Forecast Experiments Research Articles

Strongly Coupled Data Assimilation Experiments with Linearized Ocean–Atmosphere Balance Relationships

Coupled data assimilation is emerging as a target approach for Earth system prediction and reanalysis systems. Coupled data assimilation may be indeed able to minimize unbalanced air–sea initialization and maximize the intermedium propagation of observations. Here, we use a simplified framework where a global ocean general circulation model (NEMO) is coupled to an atmospheric boundary layer model [Cheap Atmospheric Mixed Layer (CheapAML)], which includes prognostic prediction of near-surface air temperature and moisture and allows for thermodynamic but not dynamic air–sea coupling. The control vector of an ocean variational data assimilation system is augmented to include 2-m atmospheric parameters. Cross-medium balances are formulated either through statistical cross covariances from monthly anomalies or through the application of linearized air–sea flux relationships derived from the tangent linear approximation of bulk formulas, which represents a novel solution to the coupled assimilation problem. As a proof of concept, the methodology is first applied to study the impact of in situ ocean observing networks on the near-surface atmospheric analyses and later to the complementary study of the impact of 2-m air observations on sea surface parameters, to assess benefits of strongly versus weakly coupled data assimilation. Several forecast experiments have been conducted for the period from June to December 2011. We find that especially after day 2 of the forecasts, strongly coupled data assimilation provides a beneficial impact, particularly in the tropical oceans. In most areas, the use of linearized air–sea balances outperforms the statistical relationships used, providing a motivation for implementing coupled tangent linear trajectories in four-dimensional variational data assimilation systems. Further impacts of strongly coupled data assimilation might be found by retuning the background error covariances.

Comparison of Nonlinear Local Lyapunov Vectors and Bred Vectors in Estimating the Spatial Distribution of Error Growth

Abstract Instabilities play a critical role in understanding atmospheric predictability and improving weather forecasting. The bred vectors (BVs) are dynamically evolved and flow-dependent nonlinear perturbations, indicating the most unstable modes of the underlying flow. Especially over smaller areas, however, BVs with different initial seeds may to some extent be constrained to a small subspace, missing potential forecast error growth along other unstable perturbation directions. In this paper, the authors study the nonlinear local Lyapunov vectors (NLLVs) that are designed to capture an orthogonal basis spanning the most unstable perturbation subspace, thus potentially ameliorating the limitation of BVs. The NLLVs are theoretically related to the linear Lyapunov vectors (LVs), which also form an orthogonal basis. Like BVs, NLLVs are generated by dynamically evolving perturbations with a full nonlinear model. In simulated forecast experiments, a set of mutually orthogonal NLLVs show an advantage in predicting the structure of forecast error growth when compared to using a set of BVs that are not fully independent. NLLVs are also found to have a higher local dimension, enabling them to better capture localized instabilities, leading to increased ensemble spread.

The Community Leveraged Unified Ensemble (CLUE) in the 2016 NOAA/Hazardous Weather Testbed Spring Forecasting Experiment

AbstractOne primary goal of annual Spring Forecasting Experiments (SFEs), which are coorganized by NOAA’s National Severe Storms Laboratory and Storm Prediction Center and conducted in the National Oceanic and Atmospheric Administration’s (NOAA) Hazardous Weather Testbed, is documenting performance characteristics of experimental, convection-allowing modeling systems (CAMs). Since 2007, the number of CAMs (including CAM ensembles) examined in the SFEs has increased dramatically, peaking at six different CAM ensembles in 2015. Meanwhile, major advances have been made in creating, importing, processing, verifying, and developing tools for analyzing and visualizing these large and complex datasets. However, progress toward identifying optimal CAM ensemble configurations has been inhibited because the different CAM systems have been independently designed, making it difficult to attribute differences in performance characteristics. Thus, for the 2016 SFE, a much more coordinated effort among many collaborators was made by agreeing on a set of model specifications (e.g., model version, grid spacing, domain size, and physics) so that the simulations contributed by each collaborator could be combined to form one large, carefully designed ensemble known as the Community Leveraged Unified Ensemble (CLUE). The 2016 CLUE was composed of 65 members contributed by five research institutions and represents an unprecedented effort to enable an evidence-driven decision process to help guide NOAA’s operational modeling efforts. Eight unique experiments were designed within the CLUE framework to examine issues directly relevant to the design of NOAA’s future operational CAM-based ensembles. This article will highlight the CLUE design and present results from one of the experiments examining the impact of single versus multicore CAM ensemble configurations.

