Shipra Jain 1,2 , Adam A Scaife - PowerPoint Presentation

1. Shipra Jain1,2, Adam A Scaife3,4, Nick Dunstone3, Doug Smith3, Saroj Kanta Mishra2 and Ruth Doherty11School of Geosciences, University of Edinburgh, UK2Indian Institute of Technology Delhi, India3Met Office, UK4University of Exeter, UKEmail: shipra.npl@gmail.comCurrent Risk of Extreme Monsoon Rainfall over India using Large Ensemble SimulationsThe UNSEEN approachSee full article here: https://iopscience.iop.org/article/10.1088/1748-9326/ab7b98

2. Indian Monsoon(Some Background information)2 Time: June-Sep70-80% of the total annual rainfallSpatial coverage: almost whole countryTime: Dec-Feb20-30% of the total annual rainfallSpatial coverage: ~20% Indian Summer Monsoon OrSouth-west MonsoonTwo monsoon seasonsIndian Winter Monsoon OrNorth-east MonsoonRegion of maximum rainfall

3. Importance of Summer Rainfall for IndiaLarge sectoral and socio-economic impacts, e.g.3DroughtsFloodsAgricultureEmploymentGrowing interest in amount of rainfall received in the summer monsoon season!

4. Factors influencing the seasonal extremes4Internal Variability: due to natural internal processes e.g. phenomena in the oceans: ENSO, IOD, NAO, PDO etc.External Variability: due to external processes e.g. changes in natural forcing, changes in anthropogenic forcingFor example: Indian drought of 2002: JJA rainfall was ~19% below normal-third largest drought over the last century, shortest monsoon on record-devastating, high losses in agriculture and economy-though extreme but not unprecedented- could have been foreseen in advance-internal variability cannot be ruled as a cause of this drought (Gadgil et al. 2002) Find full list of references here:https://iopscience.iop.org/article/10.1088/1748-9326/ab7b98

5. Given that internal variability can also lead to the extreme rainfall, the next question that comes is:What are the chances of extreme rainfall due to the internal variability in current climate?5Can we determine the chance of flood or drought using observations?-Yes! But high statistical uncertainty as observations are limited to one values per summerCan we use models to better determine the chance of flood or drought?-Yes! If model outputs are statistically indistinguishable from the observations. Can we determine the chance of unprecedented extreme rainfall, that is not yet seen in observations?-Yes! Using large ensemble of simulations from models.How? -Using UNSEEN

6. What is ‘UNSEEN’?UNprecedented Simulated Extremes using ENsemblesAdvantages:Background dynamical conditions can be identified, which is not possible with observations if extrapolated outside their rangePredictive skill over the target area is not needed6statistical framework: risk estimation using a large ensemble to sample a broad range of internal variability ApplicationsFor any given winter, there is a 7% chance that the rainfall would exceed the observed record rainfall in at least one month over south east England (Thompson et al. 2018)for each summer, there is a 10% risk of an unprecedented hot month in SE China (Thompson et al. 2019)Proposed by Thompson et al. (2018)Find full list of references here:https://iopscience.iop.org/article/10.1088/1748-9326/ab7b98

7. What is unique in our analysis?7There exist only few studies so far (<5) and all of them apply this method to the outputs of one model. First application of this method to a large ensemble of hindcasts from multiple modelsAlso the first application of UNSEEN to Indian monsoonSample a broad range of variability in similarly forced modelsIncrease number of samples as compared to any single model or obs. Examine the driving dynamical/ meteorological conditionsleading to extreme events in similarly forced models.

8. Data8Model OutputsEight Seasonal prediction systems from different modeling centresPhysical ProcessesModels: GloSea5 (UK), ECMWF-S4 (UK), CanCM4 (Canada), CFS (USA), MIROC5 (Japan), MPI (Germany), JMA (Japan), POAMA (Australia)Time period: Varying (typically ~30 years from each model)Start Date: ~1st MayLead time: 1-5 monthsNumber of ensembles: Varying (7-30)Total summer monsoon years: 3169 (189-900) RainfallSource: IMD, 1901-2013, 113 yearsSpatial domain: India (land-only), quarter degree productSSTSource: NOAA, Nino 3.4 index, 1950-2016Definition of Extremes:+10% the clim. Mean = flood-10% of the clim mean = droughtOutside the range of obs. = unprecedentedPeriod of analysis: seasonal means for June-August (summer monsoon season)Observations Rainfall: India Met. Dep. (IMD)SSTs: Nino 3.4 index from NOAA

9. Fidelity Tests9Model SelectionStep 1: Bias correctionStep 2: Resampling to form 10,000 representative time series for each model (boot strapping) by randomly selecting ensemble member for each yearMeanStandard DeviationSkewnessKurtosis

10. Step 3: Select models that are statistically indistinguishable from the observations-within 95% confidence intervalSummary of the fidelity tests10Large multimodel ensembleStep 4: Create large multimodel ensembleOut of 8 models, 5 pass the testsOut of 3169 members, we choose 1669 (from 5 models)i.e. 1669 individual realizations of summer rainfall!This is ~15x larger than full IMD obs. record

11. Good agreement with the obs. for low intensity extremes.High intensity extremes are underestimated in obs.12% chance of flood and 14% chance of drought for any given summer over India. The chance of an unprecedented drought (>23% deficit) in the current climate is ~1.6% and unprecedented flood (>16% excess) is 2.6%. Higher risk of drought than floods of same intensityDrought are more intense than floodsEstimated Risk using Large Multimodel ensemble in current climate11Even a risk of >30% drought around once in two centuries, not yet seen in observations!

12. Background Dynamical Conditions in the OceansDespite the reported non-stationarity of the observed ENSO-monsoon relationship over the last few decades (Kumar et al., 1999), ENSO provides a clear impact on risk of extremes!Maybe some impact of IOD but not significant!12Difference in SSTs for top 10 wettest and driest summersClear ENSO signal!Find full list of references here:https://iopscience.iop.org/article/10.1088/1748-9326/ab7b98

13. Influence of ENSO on risk of Flood or Drought13Asymmetry in the risk of floods and droughtsprobability of severe droughts during El Nino is higher than floods during La Nina.

14. Asymmetry in risk of extremes is related to asymmetry in ENSO14Frequency of JJA SST anomalies over the Nino 3.4 region using the ensembleEl Nino events reaching greater magnitude more frequently than La Nina events for SST anomalies exceeding 1.6 K.Higher magnitude of El Nino increases the chance of drought in models.Asymmetry between floods and droughts becomes severe with increasing ENSO intensityFractional contribution of flood and drought for each SST range

15. Key Points For any given summer: 12% chance of flood and 14% chance of drought. Dynamical conditions show a clear ENSO signal in the most extreme wet and dry summers. Droughts are more likely than floods due to the ENSO phase asymmetry and this asymmetry becomes more severe as the intensity of the ENSO events increases. The chance of an unprecedented drought (>23% deficit) in the current climate is ~1.6% and unprecedented flood (>16 excess) is 2.6%. Also a chance of record-breaking drought with ~30% deficit which is expected to occur around once every two centuries. The socio-economic impact of such a drought on India would be well-beyond anything that has occurred over the last century.This work was conducted through the Weather and Climate Science for Service Partnership (WCSSP) India, a collaborative initiative between the Met Office, supported by the UK Government’s Newton Fund, and the Indian Ministry of Earth Sciences (MoES).15Questions? See full article here: https://iopscience.iop.org/article/10.1088/1748-9326/ab7b98More Questions? Suggestions?Email: shipra.npl@gmail.com