys provide basic information for modelling and management activities, new methods are needed to speed the soil-mapping process with limited input data. In this study, the polypedon concept was used to extend the spatial representation of sampled pedons (point data) in order to train artificial neural networks (ANNs) for digital soil mapping (DSM). The input database contained 97 soil profiles belonging to 7 different soil series and 15 digital elevation model (DEM) attributes. Pedons were represented in raster format as one-cell areas. The corresponding polypedons were then spatially represented by neighbouring raster cells (e.g. 2 × 2, … up to 6 × 6 cells). The primary database contained 97 pedons (97 cells) that were extended up to 3492 cells (in the case of 6 × 6-cell regions). This approach employed test and validation areas to calculate the respective accuracies of data interpolation and extrapolation. The results showed increased accuracies in training and interpolation (test area) but a poor level of accuracy in the extrapolation process (validation area). However, the overall precision of all predictions increased considerably. Using only topographic attributes for extrapolation was not sufficient to obtain an accurate soil map. To improve prediction, other soil-forming factors, such as landforms and/or geology, should also be considered as input data in the ANN. The proposed method could help to improve existing soil maps by using DSM results in areas with limited soil data and to save time and money in soil survey work.     

Validation of an artificial neural network model for landslide susceptibility mapping

The aim of this study was to validate an artificial neural network model at Youngin, Janghung, and Boeun, Korea, using the geographic information system (GIS). The factors that influence landslide occurrence, such as the slope, aspect, curvature, and geomorphology of topography, the type, material, drainage, and effective thickness of soil, the type, diameter, age, and density of forest, distance from lineament, and land cover were either calculated or extracted from the spatial database and Landsat TM satellite images. Landslide susceptibility was analyzed using the landslide occurrence factors provided by the artificial neural network model. The landslide susceptibility analysis results were validated and cross-validated using the landslide locations as study areas. For this purpose, weights for each study area were calculated by the artificial neural network model. Among the nine cases, the best accuracy (81.36%) was obtained in the case of the Boeun-based Janghung weight, whereas the Janghung-based Youngin weight showed the worst accuracy (71.72%).     

Regional groundwater productivity potential mapping using a geographic information system (GIS) based artificial neural network model

An artificial neural network model (ANN) and a geographic information system (GIS) are applied to the mapping of regional groundwater productivity potential (GPP) for the area around Pohang City, Republic of Korea. The model is based on the relationship between groundwater productivity data, including specific capacity (SPC) and its related hydrogeological factors. The related factors, including topography, lineaments, geology, and forest and soil data, are collected and input into a spatial database. In addition, SPC data are collected from 44 well locations. The SPC data are randomly divided into a training set, to analyse the GPP using the ANN, and a test set, to validate the predicted potential map. Each factor??s relative importance and weight are determined by the back-propagation training algorithms and applied to the input factor. The GPP value is then calculated using the weights, and GPP maps are created. The map is validated using area under the curve analysis with the SPC data that have not been used for training the model. The validation shows prediction accuracies between 73.54 and 80.09?%. Such information and the maps generated from it could serve as a scientific basis for groundwater management and exploration.     

An assessment of soil erosion probability and erosion rate in a tropical mountainous watershed using remote sensing and GIS

Remote sensing data and Geographical Information System (GIS) has been integrated with the weighted index overlay (WIO) method and E ₃₀ model for the identification and delineation of soil erosion susceptibility zones and the assessment of rate of soil erosion in the mountainous sub-watershed of River Manimala in Kerala (India). Soil erosion is identified as the one of the most serious environmental problems in the human altered mountainous environment. The reliability of estimated soil erosion susceptibility and soil loss is based on how accurately the different factors were estimated or prepared. In the present analysis, factors that are considered to be influence the soil erosion are: land use/land cover, NDVI, landform, drainage density, drainage frequency, lineament frequency, slope, and relative relief. By the WIO analysis, the area is divided into zones representing low (33.30%), moderate (33.70%), and high (33%) erosion proneness. The annual soil erosion rate of the area under investigation was calculated by carefully determining its various parameters and erosion for each of the pixels were estimated individually. The spatial pattern thus created for the area indicates that the average annual rate of soil erosion in the area was ranging from 0.04 mm yr⁻¹ to 61.80 mm yr⁻¹. The high soil erosion probability and maximum erosion rate was observed in areas with high terrain alteration, high relief and slopes with the intensity and duration of heavy precipitation during the monsoons.     

