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.     

Isotopic evidence for holocene climatic changes in Bosten Lake, Southern Xinjiang, China

Based on the δ¹³C and δ¹⁸O fluctuation of lacustrine carbonate, CaCO₃ content and sporo-pollen data, a palaeoclimatic history of Bosten Lake during the Holocene has been outlined, several stages of climatic changes are divided, and the following result es are obtained: (1) Palaeoclimatic changes revealed by carbonate isotope around Bosten Lake are basically identical with that revealed by other geological records in Xinjiang. Environmental changes presented apparent Westlies Style model: during cold period, relative humidity increased, δ¹⁸O, δ¹³C and CaCO₃ appeared low; but in warm periods, the dry regime aggravated. (2) The temperature reflected by δ¹⁸O exist evident features being increase in the late period during the Holocene. Together with the δ¹³C, pollen and CaCO₃ analyses, several cold and warm phases which are of broad regional significance can be identified. The warm peaks occurred at about 11.0 ka B.P., 9.4 ka B.P., 7.5 ka B.P., 5.0 ka B.P., 3.0 ka B.P. and 2.0 ka B.P.; the cold peaks at 11.5 ka B.P., 10.5 ka B.P., 8.8 ka B.P., 5.5 ka B.P., 3.3 ka B.P., 2.2 ka B.P. and 1.5 ka B.P.. (3) Several climatic events with the nature of “abrupt climatic changes” are revealed in the periods of 11.0 ka B.P. −10.5 ka B.P., 9.4 ka B.P. −8.8 ka B.P., 5.5 ka B.P. −5.0 ka B.P. and 2.0 ka B.P. −1.5 ka B.P.. (4) The results show that carbonate isotopic record of lacustrine sediment in arid area is very sensitive to climatic changes, and may be play a very important role in understanding the features and mechanism of palaeoclimatic changes.     

The Fault Block''s Framework in Boli Basin and its Control Over the Deposition

Boli basin, between Yishu fracture belt and Dunmi fracture belt, is the biggest Mesozoic coal basin in the east of Heilongjiang Province. Now it is a fault - fold remnant basin. The basin' s shape is generally consistent with the whole distribution of the cover folds, an arc protruding southwards. The basement of the basin can be divided into three fault blocks or structural units. The formation and evoluation of the basin in Mesozoic was determined by the basement fault blocks' displacement features rusulted from by the movement of the edge faults and the main basement faults.     

Interpolating paleovegetation data with an artificial neural network approach

To drive an atmospheric general circulation model (AGCM), land surface boundary conditions like albedo and morphological roughness, which depend on the vegetation type present, have to be prescribed. For the late Quaternary there are some data available, but they are still sparse. Here an artificial neural network approach to assimilate these paleovegetation data is investigated. In contrast to a biome model the relation between climatological parameters and vegetation type is not based on biological knowledge but estimated from the available vegetation data and the AGCM climatology at the corresponding locations. For a test application, a data set for the modern vegetation reduced to the amount of data available for the Holocene climate optimum (about 6000 years B.P.) is used. From this, the neural network is able to reconstruct the complete global vegetation with a kappa value of 0.56. The most pronounced errors occur in Australia and South America in areas corresponding to large data gaps.     

南极威尔克斯地SGR钻孔处雪冰转化过程中的密度变化特征

本文着重描述了SGR钻孔处冰盖上的积雪在密实化过程中的特征变化,并对该过程进行了分段的和全面的回归分析.结果表明,冰盖密度随深度增大,但增长幅度随深度减小.作者提出密度变化减小度的概念.计算得出的所研究冰芯钻取点的密度变化减小度为-0.15kg/m~3·m~2,粒雪成冰前的密实速率平均值为4.08kg/m~3·a.本文得到的冰盖密度变化“临界点”与以往报道的有所不同.分析这一现象时,作者强调当积雪还在活动层时冰盖温度的影响,并以此解释密度剖面的异常变化以及离差的回升.特别指出,积雪的密度变化具有气候学意义,它在一定程度上能够反映出积雪形成及变化过程中气候变化的某些信息.本文由密度变化确定的钻孔点雪冰转化深度为50米.     