Predictability and Prediction of Low‐Frequency Rainfall Over the Lower Reaches of the Yangtze River Valley on the Time Scale of 20 to 30 days

AbstractThis paper presents a predictability study of the 20–30‐day low‐frequency rainfall over the lower reaches of the Yangtze River valley (LYRV). This study relies on an extended complex autoregressive (ECAR) model method, which is based on the principal components of the global 850 hPa low‐frequency meridional wind. ECAR is a recently advanced climate forecast method, based on data‐driven models. It not only reflects the lagged variations information between the leading low‐frequency components of the global circulation and rainfall in a complex space, but also displays the ability to describe the synergy variations of low‐frequency components of a climate system in a low dimensional space. A 6‐year forecast experiment is conducted on the low‐frequency rainfall over the LYRV for the extended‐range daily forecasts during 2009–2014, based on the time‐varying high‐order ECAR. These experimental results demonstrate that the useful skills of the real‐time forecasts are achieved for an extended lead‐time up to 28 days with a fifth‐order model, and are also shown to be 27‐day lead for forecasts which are initiated from weak intraseasonal oscillation (ISO). This high‐order ECAR displays the ability to significantly improve the predictions of the ISO. The analysis of the 20–30‐day ISO predictability reveals a predictability limit of about 28–40 days. Therefore, the forecast framework used in this study is determined to have the potential to assist in improving the real‐time forecasts for the 20–30‐day oscillations related to the heavy rainfall over the LYRV in summer.

An adaptive middle and long-term runoff forecast model using EEMD-ANN hybrid approach

It remains a challenge to obtain an accurate middle and long-term runoff forecast, especially in flood seasons. The forecast performance can be improved using ensemble empirical mode decomposition (EEMD) to produce cleaner signals as model inputs. In many EEMD based forecast models, the entire time series are decomposed into several sub-series, and each sub-series is divided into calibration and validation datasets and forecasted by some common models, such as artificial neural network (ANN), and finally an ensemble forecast is obtained by summing the forecasted results of each sub-series. In such a decomposition-ensemble framework, some future information is used, and thus it is a hindcast experiment. Attempts have also been made to propose a real forecast experiment, which, however, often cannot be able to adapt to the non-stationary changes in runoff due to the lack of adaptive ability. Therefore, this study tries to improve the decomposition-ensemble framework and propose an adaptive middle and long-term runoff forecast model especially for flood seasons. Unlike a hindcast experiment, it does not use any future information; and unlike conventional forecast experiments, its decomposition and forecast models are adjusted adaptively as long as new runoff information is added. EEMD is used to decompose the original time series and ANN is used to forecast each sub-series, hence the name an adaptive EEMD-ANN (AEEMD-ANN) model. This model is applied to forecast the 1-month ahead streamflow of three stations in China, and the results show that the AEEMD-ANN model can improve forecast accuracy in flood seasons, but it is not as good as ANN, adaptive neuro-fuzzy inference system (ANFIS), support vector machine (SVM) and seasonal first-order autoregressive (SAR (1)) models in dry seasons. While the conventional forecast models, especially the SAR (1) models, are more suitable in dry season. Therefore, it is recommended to use SAR (1) model in dry season and AEEMD-ANN model in flood season to forecast the monthly runoff in Yangtze River Basin.

