ATMOSPHERIC AND OCEANIC SCIENCE LETTERS, 2011, VOL. 4, NO. 3, 173180 Statistical Downscaling Prediction of Summer Precipitation in Southeastern China LIU Ying1, 2, FAN Ke1, and WANG Hui-Jun1 1 2 Nansen-Zhu International Research Centre, Institute of Atmospheric Physics, Chinese Academy of Sciences, Beijing 100029, China Graduate University of the Chinese Academy of Sciences, Beijing 100049, China Received 25 January 2011; revised 22 February 2011; accepted 22 February 2011; published 16 May 2011 Abstract A statistical downscaling approach based on multiple-linear-regression (MLR) for the prediction of summer precipitation anomaly in southeastern China was established, which was based on the outputs of seven operational dynamical models of Development of a European Multi-model Ensemble System for Seasonal to Interannual Prediction (DEMETER) and observed data. It was found that the anomaly correlation coefficients (ACCs) spatial pattern of June-July-August (JJA) precipitation over southeastern China between the seven models and the observation were increased significantly; especially in the central and the northeastern areas, the ACCs were all larger than 0.42 (above 95% level) and 0.53 (above 99% level). Meanwhile, the root-mean-square errors (RMSE) were reduced in each model along with the multi-model ensemble (MME) for some of the stations in the northeastern area; additionally, the value of RMSE difference between before and after downscaling at some stations were larger than 1 mm d1. Regionally averaged JJA rainfall anomaly temporal series of the downscaling scheme can capture the main characteristics of observation, while the correlation coefficients (CCs) between the temporal variations of the observation and downscaling results varied from 0.52 to 0.69 with corresponding variations from 0.27 to 0.22 for CCs between the observation and outputs of the models. Keywords: statistical downscaling, DEMETER, southeastern China, summer precipitation anomaly Citation: Liu, Y., K. Fan, and H.-J. Wang, 2011: Statistical downscaling prediction of summer precipitation in southeastern China, Atmos. Oceanic Sci. Lett., 4, 173–180. 1 Introduction Southeastern China is economically well developed, has a dense population, and experiences more floods in the spring and summer seasons. The climate anomaly can threaten millions of lives and property (Chen, 1991). The prediction of precipitation during June-July-August (JJA) over southeastern China is difficult due to many uncertainties. The major question in this context is regarding the method of employment of the available model outputs to increase prediction skill effectively. It is widely accepted that Atmospheric and Ocean Coupled General Circulation Models (AOGCM) models are able to reasonably simulate large-scale atmospheric characters, such Corresponding author: LIU Ying, [email protected] as geopotential height at the highest level, temperature near the surface and atmospheric circulation (Von Storch et al., 1993). However, the resolution of AOGCM output is too coarse to provide regional-scale information required for regional impact assessments (Fan et al., 2005). Our research question is regarding the method to improve the accuracy of the AOGCM output in our study region. At present, there are the following two main methods available to make up for the lack of regional scale climate change of AOGCM: a high-resolution AOGCM (Boyle, 1993); and the downscaling method (Kang et al., 2009). As the former approach requires a considerable amount of computer resources for increasing the resolution of AOGCM, the latter approach is becoming more popular. Downscaling methods can be classified into the following three main groups: Regional climate models nested in AOGCM, statistical downscaling, and dynamical-statistical downscaling. Due to convenience, relative simplicity, and less sources, the statistical downscaling method is applied widely. Essentially, the idea of the statistical downscaling technique consists of seeking relationships between the predictors (large-scale variables, i.e., geopotential height (GPH) at 500 