A deep-learning-based method,called ConvLSTMP3,is developed to predict the sea surface heights(SSHs).ConvLSTMP3 is data-driven by treating the SSH prediction problem as the one of extracting the spatial-temporal featu...A deep-learning-based method,called ConvLSTMP3,is developed to predict the sea surface heights(SSHs).ConvLSTMP3 is data-driven by treating the SSH prediction problem as the one of extracting the spatial-temporal features of SSHs,in which the spatial features are“learned”by convolutional operations while the temporal features are tracked by long short term memory(LSTM).Trained by a reanalysis dataset of the South China Sea(SCS),ConvLSTMP3 is applied to the SSH prediction in a region of the SCS east off Vietnam coast featured with eddied and offshore currents in summer.Experimental results show that ConvLSTMP3 achieves a good prediction skill with a mean RMSE of 0.057 m and accuracy of 93.4%averaged over a 15-d prediction period.In particular,ConvLSTMP3 shows a better performance in predicting the temporal evolution of mesoscale eddies in the region than a full-dynamics ocean model.Given the much less computation in the prediction required by ConvLSTMP3,our study suggests that the deep learning technique is very useful and effective in the SSH prediction,and could be an alternative way in the operational prediction for ocean environments in the future.展开更多
The dimensions of attractors and predictability are estimated from phase space trajectories of observed 500 hPa height over the Northern Hemisphere. As a first estimate the dimensions of attractors are about 11.5 and ...The dimensions of attractors and predictability are estimated from phase space trajectories of observed 500 hPa height over the Northern Hemisphere. As a first estimate the dimensions of attractors are about 11.5 and the doubling time of the initial error is 6 to 7 days for original data. But the former is shorter and the latter is longer for low frequency data set.To verify if the predictability estimated by this method and by general circulation model is identical, the doubling time of the initial error of a model data set by both methods is estimated. It is shown that the predictability obtained from phase space trajectories is overestimated to sufficient small initial error. But it is underestimated to the time being equal to the climatological RMS error.展开更多
This study evaluated the total height of trees based on diameter at breast height by using 23 widely used height-diameter non-linear regression models for mixed-species forest stands consisting of Caucasian oak,field ...This study evaluated the total height of trees based on diameter at breast height by using 23 widely used height-diameter non-linear regression models for mixed-species forest stands consisting of Caucasian oak,field maple,and hornbeam from forests in Northwest Iran.1920 trees were measured in 6 sampling plots(every sampling plot has 1 ha area).The fit of the best height–diameter models for each species were compared based on R2,Root Mean Square Error(RMSE),Akaike information criterion(AIC),standard error,and relative ranking performance criteria.In the final step,verification of results was performed by paired sample t-test to compare the observed height and estimated height.Results showed that among 23 height-diameter models,the best models were obtained from the top five ones including Modified-logistic,Prodan,Sibbesen,Burkhart,and Exponential.Comparison between the actual observed height and estimated height for Caucasian oak showed that Modified–Logistic,Prodan,Sibbesen,Burkhart,and Exponential performed better than the others,respectively(There were no statistically significant differences between observed heights and predicted height(p≥0.05)).Prodan,Modified-Logistic,Burkhart,and Loetch evaluated field maple tree height correctly,and Modified-Logistic,Burkhart,and Loetch had better fitness compared to the others for hornbeam,respectively.Although other models were introduced as appropriate criteria,they could not reliably predict the height of trees.Using the Rank analysis,the Modified-Logistic model for the Caucasian oak and Prodan model for field maple and hornbeam had the best performance.Finally,to complement the results of this study,it is suggested to assess how environmental factors such as elevation,climate parameters,forest protection policy and forest structure will modify height-diameter allometry models and will enhance the prediction accuracy of tree heights prediction in mixed stands.展开更多
基金The National Key Research and Development Program under contract Nos 2018YFC1406204 and 2018YFC1406201the Guangdong Special Support Program under contract No.2019BT2H594+5 种基金the Taishan Scholar Foundation under contract No.tsqn201812029the National Natural Science Foundation of China under contract Nos U1811464,61572522,61572523,61672033,61672248,61873280,41676016 and 41776028the Natural Science Foundation of Shandong Province under contract Nos ZR2019MF012 and 2019GGX101067the Fundamental Research Funds of Central Universities under contract Nos 18CX02152A and 19CX05003A-6the fund of the Shandong Province Innovation Researching Group under contract No.2019KJN014the Key Special Project for Introduced Talents Team of the Southern Marine Science and Engineering Guangdong Laboratory(Guangzhou)under contract No.GML2019ZD0303.
