ADVANCES IN ATMOSPHERIC SCIENCES, VOL. 35, AUGUST 2018, 918–926

• Original Paper •

Seasonal Forecasts of the Summer 2016 Yangtze River Basin Rainfall

Philip E. BETT∗1, Adam A. SCAIFE1,2, Chaofan LI3, Chris HEWITT1, Nicola GOLDING1, Peiqun ZHANG4,Nick DUNSTONE1, Doug M. SMITH1, Hazel E. THORNTON1, Riyu LU5, and Hong-Li REN4

1Met Office Hadley Centre, FitzRoy Road, Exeter EX1 3PB, UK2College of Engineering, Mathematics and Physical Sciences, University of Exeter, Exeter, Devon EX4 4QF, UK

3Center for Monsoon System Research, Institute of Atmospheric Physics, Chinese Academy of Sciences, Beijing 100029, China4Laboratory for Climate Studies, National Climate Center, China Meteorological Administration, Beijing 100081, China

5State Key Laboratory of Numerical Modeling for Atmospheric Sciences and Geophysical Fluid Dynamics,

Institute of Atmospheric Physics, Chinese Academy of Sciences, Beijing 100029, China

(Received 24 August 2017; revised 15 December 2017; accepted 24 January 2018)

ABSTRACT

The Yangtze River has been subject to heavy flooding throughout history, and in recent times severe floods such as thosein 1998 have resulted in heavy loss of life and livelihoods. Dams along the river help to manage flood waters, and areimportant sources of electricity for the region. Being able to forecast high-impact events at long lead times therefore hasenormous potential benefit. Recent improvements in seasonal forecasting mean that dynamical climate models can start tobe used directly for operational services. The teleconnection from El Nino to Yangtze River basin rainfall meant that thestrong El Nino in winter 2015/16 provided a valuable opportunity to test the application of a dynamical forecast system.This paper therefore presents a case study of a real-time seasonal forecast for the Yangtze River basin, building on previouswork demonstrating the retrospective skill of such a forecast. A simple forecasting methodology is presented, in which theforecast probabilities are derived from the historical relationship between hindcast and observations. Its performance for2016 is discussed. The heavy rainfall in the May–June–July period was correctly forecast well in advance. August sawanomalously low rainfall, and the forecasts for the June–July–August period correctly showed closer to average levels. Theforecasts contributed to the confidence of decision-makers across the Yangtze River basin. Trials of climate services such asthis help to promote appropriate use of seasonal forecasts, and highlight areas for future improvements.

Key words: seasonal forecasting, flood forecasting, Yangtze basin rainfall, ENSO, hydroelectricity

Citation: Bett, P. E., and Coauthors, 2018: Seasonal forecasts of the summer 2016 Yangtze River basin rainfall. Adv. Atmos.Sci., 35(8), 918–926, https://doi.org/10.1007/s00376-018-7210-y.

1. IntroductionThe Yangtze River basin cuts across central China, pro-

viding water, hydroelectricity and agricultural land for mil-lions of people. The Yangtze has been subject to floodingthroughout history (e.g., Plate, 2002; Yu et al., 2009), linkedto variations in the East Asian monsoon that are sometimesdriven by factors such as the El Nino–Southern Oscillation(ENSO; e.g. Zhang et al., 2016a, b). Large hydroelectricdams along the river and its tributaries, such as the ThreeGorges Dam (Jiao et al., 2013), have flood defense as theirprimary responsibility. However, by lowering the water levelbehind the dam to protect against flooding, less electricitywill be produced. There are therefore clear benefits of fore-casting impactful rainfall events at long lead times, allowingmitigation planning for flooding and electricity production.

∗ Corresponding author: Philip E. BETTEmail: philip.bett@metoffice.gov.uk

The relationship between ENSO and the East Asian mon-soon is complex and not fully understood. However, it haslong been clear that a strong El Nino peaking in winter islikely to be followed by above-average rainfall in China thefollowing summer (e.g., Stuecker et al., 2015; He and Liu,2016; Xie et al., 2016; Zhang et al., 2016a, b), although thisresponse is not symmetric under La Nina conditions (Hardi-man et al., 2017). The extreme El Nino event of 1997/98 wasfollowed by devastating floods in the Yangtze River basin(Zong and Chen, 2000; Ye and Glantz, 2005; Yuan et al.,2017): thousands of people died, millions were made home-less, and the economic losses ran into billions of CNY. Inthe subsequent years, much work has gone into better watermanagement and flood prevention, and into improving boththe accuracy and communication of climate forecasts, to pre-vent such a disaster happening again.

