Statistical Downscaling for Multi-Model Ensemble Prediction of Summer Monsoon Rainfall in the Asia-Pacific Region Using Geopotential Height Field

The 21-yr ensemble predictions of model precipitation and circulation in the East Asian and western North Pacific （Asia-Pacific） summer monsoon region （0°-50°N, 100° 150°E） were evaluated in nine different AGCM, used in the Asia-Pacific Economic Cooperation Climate Center （APCC） multi-model ensemble seasonal prediction system. The analysis indicates that the precipitation anomaly patterns of model ensemble predictions are substantially different from the observed counterparts in this region, but the summer monsoon circulations are reasonably predicted. For example, all models can well produce the interannual variability of the western North Pacific monsoon index （WNPMI） defined by 850 hPa winds, but they failed to predict the relationship between WNPMI and precipitation anomalies. The interannual variability of the 500 hPa geopotential height （GPH） can be well predicted by the models in contrast to precipitation anomalies. On the basis of such model performances and the relationship between the interannual variations of 500 hPa GPH and precipitation anomalies, we developed a statistical scheme used to downscale the summer monsoon precipitation anomaly on the basis of EOF and singular value decomposition （SVD）. In this scheme, the three leading EOF modes of 500 hPa GPH anomaly fields predicted by the models are firstly corrected by the linear regression between the principal components in each model and observation, respectively. Then, the corrected model GPH is chosen as the predictor to downscale the precipitation anomaly field, which is assembled by the forecasted expansion coefficients of model 500 hPa GPH and the three leading SVD modes of observed precipitation anomaly corresponding to the prediction of model 500 hPa GPH during a 19-year training period. The cross-validated forecasts suggest that this downscaling scheme may have a potential to improve the forecast skill of the precipitation anomaly in the South China Sea, western North Pacific and the East Asia Pacific regions, where the anomaly correlation coefficient （ACC） has been improved by 0.14, corresponding to the reduced RMSE of 10.4% in the conventional multi-model ensemble （MME） forecast.

Ensemble Simulation of Land Evapotranspiration in China Based on a Multi-Forcing and Multi-Model Approach

In order to reduce the uncertainty of offline land surface model （LSM） simulations of land evapotranspiration （ET）, we used ensemble simulations based on three meteorological forcing datasets [Princeton, ITPCAS （Institute of Tibetan Plateau Research, Chinese Academy of Sciences）, Qian] and four LSMs （BATS, VIC, CLM3.0 and CLM3.5）, to explore the trends and spatiotemporal characteristics of ET, as well as the spatiotemporal pattern of ET in response to climate factors over China's Mainland during 1982-2007. The results showed that various simulations of each member and their arithmetic mean （EnsAVlean） could capture the spatial distribution and seasonal pattern of ET sufficiently well, where they exhibited more significant spatial and seasonal variation in the ET compared with observation-based ET estimates （Obs_MTE）. For the mean annual ET, we found that the BATS forced by Princeton forcing overestimated the annual mean ET compared with Obs_MTE for most of the basins in China, whereas the VIC forced by Princeton forcing showed underestimations. By contrast, the Ens_Mean was closer to Obs_MTE, although the results were underestimated over Southeast China. Furthermore, both the Obs_MTE and Ens_Mean exhibited a significant increasing trend during 1982-98; whereas after 1998, when the last big EI Nifio event occurred, the Ens_Mean tended to decrease significantly between 1999 and 2007, although the change was not significant for Obs_MTE. Changes in air temperature and shortwave radiation played key roles in the long-term variation in ET over the humid area of China, but precipitation mainly controlled the long-term variation in ET in arid and semi-arid areas of China.

Representing Model Uncertainty by Multi-Stochastic Physics Approaches in the GRAPES Ensemble

