An evaluation of Guided Regularized Random Forest for classification and regression tasks in remote sensing     

《International Journal of Applied Earth Observation and Geoinformation》2020

New Earth observation missions and technologies are delivering large amounts of data. Processing this data requires developing and evaluating novel dimensionality reduction approaches to identify the most informative features for classification and regression tasks. Here we present an exhaustive evaluation of Guided Regularized Random Forest (GRRF), a feature selection method based on Random Forest. GRRF does not require fixing a priori the number of features to be selected or setting a threshold of the feature importance. Moreover, the use of regularization ensures that features selected by GRRF are non-redundant and representative. Our experiments based on various kinds of remote sensing images, show that GRRF selected features provides similar results to those obtained when using all the available features. However, the comparison between GRRF and standard random forest features shows substantial differences: in classification, the mean overall accuracy increases by almost 6% and, in regression, the decrease in RMSE almost reaches 2%. These results demonstrate the potential of GRRF for remote sensing image classification and regression. Especially in the context of increasingly large geodatabases that challenge the application of traditional methods.  相似文献   

Biostratigraphic and aminostratigraphic constraints on the age of the Middle Pleistocene glacial succession in north Norfolk,UK     

Richard C. Preece  Simon A. Parfitt  G. Russell Coope  Kirsty E. H. Penkman  Philippe Ponel  John E. Whittaker 《第四纪科学杂志》2009,24(6):557-580

Considerable debate surrounds the age of the Middle Pleistocene glacial succession in East Anglia following some recent stratigraphical reinterpretations. Resolution of the stratigraphy here is important since it not only concerns the glacial history of the region but also has a bearing on our understanding of the earliest human occupation of north‐western Europe. The orthodox consensus that all the tills were emplaced during the Anglian (Marine Isotope Stage (MIS) 12) has recently been challenged by a view assigning each major till to a different glacial stage, before, during and after MIS 12. Between Trimingham and Sidestrand on the north Norfolk coast, datable organic sediments occur immediately below and above the glacial succession. The oldest glacial deposit (Happisburgh Till) directly overlies the ‘Sidestrand Unio‐bed’, here defined as the Sidestrand Hall Member of the Cromer Forest‐bed Formation. Dating of these sediments therefore has a bearing on the maximum age of the glacial sequence. This paper reviews the palaeobotany and describes the faunal assemblages recovered from the Sidestrand Unio‐bed, which accumulated in a fluvial environment in a fully temperate climate with regional deciduous woodland. There are indications from the ostracods for weakly brackish conditions. Significant differences are apparent between the Sidestrand assemblages and those from West Runton, the type site of the Cromerian Stage. These differences do not result from contrasting facies or taphonomy but reflect warmer palaeotemperatures at Sidestrand and a much younger age. This conclusion is suggested by the higher proportion of thermophiles at Sidestrand and the occurrence of a water vole with unrooted molars (Arvicola) rather than its ancestor Mimomys savini with rooted molars. Amino acid racemisation data also indicate that Sidestrand is significantly younger than West Runton. These data further highlight the stratigraphical complexity of the ‘Cromerian Complex’ and support the conventional view that the Happisburgh Till was emplaced during the Anglian rather than the recently advanced view that it dates from MIS 16. Moreover, new evidence from the Trimingham lake bed (Sidestrand Cliff Formation) above the youngest glacial outwash sediments (Briton's Lane Formation) indicates that they also accumulated during a Middle Pleistocene interglacial – probably MIS 11. All of this evidence is consistent with a short chronology placing the glacial deposits within MIS 12, rather than invoking multiple episodes of glaciation envisaged in the ‘new glacial stratigraphy’ during MIS 16, 12, 10 and 6. Copyright © 2009 John Wiley & Sons, Ltd.  相似文献   

