Prediction of the Mumbai and the Mount Abu Extreme Rainfall Events during the 2017 Monsoon Season by the IITM-MME and the S2S Climate Forecast Models

EXTENDED ABSTRACT

“Prediction of the Mumbai and the Mount Abu Extreme Rainfall Events during the 2017 Monsoon Season by the IITM-MME and the S2S Climate Forecast Models”

Saloua PEATIER (Student, University Claude-Bernard, Lyon)

Dr.Rajib CHATTOPADHYAY (Scientist D, Indian Institute of Tropical Meteorology)

Introduction:

The Indian society is influenced by the annual cycle as well as the intra-seasonal variability of the summer monsoon system during June-September. Mostly because 70% of its working population depends on the agricultural activities during this season [1].

In addition to this intra-seasonal variability of monsoon (manifested as active and break spells), India has to face extreme precipitation events. Several studies [2] noticed an increase in the magnitude and the frequency of the extreme rainfall events and make the hypothesis that this increase is a consequence of the global warming. The extremes, or the high intensity precipitation events, develop in the large scale monsoon background. This could imply that there could be a hope in the outlook of high intensity rainfall events if the large scale background is predicted with greater skill [3].

A revolution in weather and climate forecasting is in progress, made possible by theoretical and technological advances [4]. In the Indian context, the multi model multi-ensemble (MME) runs at the Indian Institute of Tropical Meteorology (IITM), using NCEP-CFSv2. Incorporated as a part of the Monsoon Mission Project, this system revealed that the transition from a deterministic to a probabilistic approach and the creation of Multi-Model Multi-Ensemble forecasts allow a better prediction of the large-scale circulation of Monsoon [5]. Recently, ECMWF has started archiving forecasts from state-of-the-art climate models as a part of a new S2S project [6]. It is shown that the S2S database has potential applications in predicting the extreme events [7].

The main objective of our study is the inter comparison of IITM-MME and S2S model outlooks for the Mumbai and the Mount-Abu events (monsoon season 2017) to understand the capability and biases of current climate forecast models in predicting these events.

Data, Method and Models:

The present study used only retrospectives forecast runs from IITM-MME model, S2S ECMWF model and S2S Meteo-France model.

The IITM-MME model is a Multi-Model Multi-Ensemble forecast system based on the Climate Forecast System (CFS) models version 2 and the Global Forecast System (GFS), both from the National Centers for Environmental Prediction (NCEP). The IITM-MME contains 16 ensemble members and runs once a week [5]. The ECMWF and the Meteo-France data were downloaded from the S2S database (http://apps.ecmwf.int/datasets/data/s2s/levtype=sfc/type=cf/). Both of these models contain 51 ensemble members and are running twice a week.

For each event and for both of the models, two different lead-times were studied. The comparison was done regarding three main factors: the spatial correlation with the observations, the time evolution of the event and the Probability Distribution Function (PDF) of the rainfalls. The observation data were taken from the Global Precipitation Measurement (GPM) satellite data (https://pmm.nasa.gov/data-access).

Results:

The Mumbai flood occurred on the 29/08/2017 and had heavy consequences on the city. Regarding the observations (Figure 1), the event day seems to correspond with an active composite of the monsoon, with rainfalls over the Indian West coast and Central India. The event peak is a strong positive anomaly, corresponding to a strong active spell of the monsoon. On the six day period around the flood, the ECMWF and the IITM-MME models gave a skillful outlook of the events for the lead-times Day+5/Day+6. This observation was confirmed by the significant pattern correlation coefficients of these two models with the observations and probability distribution function (not shown). However, both of the two models had spatial biases in the location on the event, with a shift in the maximum center of rainfalls (Figure 2.a) and they underestimated the rainfall amplitude of the event (Figure 2.b).

The Mount-Abu flood occurred on the 24/07/2017 and strongly impacted the Rajasthan state. This local event did not correspond to a strong active spell and almost presented a break composite of monsoon. Similarly that for the Mumbai flood, the IITM-MME and the ECMWF models gave an actionable outlook of the event. But here also, the rainfall amplitude was underestimated and some spatial biases were found in the location of the event. It is the ECMWF model which performed the best in the estimation of the PDF and the time evolution of the event.

Conclusions:

Predictions of weather and climate are always uncertain, specially when it comes to the prediction of random high-intensity events. However, monsoon has a large-scale background which arises the extreme events. We observed an actionable outlook of the events with the ECMWF and the IITM-MME models, lead-times Day+4/Day+5. So there is a real hope to predict the extreme events by obtaining a more skillful view of the large-scale circulation. The models did not perform unevenly depending on the extreme event considered. We can formulate the hypothesis to reduce this inequality by making a Multi-Model Multi-Ensemble (MME) outlook of the S2S and IITM models.

Figure Captions:

Figure 1: GPM satellite observation data : daily rainfalls [mm.day] during the Mumbai event, from 27/08/2017 (d-2) to the 01/09/2018 (d+3).

Figure 2: The Mumbai event outlooks. (a) Time evolution of the event (17N-21N, 71E-75E). The leftmost (rightmost) plot is showing the outlook of the event for the firs (second) IC, lead-time Day+5 (Day+12). (b) Spatial biases [model-observation] of the rainfalls on the 29/082017 for the first Initial Conditions in [A] IITM-MME data, [B] ECMWF data and [C] Meteo France data.

References :

[1] Gadgil S. 2003. Annu. Rev. Earth Planet, Sci, 2003. 31:429-67.

[2] Goswami BN. & al. 2006. Science, Vol. 314,issue 5804. pp. 1442-1445.

[3] Goswami BN. & al. 2003. Geophysical Research Letters. Doi:10.1029/2002GL016734

[4] Palmer T. 2006. Chap. 1 Palmer and Hagedorn (ed), Predictability of weather and climate.. UK, Cambridge: University Press. ISBN-13: 978-0521848824

[5] Chattopadhyay R. & al. 2018. IITM Research Report RR-139. http://www.tropmet.res.in/~lip/Publication/RR-pdf/RR-139.pdf

[6] Vitart F. & al. 2017. Bull. Am. Met. Soc. 98(1). Doi: 10.1175/BAMS-D-16-0017.1

[7] Vitart F. & Robertson A. 2018. NPJ. Climate and Atmospheric Science. 1:3. Doi:10.1038/s41612-018-0013-0. (https://www.nature.com/articles/s41612-018-0013-0.pdf)