European Centre For Medium-Range Weather Forecasts Model Research Articles

Evaluation of Daily Temperature Extremes in the ECMWF Operational Weather Forecasts and ERA5 Reanalysis

In weather forecasting and climate monitoring, daily maximum and minimum air temperatures (TMAX and TMIN) are fundamental for operational and research purposes, from early warning of extreme events to climate change studies. This study provides an integrated evaluation of TMAX and TMIN from two European Centre for Medium-range Weather Forecasts (ECMWF) products: ERA5 reanalysis (1980–2019) and operational weather forecasts (2017–2021). Both products are evaluated using in situ observations from the Global Historical Climatology Network (GHCN). While the analyses span globally, emphasis is given to four key regions: Europe, East and West United States, and Australia. Results reveal a general underestimation of TMAX and overestimation of TMIN in both operational forecasts and ERA5, highlighting the limitation of the ECMWF model in estimating the amplitude of the diurnal cycle of air temperature. ERA5′s accuracy has improved over the past decade, due to enhanced constrain of land–atmosphere analysis streaming from more and higher-quality satellite data. Furthermore, ERA5 outperforms one-day-ahead weather forecasts, indicating that non-real-time dependent studies should rely on ERA5 instead of real-time operational forecasts. This study underscores the importance of ongoing research in model and data assimilation, considering the relevance of daily temperature extremes forecasting and reanalysis for operational meteorology and climate monitoring.

Evaluation of Tropopause Height from Sentinel-6 GNSS Radio Occultation Using Different Methods

The tropopause is described as the distinction between the troposphere and the stratosphere, and the tropopause height (TPH) is an indicator of climate change. GNSS Radio Occultation (RO) can monitor the atmosphere globally under all weather conditions with a high vertical resolution. In this study, four different techniques for identifying the TPH were investigated. The lapse rate tropopause (LRT) and cold point tropopause (CPT) methods are the traditional methods for determining the TPH based on temperature profiles according to the World Meteorological Organization (WMO) definition. Two advanced methods based on the covariance transform (CT) method are used to estimate the TPH from the refractivity (TPHN) and the TPH from the bending angle (TPHα). Data from the Sentinel-6 satellite were used to evaluate the different algorithms for the identification of the TPH. The analysis shows that the CPT height is greater than the LRT height and that the CPT is only valid in tropical regions. The CPT height, TPHN, and TPHα were compared with the LRT height. In general, the TPHα had the largest value, followed by the TPHN, and the LRT had the lowest value of TPH at and near the equator. In the equatorial region, the maximum TPH results from the TPHα (approximately 17.5 km), and at the poles, the minimum TPH results from the LRT (approximately 9 km). The results were also compared with the European Center for Medium-Range Weather Forecasts (ECMWF), and there was a strong correlation of 0.999 between the different approaches for identifying the TPH from RO and the ECMWF model. The identification of the TPH is critical for the transfer of mass, water, and trace gases between the troposphere and stratosphere.

The impact of Aeolus winds on near-surface wind forecasts over tropical ocean and high-latitude regions

Abstract. To detect global wind profiles and improve numerical weather prediction (NWP), the European Space Agency (ESA) launched the Aeolus satellite carrying a spaceborne Doppler wind lidar in 2018. After the successful launch, the European Centre for Medium-Range Weather Forecasts (ECMWF) performed the observing system experiments (OSEs) to evaluate the contribution of Aeolus data to NWP. This study aims to assess the impact of Aeolus wind assimilation in the ECMWF model on near-surface (10 m height) wind forecasts over tropical ocean regions by taking buoy measurements for reference and over high-latitude regions by taking weather station data for reference for the year 2020. The assessments were conducted mainly through inter-comparison analysis. The results show that Aeolus data assimilation has a limited impact on sea surface wind forecasts for tropical regions when compared with buoy measurements. For the high-latitude regions in the Northern Hemisphere, Aeolus is able to improve near-surface wind forecasts. This positive impact is more evident as the forecast time step is extended, during the first half year of 2020 and during the winter months. In addition, the v component tends to benefit more from the Aeolus observations than the u component. For the Southern Hemisphere, a few error reductions are observed but exist randomly. Overall, this in situ data-based assessment expands our understanding of the role of Aeolus data assimilation with the global NWP model in predicting near-surface wind for tropical oceans and high-latitude regions.