hPa, sea-surface pressure, and SST) and the predictands (regional-scale variables, i.e., local rainfall and temperature) to set up statistical models(Cavazos and Hewitson, 2005; Paul et al., 2008). Many researchers use different linear statistical methods to downscale interested meteorological elements (Zorita et al., 1992; Widmann and Bretherton, 2003). Multiple-linear-regression (MLR) is one of the linear statistical downscaling methods. Its principal technique is to find regression links between multi-predictors and predictands. Many recent studies have applied different statistical downscaling methods to predict regional precipitation or temperature (Fan et al., 2009; Wang and Fan, 2009). Most previous studies revealed that precipitation and temperature are the most effectively predictable by selecting different predictors as per the different regions. However, the dataset of the Development of a European Multi-model Ensemble System for Seasonal to Interannual Prediction (DEMETER) has not been applied to study summer rainfall over southeastern China by employing statistical downscaling. For this purpose, we have chosen the MLR using hindcast results of the seven models in the DEMETER, to concentrate on JJA precipitation anomaly prediction. The datasets employed in this study are described briefly in section 2, and the results are pre- 174 ATMOSPHERIC AND OCEANIC SCIENCE LETTERS sented in section 3. Section 4 presents conclusions and discussions based on the results. 2 2.1 Data and methodology Data The DEMETER project has been funded by the European Union for the period of April 2000 to September 2003. The DEMETER system comprises of seven global coupled ocean-atmosphere models initiated on 1 February, 1 May, 1 August, and 1 November initial conditions. Each hindcast of the models has been integrated for six months and comprises of an ensemble of nine members (Palmer et al., 2004). The models used in the DEMETER project are those of CERFACS (European Centre for Research and Advanced Training in Scientific Computation, France), CNRM (Centre National de Recherches Météorologiques, France), ECMWF (European Centre for Medium-Range Weather Forecasts, UK), INGV (Istituto Nazionale de Geofisica e Vulcanologia, Italy), LODYC (Laboratoire d’Océanographie Dynamique et de Climatologie, France), MPI (Max-Planck Institut für Meteorologie, Germany), and UKMO (Met Office, UK). In this paper, the multi-model ensemble (MME) is defined as the average ensemble of these seven models (More detailed information about the DEMETER can be found at the following website: http:// www.ecmwf.int/research/demeter/general/index.html). The monthly rainfall, specific humidity at 850 hPa (q850) and the geopotential height at 500 hPa (Z500) datasets in JJA for the period of 19802001 in the DEMETER project were applied. Moreover, for accuracy, the datasets of the DEMETER starting from 1 May were chosen. The monthly SST data was HadISST 1.1 monthly average sea surface temperature (HadISST1), which was obtained from the Met Office Hadley Centre, and it was composed of a 1-degree latitude by 1-degree longitude grid of a globally complete dataset. Additionally, 700 hPa GPH (Z700) in ECMWF 40-yr Reanalysis (ERA-40) were employed. The source of the observed rainfall data at 753 stations over China from 1980 to 2001 were obtained from the China Meteorological Administration National Meteorological Information Center. Antarctic Oscillation index (AAOI) was defined as the time coefficient series of the first mode of empirical orthogonal function (EOF) analysis of Z700 pole ward of 20S (Thompson and Wallace, 1998). In this paper, the southeastern China region is located in 2530°N, 110122°E and contained 59 stations. 