Artificial Intelligence

  • Display Type:
  • Text List
  • Abstract List
  • 1  A brief overview of the application of artificial intelligence to climate prediction
    YANG Shuxian LING Fenghua YING Wushan YANG song LUO Jingjia
    2022, 45(5):641-659. DOI: 10.13878/j.cnki.dqkxxb.20210623003
    [Abstract](1966) [HTML](3394) [PDF 29.15 M](3607)
    Abstract:
    In recent years, artificial intelligence (AI) has made great achievements in big data analysis in many fields.Consequently, many researchers have attempted to combine geoscience studies with AI, which has made new progress and can promote the development of Earth science.Climate prediction is closely related to human life and disaster prevention and mitigation, thus its prediction accuracy is highly important.This study briefly summarizes the recent progresses on the application of AI to climate prediction, including data assimilation, model parameterization, partial differential equation solution, prediction models, and numerical model output improvement.The results of this study demonstrate the possibility and applicability of using AI to improve climate prediction, which can significantly reduce computational costs and time.However, there are also many challenges involved in the application of AI, such as the construction of input data sets, the applicability of AI models, and their physical interpretability.Exploring and solving these difficult problems can help geoscience that involves many multi-source data to better utilize AI and thus improve climate prediction.
    2  Weather statistical downscaling using a 3D multi-scale residual Laplacian pyramid network
    KUANG Qiuming SHEN Chenkai YU Tingzhao LIU Jin
    2022, 45(5):660-673. DOI: 10.13878/j.cnki.dqkxxb.20220424002
    [Abstract](1679) [HTML](838) [PDF 27.05 M](2603)
    Abstract:
    The resolution of weather data greatly affects the judgment of meteorological service.Statistical downscaling is one of the effective methods to solve the conversion from low-resolution data by meteorological models to high-resolution data.Traditional statistical downscaling based on interpolation, reconstruction and example learning are some ways to achieve acceptable results.However, since the first application of convolutional neural network (CNN) to the statistical downscaling field, the performance of statistical downscaling has been significantly improved.However, few methods consider multi-layer images.Weather variables tend to be three-dimensional (3D), meaning that maps of the same region have altitudes, and there is a correlation among different dimensions.In this study, aiming at the improvement of spatial resolution of 3D meteorological elements, combined with the mechanisms of interaction of multi meteorological elements, multi-scale action and weather system configuration of multi barosphere meteorological elements, this paper proposes a multi-scale residual Laplacian pyramid network (MSRLapN) to perform 3D spatial downscaling of various meteorological elements.Specifically, a multi-scale resolution block (MSRB) is constructed to automatically extract prediction features from various meteorological elements in 3D space.Next, the multi-scale pyramid technology from the field of machine learning is introduced to describe the multi-scale interaction of meteorological elements.Next, the cycle iteration method of super-resolution reconstruction is used to learn and correct the error of downscaling prediction based on samples of historical data.In addition, seven cutting-edge deep learning super-resolution methods are used to perform spatial downscaling of the 3D spatial meteorological element.In the East China climate region, data for the two meteorological elements of relative humidity and wind speed are tested.The results indicate the following:(1) The MSRB module is more advanced than the linearly connected structure.(2) Considering three dimensions simultaneously enhances the effect in comparison to 2D images.(3) MSRLapN is superior to several state-of-the-art methods in terms of both quantitative assessment and visual quality.
    3  Research on storm surge floodplain prediction based on ConvLSTM machine learning
    XIE Wenhong XU Guangjun DONG Changming
    2022, 45(5):674-687. DOI: 10.13878/j.cnki.dqkxxb.20220711001
    [Abstract](1670) [HTML](1406) [PDF 26.25 M](2710)
    Abstract:
    A storm surge is the anomalous rising of the sea surface induced by intense atmospheric disturbances.Storm surges caused by tropical cyclones often cause great socio-economic, human activity and life and property hazards to coastal areas.Therefore, realizing accurate and timely storm surge floodplain prediction is critical.Numerical models are currently the primary method used to predict storm surges, and high-resolution floodplain models always need a significant investment in both research funds and processing time.The machine learning approach, which depends on the robust nonlinear mapping capability driven by data, has an edge over the conventional numerical model prediction in terms of research time and computational resource consumption.This paper uses the convolutional long-short term memory network (ConvLSTM) machine learning algorithm to predict storm surge floodplain in the Pearl River Estuary in Guangdong Province.Using the numerical model products driven by reanalysis data, the historical typhoon floodplain data set is constructed for machine learning model training, verification and testing.The paper studies two prediction techniques including the autoregressive prediction based on the sea surface height field and the prediction based on the predicted wind field and initial sea surface height field, which may realize the storm surge floodplain forecast based on data-driven scheme.Among them, the autoregressive prediction model performs better.By testing the previous model, it concludes that ConvLSTM can predict floodplains with a general error of less than 0.2 m based on the sea surface height field a few hours ago, even if the boundary conditions, topography, surface runoff and atmospheric signals are unknown.Under such conditions, the larger errors mostly occur at the coast and on both sides of the river.By analyzing the errors of the two models, it finds that adding wind field input to ConvLSTM does not significantly improve the prediction skills of the model.Further studies are required to determine the better way to train the data-driven prediction model by adding more features.
