We propose a novel machine learning approach to reconstruct meshless surface wind speed fields,i.e.,to reconstruct the surface wind speed at any location,based on meteorological background fields and geographical info...We propose a novel machine learning approach to reconstruct meshless surface wind speed fields,i.e.,to reconstruct the surface wind speed at any location,based on meteorological background fields and geographical information.The random forest method is selected to develop the machine learning data reconstruction model(MLDRM-RF)for wind speeds over Beijing from 2015-19.We use temporal,geospatial attribute and meteorological background field features as inputs.The wind speed field can be reconstructed at any station in the region not used in the training process to cross-validate model performance.The evaluation considers the spatial distribution of and seasonal variations in the root mean squared error(RMSE)of the reconstructed wind speed field across Beijing.The average RMSE is 1.09 m s^(−1),considerably smaller than the result(1.29 m s^(−1))obtained with inverse distance weighting(IDW)interpolation.Finally,we extract the important feature permutations by the method of mean decrease in impurity(MDI)and discuss the reasonableness of the model prediction results.MLDRM-RF is a reasonable approach with excellent potential for the improved reconstruction of historical surface wind speed fields with arbitrary grid resolutions.Such a model is needed in many wind applications,such as wind energy and aviation safety assessments.展开更多
Historical forest fire risk databases are vital for evaluating the effectiveness of past forest management approaches,enhancing forest fire warnings and emergency response capabilities,and accurately budgeting potenti...Historical forest fire risk databases are vital for evaluating the effectiveness of past forest management approaches,enhancing forest fire warnings and emergency response capabilities,and accurately budgeting potential carbon emissions resulting from fires.However,due to the unavailability of spatial information technology,such databases are extremely difficult to build reliably and completely in the non-satellite era.This study presented an improved forest fire risk reconstruction framework that integrates a deep learning-based time series prediction model and spatial interpolation to address the challenge in Sichuan Province,southwestern China.First,the forest fire danger index(FFDI)was improved by supplementing slope and aspect information.We compared the performances of three time series models,namely,the autoregressive integrated moving average(ARIMA),Prophet and long short-term memory(LSTM)in predicting the modified forest fire danger index(MFFDI).The bestperforming model was used to retrace the MFFDI of individual stations from 1941 to 1970.Following this,the Anusplin spatial interpolation method was used to map the distributions of the MFFDI at five-year intervals,which were then subjected to weighted overlay with the distance-to-river layer to generate forest fire risk maps for reconstructing the forest fire danger database.The results revealed LSTM as the most accurate in fitting and predicting the historical MFFDI,with a fitting determination coefficient(R^2)of 0.709,mean square error(MSE)of0.047,and validation R^2 and MSE of 0.508 and 0.11,respectively.Independent validation of the predicted forest fire risk maps indicated that 5 out of 7 historical forest fire events were located in forest fire-prone areas,which is higher than the results determined from the original FFDI(2 out of 7).This proves the effectiveness of the improved MFFDI and indicates a high level of reliability of the historical forest fire risk reconstruction method proposed in this study.展开更多
High availability of wind power data is the basis for wind power research, but there are a large number of abnormal data in actual collected data, which seriously affects analysis of wind power law and reduces predict...High availability of wind power data is the basis for wind power research, but there are a large number of abnormal data in actual collected data, which seriously affects analysis of wind power law and reduces prediction accuracy. Measured power data of wind farm are analyzed, influence of wind speed fluctuation characteristics on wind power is discussed, and abnormal points are identified for data of different wind types. The Cluster-Based Local Outlier Factor (CLOF) algorithm based on K-means is used to identify outlier abnormal points, and conditional constraints based on physical background are used to identify accumulation abnormal points. Reconstructed data segment is divided according to fluctuation of wind speed. The Bidirectional Gate Recurrent Unit (BiGRU) model with wind speed as input reconstructs fluctuation segment data, and bi-directional weighted random forest model reconstructs stationary segment data. Based on analysis of measured data of a wind farm, results show the method can effectively identify various abnormal data, and complete high-quality reconstruction of data, thereby improving accuracy of wind power prediction.展开更多
