Customization of WRF-ARW model with physical parameterization schemes for the simulation of tropical cyclones over North Indian Ocean

The convection and planetary boundary layer (PBL) processes play significant role in the genesis and intensification of tropical cyclones (TCs). Several convection and PBL parameterization schemes incorporate these processes in the numerical weather prediction models. Therefore, a systematic intercomparison of performance of parameterization schemes is essential to customize a model. In this context, six combinations of physical parameterization schemes (2 PBL Schemes, YSU and MYJ, and 3 convection schemes, KF, BM, and GD) of WRF-ARW model are employed to obtain the optimum combination for the prediction of TCs over North Indian Ocean. Five cyclones are studied for sensitivity experiments and the out-coming combination is tested on real-time prediction of TCs during 2008. The tracks are also compared with those provided by the operational centers like NCEP, ECMWF, UKMO, NCMRWF, and IMD. It is found that the combination of YSU PBL scheme with KF convection scheme (YKF) provides a better prediction of intensity, track, and rainfall consistently. The average RMSE of intensity (13?hPa in CSLP and 11?m?s^?1 in 10-m wind), mean track, and landfall errors is found to be least with YKF combination. The equitable threat score (ETS) of YKF combination is more than 0.2 for the prediction of 24-h accumulated rainfall up to 125?mm. The vertical structural characteristics of cyclone inner core also recommend the YKF combination for Indian seas cyclones. In the real-time prediction of 2008 TCs, the 72-, 48-, and 24-h mean track errors are 172, 129, and 155?km and the mean landfall errors are 125, 73, and 66?km, respectively. Compared with the track of leading operational agencies, the WRF model is competing in 24?h (116?km error) and 72?h (166?km) but superior in 48-h (119?km) track forecast.     

The Indian summer monsoon as revealed by the NCMRWF system

In this study, we present the mean seasonal features of the Indian summer monsoon circulation in the National Centre for Medium Range Weather Forecasting (NCMRWF) global data assimilation and forecast system. The large-scale budgets of heat and moisture are examined in the analyzed and model atmosphere. The daily operational analyses and forecasts (day 1 through day 5) produced for the summer seasons comprising June, July and August of 1995 and 1993 have been considered for the purpose. The principal aim of the study is two-fold. Primarily, to comprehend the influence of the systematic errors over the Indian summer monsoon, secondarily, to analyze the performance of the model in capturing the interseasonal variability. The heat and moisture balances show reduction in the influx of heat and moisture in the model forecasts compared to the analyzed atmosphere over the monsoon domain. Consequently, the diabatic heating also indicates reducing trend with increase in the forecast period. In effect, the strength of Indian summer monsoon, which essentially depends on these parameters, weakens considerably in the model forecasts. Despite producing feeble monsoon circulation, the model captures interseasonal variability realistically. Although, 1995 and 1993 are fairly normal monsoon seasons, the former received more rainfall compared to the latter in certain pockets of the monsoon domain. This is clearly indicated by the analyzed and model atmosphere in terms of energetics.     

Current status of multimodel superensemble and operational NWP forecast of the Indian summer monsoon

In the last thirty years great strides have been made by large-scale operational numerical weather prediction models towards improving skills for the medium range time-scale of 7 days. This paper illustrates the use of these current forecasts towards the construction of a consensus multimodel forecast product called the superensemble. This procedure utilizes 120 of the recent-past forecasts from these models to arrive at the training phase statistics. These statistics are described by roughly 10⁷ weights. Use of these weights provides the possibility for real-time medium range forecasts with the superensemble. We show the recent status of this procedure towards real-time forecasts for the Asian summer monsoon. The member models of our suite include ECMWF, NCEP/EMC, JMA, NOGAPS (US Navy), BMRC, RPN (Canada) and an FSU global spectral forecast model. We show in this paper the skill scores for day 1 through day 6 of forecasts from standard variables such as winds, temperature, 500 hPa geopotential height, sea level pressure and precipitation. In all cases we noted that the superensemble carries a higher skill compared to each of the member models and their ensemble mean. The skill matrices we use include the RMS errors, the anomaly correlations and equitable threat scores. For many of these forecasts the improvements of skill for the superensemble over the best model was found to be quite substantial. This real-time product is being provided to many interested research groups. The FSU multimodel superensemble, in real-time, stands out for providing the least errors among all of the operational large scale models.     