Infiltration at the Tai rain forest (Ivory Coast): Measurements and modelling     

A. Wierda  A. W. L. Veen  R. W. A. Hutjes 《水文研究》1989,3(4):371-382

Infiltration experiments have been performed at three sites along a well-known catena under virgin tropical rain forest using a portable sprinkling infiltrometer. Experimentally determined infiltration curves are presented. Infiltration curves are also simulated on the basis of the Mein-Larson equation. The parameters for this model have been obtained from the infiltration curves (saturated conductivity) and simple soil moisture determinations (fillable porosity). The agreement between experimentally determined and modelled infiltration is reasonable, provided (a) saturated conductivity as derived from the experimental data is corrected, (b) a storage parameter, also derived from the experimental data, is added to the Mein-Larson model, and (c) the decline in soil porosity with depth is either small or occurs abruptly at shallow depth. Comparison of observed infiltration rates with rainfall intensity shows that Horton Overland Flow has to occur naturally at least on the middle and lower section of the catena. Despite the fact that most parameters can be estimated in principle from basic soil data, it remains advisable to obtain sprinkling infiltrometer field measurements, because of soil variability due to dynamic surface conditions, macroporosity, air entrapment, and irregularity of the wetting front.  相似文献   

Quantitative textural analysis of packings of elongate crystals     

John F. Rudge  Marian B. Holness  Graham C. Smith 《Contributions to Mineralogy and Petrology》2008,156(4):413-429

The spatial distribution of grains in a solidifying igneous rock controls the physical properties of the crystal mush, and in turn is controlled by the rate of crystal growth and accumulation. A predominant non-spherical habit for igneous minerals brings into question the use of spherical particles in reference packings used for quantification of spatial distribution. Furthermore, variations of crystal clustering/ordering with length scale require spatial statistics which take into account the distribution of particles beyond nearest neighbours. Using random close packings of spherocylinders, we demonstrate the importance of aspect ratio for the aggregation index (usually known as R) and show that packings of spherical particles have more structure than packings of rods. The spatial distribution functions demonstrate that the plagioclase grains in the colonnade from the Holyoke basalt are clustered on a length scale of 0.5 mm. Understanding the controls on grain spatial distribution in igneous rocks will depend on the application of these techniques to well-understood environments.  相似文献   

Models of natural and human dynamics in forest landscapes: Cross-site and cross-cultural synthesis   总被引：1，自引：0，他引：1  

Miguel F. Acevedo  J. Baird Callicott  Donald Lyons  Judith Rosales  Magdiel Ablan  Giorgio Tonella  Emilio Vilanova 《Geoforum》2008,39(2):846-866

We synthesize the study of coupled natural and human systems across sites and cultures through a process of simplification and abstraction based on multiple dimensions of human-nature connectedness: satisfaction of basic needs, psycho-cultural connectedness and regulation of use of natural resources. We thus provide both a place-based and general understanding of value-driven anthropogenic environmental change and response. Two questions guide this research: what are the crucial stakeholder values that drive land use decisions and thus land cover change? And how can knowledge of these values be used to make decisions and policies that sustain both the human and natural systems in a place? To explore these questions we build simulation models of four study sites, two in the State of Texas, United States, and two in Venezuela. All include protected areas, though they differ in the specifics of vegetation and land use. In the Texas sites, relatively affluent individuals are legally converting forests to residential, commercial, and industrial uses, while in Venezuela landless settlers are extra-legally converting forests for purposes of subsistence agriculture. Contemporary modeling techniques now facilitate simulations of stakeholder and ecosystem dynamics revealing emergent patterns. Such coupled human and natural systems are currently recognized as a form of biocomplexity. Our modeling framework is flexible enough to allow adaptation to each of the study sites, capturing the essential features of the respective natural and anthropogenic land use changes and stakeholder reactions. The interactions between human stakeholders are simulated using multi-agent models that act on forest landscape models, and receive feedback of the effects of these actions on ecological habitats and hydrological response. The multi-agent models employ a formal logic-based method for the Venezuelan sites and a decision analysis approach using multi-attribute utility functions for the Texas sites, differing more in style and emphasis than in substance. Our natural-systems models are generic and can be tailored according to site-specific conditions. Similar models of tree growth and patch transitions are used for all the study sites and the differing responses to environmental variables are specified for each local species and terrain conditions.  相似文献   

Spatial Variability of Rock Depth in Bangalore Using Geostatistical, Neural Network and Support Vector Machine Models     

