Effective approaches to extending medium-term forecasting of persistent severe precipitation in regional models

WANG Dong-Hai and ZHAO Yan-Feng

aGuangdong Province Key Laboratory for Climate Change and Natural Disaster Studies, School of Atmospheric Sciences, Sun Yat-sen University,Guangzhou, China; bState Key Laboratory of Severe Weather, Chinese Academy of Meteorological Sciences, Beijing, China

1. Introduction

Persistent severe rainfall (PSR) events, with daily precipitation greater than 50 mm and durations of longer than three days (Bao 2007), are a highly damaging weather phenomenon. For example, a 12-day PSR event in the summer of 1998 caused disastrous flooding in the Yangtze River Valley, with direct economic losses of 250 billion Yuan RMB and a death toll of more than 3000 (Huang et al. 1998;Lu 2000). More recently, in January 2008, successive snow storms in southern China resulted in losses of 146 billion Yuan RMB and over 130 fatalities (Wang et al. 2009).Moreover, PSR events have been occurring with increasing frequency and at higher intensity in the last 60 years,especially since 1990 (Chen and Zhai 2013).

Regional modeling is a key method for the forecasting of mesoscale circulation and precipitation, and an important means for the forecasting of disastrous weather(Jiao et al. 2006). How to use regional models to extend the medium-term forecasting of PSR events based on improved prediction of regional atmospheric circulation is an important avenue of meteorological research, not least because of the advantages it should bring to disaster prevention and mitigation (Brunet et al. 2010). Most previous studies in this regard have focused on the simulation of atmospheric low-frequency circulation via the combination of statistical and dynamical methods (Zhang et al. 1994; Wheeler and Hendon 2004; Chen, Wei and Gong 2012; Zhu and Li 2017a, 2017b), as well as correcting the forecasting error related to atmospheric circulation (Peng,Che, and Chang 2013) and precipitation (Liu et al. 2013).Meanwhile, little research has been conducted on the use of dynamic prediction methods for PSR from the medium-range forecasting perspective.

In the last few years, based on current understanding of the formation mechanisms of PSR events (Zhao et al.2017) and the method of dynamic extended forecasting in regional models, studies have focused on analyzing the error sources of regional models and evaluating the predictability of multiscale circulation patterns, proposing improved dynamic forecasting methods for the different types of errors in regional models, and creating a theoretical framework for the dynamic extended medium-term forecasting of PSR events (Zhao et al. 2016; Zhao, Wang,and Xu 2017a, 2017b). Here, we summarize the main results from these attempts at improving dynamic extended forecasting.

2. Methods

The forecasting errors of a regional model mainly originate from the initial conditions (ICs) and the numerical forecast model itself. Methods geared toward improving these aspects operate in two main ways: minimizing the uncertainty of ICs by improving the observation and data assimilation system, and making the regional model more representative of the real atmosphere by increasing the resolution and improving the dynamical framework(Wang, Du, and Liu 2011). Most short-range forecast errors originate from the IC errors (Du 2002; Pappenberger et al. 2011), and the IC uncertainty causes the uncertainty in the weather forecast (Jung, Miller, and Palmer 2010).On the other hand, the errors of the model itself cover two main components: systematic error and random error (Reynolds, Satter field, and Bishop 2015). Among these, the systematic error results from de ficiencies in the model dynamical structure, such as the parameterization schemes, resolution, and lateral boundary conditions(LBCs). The systematic circulation errors of different predictive timescales vary in their origin (Lorenz 1982; Tibaldi and Molteni 1990; Skamarock 2004). For the forecasting of PSR events, reducing the IC and LBC errors is an effective approach to reducing the forecasting errors when using a high-resolution regional model (Zhao et al. 2016).