    4  Extended-range forecasting method of summer daily maximum temperature in the Yangtze River Basin based on convolutional neural network
    LEI Lei HSU Pang-chi GAO Qingjiu XIE Jiehong
    2022, 45(6):835-849. DOI: 10.13878/j.cnki.dqkxxb.20211101001
    [Abstract](1024) [HTML](720) [PDF 29.49 M](2606)
    Abstract:
    The Yangtze River Basin(YRB) is one of the areas with a high frequency of heatwave occurrences in China.The daily maximum temperature (Tmax) in this area shows significant low-frequency oscillation signals for (10—30 d and 30—60 d) time periods.Based on the results of the lead-lag correlation analysis between the YRB Tmax and the 10—30 d/30—60 d convection and circulation anomalies,we identify the main low-frequency signals affecting the YRB Tmax.There are three types of signals that travel in different directions:1) the eastward and southward signals from the Eurasian continent;2) circulation anomalies propagating southwestward from Northeast Asia;and 3) low-frequency convective signals propagating from the western Pacific toward East Asia.The temperature diagnostic equation results show that when the low-frequency convection/circulation anomalies approach the YRB,both the diabatic (clear-sky radiative heating) and adiabatic (associated with sinking motion) heating processes lead to variations in the YRB temperature.To identify these precursory signals objectively and efficiently,as well as consider the nonlinear interaction between YRB Tmax and the large-scale predictors,we use Convolutional Neural Network (CNN),a type of deep neural network,to train the historical data,and then develop an extended-range forecast model for YRB Tmax.The independent forecast results show that the CNN-based forecast model is capable of predicting the YRB Tmax at a 30-day lead time,with the temporal correlation coefficient between the forecast and observed Tmax of 0.63—0.70 (exceeding the 99% confidence level).The current results suggest the potential of CNN in the application of extended-range forecasting as the magnitude of error (root-mean-square error) is less than one standard deviation.
    5  A study of error correction for high-resolution gridded forecast based on a convolutional neural network in the Beijing-Tianjin-Hebei Region
    ZHANG Yanbiao SONG Linye CHEN Mingxuan HAN Lei YANG Lu
    2022, 45(6):850-862. DOI: 10.13878/j.cnki.dqkxxb.20220615001
    [Abstract](2546) [HTML](528) [PDF 22.29 M](2592)
    Abstract:
    Precise weather monitoring and accurate weather forecast are two of the most decisive factors for the success of the Winter Olympics.Considering the particularity of the 2022 Beijing Winter Olympic Games (the only Winter Olympic Games held under the climate dominated by the continental East Asian winter monsoon and the only Winter Olympic Games held in inland areas) and the rigid demand for the goal of “hundred-meter resolution and minute-updated level” high-precision forecast,the Beijing Institute of Urban Meteorology has developed a new generation of the Rapid-refresh Integrated Seamless Ensemble system—RISE—that can provide 500-and even 100-m resolution spatial grid forecast data products with 10-min updated frequency for the Beijing Winter Olympics.In order to improve the prediction accuracy of the RISE system,and considering the successful use of deep learning in the field of geoscience in recent years,this paper develops a convolution neural network-based model,Rise-Unet,using the high-resolution RISE data from 2019 to 2021 to correct the prediction results of 2-m surface temperature,2 m-relative humidity,10-m U wind speed,and 10-m V wind speed for a lead time of 4—12 hours.The root-mean-square error and mean absolute error are employed to evaluate the accuracy of the model in this study.By comparing with the original prediction results of the RISE system,it is proven that the deep learning-based model,Rise-Unet,can effectively improve the accuracy of high-resolution gridded prediction results.The method proposed in this study can be applied as the post-processing module of the RISE system,which has important scientific significance and application value for improving the grided prediction level of the RISE system as well as other high-resolution numerical weather forecasting systems.
