Assessment of extreme rainfall events for iFLOWS Mumbai in NCUM regional forecasting system

Abstract

Multiple record-breaking rainfall events were observed along the Western Ghats (WG) during the recent monsoon seasons (2019–2021). Rainfall amounts of up to > 200 mm/day (Extreme rainfall, ER) were recorded especially over the Mumbai region (19.07 N, 72.8 E) causing flooding, landslides, damage to infrastructure and loss of life. Thus, to enhance the resilience of this region by providing early warning for flooding, the National Center for Medium-Range Weather Forecasting Unified model’s regional forecasting system (NCUM-reg) provides rainfall forecasts up to 3 days (72-h), which are utilized in the integrated flood warning system hydrological model. This study focuses on evaluating the performance of NCUM-reg forecasts during ER events. For this purpose, we have systematically performed verification of regional model operational forecasts using the suite of observations (rain gauge, satellite) and newly generated NCMRWF’s regional reanalysis, Indian Monsoon Data Assimilation and Analysis (IMDAA). Key findings indicate that NCUM-reg model with explicit convection is performing well in representing the synoptic and dynamic features of the ER events similar to those observed. Quantitative assessment of the forecasts shows the strength of in-situ observations. In addition, the results summarize the importance of continuous and quality-controlled observations and stress the need for collective efforts of observations and new verification metrics (like process-oriented diagnostics) to enhance our understanding and as well as the model’s ability in forecasting such events.

Appendices

Appendix 1

Statistical scores:

$${\text{ETS}} = \frac{{{\text{hits}} - {\text{hits}}_{{{\text{random}}}} }}{{{\text{hits}} + {\text{misses}} + {\text{false}}\;{\text{alarms}} - {\text{hits}}_{{{\text{random}}}} }}$$

(1)

$${\text{hits}}_{{{\text{random}}}} = \frac{{\left( {{\text{hits}} + {\text{misses}}} \right)\left( {{\text{hits}} + {\text{fasle}}\;{\text{alarms}}} \right)}}{{{\text{total}}}}$$

(2)

ETS range from − 1/3 to 1, with 1 indicating perfect skill and 0 is no skill.

Hits, false alarms, and misses are calculated based on a 2 × 2 contingency table between forecast and observations, given below.

Forecast (F)	Observed (O)
		Yes	No	Total
	Yes	Hits	False alarms	F—Yes
	No	Misses	Correct negatives	F—No
	Total	O –Yes	O—No	Total

Appendix 2

The calculation of the atmospheric apparent heat source Q₁ and the apparent moisture sink Q₂ are given by Yanai et al. (1973):

$${\text{Q}}1 = {\text{C}}_{{\text{P}}} \left[ {\frac{{\partial {\text{T}}}}{{\partial {\text{t}}}} + {\text{V}} \cdot \nabla {\text{T}} + \left( {\frac{{\text{p}}}{{{\text{p}}0}}} \right)\upomega \frac{{\partial\uptheta }}{{\partial {\text{p}}}}} \right]$$

$${\text{Q}}2 = - {\text{L}}\left[ {\frac{{\partial {\text{q}}}}{{\partial {\text{t}}}} + {\text{V}} \cdot \nabla {\text{q}} +\upomega \frac{{\partial {\text{q}}}}{{\partial {\text{p}}}}} \right]$$

where c_p is the specific heat at constant pressure, p0 = 1000 hPa; T is the temperature; θ is the potential temperature; ω is the vertical velocity; V is the horizontal wind vector; L is a constant of the latent heat condensation; and q is the specific humidity.