PSR is different from normal rain events because the water vapor and thermodynamic conditions are produced in the context of weather systems with abnormal or less variation (Ding and Reiter 1982; Samel and Liang 2003; Niu, Zhang, and Jin 2012; Piaget et al. 2015). In the case of PSR, the rainfall duration and amount of precipitation are associated with anomalies of large-scale systems that favor the continuous con fluence of moist/warm and dry/cold air (Zhou and Yu 2005; Qian, Fu, and Yan 2007; Wang et al. 2009; Wang, Xia, and Liu 2011);also, the mesoscale convergence line is a good indicator of the area of severe precipitation (Qian, Shan, and Zhu 2012). As shown by evaluations of the forecasting of multiscale circulation patterns, large-scale circulation systems can be better predicted than smaller-scale disturbances (Lorenz 1969; Chen, Wei, and Gong 2010; Dong et al. 2015). Moreover, global models hold an advantage in predicting large-scale variation, while regional models are better in terms of simulating small-scale disturbances (Wang, Yu, and Wang 2012; Schwartz and Liu 2014; Grazzini and Vitart 2015). Thus, improving the efficiency of large-scale forecasts of the forcing fields whilst at the same time retaining the small-scale features in the regional domain is critical for better forecasting PSR events in regional models.

The methods of spectral nudging (SN), lateral boundary filtering (LBF), and updated initial conditions (UIC)have been used in the regional Weather Research and Forecasting (WRF) model for PSR forecasting (Wang et al. 2016; Zhao et al. 2016; Zhao, Wang, and Xu 2017a,2017b). SN is a scale-selective interior constraint technique (von Storch, Langenberg, and Feser 2000; Miguez-Macho, Stenchikov, and Robock 2004) for the large-scale circulation in the regional model. It con fines itself to the higher altitudes and cases where the local convection at lower levels develop freely when the large-scale systems develop to deeper levels. SN has been applied in WRF(Miguez-Macho, Stenchikov, and Robock 2004, 2005; Liu et al. 2012; Glisan et al. 2013) and many other regional climate models, in regions such as North America(Kanamaru and Kanamitsu 2007; Spero et al. 2014), western Europe (Feser 2006), and East Asia (Cha and Lee 2009;Xu and Yang 2015), and its application has been shown to significantly improve the prediction of regional climate atmospheric circulation and precipitation forecasts. Zhao et al. (2016) and Zhao, Wang, and Xu (2017a, 2017b) used SN in WRF to improve the forecasting of PSR events in southern China. The nudging experiments were mainly against the horizontal winds, geopotential height and potential temperature above the planetary boundary layer with an interval of 6 h, starting from the initial time to the end time of the forecast, and the nudging fields were from the Global Forecast System (GFS) predictions(Figure 1).

The LBF method refers to the use of low-pass filtering to retain the regional large-scale circulation from the GFS predictions (Figure 1). In Zhao et al. 2016; its application began in the third-day forecasts and harmonic filtering was selected for spatial field scale separation. The filtering wave selection was based on the dynamical features of the regional large-scale circulation for PSR events (Zhao et al.2017), and the high-frequency waves were reserved by 50%. For the ICs, the UIC method is effective at retaining the large-scale forecasts of the GFS predictions and the small-scale features of the WRF forecasts, by using multi-scale blending (MSB) (Zhao, Wang, and Xu 2017a, 2017b)for 15-day forecasts in WRF. The UIC method was applied to forecasts every three days based on the SN method, with a 12-h running-in period (Figure 1). SN was applied in the first three days, MBS was executed at 2.5 days, and then a new SN was initiated after 12 h of model adaptation.The 15-day forecasts comprised five three-day forecasts.

Figure 1. Flow diagram of the SN, LBF, and UIC forecast.

3. Sample case studies

The methods of the SN and LBF were used in Zhao et al.(2016) to forecast three PSR events during the pre- flood(0000 UTC 19 to 0000 UTC 22 May 2013) and post- flood(0000 UTC 15 to 0000 UTC 19 July 2012) season in South China, and during the Mei-yu period over the Yangtze River Valley (0000 UTC 5 to 0000 UTC 8 July 2013). The anomaly correlation coefficient (ACC) of the 500-hPa geopotential height fields for the different forecast lead times are shown in Figure 2. The improvement by the SN and SN + LBF methods during the PSR periods was re flected mainly in lower-value phases of the ACC at 1–5-day lead times(Figure 2(a)–(c)), whereas the improvement by the LBF was more obvious at 7–11-day lead times (Figure 2(d)–(f)). The averaged ACCs for PSR periods over the different forecast lead times showed that the SN and SN + LBF methods produced stable enhancement, with the SN + LBF method yielding a better forecast at 7–11-day lead times. All the improvements of the new forecasts methods were based on the better GFS forecasts.