An Analysis of Prediction Skill of Heat Waves Over India Using TIGGE Ensemble Forecasts

AbstractHeat wave has become a great concern for India in the recent years due to its disastrous impact on various sectors including health. Thus, accurate forecasts of heat wave events well in advance are required for preparing adequate mitigation strategies. The present study assesses the prediction skill of numerical weather prediction models in The Observing system Research and Predictability Experiment Interactive Grand Global Ensemble (TIGGE) experiments for predicting heat waves over India up to 7 days in advance. The models considered for this analysis are; the European Centre for Medium‐Range Weather Forecasts (ECMWF), the UK Met Office (UKMO) and National Centre for Environmental Prediction (NCEP). The model forecast verifications have been carried out for the hot weather season (April–June) over India for the period of 2008–2013. Fourteen heat wave events were identified during the study period using gridded daily maximum temperature (Tmax). The study reveals that the spatial distribution of maximum Temperature is well predicted by the TIGGE models for a forecast lead time of 1–7 days. The analysis suggested that the magnitude of heat wave events, even with a 7 days lead time, can be correctly predicted by more than 80% of ensemble members in all the TIGGE models. The prediction skill of the maximum temperatures over heat wave prone area during heat wave events is higher for ECMWF, then UKMO and NCEP models. The forecast verification analysis thus indicates that the TIGGE models are able to provide early warnings of heat waves with at least 5 days lead time.

Spatial connections in extreme precipitation events obtained from NWP forecasts: A complex network approach

Understanding the spatial variability of extreme precipitation events (EPEs) has always been a challenging task, with climate change further complicating the issue. Many studies have looked into the spatiotemporal characteristics of EPEs and the possibility of them being connected. Most of these studies are based on observation data products which inform about the events after they have occurred. Various Quantitative Precipitation Forecasts (QPFs) obtained from the Numerical Weather Prediction (NWP) models such as those of the European Centre for Medium-range Weather Forecasts (ECMWF), Japan Meteorological Agency (JMA), National Centre for Medium-Range Weather Forecasting (NCMRWF), and UK Met Office (UKMO) are available which inform about the precipitation events before they have occurred. In this study, we inter-evaluate four gridded (deterministic) QPFs over the Ganga River basin of India for their ability to detect the spatial connections among EPEs during the monsoon months, i.e., June, July, August and September (JJAS). Several experimental runs are performed in multiple time periods (13, 6 and 4 years) at various percentile thresholds (85th, 90th and 95th percentile) at two spatial resolutions of 25 and 50-km. Moreover, we also evaluate the forecast patterns of the QPF obtained from the national agency of India, i.e., NCMRWF and analyse its performance with respect to the other three selected QPFs from international agencies. For the observation/reference dataset, the gridded Indian Meteorological Department (IMD) dataset is selected. We first use the deterministic and dichotomous error statistics to evaluate the performance of NWP forecasts. Further, we employ the theory of complex networks to create networks of extreme precipitation at each grid. With this work, we aim to answer (a) can the theory of complex networks be a feasible tool for evaluating the performance of NWP forecasts in detecting extreme precipitation behaviour; (b) does the change in temporal length of the NWP forecasts influence the spatial structure of extreme precipitation as observed before, and (c) can a global NWP model, larger in length, be used as a potential substitute to the NCMRWF model. The ECMWF model performs well with respect to IMD as the average CC for TP_13 is 0.42 while they are 0.31 and 0.34 for TP_6 and TP_4, respectively. Also, we find that ECMWF can be a suitable substitute for the NCMRWF model. The identification of a substitute precipitation forecast, larger in temporal length, can be beneficial as input in hydrological models for streamflow forecasting and estimating parameters for bias correction of precipitation, among others.

