Empirical Prediction and Predictability of Dry and Wet Spells of the Indian Summer Monsoon

Abstract

n active (break) phase of intraseasonal oscillation (ISO) of the Indian summer monsoon is associated with a wet (dry) spell over the monsoon trough region and a dry (wet) spell over the equatorial Indian Ocean. Prediction of the dry and wet spells (break or active phases) of the Indian summer monsoon two to three weeks in advance is of great importance for food production and water management of the country, but is currently unavailable. While the last two decades have seen considerable advancement in our understanding of the basic characters of the monsoon ISO, the same has not been translated into predictive tools. The objective of this study is to develop a method for predicting the dry and wet spells of the monsoon about two weeks in advance. Prior to getting into the technical aspects of the prediction technique, potential for predictability of these spells is established by demonstrating that the intraseasonal fluctuations in circulation and precipitation of the Indian summer monsoon are integral parts of a low-frequency large-scale convectively coupled oscillation.

We emphasise that the day-to-day fluctuations of rainfall have predictability limited only to 2–3 days as they are governed by synoptic systems. The intraseasonal component may have predictability up to 2–3 weeks as they are part of a large-scale low-frequency oscillation. Thus, our aim is to predict only the intraseasonal envelope of rainfall variations. For this purpose, we use Climate Prediction Center Merged Analysis of Precipitation (CMAP) pentad data for the period 1979–2001. We show that CMAP pentad rainfall represents intraseasonal rainfall variability over the Indian monsoon region (including continental India) reasonably well. The first four Principal Components (PCs) are used as predictors, as the first four Empirical Orthogonal Functions (EOFs) of CMAP contain most of the low-frequency large-scale component (predictable component) of rainfall variability. In addition to CMAP rainfall, we also use daily 850 hPa vorticity and surface pressure from NCEP/NCAR reanalysis for developing the prediction model.

The potential for prediction of the monsoon ISO is explored by developing a multiple linear regression model that predicts the first four principal components (PCs) of CMAP. The first four PCs of rainfall, the first four PCs of 850 hPa vorticity, and the first two PCs of surface pressure are used as predictors. The model is developed in a stepwise manner by adding the PCs of 15–90 day filtered precipitation, vorticity, and surface pressure one by one on data over 17 northern hemisphere summers (1 June to 30 September of 1979–1995) and tested over recent five summers. The model gives optimum performance with the above-mentioned 10 predictors. Skillful predictions of the intraseasonal component of rainfall over most of the Indian monsoon region are possible up to a lead time of 15 days. The phase and northward propagation of the precipitation ISO are well predicted by the model. Predictions are found to be dependent on initial conditions. The predictions made from transition states (from dry to wet or from wet to dry) are only marginally skillful, while those made from either active (wet) or weak (dry) initial conditions are significantly skillful.

Predictions of area-averaged rainfall over three regions—the monsoon trough (70°–95°E, 15°–25°N), the equatorial central Indian Ocean (70°–95°E, 5°S–5°N), and the equatorial eastern Indian Ocean including islands (95°–110°E, 5°S–5°N)—show that the 15-day predictions starting from an active phase of the monsoon ISO (i.e., wet spell over the monsoon trough) and the 12-day predictions starting from a break phase (i.e., dry spell over the monsoon trough) are significantly skillful in all three regions. The correlation between predictions and observations during the verification period is better than 0.7 (sample size 68). The skill shows potential for application of these predictions in real-time.