Fine-scale activity patterns of large-and medium-sized mammals in a deciduous broadleaf forest in the Qinling Mountains,China

The composition of animal species and interactions among them are widely known to shape ecological communities and fine-scale(e.g.,<1 km)monitoring of animal communities is essential for understanding the relationships among animals and plants.Although the co-existence of large-and medium-sized species has been studied across different scales,research on fine-scale interactions of herbivores in deciduous broadleaf forests is limited.Camera trapping of large-and medium-sized mammals was carried out over a 1 year period within a 25 ha deciduous broadleaf forest dynamics plot in the Qinling Mountains,China.Fourteen species of large-and medium-sized mammals,including six carnivores,six ungulates,one primate and one rodent species were found.Kernel density estimations were used to analyse the diel or 24 h activity patterns of all species with more than 40 independent detections and general linear models were developed to explore the spatial relationships among the species.The combination of overlapping diel activity patterns and spatial associations showed obvious niche separation among six species:giant panda(Ailuropoda melanoleuca David),takin(Budorcas taxicolor Hodgson),Reeves’s muntjac(Muntiacus reevesi Ogilby),tufted deer(Elaphodus cephalophus Milne-Edwards),Chinese serow(Capricornis milneedwardsii David)and wild boar(Sus scrofa Linnaeus).Long-term fine-scale monitoring is useful for providing information about the co-existence of species and their interactions.The results demonstrate the importance for fine-scale monitoring of animals and plants for improving understanding of species interactions and community dynamics.

EXPERIMENTAL STUDY ON UPGRADING OF A FINE-SCALED GRAPHITE ORE BY FLOTATION

Based on the quality and analysis of the medium carbon product whose grade was 81%-82%, the emphasis of the study was placed on the orthogonal experiments of roughing flotation reagent and single factors, such as abrasive grain, pulp concentration and the rotating speed of flotation machine. The result with improvement in the grade of concentrate to 95% through the routine floatation method was achieved.

Application of FLUENT on fine-scale simulation of wind field over complex terrain

The state-of-art Computational Fluid Dynamics （CFD） codes FLUENT is applied in a fine-scale simulation of the wind field over a complex terrain. Several numerical tests are performed to validate the capability of FLUENT on describing the wind field details over a complex terrain. The results of the numerical tests show that FLUENT can simulate the wind field over extremely complex terrain, which cannot be simulated by mesoscale models. The reason why FLUENT can cope with extremely complex terrain, which can not be coped with by mesoscale models, relies on some particular techniques adopted by FLUENT, such as computer-aided design （CAD） technique, unstructured grid technique and finite volume method. Compared with mesoscale models, FLUENT can describe terrain in much more accurate details and can provide wind simulation results with higher resolution and more accuracy.

Activity-density and spatial distribution of termites on a fine-scale in a tropical rainforest in Xishuangbanna,southwest China

The community composition and activity-density of termites can influence nutrient cycling and other ecological functions.However,the spatial distribution and the activity-density of termites on a fine-scale in tropical forests are still unknown.We checked the spatial distribution patterns of the feeding groups and species of termites and their co-occurrence pattern in a 1-ha(100 m×100 m)plot,and their correlatiion with the environmental factors.We used a standard protocol to collect termite assemblages and classified them into five feeding groups based on their preferrred diet:fungus growers,litter feeders,soil feeders,soil-wood feeders,and wood feeders.We measured the environmental factors:soil pH,litter mass,aboveground plant biomass,and topographic position index(TPI).Soil-wood feeders showed the highest activity-density,followed by wood feeders,fungus growers,soil feeders,and litter feeders.Soil-wood feeders and fungus growers demonstated a strong correlation while litter feeders showed weak correlations with other feeding groups.Termite feeding groups and most of the termite species displayed a positive association with the high TPI and the low soil pH patches.Our results indicated that the examined environmental factors influenced the termite community assemblages and distribution patterns on a fine-scale in tropical rainforests.

Sampling strategy for wild soybean(Glycine soja)populations based on their genetic diversity and fine-scale spatial genetic structure

A total of 892 individuals sampled from a wild soybean population in a natural reserve near the Yellow River estuary located in Kenli of Shandong Province(China)were investigated.Seventeen SSR(simple sequence repeat)primer pairs from cultivated soybeans were used to estimate the genetic diversity of the population and its variation pattern versus changes of the sample size(sub-samples),in addition to investigating the fine-scale spatial genetic structure within the population.The results showed relatively high genetic diversity of the population with the mean value of allele number(A)being 2.88,expected heterozygosity(He)0.431,Shannon diversity index(I)0.699,and percentage of poly-morphic loci(P)100%.Sub-samples of different sizes(ten groups)were randomly drawn from the population and their genetic diversity was calculated by computer simulation.The regression model of the four diversity indexes with the change of sample sizes was computed.As a result,27-52 individuals can reach 95%of total genetic variability of the population.Spatial autocorrelation analysis revealed that the genetic patch size of this wild soybean population is about 18 m.The study provided a scientific basis for the sampling strategy of wild soybean populations.