Skills of different mesoscale models over Indian region during monsoon season: Forecast errors

Performance of four mesoscale models namely, the MM5, ETA, RSM and WRF, run at NCMRWF for short range weather forecasting has been examined during monsoon-2006. Evaluation is carried out based upon comparisons between observations and day-1 and day-3 forecasts of wind, temperature, specific humidity, geopotential height, rainfall, systematic errors, root mean square errors and specific events like the monsoon depressions.It is very difficult to address the question of which model performs best over the Indian region? An honest answer is ‘none’. Perhaps an ensemble approach would be the best. However, if we must make a final verdict, it can be stated that in general, (i) the WRF is able to produce best All India rainfall prediction compared to observations in the day-1 forecast and, the MM5 is able to produce best All India rainfall forecasts in day-3, but ETA and RSM are able to depict the best distribution of rainfall maxima along the west coast of India, (ii) the MM5 is able to produce least RMSE of wind and geopotential fields at most of the time, and (iii) the RSM is able to produce least errors in the day-1 forecasts of the tracks, while the ETA model produces least errors in the day-3 forecasts.     

Location specific forecasting of maximum and minimum temperatures over India by using the statistical bias corrected output of global forecasting system

The output from Global Forecasting System (GFS) T574L64 operational at India Meteorological Department (IMD), New Delhi is used for obtaining location specific quantitative forecast of maximum and minimum temperatures over India in the medium range time scale. In this study, a statistical bias correction algorithm has been introduced to reduce the systematic bias in the 24–120 hour GFS model location specific forecast of maximum and minimum temperatures for 98 selected synoptic stations, representing different geographical regions of India. The statistical bias correction algorithm used for minimizing the bias of the next forecast is Decaying Weighted Mean (DWM), as it is suitable for small samples. The main objective of this study is to evaluate the skill of Direct Model Output (DMO) and Bias Corrected (BC) GFS for location specific forecast of maximum and minimum temperatures over India. The performance skill of 24–120 hour DMO and BC forecast of GFS model is evaluated for all the 98 synoptic stations during summer (May-August 2012) and winter (November 2012–February 2013) seasons using different statistical evaluation skill measures. The magnitude of Mean Absolute Error (MAE) and Root Mean Squared Error (RMSE) for BC GFS forecast is lower than DMO during both summer and winter seasons. The BC GFS forecasts have higher skill score as compared to GFS DMO over most of the stations in all day-1 to day-5 forecasts during both summer and winter seasons. It is concluded from the study that the skill of GFS statistical BC forecast improves over the GFS DMO remarkably and hence can be used as an operational weather forecasting system for location specific forecast over India.     

Impact of modified physics in limited area model forecasts

A number of physical factors have been introduced to improve limited area model forecasts. The factors include land surface fluxes, shallow convection and radiation. The model including these additional physical factors (modified physics) is run for five cases of monsoon depression which made landfall over the Indian coast, and the results are compared with those of the control run. The forecasts are verified by computing the root mean square and mean errors. The differences in these skill scores between the two model runs are tested for their statistical significance. It is found that the modified physics has a statistically significant effect on the model skill with the maximum impact on the mean sea level pressure and the temperature. Detailed analyses of mean sea level pressure, wind, rainfall and temperature further confirm that the modified physics has maximum impact on mean sea level pressure and temperature and marginal impact on wind and rainfall. Furthermore, analyses of some model parameters related to physics at a grid point for one case of depression were done. The results show that the inclusion of the land surface physics, shallow convection and radiative processes have produced a better precipitation forecast over the grid point.     

Evaluation of statistical bias correction methods for numerical weather prediction model forecasts of maximum and minimum temperatures