Figure 2. Averaged anomaly correlation coefficients of the 500-hPa geopotential height fields for Domain 1 (15°–55°N, 70°–130°E) at lead times of (a) 1 day, (b) 3 days, (c) 5 days, (d) 7 days, (e) 9 days, and (f) 11 days prior to three PSR events in the pre- flood season in South China, the post- flood season in South China, and the Mei-yu period over the Yangtze–Huaihe river basin, respectively. The abscissa is the forecasting day, with the last four days for the PSR period (beginning at the dotted line). Source: Zhao et al. (2016).

Figure 3. Accumulative rainfall distribution of PSR during 0000 UTC 30 June to 0000 UTC 6 July 2016 for the observation (OBS), and in the forecast experiments at different lead times ((a1–a3) 3 days; (b1–b3) 5 days; (c1–c3) 7 days) and using the different experiment schemes((a1–c1) control (CTL); (a2–c2) SN + UIC). Panels a3–c3 are the NCEP GFS forecasts. Source: Zhao, Wang, and Xu (2017a).

The SN and UIC methods (SN + UIC) were used to investigate one of the most devastating flooding events in China since 1998: the case during 0000 UTC 30 June to 0000 UTC 6 July 2016 (Zhao, Wang, and Xu 2017a) (Figure 3). The SN + UIC approach improved the rain band’s range of this PSR event (above 100 mm) at 5–7-day lead times (Figure 3(b2)–(c2)), and the accumulated rainfall above 200 mm at the 3-day lead time (Figure 3(a2)). In addition, the larger the magnitude and longer the lead time, the more obvious the improvement. For the GFS forecasts, the rain band’s range of accumulated rainfall from 50 to 100 mm was wider than that in the observation, and the accumulated rainfall above 100 mm was not forecasted well at 3–5-day lead times (Figure 3(a3)–(b3)). The improvement by the SN + UIC method was based on the new ICs, which combines the advantages of the GFS and WRF forecasts and then improves the accumulated rainfall (especially heavy rainfall) and the rain band’s range forecasts. Furthermore, the SN + UIC method decreased the root-mean-square error (RMSE) for the related meteorological variables in the PSR period, such as the geopotential height, relative humidity, and temperature.

Numerical predictions of four PSR events during the pre- flood season in South China (case 1, 0000 UTC 12 May–0000 UTC 15 May 2011; case 2, 0000 UTC 4 June–0000 UTC 8 June 2011; case 3, 0000 UTC 6 May–0000 UTC 10 May 2013; and case 4, 0000 UTC 19 May–0000 UTC 22 May 2013) were also investigated using the SN + UIC method (Zhao, Wang, and Xu 2017b). The results showed that the SN + UIC approach improved the prediction of daily precipitation for moderate, heavy, and torrential rain(10–100 mm d?1) (Figure 4). The improvement in the 24-h precipitation threat score by using the SN + UIC method was mainly re flected at 3–7-day lead times for moderate and heavy rain (10–49.9 mm d?1) (Figure 4(b)–(d)), and achieved slightly better forecasts in terms of the relative improvement rate of RMSE for accumulated rainfall (6.2%)and relative humidity (5.67%).

4. Concluding remarks

This paper summarizes the improvements generated by a selection of methods (SN, LBF, and UIC) in extending the forecasting of PSR events in China using the WRF model.In addition, relevant case simulations are analyzed and verified.

The improvements for precipitation generated by these methods are mainly re flected at lead times of 3–7 days for moderate and heavy rain forecasts; plus, the larger the magnitude and longer the lead time, the more significant the improvement–especially when using the SN + UIC approach. For regional large-scale circulation, the improvement through use of SN is apparent mainly in the lower-value phases of the ACC at 1–5-day lead times, while the improvement via the LBF method is more obvious at 7–11-day lead times. In addition, the SN + UIC method decreases the RMSE for the geopotential height, relative humidity,and temperature in the PSR period, and the improvements for the relative humidity may make a greater contribution to the better performance of the SN + UIC method in the precipitation forecasts.

Figure 4. Averaged threat scores at lead times of (a) 1 day, (b) 3 days, (c) 5 days and (d) 7 days prior to four PSR events during the pre- flood season in South China based on different methods. Source: Zhao, Wang, and Xu (2017b).

