Dynamics of iron and aluminum storages in a subtropical forest headwater stream

The forest headwater streams are important hubs for connecting terrestrial and aquatic ecosystems,with plant litter and sediments as the major carriers for material migrations;however,until now we knew little about the dynamics of trace elements such as iron(Fe)and aluminum(Al)in forest headwater streams.Here,we quantitatively identified the spatiotemporal dynamics of Fe and Al storages in plant litter and sediments and their influencing factors in a subtropical forest headwater stream,and assessed the potential pollution risk.The results showed that:(1)the mean concentrations of Fe and Al in plant litter(sediments)were 5.48 and 8.46(7.39 and 47.47)g·kg^(-1),and the mean storages of Fe and Al in plant litter(sediments)were 0.26 and 0.43(749.04 and 5030.90)g·m^(-2),respectively;(2)the storages of Fe and Al in plant litter and sediments significantly fluctuated from January to December,and showed a decreasing pattern from the source to mouth;and(3)storages of Fe and Al had no significant correlation with riparian forest type and the present of tributary and the Fe and Al storages in plant litter were mainly affected by water temperature and water alkalinity,and their storages in sediments were mainly affected by water temperature and frequency of rainfall;and(4)there were no anthropogenic pollution in Fe and Al in the forest headwater stream.Our study revealed the primary factors of concentrations and storages of Fe and Al in plant litter and sediments in a forest headwater stream,which will improve our understanding of the role of headwater streams in forest nutrient storage and cycling along with hydrological processes.

Comparing Freshwater Benthic Macroinvertebrate Communities in Forest and Urban Streams of the Coastal Ecological Region of Cameroon

In this study conducted in the coastal zone of Cameroon, biological indices and functional feeding groups of benthic macroinvertebrates were used to assess the health status of two urban streams. For a better diagnosis, two streams located in coastal forest zone were used as a reference. Benthic macroinvertebrates were sampled monthly over a 3-month period (from May to July 2017) in six urban stations and six forest stations. Measurements of the physicochemical variables were done simultaneously. Physicochemical analysis revealed that urban streams are strongly polluted with high content of decaying organic matters, while forest streams are slightly polluted as indicated by the Principal Component Analysis. Concerning benthic macroinvertebrates, urban streams are poorly diversified with the proliferation of taxa tolerant to water pollution and belonging to the functional feeding groups of collectors-gatherers. Inversely, forest streams are more diversified and dominated by sensitive taxa, most belonging to the functional feeding groups of predators and shredders. These marked differences between biological indices and feeding mode of benthic macroinvertebrates in forest and urban rivers confirm the reliability of benthic macroinvertebrates as good indicators of freshwater ecosystem in the coastal zone of Cameroon.

Random Forest Based Very Fast Decision Tree Algorithm for Data Stream

The Very Fast Decision Tree(VFDT)algorithm is a classification algorithm for data streams.When processing large amounts of data,VFDT requires less time than traditional decision tree algorithms.However,when training samples become fewer,the label values of VFDT leaf nodes will have more errors,and the classification ability of single VFDT decision tree is limited.The Random Forest algorithm is a combinational classifier with high prediction accuracy and noise-tol-erant ability.It is constituted by multiple decision trees and can make up for the shortage of single decision tree.In this paper,in order to improve the classification accuracy on data streams,the Random Forest algorithm is integrated into the process of tree building of the VFDT algorithm,and a new Random Forest Based Very Fast Decision Tree algorithm named RFVFDT is designed.The RFVFDT algorithm adopts the decision tree building criterion of a Random Forest classifier,and improves Random Forest algorithm with sliding window to meet the unboundedness of data streams and avoid process delay and data loss.Experimental results of the classification of KDD CUP data sets show that the classification accuracy of RFVFDT algorithm is higher than that of VFDT.The less the samples are,the more obvious the advantage is.RFVFDT is fast when running in the multithread mode.