To represent model uncertainties more comprehensively,a stochastically perturbed parameterization(SPP)scheme consisting of temporally and spatially varying perturbations of 18 parameters in the microphysics,convection,boundary layer,and surface layer parameterization schemes,as well as the stochastically perturbed parameterization tendencies(SPPT)scheme,and the stochastic kinetic energy backscatter(SKEB)scheme,is applied in the Global and Regional Assimilation and Prediction Enhanced System-Regional Ensemble Prediction System(GRAPES-REPS)to evaluate and compare the general performance of various combinations of multiple stochastic physics schemes.Six experiments are performed for a summer month(1-30 June 2015)over China and multiple verification metrics are used.The results show that:(1)All stochastic experiments outperform the control(CTL)experiment,and all combinations of stochastic parameterization schemes perform better than the single SPP scheme,indicating that stochastic methods can effectively improve the forecast skill,and combinations of multiple stochastic parameterization schemes can better represent model uncertainties;(2)The combination of all three stochastic physics schemes(SPP,SPPT,and SKEB)outperforms any other combination of two schemes in precipitation forecasting and surface and upper-air verification to better represent the model uncertainties and improve the forecast skill;(3)Combining SKEB with SPP and/or SPPT results in a notable increase in the spread and reduction in outliers for the upper-air wind speed.SKEB directly perturbs the wind field and therefore its addition will greatly impact the upper-air wind-speed fields,and it contributes most to the improvement in spread and outliers for wind;(4)The introduction of SPP has a positive added value,and does not lead to large changes in the evolution of the kinetic energy(KE)spectrum at any wavelength;(5)The introduction of SPPT and SKEB would cause a 5%-10%and 30%-80%change in the KE of mesoscale systems,and all three stochastic schemes(SPP,SPPT,and SKEB)mainly affect the KE of mesoscale systems.This study indicates the potential of combining multiple stochastic physics schemes and lays a foundation for the future development and design of regional and global ensembles.

Improving Multi-model Ensemble Probabilistic Prediction of Yangtze River Valley Summer Rainfall

Seasonal prediction of summer rainfall over the Yangtze River valley（YRV） is valuable for agricultural and industrial production and freshwater resource management in China, but remains a major challenge. Earlier multi-model ensemble（MME） prediction schemes for summer rainfall over China focus on single-value prediction, which cannot provide the necessary uncertainty information, while commonly-used ensemble schemes for probability density function（PDF） prediction are not adapted to YRV summer rainfall prediction. In the present study, an MME PDF prediction scheme is proposed based on the ENSEMBLES hindcasts. It is similar to the earlier Bayesian ensemble prediction scheme, but with optimization of ensemble members and a revision of the variance modeling of the likelihood function. The optimized ensemble members are regressed YRV summer rainfall with factors selected from model outputs of synchronous 500-h Pa geopotential height as predictors. The revised variance modeling of the likelihood function is a simple linear regression with ensemble spread as the predictor. The cross-validation skill of 1960–2002 YRV summer rainfall prediction shows that the new scheme produces a skillful PDF prediction, and is much better-calibrated, sharper, and more accurate than the earlier Bayesian ensemble and raw ensemble.

A Bayesian Scheme for Probabilistic Multi-Model Ensemble Prediction of Summer Rainfall over the Yangtze River Valley

A Bayesian probabilistic prediction scheme of the Yangtze River Valley (YRV) summer rainfall is proposed to combine forecast information from multi-model ensemble dataset provided by ENSEMBLES project.Due to the low forecast skill of rainfall in dynamic models,the time series of regressed YRV summer rainfall are selected as ensemble members in the new scheme,instead of commonly-used YRV summer rainfall simulated by models.Each time series of regressed YRV summer rainfall is derived from a simple linear regression.The predictor in each simple linear regression is the skillfully simulated circulation or surface temperature factor which is highly linear with the observed YRV summer rainfall in the training set.The high correlation between the ensemble mean of these regressed YRV summer rainfall and observation benefit extracting more sample information from the ensemble system.The results show that the cross-validated skill of the new scheme over the period of 1960 to 2002 is much higher than equally-weighted ensemble,multiple linear regression,and Bayesian ensemble with simulated YRV summer rainfall as ensemble members.In addition,the new scheme is also more skillful than reference forecasts (random forecast at a 0.01 significance level for ensemble mean and climatology forecast for probability density function).