Glacial lakes and glacial lake outburst flood in a Himalayan basin using remote sensing and GIS

Glacial hazards relate to hazards associated with glaciers and glacial lakes in high mountain areas and their impacts downstream. The climatic change/variability in recent decades has made considerable impacts on the glacier life cycle in the Himalayan region. As a result, many big glaciers melted, forming a large number of glacial lakes. Due to an increase in the rate at which ice and snow melted, the accumulation of water in these lakes started increasing. Sudden discharge of large volumes of water with debris from these lakes potentially causes glacial lake outburst floods (GLOFs) in valleys downstream. Outbursts from glacier lakes have repeatedly caused the loss of human lives as well as severe damage to local infrastructure. Monitoring of the glacial lakes and extent of GLOF impact along the downstream can be made quickly and precisely using remote sensing technique. A number of hydroelectric projects in India are being planned in the Himalayan regions. It has become necessary for the project planners and designers to account for the GLOF also along with the design flood for deciding the spillway capacity of projects. The present study deals with the estimation of GLOF for a river basin located in the Garwhal Himalaya, India. IRS LISSIII data of the years 2004, 2006 and 2008 have been used for glacial lake mapping, and a total of 91 lakes have been found in the year 2008, and out of these, 45 lakes are having area more than 0.01?km². All the lakes have been investigated for vulnerability for potential bursting, and it was found that no lake is vulnerable from GLOF point of view. The area of biggest lake is 0.193, 0.199 and 0.203?km² in the years 2004, 2006 and 2008, respectively. Although no lake is potentially hazardous, GLOF study has been carried out for the biggest lake using MIKE 11 software. A flood of 100-year return period has been considered in addition to GLOF. The flood peak at catchment outlet comes out to be 993.74, 1,184.0 and 1,295.58 cumec due to GLOF; 3,274.74, 3,465.0 and 3,576.58 cumec due to GLOF; and 100-year return flood together considering breach width of 40, 60 and 80?m, respectively.     

Estimation of soil compaction parameters by using statistical analyses and artificial neural networks

This study presents the application of different methods (simple–multiple analysis and artificial neural networks) for the estimation of the compaction parameters (maximum dry unit weight and optimum moisture content) from classification properties of the soils. Compaction parameters can only be defined experimentally by Proctor tests. The data collected from the dams in some areas of Nigde (Turkey) were used for the estimation of soil compaction parameters. Regression analysis and artificial neural network estimation indicated strong correlations (r ² = 0.70–0.95) between the compaction parameters and soil classification properties. It has been shown that the correlation equations obtained as a result of regression analyses are in satisfactory agreement with the test results. It is recommended that the proposed correlations will be useful for a preliminary design of a project where there is a financial limitation and limited time.     

Predicting moisture content of soil from thermal properties using artificial neural network

Monitoring of soil moisture contents is an important practice for irrigation water management. The benefit of periodic soil water content data includes improved irrigation scheduling in order to optimize water usage for improved crop productivity. However, the in situ equipment for measuring soil water contents have high maintenance and operation cost and are highly affected by neighboring soil conditions, and some have overwhelming calibration and data interpretation, whereas the common standard laboratory procedure requires much effort and can be time-consuming for large dataset. The objective of this study is to evaluate the applicability of artificial neural network (ANN) to predict moisture content of soil using available or measured thermal properties (thermal conductivity, thermal diffusivity, specific heat, and temperature) of soil. We used both multilayered perception (MLP) and radial basis function (RBF) types of ANN. The study area is a farmland situated within the premises of the University of Ibadan campus. Thermal properties were measured with KD2 Pro at 42 points along seven transects. Soil samples were also collected at these points to determine their moisture contents in the laboratory. ANN analysis carried out effectively predicted the soil moisture content with very low root-mean-square error (RMSE) and high correlation coefficient (R) of approximately 0.9 for the two methods evaluated. The overall results suggest that ANN can be incorporated to predict the moisture content of soil in this area where thermal properties are known.     

Storm surge prediction using an artificial neural network model and cluster analysis

In this study, an artificial neural network model was developed to predict storm surges in all Korean coastal regions, with a particular focus on regional extension. The cluster neural network model (CL-NN) assessed each cluster using a cluster analysis methodology. Agglomerative clustering was used to determine the optimal clustering of 21 stations, based on a centroid-linkage method of hierarchical clustering. Finally, CL-NN was used to predict storm surges in cluster regions. In order to validate model results, sea levels predicted by the CL-NN model were compared with results using conventional harmonic analysis and the artificial neural network model in each region (NN). The values predicted by the NN and CL-NN models were closer to observed data than values predicted using harmonic analysis. Data such as root mean square error and correlation coefficient varied only slightly between CL-NN and NN model results. These findings demonstrate that cluster analysis and the CL-NN model can be used to predict regional storm surges and may be used to develop a forecast system.     