Seasonal rainfall forecasts across China have long beenproduced based on statistical relationships with large-scaleclimate phenomena, rather than forecasting precipitation di-

© Copyright [2018], British Crown (administered by Met Office); Chaofan LI, Peiqun ZHANG, Riyu LU, and Hong-Li REN.

AUGUST 2018 BETT ET AL. 919

rectly from dynamical models. For example, Zhu et al.(2008) and Li and Lin (2015) both examined the skill of 500hPa geopotential height (z500) data, from multi-model ensem-bles of dynamical seasonal forecast systems, for forecastingsummer monsoon and Yangtze river valley rainfall, respec-tively. Tung et al. (2013), however, found that using sealevel pressure performed better than using z500 when forecast-ing station-scale summer rainfall in southern China. Kwon etal. (2009), Peng et al. (2014), Wu and Yu (2016), Xing etal. (2016) and Zhang et al. (2016b) all investigated differentstatistical approaches to forecasting summer precipitation inChina, based on various observational indices derived fromsea surface temperatures (SSTs), air temperature and pres-sure. In many cases, these showed an improvement overdynamical models. Wang et al. (2013) found that both dy-namical models and a statistical model based on SSTs andpressure were able to predict the variability in the West Pa-cific subtropical high, which was itself shown to be a goodpredictor of East Asian summer monsoon rainfall. Statisti-cal downscaling techniques have also been shown to improvepredictions of summer precipitation in China over global dy-namical forecast models (e.g., Ke et al., 2011; Liu and Fan,2012).

Recent advances in the dynamical seasonal forecast sys-tem developed at the UK Met Office, GloSea5 (MacLachlanet al., 2015), have resulted in the development of operationaland prototype climate services for the UK in many sectors(e.g., Svensson et al., 2015; Palin et al., 2016; Clark et al.,2017). Recent work has shown that GloSea5 also has usefullevels of skill for various processes in China (Bett et al., 2017;Lu et al., 2017), including for summer precipitation over theYangtze River basin (Li et al., 2016), without having to usestatistical models based on larger-scale drivers.

In parallel to these findings, Golding et al. (2017) demon-strated that there was a clear demand from users for improvedseasonal forecasts for the Yangtze, both from the flood riskand hydropower production communities. The very strong ElNino that developed during the winter of 2015/16 (Zhai et al.,2016) provided a perfect opportunity to develop a trial oper-ational seasonal forecast using GloSea5 for the subsequentsummer of 2016.

We therefore produced forecasts for the upcoming three-month period each week, from February (forecasting March–April–May) to the end of July 2016 (forecasting August–September–October); our focus, however, was on forecast-ing the June–July–August (JJA) period, as that was whereLi et al. (2016) had demonstrated skill. In the last week ofeach month, a forecast for the coming season was issued bythe Met Office to the China Meteorological Administration(CMA).

In this paper, we describe the observed rainfall in theYangtze region in summer 2016, and assess how the real-timeforecasts for May–June–July (MJJ) and JJA performed, witha range of lead times from zero to three months. We describein section 2 the datasets used, and in section 3 our forecastproduction methodology. In section 4 we compare the fore-casts to the observed behavior, and in section 5 discuss pos-

sible future developments.

2. DatasetsThe current operational version of GloSea5 (MacLachlan

et al., 2015) is based on the Global Coupled 2 (GC2) config-uration of the HadGEM3 global climate model, described indetail in Williams et al. (2015) and references therein. WithinHadGEM3-GC2, the atmospheric component [the Met Of-fice Unified Model (Walters et al., 2017)] is coupled to theJULES land surface model (Best et al., 2011), the NEMOocean model (Madec, 2008; Megann et al., 2014) and theCICE sea ice model (Hunke and Lipscomb, 2010; Rae et al.,2015). The atmosphere is modelled on a grid of 0.83 in lon-gitude and 0.55 in latitude, with 85 levels vertically, includ-ing a well-resolved stratosphere; the ocean model is modelledon a 0.25 horizontal grid, with 75 levels vertically.

Using this configuration, GloSea5 runs operationally,producing both forecasts and corresponding hindcasts (in-tended to bias-correct the forecasts). Each day, two initializedforecasts are produced, running out to seven months. To pro-duce a complete forecast ensemble for a given start date, thelast three weeks of individual forecasts are collected togetherto form a 42-member lagged forecast ensemble.