2.2 VOL. 4 Methodology The precipitation in southeastern China has a close relationship with large-scale background atmospheric circulation and SST. When the weakening of the South China Sea sub-high is delayed, abnormal floods prevail from the middle and lower reaches of the Yangtze River up to southeastern China (Wan et al., 2008). SST in a key region, for example, northern Atlantic and tropical eastern Pacific, is also an important factor influencing southeastern China precipitation, while the central region of southeastern China is the most affected by the SST (Chen et al., 2003; Lang and Wang, 2010; Wang and Chen, 2004). Abnormal atmospheric circulation over the middle and lower reaches of the Yangtze River and southeastern China followed by rainstorm and heavy rainstorm in these regions led to the invalidation of climate prediction in the summer of 1999 (Wang and Zhang, 2000). Mid-high latitude at 500 hPa geopotential height is another key factor that influences summer precipitation over eastern China (Pan et al., 2004; Wang, 2000; Zhou and Wang, 2006). Disturbance in mid-latitude atmosphere is an effective factor for the precipitation in eastern China (Huang et al., 2003). Water vapor in the South China Sea (SCS), the Bay of Bengal, and the southwestern side of the western Pacific subtropical high are the key regions that influence China’s summer monsoon precipitation. The vapor fluxes merge in the Yangtze River basin, which forces summer rainfall in this area (Huang et al., 1998). An increase of the vapor flux in the SCS and the Bay of Bengal will increase rainfall in south and southeastern China (Liu, 2006). AAO is the key factor that impacts the climate and can play an important part in JJA precipitation in the Yangtze River and the Jianghuai River basin (Gao et al., 2003; Wang and Fan, 2005; Fan, 2006; Fan and Wang, 2006; Sun et al., 2009; Zhu, 2009). Therefore, for this paper, we chose the current spring AAOI as a predictor. Based on previous research, we chose q850 as one of predictors in the Asian JJA monsoon region. The other predictors were area-average December-January-February (DJF) SST and area-average Z500 (JJA). The influential areas of these predictors are chosen by the most significant correlation coefficients (CCs) between the precipitation (JJA) anomaly in southeastern China and the global fields of Z500 (JJA), q850 (JJA), and SST (DJF). Predictors are described in Table 1. The correlation coefficients’ relationship between the predictors and the regionally averaged JJA precipitation over southeastern China for the period of 19802001 are listed in Table 2. The results Table 1 Predictors and their key regions where the correlation coefficients between regionally average summer precipitation over southeastern China and global predictors were significant. Predictors Key regions Z500 (JJA) of models Baikal region (37.555°N, 80130°E) q850 (JJA) of models East Asian summer monsoon region (530°N, 60160°E) Kuroshio region (20.534.5°N, 120.5140.5°E), North Atlantic (30.545.5°N, 40.5°W10.5°E) South Pacific (19.54.5°S, 105.5170.5°W) Observed SST (DJF) NO. 3 LIU ET AL.: STATISTICAL DOWNSCALING, SUMMER PRECIPITATION, SOUTHEASTERN CHINA Table 2 Correlation coefficients between the predictors at the different regions of observation and regionally averaged JJA precipitation in southeastern China during 19802001. Z500 q850 AAOI SSTKU SSTNA SSTSP 0.50 0.39 0.54 0.56 0.61 0.40 The underlined CCs indicate above 90% significance level. SSTKU, SSTNA, and SSTSP mean sea surface temperature in the areas of Kuroshio, North Atlantic and South Pacific. indicate that the correlation coefficients of all predictors are significant. The downscaling technique in this paper is carried by using the leaving-one-year-out cross validation method (Michaelsen, 1987). Every target year was picked out from the 22-year period (19802001), the MLR fitting equations between the predictors and predictands were carried out for the residual 21-year period, and every year acted as a predictand for one year. In addition, we established a MLR equation for every station in the southeastern China region. This MLR equation is based on the previous six predictors and can be expressed as: yit=b0+b1Z500+b2q850+b3AAOI+b4SSTKU+ b5SSTNA+b6SSTSP, (1) where i is number of stations, t is forecast time, b0, b1, b2, b3, b4, b5, and b6 are constant and coefficients for predictors. And the formula for the root mean square errors (RMSE) is given by: T RMSE y t y 0t 2 t 1 , (2) N where yt is hindcasts of models or downscaling; y0t is observation data; N and T present total numbers of stations in the southeastern China region and forecast time length, respectively. 