    6  Machine learning and its potential application to climate prediction
    HE Shengping WANG Huijun LI Hua ZHAO Jiazhen
    2021, 44(1):26-38. DOI: 10.13878/j.cnki.dqkxxb.20201125001
    [Abstract](2342) [HTML](4672) [PDF 30.42 M](4734)
    Abstract:
    After two "Artificial Intelligence winters",machine learning has become a subject of intense of media hype and come up in countless articles,showing a promising future.Machine learning has gained a big success in image recognition and speech recognition systems.Refining key message and dominant features from the train datasets and making accurate prediction on the never-seen-before datasets are the major task and the ultimate goal of machine learning,respectively.From this perspective,it's feasible to integrate machine learning into climate prediction.Beginning with a simple example on finding the weights of a linear fitting,this study shows how machine learning updates weights through gradient descent algorithm and eventually obtains the linear fitting line.Next,this study illustrates the architecture of neural network and uses neural network algorithm to learn the true curve fitting a non-linear function.In the end,this study elaborates the architecture of deep learning such as convolutional neural network,and uses convolutional neural network model to hindcast winter monthly surface air temperature anomalies in East Asia.The results by deep learning are further compared with the hindcast by dynamical model-CanCM4i.This study will help to understand the fundamental of machine learning and provides insights how to integrate machine learning into climate prediction.
    7  Analysis of the application of traffic visibility estimation models in different scenarios
    LIU Qiyang QIAO Fengxue CHEN Bo SONG Zhichao JI Renjie WEI Chaoshi
    2022, 45(2):179-190. DOI: 10.13878/j.cnki.dqkxxb.20210731001
    [Abstract](2069) [HTML](1226) [PDF 8.25 M](2694)
    Abstract:
    Visibility is an important physical quantity that reflects the degree of atmospheric transparency, and is closely related to people's daily life and traffic travel.In this study, in order to make the estimation of visibility more flexible and efficient, three visibility estimation models are constructed and improved for different scenarios, and the respective applicability, advantages and disadvantages of the different models are analyzed.First, the visibility estimation is performed based on meteorological station observations, using correlation coefficient matrix and feature importance analysis to filter out the three variables of relative humidity, temperature and horizontal wind speed, and both day and night are considered to build a ternary cubic polynomial fitting model, which improves the overall fitting ability.Second, the deep learning model of visibility performs estimation based on images, and the scale invariant feature change method is used to extract the feature vector of key points of images, as the training of fully connected neural network model.Next, as the training data of the fully connected neural network model, the computational cost is reduced and the stability of the model is improved.Third, the inverse model of visibility estimation based on height highway images, according to the dark channel a priori theory and basic equation of visibility measurement, the atmospheric luminosity and transmittance are calculated, and the visibility of the monocular images is obtained based on the image distance information.The method does not require pre-set target and camera parameters, nor does it require training samples.The three visibility estimation models can be adapted to different scenarios, and can reduce the dependence on observation equipment.
    8  Multimodel ensemble forecasts of surface air temperature over China based on deep learning approach
    ZHI Xiefei WANG Tian JI Yan
    2020, 43(3):435-446. DOI: 10.13878/j.cnki.dqkxxb.20200219003
    [Abstract](1615) [HTML](0) [PDF 5.85 M](2990)
    Abstract:
    Based on the 1-7 days ensemble forecasts of the European Centre for Medium-Range Weather Forecasts (ECMWF),US National Centers for Environmental Prediction (NCEP),and the Japan Meteorological Agency (JMA),the UK Met Office (UKMO) as well as the Korea Meteorological Administration (KMA) in the TIGGE datasets,the multimodel ensemble forecasts of the surface air temperature in China and its adjacent area during the period from 1 January to 30 September 2015 were conducted by using long-term memory (LSTM) neural networks,neural networks (NN),bias-removed ensemble mean (BREM) and the superensemble (SUP) with sliding training period for the forecast period from September 5 to 30,2015.The results showed that the BREM forecast was no better than the ECMWF forecast due to the impact of low skill model forecasts among the five models.The forecast skill of SUP was better than that of all the single models.For 24-144h forecasts,the root mean square error (RMSE) of SUP was significantly smaller than that of ECMWF forecast.As the forecast lead-time increased,the RMSE increased as well.The forecast skill of NN was roughly equivalent to that of SUP.Overall,the LSTM approach showed the best forecast performance,especially when the forecast lead-time was longer than 72 h,the RMSE of the LSTM forecast was considerably smaller than that of ECMWF,BREM,NN,and SUP forecasts.The LSTM neural networks approach significantly reduced the forecast RMSE of the surface air temperature in the northwestern,northern,northeastern,southwestern,and southern China.However,the RMSE of the LSTM forecast was relatively larger in southern Xinjiang area compared with ECMWF forecast.