Case studies show that achieving a more efficient use of large-scale forecasts of the global model whilst at the same time retaining the small-scale features in the regional domain is critical for better forecasting PSR events in China using a regional model. In view of the universality of the principles behind the improvements generated by the methods mentioned in this paper, it should be possible to apply them in other regional models for extending the forecasting range for PSR events and other disastrous weather, thus further enhancing disaster prevention and mitigation capabilities. Future work should focus on identifying the optimum con figuration of parameterization schemes and investigating in detail the function of the methods mentioned here, as well as designing a new skill score that can be used for better quantitative verification and analysis. Finally, more cases and long-term statistical studies in different areas with more in-depth dynamic and thermodynamic analysis are needed to fully assess the advantages of these methods of improvement.

Disclosure statement

No potential conflict of interest was reported by the authors.

Funding

This study was jointly supported by the National Natural Science Foundation of China [grant number 41775097], [grant number 91437221], the National Key Basic Research Program of China [grant number 2012CB417204], and the China Special Fund for Meteorological Research in the Public Interest [grant number GYHY201506002].

References

Bao, M. 2007. “The Statistical Analysis of the Persistent Heavy Rain in the Last 50 Years over China and Their Backgrounds on the Large-scale Circulation.”Chinese Journal of Atmospheric Sciences (in Chinese)31: 779–792.

Brunet, G., M. Shapiro, B. Hoskins, M. Moncrieff, R. Dole, G. N.Kiladis, B. Kirtman, et al. 2010. “Collaboration of the Weather and Climate Communities to Advance Subseasonal-to-Seasonal Prediction.”Bulletin of the American Meteorological Society91: 1397–1406.

Cha, D. H., and D. K. Lee. 2009. “Reduction of Systematic Errors in Regional Climate Simulations of the Summer Monsoon over East Asia and the Western North Pacific by Applying the Spectral Nudging Technique.”Journal of Geophysical Research114: D14108.

Chen, G. J., F. Y. Wei, and Y. F. Gong. 2012. “An Extended-range Forecast Method for the Persistent Heavy Rainfall over the Yangtze Huaihe River Valley in Summer Based on the Lowfrequency Oscillation Characteristics.”Chinese Journal of Atmospheric Sciences (in Chinese)36: 633–644.

Chen, G. J., F. Y. Wei, and Y. F. Gong. 2010. “Assessing the Extended Range Forecast Error of NCEP/CFS in the Summer of East Asia.”Journal of Applied Meteorological Science (in Chinese)21:659–670.

Chen, Y., and P. M. Zhai. 2013. “Persistent Extreme Precipitation Events in China during 1951–2010.”Climate Research57:143–155.

Ding, Y. H., and E. R. Reiter. 1982. “A Relationship between Planetary Waves and Persistent Rain- and Thunderstorms in China.”Theoretical & Applied Climatology31: 221–252.

Dong, Y., S. D. Liu, D. H. Wang, Y. F. Zhao. 2015. “Assessment on Forecasting Skills of GFS Model for Two Persistent Rainfalls over Southern China GFS.”Meteorological Monthly Science (in Chinese)41: 45–51.

Du, J. 2002. “Present Situation and Prospects of Ensemble Numerical Prediction.”Journal of Applied Meteorological Science (in Chinese)13: 16–28.

Feser, F. 2006. “Enhanced Detectability of Added Value in Limited-area Model Results Separated into Different Spatial Scales.”Monthly Weather Review134: 2180–2190.

Glisan, J. M., W. J. Gutowski, J. J. Cassano, M. E. Higgins. 2013.“Effects of Spectral Nudging in WRF on Arctic Temperature and Precipitation Simulations.”Journal of Climate26: 3985–3999.

Grazzini, F., and F. Vitart. 2015. “Atmospheric Predictability and Rossby Wave Packets.”Quarterly Journal of the Royal Meteorological Society141: 2793–2802.

Huang, R. H., Y. H. Xu, P. F. Wang, L. T. Zhou. 1998. “The Features of the Catastrophic Flood over the Changjiang River Basin during the Summer of 1998 and Cause Exploration.”Climatic and Environmental Research (in Chinese)3: 300–313.

Jiao, H. Y., J. D. Gong, B. Zhou, S. R. Zhao. 2006. “An Overview of the Development of Weather Forecasting.”Journal of Applied Meteorological Science (in Chinese)17: 594–602.

Jung, T., M. Miller, and T. Palmer. 2010. “Diagnosing the Origin of Extended-range Forecast Errors.”Monthly Weather Review138: 2434–2446.