3 3.1 Results Model evaluation Table 3 presents the model evaluation results. The CCs are calculated between the ERA-40 reanalysis data and the DEMETER data. For models CERFACS, CNRM, INGV, and UKMO, Z500 and q850 revealed convincingly high CCs (from 0.42 to 0.65, exceeding 95% significance level). One predictor’s CCs reached 90% significance level for models ECMWF and LODYC, while only MPI exhibited low CCs at 0.29 and 0.15. As seen from Table 3, most models produced satisfactory results. Hence, Z500 and q850 out of the seven individual models would be con- 175 sidered in the MLR scheme. 3.2 Relationship of rainfall anomaly between observation and models or downscaling approach To examine the efficiency of the downscaling framework in predicting JJA precipitation in southeastern China, we calculated the anomaly correlation coefficients (ACCs) between the observed and model-predicted or MLR downscaling predicted JJA precipitation (Fig. 1), in which the left and right represent before and after the downscaling scheme, respectively. As Fig. 1 demonstrates, most of the models have no prediction skills for summer rainfall over southeastern China, and ACCs of most parts of southeastern China were negative for the models CERFACS, INGV, LODYC, MPI, and UKMO, with a minimum and maximum value of 0.6 and 0.2, respectively. The significantly positive value for models CNRM and ECMWF could be as high as 0.42, but the areas were very small. The MME also shows some prediction skill for the northwestern area. After performing the statistical downscaling scheme, ACCs over southeastern China were increased significantly for all the models, and the MME as ACCs in each model and the MME become positive for most stations (above 95% significance level), especially for the central and northeastern parts (Fig. 1., the right column). Models INGV, MPI, UKMO, and MME, have the areas in which the maximum ACCs reached 0.8, much larger than the 99% significance level. Meanwhile, for CERFACS, CNRM, LODYC, and MME, their values increased from 0.2 (see Fig. 1, the left column) to 0.42 (significant at 95% level) in the northeastern part of the entire region. Additionally, as the results of ECMWF and INGV indicate, the ACCs achieved a value of 0.42 as compared to 0.4 in the central part of southeastern China. 3.3 Rainfall RMSE of before and after downscaling The bias between the observation and models’ prediction before and after statistical downscaling was measured by RMSE, with large RMSE differences between before and after downscaling, indicating better prediction skills. In this study, RMSE difference signifies that the RMSE of the model distracted the RMSE of the downscaling scheme. Fig. 2 presents the spatial patterns of RMSE for precipitation anomaly differences between before and after downscaling. As compared to the models’ prediction, the downscaling scheme improved the predicted skill in most parts of the northern region over southeastern China for all the models and MME, and the large differences were centralized in the northeastern area, indicating im- Table 3 Correlation coefficients between regionally averaged predictors from seven DEMETER models, the MME, and ERA-40s for the 22-year period (19802001). CERFACS CNRM ECMWF INGV LODYC MPI UKMO MME Z500 0.65 0.45 0.76 0.45 0.30 0.29 0.64 0.71 q850 0.48 0.60 0.27 0.48 0.35 0.15 0.42 0.51 The underlined CCs indicate above 90% significance level. 