Indian Ocean Dipole can be better predicted thru machine learning, say researchers
Researchers in Japan and The Netherlands have, for the first time, used machine learning techniques, in particular artificial neural networks (ANNs), to predict the Indian Ocean Dipole (IOD), a positive phase of which has affected weather and climate in India and Australia in a spectacular fashion so far in 2019-20.
Positive, negative phases
The IOD has both positive and negative phases, and signals large socio-economic impacts on many countries and hence predicting the IOD well in advance will benefit the affected societies, note authors JV Ratnam and Swadhin K Behera (Application Laboratory, Japan Agency for Marine-Earth Science and Technology, Yokohama) and HA Dijkstra (Institute for Marine and Atmospheric Research Utrecht, Utrecht University in The Netherlands) in a paper published by Nature.
The IOD is a mode of climate variability observed in the Indian Ocean sea surface temperature anomalies with one pole in Sumatra (Indonesia) and the other near East Africa. Therefore, the IOD is represented by an index derived from the gradient between the western equatorial Indian Ocean and the south-eastern equatorial Indian Ocean. It starts sometime in May-June, peaks in September-October and ends in November (2019’s rather strong positive phase of the IOD lasted into early January of 2020).
In a positive IOD phase, the western part of the Indian Ocean (closer to East Africa where the monsoon winds turn as south-westerly winds towards India) warms up relative to the eastern basin, beefing up the incoming monsoon flows. These conditions are more or less reversed during a negative IOD phase.
The IOD is also known to affect the climates of other parts of the world, including Sri Lanka, the Maritime Continent (Indonesia, et al), Japan, East Africa and Europe through atmospheric teleconnections. The climate of Australia and the Maritime Continent also are affected by the cool (warm) SST anomalies over the South-East Indian Ocean region during the positive (negative) phase of the IOD.
The anomalously cool (warm) waters around Australia and the Maritime Continent during the positive (negative) phase of IOD reduce (enhance) rainfall over those countries. The IOD also has a remote effect on the climate of Japan through modification of the Pacific-Japan teleconnection and it is also known to affect the summers of Europe due to the atmospheric teleconnections as a response to the IOD.
Wetter India, dry and hot Australia
In recent years, it has been found that the spatial distribution of the summer (monsoon) rainfall over India is affected by IOD during its various phases. During the positive IOD phase, India experiences anomalously high rainfall along the latitude belts covering Central India and during the negative phase of the IOD, the rainfall is anomalously high along the longitudinal belt with the western part of the country receiving high rainfall.
The extended South-West monsoon (June-September) of 2019 in India had a lag effect on the Australian monsoon (delayed to this day), which is thought to have aided and abetted the devastating bush/forest fire in the island-continent. Owing to its large impacts, previous studies have addressed the predictability of the IOD using modern coupled climate models. Various forecasting centres try to predict IOD using the coupled climate models at seasonal time scales. Such dynamical models are promising but are dependent on large computational as well as human resources.
Machine learning to the fore
But in the instant case, researchers in Japan and The Netherlands tried to complement those efforts with a simpler model based on the machine learning technique of ANNs. ANNs are tools used in machine learning which mimic the functioning of neurons in the human brain. Similar to the human brain, the ANN also learns from past data and makes decisions for the future.
An ANN consists of input, output and hidden layers. ANNs have been used in many fields for classification and regression studies to model processes. The correlation analysis of the IOD index indicated that a single ANN model is not suitable for forecasting the IOD index for all the months from May to November. So, researchers developed ANN models for forecasting the IOD index for every month from May to November.
The results were compared with persistence forecasts and also IOD index forecasts derived from the ensemble mean sea surface temperature anomalies of seven models within the North American Multi-Model Ensemble (NMME), an experimental multi-model seasonal forecasting system consisting of coupled models from the US and Canada. The ANN and NMME model results were compared with persistence forecasts to check if the models have skill higher than just persistence of the IOD index of February-April to May-November.
Superior forecast skills
The IOD forecasts were generated for May to November from February-April conditions. The attributes for the ANNs were derived from sea surface temperature and conditions in the upper levels of the atmosphere using a correlation analysis for the period 1949–2018.
An ensemble of ANN forecasts indicates the machine learning-based ANN models to be capable of forecasting the IOD index well in advance with excellent skills. The forecast skills are much superior to the skills obtained from the persistence forecasts that one would guess from the observed data. The ANN models also performed far better than the models of the NMME with higher correlation coefficients and lower root mean square errors for all the target months of May-November.