At the same time, an ensemble of hindcasts is producedeach week. As described by MacLachlan et al. (2015), threemembers are run from each of four fixed initialization datesin each month (the 1st, 9th, 17th and 25th), for each of the14 years covering 1996–2009. The full hindcast ensemble ismade by collecting together the four hindcast dates nearest tothe forecast start date, yielding a 12-member, 14-year hind-cast. Note that the hindcast was extended at the end of April2016 to cover 23 years (1993–2015).

This operational hindcast is not intended to be used forskill assessments: with only 12 members, skill estimateswould be biased low (Scaife et al., 2014). However, a sep-arate, dedicated hindcast was produced for skill assessment,with 24 members and 20 years. Using that hindcast, we find acorrelation skill of 0.56 for summer Yangtze rainfall, statisti-cally indistinguishable from the previous value of 0.55 foundby Li et al. (2016).

We use precipitation data from the Global PrecipitationClimatology Project (GPCP) as our observational dataset.This is derived from both satellite data and surface raingauges, covering the period from 1979 to the present at aspatial resolution of 2.5 (Adler et al., 2003). The verifica-tion we present here uses version 2.3 of the data (Adler et al.,2016). Only version 2.2 was available when we started ouroperational trial, although we have confirmed that the choiceof version 2.2 or 2.3 makes negligible difference to our fore-casts or results.

3. Forecast productionTypically, when producing a seasonal forecast, the dis-

tribution of forecast ensemble members is used to repre-

920 YANGTZE SEASONAL FORECASTS VOLUME 35

sent the forecast probability distribution directly. However,experience has shown that the GloSea5 ensemble membersmay contain anomalously small signals, such that the pre-dictable signal only emerges through averaging a large en-semble (Eade et al., 2014; Scaife et al., 2014). While this ef-fect is less pronounced in subtropical regions like the YangtzeBasin, it is still present (Li et al., 2016).

We therefore implemented a simple precipitation fore-casting methodology, based entirely on the historical rela-tionship between the hindcast ensemble means and the ob-served precipitation, averaged over the Yangtze River basinregion (25–35N, 91–122E), following Li et al. (2016),for the season in question. The prediction intervals, derivedfrom the linear regression of the hindcasts to the observations(e.g., Wilks, 2011), provide a calibrated forecast probabilitydistribution.

This is illustrated in Fig. 1, where we show the precipita-tion forecasts issued in late April for MJJ, and in late May forJJA. The distribution of hindcasts and observations is shownas a scatter plot, with the ensemble mean forecast also in-cluded as a green circle. The uncertainty in the linear regres-sion (gray) determines the forecast probabilities (green bars).The GloSea5 data are shown in standardized units—that is,the anomaly of each year from the mean, as a fraction of thestandard deviation of hindcast ensemble means. The obser-vations on the vertical axis are presented as seasonal meansof monthly precipitation totals. The relationship with ENSOis indicated though color-coding of the hindcast points: yearsare labelled as El Nino (red) or La Nina (blue) accordingto whether their Oceanic Nino Index (http://www.cpc.noaa.

gov/products/analysis monitoring/ensostuff/ensoyears.shtml),based on observed SST anomalies in the Nino3.4 region, isabove 0.5 K or below −0.5 K, respectively.

Forecasts like those shown in Fig. 1 were produced eachMonday starting in February 2016, using the forecast modelruns initialized each day of the preceding three weeks to gen-erate the 42-member ensemble, and the four nearest weeksof hindcast runs for the 12-member hindcast ensemble. Theforecast produced near the end of each month was issued tothe CMA: the MJJ release was produced on 25 April and theJJA release on 23 May.

It is important to note that, due to the linear regressionmethod we employ, our forecast probabilities are explicitlylinked to both the hindcasts and the observations. The cor-relations between hindcasts and observations are biased lowdue to the smaller size of the hindcast ensemble comparedto the forecast ensemble—a larger hindcast ensemble wouldnot necessarily alter the gradient of the linear regression, butwould reduce its uncertainty. Our forecast probabilities aretherefore conservative (likely to be too small).

The forecast information provided was designed to showvery clearly and explicitly the uncertainties in the forecastsystem, to prevent overconfidence on the part of potentialdecision-makers. In addition to the scatterplot showing theforecast and the historical relationship (Fig. 1), we also pro-vided the probability of above-average precipitation as a“headline message”. This was accompanied by a contingencytable showing the hit rate and false alarm rate for above-average forecasts over the hindcast period. For the MJJ andJJA forecasts, these are shown in Tables 1 and 2.