    9  Classification and application of highway visibility based on deep learning
    HUANG Liang ZHANG Zhendong XIAO Pengfei SUN Jiaqing ZHOU Xuecheng
    2022, 45(2):203-211. DOI: 10.13878/j.cnki.dqkxxb.20220104002
    [Abstract](1679) [HTML](1271) [PDF 4.59 M](2859)
    Abstract:
    Taking VGG16 as the benchmark model, integrating batch normalization, global average pooling and joint loss function, this paper proposed a highway fog visibility classification method based on the convolutional neural network.The experimental results show that the average recognition accuracy of the improved neural network model is 83.9%, which has higher accuracy and better convergence than other models.After the model is encapsulated into the business system for operational verification, the average recognition accuracy can reach 84.9%, and the recognition performance in the daytime is better than that at night.A dynamic generation and elimination process of agglomerate fog in Beijing-Shanghai Expressway on April 4, 2019 was monitored by the business system.The agglomerate fog process has the characteristics of fast movement, small range and short survival time.The application of the system can provide technical support for the traffic management department to deal with the intelligent management and control and decision-making scheduling when the fog occurs.
    10  Joint retrieval of soil moisture from Sentinel-1 and Sentinel-2 remote sensing data based on neural network algorithm
    WU Shanyu BAO Yansong LI Yefei WU Ying
    2021, 44(4):636-644. DOI: 10.13878/j.cnki.dqkxxb.20190419001
    [Abstract](1586) [HTML](2329) [PDF 6.17 M](2363)
    Abstract:
    Soil moisture is an important parameter of ecological environment and an important part of water cycle.The retrieval of surface soil moisture based on multi-source remote sensing data is a hotspot and trend in recent years.As a new generation of Sentinel satellites, the Sentinel-1 SAR data combined with the Sentinel-2 optical data have broad application prospects.Taking Salamanca, Spain as the research area, a BP neural network soil moisture retrieval model is constructed by combining the Sentinel-1 backscatter coefficient and incidence angle information, the vegetation index extracted from the Sentinel-2 optical data, and the ground observation data, and the model is applied to retrieve the soil moisture in the area.Finally, the model retrieval results are tested and evaluated.Results show that:(1) Based on the Sentinel-1 satellite VV and VH polarization radar backscatter coefficients and radar incidence angles and the Sentinel-2 vegetation index data, the BP neural network soil moisture retrieval model can realize high-precision retrieval of soil moisture in Salamanca area;(2) In the joint retrieval of soil moisture of optical and microwave data in vegetation coveragearea, the NDVI, NDWI1 and NDWI2 indices from the Sentinel-2 can be used to weaken the influence of vegetation on soil moisture retrieval, but the NDWI1 based on SWRI1 band can obtain more accurate soil moisture retrieval results (RMSE=0.049 cm3/cm3, ubRMSE=0.048 cm3/cm3, Bias=0.008 cm3/cm3, r=0.681);(3) Comparing with the Sentinel-1 VH polarization model, the Sentinel-1 VV polarization model shows greater advantages in soil moisture, indicating that the Sentinel-1 VV polarization model is more suitable for soil moisture retrieval.
    11  Prediction of the smoothed monthly mean sunspot area based on neural network
    DING Liu-guan LAN Ru-shi JIANG Yong PENG Jian-dong
    2012, 35(4):508-512.
    [Abstract](1417) [HTML](0) [PDF 782.32 K](2985)
    Abstract:
    Sunspot area is an important feature to measure the solar activities.Prediction of sunspot area can provide useful information for solar activities and space weather studies,etc.In this paper,we propose a smoothed monthly mean sunspot area prediction method by using an artificial neural network.The prediction model is built by training the area data before the twentieth solar cycle,and then it is used to forecast the data after the twenty-first solar cycle.We also consider the influence of different training steps and prediction steps respectively.The proposed method is able to exactly forecast the sunspot area of the next month,and the relative errors for different training steps are all less than 5%.However,the relative error will get larger if the prediction time is longer.