Forecasting Imports with Information from Abroad

Globalization has led to huge increases in import volumes, increasing the importance of imports for total output. Since imports are a volatile component, they are difficult to forecast and strongly influence the forecast accuracy of gross domestic product. We introduce the first leading indicator constructed to forecast import growth, the Import Climate. It builds on the idea that the import demand of the domestic country should be reflected in the expected export developments of its main trading partners. A foreign country’s expected exports are, in turn, determined by its trading partners’ business and consumer confidence and its own price competitiveness. In a real-time forecasting experiment, the Import Climate outperforms standard business cycle indicators at short horizons for France, Germany, Italy, and the United States for the first release of data. For Spain and the United Kingdom, our indicator works particularly well with the latest vintage of data.

Прогнозирование текущей динамики ВВП на основе данных поисковых запросов

In order to conduct a conscious economic policy, timely assessment of the main economic indicators, viz GDP, is needed. In this paper the authors justify the reason why an inclusion of query search data may help to produce better nowcasts compared to the currently used Russian GDP models, which are built on the data from official statistical services. The authors also check, whether this hypothesis holds true in the real-time forecasting experiment. For this purpose the authors suggest two competing dynamic factor models: the one, which includes data on the query search frequency, and the other one, which excludes it. The models show that the inclusion of query search data does not change the forecast performance of the model built only upon official economic indicators. At the same time, both models have produced more accurate nowcasts of Russian GDP then AR(1) model did. Finally, the authors try to explain the resulting irrelevance of query search data in nowcasting GDP. The article discusses both fundamental reasons and the pitfalls of the methodology used in this paper, which could have led to such result.

Using NWP to assess the influence of the Arctic atmosphere on midlatitude weather and climate

The influence of the Arctic atmosphere on Northern Hemisphere midlatitude tropospheric weather and climate is explored by comparing the skill of two sets of 14-day weather forecast experiments using the ECMWF model with and without relaxation of the Arctic atmosphere towards ERA-Interim reanalysis data during the integration. Two pathways are identified along which the Arctic influences midlatitude weather: a pronounced one over Asia and Eastern Europe, and a secondary one over North America. In general, linkages are found to be strongest (weakest) during boreal winter (summer) when the amplitude of stationary planetary waves over the Northern Hemisphere is strongest (weakest). No discernible Arctic impact is found over the North Atlantic and North Pacific region, which is consistent with predominantly southwesterly flow. An analysis of the flow-dependence of the linkages shows that anomalous northerly flow conditions increase the Arctic influence on midlatitude weather over the continents. Specifically, an anomalous northerly flow from the Kara Sea towards West Asia leads to cold surface temperature anomalies not only over West Asia but also over Eastern and Central Europe. Finally, the results of this study are discussed in the light of potential midlatitude benefits of improved Arctic prediction capabilities.

Potential predictability and forecast skill in ensemble climate forecast: a skill-persistence rule

This study investigates the factors relationship between the forecast skills for the real world (actual skill) and perfect model (perfect skill) in ensemble climate model forecast with a series of fully coupled general circulation model forecast experiments. It is found that the actual skill for sea surface temperature (SST) in seasonal forecast is substantially higher than the perfect skill on a large part of the tropical oceans, especially the tropical Indian Ocean and the central-eastern Pacific Ocean. The higher actual skill is found to be related to the higher observational SST persistence, suggesting a skill-persistence rule: a higher SST persistence in the real world than in the model could overwhelm the model bias to produce a higher forecast skill for the real world than for the perfect model. The relation between forecast skill and persistence is further proved using a first-order autoregressive model (AR1) analytically for theoretical solutions and numerically for analogue experiments. The AR1 model study shows that the skill-persistence rule is strictly valid in the case of infinite ensemble size, but could be distorted by sampling errors and non-AR1 processes. This study suggests that the so called “perfect skill” is model dependent and cannot serve as an accurate estimate of the true upper limit of real world prediction skill, unless the model can capture at least the persistence property of the observation.