Factors Impacting Water Quality of a Small Stream in a Mountain Forest

Recent improvements to sewerage systems have meant that the relative contribution of point sources of pollution to the overall pollutant loads of streams has decreased markedly. Consequently, the potential increase in non-point source pollution means that it may be necessary to focus more on pollution loads from non-point sources in the future. This study examined pollution loads from non-point sources in streams flowing through forested areas. In addition, the relationship between runoff and pollution loads was also clarified. The small streams in the Tohoku University Botanical Gardens, in Sendai city, Japan, were sampled during a dry weather period and their water quality parameters were characterized. Chemical Oxygen Demand （COD） concentration increased with distance downstream, possibly because the soils in downstream areas contained high amounts of organic matter. Conversely, the concentration of nitrate nitrogen （NO3-N） decreased with distance downstream, probably because upstream soils were generally in an oxidized state while those further downstream were reduced. COD concentration increased with air temperature, while NO3-N levels decreased with an increase in air temperatures.

Spatiotemporal Variation of Benthic Macroinvertebrates in Some Tropical Forest Streams of the Nyong Catchment (Cameroon)

With the aim of assessing the benthic macroinvertebrates’ diversity, a study was carried out in some tropical forest streams of the Nyong River catchment in Cameroon from February 2019 to February 2020. A total of 167 samples were carried out from 13 stations during 13 months. Some environmental variables were measured. These parameters varied more or less from station to station but significantly at the temporal level. In this study, 13,690 benthic macroinvertebrates belonging to 4 phyla, 7 classes, 16 orders and 93 families were collected. The benthic macroinvertebrates were more abundant and more diversified in the stations whose waters are well oxygenated and present a moderate current compared to the stations with a very weak current. The diversity varied significantly from 1.33 ± 0.14 bits/ind to 2.00 ± 0.35 bits/ind and the high values were found in stations with multiple substrates and well-oxygenated waters. Temporally, the diversity varied significantly from 1.10 ± 0.16 bits/ind in NM sampling station in February 2020 to 1.87 ± 0.1 bits/ind in the OB sampling station in September. In addition, the settlement was more abundant during the short dry season, more precisely during the month of August (1471 individuals) but richer during the long dry season during the month of February (54 families). The distribution of the abundances of the benthic fauna in the different stations made it possible to identify five typological groups using the rarefaction curves, the ascending hierarchical classification and the principal component analyses. Each of these five groups is characterized by a specific taxonomic richness, composition and abundance.

Influence of Tea Plantations, Forest and Mixed Farming on Stream Flow and Sediment Loads, Case of Sondu Miriu River Basin, Kenya

The changing patterns of land cover and land use in the tropical river basin over time are critical. The hydrological phenomena at basin and sub basin scale are affected positively or negatively by dynamics of the land cover and land use patterns. Hence identifying causes and driving factors aid in taking appropriate measures to avert the impacts. This study determined the influences of sub basins dominated by tea plantations, forests and agricultural land uses in terms of streamflow and sediment flux variability in Sondu Miriu River Basin in Kenya, East Africa. Field-based investigations were conducted through sampling of flow velocities, turbidity and TSSC obtained from existing River Gauging Stations established within the three sub basins. The sub basin dominated by mixed farming land cover exhibits high turbidity approximately 620 NTU and high levels of total suspended sediment concentration (TSSC) of the order of 630 mg/l in wet seasons. The turbidity levels and TSSC were low in sub basins dominated by forest and tea plantations with approximately mean value of 17 - 29 NTU and 0.019 g/l. The sediment loads in sub basin dominated by mixed farming in the pre planting season in January to February were about 900 tonnes/day higher than that in crop growing season. In sub basins dominated by forest cover and tea plantations, sediment loads were low ranging between 2 - 7 tonnes/day. The relationship between stream flows and area under tea plantations, forests and mixed farming ranged between R2 of 0.025 and 0.16. Tea plantations and forests influence the stream flows and sediment yields in long term duration while in mixed farming variations were observed seasonally. The strong relationships between rainfall and stream flows at the sub basins ranging between R2 of 0.84 and 0.97 revealed the significance of rainfall in hydrologic response of the Sondu Miriu River Basin.

