基于深度学习的天气雷达回波序列外推及效果分析

天气雷达探测资料是进行强对流天气临近预报的主要参考数据。针对传统雷达回波外推方法中存在资料信息利用率不足和外推时效有限的问题,文中利用神经网络进行雷达回波的外推、利用预测神经网络模型进行2 h以内的回波变化预报。回波外推问题的关键是回波时、空序列预测问题,该网络具有解决时间记忆问题的长、短时记忆单元（Long Short-Term Memory,LSTM）和提取空间特征的卷积模块。应用福建、江苏和河南多年的雷达探测资料构造训练和测试数据集。为消除降水的不平衡和提高对强回波的预报准确率,网络采用带权重的损失函数进行训练。对光流法和预测神经网络进行测试集检验以及个例分析,结果表明,在相同外推时效和检验反射率阈值的情况下,预测神经网络的临界成功指数、命中率均高于光流法,虚警率低于光流法。不同类型降水预测神经网络的SSIM值（structural similarity）均高于光流法,且层状云降水的SSIM值比对流云降水的大。因此,预测神经网络对强回波的预报能力高于光流法;在预报时效性上,预测神经网络模型具有一定的优越性;预测神经网络对层状云降水预报的准确率比对流云降水的高。      

Developing synergy regression models with space-borne ALOS PALSAR and Landsat TM sensors for retrieving tropical forest biomass

Forest stand biomass serves as an effective indicator for monitoring REDD (reducing emissions from deforestation and forest degradation). Optical remote sensing data have been widely used to derive forest biophysical parameters inspite of their poor sensitivity towards the forest properties. Microwave remote sensing provides a better alternative owing to its inherent ability to penetrate the forest vegetation. This study aims at developing optimal regression models for retrieving forest above-ground bole biomass (AGBB) utilising optical data from Landsat TM and microwave data from L-band of ALOS PALSAR data over Indian subcontinental tropical deciduous mixed forests located in Munger (Bihar, India). Spatial biomass models were developed. The results using Landsat TM showed poor correlation (R² = 0.295 and RMSE = 35 t/ha) when compared to HH polarized L-band SAR (R² = 0.868 and RMSE = 16.06 t/ha). However, the prediction model performed even better when both the optical and SAR were used simultaneously (R² = 0.892 and RMSE = 14.08 t/ha). The addition of TM metrics has positively contributed in improving PALSAR estimates of forest biomass. Hence, the study recommends the combined use of both optical and SAR sensors for better assessment of stand biomass with significant contribution towards operational forestry.     

Missing basics: a study on sanitation and women’s health in urban slums in Lucknow,India

GeoJournal - Sanitation is a multidimensional concept alluding primarily to provision of services for safe disposal of human excreta, provision of clean potable water as well as maintenance of...     

Topographical features of rainfall-triggered landslides in Mon State,Myanmar, August 2019: spatial distribution heterogeneity and uncommon large relative heights

Continuous 5-day (August 4–9, 2019) torrential rainfall in the monsoon season triggered more than 90 landslides on northwest-southeast extended mountain range of Mon State, Myanmar. In this study, remote sensing images, DEM, and limited fieldworks were used to create the landslide inventory. The topography features of these landslides are analyzed via ArcGIS. The largest one occurred on 9 August 2019 and caused 75 deaths and 27 buildings were damaged. This landslide occurred on gentle topography (slope angle, 23°) with long run-out, in which the angle of reach was relatively low (10°). The volume was 111,878 m³ was mainly composed of weathered granite and red soil and the sliding depth was approximately 7.5 m. Topographic characteristics including the relative slope height, angle of reach, and slope angle of source area of 35 landslides with areas?>?4000 m² were analyzed. The spatial distribution characteristics and topographic features of the 35 landslides below are distinguished: (1) the concentration of most of landslides on southwest-facing slopes showing the heterogeneous spatial distribution of landslide; (2) an uncommon landslide distribution in which more than half of landslide originates from upper slope; (3) the range of the angle of the source area (17°–38°) compatible with the internal friction angle of soils in tropical regions (17°–33°); and (4) the tangent of the angle of reach is generally smaller than 0.5 (angle of reach?<?27°) shows a relative high mobility and the relation between landslide mobility and the slope angle of the landslide source area is similar to the one of earthquake-triggered landslides, even though the triggering mechanism, landslide type, and landslide volume are dramatically different.