Validation of the effects of temperature simulated by multi-model ensemble and prediction of mean temperature changes for the next three decades in China

Using series of daily average temperature observations over the period of 1961-1999 of 701 meteorological stations in China, and simulated results of 20 global climate models （such as BCCR_BCM2.0, CGCM3T47） during the same period as the observation, we validate and analyze the simulated results of the models by using three factor statistical method, achieve the results of mul- ti-model ensemble, test and verify the results of multi-model ensemble by using the observation data during the period of 1991-1999. Finally, we analyze changes of the annual mean temperature result of multi-mode ensemble prediction for the period of 2011-2040 under the emission scenarios A2, A1B and B 1. Analyzed results show that： （1） Global climate models can repro- duce Chinese regional spatial distribution of annual mean temperature, especially in low latitudes and eastern China. （2） With the factor of the trend of annual mean temperature changes in reference period, there is an obvious bias between the model and the observation. （3） Testing the result of multi-model ensemble during the period of 1991-1999, we can simulate the trend of temper- ature increase. Compared to observation, the result of different weighing multi-model ensemble prediction is better than the same weighing ensemble. （4） For the period of 20ll-2040, the growth of the annual mean temperature in China, which results from multi-mode ensemble prediction, is above 1℃. In the spatial distribution of annual mean temperature, under the emission scenarios of A2, A1B and B 1, the trend of growth in South China region is the smallest, the increment is less than or equals to 0.8℃; the trends in the northwestern region and south of the Qinghai-Tibet Plateau are the largest, the increment is more than 1℃.

Covid-19 Diagnosis Using a Deep Learning Ensemble Model with Chest X-Ray Images

Covid-19 is a deadly virus that is rapidly spread around the world towards the end of the 2020.The consequences of this virus are quite frightening,especially when accompanied by an underlying disease.The novelty of the virus,the constant emergence of different variants and its rapid spread have a negative impact on the control and treatment process.Although the new test kits provide almost certain results,chest X-rays are extremely important to detect the progression and degree of the disease.In addition to the Covid-19 virus,pneumonia and harmless opacity of the lungs also complicate the diagnosis.Considering the negative results caused by the virus and the treatment costs,the importance of fast and accurate diagnosis is clearly seen.In this context,deep learning methods appear as an extremely popular approach.In this study,a hybrid model design with superior properties of convolutional neural networks is presented to correctly classify the Covid-19 disease.In addition,in order to contribute to the literature,a suitable dataset with balanced case numbers that can be used in all artificial intelligence classification studies is presented.With this ensemble model design,quite remarkable results are obtained for the diagnosis of three and four-class Covid-19.The proposed model can classify normal,pneumonia,and Covid-19 with 92.6%accuracy and 82.6%for normal,pneumonia,Covid-19,and lung opacity.

基于多LS-SVM集成模型的锅炉NO_x排放量建模

为了提高电站锅炉氮氧化物(NO_x)排放量预测模型的精度,提出了一种基于多最小二乘支持向量机(LS-SVM)集成模型的NO_x排放量建模方法。首先按照NO_x排放量由低到高将数据空间初步划分为低、中、高3个子空间,然后依据输入变量与NO_x相关性分析来确定输入变量的权重,通过筛选得到主要的特征变量。在此基础之上,采用有监督的遗传算法-软模糊聚类(GA-SFCM)方法,获得各数据子空间的聚类中心及其相应的样本隶属度,通过融合隶属度的最小二乘法对各子空间LS-SVM模型进行集成。仿真结果表明,通过筛选参与聚类的变量提高了聚类性能和模型精度,采用有监督的GA-SFCM算法进行聚类,降低了聚类复杂度,建立的多LS-SVM集成模型比单一LS-SVM模型有更好的泛化能力。

The China Multi-Model Ensemble Prediction System and Its Application to Flood-Season Prediction in 2018

Multi-model ensemble prediction is an effective approach for improving the prediction skill short-term climate prediction and evaluating related uncertainties. Based on a combination of localized operation outputs of Chinese climate models and imported forecast data of some international operational models, the National Climate Center of the China Meteorological Administration has established the China multi-model ensemble prediction system version 1.0 (CMMEv1.0) for monthly-seasonal prediction of primary climate variability modes and climate elements. We verified the real-time forecasts of CMMEv1.0 for the 2018 flood season (June-August) starting from March 2018 and evaluated the 1991-2016 hindcasts of CMMEv1.0. The results show that CMMEv1.0 has a significantly high prediction skill for global sea surface temperature (SST) anomalies, especially for the El Nino-Southern Oscillation (ENSO) in the tropical central-eastern Pacific. Additionally, its prediction skill for the North Atlantic SST triple (NAST) mode is high, but is relatively low for the Indian Ocean Dipole (IOD) mode. Moreover, CMMEv1.0 has high skills in predicting the western Pacific subtropical high (WPSH) and East Asian summer monsoon (EASM) in the June-July-August (JJA) season. The JJA air temperature in the CMMEv1.0 is predicted with a fairly high skill in most regions of China, while the JJA precipitation exhibits some skills only in northwestern and eastern China. For real-time forecasts in March-August 2018, CMMEv1.0 has accurately predicted the ENSO phase transition from cold to neutral in the tropical central-eastern Pacific and captures evolutions of the NAST and IOD indices in general. The system has also captured the main features of the summer WPSH and EASM indices in 2018, except that the predicted EASM is slightly weaker than the observed. Furthermore, CMMEv1.0 has also successfully predicted warmer air temperatures in northern China and captured the primary rainbelt over northern China, except that it predicted much more precipitation in the middle and lower reaches of the Yangtze River than observation.