176 ATMOSPHERIC AND OCEANIC SCIENCE LETTERS VOL. 4 Figure 1 ACCs between the observed and models-hindcast JJA precipitation for the 22-year period (19802001). The left column represent the before downscaling scheme, and the right column represent the after downscaling scheme. Areas exceeding 95% significance level are shaded. NO. 3 LIU ET AL.: STATISTICAL DOWNSCALING, SUMMER PRECIPITATION, SOUTHEASTERN CHINA 177 Figure 1 (Continued) proved prediction skills. From the aforementioned analysis, these typical spatial distributions are similar to the ACCs (Fig. 1), namely that stations in the northeastern part were responding better to the predictors. Models LODYC and UKMO presented improved results as compared to the other models. Although Z500 and q850’s predicted skill for the LODYC were not high (Table 3), the downscaling result was superior, suggesting the role of preceding SST factors on the JJA precipitation in the key regions over southeastern China is very important. Meanwhile, about one-quarter of the stations increased the predicted level in most models and the MME. However, UKMO emerged as the best performer, and its RMSE at more than half the stations (33 stations) was larger than zero. However, ECMWF and INGV were the worst performers, with larger than zero RMSE values recorded at no more than 12 and 13 stations, respectively (Table 4). 3.4 Temporal variations of the regionally averaged JJA precipitation anomaly The above results exhibit the station-to-station MLR scheme’ predicted capability of JJA precipitation anomaly for the period of 19802001. In addition to these, we evaluated the scheme performance on an average over the entire region. For the observation data prior to 2000, the trend of JJA precipitation anomaly over the southeastern China was rising with a subsequent decrease after 2000 (see Fig. 3). All of the models’ outputs failed to capture the main variation in the tendency and the year-to-year variability. In contrast, the downscaling method was able to successfully record the entire increasing trend closer to the observations, particularly CERFACS, CNRM, UKMO, and MME. The correlation coefficients between the temporal variation of observation and the downscaling varies from 0.52 to 0.69, with corresponding values of 0.27 to 0.22 for the observation and the original output of models (CCs are not shown; refer to Fig. 3). However, for 1991, 1994, 1998, and 2001, the downscaling method underestimated or overestimated the precipitation value. This was unavoidable, as the cross-validation downscaling model can slightly reproduce extreme precipitation events. All of these analyses suggest that the six predictors applied to the downscaling scheme are appropriate for predicting JJA precipitation anomaly over southeastern China, especially in the central and northeastern parts. Therefore, it might be possible to increase JJA precipitation anomaly prediction skills for individual models including the MME, by adopting this downscaling scheme. 4 Conclusions and discussions In this study, statistical downscaling based on the DEMETER multimodel outputs was used to predict JJA rainfall anomaly over southeastern China. In particular, Table 4 Numbers of stations for RMSE differences between before and after downscaling above zero. CERFACS CNRM ECMWF INGV LODYC MPI UKMO MME Total 20 17 12 13 21 14 33 19 d≥1.0 1 0 0 1 2 2 2 1 0.5≤d<1.0 3 3 3 3 2 1 9 4 0≤d<0.5 16 14 9 9 17 11 22 14 178 ATMOSPHERIC AND OCEANIC SCIENCE LETTERS VOL. 4 Figure 2 Spatial distribution of JJA precipitation anomaly’s RMSE differences between before and after downscaling. The solid and hollow circles denote the positive and negative differences, respectively. Units: mm d1. Z500 and q850 from seven different global models were used as predictors. Downscaling has been shown to considerably outperform global models in predicting regional precipitation. To search for appropriate predictors in an optimal window, we calculated the CCs between the observed JJA precipitation anomaly series over southeastern China and the predictors globally. These six predictors not only included preceding factors (SST and AAOI), but also current summer variables (Z500 and q850). On the basis of these six predictors in an optimal window, the MLR downscaling scheme was found to adopt the leaving-oneyear-out cross-validation method, and subsequently, we used this fitting downscaling scheme for every station to predict the precipitation anomaly over southeastern China for each year. Examinations of the downscaling scheme using the DEMETER multi-model hindcast datasets were performed. The results indicate that