A study on loss function against data imbalance in deep learning correction of precipitation forecasts

Sample imbalance has always been an important issue to be solved in heavy precipitation forecasts. Introducing appropriate constraints into the loss function of a machine learning model can mitigate the negative impact of sample imbalance on model training. This study explores the role of Dice loss function in dealing with sample imbalance and verifies its application on the correction of heavy precipitation forecasts using the U-Net neural network. On this basis, the application of ordinal classification in forecasts correction is further verified. The results show that the concept of Dice loss is highly similar to that of threat score, which can suppress the negative impact of sample imbalance on heavy precipitation forecasts by ignoring most of the events that are insensitive to heavy precipitation forecasts. Model correction experiments further indicate that the correction models trained with weighted cross entropy, differentiable threat score loss and Dice loss (WCE, DTS and Dice models) improve the precipitation forecasts of ECMWF (European Centre for Medium-Range Weather Forecasts) model. For the forecasts of all precipitation levels, Dice model effectively suppresses false alarms while improving hits, whereas WCE model greatly promotes false alarms while significantly improving hits. In contrast, DTS model significantly suppresses large precipitation false alarms while improving its hits, but causes substantial small precipitation false alarms while significantly promoting its hits. As a result, for large precipitation forecasts, both Dice and DTS models perform significantly better than WCE model, while for small precipitation forecasts, Dice model performs better than DTS and WCE models. Furthermore, the correction model combining ordinal classification and Dice loss exhibits the best forecasting skill. This model further suppresses the false alarms of large precipitation while improving its hits.

Evaluation of the Forecast Performance for Week-2 Winter Surface Air Temperature from the Model for Prediction Across Scales–Atmosphere (MPAS-A)

Abstract The forecast skill for week-2 wintertime surface air temperature (SAT) over the Northern Hemisphere by the Model for Prediction Across Scales–Atmosphere (MPAS-A) is evaluated and compared with operational forecast systems that participate in the Subseasonal to Seasonal Prediction project (S2S). An intercomparison of the MPAS against the China Meteorological Administration (CMA) model and the European Centre for Medium-Range Weather Forecasts (ECMWF) model was performed using 10-yr reforecasts. Comparing the forecast skill for SAT and atmospheric circulation anomalies at a lead of 2 weeks among the three models, the MPAS shows skill lower than the ECMWF model but higher than the CMA model. The gap in skills between the MPAS model and CMA model is not as large as that between the ECMWF model and MPAS model. Additionally, an intercomparison of the MPAS model against 10 S2S models is presented by using real-time forecasts since 2016 stored in the S2S database. The results show that the MPAS model has forecast skill for week-2 to week-4 wintertime SAT comparable to that in most S2S models. The MPAS model tends to be at an intermediate level compared to current operational forecast models.

Validation of the Aeolus L2B Rayleigh winds and ECMWF short‐range forecasts in the upper troposphere and lower stratosphere using Loon super pressure balloon observations

AbstractThe novel Aeolus satellite, which carries the first Doppler wind lidar providing profiles of horizontal line‐of‐sight (HLOS) winds, addresses a significant gap in direct wind observations in the global observing system. The gap is particularly critical in the tropical upper troposphere and lower stratosphere (UTLS). This article validates the Aeolus Rayleigh–clear wind product and short‐range forecasts of the European Centre for Medium‐Range Weather Forecasts (ECMWF) with highly accurate winds from the Loon super pressure balloon network at altitudes between 16 and 20 km. Data from 229 individual balloon flights are analysed, applying a collocation criterion of 2 hr and 200 km. The comparison of Aeolus and Loon data shows systematic and random errors of 0.31 and 6.37 ms, respectively, for the Aeolus Rayleigh–clear winds. The horizontal representativeness error of Aeolus HLOS winds (nearly the zonal wind component) in the UTLS ranges from 0.6–1.1 ms depending on the altitude. The comparison of Aeolus and Loon datasets against ECMWF model forecasts suggests that the model systematically underestimates the HLOS winds in the tropical UTLS by about 1 ms. While Aeolus winds are currently considered as point winds by the ECMWF data assimilation system, the results of the present study demonstrate the need for a more realistic HLOS wind observation operator for assimilating Aeolus winds.