     

Delineation of the 85°E ridge and its structure in the Mahanadi Offshore Basin,Eastern Continental Margin of India (ECMI), from seismic reflection imaging

The passive Eastern Continental Margin of India (ECMI) evolved during the break up of India and East Antarctica in the Early Cretaceous. The 85°E ridge is a prominent linear aseismic feature extending from the Afanasy Nikitin Seamounts northward to the Mahanadi basin along the ECMI. Earlier workers have interpreted the ridge to be a prominent hot spot trail. In the absence of conclusive data, the extension of the ridge towards its northern extremity below the thick Bengal Fan sediments was a matter of postulation. In the present study, interpretation of high resolution 2-D reflection data from the Mahanadi Offshore Basin, located in the northern part of the ridge, unequivocally indicates continuation of the ridge across the continent–ocean boundary into the slope and shelf tracts of the ECMI. Its morphology and internal architecture suggest a volcanic plume related origin that can be correlated with the activity of the Kerguelen hot spot in the nascent Indian Ocean. In the continental region, the plume related volcanic activity appears to have obliterated all seismic features typical of continental crust. The deeper oceanic crust, over which the hot spot plume erupted, shows the presence of linear NS aligned basement highs, corresponding with the ridge, underlain by a depressed Moho discontinuity. In the deep oceanic basin, the ridge influences the sediment dispersal pattern from the Early Cretaceous (?)/early part of Late Cretaceous times till the end of Oligocene, which is an important aspect for understanding the hydrocarbon potential of the basin.     

Dynamical problem of micropolar viscoelasticity

The dynamic problem in micropolar viscoelastic medium has been investigated by employing eigen value approach after applying Laplace and Fourier transformations. An example of infinite space with concentrated force at the origin has been presented to illustrate the application of the approach. The integral transforms have been inverted by using a numerical technique to obtain the displacement components, force stresses, couple stress and microrotation in the physical domain. The results for these quantities are given and illustrated graphically.     

Time-Variations of D-Region Electron Densities, and Comparisons With Model Computations

An extensive series of incoherent scatter studies of the ionospheric D-region was carried out at the Arecibo radar facility during 1978 and 1979. They included several full-day sequences of electron density measurements over a range of altitudes, and also included a sequence during the serendipitous occurrence of a large solar flare. For the solar flare event simultaneous data on solar X-ray fluxes in several wavelength bands were available from the GOES-2 and ISEE-3 satellites. In the course of development of a large ionospheric computer model at Los Alamos we have used the solar flare data as a reality check. The solar X-ray flux data were used as inputs for computing ionization rates. The model computer includes 999 chemical reactions, and also includes diffusion and transport processes. In the course of the flare studies we used the data comparisons to adjust the values of three chemical rate coefficients that were poorly known. With those adjustments the model computations fitted the data quite well. Subsequent to the flare analysis we have been using the same model with some minor updates to compute the expected diurnal variations of the ambient D-region under conditions chosen to match those existing at the times of the incoherent scatter measurements. Comparisons of the computations and the data will be shown, and the relative importance of the several separate ionization processes will be discussed. We also compare model results with experimental data on concentrations of NO.     

Variation of δ18O, δD and 3H in karst springs of south Kashmir,western Himalayas (India)

Water samples were collected from cold and warm karst springs for stable isotopes (δ¹⁸O and δD) and ³H from SE of Kashmir valley (western Himalayas) to distinguish the sources of recharge and infer their recharge areas. The spring water samples were most depleted in heavier isotopes in May (average δ¹⁸O: ?8.87‰ and δD: ?50.3‰) and enriched in September (average δ¹⁸O: ?7.58‰ and δD: ?48.1‰). The depleted ¹⁸O and ²H of spring waters bear the signatures of winter precipitation while as the enriched ¹⁸O and ²H of spring waters bear the signature of summer rainfall. D‐excess and ³H corroborate with the stable isotope results that the spring flow in spring season (May) and autumn (September) is dominantly controlled by the melting of winter snowmelt and summer rainfall, respectively. The results showed that unlike δD, the δ¹⁸O value in the karst spring waters decreases in January suggesting δ¹⁸O shift. The spring water samples also fall above the Local Meteoric Water Line and Global Meteoric Water Line indicating the δ¹⁸O shift due to interaction of groundwater with the host carbonate rocks during its traverse. The mean elevation of the recharge areas of the springs using δ¹⁸O and δD tracers was also estimated. Copyright © 2014 John Wiley & Sons, Ltd.