Role of land state in a high resolution mesoscale model for simulating the Uttarakhand heavy rainfall event over India

Abstract

In 2013, Indian summer monsoon witnessed a very heavy rainfall event (>30 cm/day) over Uttarakhand in north India, claiming more than 5000 lives and property damage worth approximately 40 billion USD. This event was associated with the interaction of two synoptic systems, i.e., intensified subtropical westerly trough over north India and north-westward moving monsoon depression formed over the Bay of Bengal. The event had occurred over highly variable terrain and land surface characteristics. Although global models predicted the large scale event, they failed to predict realistic location, timing, amount, intensity and distribution of rainfall over the region. The goal of this study is to assess the impact of land state conditions in simulating this severe event using a high resolution mesoscale model. The land conditions such as multi-layer soil moisture and soil temperature fields were generated from High Resolution Land Data Assimilation (HRLDAS) modelling system. Two experiments were conducted namely, (1) CNTL (Control, without land data assimilation) and (2) LDAS, with land data assimilation (i.e., with HRLDAS-based soil moisture and temperature fields) using Weather Research and Forecasting (WRF) modelling system. Initial soil moisture correlation and root mean square error for LDAS is 0.73 and 0.05, whereas for CNTL it is 0.63 and 0.053 respectively, with a stronger heat low in LDAS. The differences in wind and moisture transport in LDAS favoured increased moisture transport from Arabian Sea through a convectively unstable region embedded within two low pressure centers over Arabian Sea and Bay of Bengal. The improvement in rainfall is significantly correlated to the persistent generation of potential vorticity (PV) in LDAS. Further, PV tendency analysis confirmed that the increased generation of PV is due to the enhanced horizontal PV advection component rather than the diabatic heating terms due to modified flow fields. These results suggest that, two different synoptic systems merged by the strong interaction of moving PV columns resulted in the strengthening and further amplification of the system over the region in LDAS. This study highlights the importance of better representation of the land surface fields for improved prediction of localized anomalous weather event over India.

Appendix 1

The verification measures for categorical variables like ETS, POD, FAR and continuous variables like RMSE, correlation, mean error and multiplicative bias are carried out by means of Model Evaluation Tools (MET) statistical package. The statistics for categorical variables are formulated using the contingency table shown in table A1. In this table, f represents the forecasts and o represents the observations and the two possible forecast and observation values are represented by the values 0 and 1.

Table A1 HRLDAS model data details.

The values in table A1 are counts of the number of occurrences of the four possible combinations of forecasts and observations. The n _{i j} values in the table represent the counts in each forecast-observation category, where i represents the forecast and j represents the observations. The ‘.’ symbols in the total cells represent sums across categories.

• 1) Equitable Threat Score (ETS)

ETS (also known as Gilbert Skill Score) is based on the Critical Success Index (CSI), corrected for the number of hits that would be expected by chance.

$$\text{ETS}=\frac{n_{11}-C_{1}}{n_{11}+n_{10}+n_{01}-C_{1}}$$

$$C_{1}=\frac{{(n}_{11}+n_{10})\times \left( n_{11}+n_{01}\right)}{T}=\frac{n_{1.}n_{.1}}{T}. $$

ETS values range from –1/3 to 1. A no-skill forecast would have ETS = 0; a perfect forecast would have ETS = 1.

$$\text{CSI}=\frac{n_{11}}{n_{11}+n_{10}+n_{01}} $$

CSI is the ratio of the number of times the event n ₁₁ + n ₁₀ + n ₀₁ was correctly forecasted to occur to the number of times it was either forecasted or occurred. CSI ignores the ‘correct rejections’ category (i.e., n ₀₀). CSI is also known as the Threat Score (TS).

• 2) Probability of detection (POD)

POD is defined as:

$$\text{POD}=\frac{n_{11}}{n_{11}+n_{01}}=\frac{n_{11}}{n_{.1}}. $$

It is the fraction of events that were correctly forecasted to occur and often known as hit rate. POD ranges from 0 to 1 and for a perfect forecast POD = 1.

• 3) False alarm ratio (FAR)

FAR is defined as:

$$\text{FAR}=\frac{n_{10}}{n_{11}+n_{10}}=\frac{n_{10}}{n_{11.}}.$$

It is the proportion of forecasts of the event occurring for which the event did not occur. FAR ranges from 0 to 1 and FAR = 0 for a perfect forecast.

• 4) Root-mean-square error (RMSE)

RMSE is defined as the square root of Mean Square Error (MSE),

$$\text{RMSE}=\sqrt {\text{MSE}} $$

where MSE measures the average squared error of the forecasts

$$\text{MSE}=\frac{1}{n}\sum\limits_{i=1}^{n} \left( f_{i}-O_{i} \right)^{2}. $$

• 5) Correlation

Correlation measures the strength of linear association between observation and forecast. It is defined as:

$$\text{Correlation}=\frac{{\sum}_{i=1}^{T} {\left( f_{i}-\bar{f} \right)\left( O_{i}-\bar{O} \right)} }{\sqrt{\sum \left( f_{i}-\bar{f} \right)^{2}} \sqrt{\sum \left( O_{i}-\bar{O} \right)^{2}} } $$

where $\bar {f}$ represents the forecast mean and $\bar {O}$ represents observation mean. Correlation varies between −1 and 1. Value of 1 indicates perfect correlation and Value of –1 indicates perfect negative correlation. Value of 0 indicates that the forecasts and observations are not correlated.

• 6) Mean error

Mean error is a measure of overall bias for continuous variables and is the bias itself. It is defined as:

$$\text{ME}=\frac{1}{n}\sum\limits_{i=1}^{n} \left( f_{i}-O_{i} \right).$$

• 7) Multiplicative bias

Multiplicative bias is the ratio of the means of forecast and observations and is defined as:

$$\text{MBIAS}=\frac{\bar{f}}{\bar{O}}. $$