Exploring the long-term changes in the Madden Julian Oscillation using machine learning

Abstract

The Madden Julian Oscillation (MJO), the dominant subseasonal variability in the tropics, is widely represented using the Real-time Multivariate MJO (RMM) index. The index is limited to the satellite era (post-1974) as its calculation relies on satellite-based observations. Oliver and Thompson (J Clim 25:1996–2019, 2012) extended the RMM index for the twentieth century, employing a multilinear regression on the sea level pressure (SLP) from the NOAA twentieth century reanalysis. They obtained an 82.5% correspondence with the index in the satellite era. In this study, we show that the historical MJO index can be successfully reconstructed using machine learning techniques and improved upon. We obtain a significant improvement of up to 4%, using the support vector regressor (SVR) and convolutional neural network (CNN) methods on the same set of predictors used by Oliver and Thompson. Based on the improved RMM indices, we explore the long-term changes in the intensity, phase occurrences, and frequency of the winter MJO events during 1905–2015. We show an increasing trend in MJO intensity (22–27%) during this period. We also find a multidecadal change in MJO phase occurrence and periodicity corresponding to the Pacific Decadal Oscillation (PDO), while the role of anthropogenic warming cannot be ignored.

A Current Approach to Mean Temperature Trends and Relationships With Teleconnection Patterns in California (U.S.)

“…11) shows highest value of correlation in November 41.3% and January 19.8% of the meteorological stations. One of the reasons of this results that are consistent with other research, could be that RMM1 describes the situation when a MJO produces an enhanced convection at the Maritime Continent while RMM2 has enhanced convection over the Paci c Ocean90,91 closer to California State. In addition it is remarkable that the correlated stations with temperatures, in June, are located mainly both in coastal areas of San Francisco, Santa Barbara and Los Angeles and mountainous areas of Auburn, Oroville and Susanville Dasgupta et al (2020).…”

“…One of the reasons of this results that are consistent with other research, could be that RMM1 describes the situation when a MJO produces an enhanced convection at the Maritime Continent while RMM2 has enhanced convection over the Paci c Ocean90,91 closer to California State. In addition it is remarkable that the correlated stations with temperatures, in June, are located mainly both in coastal areas of San Francisco, Santa Barbara and Los Angeles and mountainous areas of Auburn, Oroville and Susanville Dasgupta et al (2020). found that the occurrences of MJO activity at RMM phase locations 4, 5 and 6 during boreal winter are related to the PDO index, particularly in the negative phases.…”

Machine learning to extend and understand the sources and limits of water cycle predictability on subseasonal-to-decadal timescales in the Earth system

“…Synthetic data can also be generated using ML applied to observations and reanalysis to extend the availability of labeled data to train a CNN on S2D timescales. Dasgupta et al (2020) recently demonstrated this technique with a CNN-based approach that improved historical reconstructions of the MJO, which was then used to explore its decadal variability. This type of semi-supervised approach can extend S2D predictability by increasing the availability of labeled data.…”

Deep learning–based downscaling of summer monsoon rainfall data over Indian region

“…There could be various reasons for these issues, particularly in India, such as the complex, non-linear and turbulent weather in the tropical regions and the usage of parameterization schemes generating precipitation in the model. Other than numerical models, statistical and feature selection-based Artificial Neural Network (ANN) models have been used in the past to predict rainfall in different time scales with some success [9], [10] . These models employ two popular concepts: feature selection and then prediction using statistical or simple machine learning algorithms [9], [11]- [13].…”