Historical forest and grassland cover changes not only are critical indicators for quantifying ecological and environmental change processes but also serve as fundamental data for long-term climate change simulations ...Historical forest and grassland cover changes not only are critical indicators for quantifying ecological and environmental change processes but also serve as fundamental data for long-term climate change simulations and terrestrial ecosystem carbon emission assessments.However,because of limitations in historical data,quantitative estimations and spatially gridded reconstructions of these changes remain challenging,necessitating further methodological exploration.This study focused on China's present-day land area over the past millennium,objectively capturing the characteristics and drivers of forest and grassland cover changes.On this basis,using the forest transition theory and the space-for-time substitution method,we depicted the historical deforestation process as an inverted“S”curve and developed a model to reconstruct historical forest area changes based on the functional relationship between the forest area and population size dynamics.Subsequently,a gridded forest allocation model was established on the basis of deforestation tendencies.For the grassland cover,we implemented regionspecific methods,such as the cropland area deduction method and the habitat constraint method,to quantitatively reconstruct historical changes.Consequently,we obtained provincial forest and grassland area changes over the past millennium and mapped 10-km-resolution gridded data of forest and grassland cover.The results indicated the following.(1)The methods developed using population data as a proxy objectively reproduced the spatiotemporal evolution of forest and grassland cover in China over the past millennium.These feasible methods offer a novel pathway for the quantitative reconstruction of historical forest and grassland cover changes.(2)The data indicated that China's forest area generally decreased over the past millennium,characterized by a“decrease-then-increase”pattern.The forest area experienced three distinct phases:a slow decline(AD 1000–1650),a rapid decline(AD 1650–1960),and a gradual recovery(AD 1960–2000).The area decreased from 298 million hectares(Mha)in AD 1000 to 89 Mha in AD 1960 before increasing to 153 Mha in AD 2000.Spatially,deforestation began in the middlelower reaches of the Yellow River and gradually expanded to the middle-lower reaches of the Yangtze River,the southern coastal areas of China,southwest China,and northeast China,with the forest cover declining by 27%,40%,58%,55%,and 35%in these regions,respectively.(3)China's grassland area has shown a continuous decline over the past millennium with three phases:stable fluctuation(AD 1000–1600),slow decline(AD 1600–1900),and rapid decline(AD 1900–2000).The grassland area decreased from 305 Mha in AD 1000 to 277 Mha in AD 2000.Notably,zonal grassland areas in Northeast China,Inner Mongolia,Gan-Ning,Qinghai,Xinjiang,and Xizang decreased by 28 Mha over the millennium,whereas nonzonal secondary grassland areas in the hilly and mountainous areas of eastern and southern China increased by 0.3 Mha.展开更多
森林三维场景的模拟有助于在结构尺度上研究森林生态系统结构与生态系统多样性间的联系。激光雷达点云数据的立方体体素化是一种常用的森林三维重构方法,但存在易高估叶片面积等缺点。鉴于此,提出一种改进的薄片化旋转化体素方法,并与...森林三维场景的模拟有助于在结构尺度上研究森林生态系统结构与生态系统多样性间的联系。激光雷达点云数据的立方体体素化是一种常用的森林三维重构方法,但存在易高估叶片面积等缺点。鉴于此,提出一种改进的薄片化旋转化体素方法,并与传统的立方体体素化方法进行对比研究,探讨二者模拟森林三维结构的精度差异。研究使用虚拟和部分真实的森林场景数据,借助HELIOS++(Heidelberg LiDAR operations simulator)、VBRT(voxel-based radiative transfer)和PBRT(physically based ray tracer)模型,模拟机载激光雷达点云数据及多光谱影像。结果表明:与立方体体素化相比,薄片化及随机旋转后的树叶模型可显著改善树木冠层的真实感,且在点云数据高度分布、冠层覆盖度等核心指标上与基准数据表现出更高的相关度。研究提出的改进体素化方法优化了虚拟遥感数据的视觉效果和准确性,有助于森林精细化三维重构,在森林辐射传输研究方面具有良好的应用潜能。展开更多
High coverage of Pinus massoniana forest on low mountains in Eastern China at present was studied in this paper. This forest is threatened by plant diseases, especially pines wilt, and needs to be restored urgently. S...High coverage of Pinus massoniana forest on low mountains in Eastern China at present was studied in this paper. This forest is threatened by plant diseases, especially pines wilt, and needs to be restored urgently. Species of later successional stage or climax communities were retained or introduced to the forest through reconstruction according to vegetation ecology theory, so as to restore it quickly to zonal evergreen broad-leaved forest. It formed an evergreen broad-leaved sub-tree layer of 2～3m high dominated by Schima superba from a shrub layer of 57m high after 3 years of reconstruction. The questions of restoration were discussed in this paper.展开更多
基金supported by the Strategic Priority Research Program of the Chinese Academy of Sciences(Grant No.XDA19030402)the Key Special Projects for International Cooperation in Science and Technology Innovation between Governments(Grant No.2017YFE0133600the Beijing Municipal Natural Science Foundation Youth Project 8214066:Application Research of Beijing Road Visibility Prediction Based on Machine Learning Methods.