Improving the prediction of the Madden–Julian Oscillation of the ECMWF model by post-processing

Abstract. The Madden–Julian Oscillation (MJO) is a major source of predictability on the sub-seasonal (10 to 90 d) timescale. An improved forecast of the MJO may have important socioeconomic impacts due to the influence of MJO on both tropical and extratropical weather extremes. Although in the last decades state-of-the-art climate models have proved their capability for forecasting the MJO exceeding the 5-week prediction skill, there is still room for improving the prediction. In this study we use multiple linear regression (MLR) and a machine learning (ML) algorithm as post-processing methods to improve the forecast of the model that currently holds the best MJO forecasting performance, the European Centre for Medium-Range Weather Forecasts (ECMWF) model. We find that both MLR and ML improve the MJO prediction and that ML outperforms MLR. The largest improvement is in the prediction of the MJO geographical location and intensity.

Diagnostic Analysis of Multimodel Rainstorm Forecast for Cases Based on MODE Method

It is very important to analyze the performance of the model rainstorm forecast for understanding the intensity and position deviation of the model precipitation and improving the forecast ability. This paper uses classical scoring and the MODE (Method for Object-Based Diagnostic Evaluation) method to evaluate the forecast performance of different products. The forecast and observation data used in the article mainly include CMA (China Meteorological Administration) multi-source merged precipitation analysis, the precipitation forecast of ECMWF (European Centre for Medium-Range Weather Forecasts), CMA-Meso (Mesoscale model forecast of CMA) and SCMOC (National Meteorological Center grid precipitation forecast guidance product) data. At the same time, the possible correction method of heavy rainfall area is explored by using the high and low-level circulation configuration of the ECMWF model. The main conclusions are as follows: ① MODE spatial verification shows that the number, intensity, area and location of ECMWF rainstorm precipitation objects match the observed precipitation best, which is obviously better than SCMOC and CMA-Meso precipitation forecasts. There are significantly fewer SCMOC rainstorm precipitation objects, and the area of each single precipitation object is significantly larger, which often fails to report the small area objects of convective precipitation. ② The reason for the high TS score of SCMOC is that it reduces the number of small area rainstorm objects and avoids the “double punishment” phenomenon caused by position forecast error, which leads to the failure of SCMOC in local rainstorm forecasts. ③ Analyzing the relationship between the circulation situation of the ECMWF model and the location of rainstorm forecasts by the model, it is found that the location of the rainstorm area is consistent with the upper circulation system, especially with the strong rising area of the vertical velocity of 700 hPa and the high value area of the specific humidity of 850 hPa. When the rainstorm area coincides with the upper air system but is not consistent with the ground convergence area and the high value area of velocity potential, the rainstorm location often has a large deviation. The location of the surface convergence area can be used as a reference to improve the performance of the rainstorm forecast.

Factors determining the subseasonal prediction skill of summer extreme rainfall over southern China

The occurrence of summer extreme rainfall over southern China (SCER) is closely related to the boreal summer intraseasonal oscillation (BSISO), and whether operational models can reasonably predict the BSISO evolution and its modulation on SCER probability is crucial for disaster prevention and mitigation. Here, we find that the skill of subseasonal-to-seasonal (S2S) operational models in predicting the first component of BSISO (BSISO1) might determine the forecast skill of SCER. A systematic assessment is conducted on the reforecast data from two operational models that participated in the S2S project, i.e., the model of European Centre for Medium-Range Weather Forecasts (ECMWF) and the model of China Meteorological Administration (CMA). The results show that the ECMWF model can yield skillful prediction of the BSISO1 index up to 24 days in advance, while the skill of the CMA model is about 10 days. Accordingly, the SCER occurrence is correctly predicted by ECMWF (CMA) model at a forecast lead time of ~ 14 (7) days. The diagnostic results of modeled moisture processes further suggest that the anomalous moisture convergence (advection) induced by the BSISO1 activity serves as the primary (secondary) source of subseasonal predictability of SCER. With better prediction of the moisture convergence anomaly in the specific phases of BSISO1, higher skills can be obtained in the probability prediction of SCER. The present study implies that a further improvement in predicting the BSISO and its related moisture processes is crucial to promoting the subseasonal prediction skill of SCER probability.