A convection-allowing ensemble forecast based on the breeding growth mode and associated optimization of precipitation forecast

A convection-allowing ensemble forecast experiment on a squall line was conducted based on the breeding growth mode (BGM). Meanwhile, the probability matched mean (PMM) and neighborhood ensemble probability (NEP) methods were used to optimize the associated precipitation forecast. The ensemble forecast predicted the precipitation tendency accurately, which was closer to the observation than in the control forecast. For heavy rainfall, the precipitation center produced by the ensemble forecast was also better. The Fractions Skill Score (FSS) results indicated that the ensemble mean was skillful in light rainfall, while the PMM produced better probability distribution of precipitation for heavy rainfall. Preliminary results demonstrated that convection-allowing ensemble forecast could improve precipitation forecast skill through providing valuable probability forecasts. It is necessary to employ new methods, such as the PMM and NEP, to generate precipitation probability forecasts. Nonetheless, the lack of spread and the overprediction of precipitation by the ensemble members are still problems that need to be solved.

Assimilation Impact of Different GPS Analysis Methods on Precipitation Forecast: A Heavy Rainfall Case Study of Kani City, Gifu Prefecture on July 15, 2010

In this study, we examined variations in predicted precipitable water produced from different Global Positioning System (GPS) zenith delay methods, and assessed the corresponding difference in predicted rainfall after assimilating the obtained precipitable water data. Precipitable water data estimated from the GPS and three-dimensional horizontal wind velocity field derived from the X-band dual polarimetric radar were assimilated in CReSS and rainfall forecast experiments were conducted for the heavy rainfall system in Kani City, Gifu Prefecture on July 15, 2010. In the GPS analysis, a method to simultaneously estimate coordinates and zenith delay, i.e., the simultaneous estimation method, and a method to successively estimate coordinates and zenith delay, i.e., the successive estimation method, were used to estimate precipitable water. The differences generated from using predicted orbit data provided in pseudo-real time from the International GNSS (Global Navigation Satellite System) Service for geodynamics (IGS) versus precise orbit data released after a 10-day delay were examined. The change in precipitable water due to varying the analysis methods was larger than that due to the type of satellite orbit information. In the rainfall forecast experiments, those using the successive estimation method results had a better precision than those using the simultaneous estimation method results. Both methods that included data assimilation had higher rainfall forecast precisions than the forecast precision without precipitable water assimilation. Water vapor obtained from GPS analysis is accepted as important in rainfall forecasting, but the present study showed additional improvements can be attained from incorporating a zenith delay analysis method.

Assimilating Doppler radar observations with an ensemble Kalman filter for convection-permitting prediction of convective development in a heavy rainfall event during the pre-summer rainy season of south China

This study examines the effectiveness of an ensemble Kalman filter based on the weather research and forecasting model to assimilate Doppler-radar radial-velocity observations for convection-permitting prediction of convection evolution in a high-impact heavy-rainfall event over coastal areas of South China during the pre-summer rainy season. An ensemble of 40 deterministic forecast experiments (40 DADF) with data assimilation (DA) is conducted, in which the DA starts at the same time but lasts for different time spans (up to 2 h) and with different time intervals of 6, 12, 24, and 30 min. The reference experiment is conducted without DA (NODA).To show more clearly the impact of radar DA on mesoscale convective system (MCS) forecasts, two sets of 60-member ensemble experiments (NODA EF and exp37 EF) are performed using the same 60-member perturbed-ensemble initial fields but with the radar DA being conducted every 6 min in the exp37 EF experiments from 0200 to 0400 BST. It is found that the DA experiments generally improve the convection prediction. The 40 DADF experiments can forecast a heavy-rain-producing MCS over land and an MCS over the ocean with high probability, despite slight displacement errors. The exp37 EF improves the probability forecast of inland and offshore MCSs more than does NODA EF. Compared with the experiments using the longer DA time intervals, assimilating the radial-velocity observations at 6-min intervals tends to produce better forecasts. The experiment with the longest DA time span and shortest time interval shows the best performance. However, a shorter DA time interval (e.g., 12 min) or a longer DA time span does not always help. The experiment with the shortest DA time interval and maximum DA window shows the best performance, as it corrects errors in the simulated convection evolution over both the inland and offshore areas. An improved representation of the initial state leads to dynamic and thermodynamic conditions that are more conducive to earlier initiation of the inland MCS and longer maintenance of the offshore MCS.