Statistical bias correction methods for numerical weather prediction (NWP) forecasts of maximum and minimum temperatures over India in the medium-range time scale (up to 5 days) are proposed in this study. The objective of bias correction is to minimize the systematic error of the next forecast using bias from past errors. The need for bias corrections arises from the many sources of systematic errors in NWP modeling systems. NWP models have shortcomings in the physical parameterization of weather events and have the inability to handle sub-grid phenomena successfully. The statistical algorithms used for minimizing the bias of the next forecast are running-mean (RM) bias correction, best easy systematic estimator, simple linear regression and the nearest neighborhood (NN) weighted mean, as they are suitable for small samples. Bias correction is done for four global NWP model maximum and minimum temperature forecasts. The magnitude of the bias at a grid point depends upon geographical location and season. Validation of the bias correction methodology is carried out using daily observed and bias-corrected model maximum and minimum temperature forecast over India during July–September 2011. The bias-corrected NWP model forecast generally outperforms direct model output (DMO). The spatial distribution of mean absolute error and root-mean squared error for bias-corrected forecast over India indicate that both the RM and NN methods produce the best skill among other bias correction methods. The inter-comparison reveals that statistical bias correction methods improve the DMO forecast in terms of accuracy in forecast and have the potential for operational applications.     

Experimental real-time multi-model ensemble (MME) prediction of rainfall during monsoon 2008: Large-scale medium-range aspects

Realistic simulation/prediction of the Asian summer monsoon rainfall on various space–time scales is a challenging scientific task. Compared to mid-latitudes, a proportional skill improvement in the prediction of monsoon rainfall in the medium range has not happened in recent years. Global models and data assimilation techniques are being improved for monsoon/tropics. However, multi-model ensemble (MME) forecasting is gaining popularity, as it has the potential to provide more information for practical forecasting in terms of making a consensus forecast and handling model uncertainties. As major centers are exchanging model output in near real-time, MME is a viable inexpensive way of enhancing the forecasting skill and information content. During monsoon 2008, on an experimental basis, an MME forecasting of large-scale monsoon precipitation in the medium range was carried out in real-time at National Centre for Medium Range Weather Forecasting (NCMRWF), India. Simple ensemble mean (EMN) giving equal weight to member models, bias-corrected ensemble mean (BCEMn) and MME forecast, where different weights are given to member models, are the products of the algorithm tested here. In general, the aforementioned products from the multi-model ensemble forecast system have a higher skill than individual model forecasts. The skill score for the Indian domain and other sub-regions indicates that the BCEMn produces the best result, compared to EMN and MME. Giving weights to different models to obtain an MME product helps to improve individual member models only marginally. It is noted that for higher rainfall values, the skill of the global model rainfall forecast decreases rapidly beyond day-3, and hence for day-4 and day-5, the MME products could not bring much improvement over member models. However, up to day-3, the MME products were always better than individual member models.     

High resolution numerical weather prediction over the Indian subcontinent

In this study, the Florida State University Global Spectral Model (FSUGSM), in association with a high-resolution nested regional spectral model (FSUNRSM), is used for short-range weather forecasts over the Indian domain. Three-day forecasts for each day of August 1998 were performed using different versions of the FSUGSM and FSUNRSM and were compared with the observed fields (analysis) obtained from the European Center for Medium Range Weather Forecasts (ECMWF). The impact of physical initialization (a procedure that assimilates observed rain rates into the model atmosphere through a set of reverse algorithms) on rainfall forecasts was examined in detail. A very high nowcasting skill for precipitation is obtained through the use of high-resolution physical initialization applied at the regional model level. Higher skills in wind and precipitation forecasts over the Indian summer monsoon region are achieved using this version of the regional model with physical initialization. A relatively new concept, called the ‘multimodel/multianalysis superensemble’ is described in this paper and is applied for the wind and precipitation forecasts over the Indian subcontinent. Large improvement in forecast skills of wind at 850 hPa level over the Indian subcontinent is shown possible through the use of the multimodel superensemble. The multianalysis superensemble approach that uses the latest satellite data from the Tropical Rainfall Measuring Mission (TRMM) and the Defense Meteorological Satellite Program (DMSP) has shown significant improvement in the skills of precipitation forecasts over the Indian monsoon region.     

Impact of gridpoint statistical interpolation scheme over Indian region

An analysis system experiment was conducted for the month of June 2008 with Gridpoint Statistical Interpolation (GSI) analysis scheme using NCMRWF’s (National Centre for Medium Range Weather Forecasting) T254L64 model. Global analyses were carried out for all days of the month and respective forecast runs are made up to 120-hr. These analyses and forecasts are inter-compared with the operational T254L64 model outputs which uses Spectral Statistical Interpolation (SSI) analysis scheme. The prime objective of this study is to assess the impact of GSI analysis scheme with special emphasis on Indian summer monsoon as compared to SSI.     