Determination and application of the weights for landslide susceptibility mapping using an artificial neural network

The purpose of this study is the development, application, and assessment of probability and artificial neural network methods for assessing landslide susceptibility in a chosen study area. As the basic analysis tool, a Geographic Information System (GIS) was used for spatial data management and manipulation. Landslide locations and landslide-related factors such as slope, curvature, soil texture, soil drainage, effective thickness, wood type, and wood diameter were used for analyzing landslide susceptibility. A probability method was used for calculating the rating of the relative importance of each factor class to landslide occurrence. For calculating the weight of the relative importance of each factor to landslide occurrence, an artificial neural network method was developed. Using these methods, the landslide susceptibility index (LSI) was calculated using the rating and weight, and a landslide susceptibility map was produced using the index. The results of the landslide susceptibility analysis, with and without weights, were confirmed from comparison with the landslide location data. The comparison result with weighting was better than the results without weighting. The calculated weight and rating can be used to landslide susceptibility mapping.     

Application of probabilistic-based frequency ratio model in groundwater potential mapping using remote sensing data and GIS

The main goal of this study is to investigate the application of the probabilistic-based frequency ratio (FR) model in groundwater potential mapping at Langat basin in Malaysia using geographical information system. So far, the approach of probabilistic frequency ratio model has not yet been used to delineate groundwater potential in Malaysia. Moreover, this study includes the analysis of the spatial relationships between groundwater yield and various hydrological conditioning factors such as elevation, slope, curvature, river, lineament, geology, soil, and land use for this region. Eight groundwater-related factors were collected and extracted from topographic data, geological data, satellite imagery, and published maps. About 68 groundwater data with high potential yield values of ≥11 m³/h were randomly selected using statistical software of SPSS. Then, the groundwater data were randomly split into a training dataset 70 % (48 borehole data) for training the model and the remaining 30 % (20 borehole data) was used for validation purpose. Finally, the frequency ratio coefficients of the hydrological factors were used to generate the groundwater potential map. The validation dataset which was not used during the FR modeling process was used to validate the groundwater potential map using the prediction rate method. The validation results showed that the area under the curve for frequency model is 84.78 %. As far as the performance of the FR approach is concerned, the results appeared to be quite satisfactory, i.e., the zones determined on the map being zones of relative groundwater potential. This information could be used by government agencies as well as private sectors as a guide for groundwater exploration and assessment in Malaysia.     

Urban development and flooding in Houston Texas, inferences from remote sensing data using neural network technique

The effects of urban development on the natural ecosystem and its link to the increased flooding in Houston, Texas were evaluated. Houston is suitable for this type of analysis due to its 1.95 million population, large geographic area and fast growth rate. Using neural network techniques, four Landsat Thematic Mapper images were grouped into five land use classes for the period 1984 to 2003: vegetation, bare ground, water, concrete and asphalt. Results show that asphalt and concrete increased 21% in the time period 1984–1994, 39% in 1994–2000 and 114%, from 2000 to 2003, while vegetation suffered an overall decrease. When change detection data are compared with runoff ratio data, a relationship between increased runoff and urban development is apparent, which indicates increased chances of flooding. Initial results of this work are made available to the public in GIS format via internet using Arc Internet Map Server (ArcIMS) at .     

Evaluation of effective management plan for an agricultural watershed using AVSWAT model,remote sensing and GIS

In the present investigation, an effort has been made to identify the critical sub-watersheds for the development of best management plan for a small watershed of Eastern India using a hydrological model, namely, AVSWAT2000. A total of 180 combinations of various management treatments including crops (rice, maize ground nut and soybean), tillage (zero, conservation, field cultivator, mould board plough and conventional practices) and fertilizer levels (existing half of recommended and recommended) have been evaluated. The investigation reveled that rice cannot be replaced by other crops such as groundnut, maize, mungbean, sorghum and soybean since comparatively these crops resulted in higher sediment yield. The tillage practices with disk plough have been found to have more impact on sediment yield and nutrient losses than conventional tillage practices for the existing level of fertilizer. Sediment yield decreased in the case of zero tillage, conservation tillage, field cultivator, moldboard plough, and conservation tillage as compare to conventional tillage. Lowest NO₃–N loss was observed in zero tillage in all the fertilizer treatments, whereas field cultivator, moldboard plough and disk plough resulted in increase of NO³–N loss. As compared to conventional tillage, the losses of soluble phosphorus were increased in moldboard plough. The losses of organic nitrogen were also increased as fertilizer dose increased. After zero tillage the conservation tillage preformed better in all the fertilizer treatments as per loss of organic nitrogen and organic phosphorus is concerned. It can be concluded that the sediment yield was found to be the highest in the case of disk plough followed by moldboard plough, field cultivator, conventional tillage, field cultivator and least in zero tillage practices. The nutrient losses were found to be in different order with tillage practices, resulted highest in disk plough tillage practices. In view of sediment yield and nutrient losses, the conservation tillage practice was found to be the best as the sediment yield is less than the average soil loss whereas nutrient loss is within the permissible limit.     