Hurricane Harvey provides lab for U.S. forecast experiments.

New model and satellites test their mettle against the storm.

Hurricane Harvey provides lab for U.S. forecast experiments

Hurricane Harvey provides lab for U.S. forecast experiments

The Role of Forcings in the Twentieth-Century North Atlantic Multidecadal Variability: The 1940–75 North Atlantic Cooling Case Study

Abstract Results from a study inspecting the origins of multidecadal variability in the North Atlantic sea surface temperature (NASST) are presented. The authors target in particular the 1940–75 “warm-to-cold” transition, an event that is generally framed in the context of the longer-term Atlantic multidecadal variability (AMV) cycle, in turn associated with the Atlantic meridional overturning circulation (AMOC) internal variability. Here the authors examine the ability of uninitialized, historical integrations from the phase 5 of the Coupled Model Intercomparison Project (CMIP5) archive to retrospectively reproduce this specific episode of twentieth-century climatic history, under a hierarchy of forcing conditions. For this purpose, both standard and so-called historical Misc CMIP5 simulations of the historical climate (combining selected natural and anthropogenic forcings) are exploited. Based on this multimodel analysis, evidence is found for a significant influence of anthropogenic agents on multidecadal sea surface temperature (SST) fluctuations across the Atlantic sector, suggesting that anthropogenic aerosols and greenhouse gases might have played a key role in the 1940–75 North Atlantic cooling. However, the diagnosed forced response in CMIP5 models appears to be affected by a large uncertainty, with only a limited subset of models displaying significant skill in reproducing the mid-twentieth-century NASST cooling. Such uncertainty originates from the existence of well-defined behavioral clusters within the analyzed CMIP5 ensembles, with the bulk of the models splitting into two main clusters. Such a strong polarization calls for some caution when using a multimodel ensemble mean in climate model analyses, as averaging across fairly distinct model populations may result, through mutual cancellation, in a rather artificial description of the actual multimodel ensemble behavior. A potentially important role for both anthropogenic aerosols and greenhouse gases with regard to the observed North Atlantic multidecadal variability has clear implications for decadal predictability and predictions. The uncertainty associated with alternative aerosol and greenhouse gas emission scenarios should be duly accounted for in designing a common protocol for coordinated decadal forecast experiments.

The HIRLAM fast radiation scheme for mesoscale numerical weather prediction models

Abstract. This paper provides an overview of the HLRADIA shortwave (SW) and longwave (LW) broadband radiation schemes used in the HIRLAM numerical weather prediction (NWP) model and available in the HARMONIE-AROME mesoscale NWP model. The advantage of broadband, over spectral, schemes is that they can be called more frequently within the model, without compromising on computational efficiency. In mesoscale models fast interactions between clouds and radiation and the surface and radiation can be of greater importance than accounting for the spectral details of clear-sky radiation; thus calling the routines more frequently can be of greater benefit than the deterioration due to loss of spectral details. Fast but physically based radiation parametrizations are expected to be valuable for high-resolution ensemble forecasting, because as well as the speed of their execution, they may provide realistic physical perturbations. Results from single-column diagnostic experiments based on CIRC benchmark cases and an evaluation of 10 years of radiation output from the FMI operational archive of HIRLAM forecasts indicate that HLRADIA performs sufficiently well with respect to the clear-sky downwelling SW and longwave LW fluxes at the surface. In general, HLRADIA tends to overestimate surface fluxes, with the exception of LW fluxes under cold and dry conditions. The most obvious overestimation of the surface SW flux was seen in the cloudy cases in the 10-year comparison; this bias may be related to using a cloud inhomogeneity correction, which was too large. According to the CIRC comparisons, the outgoing LW and SW fluxes at the top of atmosphere are mostly overestimated by HLRADIA and the net LW flux is underestimated above clouds. The absorption of SW radiation by the atmosphere seems to be underestimated and LW absorption seems to be overestimated. Despite these issues, the overall results are satisfying and work on the improvement of HLRADIA for the use in HARMONIE-AROME NWP system is ongoing. In a HARMONIE-AROME 3-D forecast experiment we have shown that the frequency of the call for the radiation parametrization and choice of the parametrization scheme makes a difference to the surface radiation fluxes and changes the spatial distribution of the vertically integrated cloud cover and precipitation.