Fig. 1. Forecasts for MJJ (produced 25 April 2016) and JJA (produced 23 May 2016), as labeled, using GPCP obser-vations. Observation/hindcast points are color-coded according to their observed winter ENSO index: red points areEl Nino years, blue points are La Nina years, and gray points are neutral. The horizontal width of the green forecastbars is the standard error on the ensemble mean, i.e., the forecast ensemble spread divided by the number of ensemblemembers. The 75% and 95% prediction intervals are shown as gray shading. The variability in the observations isindicated by the pink horizontal dotted lines, at ±1 and 2 standard deviations. The correlation r between hindcast andobservations is marked on each panel (coincidentally the same when rounded).

AUGUST 2018 BETT ET AL. 921

Table 1. Contingency table for forecasts of above-average precipi-tation for the Yangtze region in MJJ, produced on 25 April 2016.The event counts are based on the GPCP observations and ensemblemean hindcasts shown in Fig. 1. The hit rate is the ratio of the num-ber of hits to the number of times above-average conditions wereobserved. The false alarm rate is the ratio of the number of falsealarms to the total number of observed below-average years.

Observed

Above-average precipitation Yes No

Predicted Yes 6 4Hits False alarms

No 4 9Misses Correct rejections

Hit rate: 60%False alarm rate: 30%

Table 2. Contingency table for forecasts of above-average precipi-tation in JJA, produced on 23 May 2016, similar to Table 1.

Observed

Above-average precipitation Yes No

Predicted Yes 9 3Hits False alarms

No 3 8Misses Correct rejections

Hit rate: 75%False alarm rate: 25%

To assess the sensitivity of our results to individual years,we have performed leave-one-out cross-validation for theMJJ and JJA forecasts. We find that the correlation betweenhindcasts and observations in the case of each left-out yeardoes not vary much: 75% of the cases have correlations be-tween 0.41 and 0.47. However, leaving out 1998 does re-duce the performance, as expected: the correlation over theremaining 22 years in that case reduces to r = 0.37 (MJJ) andr = 0.24 (JJA), and the observed value falls outside the 95%prediction range of the forecasts; our procedure does requiresimilar signals to be present in the hindcast period in orderto calibrate the forecasts. Note that this cross-validation pro-cedure is not directly analogous to our actual forecasts: withonly 12 members per year, the hindcast ensemble means aremuch more uncertain than our actual 42-member forecasts for2016, and our cross-validation does not account for this.

4. ResultsThe observed precipitation in May, June, July and Au-

gust 2016 is shown in Fig. 2. We use standardized units hereto show the precipitation anomaly relative to the historicalvariability over the hindcast period (1993–2015). It is clearthat the most anomalously high rainfall was in May and June,

and largely in the eastern half of the basin. July was close tonormal overall when considering the box we were forecastingfor, although there were disastrous floods further north. Au-gust had anomalously low rainfall across most of the region.Yuan et al. (2017) examined the observed summer 2016 rain-fall in China and the Yangtze River basin in detail, includingits relationship to larger-scale drivers: the anomalously lowrainfall in August 2016 is in marked contrast to the situationin 1998, and is related to the behavior of Indian Ocean tem-peratures and the Madden–Julian Oscillation (MJO) duringthe summer.

Figures 3 and 4 show the three-month mean precipitationanomalies for MJJ and JJA respectively, for both GPCP andthe forecast averages from the GloSea5 model output. Whilewe do not expect the spatial patterns to match in detail [con-sidering the skill maps of Li et al. (2016)], the overall signalis similar to the observations, with stronger anomalous pre-cipitation in the eastern region in MJJ, and closer-to-averageprecipitation in JJA.

We examine our forecasts for the Yangtze basin box morequantitatively in Figs. 5 and 6, where we show the variationwith lead time of the hindcast–observation correlation, the2016 forecast signal, and the probability of above-averageprecipitation, for MJJ and JJA respectively. Neither thehindcast–observation correlation nor the forecast signal varysignificantly with lead time; indeed, they are remarkably con-sistent back to three months before the forecast season, andwhen the 23-year hindcast is introduced at the end of April.

The forecasts did a good job of giving an indication ofprecipitation in the coming season. For MJJ, the forecastgave a high probability of above-average precipitation (80%),and it was observed to be above average. In JJA, the meanprecipitation was observed to be slightly below average, dueto the strong drier-than-average signal in August, although itwas within a standard deviation of the interannual variabil-ity. While our forecast marginally favored wetter than aver-age conditions (65% probability of above-average rainfall), itwas correctly near to the long-term mean, and the observedvalue was well within the forecast uncertainties.