New perspective in statistical modeling of wall-bounded turbulence

Despite dedicated effort for many decades,statistical description of highly technologically important wall turbulence remains a great challenge.Current models are unfortunately incomplete,or empirical,or qualitative.After a review of the existing theories of wall turbulence,we present a new framework,called the structure ensemble dynamics （SED）,which aims at integrating the turbulence dynamics into a quantitative description of the mean flow.The SED theory naturally evolves from a statistical physics understanding of non-equilibrium open systems,such as fluid turbulence, for which mean quantities are intimately coupled with the fluctuation dynamics.Starting from the ensemble-averaged Navier-Stokes（EANS） equations,the theory postulates the existence of a finite number of statistical states yielding a multi-layer picture for wall turbulence.Then,it uses order functions（ratios of terms in the mean momentum as well as energy equations） to characterize the states and transitions between states.Application of the SED analysis to an incompressible channel flow and a compressible turbulent boundary layer shows that the order functions successfully reveal the multi-layer structure for wall-bounded turbulence, which arises as a quantitative extension of the traditional view in terms of sub-layer,buffer layer,log layer and wake. Furthermore,an idea of using a set of hyperbolic functions for modeling transitions between layers is proposed for a quantitative model of order functions across the entire flow domain.We conclude that the SED provides a theoretical framework for expressing the yet-unknown effects of fluctuation structures on the mean quantities,and offers new methods to analyze experimental and simulation data.Combined with asymptotic analysis,it also offers a way to evaluate convergence of simulations.The SED approach successfully describes the dynamics at both momentum and energy levels, in contrast with all prevalent approaches describing the mean velocity profile only.Moreover,the SED theoretical framework is general,independent of the flow system to study, while the actual functional form of the order functions may vary from flow to flow.We assert that as the knowledge of order functions is accumulated and as more flows are analyzed, new principles（such as hierarchy,symmetry,group invariance,etc.） governing the role of turbulent structures in the mean flow properties will be clarified and a viable theory of turbulence might emerge.

Probabilistic Seasonal Prediction of Summer Rainfall over East China Based on Multi-Model Ensemble Schemes

The skill of probability density function （PDF） prediction of summer rainfall over East China using optimal ensemble schemes is evaluated based on the precipitation data from five coupled atmosphere-ocean general circulation models that participate in the ENSEMBLES project. The optimal ensemble scheme in each region is the scheme with the highest skill among the four commonly-used ones： the equally-weighted ensemble （EE）, EE for calibrated model-simulations （Cali-EE）, the ensemble scheme based on multiple linear regression analysis （MLR）, and the Bayesian ensemble scheme （Bayes）. The results show that the optimal ensemble scheme is the Bayes in the southern part of East China; the Cali-EE in the Yangtze River valley, the Yangtze-Huaihe River basin, and the central part of northern China; and the MLR in the eastern part of northern China. Their PDF predictions are well calibrated, and are sharper than or have approximately equal interval-width to the climatology prediction. In all regions, these optimal ensemble schemes outperform the climatology prediction, indicating that current commonly-used multi-model ensemble schemes are able to produce skillful PDF prediction of summer rainfall over East China, even though more information for other model variables is not derived.

Evaluating the Formation Mechanisms of the Equatorial Pacific SST Warming Pattern in CMIP5 Models

Based on the historical and RCP8.5 runs of the multi-model ensemble of 32 models participating in CMIP5, the present study evaluates the formation mechanisms for the patterns of changes in equatorial Pacific SST under global warming. Two features with complex formation processes, the zonal E1 Nifio-like pattern and the meridional equatorial peak warm- ing （EPW）, are investigated. The climatological evaporation is the main contributor to the E1 Nifio-like pattern, while the ocean dynamical thermostat effect plays a comparable negative role. The cloud-shortwave-radiation-SST feedback and the weakened Walker circulation play a small positive role in the E1 Nifio-like pattern. The processes associated with ocean dynamics are confined to the equator. The climatological evaporation is also the dominant contributor to the EPW pattern, as suggested in previous studies. However, the effects of some processes are inconsistent with previous studies. For example, changes in the zonal heat advection due to the weakened Walker circulation have a remarkable positive contribution to the EPW pattern, and changes in the shortwave radiation play a negative role in the EPW pattern.