基于多模态注意力网络的红外人体行为识别方法

深度学习网络对红外单一模态数据的学习表征能力具有一定的局限性,针对该问题,文中提出了基于多模态注意力网络的红外人体行为识别方法。由于深度学习网络模型无法直接对视频信息进行训练和分类,首先,通过预处理模块将得到的视频信息预处理成红外视图,再将得到的红外视图通过Sobel算子和基于L 1范数的全变分光流法分别提取红外视图的边缘信息和光流信息得到边缘视图和光流视图;其次,将红外视图、边缘视图、光流视图分别输入融合注意力机制模块的三流网络中进行特征学习;然后,对三流网络中每个网络提取的多模态特征进行融合;最后,将融合得到的特征向量输入随机森林进行训练和分类。在公开数据集NTU RGB+D和自建数据集上进行实验,结果表明了所提方法具有不错的识别效果。

Concentration variations of several ions in stream after a wildfire

In May 2006, a high intensity wildfire occurred in Songling forest region in Daxing＇an Mountains, China. The concentration changes of eight ions （K^＋, Na^＋, Ca^2＋, Mg^2＋, Cl^-, Br^-, NO3^- and SO4^2-） were measured in burned and tmbumed streams after fire from May to Oct., 2006. Results show that the most ions flux were higher in burned stream than that in unburned stream during the sampling period, and the greatest concentrations of most ions transported from burned stream occurred in July. After fire, the most amplitude chemical ion was Ca^2＋, whose average concentration was 5.50 mg·L^-1 higher than that in unburned stream, and the total concentration of every chemical ion presents a trend Ca^2＋〉SO4^2- 〉Na^＋〉Mg^2＋〉NO3^- . The average concentrations of Ca^2＋, SO4^2- , Na^＋＇Mg^2＋,NO3^- showed an increase trend, but those of K^＋, Cl^- , Br had a decreased trend. SO4^2- had the largest loss among these anions, followed by NO3^-. Overall, the increase degree of cation was greater than that of anion after burning.

基于闭合频繁模式的半随机森林数据流分类算法

提出了一种基于闭合频繁模式的半随机森林数据流分类算法(Semi-Random Forest based on Closed Frequent Pattern,SRFCFP),以解决数据流中噪声和概念漂移问题。SRFCFP利用闭合频繁模式对数据流进行表示,去除冗余信息和噪声,突出数据特征。采用半随机森林建立分类模型,并通过基于时间衰减的模式集更新机制适应数据流的无限性。为了检测概念漂移并及时适应,引入了一种模式集差异性度量方式,用于测量数据分布变化。实验结果表明,在MOA平台下使用真实和合成数据集,SRFCFP在平均精度上超越了相关对比算法,并能有效处理数据流中的概念漂移和噪声问题。

基于机器学习算法的服装直播销量预测模型

为了完善直播销量预测的指标体系,提高直播销量预测的准确率,本文对比多种机器学习算法,分析服装直播销量的影响因素并预测服装直播销量。首先,通过文献分析选取直播销量的影响因素,并使用Spearman相关系数结合显著性进行筛选。其次,通过不同机器学习算法建立服装销量预测模型。选用R 2、MAE、RMSE及MAPE为评价指标,采用5折交叉验证的方法,测试各模型性能。研究结果表明:主播粉丝数、主播近30天场均观看人次、主播近30天场均坑产、产品价格、产品讲解时长、产品近30天历史销量、品牌粉丝数、品牌近30天历史销量、折扣这9个因素之间共线性较弱且与直播销量之间的相关性显著,可作为预测模型中的影响因素;预测算法中K近邻算法和随机森林算法的表现较好,R 2均大于0.98,MAPE均在30.5%以内。预测结果可帮助零售商规划库存,调整生产计划,为产品采购、定价、推广提供数据支持。