文摘We propose a novel machine learning approach to reconstruct meshless surface wind speed fields,i.e.,to reconstruct the surface wind speed at any location,based on meteorological background fields and geographical information.The random forest method is selected to develop the machine learning data reconstruction model(MLDRM-RF)for wind speeds over Beijing from 2015-19.We use temporal,geospatial attribute and meteorological background field features as inputs.The wind speed field can be reconstructed at any station in the region not used in the training process to cross-validate model performance.The evaluation considers the spatial distribution of and seasonal variations in the root mean squared error(RMSE)of the reconstructed wind speed field across Beijing.The average RMSE is 1.09 m s^(−1),considerably smaller than the result(1.29 m s^(−1))obtained with inverse distance weighting(IDW)interpolation.Finally,we extract the important feature permutations by the method of mean decrease in impurity(MDI)and discuss the reasonableness of the model prediction results.MLDRM-RF is a reasonable approach with excellent potential for the improved reconstruction of historical surface wind speed fields with arbitrary grid resolutions.Such a model is needed in many wind applications,such as wind energy and aviation safety assessments.
基金the following grants:The National Key R&D Program of China(2019YFA0606600)the Natural Science Foundation of China(31971577)the Priority Academic Program Development of Jiangsu Higher Education Institutions(PAPD)。
文摘Historical forest fire risk databases are vital for evaluating the effectiveness of past forest management approaches,enhancing forest fire warnings and emergency response capabilities,and accurately budgeting potential carbon emissions resulting from fires.However,due to the unavailability of spatial information technology,such databases are extremely difficult to build reliably and completely in the non-satellite era.This study presented an improved forest fire risk reconstruction framework that integrates a deep learning-based time series prediction model and spatial interpolation to address the challenge in Sichuan Province,southwestern China.First,the forest fire danger index(FFDI)was improved by supplementing slope and aspect information.We compared the performances of three time series models,namely,the autoregressive integrated moving average(ARIMA),Prophet and long short-term memory(LSTM)in predicting the modified forest fire danger index(MFFDI).The bestperforming model was used to retrace the MFFDI of individual stations from 1941 to 1970.Following this,the Anusplin spatial interpolation method was used to map the distributions of the MFFDI at five-year intervals,which were then subjected to weighted overlay with the distance-to-river layer to generate forest fire risk maps for reconstructing the forest fire danger database.The results revealed LSTM as the most accurate in fitting and predicting the historical MFFDI,with a fitting determination coefficient(R^2)of 0.709,mean square error(MSE)of0.047,and validation R^2 and MSE of 0.508 and 0.11,respectively.Independent validation of the predicted forest fire risk maps indicated that 5 out of 7 historical forest fire events were located in forest fire-prone areas,which is higher than the results determined from the original FFDI(2 out of 7).This proves the effectiveness of the improved MFFDI and indicates a high level of reliability of the historical forest fire risk reconstruction method proposed in this study.
基金supported by the National Key R&D Program of China“Technology and application of wind power/photovoltaic power prediction for promoting renewable energy consumption”(2018YFB0904200).
文摘High availability of wind power data is the basis for wind power research, but there are a large number of abnormal data in actual collected data, which seriously affects analysis of wind power law and reduces prediction accuracy. Measured power data of wind farm are analyzed, influence of wind speed fluctuation characteristics on wind power is discussed, and abnormal points are identified for data of different wind types. The Cluster-Based Local Outlier Factor (CLOF) algorithm based on K-means is used to identify outlier abnormal points, and conditional constraints based on physical background are used to identify accumulation abnormal points. Reconstructed data segment is divided according to fluctuation of wind speed. The Bidirectional Gate Recurrent Unit (BiGRU) model with wind speed as input reconstructs fluctuation segment data, and bi-directional weighted random forest model reconstructs stationary segment data. Based on analysis of measured data of a wind farm, results show the method can effectively identify various abnormal data, and complete high-quality reconstruction of data, thereby improving accuracy of wind power prediction.