Comparison of the Analyses and Forecasts of a Tornadic Supercell Storm from Assimilating Phased-Array Radar and WSR-88D Observations

Abstract NOAA’s National Severe Storms Laboratory is actively developing phased-array radar (PAR) technology, a potential next-generation weather radar, to replace the current operational WSR-88D radars. One unique feature of PAR is its rapid scanning capability, which is at least 4–5 times faster than the scanning rate of WSR-88D. To explore the impact of such high-frequency PAR observations compared with traditional WSR-88D on severe weather forecasting, several storm-scale data assimilation and forecast experiments are conducted. Reflectivity and radial velocity observations from the 22 May 2011 Ada, Oklahoma, tornadic supercell storm are assimilated over a 45-min period using observations from the experimental PAR located in Norman, Oklahoma, and the operational WSR-88D radar at Oklahoma City, Oklahoma. The radar observations are assimilated into the ARPS model within a heterogeneous mesoscale environment and 1-h ensemble forecasts are generated from analyses every 15 min. With a 30-min assimilation period, the PAR experiment is able to analyze more realistic storm structures, resulting in higher skill scores and higher probabilities of low-level vorticity that align better with the locations of radar-derived rotation compared with the WSR-88D experiment. Assimilation of PAR observations for a longer 45-min time period generates similar forecasts compared to assimilating WSR-88D observations, indicating that the advantage of rapid-scan PAR is more noticeable over a shorter 30-min assimilation period. An additional experiment reveals that the improved accuracy from the PAR experiment over a shorter assimilation period is mainly due to its high-temporal-frequency sampling capability. These results highlight the benefit of PAR’s rapid-scan capability in storm-scale modeling that can potentially extend severe weather warning lead times.

Comparison of Next-Day Probabilistic Severe Weather Forecasts from Coarse- and Fine-Resolution CAMs and a Convection-Allowing Ensemble

Abstract Given increasing computing power, an important question is whether additional computational resources would be better spent reducing the horizontal grid spacing of a convection-allowing model (CAM) or adding members to form CAM ensembles. The present study investigates this question as it applies to CAM-derived next-day probabilistic severe weather forecasts created by using forecast updraft helicity as a severe weather proxy for 63 days of the 2010 and 2011 NOAA Hazardous Weather Testbed Spring Forecasting Experiments. Forecasts derived from three sets of Weather Research and Forecasting Model configurations are tested: a 1-km deterministic model, a 4-km deterministic model, and an 11-member, 4-km ensemble. Forecast quality is evaluated using relative operating characteristic (ROC) curves, attributes diagrams, and performance diagrams, and forecasts from five representative cases are analyzed to investigate their relative quality and value in a variety of situations. While no statistically significant differences exist between the 4- and 1-km deterministic forecasts in terms of area under ROC curves, the 4-km ensemble forecasts offer weakly significant improvements over the 4-km deterministic forecasts over the entire 63-day dataset. Further, the 4-km ensemble forecasts generally provide greater forecast quality relative to either of the deterministic forecasts on an individual day. Collectively, these results suggest that, for purposes of improving next-day CAM-derived probabilistic severe weather forecasts, additional computing resources may be better spent on adding members to form CAM ensembles than on reducing the horizontal grid spacing of a deterministic model below 4 km.