the downscaling scheme can substantially improve JJA precipitation anomaly predictions in southeastern China. ACCs, RMSE, and regionally averaged precipitation temporal series predicted by the downscaling scheme were carried out for comparisons with the models’ outputs. All seven models and the MME revealed increased ACCs between the downscaling predictions and the observations during 19802001, especially for the central and northeastern regions of southeastern China. Differences between the models’ predictions and the downscaling scheme results showed that the RMSE decreased mainly in the central and northeastern parts, similar to ACCs distribution. Specifically, the downscaling framework is preferable for stations located in the central and NO. 3 LIU ET AL.: STATISTICAL DOWNSCALING, SUMMER PRECIPITATION, SOUTHEASTERN CHINA 179 Figure 3 Interannual variation of regionally averaged JJA precipitation anomaly during 19802001. A short dash line with a hollow circle indicates the observation result, while the solid line with a solid dot and a solid line with a solid triangle represent the value of the model and the downscaling scheme, respectively. Units: mm d1. northeastern of southeastern China. The downscaling operation can reproduce the linear trend of regionally averaged JJA precipitation anomaly temporal variation, with the exception of missing some extreme precipitation year. On the whole, the RMSE exhibited decreased values for all the seven models. Therefore, statistical downscaling can be a powerful method for improving JJA precipitation anomaly prediction over southeastern China effectively. Due to the imperfect models’ outputs of Z500 and q850 and the limitation of the dataset’s time scale, the downscaling performances was not quite acceptable for some of the models. In future, the downscaling scheme may be improved by using long model outputs and more predictors. Acknowledgements. This research is jointly supported by the special Fund for Public Welfare Industry (Meteorology) (Grant No. GYHY200906018), the National Basic Research Program of China (Grant Nos. 2010CB950304 and 2009CB421406), and the Knowledge Innovation Program of the Chinese Academy of Sciences (Grant No. KZCX2-YW-QN202). References Boyle, J. S., 1993: Sensitivity of dynamical quantities to horizontal resolution for a climate simulation using the ECMWF (cycle 33) Model, J. Climate, 6(5), 796815. Cavazos, T., and B. C. Hewitson, 2005: Performance of statistical downscaling of daily precipitation, Climate Res., 28, 95107. Chen, J. Y., 1991: Analysis and Long-Term Prediction Research on Drought-Flood in China (in Chinese), Chinese Agriculture Press, 180 ATMOSPHERIC AND OCEANIC SCIENCE LETTERS Beijing, 14pp. Chen, S. D., Q. Q. Wang, and Y. F. Qian, 2003: Preliminary discussions of basic climatic characteristics of precipitation during seasons in regions south of Yangtze River and its relationship with SST anomaly, J. Tropical Meteor. (in Chinese), 9(2), 191200. Fan, K., 2006: Atmospheric circulation anomalies in the Southern Hemisphere and summer rainfall over Yangtze River Valley, Chinese J. Geophys. (in Chinese), 49(3), 672679. Fan, K., M. J. Lin, and Y. Z. Gao, 2009: Forecasting the summer rainfall in North China using the year-to-year increment approach, Sci. China-Earth Sci., 52(4), 532539. Fan, K., and H. J. Wang, 2006: Antarctic Oscillation anomaly and forecast of 2006 summer precipitation in eastern China, J. Appl. Meteor. Sci. (in Chinese), 17(3), 383384. Fan, L. J., C. B. Fu, and D. L. Chen, 2005: Review on creating future climate change scenarios by statistical downscaling techniques, Adv. Earth Sci. (in Chinese), 20(3), 320329. Gao, H., F. Xue, and H. J. Wang, 2003: Influence of interannual variability of Antarctic oscillation on mei-yu along the Yangtze and Huaihe River valley and its importance to prediction, Chinese Sci. Bull., 48(suppl.), 8792. Huang, R. H., J. L. Chen, L. T. Zhou, et al., 2003: Studies on the relationship between the severe climate disasters in China and the East Asia climate system, Chinese J. Atmos. Sci. (in Chinese), 27(4), 770787. Huang, R. H., Z. Z. Zhang, G. Huang, et al., 1998: Characteristics of the water vapor transport in East Asian monsoon region and its difference from that in South Asian monsoon region in summer, Chinese J. Atmos. Sci. (in Chinese), 22(4), 460469. Kang, H., C.