    12  Probabilistic precipitation forecast in East and South China based on neural network and geographic information
    ZHI Xiefei ZHANG Kejun TIAN Ye JI Yan
    2021, 44(3):381-393. DOI: 10.13878/j.cnki.dqkxxb.20210117001
    [Abstract](1059) [HTML](968) [PDF 4.49 M](2498)
    Abstract:
    With the increasing impact of human activities on climate change,the extreme weather events such as extreme precipitation occur more frequently and people pay more attention on probabilistic precipitation forecast.Since there is still a large error in precipitation ensemble forecast,it is of great significance to calibrate the forecast.Based on the daily 24 h accumulated precipitation forecasts obtained from the global ensemble forecast system of ECMWF (the European Centre for Medium-Range Weather Forecasts) with 24—168 h forecast lead times in East and South China from 8 February 2015 to 31 December 2016,NN (neutral network) model and NN-GI (neutral network-geographic information) model using feedforward neural network were established to improve probabilistic precipitation forecast and evaluate the results before and after calibration.Results show that after the correction of NN model and NN-GI model,the precipitation probabilistic forecasts are improved obviously.Compared with ECMWF raw ensemble forecasts,CRPSs of precipitation probabilistic forecasts from NN model and NN-GI model with 168 h forecast lead time decrease by around 16.00% and 21.27%,respectively.Meanwhile,compared with NN model,NN-GI model takes into account the geographic information difference of each grid point,and the overall improvement of forecasting skills in the region is better.However,NN-GI model has better performance,indicating that the machine leaning approach can improve the probabilistic forecast of the precipitation more significantly by taking into account the geographic information of each grid point in the model.
    13  A study on the artificial intelligence nowcasting based on generative adversarial networks
    CHEN Yuanzhao LIN Liangxun WANG Rui LANG Hongping YE Yunming CHEN Xunlai
    2019, 42(2):311-320. DOI: 10.13878/j.cnki.dqkxxb.20190117001
    [Abstract](2956) [HTML](0) [PDF 10.19 M](3446)
    Abstract:
    Artificial intelligence nowcasting based on generative adversarial networks (GAN) has been conducted by using abundant radar echo images from 12 S-band Doppler radars in Guangdong province during the period from 2015 to 2017.Radar echo images were convoluted for 5 times in order to build the initial forecasting model.Afterwards,several confrontation trainings took place between the model images and real radar echo images,resulting in the loss function.The model was optimized constantly.Given that the model images were similar to the real radar echo images,the outputs of optimum model would be used for nowcasting.The experiments of four precipitation events in Guangdong province during 2018 suggested that the 60 min forecasted position,shape and intensity of radar echo in convective systems by GAN mostly coincide with the observations.However,the forecasted area of strong radar echo is larger than that of the observed radar echo.Furthermore,the GAN method could not forecast the precipitation caused by stratus clouds well.The GAN method could forecast moderate radar echoes quite well,while its forecast capability for strong radar echoes needs to be improved.
    14  A statistical simulation study on spring-summer precipitation in Jilin Province using self-organizing maps
    WU Xianghua MENG Fangxiu XIONG Pingping YU Huaying YAN Ni LIU Weiqi
    2018, 41(6):829-837. DOI: 10.13878/j.cnki.dqkxxb.20170507005
    [Abstract](1493) [HTML](0) [PDF 1.27 M](2625)
    Abstract:
    Based on the daily ground observations in meteorological stations of Jilin Province during April-July 1997-2015,taking temperature,air pressure,relative humidity,water vapor pressure and wind speed as covariates,this paper established a statistical prediction model of daily precipitation based on self-organizing maps(SOM).This paper studied major synoptic patterns in Jilin Province and the relationship between daily precipitation and the patterns,and based on this relationship,proposed a Monte Carlo simulation method for daily precipitation.Results demonstrate that SOM has high classification quality of synoptic patterns,and the accumulative probability distributions of adjacent synoptic patterns are similar,while those of synoptic patterns far away are quite different.The correlation coefficient between the probability of no precipitation and the corresponding width of daily precipitation interval in the synoptic patterns is -0.94,and the significance level is less than 0.01.According to the accumulative probability distribution of precipitation,20 types of synoptic patterns are divided into four categories,which match the occurrence rate of precipitation and the daily precipitation.On this basis,this paper carried out Monte Carlo simulation of daily precipitation in 24 stations of Jilin Province,and analyzed the forecast performance.The median values of MAE(mean absolute error),RMSE(root mean square error),SBrier and Ssig are 3.12 mm,6.13 mm,0.06 and 0.51,respectively,which indicates that the method has a good forecast performance in general.The distribution of MAE and RMSE is large in the southeast and small in the northwest,and all stations have smaller errors after removing the effect of the natural fluctuation of precipitation.SBrier and Ssig have no obvious spatial distribution characteristics.
