After scrolling through numerous social media posts documenting ice and snow tourism activities,Sun Qi and her friends couldn't wait to visit the frosty city of Harbin in northeast China’s Heilongjiang Province.After all,as a staff member at a university in Zhejiang Province,she has the opportunity to travel during the ongoing winter break.
Check dams are widely constructed on China's Loess Plateau,which had a total number of 58,776 by the end of 2019.Great achievements in check dam construction have been gained regarding the economic and environmental impacts.This study reviews the remarkable benefits of check dams on the land reclamation and environmental improvement on the Loess Plateau,and sediment reduction entering the Yellow River.However,the flood incidents on check dams have been frequently reported for the past decades,which has attracted more attention in the context of climate change and extreme rainfall events recently.Advances in the flood migration techniques achieved by the research group led by the first author have been highlighted to migrate the breach risk of check dams due to floods.The“family tree method”has been proposed to determine the survival status and critical rainfall threshold of each check dam in the complicated dam system.An updated dam breach flood evaluation framework and the corresponding numerical algorithm(i.e.,DB-IWHR)have been developed.Moreover,innovative types of water-release facilities for check dams,including geobag stepped spillway and prestressed concrete cylinder pipe in the underlying conduit,have been proposed and developed.Finally,the perspectives concerning the check dam construction on the Loess Plateau have been put forward.
随着定位技术和传感器的高速发展,用户移动轨迹数据日渐丰富,但大多分散在不同平台上。为了全面利用这些数据并准确反映用户的真实行为,对轨迹用户匹配的研究变得至关重要。该任务旨在从海量签到轨迹数据中精准关联用户身份。近年来,研究者们尝试运用循环神经网络、注意力机制等方法深入挖掘轨迹数据。然而,当前方法在处理用户签到轨迹时面临两大挑战:一是签到数据中有限的时空特征不足以从主观和客观两个角度全面地建模签到点信息,二是用户的签到轨迹往往围绕着一个特定的主题。针对这两点挑战,提出了一种基于自然语言增强的轨迹用户匹配模型(Natural Language Augmented Trajectory User Link,NLATUL)。首先,设计了一套自然语言模板与软提示令牌来描述签到轨迹,并使用语言模型来理解签到点中的主观意图,融合用户的时空状态,提供了一种充分从主观与客观两个方面建模签到点的方法;在此基础上,通过提示学习的方法推理签到轨迹的主题,并对建模的签到点表示的轨迹进行双向编码,通过签到轨迹主题与签到轨迹编码的结合实现对用户签到轨迹的准确理解。在两个真实世界签到数据集上验证的实验结果表明,NLATUL能够更准确地匹配签到轨迹与其对应的用户。
