基于空间回归模型的中国碳排放空间差异模拟

以中国碳排放数据库的多年分省碳排放数据为基础,选取植被指数NDVI、夜间灯光影像DMSP-OLS、生产总值、工业总产值和人口数5项指标,采用1 km×1 km格网,运用空间滞后回归模型对中国2000年和2013年的碳排放进行时空分布模拟,并对空间模拟结果进行误差分析及纠正,获得地均碳排放强度的空间分布图,以弥补传统碳排放统计空间展示效果差、空间分辨率低等缺陷.结果发现,中国碳排放的空间分布格局差异显著,地均碳排放强度分布呈现由东部地区向中西部地区逐渐递减的阶梯状区域分布差异,东部地区的碳排放高值聚集区域大致呈现面状聚集分布特征,区域之间碳排放的差异较小,西部省会城市、工矿城市与农村地区碳排放量差异显著.2013年的地均碳排放强度明显高于2000年,且高值区域的分布范围明显扩大,低值区域分布范围面积明显减少.     

三峡水库蓄水对消落带土壤Cu、Zn、Cr、Cd含量的影响

为揭示水位涨落形成系列淹水梯度对消落带土壤重金属含量的影响,论文采集三峡水库忠县消落带土壤,分析了不同淹水深度土壤Cu、Zn、Cr、Cd含量,并探讨了这些重金属元素物质来源及其与三峡水库水位涨落关系。结果表明：1）不同淹水深度下消落带土壤重金属含量差异较大,长期淹水土壤Cu、Zn、Cd含量远远高于短期淹水和未淹水土壤,Cr则在不同淹水深度略有变化;2）除Cr相对独立外,不同淹水深度下土壤SOM、pH与土壤Cu、Zn、Cd含量密切相关,但pH、SOM间仅有微弱的相关关系;3）因子分析中因子1和因子2对Cu、Zn、Cr、Cd的累计贡献率为93.47%,Cu、Zn、Cd在第一轴载荷值为0.962、0.967、0.925,Cr在第二轴载荷值为0.998。研究发现：1）重金属输入来源和流向异质性是决定不同淹水深度Cu、Zn、Cr、Cd含量差异的重要原因,长期淹水显著增加了土壤Cu、Zn、Cd含量,但对Cr效应并不明显;2）区域土壤内Cr主要来源为自然源的土壤侵蚀和母岩风化,Cu、Zn、Cd来源于人为输入的库岸输入和长江上游来水悬浮颗粒沉积,应引起足够重视。     

基于BP神经网络和SVR的Fund?o尾矿坝排水数据预测对比研究

针对尾矿库运行过程中安全预警问题,选取2015年巴西Samarco铁矿溃坝事故案例,研究BP神经网络和SVR方法在排水数据预测的适用性。综合分析了排水数据的复杂且非线性的特点,以库水位、降雨量和干滩长度为输入特征,采用上述2个模型对尾矿坝排水数据进行预测。研究结果表明:基于BP神经网络预测结果的最大相对误差不高于4.35%;基于SVR算法的最大相对误差不高于9.21%;Fundo坝的排水预测结果是可行的,BP神经网络的预测精度更高,而SVR模型的运算速度更快。研究结果可为矿山安全工作的快速响应和溃坝预警提供信息支撑和参考依据。     

基于地理加权回归的NO2排放预测模型

为了探究影响NO2排放的交通相关因素,以交通分析小区为基本单位,提取了美国洛杉矶城市的人口特征、道路网络特征、交通状况等数据,采用地理加权回归模型(GWR)分析各交通相关因素对NO2排放的影响,从而建立交通小区NO2排放预测模型。结果表明,交通小区路网密度、交通小区机动车吸引量、通勤时间在30~60 min的工作人数与NO2的排放呈正相关,表明3种影响因素的增加会造成NO2排放的增加;而交通小区在家工作的人数、慢行交通(步行、自行车等绿色出行方式)吸引量与NO2的排放呈负相关,表明适当鼓励在家工作的新型办公方式、鼓励居民出行选择慢行交通,能有效减少交通小区NO2的排放。     

醇酚类化合物毒性的QSAR研究

化合物毒性与描述符通常呈现为非线性关系,量子化学计算的化合物分子描述符中包含诸多无关特征与冗余特征.最大相关最小冗余(m RMR)是应用较广泛的特征选择方法,但当前的m RMR对连续型因变量不适用,且存在相关性测度与冗余性测度不可比的缺陷.定量构效关系(QSAR)研究中因变量(毒性)与自变量(描述符)多为连续型变量,本文以非线性的距离相关系数(d Cor)取代线性的Pearson相关系数(R),在非线性条件下实现了相关性测度与冗余性测度可比,由此提出了新的特征选择方法 m RMR-d Cor.3个醇酚类化合物毒性QSAR数据集的分析表明,基于m RMR-d Cor选择特征的支持向量回归(SVR)模型独立预测Q2分别为0.954、0.941、0.981,明显优于参比模型与文献报道,m RMR-d Cor选择的多数保留分子描述符得到文献报道支持.m RMR-d Cor在化合物QSAR、定量构质关系等研究中有广泛应用前景.     

Analysis of risk factors affecting driver injury and crash injury with drivers under the influence of alcohol (DUI) and non-DUI

Objective: The objective of this research was to study risk factors that significantly influence the severity of crashes for drivers both under and not under the influence of alcohol.

Methods: Ordinal logistic regression was applied to analyze a crash data set involving drivers under and not under the influence of alcohol in China from January 2011 to December 2014.