Unemployment Rates Forecasts – Unobserved Component Models Versus SARIMA Models In Central And Eastern European Countries

In this paper we compare the accuracy of unemployment rates forecasts of eight Central and Eastern European countries. The unobserved component models and seasonal ARIMA models are used within a rolling short-term forecast experiment as an out-of-sample test of forecast accuracy. We find that unemployment rates present clear unconditional asymmetry in three out of eight countries. Half the cases there is no difference between forecasting accuracy of the methods used in the study. In the remaining, a proper specification of seasonal ARIMA model allows to generate better forecasts than from unobserved component models. The forecasting accuracy deteriorates in periods of rapid upward and downward movement and improves in periods of gradual change in the unemployment rates.

An ensemble forecast method combining CNOPs with BVs

Numerical forecasting can estimate the future state of the atmosphere or oceans. Because the atmosphere and oceans are complex and nonlinear systems, very small errors in the initial field can be amplified nonlinearly during integration of the numerical model, which may ultimately result in large errors. The true initial field cannot be obtained, and it is not possible to construct a perfect model that simulates the development of the atmospheric or oceanic state. Therefore, initial uncertainties and system instabilities result in large forecast errors. However, ensemble forecasting can be used to estimate the forecasting uncertainty. The ensemble initial perturbation has a crucial impact on the ensemble forecasting result. To obtain the initial perturbation that can capture initial field uncertainties and system instabilities, conditional nonlinear optimal perturbation (CNOP) is applied to generate initial perturbations for ensemble forecasting. CNOP is the initial perturbation that satisfies certain physical constraints and causes the largest forecast error at forecast time. More specifically, it is a generalization of the singular vector (SV) in nonlinear field. Using the Lorenz-96 model, we generate initial perturbations by combining CNOPs and breeding vectors (BVs), and then compare it with the traditional breeding growing modes (BGM) method. The spectral projected gradient 2 (SPG2) optimal algorithm is used to compute CNOPs based on basic states, and we sort the BVs according to the corresponding L 2 norm magnitude. We design two ensemble samples for the forecast experiments: sample 1 (S1) and sample 2 (S2). Ensemble initial perturbations all consist of BVs in S1. CNOPs and BVs are combined to form ensemble initial perturbations in S2; we use CNOPs to replace the several larger BVs in S1 according to the BVs’ L 2 norm magnitude order, and keep the remaining BVs unchanged. Both S1 and S2 are consist of 17 ensemble members, including one control forecast. Under the perfect model assumption, many ensemble forecasting numerical experiments are run on the Lorenz-96 model. The root mean square error (RMSE) and anomaly correlation coefficient (ACC) are used to evaluate the forecast results. We find that the forecast skill of S2 is higher than that of S1 in most experiments. This result suggests that initial perturbations generated using the new method better overcome the adverse impact from initial field uncertainty and system instability, and provide higher ensemble forecasting skill than the BGM method. When the initial analysis errors are fast-growing, ensemble forecasts generated by the new method can improve the control forecasts results. Furthermore, the rate of improvement increases with increasing forecast time, and is more obvious in the medium range (6–14 d). Ensemble forecasts generated using the new method can extend the effective forecast time about four days relative to the control forecasts based on the standard that the effective ACC forecast is not <0.6. Nevertheless, for slowly-growing initial analysis errors, ensemble forecasts generated using the new and BGM method may be worse than the control forecasts. The initial analysis errors that show much faster growth behavior are easily influenced by nonlinearity, and extreme events due to strong nonlinear instability may be more likely to cause fast growth of initial analysis errors. Therefore, ensemble forecasts generated by combining CNOPs and BVs will have higher forecasting skill than the BGM method, and can improve the forecast effect of control forecast.