川西高寒森林溪流好氧不产氧光合细菌多样性

为了探明好氧不产氧光合菌(AAPB)在川西高寒森林溪流极端环境中的生态分布和群落结构及其环境影响因子,本实验结合同步温度动态和水环境特征监测,采用实时荧光定量PCR和克隆文库的方法,检测了川西高寒森林溪流季节性冻融循环3个关键时期(冻结前期、深冻期和融化期)AAPB的丰度和多样性变化及其与环境因子之间的相关性.结果表明:AAPB丰度在冻结前期非常低,仅(2.80±0.19)×10~4bp/mL;深冻期迅速增加到(4.65±0.30)×10~4bp/mL;融化期达到(5.79±0.19)×104~bp/mL.系统发育分析显示,冻结前期AAPB类群主要为α-变形菌纲细菌;而深冻期和融化期以β-变形菌纲细菌占据主要地位.相关性分析表明,在季节性冻融循环过程中,叶绿素a、可溶性有机碳和温度是AAPB菌群丰度和多样性的主要影响因子.实验结果表明冬季的川西高寒森林溪流中,AAPB类群的丰度低,但是它们高度动态变化,并且与其他淡水环境中的该类群的同源性低.

高寒森林溪流微生物群落结构的季节性变化

高寒森林溪流不仅是区域河流的源头,而且是联系陆地与水域的生态纽带。微生物活动可能成为控制溪流生态系统过程的关键因子,但其结构与动态过程缺乏必要关注。因此,结合同步温度动态监测,采用实时荧光定量PCR和DGGE技术,在2014年到2015年冻融季节和生长季节关键时期对比研究了川西高寒森林溪流和森林林下土壤中微生物群落的动态特征。研究结果发现,高寒森林溪流具有较低的真菌和细菌群落丰度;与森林土壤相同,溪流在冻融季节表现出相对生长季节更高的真菌/细菌比,而且从冻融季节到生长季节,溪流微生物丰度动态也表现出明显的季节性变化特征。与森林土壤不同的是,溪流中细菌和真菌的丰度及其Shannon-Wiener多样性指数的最高值均出现在生长季节而不是冬季冻融季节,并且溪流中细菌丰度在季节性变化的不同时期具有显著差异(P<0.05)。此外,森林土壤细菌类群以芽孢杆菌属(Bacillus sp.)比例相对较高,真菌类群则以格孢菌属(Pleosporales sp.)、曲霉属(Aspergillus sp.)和其他一些子囊菌门(Ascomycota)的类群为优势;而溪流细菌类群以红球菌属(Rhodococcus sp.)为主,真菌类群则以曲霉属和空团菌属(Cenococcum sp.)为主。同时,季节性变化中温度、p H、水溶性有机碳和溶解氧等环境因子可显著影响溪流微生物群落结构及其组成,这些环境因子在高寒森林溪流微生物群落的季节性变化过程中具有重要的作用。

GIS林相专题图的建立及在护岸公益林规划中的应用

对帽儿山的河流、施业区、林班、小班建立了ＧＩＳ专题库．对各小班的资源数据库均并入了ＧＩＳ的 属性表。为河流、集水区、林班建立了统一、灵活、多用途的编号系统。应用ＧＩＳ对集水区及护岸公益林进行分 级配套规划。提出了每米护岸公益林占地率的指标。并根据此指标对帽儿山北林施业区的部分河流划分了护岸公 益林，规划结果直接成图。

长白山二道白河森林流域溪流倒木调查研究

溪流倒木是森林生态系统对水生态系统最重要、最直观的输入和干扰之一 ,也是两系统之间的主要联结 ,对于溪流生态系统的稳定、水生生物多样性、河槽形态及其变化过程有着重要的作用。重点对长白山北坡溪流倒木现存量进行了调查和研究 ,在调查的红松阔叶林植被带内 45 0 0 m长河道内 ,共发现溪流倒木 42 5株 ,分属于 1 7个树种 ;其中 、 级腐烂占相当大的比重 ,与林地倒木 、 级腐烂占优有所不同 ,其原因可能与分解环境的不同有关。所有溪流倒木的总材积为 77.98m3,故溪流倒木的现存量为 1 .73 3 m3/1 0 0 m和 1 0 .83 m3/hm2 。溪流倒木的树种组成和不同树种的材积与河岸带植被密切相关 ,但存在差异。研究表明林分形成倒木并进入河流在时间上可能是均匀或随机的 ,但不同树种间 ,其形成倒木并进入河流时的树木材积或生长年龄存在较大差异。溪流倒木和林地活立木的个体数量的径级分布基本上为反 J型 ,而它们材积的径级分布均为典型的