Kanamaru, H., and M. Kanamitsu. 2007. “Scale-selective Bias Correction in a Downscaling of Global Analysis Using a Regional Model.”Monthly Weather Review135: 334–350.

Liu, L., J. Chen, L. Cheng, C. Z. Lin, and Z. P. Wu. 2013. “Study of the Ensemble-Based Forecast of Extremely Heavy Rainfalls in China:Experiments for July 2011 Cases.”Acta Meteorologica Sinica (in Chinese)71: 853–866.

Liu, P., A. P. Tsimpidi, Y. Hu, B. Stone, A. G. Russell, and A. Nenes.2012. “Differences between Downscaling with Spectral and Grid Nudging Using WRF.”Atmospheric Chemistry and Physics12: 3601–3610.

Lorenz, E. N. 1969. “The Predictability of a Flow Which Possesses Many Scales of Motion.”Tellus21: 289–307.

Lorenz, E. N. 1982. “Atmospheric Predictability Experiments with a Large Numerical Model.”Tellus34: 505–513.

Lu, R. Y. 2000. “Anomalies in the Tropics Associated with the Heavy Rainfall in East Asia during the Summer of 1998.”Advances in Atmospheric Sciences17: 205–220.

Miguez-Macho, G., G. L. Stenchikov, and A. Robock. 2004.“Spectral Nudging to Eliminate the Effects of Domain Position and Geometry in Regional Climate Model Simulations.”Journal of Geophysical Research109: 1025–1045.

Miguez-Macho, G., G. L. Stenchikov, and A. Robock. 2005.“Regional Climate Simulations over North America:Interaction of Local Processes with Improved Large-scale Flow.”Journal of Climate18: 1025–1045.

Niu, R. Y., Z. G. Zhang, and R. H. Jin. 2012. “The Atmospheric Circulation Features of Two Persistent Heavy Rainfalls over Southern China in the Summer of 2010.”Journal of Applied Meteorological Science (in Chinese)23: 385–394.

Pappenberger, F., H. Cloke, A. Persson, and D. Demeritt. 2011.“‘HESS Opinions’ on Forecast (in) Consistency in a Hydrometeorological Chain. Curse or Blessing?.”Hydrology and Earth System Sciences15: 2391–2400.

Peng, X., Y. Che, and J. Chang. 2013. “A Novel Approach to Improve Numerical Weather Prediction Skills by Using Anomaly Integration and Historical Data.”Journal of Geophysical Research Atmospheres118: 8814–8826.

Piaget, N., P. Froidevaux, P. Giannakaki, F. Gierth, O. Martius,M. Riemer, G. Wolf, C. M. Grams. 2015. “Dynamics of a Local Alpine Flooding Event in October 2011. Moisture Source and Large-Scale Circulation.”Quarterly Journal of the Royal Meteorological Society141: 1922–1937.

Qian, W. H., J. Fu, and Z. Yan. 2007. “Decrease of Light Rain Events in Summer Associated with a Warming Environment in China during 1961–2005.”Geophysical Research Letters34: L11705.

Qian, W. H., X. L. Shan, and Y. F. Zhu. 2012. “Capability of Regionalscale Transient Wind Anomalies to Indicate Regional Heavy Rains.”Chinese Journal of Geophysics (in Chinese)55: 1513–1522.

Reynolds, C. A., E. A. Satter field, and C. H. Bishop. 2015. “Using Forecast Temporal Variability to Evaluate Model Behavior.”Monthly Weather Review143: 4785–4804.

Samel, A. N., and X. Z. Liang. 2003. “Understanding Relationships between the 1998 Yangtze River Flood and Northeast Eurasian Blocking.”Climate Research23: 149–158.

Schwartz, C. S., and Z. Liu. 2014. “Convection-permitting Forecasts Initialized with Continuously Cycling Limited-area 3DVAR, Ensemble Kalman Filter, and ‘Hybrid’ Variationalensemble Data Assimilation Systems.”Monthly Weather Review142: 716–738.

Skamarock, W. C. 2004. “Evaluating Mesoscale NWP Models Using Kinetic Energy Spectra.”Monthly Weather Review132:3019–3032.