Comparison of BLSTM-Attention and BLSTM-Transformer Models for Wind Speed Prediction


 
 
 Accurate estimation of wind speed is essential for many meteorological applications. A novel short-term wind speed prediction method of Bi-directional LSTM and Transformer Network (BLSTM-TRA) model is proposed by combining the Transformer model and LSTM model, and a hybrid model of Bi-directional Long Short-term Memory and Attention Network (BLSTM-ATT) is proposed based on Attention mechanism and LSTM model. The proposed BLSTM-ATT and BLSTM-TRA model are used for predicting the wind speed of seven meteorological stations in Qingdao. In combination with historical ob- servation data, the proposed models outperform the Numerical Weather Prediction (NWP) system of European Centre for Medium-Range Weather Forecasts (ECMWF). By comparing the results of BLSTM-ATT, BLSTM-TRA and ECMWF forecast model, RMSE and MAE of BLSTM-ATT are reduced by 44.7% and 50.3% on average, respectively, as well as an average decrease of 43.0% in the RMSE, an average decrease of 47.4% in the MAE of the BLSTM-TRA model. This demonstrates that the BLSTM-ATT model and the BLSTM-TRA model are more accurate than the ECMWF model in wind speed prediction.
 
 

What do large‐scale patterns teach us about extreme precipitation over the Mediterranean at medium‐ and extended‐range forecasts?

AbstractExtreme precipitation events (EPEs) can have devastating consequences, such as floods and landslides, posing a great threat to society and economy. Predicting such events long in advance can support the mitigation of negative impacts. Here, we focus on EPEs over the Mediterranean, a region that is frequently affected by such hazards. Previous work identified strong connections between localized EPEs and large‐scale atmospheric flow patterns, affecting weather over the entire Mediterranean. We analyse the predictive skill of these patterns in the European Centre for Medium‐Range Weather Forecasts (ECMWF) extended range forecasts and assess if and where these patterns can be used for indirect predictions of EPEs, using the Brier skill score. The results show that the ECMWF model provides skilful predictions of the Mediterranean patterns up to 2 weeks in advance. Moreover, using the forecasted patterns for indirect predictability of EPEs outperforms the reference score up to ∼10 days lead time for many locations. For high orography locations or coastal areas in particular, like parts of western Turkey, western Balkans, Iberian Peninsula, and Morocco, this limit extends to 11–14 days lead time. This study demonstrates that the connections between localized EPEs and large‐scale patterns over the Mediterranean extend the forecasting horizon of the model by over 3 days in many locations, in comparison with forecasting based on the predicted precipitation. Thus, it is beneficial to use the predicted patterns rather than the predicted precipitation at longer lead times for EPE forecasting. The model's performance is also assessed from a user perspective, showing that the EPE forecasting based on the patterns increases the economic benefits at medium/extended range lead times. Such information could support higher confidence in the decision‐making of various users; for example, the agricultural sector and (re)insurance companies.

The Performance of ECMWF Subseasonal Forecasts to Predict the Rainy Season Onset Dates in Vietnam

Abstract The onset of the rainy season is an important date for the mostly rain-fed agricultural practices in Vietnam. Subseasonal to seasonal (S2S) ensemble hindcasts from the European Centre for Medium-Range Weather Forecasts (ECMWF) are used to evaluate the predictability of the rainy season onset dates (RSODs) over five climatic subregions of Vietnam. The results show that the ECMWF model reproduces well the observed interannual variability of RSODs, with a high correlation ranging from 0.60 to 0.99 over all subregions at all lead times (up to 40 days) using five different RSOD definitions. For increasing lead times, forecasted RSODs tend to be earlier than the observed ones. Positive skill score values for almost all cases examined in all subregions indicate that the model outperforms the observed climatology in predicting the RSOD at subseasonal lead times (∼28–35 days). However, the model is overall more skillful at shorter lead times. The choice of the RSOD criterion should be considered because it can significantly influence the model performance. The result of analyzing the highest skill score for each subregion at each lead time shows that criteria with higher 5-day rainfall thresholds tend to be more suitable for the forecasts at long lead times. However, the values of mean absolute error are approximately the same as the absolute values of the mean error, indicating that the prediction could be improved by a simple bias correction. The present study shows a large potential to use S2S forecasts to provide meaningful predictions of RSODs for farmers.