Adaptive integration of local region information to detect fine-scale brain activity patterns

With the rapid development of functional magnetic resonance imaging (fMRI) technology, the spatial resolution of fMRI data is continuously growing. This pro- vides us the possibility to detect the fine-scale patterns of brain activities. The es- tablished univariate and multivariate methods to analyze fMRI data mostly focus on detecting the activation blobs without considering the distributed fine-scale pat- terns within the blobs. To improve the sensitivity of the activation detection, in this paper, multivariate statistical method and univariate statistical method are com- bined to discover the fine-grained activity patterns. For one voxel in the brain, a local homogenous region is constructed. Then, time courses from the local ho- mogenous region are integrated with multivariate statistical method. Univariate statistical method is finally used to construct the interests of statistic for that voxel. The approach has explicitly taken into account the structures of both activity pat- terns and existing noise of local brain regions. Therefore, it could highlight the fine-scale activity patterns of the local regions. Experiments with simulated and real fMRI data demonstrate that the proposed method dramatically increases the sensitivity of detection of fine-scale brain activity patterns which contain the subtle information about experimental conditions.

GPS-based fine-scale mapping surveys for schistosomiasis assessment: a practical introduction and documentation of field implementation

Background Fine-scale mapping of schistosomiasis to guide micro-targeting of interventions will gain importance in elimination settings,where the heterogeneity of transmission is often pronounced.Novel mobile applications offer new opportunities for disease mapping.We provide a practical introduction and documentation of the strengths and shortcomings of GPS-based household identification and participant recruitment using tablet-based applications for fine-scale schistosomiasis mapping at sub-district level in a remote area in Pemba,Tanzania.Methods A community-based household survey for urogenital schistosomiasis assessment was conducted from November 2020 until February 2021 in 20 small administrative areas in Pemba.For the survey,1400 housing structures were prospectively and randomly selected from shapefile data.To identify pre-selected structures and collect survey-related data,field enumerators searched for the houses’geolocation using the mobile applications Open Data Kit(ODK)and MAPS.ME.The number of inhabited and uninhabited structures,the median distance between the pre-selected and recorded locations,and the dropout rates due to non-participation or non-submission of urine samples of sufficient volume for schistosomiasis testing was assessed.Results Among the 1400 randomly selected housing structures,1396(99.7%)were identified by the enumerators.The median distance between the pre-selected and recorded structures was 5.4 m.A total of 1098(78.7%)were residential houses.Among them,99(9.0%)were dropped due to continuous absence of residents and 40(3.6%)households refused to participate.In 797(83.1%)among the 959 participating households,all eligible household members or all but one provided a urine sample of sufficient volume.Conclusions The fine-scale mapping approach using a combination of ODK and an offline navigation application installed on tablet computers allows a very precise identification of housing structures.Dropouts due to non-residential housing structures,absence,non-participation and lack of urine need to be considered in survey designs.Our findings can guide the planning and implementation of future household-based mapping or longitudinal surveys and thus support micro-targeting and follow-up of interventions for schistosomiasis control and elimination in remote areas.

Observation of a Straight-Line Wind Case Caused by a Gust Front and Its Associated Fine-Scale Structures

A straight-line wind case was observed in Tianjin on 13 June 2005,which was caused by a gust front from a squall line.Mesoscale analyses based on observations from in-situ surface stations,sounding,and in-situ radar as well as fine-scale analyses based on observation tower data were performed.The mesoscale characteristics of the gust front determined its shape and fine-scale internal structures.Based on the scale and wavelet analyses,the fine-scale structures within the gust front were distinguished from the classical mesoscale structures,and such fine-scale structures were associated with the distribution of straight-line wind zones.A series of cross-frontal fine-scale circulations at the lowest levels of the gust front was discovered,which caused a relatively weak wind zone within the frontal strong wind zone.The downdraft at the rear of the head region of the gust front was more intense than in the classical model,and similar to the microburst,a series of vertical vortices propagated from the rear region to the frontal region.In addition,strong tangential fine-scale instability was detected in the frontal region.Finally,a fine-scale gust front model with straight-line wind zones is presented.