5. Discussion and conclusionsThe heavy rainfall in the Yangtze River region in early

summer 2016 was at a similar level to that of 1998, andcaused heavy flooding (WMO, 2017; Yuan et al., 2017).While deaths due to the flooding were roughly an order ofmagnitude fewer than those caused by the 1998 floods (i.e.,hundreds rather than thousands of lives), the economic lossesnevertheless ran into tens of billions of CNY. Furthermore,it was reported that insurance claims, mostly from agricul-tural losses, amounted to less than 2% of the total economicloss, suggesting significant levels of underinsurance (Podlahaet al., 2016). The prior experience of the 1998 El Nino-enhanced flooding, and the high levels of awareness of thestrong El Nino in winter 2015/16, meant that dams along theYangtze were prepared in anticipation of high levels of rain-

922 YANGTZE SEASONAL FORECASTS VOLUME 35

Fig. 2. Observed precipitation from GPCP (version 2.3) for May, June, July and August (as labeled), in standardized units withrespect to the 1993–2015 period. The Yangtze box used for the forecasts is marked as a red rectangle, with a pink polygonshowing the physical Yangtze River catchment. Major rivers are marked in blue.

Fig. 3. Mean precipitation for 2016-MJJ in the GPCP observations and GloSea5 forecast signal, (as labeled), in standardizedunits. The GloSea5 data have been regridded to match the lower-resolution observations.

fall. Our forecasts from GloSea5, produced using the simplemethodology described here, contributed to the confidence of

users adapting to the impending rainfalla.Our verification has shown that our forecasts gave a good

aGolding, N., C. Hewitt, P. Bett, M. Liu, and P. Zhang: Co-Development of a Seasonal Forecast Climate Service: Supporting flood risk management forthe Yangtze River Basin. (in preparation)

AUGUST 2018 BETT ET AL. 923

Fig. 4. Mean precipitation for 2016-JJA in the GPCP observations and GloSea5 forecast signal, (as labeled), in standardizedunits. The GloSea5 data have been regridded to match the lower-resolution observations.

Fig. 5. Time series showing the behavior of the MJJ forecasts and hindcasts with lead time: (a)Correlation between observations and the operational hindcasts available each week. The finalpoint was produced using 23 years, whereas only 14 were available before that. The shadingindicates 95% confidence intervals using the Fisher Z-transformation. (b) The forecast signalshown as 95% and 75% prediction intervals (boxes) and the ensemble mean (blue line). Theobserved precipitation is marked as an orange horizontal line from May. The observed his-torical mean and standard deviation over the hindcast period are marked as a dashed line andorange shading respectively. (c) The forecast probability of above-average precipitation. Thefinal forecast issued for MJJ, produced 25 April, is highlighted with a gray vertical bar.

924 YANGTZE SEASONAL FORECASTS VOLUME 35

Fig. 6. Time series showing the behavior of the JJA forecasts and hindcasts with lead time, fol-lowing the same format as Fig. 5. In (a) (hindcast–observations correlation), the line becomesthicker when 23 years of hindcasts are available. We mark with a blue cross and error bar thecorrelation skill derived from the assessment hindcast (see text for details). The final forecastissued for JJA, on 23 May, is highlighted across all panels.

indication of the observed levels of precipitation for bothMJJ and JJA averages over the large Yangtze Basin region.A greater degree of both spatial and temporal resolution—splitting the basin into upper and lower sections, and pro-ducing additional forecasts at a monthly timescale—wouldof course be preferable to users. However, smaller regionsand shorter time periods may well be less skillful, so furtherwork is needed to assess how best to achieve skillful forecastsin these cases.

One significant improvement would be to increase theensemble size of the hindcast. During 2017 the GloSea5system was changed from three hindcast members per startdate to seven. This could result in noticeable improve-ments in forecasts like those described here, as the hindcast–observations relationship will be less uncertain, especiallywhen a predictable signal is present, such as from El Nino.Improvements in the underlying climate model, such as toparametrized convective precipitation, and the simulation ofthe monsoon and features like the MJO, could also improvethe forecast skill.

We will be issuing forecasts again in 2017. However,

unlike 2016, in 2017 there are no strong drivers, such as ElNino. Nevertheless, understanding the behavior of the fore-cast system under such conditions will be informative, forboth the users and the producers of the forecasts. Ultimately,trial climate services such as this help to drive forecast devel-opment, improve understanding of forecast uncertainties, andpromote careful use by stakeholders in affected areas.