A Study on the Impact of Parameterization of Physical Processes on Prediction of Tropical Cyclones over the Bay of Bengal With NCAR/PSU Mesoscale Model

Prediction of the track and intensity of tropical cyclones is one of the most challenging problems in numerical weather prediction (NWP). The chief objective of this study is to investigate the performance of different cumulus convection and planetary boundary layer (PBL) parameterization schemes in the simulation of tropical cyclones over the Bay of Bengal. For this purpose, two severe cyclonic storms are simulated with two PBL and four convection schemes using non-hydrostatic version of MM5 modeling system. Several important model simulated fields including sea level pressure, horizontal wind and precipitation are compared with the corresponding verification analysis/observation. The track of the cyclones in the simulation and analysis are compared with the best-fit track provided by India Meteorological Department (IMD). The Hong-Pan PBL scheme (as implemented in NCAR Medium Range Forecast (MRF) model) in combination with Grell (or Betts-Miller) cumulus convection scheme is found to perform better than the other combinations of schemes used in this study. Though it is expected that radiative processes may not have pronounced effect in short-range forecasts, an attempt is made to calibrate the model with respect to the two radiation parameterization schemes used in the study. And the results indicate that radiation parameterization has noticeable impact on the simulation of tropical cyclones.     

基于数值天气预报产品的气象水文耦合径流预报

以福建金溪池潭水库流域为例,采用TIGGE数据中心的ECMWF、UKMO、NCEP等7种模式控制预报产品驱动新安江模型,开展径流集合预报。通过集合挑选、多模式集成前处理以及基于BMA模型的后处理等过程,探讨不同处理方案和初始集合质量对气象水文耦合径流预报精度及不确定性的影响。结果表明,不同的处理方案均能有效提高径流预报的精度和稳定性,同时进行前处理和后处理能从降低误差输入和控制误差输出两方面减小预报误差,相对于其他方案表现更好。初始集合质量对气象水文耦合径流集合预报有一定影响,但前处理或后处理对预报误差的有效控制使得该影响并不显著。总体而言,前处理和后处理过程是提高气象水文耦合径流预报准确性和可靠性必不可少的环节,应予以重视。     

Impact of physical parameterization schemes on numerical simulation of super cyclone Gonu

The objective of this study is to investigate in detail the sensitivity of cumulus, planetary boundary layer and explicit cloud microphysics parameterization schemes on intensity and track forecast of super cyclone Gonu (2007) using the Pennsylvania State University-National Center for Atmospheric Research Fifth-Generation Mesoscale Model (MM5). Three sets of sensitivity experiments (totally 11 experiments) are conducted to examine the impact of each of the aforementioned parameterization schemes on the storm’s track and intensity forecast. Convective parameterization schemes (CPS) include Grell (Gr), Betts–Miller (BM) and updated Kain–Fritsch (KF2); planetary boundary layer (PBL) schemes include Burk–Thompson (BT), Eta Mellor–Yamada (MY) and the Medium-Range Forecast (MRF); and cloud microphysics parameterization schemes (MPS) comprise Warm Rain (WR), Simple Ice (SI), Mixed Phase (MP), Goddard Graupel (GG), Reisner Graupel (RG) and Schultz (Sc). The model configuration for CPS and PBL experiments includes two nested domains (90- and 30-km resolution), and for MPS experiments includes three nested domains (90-, 30- and 10-km grid resolution). It is found that the forecast track and intensity of the cyclone are most sensitive to CPS compared to other physical parameterization schemes (i.e., PBL and MPS). The simulated cyclone with Gr scheme has the least forecast track error, and KF2 scheme has highest intensity. From the results, influence of cumulus convection on steering flow of the cyclone is evident. It appears that combined effect of midlatitude trough interaction, strength of the anticyclone and intensity of the storm in each of these model forecasts are responsible for the differences in respective track forecast of the cyclone. The PBL group of experiments has less influence on the track forecast of the cyclone compared to CPS. However, we do note a considerable variation in intensity forecast due to variations in PBL schemes. The MY scheme produced reasonably better forecast within the group with a sustained warm core and better surface wind fields. Finally, results from MPS set of experiments demonstrate that explicit moisture schemes have profound impact on cyclone intensity and moderate impact on cyclone track forecast. The storm produced from WR scheme is the most intensive in the group and closer to the observed strength. The possible reason attributed for this intensification is the combined effect of reduction in cooling tendencies within the storm core due to the absence of melting process and reduction of water loading in the model due to absence of frozen hydrometeors in the WR scheme. We also note a good correlation between evolution of frozen condensate and storm intensification rate among these experiments. It appears that the Sc scheme has some systematic bias and because of that we note a substantial reduction in the rain water formation in the simulated storm when compared to others within the group. In general, it is noted that all the sensitivity experiments have a tendency to unrealistically intensify the storm at the later part of the integration phase.     