-K. Park, S. N. Hameed, et al., 2009: Statistical downscaling of precipitation in Korea using multimodel output variables as predictors, Mon. Wea. Rev., 137, 19281938. Lang, X. M. and H. J. Wang, 2010: Improving extraseasonal summer rainfall prediction by merging information from GCMs and observations, Wea. Forecasting, 25, 12631274. Liu, X. D., 2006: Diagnosis and simulation of influence of vapor transportation to summer rain’s type changes in eastern, Meteor. Disaster Reduce Res. (in Chinese), 29(4), 1216. Michaelsen, J., 1987: Cross-validation in statistical climate forecast models, J. Appl. Meteor. Sci., 26(11), 15891600. Palmer, T. N., A. Alessandri, U. Andersen, et al., 2004: Development of a European Multimodel Ensemble System for Seasonal-to-Interannual Prediction (DEMETER), Bull. Amer. Meteor. Soc., 85, 853872. Pan, J., P. X. Wang, and L. R. Ji, 2004: Study on the summer persistent circulation pattern features over Asian-European mid-high latitude part1: Circulation pattern index and persistent, Scientia Meteor. Sinica (in Chinese), 24(2), 127136. Paul, S., C. M. Liu, J. M. Chen, et al., 2008: Development of a sta- VOL. 4 tistical downscaling model for projecting monthly rainfall over East Asia from a general circulation model output, J. Geophys. Res., 113, D15117, doi:10.1029/2007JD009472. Sun, J. Q., H. J. Wang, and W. Yuan, 2009: A possible mechanism for the co-variability of the boreal spring Antarctic Oscillation and the Yangtze River valley summer rainfall, Int. J. Climatol., 29, 12761284. Thompson, D. W. J., and J. M. Wallace, 1998: The Arctic Oscillation signature in the wintertime geopotential height and temperature fields, Geophys. Res. Lett., 25, 12971300. Von Storch, H., E. Zorita, and U. Cubasch, 1993: Downscaling of global climate change estimates to regional scales: An application to Iberian rainfall in wintertime, J. Climate, 6(6), 11611171. Wan, R. J., T. M. Wang, and G. X. Wu, 2008: Temporal variations of the spring persistent rains and SCS subtropical high and their correlations to the circulation and precipitation of the East Asia summer monsoon, Acta Meteor. Sinica (in Chinese), 66(5), 800807. Wang, H. J., 2000: Characteristics of the atmospheric general circulation in three flood years in China, Quart. J. Appl. Meteor. (in Chinese), 11(Suppl.), 7886. Wang, H. J., and K. Fan, 2005: Central-North China precipitation as reconstructed from the Qing dynasty: Signal of the Antarctic Atmospheric Oscillation, Geophys. Res. Lett., 32, L24705, doi:10.1029/2005GL024562. Wang, H. J., and K. Fan, 2009: A new scheme for improving the seasonal prediction of summer precipitation anomalies, Wea. Forecasting, 24(2), 548554. Wang, H. J., and J. Y. Zhang, 2000: The reason for continuous strong plum rains over middle and lower Yangtze River valley and Southeastern China region, Meteor. Sci. Technol. (in Chinese), 28(2), 2023. Wang, Q. Q., and S. D. Chen, 2004: SVD analysis of the relationship between southeastern China rainy season precipitation and sea surface temperature in the tropical oceans, Arid Meteor. (in Chinese), 22(3), 1116. Widmann, M., and C. S. Bretherton, 2003: Statistical precipitation downscaling over Northwestern United States using numetically simulated precipitation as a predictor, J. Climate, 16, 799816. Zhou, B. T., and H. J. Wang, 2006: Relationship between the boreal spring Hadley circulation and the summer precipitation in the Yangtze River valley, J. Geophys. Res., 111, D16109, doi:10. 1029/2005JD0070006 Zhu, Y. L., 2009: The Antarctic oscillation—East Asian summer monsoon connections in NCEP-1 and ERA-40, Adv. Atmos. Sci., 26(4), 707716. Zorita, E., V. Kharin, and H. Von Storch, 1992: The atmospheric circulation and sea surface temperature in the North Atlantic area in winter: Their interaction and relevance for Iberian precipitation, J. Climate, 5, 10971108.
© Copyright 2026 Paperzz