基金supported by the National Key Research and Development Program of China(Grant No.2017YFA0603304)the National Natural Science Foundation of China(Grant No.42201263)。
文摘Historical forest and grassland cover changes not only are critical indicators for quantifying ecological and environmental change processes but also serve as fundamental data for long-term climate change simulations and terrestrial ecosystem carbon emission assessments.However,because of limitations in historical data,quantitative estimations and spatially gridded reconstructions of these changes remain challenging,necessitating further methodological exploration.This study focused on China's present-day land area over the past millennium,objectively capturing the characteristics and drivers of forest and grassland cover changes.On this basis,using the forest transition theory and the space-for-time substitution method,we depicted the historical deforestation process as an inverted“S”curve and developed a model to reconstruct historical forest area changes based on the functional relationship between the forest area and population size dynamics.Subsequently,a gridded forest allocation model was established on the basis of deforestation tendencies.For the grassland cover,we implemented regionspecific methods,such as the cropland area deduction method and the habitat constraint method,to quantitatively reconstruct historical changes.Consequently,we obtained provincial forest and grassland area changes over the past millennium and mapped 10-km-resolution gridded data of forest and grassland cover.The results indicated the following.(1)The methods developed using population data as a proxy objectively reproduced the spatiotemporal evolution of forest and grassland cover in China over the past millennium.These feasible methods offer a novel pathway for the quantitative reconstruction of historical forest and grassland cover changes.(2)The data indicated that China's forest area generally decreased over the past millennium,characterized by a“decrease-then-increase”pattern.The forest area experienced three distinct phases:a slow decline(AD 1000–1650),a rapid decline(AD 1650–1960),and a gradual recovery(AD 1960–2000).The area decreased from 298 million hectares(Mha)in AD 1000 to 89 Mha in AD 1960 before increasing to 153 Mha in AD 2000.Spatially,deforestation began in the middlelower reaches of the Yellow River and gradually expanded to the middle-lower reaches of the Yangtze River,the southern coastal areas of China,southwest China,and northeast China,with the forest cover declining by 27%,40%,58%,55%,and 35%in these regions,respectively.(3)China's grassland area has shown a continuous decline over the past millennium with three phases:stable fluctuation(AD 1000–1600),slow decline(AD 1600–1900),and rapid decline(AD 1900–2000).The grassland area decreased from 305 Mha in AD 1000 to 277 Mha in AD 2000.Notably,zonal grassland areas in Northeast China,Inner Mongolia,Gan-Ning,Qinghai,Xinjiang,and Xizang decreased by 28 Mha over the millennium,whereas nonzonal secondary grassland areas in the hilly and mountainous areas of eastern and southern China increased by 0.3 Mha.
文摘森林三维场景的模拟有助于在结构尺度上研究森林生态系统结构与生态系统多样性间的联系。激光雷达点云数据的立方体体素化是一种常用的森林三维重构方法,但存在易高估叶片面积等缺点。鉴于此,提出一种改进的薄片化旋转化体素方法,并与传统的立方体体素化方法进行对比研究,探讨二者模拟森林三维结构的精度差异。研究使用虚拟和部分真实的森林场景数据,借助HELIOS++(Heidelberg LiDAR operations simulator)、VBRT(voxel-based radiative transfer)和PBRT(physically based ray tracer)模型,模拟机载激光雷达点云数据及多光谱影像。结果表明:与立方体体素化相比,薄片化及随机旋转后的树叶模型可显著改善树木冠层的真实感,且在点云数据高度分布、冠层覆盖度等核心指标上与基准数据表现出更高的相关度。研究提出的改进体素化方法优化了虚拟遥感数据的视觉效果和准确性,有助于森林精细化三维重构,在森林辐射传输研究方面具有良好的应用潜能。
文摘High coverage of Pinus massoniana forest on low mountains in Eastern China at present was studied in this paper. This forest is threatened by plant diseases, especially pines wilt, and needs to be restored urgently. Species of later successional stage or climax communities were retained or introduced to the forest through reconstruction according to vegetation ecology theory, so as to restore it quickly to zonal evergreen broad-leaved forest. It formed an evergreen broad-leaved sub-tree layer of 2～3m high dominated by Schima superba from a shrub layer of 57m high after 3 years of reconstruction. The questions of restoration were discussed in this paper.