基于多尺度特征深度神经网络的不同产地山楂细粒度图像识别

中药是中医治疗疾病的主要途径,也是我国中医药事业传承与创新发展的物质基础,其真伪优劣也会直接影响中医临床的疗效,因此研究科学合理且高效的中药材质量检测方法符合当前行业热点.山楂作为中国著名的药食两用类药材,在烹饪和治疗中具有保护心血管、降低血压的作用,被广泛应用;但由于自然环境与栽培条件的不同,不同产地的山楂易被混淆从而对品质产生影响.尽管化学、生物鉴定的方法广泛而重要,但专业门槛高,耗时较长;且传统图像处理方法容易受外在环境因素干扰,可靠性差.因此亟待研究快速准确的方法以实现山楂产地的精准鉴别;受CoAtNet与Swin-Transformer网络启发,本文结合MBConv模块中深度可分离卷积网络对局部信息建模的特点与Swin Transformer模块多层次结构可弥补网络非局部性损失的特性,提出一种多尺度特征的混合神经网络模型,通过获取图像不同层级特征,将获取的形状、颜色与纹理等浅层特征作为先验知识与高层级语义信息进行特征融合,研究了一种快速有效的识别方法以实现对不同产地山楂的有效鉴别;此外,本文提出一种新的局部空间注意力机制,通过形成通道注意力模块联合空间注意力模块的新结构,实现对图像细粒度特征的聚焦与学习.实验结果表明,本文所提出的方法有最高的鉴别准确率为89.306%,优于其他基线模型.实践证明,本文的研究提高中药材鉴别的科技水平,拓宽传统中医药的研究思路.

基于细粒度特征的面料图像检索

面料图像检索对于纺织工厂面料库存和样品管理意义重大,但面料外观的多样性以及织物纹理的精细性,使得在面料检索时面料的特征提取较困难。该研究提出一种基于细粒度特征的面料图像检索算法。该算法使用坐标注意(coordinate attention,CA)模块来提取图像的精准位置信息,并将缩放系数法用于在宽度和高度方面整体缩放MobileNetV3的网络结构以减少模型参数数量,达到减少网络训练时间的目的。据此筛选出提取面料图像细粒度特征的最佳模型,在面料图像数据集(fabric image dataset,FID)上进行面料检索实验。结果表明,该算法有效提高了面料图像细粒度特征提取的准确性,检索精度达到91.82%,浮点运算数达到175.34 MB。检索精度比MobileNetV3原模型提高了13.49个百分点,同时减少了网络训练时间,速度提高了25.14%。该算法具有实际应用价值。

基于特征增强和分组模块的车型精细识别

针对车型种类多、差异小,模型复杂,识别精度低的问题,提出一种基于特征增强和分组模块的车型精细识别方法,在ResNet网络基础上改进,在卷积块中加入多尺度通道域和空间域的注意力机制,增强对重要的特征提取,并将多通道特征图进行分组,根据分组损失函数不断优化分组,通过加权方式结合KL(Kullback-Leibler)散度损失函数和交叉熵损失函数,有助于网络学习类内差异小、类间差异大的特征。该方法在Stanford cars-196数据集和自制数据集上进行测试,验证了所提模型的有效性。

基于小波多尺度分析大气边界层高度提取方法研究

大气边界层物理演变特征及复杂边界层结构对大气污染过程的影响机制研究中,迫切需要边界层精细化结构的探测手段。通常使用梯度法和小波协方差变换法的方式进行边界层高度提取,但由于该方式容易受到噪声和气溶胶层结构的影响,实验误差较大。提出使用小波多尺度分析方法细化边界层特征结构,从而筛选出有效的细节信息,提高对探索边界层高度的准确性。此外,基于GBQ L-01激光雷达设备实测合肥地区全天气溶胶分布情况,使用小波多尺度分析方法分析对应的平行分量距离平方矫正信号、垂直分量距离平方矫正信号和退偏振比全天分布情况。实验结果表明,该方法准确性高,与梯度法、小波协方差法相比具有更高的稳定性与连续性。对特殊时间点进行数据分析,明确主要影响边界层高度的主要因素,并确定各个时间段的大气边界层高度值。