文摘A deep-learning-based method,called ConvLSTMP3,is developed to predict the sea surface heights(SSHs).ConvLSTMP3 is data-driven by treating the SSH prediction problem as the one of extracting the spatial-temporal features of SSHs,in which the spatial features are“learned”by convolutional operations while the temporal features are tracked by long short term memory(LSTM).Trained by a reanalysis dataset of the South China Sea(SCS),ConvLSTMP3 is applied to the SSH prediction in a region of the SCS east off Vietnam coast featured with eddied and offshore currents in summer.Experimental results show that ConvLSTMP3 achieves a good prediction skill with a mean RMSE of 0.057 m and accuracy of 93.4%averaged over a 15-d prediction period.In particular,ConvLSTMP3 shows a better performance in predicting the temporal evolution of mesoscale eddies in the region than a full-dynamics ocean model.Given the much less computation in the prediction required by ConvLSTMP3,our study suggests that the deep learning technique is very useful and effective in the SSH prediction,and could be an alternative way in the operational prediction for ocean environments in the future.
文摘The dimensions of attractors and predictability are estimated from phase space trajectories of observed 500 hPa height over the Northern Hemisphere. As a first estimate the dimensions of attractors are about 11.5 and the doubling time of the initial error is 6 to 7 days for original data. But the former is shorter and the latter is longer for low frequency data set.To verify if the predictability estimated by this method and by general circulation model is identical, the doubling time of the initial error of a model data set by both methods is estimated. It is shown that the predictability obtained from phase space trajectories is overestimated to sufficient small initial error. But it is underestimated to the time being equal to the climatological RMS error.
文摘This study evaluated the total height of trees based on diameter at breast height by using 23 widely used height-diameter non-linear regression models for mixed-species forest stands consisting of Caucasian oak,field maple,and hornbeam from forests in Northwest Iran.1920 trees were measured in 6 sampling plots(every sampling plot has 1 ha area).The fit of the best height–diameter models for each species were compared based on R2,Root Mean Square Error(RMSE),Akaike information criterion(AIC),standard error,and relative ranking performance criteria.In the final step,verification of results was performed by paired sample t-test to compare the observed height and estimated height.Results showed that among 23 height-diameter models,the best models were obtained from the top five ones including Modified-logistic,Prodan,Sibbesen,Burkhart,and Exponential.Comparison between the actual observed height and estimated height for Caucasian oak showed that Modified–Logistic,Prodan,Sibbesen,Burkhart,and Exponential performed better than the others,respectively(There were no statistically significant differences between observed heights and predicted height(p≥0.05)).Prodan,Modified-Logistic,Burkhart,and Loetch evaluated field maple tree height correctly,and Modified-Logistic,Burkhart,and Loetch had better fitness compared to the others for hornbeam,respectively.Although other models were introduced as appropriate criteria,they could not reliably predict the height of trees.Using the Rank analysis,the Modified-Logistic model for the Caucasian oak and Prodan model for field maple and hornbeam had the best performance.Finally,to complement the results of this study,it is suggested to assess how environmental factors such as elevation,climate parameters,forest protection policy and forest structure will modify height-diameter allometry models and will enhance the prediction accuracy of tree heights prediction in mixed stands.