Assessment of Arctic sea ice simulations in CMIP5 models using a synthetical skill scoring method

The Arctic sea ice cover has declined at an unprecedented pace since the late 20th century. As a result, the feedback of sea ice anomalies for atmospheric circulation has been increasingly evidenced. While climatic models almost consistently reproduced a decreasing trend of sea ice cover, the reported results show a large distribution. To evaluate the performance of models for simulating Arctic sea ice cover and its potential role in climate change, this study constructed a reasonable metric by synthesizing both linear trends and anomalies of sea ice. This study particularly focused on the Barents Sea and the Kara Sea, where sea ice anomalies have the highest potential to affect the atmosphere. The investigated models can be grouped into three categories according to their normalized skill scores. The strong contrast among the multi-model ensemble means of different groups demonstrates the robustness and rationality of this method. Potential factors that account for the different performances of climate models are further explored. The results show that model performance depends more on the ozone datasets that are prescribed by the model rather than on the chemical representation of ozone.

Seasonal Prediction of June Rainfall over South China:Model Assessment and Statistical Downscaling

The performances of various dynamical models from the Asia-Pacific Economic Cooperation（APEC） Climate Center（APCC） multi-model ensemble（MME） in predicting station-scale rainfall in South China（SC） in June were evaluated.It was found that the MME mean of model hindcasts can skillfully predict the June rainfall anomaly averaged over the SC domain.This could be related to the MME＇s ability in capturing the observed linkages between SC rainfall and atmospheric large-scale circulation anomalies in the Indo-Pacific region.Further assessment of station-scale June rainfall prediction based on direct model output（DMO） over 97 stations in SC revealed that the MME mean outperforms each individual model.However,poor prediction abilities in some in-land and southeastern SC stations are apparent in the MME mean and in a number of models.In order to improve the performance at those stations with poor DMO prediction skill,a station-based statistical downscaling scheme was constructed and applied to the individual and MME mean hindcast runs.For several models,this scheme can outperform DMO at more than 30 stations,because it can tap into the abilities of the models in capturing the anomalous Indo-Paciric circulation to which SC rainfall is considerably sensitive.Therefore,enhanced rainfall prediction abilities in these models should make them more useful for disaster preparedness and mitigation purposes.

基于多模型融合的中长期径流集成预测方法

中长期水文预报是流域水资源规划与合理配置的重要依据。为提高中长期径流预测精度,提出了一种基于多模型融合的水库中长期径流集成预测方法。该方法将ARMA、BP、LSTM、RF和SVR等5个异质预测模型进行融合,同时采用超参数优化方法确定各模型的最优参数。将其用于青海省龙羊峡水库的中长期径流预报中,结果表明,通过Stacking融合算法建立的集成预测模型相较于单一模型,取得了更高的预测精度(R2值由0.71提升至0.82)。此方法可为提升流域中长期径流预测精度提供一定参考。

MSH-YOLOv8:融合尺度重建的蘑菇小目标检测方法

[目的/意义]为了解决图像尺寸变化和目标尺度变换共存对小目标检测精度的影响问题,本研究提出了一种新的检测模型:Multi-Strategy Handling YOLOv8(MSH-YOLOv8)。[方法]该模型在YOLOv8的基础上增加一个检测头,以提高小尺度目标敏感度;引入Swin Transformer的检测结构到头部网络,以减少计算冗余;引入包含可变形卷积的C2f_Deformable Convolutionv4(C2f_DCNv4)结构和Swin Transformer编码器结构重构YOLOv8主干网络,优化并增强其特征传递和提取能力,提高小目标敏感度;采用基于规范化的注意力模块(Normalizationbased Attention Module,NAM)优化网络检测速度和准确性;用Wise-Intersection over Union Loss(WIoU)代替原损失函数,以提高训练效果和收敛速度;在后处理阶段应用分辨率动态训练、多尺度测试、软非极大值抑制算法(Soft-Non-Maximum Suppression,Soft-NMS)、加权边界框融合算法(Weighted Boxes Fusion,WBF)等方法,提高尺度变化下小目标检测效果。以蘑菇为研究对象,在开放数据集Fungi上开展实验。[结果和讨论]MSH-YOLOv8的平均正确率(Average Precision50,AP50)和AP@50-95分别达到了98.49%和75.29%,其中小目标检测指标值APs达39.73%。相较于主流模型YOLOv8,三项指标分别提高了2.34%,4.06%和8.55%;相较于优秀模型Transformer Prediction Heads-YOLOv5(TPH-YOLOv5),三项指标分别提高了2.14%,2.76%和6.89%。[结论]本研究提出的MSH-YOLOv8改进方法可在图像尺寸变化与目标尺度变化条件下有效提高小目标的检测效果。