时空多尺度关联特征融合的二维卷积网络细粒度动作识别模型

针对传统二维(2D)卷积网络提取时空特征尺度单一以及对细粒度动作数据集中帧与帧之间的远程时间关联信息利用不足的问题,本文提出时空多尺度关联特征融合的2D卷积网络细粒度动作识别模型。首先,为建模视频多尺度空间关联以加强对细粒度视频数据的空间表征能力,模型使用多尺度“特征压缩、特征激发”方式,使网络所提取空间特征更加丰富有效。然后,为充分利用细粒度视频数据时间维度上的运动信息,本文引入时间窗口自注意力机制,利用自注意力机制强大的远程依赖建模能力同时只在时间维度上进行自注意力操作,以较低计算成本建模远程时间依赖关系。最后,考虑到所提取时空特征对不同类型动作分类的贡献不均等,本文引入自适应特征融合模块,为特征动态赋予不同权重实现自适应特征融合。模型在2个细粒度动作识别数据集Diving48和Something-somethingV1上识别准确率分别达到86.0%和46.9%,分别使原始主干网络识别准确率提升3.8%和1.3%。实验结果表明,在只使用视频帧信息作为输入的情况下,本模型达到与现有基于Transformer和三维卷积神经网络(3D CNN)算法相当的识别准确率。

邻体竞争对城市森林微景观中针叶树视觉形态性状的影响

树木视觉形态性状是城市绿地微景观美学质量的重要影响因素之一,树木视觉形态性状的变化与其周围邻体木的竞争作用息息相关,但邻体竞争对树木视觉性状的作用机制尚不明确。研究于2022年8—10月对北京市城市公园中常见的针叶树种白皮松、侧柏、油松、圆柏展开调查,从树冠形态、干冠协调、树干形态3个方面构建了9个树木视觉形态性状指标,采用3个不同的竞争指标分析邻体竞争对针叶树种视觉形态性状的影响。结果表明,针叶树的树冠形态对邻体竞争的响应比较敏感,竞争中的白皮松、侧柏、油松、圆柏偏冠指数与孤立木相比分别提高了16.95%、28.95%、22.76%、17.67%;树冠缺失率分别提高了3.92%、6.09%、4.87%、4.95%。与孤立木相比,部分针叶树种高径比、分枝角变异度在多侧竞争环境中显著提高,而树冠舒展度则显著降低。邻体竞争强度越大,针叶树树冠的偏移与缺失程度越大。当针叶树受到强烈的侧方竞争时,对象木树冠和树干的径向生长显著受阻,表现为树冠舒展度大幅度下降、高径比显著提高,使树冠向细高方向发展。当针叶树上方的生长空间被占据时,其轴向生长同样受到严重阻碍,树木的冠径比和高径比维持较稳定状态。总体而言,四种针叶树种在视觉形态上对竞争胁迫的响应具有一定差异性,其中油松最为敏感,圆柏次之。综上,在城市森林微景观中,邻体竞争会导致针叶树种的树冠及树形发生明显变化,这种变化主要受综合资源竞争的影响,与其周围潜在生长空间的大小及对称性有关。在城市森林景观营建时,建议将针叶树栽植在对称的竞争环境中,但是其邻体木不宜过高,通过四周邻体木的适度竞争,能够促进针叶树的轴向生长,同时降低树冠偏移或变形的风险,提高其视觉美学效果。

基于细粒度识别的即插性多粒度特征融合算法

为丰富卷积神经网络下细粒度图像特征的表征信息,进一步扩大类间差异缩小类内差异,提出一种基于多种粒度图像训练的特征融合方式,挖掘图像的细化特征。通过逐渐改变输入图像的粒度值,构造一个包含多种粒度信息的网络模型;提取多粒度多尺度图像特征,与原始特征进行融合,完成最终的分类预测。无需引入其它辅助网络,在没有显著增加模型参数的情况下,融合不同粒度特征。实验结果表明,其分类准确度高于只包含单一粒度图像的训练结果,验证了该方法能有效丰富特征信息。

含细观缺陷固体推进剂非冲击点火性能研究进展

含有细观缺陷(微裂纹、微孔洞)的固体推进剂在受到低压长脉冲的载荷时可能发生意外起爆事故,严重影响着固体推进剂的安全性。从固体推进剂的点火机理出发,综述了固体推进剂中细观缺陷观测实验与表征方法及非冲击点火试验;梳理了含细观缺陷固体推进剂非冲击点火数值模拟的相关研究手段及成果;并基于固体推进剂内细观缺陷类型,系统地总结了细观缺陷对固体推进剂非冲击点火性能的影响,对含细观缺陷固体推进剂非冲击点火性能的数值模拟研究及非冲击点火的试验设备发展作了展望。