Acknowledgements. This work and its contributors (PB, AS,ND, DS, CH, NG) were supported by the UK–China Research &Innovation Partnership Fund through the Met Office Climate Sci-ence for Service Partnership China as part of the Newton Fund. CLand RL were supported by the National Natural Science Founda-tion of China (Grant No. 41320104007). HR was supported by theProject for Development of Key Techniques in Meteorological Op-eration Forecasting (Grant No. YBGJXM201705). The trial forecastservice was first suggested by AS in 2015. The GPCP precipita-tion data were provided by the NOAA/OAR/ESRL PSD, Boulder,Colorado, USA, via their website at http://www.esrl.noaa.gov/psd/.The Yangtze River basin shapefile used in the maps was ob-tained from http://worldmap.harvard.edu/data/geonode:ch wtrshed

AUGUST 2018 BETT ET AL. 925

30mar11 and is based on the watersheds shown in the China Envi-ronmental Atlas (2000), © Chinese Academy of Sciences, Environ-mental Data Center.

REFERENCES

Adler, R., and Coauthors, 2016: The new version 2.3 of the GlobalPrecipitation Climatology Project (GPCP) monthly analy-sis product. University of Maryland. [Available online fromhttp://eagle1.umd.edu/GPCP ICDR/GPCP Monthly.html]

Adler, R. F., and Coauthors, 2003: The version-2 Global Pre-cipitation Climatology Project (GPCP) monthly precipita-tion analysis (1979–Present). Journal of Hydrometeorology,4, 1147–1167, https://doi.org/10.1175/1525-7541(2003)004<1147:TVGPCP>2.0.CO;2.

Best, M. J., and Coauthors, 2011: The Joint UK Land Environ-ment Simulator (JULES), model description—Part 1: Energyand water fluxes. Geoscientific Model Development, 4, 677–699, https://doi.org/10.5194/gmd-4-677-2011.

Bett, P. E., and Coauthors, 2017: Skill and reliability of sea-sonal forecasts for the Chinese energy sector. Journal of Ap-plied Meteorology and Climatology, 56, 3099–3114, https://doi.org/10.1175/jamc-d-17-0070.1.

Clark, R. T., P. E. Bett, H. E. Thornton, and A. A. Scaife, 2017:Skillful seasonal predictions for the European energy in-dustry. Environmental Research Letters, 12, 024002, https://doi.org/10.1088/1748-9326/aa57ab.

Eade, R., D. Smith, A. Scaife, E. Wallace, N. Dunstone, L. Her-manson, and N. Robinson, 2014: Do seasonal-to-decadal cli-mate predictions underestimate the predictability of the realworld? Geophys. Res. Lett., 41, 5620–5628, https://doi.org/

10.1002/2014gl061146.Golding, N., C. Hewitt, P. Q. Zhang, P. Bett, X. Y. Fang, H. Z. Hu,

and S. Nobert, 2017: Improving user engagement and up-take of climate services in China. Climate Services, 5, 39–45,https://doi.org/10.1016/j.cliser.2017.03.004.

Hardiman, S., and Coauthors, 2017: The asymmetric response ofYangtze River basin summer rainfall to El Nino/La Nina. En-vironmental Research Letters, https://doi.org/10.1088/1748-9326/aaa172. (in press)

He, J. H., and B. Q. Liu, 2016: The East Asian subtropical sum-mer monsoon: Recent progress. Journal of MeteorologicalResearch, 30, 135–155, https://doi.org/10.1007/s13351-016-5222-z.

Hunke, E. C., and W. H. Lipscomb, 2010: CICE: The Los AlamosSea Ice model documentation and software user’s manual,version 4.1. Report LA-CC-06-012, Los Alamos NationalLaboratory. [Available online from http://oceans11.lanl.gov/

trac/CICE]Jiao, M. Y., and Coauthors, 2013: Addressing the potential climate

effects of China’s Three Gorges Project. WMO Bulletin, 62(Special Issue), 49–53. [Available online from http://library.wmo.int/opac/index.php?lvl=bulletin display&id=2738]

Ke, Z. J., P. Q. Zhang, L. J. Chen, and L. M. Du, 2011: An exper-iment of a statistical downscaling forecast model for summerprecipitation over China. Atmospheric and Oceanic ScienceLetters, 4, 270–275, https://doi.org/10.1080/16742834.2011.11446941.