Sparse Gaussian process regression for multi-step ahead forecasting of wind gusts combining numerical weather predictions and on-site measurements

Accurate forecasts of wind gusts are crucially important for wind power generation, severe weather warnings, and the regulation of vehicle speed. To improve the short-term and long-term forecasting accuracy, the sparse Gaussian process regression (GPR) model that reduces the complexity of full GPR is employed for wind gust forecasting by combining numerical weather prediction (NWP) data and on-site measurements. Historical measurements of wind gusts and the European Centre for Medium-Range Weather Forecasts (ECMWF) data are used as inputs of sparse GPR models. In particular, the historical wind gust input allows the sparse GPR model to adapt to the local change of wind speed at specific locations. A moving window strategy is introduced to perform multi-step forecasting and reduce the size of training data. The feasibility of the proposed method is illustrated by measurements collected from the outdoor competition venues in the 2022 Winter Olympics. The presented approach is then compared with the ECMWF, GPR, random forest, ECMWF-Sparse GPR, and ECMWF-MLR models. The results indicate that the proposed method exhibits the best forecasting performance than other models, and it improves the forecasting accuracy in both short-term and long-term time scales.

Upgrading Land-Cover and Vegetation Seasonality in the ECMWF Coupled System: Verification With FLUXNET Sites, METEOSAT Satellite Land Surface Temperatures, and ERA5 Atmospheric Reanalysis.

In this study, we show that limitations in the representation of land cover and vegetation seasonality in the European Centre for Medium‐Range Weather Forecasting (ECMWF) model are partially responsible for large biases (up to ∼10°C, either positive or negative depending on the region) on the simulated daily maximum land surface temperature (LST) with respect to satellite Earth Observations (EOs) products from the Land Surface Analysis Satellite Application Facility. The error patterns were coherent in offline land‐surface and coupled land‐atmosphere simulations, and in ECMWF's latest generation reanalysis (ERA5). Subsequently, we updated the ECMWF model's land cover characterization leveraging on state‐of‐the‐art EOs—the European Space Agency Climate Change Initiative land cover data set and the Copernicus Global Land Services leaf area index. Additionally, we tested a clumping parameterization, introducing seasonality to the effective low vegetation coverage. The updates reduced the overall daily maximum LST bias and unbiased root‐mean‐squared errors. In contrast, the implemented updates had a neutral impact on daily minimum LST. Our results also highlighted the complex regional heterogeneities in the atmospheric sensitivity to land cover and vegetation changes, particularly with issues emerging over eastern Brazil and northeastern Asia. These issues called for a re‐calibration of model parameters (e.g., minimum stomatal resistance, roughness length, rooting depth), along with a revision of several model assumptions (e.g., snow shading by high vegetation).

How reliable are TIGGE daily deterministic precipitation forecasts over different climate and topographic conditions of Iran?

AbstractAccurate forecasting of precipitation has been a very crucial issue for developing meteorological and hydrological early warning systems. Ensemble forecasts using numerical weather prediction models are promising for these purposes. This research addressed the reliability of global precipitation forecasts of the THORPEX (The Observing System Research and Predictability Experiment) Interactive Grand Global Ensemble (TIGGE) archive over distinct climate and topographic regions of Iran. Besides, post‐processing the raw forecasts and assessing TIGGE datasets' performance in predicting extreme rainfall events, at the thresholds of 10–20, 20–50, 50–100, and higher than 100 mm, are other objectives of this study. Results indicated that the European Centre for Medium‐Range Weather Forecasts (ECMWF) and Meteo France forecasts, with the mean correlation coefficients (CCs) of 0.64 and 0.61, were closer to the ground‐gauge observations, especially in different types of humid zones. On the contrary, in semi‐arid and extra‐arid climate zones, the CC value was relatively low. Furthermore, Meteo France, Korea Meteorological Administration, Japan Meteorological Agency and ECMWF outperformed other models in detecting rainy or non‐rainy days. Also, by increasing the elevation, the correlation between forecasts and in situ observations increased significantly, and the percentage increment of CC for ECMWF and Meteo France models is higher than 38%. Moreover, the skill of predictions for extreme events (higher than 50 mm) decreased meaningfully, while at the thresholds of 10–20 and 20–50 mm, ECMWF and China Meteorological Administration outperformed other models. Finally, post‐processing of raw forecasts using quantile mapping improved the CC and relative bias metrics up to 30% and 51%, respectively, especially in extra‐arid, semi‐arid and Mediterranean climate zones.