广东省罗坑自然保护区鳄蜥对溪沟林型的选择

2006年5—9月在广东省罗坑自然保护区,对大竹园管护站所辖区域内鳄蜥栖息溪沟的林型进行了调查,采用选择指数法研究了鳄蜥对溪沟林型的选择性。共调查了154只鳄蜥的生境特征,并对5种溪沟林型(常绿阔叶林、针阔混交林、针叶林、竹林和灌草)进行了样方调查。结果表明:鳄蜥对溪沟林型的选择是非随机的,偏好常绿阔叶林,对竹林接近于随机选择,回避其它林型;鳄蜥选择常绿阔叶林可能与其植被盖度较高、涵养水源能力强以及土壤有机质含量高有关。建议加强对常绿阔叶林的保护工作,以更好地保护鳄蜥。

森林和沼泽对溪流水化学特征的影响

以小兴安岭北部公别拉河上游为研究区,于2004年7～9月对森林溪流和沼泽溪流水样进行水化学特征对比分析.结果表明,森林和沼泽溪流水化学类型均为重碳酸盐类钙组Ⅰ型水（CCaⅠ）.森林溪流水的pH、矿化度、总硬度、HCO3^-、SO4^2-、Ca^2＋、Mg^2＋、Fe均低于沼泽溪流,而总氮、总磷、Cl^-、K^＋、Na^＋则高于沼泽溪流.森林溪流和沼泽溪流中重金属元素Fe、Mn、Cu、Zn、Cd、Hg和Pb含量较低,均未超过我国Ⅰ类地表水环境质量标准.森林溪流中总氮含量为（0.27±0.04） mg·L^-1、总磷含量为（0.040±0.005） mg·L^-1,明显高于沼泽溪流中总氮含量（（0.21±0.02） mg·L^-1）和总磷含量（（0.025±0.004） mg·L^-1）,沼泽湿地对N、P有较强的储存和吸附能力,且对NH4^＋ -N的吸附作用远大于对NO3^- -N的吸附.沼泽溪流中Fe含量为（0.26±0.05） mg·L^-1,显著高于森林溪流Fe含量,沼泽湿地对Fe起到还原释放作用.

不同森林群落类型溪流水化学特征的季节动态

以小兴安岭凉水自然保护区内的阔叶红松林、云冷杉林和落叶松人工林为研究对象,于2006年3—10月,分析了其溪流水化学特征的动态变化.结果表明:不同月份3种森林群落溪流水的主要阳离子含量均表现为Ca2+>Na+>K+>Mg2+,主要阴离子含量均为HCO3->SO42->NO3->C l-;不同群落类型的主要离子含量影响显著,3种森林群落溪流水中Na+、Ca2+、Mg2+、Fe2+和Fe3+平均含量为云冷杉林>落叶松人工林>阔叶红松林,而K+为落叶松人工林>云冷杉林>阔叶红松林;主要阴离子平均含量均以落叶松人工林溪流水中为最高.

西双版纳热带季节雨林林下土壤蒸发的稳定性同位素分析

利用2002年1月～2003年12月在西双版纳热带季节雨林样地测定的大气降雨、林冠穿透雨、林冠滴落雾水、溪流水水量及其氢、氧稳定性同位素比率δD和δ18O的测定结果,分析了热带季节雨林林下土壤的蒸发率.结果表明,热带雨林在雨水偏多的2002年和雨水偏少的2003年的蒸发散分别为1186 mm和987 mm,明显低于世界其它湿润热带雨林的蒸发散值;雾水的稳定性同位素比值明显重于雨水,表明雾水中包含区域再循环的蒸发水汽;由于森林林冠和林下土壤的蒸发作用,穿透雨水和溪水的稳定性同位素值又明显重于雨水的相应值.根据溪流水和穿透水（穿透雨水＋滴落雾水）的同位素差值,利用平衡状态下的瑞利蒸馏方程计算的森林土壤蒸发率表明,这两年的森林总蒸发散中分别有5.1%和3.1%来自林下土壤的蒸发.分析认为,持久、浓重的辐射雾是导致西双版纳地区热带雨林蒸发散和土壤蒸发率较低的重要因子.