基于机器学习方法的上海市暴雨内涝灾情预测模型研究

本文基于区域灾害系统理论,综合考虑致灾因子、暴露度和脆弱性,提出了一套暴雨内涝灾情预测指标体系;在此基础上利用上海市应急联动中心110接报暴雨内涝灾情数据,构建了暴雨内涝灾情预测BP模型(Back Propagation Model)和XGBoost模型(Extreme Gradient Boosting Model),并对比分析了预测模型效果,实现对上海市暴雨过程内涝灾情数量预测;最后对内涝灾情影响等级进行阈值划分,以期为暴雨内涝影响预报与风险预警业务、服务及灾害管理提供技术支撑。结果表明:1)综合考虑致灾因子、暴露度、脆弱性指标且不经主成分分析降维的指标组合作为暴雨内涝灾情预测指标体系时,BP模型和XGBoost模型的预测精度最优;2)全量样本XGBoost模型总体表现最优,暴雨内涝灾情的右偏分布和内涝灾情的异常高值均对XGBoost模型预测误差有不同程度的贡献;3)综合评价法在历史灾情百分位法和模拟灾情百分位法基础上,结合多年业务实践经验和用户对于110灾情的处置承受力对暴雨内涝灾情进行阈值划分,在实际应用中具有一定参考意义。     

基于SSA-XGBoost方法的降水变化趋势预测模型

全球变暖预期下不同区域的未来降水变化,是政府和公众都关心的重要问题,也是全球变化研究的前沿科学问题。预测模型的建立是预测/预估研究的重点和难点,现今不稳定的气候背景对预测模型的精准度提出了更高的要求。为了解决传统方法对长期时间序列预测效果欠佳的问题,本文以泰国南部洞穴石笋δ¹⁸O重建的过去270多年（公元1773—2004年）的降水记录为数据集,提出了SSA-XGBoost预测模型。对原始数据去趋势预处理后,采用奇异谱分析法（SSA）提取前部分数据（1773—1964年）的振荡成分以确定数据的最佳谐波个数,并进行准确的周期信号分量分解;之后用去趋势数据减去周期信号得到随机信号,再利用XGBoost模型对随机项进行预测;最后将预测的序列、趋势曲线和周期信号延拓结果相叠加得到最终的预测数据（1965—2004年）。与其他四种模型（XGBoost、ARIMA、SSA-ARIMA、LightGBM）的预测结果相比,SSA-XGBoost的预测结果与真实值最相近,且MAE和RMSE均最小,R²也更接近1,说明该模型具有更高的精度和稳定性。该研究对于泰国南部等热带地区未来的降水变化趋势预测具有较好的指导意义,也可为其他长时间序列的预估研究提供借鉴。     

基于GA参数寻优的决策树支持向量机生态环境质量评价方法

选取决策树作为支持向量机多类分类方法,选择径向基核函数建立了生态环境质量决策树支持向量机评价模型,基于遗传算法实现了惩罚因子、核函数参数的自适应优选,并运用建立的模型对我国主要省市生态环境质量进行了评价。研究结果表明,该方法可以较好地实现生态环境质量评价。     

环境DNA宏条形码监测水生态系统变化与健康状态

开展快速可靠的水生态监测并预测其变化趋势,对保护水生态环境具有重要的价值。近年来,环境DNA宏条形码技术(简称环境DNA技术)的快速发展弥补了传统形态学生物监测的缺陷,显著提升了水生生物群落的监测能力。与机器学习、遥感和云服务等技术结合,环境DNA技术不仅能大尺度、高频率、高灵敏度、自动化地获取生态监测信息,而且能准确地识别水生态系统的变化趋势,进而改变对水生态系统的认识与管理方式。因此,研究着重总结了环境DNA技术在水生态监测中的应用,分析了环境DNA技术与机器学习、卫星遥感等跨学科合作的潜在机遇,基于环境DNA技术简单、便捷的优势,提出了社会公民参与水环境保护的生态监测新思路。     

Using a social learning configuration to increase Vietnamese smallholder farmers’ adaptive capacity to respond to climate change

Social learning is crucial for local smallholder farmers in developing countries to improve their adaptive capacity and to adapt to the current and projected impacts of climate change. While it is widely acknowledged that social learning is a necessary condition for adaptation, few studies have systematically investigated under which conditions particular forms of social learning are most successful in improving adaptive capacity of the most vulnerable groups. This study aims to design, implement and evaluate a social learning configuration in a coastal community in Vietnam. We make use of various methods during four workshop-based interventions with local smallholder farmers: interviews with key farmers and commune leaders, farmer-to-farmer learning, participatory observations and focus group discussions. The methods for evaluation of social learning configuration include in-depth interviews, focus group discussions and structured survey interviews. Our findings show that the social learning configuration used in this study leads to an increased problem ownership, an enhanced knowledge-base with regard to climate change impacts and production adaptation options, improved ability to see connections and interdependencies and finally, strengthened relationships and social cohesion. The results suggest that increased social learning in the community leads to increase in adaptive capacity of smallholder farmers and improves both their economic and environmental sustainability. We discuss the key lessons for designing learning configurations that can successfully enhance adaptive capacity and smallholder farmers’ agency and responsiveness to the challenges posed by climate change impacts.     