Prioritization of sub-watersheds for conservation measures in a semi arid watershed using remote sensing and GIS

Land and water resources development plans are generally adopted at watershed level. Delineation of watersheds and their prioritization within large river basins requires host of terrain parameters to be studied and analysed. Chopan watershed in Central India has been studied for sub-watershed delineation and prioritization based on drainage morphometry, land use/land cover and sediment yield index analysis using remote sensing and GIS techniques. The watershed was demarcated into five sub-watersheds on the basis of drainage flow directions, contour value, slope, elevation. Geocoded satellite data of 1989 and 2001 on 1:50 000 scale were visually interpreted to prepare land use/land cover and drainage maps which were later digitized using Arcview/ArcGIS. Linear and shape aspects of the sub-watersheds were computed and used for prioritization. The results show widespread variation in drainage characteristics, land cover changes and sediment yield rates across sub-watersheds. On the basis of morphometric, land use/land cover change and sediment yield index, sub-watersheds were grouped into low, medium and high priority. A correlation of results show that SW1 and SW5 are common sub-watersheds falling under high and low priority based on morphometric, land use change analysis and SYI. The priority list of sub-watersheds will be crucial for decision making and implementation of land and water resource conservation projects.     

Determination of importance for comprehensive topographic factors on landslide hazard mapping using artificial neural network

A landslide is one of the natural disasters that occur in Malaysia. In addition to the geological factor and the rain as triggering factor, topographic factors such as elevation, slope angle, slope aspect, and curvature are considered as the main causes of landslides. The study in this paper was conducted in three stages. The first stage involved the extraction of extra topographic factors. Previous landslide studies had identified only four of the topographic factors. However, eight new additional factors have also been identified in this study. They are general curvature, longitudinal curvature, tangential curvature, cross-section curvature, surface area, diagonal line length, surface roughness, and rugosity. At this stage, 13 factors were extracted from the digital elevation model. The second stage involved specifying the importance of each factor. The multilayer perceptron network and backpropagation algorithm were used to specify the weight of each factor. Results were verified using the receiver operating characteristics based on the area under the curve method in the third stage. The results indicated 76.07 % accuracy in predicting of landslides, with slope angle as the most important factor while the tangential curvature has the least importance.     

Soil erosion modeling of a Himalayan watershed using RS and GIS

Employing the remote sensing (RS) and geographical information system (GIS), an assessment of sediment yield from Dikrong river basin of Arunachal Pradesh (India) has been presented in this paper. For prediction of soil erosion, the Morgan-Morgan and Finney (MMF) model and the universal soil loss equation (USLE) have been utilized at a spatial grid scale of 100 m × 100 m, an operational unit. The average annual soil loss from the Dikrong river basin is estimated as 75.66 and 57.06 t ha⁻¹ year⁻¹ using MMF and USLE models, respectively. The watershed area falling under the identified very high, severe, and very severe zones of soil erosion need immediate attention for soil conservation.     

Correction of soil parameters in calculation of embankment settlement using a BP network back-analysis model

Finite element method (FEM) have been widely used for the calculation of settlement of embankment on soft soils in the last decade. However, due to the complexity of construction, spatial inhomogeneity of soils, as well as sensitivity of numerical results to the variation of soil parameters, large discrepancy typically exists between numerical outputs and field observations. This paper presents a novel method, combining FEM and an improved back-propagation (BP) neural network, for correction of soil parameters in numerical prediction of embankment settlement. Duncan–Chang hyperbolic soil model is adopted with the sensitivity of eight constitutive parameters numerically investigated. The soil parameters with large sensitivity are identified, and together with the representative settlements, are used for the training of the improved BP neural network which, once established, generates correction factors of soil parameters for subsequent more accurate FEM forward predictions. It is demonstrated that the proposed numerical back-analysis framework is very efficient in practical engineering applications to calculate highway settlement.