Late Spring and Summer Subseasonal forecasts in the Northern Hemisphere midlatitudes: biases and skill in the ECMWF model

AbstractSubseasonal forecasts lie between medium-range and seasonal time scales with an emerging attention due to the relevance in society and by the scientific challenges involved. This study aims to (i) evaluate the development of systematic errors with lead time in the European Centre for Medium-Range Weather Forecasts (ECMWF) ensemble forecasts of surface-related variables during late spring and summer, and (ii) investigate potential relationships between the systematic errors and predictive skill. The evaluation is performed over the northern hemisphere midlatitudes, focusing on several regions with different climate characteristics. The results indicate five key bias patterns: (i) cold bias of daily maximum temperature (mx2t) in the April-May forecasts at all lead times in most regions; (ii) central North America with a warm bias mostly in the daily minimum temperature (mn2t); (iii) East of Caspian Sea region with a warm and dry bias; (iv) Western and Mediterranean Europe with a cold bias in mn2t mainly in April-May forecasts and (v) continental Europe with a cold bias in the mx2t and warm bias of mn2t in the June-July forecasts. We also found substantial deviations of soil moisture and terrestrial water storage variation in most regions compared to ECMWF ERA5 reanalysis. Despite the large differences in the systematic error characteristics among the different regions, there is little relation to the skill of the subseasonal forecasts. The systematic temperature biases require further attention from model developers as diurnal cycle improvements could enhance some of the potential predictability coming from the long-memory effect of soil moisture.

Impact of the Boreal Summer Intraseasonal Oscillation on Typhoon Tracks in the Western North Pacific and the Prediction Skill of the ECMWF Model

AbstractThe tropical intraseasonal oscillation (ISO) strongly influences tropical cyclone (TC) genesis. However, little is known concerning its impact on TC tracks. Here, we examine how TC tracks in the western North Pacific are modulated by the boreal summer ISO (BSISO) during each month of the TC season (June–October) using the best track data and how well the modulation are reproduced in the European Centre for Medium‐Range Weather Forecasts (ECMWF) model forecasts. The results reveal that the impact of the BSISO on the typhoon tracks varies month to month. The ECMWF model successfully reproduces this impact up to a month in advance. A simple advection model shows that the reproduction of the modulations in the TC genesis location leads to accurate TC track forecasts. These results suggest that the BSISO is one of the major sources of TC track predictability at the subseasonal time scale (2 weeks to 2 months).

A revised forecast method of ECMWF precipitation based on CNN feature extraction

The paper proposes a method for correcting the precipitation forecast of the European Centre for Medium-Range Weather Forecasts (ECMWF) model. The method first uses the ECMWF physical quantity field to draw the flow field diagram in the form of weather, and uses it as the input of the convolutional neural network(CNN) and performs feature extraction on it. Secondly, select several characteristic factors that are highly correlated with the forecast object from the many characteristic factors. Finally, the selected characteristic factors and the precipitation forecast factors of the ECMWF model are used as the input factors of the random forest regression model for modeling and forecasting. Through the correction forecast test of the next 24h precipitation at 10 forecast test stations, the results show that the ECMWF precipitation correction forecast method proposed in this paper is 15% less than the forecast method using ECMWF interpolation to the station on the MAE and RMSE indicators. And 17%. At the same time, on the TS scores of rainstorms above 24h, the revised forecast method proposed in this paper has a false alarm rate significantly lower than the forecast method of ECMWF interpolation to the station.