T. G. Sitharam  Pijush Samui  P. Anbazhagan 《Geotechnical and Geological Engineering》2008,26(5):503-517

Geospatial technology is increasing in demand for many applications in geosciences. Spatial variability of the bed/hard rock is vital for many applications in geotechnical and earthquake engineering problems such as design of deep foundations, site amplification, ground response studies, liquefaction, microzonation etc. In this paper, reduced level of rock at Bangalore, India is arrived from the 652 boreholes data in the area covering 220 km². In the context of prediction of reduced level of rock in the subsurface of Bangalore and to study the spatial variability of the rock depth, Geostatistical model based on Ordinary Kriging technique, Artificial Neural Network (ANN) and Support Vector Machine (SVM) models have been developed. In Ordinary Kriging, the knowledge of the semi-variogram of the reduced level of rock from 652 points in Bangalore is used to predict the reduced level of rock at any point in the subsurface of the Bangalore, where field measurements are not available. A new type of cross-validation analysis developed proves the robustness of the Ordinary Kriging model. ANN model based on multi layer perceptrons (MLPs) that are trained with Levenberg–Marquardt backpropagation algorithm has been adopted to train the model with 90% of the data available. The SVM is a novel type of learning machine based on statistical learning theory, uses regression technique by introducing loss function has been used to predict the reduced level of rock from a large set of data. In this study, a comparative study of three numerical models to predict reduced level of rock has been presented and discussed.  相似文献   

Generation of random waves in time-dependent extended mild-slope equations using a source function method     

Gunwoo Kim  Changhoon Lee  Kyung-Duck Suh   《Ocean Engineering》2006,33(14-15):2047-2066

We develop techniques of numerical wave generation in the time-dependent extended mild-slope equations of Suh et al. [1997. Time-dependent equations for wave propagation on rapidly varying topography. Coastal Engineering 32, 91–117] and Lee et al. [2003. Extended mild-slope equation for random waves. Coastal Engineering 48, 277–287] for random waves using a source function method. Numerical results for both regular and irregular waves in one and two horizontal dimensions show that the wave heights and the frequency spectra are properly reproduced. The waves that pass through the wave generation region do not cause any numerical disturbances, showing usefulness of the source function method in avoiding re-reflection problems at the offshore boundary.  相似文献   

Dynamic properties of green water event in the overtopping of extreme waves on a fixed dock     

Nobuhito Mori  Daniel T. Cox 《Ocean Engineering》2003,30(16):2021-2052

A statistical model is developed to predict wave overtopping volume and rate of extreme waves on a fixed deck. The probability density function for the volume and rate of overtopping water are formulated based on the truncated Weibull distribution with the assumption of local sinusoidal profile for small amplitude waves. Sensitivity to the wave nonlinearity parameter and deck clearance is discussed. The statistical model is compared to laboratory data of the instantaneous free surface elevation measured in front of a fixed deck, and overtopping volume and overtopping rate measured at the leading edge of the deck. The statistical theory compared well with the measured exceedance probability seaward of the deck. The model prediction of the exceedance probability of deck overtopping gave qualitatively good agreement for large overtopping values.  相似文献   

均匀水底上二维随机波面的二阶谱   总被引：1，自引：1，他引：1       下载免费PDF全文

孙孚  丁平兴  余宙文 《海洋与湖沼》1993,24(1):1-6

本文用摄动方法导出了均匀水底上精确至二阶的二维随机波动解。依据此解并将二阶谱定义为三阶矩函数的Fourier变换,首次给出了精确至二阶的二维随机波面二阶谱的理论表达式并简略地考虑了它的各个组成部分的物理意义。  相似文献   

Geoacoustic Inversion via Genetic Algorithm and Its Application to Manganese Sediment Identification     

Woojae Seong  Cheolsoo Park 《Marine Georesources & Geotechnology》2001,19(1):37-50

An acoustic inversion method using a wide-band signal and two near field receivers is proposed and applied to multiple layered seabed models including a manganese sediment. The inversion problem can be formulated into a probabilistic model comprised of signals, a forward model, and additive noise. The forward model simulates wide-band signals, such as chirp signals, and is chosen to be the source-waveletconvolution plane wave modeling method. The wavelet matching technique, using weighted least-squares fitting, estimates the sediment sound-speed and thickness on which determination of the possible numerical ranges for a priori uniform distribution is based. The genetic algorithm is applied to a global optimization problem to find a maximum a posteriori solution for determined a priori search space. Here the object function is defined by an L 2 norm of the difference between measured and modeled signals. Not only the marginal pdf but also its statistics are calculated by numerical evaluation of integrals using the samples selected during importance sampling process of the genetic algorithm.  相似文献