Palynofacies and sea-level changes in the Upper Cretaceous of the Vocontian Basin, southeast France

Upper Cretaceous platform carbonates of the Vocontian Basin (southeastern France) have been investigated in a cross-section from the proximal deposits exposed in the lower Rhône Valley to the distal part of the basin in the Southern Subalpine Ranges north of Nice. The stratigraphic interval studied in detail spans the uppermost Turonian and Coniacian.Palynofacies patterns were used to detect eustatic signals at a third-order scale and are the tool for correlation of proximal and distal platform deposits. The organic constituents observed in the studied samples have been grouped into a continental fraction, including higher plant debris (phytoclasts) and sporomorphs, and a marine fraction with dinoflagellate cysts, acritarchs, prasinophytes, and foraminiferal test linings. The main factors influencing the stratigraphic and spatial distribution of land-derived, allochthonous, and marine, relatively autochthonous, organic particles are the proximity of land, the organic productivity, the degree of biodegradation and the hydrodynamic conditions of the depositional system. Palynofacies parameters used for the sequence stratigraphic interpretation are: (1) the ratio of continental to marine constituents (CONT/MAR ratio); (2) the ratio of opaque to translucent phytoclasts (OP/TR ratio); (3) the phytoclast particle size and shape; and (4) the relative proportion and species diversity of marine plankton. Ternary diagrams illustrating significant proximality changes were used to decipher transgressive-regressive trends within the succession.High amounts of translucent phytoclasts and decreasing values of the CONT/MAR ratio occur during the phase of relative sea-level rise in the upper Turonian. The stratigraphic interval of maximum flooding around the Turonian/Coniacian boundary is marked by the highest abundance and species diversity of dinoflagellate cysts, and by high percentages of opaque, equidimensional particles within the phytoclast group. The OP/TR ratio is still high within the lower Coniacian representing the early highstand deposits, whereas the relative abundance of marine constituents is again decreasing. Sedimentary organic matter of the upper Coniacian is dominated by large, blade-shaped, mainly opaque phytoclasts, which are a characteristic palynofacies signature of late highstand deposits.The present study demonstrates the high potential of palynofacies analysis in high-resolution stratigraphy and correlation of sedimentary series of shallow epeiric seas.     

Impact of land use covers upon karst processes in a typical Fengcong depression system of Nongla, Guangxi, China

The direction and intensity of karst processes can be deeply affected by soil physical and chemical variations which were resulted from land use. Taking Nongla Fengcong depression area, Mashan County, Guangxi as an example, authors discussed the impact of land use on karst processes based on the data of field limestone tablet. The results showed that the corrosional rates at varied soil depth are quite different. Corrosional rate in woodland and orchard is mostly bigger than 20 mg/a, which is much higher than that in tilled land and shrub. Generally, corrosional rate decreased from orchard, woodland, tilled land, fallow land and shrub successively, in which soil organic matter (OM) and soil pH are two major controlling factors: corrosion process is controlled remarkably by soil OM in woodland and orchard. The higher the organic matter content is and the less the pH value is, the higher the corrosional rate is. Owing to lower organic matter content, the corrosional rate is mainly affected by soil CO₂ in tilled land and shrub.     

Plant distribution at the mobile dune scale and its relevance to soil properties and topographic features

A survey was conducted in an 11-year recovery mobile dune (RMD11) and a 20-year recovery mobile dune (RMD20), in Horqin Sandy Land, Northern China, to determine plant distribution at the mobile dune scale and its relevance to soil properties and topographic features. The results showed that (1) vegetation cover and species number increased from dune top to bottom in the restoration process of mobile dune; (2) the average value of soil organic C, total N, pH, relative height of sampling site, very fine sand content and soil water contents (40−60 and 60−80 cm) of RMD11 were less than that of RMD20, respectively, and there were significant differences (P < 0.05) between the two dunes; (3) soil resources were redistributed by shrub restoration and relative height of sampling site on dune. The distribution of sand pioneer plant, Agriophyllum squarrosum, was positively related to the relative height of sampling site and soil water content, while that of other herbaceous plants was positively related to soil nutrients in the restoration process of mobile dune. These results suggest that at mobile dune scale, plant distributions are determined by a combination of soil properties and topographic feature. Much effort should be made to preserve the interdune lowland and to improve the level of soil nutrients on mobile dune.     

Suitability analysis for sand and gravel extraction site location in the context of a sustainable development in the surroundings of Zaragoza (Spain)

Zaragoza city is located in the central Ebro Basin, in the Iberian Peninsula. The fluvial terraces formed by the Ebro River present a valuable resource of sand and gravel deposits. However, taking advantage of these available resources implies conflicts with other land use interests like urban and industrial development as well as agricultural use, which has also traditionally occupied the alluvial terraces. These deposits represent a substantial groundwater resource that should be preserved for future generations. The development of spatial decision support systems (SDSS) has greatly assisted efforts for solving land-use conflicts. These systems combine the benefits of geographic information systems (GIS) and decision support methodologies and are therefore suitable to manage sustainable development of urban areas. In this contribution, an extraction suitability map taking into consideration a variety of environmental criteria is created with the help of a SDSS. The method used is the analytical hierarchy process which is integrated in ArcGIS. Areas most suitable to sand and gravel extraction are located in the high terraces, and in those terraces covered by pediments where the thickness of resource is relatively high. These areas are far from valuable natural areas, outside areas most vulnerable to groundwater contamination, and beneath soils with poor irrigation characteristics.