    15  Forecast Models for Fujian Rainy Season Drought/Flood Based on BP and Elman Neural Networks
    WANG Yan-jiao DENG Zi-wang WANG Yao-ting SONG De-zhong
    2004, 27(6):776-783.
    [Abstract](872) [HTML](0) [PDF 372.33 K](2536)
    Abstract:
    The forecasting models of momentum BP(MBP)and Elman neural networks are developed for Fujian rainy season drought/flood prediction,and the abilities and differences of the two types of models are compared.Results suggest that the forecasting model of MBP,especially the Elman neural network which has the character of local feedback,have better fitting precision and forecast accuracy.Additionally the forecasting abilities of the two kinds of models are worse for the drought/flood grades of 2 and 4,but best for the drought/flood grades of 3.
    16  Application of downscaling forecast for the North of Zhejiang precipitation in summer based on the BP neural network model
    LI Yuejun GUO Pinwen
    2017, 40(3):425-432. DOI: 10.13878/j.cnki.dqkxxb.20140324001
    [Abstract](1238) [HTML](0) [PDF 2.51 M](2826)
    Abstract:
    Based on the daily 500 hPa geopotential height data between June and August, 2007—2012, the historical reanalysis grid data of NCEP global 2.5°×2.5°and the daily precipitation data of 158 meteorological stations in north of Zhejiang province, the relationships between local precipitation and large-scale precipitation in different atmospheric circulations are studied in this paper.The BP neural network combined with 4 forecasting objects and corresponding predictor variables in different circulations are employed to design 4 downscaling function models to approximate the precipitation data.The 4 models are used to simulate and forecast the daily precipitation data of 158 meteorological stations in north of Zhejiang province, and the results show that the BP neural network model with 2 hidden layers has good simulation accuracy.Through Jenkinson atmospheric circulation to classify the precipitation into SE(SE type), NW(NW type), C(C type) and SW(SW type), NW type and C type generally outperform the SW type and SE type in simulation of the extreme precipitation.Compared with the area of Ningbo and Zhoushan, other areas of north Zhejiang reflect the greater error value from 4 atmospheric circulations.The prediction accuracy of the downscaling model is the best of three types of rainstorm forecast after categorizing rainfall into different levels.
    17  Investigate on the pre-assessment of typhoon disaster in Ningbo based on BP neural network
    CHEN Youli ZHU Xianchun HU Bo GU Xiaoli
    2018, 41(5):668-675. DOI: 10.13878/j.cnki.dqkxxb.20180523001
    [Abstract](1595) [HTML](0) [PDF 1.22 M](2967)
    Abstract:
    Expending 58 typhoon cases that had the considerable effect on Ningbo and had finish catastrophe records from 1949 to 2015. In view of the information of the calamity, the comprehensive correlation degree of typhoon disaster (Roj) was set up by utilizing the grey relational investigation technique. Choosing the disaster-causing factors of typhoon and Roj that point build disaster pre-assessment technique of typhoon disaster by utilizing BP neural network (BP). The outcomes demonstrated that, the severity of typhoon which evaluated by Roj is reasonable and available. There is a significant correlation between typhoon disaster risk factors and disaster assessment indicators as well as Roj. The pre-evaluation model of BP is useful for predicting typhoon disaster;the correlation coefficient linking the simulated value and the actual value of the training set and the test set respectively reached 0.94 and 0.896 and both achieved the confidence interval of 0.01. The consensus rate of the disaster level forecast of the training set and the test set is 85.3% and 77.8% respectively. This investigate outcomes could provide scientific premise to counter the typhoon work of government decision-making divisions.

    Address:No.219, Ningliu Road, Nanjing, Jiangsu, China

    Postcode:210044

    Tel:025-58731158