秸秆还田与氮肥施用是农田生态系统中碳氮元素的两大主要补给途径,其在调控稻田甲烷(CH_(4))和氧化亚氮(N_(2)O)排放以及水稻产量方面具有重要作用。以往的研究主要关注秸秆还田或氮肥施用单因素对稻田温室气体排放的影响,而双因素互作...秸秆还田与氮肥施用是农田生态系统中碳氮元素的两大主要补给途径,其在调控稻田甲烷(CH_(4))和氧化亚氮(N_(2)O)排放以及水稻产量方面具有重要作用。以往的研究主要关注秸秆还田或氮肥施用单因素对稻田温室气体排放的影响,而双因素互作对甲烷和氧化亚氮排放的影响尚未明确。同时,在秸秆还田条件下如何进行合理的氮肥施用鲜有深入研究。本研究基于3个氮肥处理(0、180、360 kg N/hm^(2))和3个秸秆还田处理(0、2.25、3.75 t/hm^(2))进行多年水稻田间定位试验,研究结果表明:CH_(4)季节累积排放随秸秆还田量增加而增加,与施氮量无显著正相关关系;N_(2)O季节累积排放随施氮量增加而增加,与秸秆还田量无显著正相关关系;秸秆还田对于产量的影响具有不确定性,两年均在秸秆不还田+不施氮处理(S0N0)出现最低产量,2021与2022年最低产量分别为5740.64和4903.75 kg/hm^(2)。2021与2022年最高产量分别在秸秆不还田+高氮(S0N_(2))和高量秸秆还田+高氮(S2N_(2))出现,分别为10938.48和10384.83 kg/hm^(2)。同时,本研究发现在低量秸秆还田条件下,在碳足迹(CF, Carbon Footprint)方面,施氮量为251 kg N/hm^(2)时碳足迹达到最低点,为1.01 kg C/kg;而在生态经济净收益(NEEB, Net Ecosystem Economic Benefits)方面,施氮量为294 kg N/hm^(2)时生态经济净收益达到最高点,为11778.15元/hm^(2)。为协同生态经济净收益与碳排放,在低量秸秆还田(S1)下,配合251—294 kg N/hm^(2)的施氮量为最优施肥方案。研究结果为指导稻田温室气体减排、实现稻田碳中和以及农田管理提供了理论支撑,为实现水稻的高产稳产与低碳生产科学依据。展开更多
针对遥感建筑物图像中建筑物大小不一、边缘模糊导致精度不高的问题,提出一种双分支并行融合注意力机制的网络模型TC-UNet++。针对卷积神经网络擅长提取局部特征,难以捕获全局信息的特点,引入Transformer结构以解决全局信息丢失的问题...针对遥感建筑物图像中建筑物大小不一、边缘模糊导致精度不高的问题,提出一种双分支并行融合注意力机制的网络模型TC-UNet++。针对卷积神经网络擅长提取局部特征,难以捕获全局信息的特点,引入Transformer结构以解决全局信息丢失的问题。对于两种结构的特征维度和通道数不匹配的问题,设计一种TC(Transformer to CNN)模块以交互的方式融合不同分辨率下局部与全局特征。引入坐标注意力机制,根据像素在图像中的位置信息,定位和识别建筑物。实验结果表明,TC-UNet++在WHU数据集上交互比、准确率、总精度分别达到了93.1%、95.9%、98.8%,在不显著增加参数的情况下,展现出良好的有效性。展开更多
From December 3 to 12,2017,Zhangzhou experienced a haze weather process lasting for 10 days.In this paper,based on surface meteorological observation and environmental monitoring data,statistical comparison,Eta correl...From December 3 to 12,2017,Zhangzhou experienced a haze weather process lasting for 10 days.In this paper,based on surface meteorological observation and environmental monitoring data,statistical comparison,Eta correlation coefficient,clustering analysis and other methods were adopted to analyze the causes of haze weather in the whole process from clean air (December 2) to haze generation,maintenance (December 3-12) and dissipation (December 13).The results showed that two blasts of strong cold air invaded the city during the haze formation and maintenance,and regional pollution transport and static and stable weather were the main reasons for the continuous high concentration of particulate matter,the occurrence of mild pollution from PM 10 and PM 2.5 in some periods and mild-heavy haze weather lasting for 10 days.In this process,PM concentration had a strong negative correlation with air temperature and visibility,and the absolute values of the correlation coefficients were between 0.6 and 0.8.PM concentration correlated with relative humidity positively and with wind speed weakly.Regional transport of particulate matter and ozone was conducted in the form of northeast wind and southeast wind backflow.At night and in the morning,under the influence of static and stable weather and long-term sea and land breeze (6 d),average wind speed was less than 1.5 m/s,and the frequency of static wind and small wind exceeded 50%,while there was no precipitation.At 08:00,in the middle and low altitudes,the inversion layer appeared in one week,and cloud cover was large,while the height of mixed layer was below 600 m.In the whole process,it was in the weak divergence field under cold high pressure,and the overall diffusion,dilution and cleaning ability of the atmosphere were poor.展开更多