基于数据同化系统的作物产量预测研究进展

数据同化系统融合了遥感数据和作物生长模型的优势,是实时监测农业生产状况的有力手段。本文在简要介绍作物产量遥感估测方法的基础上,重点对数据同化算法的发展情况、多源遥感数据在数据同化上的应用潜力、数据同化系统的不确定性以及数据同化系统的尺度效应4方面进行论述。并且针对农业应用现状,提出未来应充分挖掘多源遥感数据、多作物生长模型集合和数据算法的优势,最终实现以机理模型为纽带的作物估产模式,并为制定田间管理策略、规划粮食产业布局和制定进出口贸易政策提供有力的数据和技术支撑。

低压货舱多参数火灾探测集成模型的优化选择

针对低压飞机货舱环境下单模型的烟雾探测算法不灵敏的问题,给出了一种基于二次优化选择(Quadratic Optimization Choice,QOC)策略的集成分类模型。首先,对多种火灾特征参数包括CO体积分数、温度、湿度、红外和蓝光波长光散射功率信号、烟颗粒索特平均直径及对应增长率进行增益评估,筛选出关联度高的参数作为属性,通过特征工程和性能对候选分类器进行排序,然后采用QOC策略和软投票法集成机制确定次级分类器,最后指定多层感知器(Multilayer Perceptron,MLP)作为元分类器的模型集成方法,以提高烟雾探测模型在真实飞行环境的准确性和鲁棒性。模型性能将基于精确率、召回率和F1、F2、F3指标进行比较。结果表明,集成模型应用于60 kPa低压环境烟雾探测结果优于K邻近算法(K Nearest Neighbor,KNN)和MLP,对榉木和航空汽油可燃物分别具有0.9724和0.9601的分类精确率,较原始算法KNN分别提高了0.0872和0.0626,较原始算法MLP分别提高了0.0368和0.1822,集成模型具有更好的性能。

两阶段模型协同搜索的昂贵多目标进化优化

近年来,昂贵多目标优化问题的求解获得了越来越多的关注.然而,随着决策空间维度的升高,模型的有效性和准确性很难保证.因此,本文提出了一种两阶段模型协同搜索的昂贵多目标进化优化.在该方法中,每轮种群进化前构建全局模型,以辅助加快对最优解集的搜索.随后,利用搜索到的种群选择其邻域样本训练局部模型,对二者集成辅助算法进行进一步搜索.最后,提出基于不确定度的填充采样策略选点,进行真实评价.为了验证算法的有效性,将本文算法与4个算法分别在DTLZ和MaF测试集以及两个实际问题上进行比较,实验结果表明其具有良好的性能.

基于事件最大边界的密集视频描述方法

针对基于集合预测的密集视频描述方法由于缺乏显式的事件间特征交互且未针对事件间差异训练模型而导致的模型重复预测事件或生成语句雷同问题,提出一种基于事件最大边界的密集视频描述(dense video captioning based on event maximal margin,EMM-DVC)方法。事件边界是包含事件间特征相似度、事件在视频中时间位置的距离、生成描述多样性的评分。EMM-DVC通过最大化事件边界,使相似预测结果的距离远且预测结果和实际事件的距离近。另外,EMM-DVC引入事件边界距离损失函数,通过扩大事件边界距离,引导模型关注不同事件。在ActivityNet Captions数据集上的实验证明,EMM-DVC与同类密集视频描述模型相比能生成更具多样性的描述文本,并且与主流密集视频描述模型相比,EMM-DVC在多个指标上达到最优水平。