Results: Four risk factors were found to be significantly associated with the severity of driver injury, including crash partner and intersection type. Age group was found to be significantly associated with the severity of crashes involving drivers under the influence of alcohol. Crash partner, intersection type, lighting conditions, gender, and time of day were found to be significantly associated with severe driver injuries, the last of which was also significantly associated with severe crashes involving drivers not under the influence of alcohol.

Conclusions: This study found that pedestrian involvement decreases the odds of severe driver injury when a driver is under the influence of alcohol, with a relative risk of 0.05 compared to the vehicle-to-vehicle group. The odds of severe driver injury at T-intersections were higher than those for traveling along straight roads. Age was shown to be an important factor, with drivers 50–60 years of age having higher odds of being involved in severe crashes compared to 20- to 30-year-olds when the driver was under the influence of alcohol.

When the driver was not under the influence of alcohol, drivers suffered more severe injuries between midnight and early morning compared to early nighttime. The vehicle-to-motorcycle and vehicle-to-pedestrian groups experienced less severe driver injuries, and vehicle collisions with fixed objects exhibited higher odds of severe driver injury than did vehicle-to-vehicle impacts. The odds of severe driver injury at cross intersections were 0.29 compared to travel along straight roads. The odds of severe driver injury when street lighting was not available at night were 3.20 compared to daylight. The study indicated that female drivers are more likely to experience severe injury than male drivers when not under the influence of alcohol. Crashes between midnight and early morning exhibited higher odds of severe injury compared to those occurring at other times of day.

The identification of risk factors and a discussion on the odds ratio between levels of the impact of the driver injury and crash severity may benefit road safety stakeholders when developing initiatives to reduce the severity of crashes.     

Modeling freshwater fish distributions using multiscale landscape data: A case study of six narrow range endemics

Species distribution models (SDMs) have become integral tools in scientific research and conservation planning. Despite progress in the assessment of various statistical models for use in SDMs, little has been done in way of evaluating appropriate ecological models. In this paper, we evaluate the multiscale filter framework as a suitable theoretical model for predicting freshwater fish distributions in the upper Green River system (Ohio River drainage), USA. The spatial distributions of six fishes with contrasting biogeographies were modeled using boosted regression trees and multiscale landscape data. Species biogeography did not appear to affect predictive performance and all models performed well statistically with receiver operating characteristic area under the curve (AUC) ranging from 0.87 to 0.98. Predictive maps show accurate estimations of ranges for five of six species based on historical collections. The relative influence of each type of environmental feature and spatial scale varied markedly with between species. A hierarchical effect was detected for narrowly distributed species. These species were highly influenced by soil composition at larger spatial scales and land use/land cover (LULC) patterns at more proximal scales. Conversely, LULC pattern was the most influential feature for widely distributed at all spatial scales. Using multiscale data capable of capturing hierarchical landscape influences allowed production of accurate predictive models and provided further insight into factors controlling freshwater fish distributions.     

Scaling down the “Netherlands Fallacy”: a local-level quantitative study of the effect of affluence on the carbon footprint across the United States

Much quantitative research examining the determinants of the ecological footprint has been conducted cross-nationally, where data on cross-boundary flows have been readily available. While local-level studies of the footprint do exist for specific localities, most quantitative research at this scale has examined direct environmental impacts attributed to the internal activities of the locality, for instance, carbon emissions. Our analysis builds on this previous work by exploiting a local-level carbon footprint dataset with coverage for 28,321 zip codes across the United States. Following prior research, we focus on the effect of local affluence, measured in terms of median household income. In spatial regression models, we regress the per capita carbon footprint on local affluence, controlling for a variety of other factors. Consistent with previous work, we find that affluence is positively correlated the carbon footprint and there is no evidence of an environmental Kuznets curve. In the conclusion, we review the results of the study and discuss their implications for policy, specifically in terms of cross-boundary environmental problems.     

土地质量对农业劳动生产率的影响——来自六省县级数据的经验证据

本文利用国土资源部农用地分等定级与估价项目的土地自然质量指数数据,结合2004年的县级统计数据,建立改造的C-D生产函数,通过多元回归分析定量考察了土地质量对农业劳动生产率的影响。最终得到如下研究结果:对所用的县,土地质量指数每增加1%,农业劳动生产率增加0.336%;分地区的结果显示,土地质量和化肥投入在东部地区为替代关系,在西部地区为互补关系。研究结论:土地质量在很大程度上影响着农业劳动生产率,应加强对土地质量的保护和改善。     

A distributional analysis of the socio-ecological and economic determinants of forest carbon stocks

Forest carbon (C) sequestration is being actively considered by several states as a way to cost-effectively comply with the forthcoming United States (US) Environmental Protection Agency’s rule that will reduce power plant C emissions by 32% of 2005 levels by 2030. However, little is known about the socio-ecological and distributional effects of such a policy. Given that C is heterogeneous across the landscape, understanding how social, economic, and ecological changes affect forest C stocks and sequestration is key for developing forest management policies that offset C emissions. Using Florida US as a case study, we use US National Forest Inventory Analysis and Census Bureau data in both linear regression and quantile regression analyses to examine the socio-ecological and economic determinants of forest C stocks and its relationship with differing communities. Quantile regression findings demonstrate nonlinearity in the effects of key determinants, which highlight the limitations of regularly used mean-based regression analyses. We also found that forest basal area, site quality, stand size, and stand age are significant ecological predictors of carbon stocks, with a positive and increasing effect on upper quantiles where C stocks are greater. The effect of education was generally positive and mostly significant at upper quantiles, while the effects of income and locations with predominantly minority residents (as compared to whites) were negative. Upper quantiles were also affected by population age. Our findings underscore the importance of considering the broader socio-ecological and economic implications of compliance strategies that target the management of forests for carbon sequestration and other ecosystem services.