文摘秸秆还田与氮肥施用是农田生态系统中碳氮元素的两大主要补给途径,其在调控稻田甲烷(CH_(4))和氧化亚氮(N_(2)O)排放以及水稻产量方面具有重要作用。以往的研究主要关注秸秆还田或氮肥施用单因素对稻田温室气体排放的影响,而双因素互作对甲烷和氧化亚氮排放的影响尚未明确。同时,在秸秆还田条件下如何进行合理的氮肥施用鲜有深入研究。本研究基于3个氮肥处理(0、180、360 kg N/hm^(2))和3个秸秆还田处理(0、2.25、3.75 t/hm^(2))进行多年水稻田间定位试验,研究结果表明:CH_(4)季节累积排放随秸秆还田量增加而增加,与施氮量无显著正相关关系;N_(2)O季节累积排放随施氮量增加而增加,与秸秆还田量无显著正相关关系;秸秆还田对于产量的影响具有不确定性,两年均在秸秆不还田+不施氮处理(S0N0)出现最低产量,2021与2022年最低产量分别为5740.64和4903.75 kg/hm^(2)。2021与2022年最高产量分别在秸秆不还田+高氮(S0N_(2))和高量秸秆还田+高氮(S2N_(2))出现,分别为10938.48和10384.83 kg/hm^(2)。同时,本研究发现在低量秸秆还田条件下,在碳足迹(CF, Carbon Footprint)方面,施氮量为251 kg N/hm^(2)时碳足迹达到最低点,为1.01 kg C/kg;而在生态经济净收益(NEEB, Net Ecosystem Economic Benefits)方面,施氮量为294 kg N/hm^(2)时生态经济净收益达到最高点,为11778.15元/hm^(2)。为协同生态经济净收益与碳排放,在低量秸秆还田(S1)下,配合251—294 kg N/hm^(2)的施氮量为最优施肥方案。研究结果为指导稻田温室气体减排、实现稻田碳中和以及农田管理提供了理论支撑,为实现水稻的高产稳产与低碳生产科学依据。
文摘针对遥感建筑物图像中建筑物大小不一、边缘模糊导致精度不高的问题,提出一种双分支并行融合注意力机制的网络模型TC-UNet++。针对卷积神经网络擅长提取局部特征,难以捕获全局信息的特点,引入Transformer结构以解决全局信息丢失的问题。对于两种结构的特征维度和通道数不匹配的问题,设计一种TC(Transformer to CNN)模块以交互的方式融合不同分辨率下局部与全局特征。引入坐标注意力机制,根据像素在图像中的位置信息,定位和识别建筑物。实验结果表明,TC-UNet++在WHU数据集上交互比、准确率、总精度分别达到了93.1%、95.9%、98.8%,在不显著增加参数的情况下,展现出良好的有效性。
基金Supported by National Natural Science Foundation of China(U14052235)National Natural Science Foundation of Zhangzhou(ZZ2018J03)Shengzhou Environmental Protection Bureau([350600]ZY[GK]2018003-1)
文摘From December 3 to 12,2017,Zhangzhou experienced a haze weather process lasting for 10 days.In this paper,based on surface meteorological observation and environmental monitoring data,statistical comparison,Eta correlation coefficient,clustering analysis and other methods were adopted to analyze the causes of haze weather in the whole process from clean air (December 2) to haze generation,maintenance (December 3-12) and dissipation (December 13).The results showed that two blasts of strong cold air invaded the city during the haze formation and maintenance,and regional pollution transport and static and stable weather were the main reasons for the continuous high concentration of particulate matter,the occurrence of mild pollution from PM 10 and PM 2.5 in some periods and mild-heavy haze weather lasting for 10 days.In this process,PM concentration had a strong negative correlation with air temperature and visibility,and the absolute values of the correlation coefficients were between 0.6 and 0.8.PM concentration correlated with relative humidity positively and with wind speed weakly.Regional transport of particulate matter and ozone was conducted in the form of northeast wind and southeast wind backflow.At night and in the morning,under the influence of static and stable weather and long-term sea and land breeze (6 d),average wind speed was less than 1.5 m/s,and the frequency of static wind and small wind exceeded 50%,while there was no precipitation.At 08:00,in the middle and low altitudes,the inversion layer appeared in one week,and cloud cover was large,while the height of mixed layer was below 600 m.In the whole process,it was in the weak divergence field under cold high pressure,and the overall diffusion,dilution and cleaning ability of the atmosphere were poor.