Kwon, H. H., C. Brown, K. Q. Xu, and U. Lall, 2009: Seasonal andannual maximum streamflow forecasting using climate infor-mation: Application to the Three Gorges Dam in the Yangtze

River basin, China. Hydrological Sciences Journal, 54, 582–595, https://doi.org/10.1623/hysj.54.3.582.

Li, C. F., and Coauthors, 2016: Skillful seasonal predictionof Yangtze river valley summer rainfall. Environmental Re-search Letters, 11, 094002, https://doi.org/10.1088/1748-9326/11/9/094002.

Li, F., and Z. D. Lin, 2015: Improving multi-model ensembleprobabilistic prediction of Yangtze River valley summer rain-fall. Adv. Atmos. Sci., 32, 497–504, https://doi.org/10.1007/

s00376-014-4073-8.Liu, Y., and K. Fan, 2012: Improve the prediction of summer

precipitation in the Southeastern China by a hybrid statisti-cal downscaling model. Meteor. Atmos. Phys., 117, 121–134,https://doi.org/10.1007/s00703-012-0201-0.

Lu, B., A. A. Scaife, N. Dunstone, D. Smith, H. L. Ren, Y. Liu, andR. Eade, 2017: Skillful seasonal predictions of winter precip-itation over southern China. Environmental Research Letters,12, 074021, https://doi.org/10.1088/1748-9326/aa739a.

MacLachlan, C., and Coauthors, 2015: Global Seasonal forecastsystem version 5 (GloSea5): A high-resolution seasonal fore-cast system. Quart. J. Roy. Meteor. Soc., 141, 1072–1084,https://doi.org/10.1002/qj.2396.

Madec, G., 2008: NEMO ocean engine. Note du Pole demodelisation, Institut Pierre-Simon Laplace (IPSL). France,No. 27. [Available online from http://www.nemo-ocean.eu/

About-NEMO/Reference-manuals]Megann, A., and Coauthors, 2014: GO5.0: The joint NERC-

Met Office NEMO global ocean model for use in coupledand forced applications. Geoscientific Model Development, 7,1069–1092, https://doi.org/10.5194/gmd-7-1069-2014.

Palin, E. J., A. A. Scaife, E. Wallace, E. C. D. Pope, A. Arribas,and A. Brookshaw, 2016: Skillful seasonal forecasts of win-ter disruption to the U.K. transport system. Journal of AppliedMeteorology and Climatology, 55, 325–344, https://doi.org/

10.1175/jamc-d-15-0102.1.Peng, Z. L., Q. J. Wang, J. C. Bennett, P. Pokhrel, and Z. R. Wang,

2014: Seasonal precipitation forecasts over China usingmonthly large-scale oceanic-atmospheric indices. J. Hydrol.,519, 792–802, https://doi.org/10.1016/j.jhydrol.2014.08.012.

Plate, E. J., 2002: Flood risk and flood management. J. Hydrol.,267, 2–11, https://doi.org/10.1016/s0022-1694(02)00135-x.

Podlaha, A., S. Bowen, and C. Darbinyan, 2016: July 2016 GlobalCatastrophe Recap. Aon Benfield Impact Forecasting, [Avail-able online from http://thoughtleadership.aonbenfield.com/

sitepages/display.aspx?tl=601]Rae, J. G. L., H. T. Hewitt, A. B. Keen, J. K. Ridley, A. E. West,

C. M. Harris, E. C. Hunke, and D. N. Walters, 2015: De-velopment of the Global Sea Ice 6.0 CICE configuration forthe Met Office Global Coupled model. Geoscientific ModelDevelopment, 8, 2221–2230, https://doi.org/10.5194/gmd-8-2221-2015.

Scaife, A. A., and Coauthors, 2014: Skillful long-range predic-tion of European and North American winters. Geophys. Res.Lett., 41, 2514–2519, https://doi.org/10.1002/2014gl059637.

Stuecker, M. F., F. F. Jin, A. Timmermann, and S. McGre-gor, 2015: Combination mode dynamics of the AnomalousNorthwest Pacific Anticyclone. J. Climate, 28, 1093–1111,https://doi.org/10.1175/jcli-d-14-00225.1.

Svensson, C., and Coauthors, 2015: Long-range forecasts ofUK winter hydrology. Environmental Research Letters, 10,064006, https://doi.org/10.1088/1748-9326/10/6/064006.

Tung, Y. L., C. Y. Tam, S. J. Sohn, and J. L. Chu, 2013: Improv-

926 YANGTZE SEASONAL FORECASTS VOLUME 35

ing the seasonal forecast for summertime South China rainfallusing statistical downscaling. J. Geophys. Res. Atmos., 118,5147–5159, https://doi.org/10.1002/jgrd.50367.