Investigating origin of the inadequate medium range predictability of the lower tropospheric ultra-long waves in tropics

The predictability of planetary or ultra-long scale waves is limited by the large growth of errors in these scales in almost all the medium range forecast models. Understanding the cause for the enormous build up of error is, therefore, a necessary task for improving the prediction of planetary waves. A diagnostic analysis of the systematic error energetics has been performed in the Global Forecast System model to investigate the reasons for poor predictability of the lower tropospheric ultra-long waves (wavenumber bands 1–4) in tropics using the analysis-forecast system of horizontal wind field at 850 hPa level during the boreal summer period. For this purpose, systematic error energy is computed in spatial as well as in wavenumber domain. Non-linear inter-scale transfer of error has been formulated and evaluated through energy exchanges among participating triads. The study reveals that the error is generated in the prognostic model initially with a small magnitude at the different locations around tropical convergence zone (TCZ) attributed to the inaccuracy in representing different physical processes like cumulus convection applied in the model. At subsequent evolution of forecasts, error increases and spreads along the TCZ due to its non-linear advection to the higher scales and eventually to the ultra-long scales attributed to the inherent dynamics of the model evaluated through the process of wave-wave exchange of error energy in terms of the triad interactions. The continuous generation and then, non-linear propagation of error up to the planetary scales in the course of prediction increase the uncertainty in ultra-long scales which actually inhibit to predict accurately the planetary scale waves in tropics during medium range forecasts. This work suggests caveats to the modeler’s community in the predictability study of tropical ultra-long waves.     

One step back for a leap forward: toward operational measurements of elements at risk

This study presents the evaluation of 1 year of operational lightning forecasts provided for Europe, using the Weather Research and Forecasting model coupled with a cloud-top height-based lightning parameterization scheme. Three different convective parameterization schemes were employed for parameterizing sub-grid cloud-top heights and consequently driving the lightning scheme. Triggering of the lightning scheme was controlled by means of a model-resolved microphysics-based masking filter, while the formulation for deriving lightning flash rates was also modified, assuming a single “marine” equation instead of the original equations discriminating between continental and marine lightning. Gridded lightning observations were used for evaluating model performance on a dichotomous decision basis. Analysis showed that the lightning scheme is sensitive to the parameterization of convection. In particular, the Kain–Fritsch convective scheme was found to outperform the Grell–Devenyi and Grell–Freitas schemes, showing a statistically significant better performance with respect to lightning prediction. This was most evident during the warm season, while smaller differences among the schemes were recorded during the cold season. Further, for all examined convective schemes, it was found that the application of the masking filter is desirable for improving model performance in terms of lightning forecasting. Last, the reported results revealed that the refinement of the formulation of the lightning parameterization scheme, adhering to a “global” marine equation instead of distinguishing between land and sea lightning, may be necessary in order to obtain reliable lightning forecasts.     

Variation of coastal atmospheric boundary layer characteristics with convective activity along the west coast of India during the Arabian Sea Monsoon Experiment (ARMEX) 2002

This paper investigates the characteristic features of the coastal atmospheric boundary layer (CABL) along the west coast of India during the south-west monsoon (SWM) 2002. Extensive surface and upper-air findings were obtained during the same period from the Arabian Sea Monsoon Experiment (ARMEX; 15th June to 15th August 2002) 2002. The operational general circulation model (GCM) of the National Centre for Medium Range Weather Forecasting (NCMRWF) was used in this study to see the spatial variation of the CABL during two specific convective episodes that led to heavy rainfall along the west coast of India. The impact of a non-local closure (NLC) scheme employed in the NCMRWF GCM was carried out in simulating the CABL. The same episodes were also simulated using a similar parameterization scheme employed in the high resolution mesoscale modelling system (MM5). The diurnal variation of CABL is better represented from MM5 simulation. Comparing the MM5 simulation with that of the coarser grid NCMRWF GCM, we observed that the NCMRWF GCM underestimates the values of both latent heat flux (LHF) and the coastal atmospheric boundary layer height (CABLH). Results from MM5 therefore indicate that the best way to move forward in addressing the short-comings of coarse grid-scale GCMs is to provide a parameterization of the diurnal effects associated with convection processes.     

