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

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.     

Multiple longshore sand bars in the upper Chesapeake Bay

In the upper Chesapeake Bay (Maryland, U.S.A.) field surveys were conducted at 18 multiple longshore sand bar sites. The multiple bar systems were found in water depths less than approximately 2 m (mean sea level), and exhibited mild bottom slopes of 0·0052 or less. The number of bars composing each system ranged from four to 17 and the spacing between the crests typically increased in the offshore direction, ranging from 12 to 70 m. Bar height also typically increased with distance offshore and ranged from 0·03 to 0·61 m. A grain size analysis of crest and trough sediment did not reveal any significant differences and the sediment was categorized as ‘fine sand’. A review of the literature data indicated that the Chesapeake Bay multiple bars possessed similar characteristics to those found in Gelding Bay (Baltic Sea); similarities in fetch, wave height and tidal range between the two bays may account for this finding. The surf-scaling parameter indicated that the multiple bar systems were extremely dissipative with regard to wave energy, and wave height appeared to be an important factor in controlling bar spacing and bar height. A multiple wave break point hypothesis was discussed as a possible mechanism for the formation of Chesapeake Bay multiple longshore bars, and limited observational evidence appeared to support such a mechanism.     

Periodical light variations of t tauri stars as a result of disk accretion

I argue that temperatures of spots, responsible for observed periodical light variations of T Tauri stars (TTS), are not known with reliable accuracy to discriminate between chromospheric and accretion theories of TTS 's phenomenon. The hypothesis is set up that spots on classical TTS (CTTS) are due to heating of stellar surface by radiation from a collisional accretion shock, whereas spots on weak line TTS (WTTS), at least in some cases, are connected with a collisionless accretion shock rather than chromospheric activity. Possible scenarios of WTTS interaction with circumstellar matter are discussed.     

An HLLC Riemann solver for relativistic flows – II. Magnetohydrodynamics

An approximate Riemann solver for the equations of relativistic magnetohydrodynamics (RMHD) is derived. The Harten–Lax–van Leer contact wave (HLLC) solver, originally developed by Toro, Spruce and Spears, generalizes the algorithm described in a previous paper to the case where magnetic fields are present. The solution to the Riemann problem is approximated by two constant states bounded by two fast shocks and separated by a tangential wave. The scheme is Jacobian-free, in the sense that it avoids the expensive characteristic decomposition of the RMHD equations and it improves over the HLL scheme by restoring the missing contact wave.
Multidimensional integration proceeds via the single step, corner transport upwind (CTU) method of Colella, combined with the constrained transport (CT) algorithm to preserve divergence-free magnetic fields. The resulting numerical scheme is simple to implement, efficient and suitable for a general equation of state. The robustness of the new algorithm is validated against one- and two-dimensional numerical test problems.