基于多尺度YOLOv5的交通标志检测

针对小目标交通标志检测存在的检测精度低、漏检率高等问题,提出了一种基于多尺度融合的YOLOv5改进算法。在主干网络后输出4个有效特征层以便更好地融合多尺度信息,在主干网络输出的3个特征层中添加改进的多尺度融合注意力机制CBAM_U,以提升网络的检测能力;在Path Aggregation Network(PANet)下采样过程中添加Fusion模块,促进不同感受野下特征的细融合;在YOLOHand前加入Adaptively Spatial Feature Fusion(ASFF)模块解决特征金字塔融合的不一致性,进一步提升网络的表达能力。实验结果表明,提出的方法相比于原始YOLOv5网络在CCTSDB数据集中mAP@0.5提升了3.07%,召回率提升了3.83%,查准率提升了1.64%,F1-Score提升了2.66%,相比于其他检测算法,改进后的YOLOv5算法在复杂场景中具有更好的鲁棒性。

青藏铁路沿线气候特征精细化模拟研究

青藏铁路格拉段(以下简称青藏铁路),特别是从西大滩至安多穿越550 km多年冻土区的路段,沿线有着复杂的地形地貌和冻土环境。近年来随着气候变暖和多年冻土退化,对于建造在多年冻土之上的青藏铁路的维护需求剧增,为了能充分捕捉青藏铁路沿线地区复杂的地形地貌对局地气候变化的影响,以期对青藏铁路的运行维护提供理论支持。本文利用铁路沿线站点观测数据,并以ERA5数据驱动WRF模式进行网格距离为10 km×10 km的动力降尺度模拟。结果显示,青藏铁路沿线的六个站点自1998-2020年普遍呈现增温趋势,增温率最低的站点为0.27℃·(10a)^(-1),最高为0.56℃·(10a)^(-1);WRF模式对表面气温的模拟在年平均气温结果与观测数据略有差异,在夏季和秋季的模拟结果较好,夏季相关系数达到0.95以上,秋季在0.80以上,春季和冬季较差;WRF降水模拟的结果显示,Nudging方法有效改进了青藏高原夏季降水的湿偏差,青藏铁路北段至铁路南段降水逐渐增加,且在铁路中部降水出现极大值。WRF模式在青藏高原的温度和降水模拟上仍存在着一定的冷偏差和湿偏差,寻找新的方法或利用高质量驱动数据来驱动模式将可能对青藏高原地区的降尺度模拟结果有更进一步的改善。

小尺度高寒湿地放牧干扰格局——以纳帕海湿地为例

高寒湿地放牧干扰强烈,是其生态系统退化的主因。湿地植被生物量空间格局不仅能够反映自然因素的作用,也能够体现地域放牧等人类活动的主要特征,因此常作为表征放牧干扰格局的主要度量方法。但尺度效应导致的植被生物量影响要素在不同空间尺度下存在较大差异,尤其是针对面积为几十至几百km^(2)范围的小尺度高寒湿地而言,自然因素和放牧等人为干扰对其植被地上生物量的共同作用特征仍不明确,缺乏针对该类型湿地的合理评估模式。以滇西北国际重要湿地纳帕海为例,在排除其它人为干扰的条件下,通过植被地上生物量的空间异质性特征来表征其放牧干扰格局,并以该案例为基础提出一种小尺度高寒湿地放牧干扰格局评估模式。研究尝试基于详细地面植被调查,获得潜在植被地上生物量,利用调查同时期高分定量遥感方法(法国PLEIADES卫星数据结合实测植被地上生物量建模)获得实际植被地上生物量,并使用潜在与实际植被地上生物量之差的空间异质性特征表示纳帕海湿地放牧干扰格局。该评估模式可以有效减少对于小尺度湿地放牧干扰数据的依赖,同时减少牛羊马等大牲畜散养带来的空间分布随机性。研究结果显示纳帕海湿地放牧干扰对湿地植被地上生物量的影响严重,影响比例约为93.5%。研究发现水文等自然因素和放牧等人为干扰分别决定纳帕海湿地植被地上生物量的上限和下限,并基于此提出四方面干扰调控策略。

鲁西北潘店重磁异常带矽卡岩型铁矿精细找矿方向——来自广域电磁法深部找矿的启示

山东省鲁西北潘店地区超深覆盖区航磁和重力异常明显,通过系统的重磁异常查证工作,在研究区重磁异常带开展的多个钻孔均取得了深部矽卡岩型铁矿床找矿突破,但由于重磁异常带范围较大且其特征不能直接指示找矿方向,造成部分钻孔不见矿或见矿情况差。因此,为了更好地在重磁异常带精细圈定成矿位置,实现新的更大的找矿大突破,本文旨在已查证的潘店重磁异常带开展广域电磁测深(WFEM)工作,结合钻探验证情况,圈定深覆盖区矽卡岩型铁矿深部成矿区间及成矿有利结构面,提出深部精细找矿方向。