Application and Verification of Joint Probability Method in Potential Forecast for Severe Convective Weather in Anhui Province

In consideration of large uncertainties in severe convective weather forecast, ensemble forecasting is a dynamic method developed to quantitatively estimate forecast uncertainty. Based on ensemble output, joint probability is a post-processing method to delineate key areas where weather event may actually occur by taking account of the uncertainty of several important physical parameters. An investigation of the environments of little rainfall convection and strong rainfall convection from April to September (warm season) during 2009-2015 was presented using daily disastrous weather data, precipitation data of 80 stations in Anhui province and NCEP Final Analysis (FNL) data. Through ingredients-based forecasting methodology and statistical analysis,four convective parameters characterizing two types of convection were obtained, respectively, which were used to establish joint probability forecasting together with their corresponding thresholds. Using the ECMWF ensemble forecast and observations from April to September during 2016-2017, systematic verification mainly based on ROC and case study of different weather processes were conducted. The results demonstrate that joint probability method is capable of discriminating little rainfall convection and non-convection with comparable performance for different lead times, which is more favorable to identifying the occurrence of strong rainfall convection. The joint probability of little rainfall convection is a good indication for the occurrence of regional or local convection, but may produce some false alarms. The joint probability of strong rainfall convection is good at indicating regional concentrated short-term heavy precipitation as well as local heavy rainfall. There are also individual missing reports in this method, and in practice, 10% can be roughly used as joint probability threshold to achieve relative high TS score. Overall, ensemble-based joint probability method can provide practical short-term probabilistic guidance for severe convective weather.     

Mesoscale model forecast verification during monsoon 2008

There have been very few mesoscale modelling studies of the Indian monsoon, with focus on the verification and intercomparison of the operational real time forecasts. With the exception of Das et al (2008), most of the studies in the literature are either the case studies of tropical cyclones and thunderstorms or the sensitivity studies involving physical parameterization or climate simulation studies. Almost all the studies are based on either National Center for Environmental Prediction (NCEP), USA, final analysis fields (NCEP FNL) or the reanalysis data used as initial and lateral boundary conditions for driving the mesoscale model.     

Research Progress on Physical Parameterization Schemes in Numerical Weather Prediction Models

Atmospheric physics in numerical weather prediction model which predominantly determines the evolution of atmospheric processes is mainly described by physical parameterization. As a result, the development of physical parameterization has been a hot research issue in the area of numerical prediction for a long time. In this regard, the theoretical background and history of physical parameterization schemes for convection, microphysics, and planetary boundary layer, were reviewed in this study. It is suggested that the advance of physical parameterization for the model with high-resolution grid spaces should be considered as a principle issue for numerical model development in the future. Although the gird spaces in current operational numerical models generally decrease toward 10 km owing to the rapid development of high-performance computation, yet most of these schemes are designed for coarse grid spaces. Because of this kind of deficiency, the theoretical basis of these schemes inevitably faces controversy. Directions for development of physical parameterization were also suggested according to the trends of research in numerical prediction.     

订正与集成多模式的中国季度降水预测

针对两个最新换代的季度集合预测系统对中国季度降水预测中存在的系统缺陷,应用改进的贝叶斯联合概率模型（BJP）加以订正。对订正后的单一模式概率预测应用一种混合模型贝叶斯模型平均（BMA）方法加以集成,以综合各模式的优势来提高中国季度降水预测技巧。结果表明:BJP模型可有效地消除集合模式预测的系统偏差,同时大幅提高了概率预测的可靠性。经过订正的欧洲中尺度天气预报中心的 System4预测在许多季度在中国的很大区域范围内都显示出了一定的预测技巧;而澳洲气象局的POAMA2.4预测只在个别季度局部范围内具有技巧。使用BMA对订正后的单一模式预测进行集成可显著提高对中国季度降水预测的精度,相比单一模式预测,技巧得分为正值的网格百分率分别提高了13.3%和20.0%。