Walters, D., and Coauthors, 2017: The Met Office Unified ModelGlobal Atmosphere 6.0/6.1 and JULES Global Land 6.0/6.1configurations. Geoscientific Model Development, 10, 1487–1520, https://doi.org/10.5194/gmd-10-1487-2017.

Wang, B., B. Q. Xiang, and J. Y. Lee, 2013: Subtropical high pre-dictability establishes a promising way for monsoon and trop-ical storm predictions. Proceedings of the National Academyof Sciences of the United States of America, 110, 2718–2722,https://doi.org/10.1073/pnas.1214626110.

Wilks, D. S., 2011: Statistical forecasting. International Geo-physics, 100, 215–300, https://doi.org/10.1016/b978-0-12-385022-5.00007-5.

Williams, K. D., and Coauthors, 2015: The Met Office GlobalCoupled model 2.0 (GC2) configuration. Geoscientific ModelDevelopment, 8, 1509–1524, https://doi.org/10.5194/gmd-8-1509-2015.

WMO, 2017: WMO statement on the state of the global climate in2016. WMO-No. 1189. Geneva: World Meteorological Orga-nization. [Available online from http://library.wmo.int/opac/

doc num.php?explnum id=3414]Wu, Z. W., and L. L. Yu, 2016: Seasonal prediction of the East

Asian summer monsoon with a partial-least square model.Climate Dyn., 46, 3067–3078, https://doi.org/10.1007/

s00382-015-2753-4.Xie, S. P., Y. Kosaka, Y. Du, K. M. Hu, J. S. Chowdary, and

G. Huang, 2016: Indo-western Pacific ocean capacitor andcoherent climate anomalies in post-ENSO summer: A re-view. Adv. Atmos. Sci., 33, 411–432, https://doi.org/10.1007/

s00376-015-5192-6.Xing, W., B. Wang, and S. Y. Yim, 2016: Long-lead sea-

sonal prediction of China summer rainfall using an EOF-PLSregression-based methodology. J. Climate, 29, 1783–1796,https://doi.org/10.1175/jcli-d-15-0016.1.

Ye, Q., and M. H. Glantz, 2005: The 1998 Yangtze floods: The useof short-term forecasts in the context of seasonal to interan-nual water resource management. Mitigation and AdaptationStrategies for Global Change, 10, 159–182, https://doi.org/

10.1007/s11027-005-7838-7.Yu, F. L., Z. Y. Chen, X. Y. Ren, and G. F. Yang, 2009: Anal-

ysis of historical floods on the Yangtze River, China: Char-acteristics and explanations. Geomorphology, 113, 210–216,https://doi.org/10.1016/j.geomorph.2009.03.008.

Yuan, Y., H. Gao, W. J. Li, Y. J. Liu, L. J. Chen, B. Zhou, and Y.H. Ding, 2017: The 2016 summer floods in China and asso-ciated physical mechanisms: A comparison with 1998. Jour-nal of Meteorological Research, 31, 261–277, https://doi.org/

10.1007/s13351-017-6192-5.Zhai, P. M., and Coauthors, 2016: The strong El Nino of 2015/16

and its dominant impacts on global and China’s climate. Jour-nal of Meteorological Research, 30, 283–297, https://doi.org/

10.1007/s13351-016-6101-3.Zhang, W. J., H. Y. Li, M. F. Stuecker, F. F. Jin, and A. G.

Turner, 2016a: A new understanding of El Nino’s impactover East Asia: Dominance of the ENSO combination mode.J. Climate, 29, 4347–4359, https://doi.org/10.1175/jcli-d-15-0104.1.

Zhang, W. J., and Coauthors, 2016b: Unraveling El Nino’s impacton the East Asian Monsoon and Yangtze River summer flood-ing. Geophys. Res. Lett., 43, 11 375–11 382, https://doi.org/

10.1002/2016gl071190.Zhu, C., C. K. Park, W. S. Lee, and W. T. Yun, 2008: Sta-

tistical downscaling for multi-model ensemble prediction ofsummer monsoon rainfall in the Asia-Pacific region usinggeopotential height field. Adv. Atmos. Sci., 25, 867–884,https://doi.org/10.1007/s00376-008-0867-x.

Zong, Y. Q., and X. Q. Chen, 2000: The 1998 flood on the Yangtze,China. Natural Hazards, 22, 165–184, https://doi.org/10.1023/

a:1008119805106.