Machine learning and theoretical analysis release the non-linear relationship among ozone, secondary organic aerosol and volatile organic compounds

Fine particulate matter (PM_2.5) and ozone (O₃) pollutions are prevalent air quality issues in China. Volatile organic compounds (VOCs) have significant impact on the formation of O₃ and secondary organic aerosols (SOA) contributing PM_2.5. Herein, we investigated 54 VOCs, O₃ and SOA in Tianjin from June 2017 to May 2019 to explore the non-linear relationship among O₃, SOA and VOCs. The monthly patterns of VOCs and SOA concentrations were characterized by peak values during October to March and reached a minimum from April to September, but the observed O₃ was exactly the opposite. Machine learning methods resolved the importance of individual VOCs on O₃ and SOA that alkenes (mainly ethylene, propylene, and isoprene) have the highest importance to O₃ formation; alkanes (C_n, n ≥ 6) and aromatics were the main source of SOA formation. Machine learning methods revealed and emphasized the importance of photochemical consumptions of VOCs to O₃ and SOA formation. Ozone formation potential (OFP) and secondary organic aerosol formation potential (SOAFP) calculated by consumed VOCs quantitatively indicated that more than 80% of the consumed VOCs were alkenes which dominated the O₃ formation, and the importance of consumed aromatics and alkenes to SOAFP were 40.84% and 56.65%, respectively. Therein, isoprene contributed the most to OFP at 41.45% regardless of the season, while aromatics (58.27%) contributed the most to SOAFP in winter. Collectively, our findings can provide scientific evidence on policymaking for VOCs controls on seasonal scales to achieve effective reduction in both SOA and O₃.     

海洋环境耐蚀结构钢研究进展

以海洋装备主体材料钢铁作为研究对象,概述了结构钢在海洋环境下的腐蚀与防护研究进展。叙述了结构钢在海洋环境不同区域的各种因素影响下的腐蚀行为和机理。阐述了耐蚀结构钢材料的研究现状和国内外海洋环境用典型耐蚀低合金钢,目前主要通过微合金化技术和组织调控技术提高合金钢材料的耐蚀性。对比国外相关研究,我国耐蚀结构钢开发相对滞后,除了利用微合金化技术、组织调控技术这2种方法提高结构钢材料的耐蚀性外,采用传统腐蚀研究方法结合基于机器学习技术的新型腐蚀研究方法,挖掘海洋环境因素、材料成分和组织结构因素对结构钢腐蚀速率的影响规律,对加强结构钢腐蚀机理研究、加快新型耐蚀结构钢的开发、延长海洋装备的服役寿命具有重要意义。     

基于机器学习利用低成本传感器数据构建城市近地面PM2.5浓度的高分辨率空间分布

PM_2.5是当前影响中国城市空气质量的一种重要污染物,它对人体健康有不利影响,高时空间分辨率的PM_2.5浓度分布是城市大气污染 精细化防控和健康影响评估的重要基础.低成本传感器在中国城市空气质量监测领域得到越来越多的应用,但由于其数据质量不够高,其价值没有得到充分的挖掘.本研究以在长三角地区蚌埠市城区作为研究案例,首先基于国控站监测数据对低成本传感器观测数据进行校准.接着通过6种机器学习模型的预测性能的比较选取随机森林作为本研究的核心方法.然后利用随机森林方法结合土地利用类型、人口密度和通过WRF模式模拟得到的气象数据、监测数据等多源数据,构建了时间分辨率为1 h、空间分辨率为0.005°（接近500 m）的近地面PM_2.5浓度的时空分布.最后对几种方法的结果进行了比较和评估.反演得到的高分辨率PM_2.5浓度的10倍交叉验证结果显示,空间R²=0.95,RMSE=9.32 μg·m^-3,时间R²=0.88,RMSE=9.24 μg·m^-3.结果表明,低成本传感器观测数据可以帮助人们构建具有更高精度的高时空分辨率的PM_2.5浓度 分布,使得低成本传感器数据的时空特征价值得到更好的利用,同时也可为低成本传感器更好地服务于空气质量精细化管理提供一种新的 应用思路.