Spero, T. L., M. J. Otte, J. H. Bowden, and C. G. Nolte. 2014.“Improving the Representation of Clouds, Radiation, and Precipitation Using Spectral Nudging in the Weather Research and Forecasting Model.”Journal of Geophysical Research Atmospheres119: 11682–11694.

von Storch, H., H. Langenberg, and F. Feser. 2000. “A Spectral Nudging Technique for Dynamical Downscaling Purposes.”Monthly Weather Review128: 3664–3673.

Tibaldi, S., and F. Molteni. 1990. “On the Operational Predictability of Blocking.”Tellus A42: 343–365.

Wang, D. H., J. Du, and C. J. Liu. 2011. “Recognizing and Dealing with the Uncertainty in Weather and Climate Forecasts.”Meteorological Monthly (in Chinese)37: 385–391.

Wang, D. H., C. J. Liu, Y. Liu, F. Y. Wei, N. Zhao, Z. N. Jiang, Y. Ying,et al. 2009. “A Preliminary Analysis of Features and Causes of the Snow Storm Event over the Southern Areas of China in January 2008.”Journal of Meteorological Research23: 374–386.

Wang, D. H., R. D. Xia, and Y. Liu. 2011. “A Preliminary Study of the Flood Causing Rainstorm during the First Rainy Season in South China in 2008.”Acta Meteorologica Sinica (in Chinese)69: 137–148.

Wang, S. L., X. D. Xu, H. W. Kang, S. Zhang, and X. Zhang. 2016.“Simulation of Continuous Rainfall over South China in Early 2008 with the Spectral Nudging Method and the Periodicity Characteristics of the Water Vapor Channel.”Chinese Journal of Atmospheric Sciences (in Chinese)40: 476–488.

Wang, S. Z., E. T. Yu, and H. J. Wang. 2012. “A Simulation Study of a Heavy Rainfall Process over the Yangtze River Valley Using the Two-way Nesting Approach.”Advances in Atmospheric Sciences29: 731–743.

Wheeler, M. C., and H. H. Hendon. 2004. “An All-season Realtime Multivariate MJO Index: Development of an Index for Monitoring and Prediction.”Monthly Weather Review132:1917–1932.

Xu, Z., and Z. L. Yang. 2015. “A New Dynamical Downscaling Approach with GCM Bias Corrections and Spectral Nudging.”Journal of Geophysical Research Atmospheres120: 3063–3084.

Zhang, J. J., W. J. Li, X. D. Xu, and J. Miao. 1994. “The Experiment of DERF with T42L9 Model for DEKAD and Monthly Mean Circulation Anomaly during the Summer Heavy Rainfall Period in 1991.”Acta Meteorologica Sinica (in Chinese)52:180–186.

Zhao, Y. F., D. H. Wang, Z. M. Liang, and J. J. Xu. 2016. “Improving Numerical Experiments on Persistent Severe Rainfall Events in Southern China Using Spectral Nudging and Filtering Schemes.”Quarterly Journal of the Royal Meteorological Society142: 3115–3127.

Zhao, Y. F., D. H. Wang, Z. M. Liang, and J. J. Xu. 2017. “On the Dynamics of the Large-scale Circulation during the Persistent Severe Rainfall Events in Southern China.”Journal of the Meteorological Society of Japan95: 111–125.

Zhao, Y. F., D. H. Wang, and J. J. Xu. 2017a. “Improving the Regional Model Forecasting of Persistent Severe Rainfall over the Yangtze River Valley Using the Spectral Nudging and Update Cycle Methods: A Case Study.”Atmospheric Science Letters18: 96–102.

Zhao, Y. F., D. H. Wang, and J. J. Xu. 2017b. “An Attempt to Improve the Forecasting of Persistent Severe Rainfall Using the Spectral Nudging and Update Cycle Methods.”Weather and Forecasting32: 713–723.

Zhou, T. J., and R. C. Yu. 2005. “Atmospheric Water Vapor Transport Associated with Typical Anomalous Summer Rainfall Patterns in China.”Journal of Geophysical Research Atmospheres110: 211–211.

Zhu, Z., and T. Li. 2017a. “Extended-Range Forecasting of Chinese Summer Surface Air Temperature and Heat Waves.”Climate Dynamics: 1–15. doi:10.1007/s00382-017-3733-7.

Zhu, Z., and T. Li. 2017b. “Statistical Extended-Range Forecast of Winter Surface Air Temperature and Extremely Cold Days over China.”Quarterly Journal of the Royal Meteorological Society704 (143): 1528–1538.