arXiv subjects

Mitsue Den

Publications and source records attributed to Mitsue Den.

6 recordsLinked to original sources

arXiv2021

Operational solar flare prediction model using Deep Flare Net

We developed an operational solar flare prediction model using deep neural networks, named Deep Flare Net (DeFN). DeFN can issue probabilistic forecasts of solar flares in two categories, such as >=M-class and =C-class and =M-class flares and TSS = 0.63 for >=C-class flares. For comparison, we evaluated the operationally forecast results from January 2019 to June 2020. We found that operational DeFN forecasts achieved TSS = 0.70 (0.84) for >=C-class flares with the probability threshold of 50 (40)%, although there were very few M-class flares during this period and we should continue monitoring the results for a longer time. Here, we adopted a chronological split to divide the database into two for training and testing. The chronological split appears suitable for evaluating operational models. Furthermore, we proposed the use of time-series cross-validation. The procedure achieved TSS = 0.70 for >=M-class flares and 0.59 for >=C-class flares using the datasets obtained from 2010 to 2017.

astro-ph.SR↗

arXiv2020

Validation of Coronal Mass Ejection Arrival-Time Forecasts by Magnetohydrodynamic Simulations based on Interplanetary Scintillation Observations

Coronal mass ejections (CMEs) cause various disturbances of the space environment; therefore, forecasting their arrival time is very important. However, forecasting accuracy is hindered by limited CME observations in interplanetary space. This study investigates the accuracy of CME arrival times at the Earth forecasted by three-dimensional (3D) magnetohydrodynamic (MHD) simulations based on interplanetary scintillation (IPS) observations. In this system, CMEs are approximated as spheromaks with various initial speeds. Ten MHD simulations with different CME initial speed are tested, and the density distributions derived from each simulation run are compared with IPS data observed by the Institute for Space-Earth Environmental Research (ISEE), Nagoya University. The CME arrival time of the simulation run that most closely agrees with the IPS data is selected as the forecasted time. We then validate the accuracy of this forecast using 12 halo CME events. The average absolute arrival-time error of the IPS-based MHD forecast is approximately 5.0 h, which is one of the most accurate predictions that ever been validated, whereas that of MHD simulations without IPS data, in which the initial CME speed is derived from white-light coronagraph images, is approximately 6.7 h. This suggests that the assimilation of IPS data into MHD simulations can improve the accuracy of CME arrival-time forecasts. The average predicted arrival times are earlier than the actual arrival times. These early predictions may be due to overestimation of the magnetic field included in the spheromak and/or underestimation of the drag force from the background solar wind, the latter of which could be related to underestimation of CME size or background solar wind density.

astro-ph.SR↗

arXiv2020

Reliable Probability Forecast of Solar Flares: Deep Flare Net-Reliable (DeFN-R)

We developed a reliable probabilistic solar flare forecasting model using a deep neural network, named Deep Flare Net-Reliable (DeFN-R). The model can predict the maximum classes of flares that occur in the following 24 h after observing images, along with the event occurrence probability. We detected active regions from 3x10^5 solar images taken during 2010-2015 by Solar Dynamic Observatory and extracted 79 features for each region, which we annotated with flare occurrence labels of X-, M-, and C-classes. The extracted features are the same as used by Nishizuka et al. (2018); for example, line-of-sight/vector magnetograms in the photosphere, brightening in the corona, and the X-ray emissivity 1 and 2 h before an image. We adopted a chronological split of the database into two for training and testing in an operational setting: the dataset in 2010-2014 for training and the one in 2015 for testing. DeFN-R is composed of multilayer perceptrons formed by batch normalizations and skip connections. By tuning optimization methods, DeFN-R was trained to optimize the Brier skill score (BSS). As a result, we achieved BSS = 0.41 for >=C-class flare predictions and 0.30 for >=M-class flare predictions by improving the reliability diagram while keeping the relative operating characteristic curve almost the same. Note that DeFN is optimized for deterministic prediction, which is determined with a normalized threshold of 50%. On the other hand, DeFN-R is optimized for a probability forecast based on the observation event rate, whose probability threshold can be selected according to users' purposes.

astro-ph.SR↗

arXiv2019

Development of a coronal mass ejection arrival time forecasting system using interplanetary scintillation observations

Coronal mass ejections (CMEs) cause disturbances in the environment of the Earth when they arrive at the Earth. However, the prediction of the arrival of CMEs still remains a challenge. We have developed an interplanetary scintillation (IPS) estimation system based on a global magnetohydrodynamic (MHD) simulation of the inner heliosphere to predict the arrival time of CMEs. In this system, the initial speed of a CME is roughly derived from white light coronagraph observations. Then, the propagation of the CME is calculated by a global MHD simulation. The IPS response is estimated by the three-dimensional density distribution of the inner heliosphere derived from the MHD simulation. The simulated IPS response is compared with the actual IPS observations made by the Institute for Space-Earth Environmental Research, Nagoya University, and shows good agreement with that observed. We demonstrated how the simulation system works using a halo CME event generated by a X9.3 flare observed on September 5, 2017. We find that the CME simulation that best estimates the IPS observation can more accurately predict the time of arrival of the CME at the Earth. These results suggest that the accuracy of the CME arrival time can be improved if our current MHD simulations include IPS data.

astro-ph.SR↗

arXiv2018

Deep Flare Net (DeFN) model for solar flare prediction

We developed a solar flare prediction model using a deep neural network (DNN), named Deep Flare Net (DeFN). The model can calculate the probability of flares occurring in the following 24 h in each active region, which is used to determine the most likely maximum classes of flares via a binary classification (e.g., >=M class versus =C class versus =10^7 K) and the X-ray and 131 A intensity data 1 and 2 h before an image. For operational evaluation, we divided the database into two for training and testing: the dataset in 2010-2014 for training and the one in 2015 for testing. The DeFN model consists of deep multilayer neural networks, formed by adapting skip connections and batch normalizations. To statistically predict flares, the DeFN model was trained to optimize the skill score, i.e., the true skill statistic (TSS). As a result, we succeeded in predicting flares with TSS=0.80 for >=M-class flares and TSS=0.63 for >=C-class flares. Note that in usual DNN models, the prediction process is a black box. However, in the DeFN model, the features are manually selected, and it is possible to analyze which features are effective for prediction after evaluation.

astro-ph.SR↗

arXiv2017

Verification of operational solar flare forecast: Case of Regional Warning Center Japan

In this article, we discuss a verification study of an operational solar flare forecast in the Regional Warning Center (RWC) Japan. The RWC Japan has been issuing four-categorical deterministic solar flare forecasts for a long time. In this forecast verification study, we used solar flare forecast data accumulated over 16 years (from 2000 to 2015). We compiled the forecast data together with solar flare data obtained with the Geostationary Operational Environmental Satellites (GOES). Using the compiled data sets, we estimated some conventional scalar verification measures with 95% confidence intervals. We also estimated a multi-categorical scalar verification measure. These scalar verification measures were compared with those obtained by the persistence method and recurrence method. As solar activity varied during the 16 years, we also applied verification analyses to four subsets of forecast-observation pair data with different solar activity levels. We cannot conclude definitely that there are significant performance difference between the forecasts of RWC Japan and the persistence method, although a slightly significant difference is found for some event definitions. We propose to use a scalar verification measure to assess the judgment skill of the operational solar flare forecast. Finally, we propose a verification strategy for deterministic operational solar flare forecasting.

astro-ph.SR↗