Predicting maximum temperatures over India 10-days ahead using machine learning models.

In the months of March-June, India experiences high daytime temperatures (Tmax), which sometimes lead to heatwave-like conditions over India. In this study, 10 different machine learning models are evaluated for their ability to predict the daily Tmax anomalies 10 days ahead in the months of March-June. Several model experiments were carried out to identify an optimal model to predict daily Tmax anomalies over India. The results indicate that the AdaBoost regressor with Multi-layer Perceptron as the base estimator is an optimal model to predict the Tmax anomalies over India in the months of March-June. The optimal model predictions are benchmarked against 10-day persistence predictions and the predictions from the Climate Forecast System (CFS) reforecast. The results indicate that the machine learning model skill is higher than persistence and comparable to CFS reforecast 10-day predictions in April and May. In March and June, the machine learning models have low skill scores and perform no better than persistence. These results indicate that the machine learning models are promising tools to predict the surface air maximum temperature anomalies over India in April and May and can complement predictions from more sophisticated numerical models.

A machine learning based prediction system for the Indian Ocean Dipole.

The Indian Ocean Dipole (IOD) is a mode of climate variability observed in the Indian Ocean sea surface temperature anomalies with one pole off Sumatra and the other pole near East Africa. An IOD event starts sometime in May-June, peaks in September-October and ends in November. Through atmospheric teleconnections, it affects the climate of many parts of the world, especially that of East Africa, Australia, India, Japan, and Europe. Owing to its large impacts, previous studies have addressed the predictability of the IOD using state of the art coupled climate models. Here, for the first-time, we predict the IOD using machine learning techniques, in particular artificial neural networks (ANNs). The IOD forecasts are generated for May to November from February-April conditions. The attributes for the ANNs are derived from sea surface temperature, 850 hPa and 200 hPa geopotential height anomalies, using a correlation analysis for the period 1949-2018. An ensemble of ANN forecasts is generated using 500 samples with replacement using jackknife approach. The ensemble mean of the IOD 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 perform far better than the models of the North American Multi-Model Ensemble (NMME) with higher correlation coefficients and lower root mean square errors (RMSE) for all the target months of May-November.

Long-lead Prediction of ENSO Modoki Index using Machine Learning algorithms.

The focus of this study is to evaluate the efficacy of Machine Learning (ML) algorithms in the long-lead prediction of El Niño (La Niña) Modoki (ENSO Modoki) index (EMI). We evaluated two widely used non-linear ML algorithms namely Support Vector Regression (SVR) and Random Forest (RF) to forecast the EMI at various lead times, viz. 6, 12, 18 and 24 months. The predictors for the EMI are identified using Kendall's tau correlation coefficient between the monthly EMI index and the monthly anomalies of the slowly varying climate variables such as sea surface temperature (SST), sea surface height (SSH) and soil moisture content (SMC). The importance of each of the predictors is evaluated using the Supervised Principal Component Analysis (SPCA). The results indicate both SVR and RF to be capable of forecasting the phase of the EMI realistically at both 6-months and 12-months lead times though the amplitude of the EMI is underestimated for the strong events. The analysis also indicates the SVR to perform better than the RF method in forecasting the EMI.

Publisher Correction: Malaria predictions based on seasonal climate forecasts in South Africa: A time series distributed lag nonlinear model.

An amendment to this paper has been published and can be accessed via a link at the top of the paper.

Improving seasonal forecasts of air temperature using a genetic algorithm.

Seasonal forecasts of air-temperature generated by numerical models provide guidance to the planners and to the society as a whole. However, generating accurate seasonal forecasts is challenging mainly due to the stochastic nature of the atmospheric internal variability. Therefore, an array of ensemble members is often used to capture the prediction signals. With large spread in the prediction plumes, it becomes important to employ techniques to reduce the effects of unrealistic members. One such technique is to create a weighted average of the ensemble members of seasonal forecasts. In this study, we applied a machine learning technique, viz. a genetic algorithm, to derive optimum weights for the 24-ensemble members of the coupled general circulation model; the Scale Interaction Experiment-Frontier research center for global change version 2 (SINTEX-F2) boreal summer forecasts. Our analysis showed the technique to have significantly improved the 2m-air temperature anomalies over several regions of South America, North America, Australia and Russia compared to the unweighted ensemble mean. The spatial distribution of air temperature anomalies is improved by the GA technique leading to better representation of anomalies in the predictions. Hence, machine learning techniques could help in improving the regional air temperature forecasts over the mid- and high-latitude regions where the model skills are relatively modest.

Malaria predictions based on seasonal climate forecasts in South Africa: A time series distributed lag nonlinear model.

Although there have been enormous demands and efforts to develop an early warning system for malaria, no sustainable system has remained. Well-organized malaria surveillance and high-quality climate forecasts are required to sustain a malaria early warning system in conjunction with an effective malaria prediction model. We aimed to develop a weather-based malaria prediction model using a weekly time-series data including temperature, precipitation, and malaria cases from 1998 to 2015 in Vhembe, Limpopo, South Africa and apply it to seasonal climate forecasts. The malaria prediction model performed well for short-term predictions (correlation coefficient, r > 0.8 for 1- and 2-week ahead forecasts). The prediction accuracy decreased as the lead time increased but retained fairly good performance (r > 0.7) up to the 16-week ahead prediction. The demonstration of the malaria prediction process based on the seasonal climate forecasts showed the short-term predictions coincided closely with the observed malaria cases. The weather-based malaria prediction model we developed could be applicable in practice together with skillful seasonal climate forecasts and existing malaria surveillance data. Establishing an automated operating system based on real-time data inputs will be beneficial for the malaria early warning system, and can be an instructive example for other malaria-endemic areas.

Quasi-asymmetric response of the Indian summer monsoon rainfall to opposite phases of the IOD.

The El Niño/Southern Oscillation has been traditionally linked to the extremes in the Indian summer monsoon rainfall (ISMR) affecting more than a billion people in the region. This trans-oceanic influence is seen to be moderated by the Indian Ocean Dipole (IOD) phenomenon in recent decades. In the presence of a positive IOD (pIOD), the otherwise subdued ISMR in an El Niño year remains close to normal even in the face of record breaking El Niños. While this general influence of pIOD on ISMR is understood, the influence of negative IOD (nIOD) on ISMR is not yet recognized. In this study, it is revealed that those opposite phases of IOD are associated with distinct regional asymmetries in rainfall anomalies. The pIOD is associated with a tripolar pattern in rainfall anomalies with above normal rainfall in central parts of India and below normal rainfall to north and south of it. Conversely, the nIOD is associated with a zonal dipole having above (below) normal rainfall on the western (eastern) half of the country. This spatial quasi-asymmetry arises from the differences in the atmospheric responses and the associated differences in moisture transports to the region during contrasting phases of the IOD.

ENSO's far reaching connection to Indian cold waves.

During boreal winters, cold waves over India are primarily due to transport of cold air from higher latitudes. However, the processes associated with these cold waves are not yet clearly understood. Here by diagnosing a suite of datasets, we explore the mechanisms leading to the development and maintenance of these cold waves. Two types of cold waves are identified based on observed minimum surface temperature and statistical analysis. The first type (TYPE1), also the dominant one, depicts colder than normal temperatures covering most parts of the country while the second type (TYPE2) is more regional, with significant cold temperatures only noticeable over northwest India. Quite interestingly the first (second) type is associated with La Niña (El Niño) like conditions, suggesting that both phases of ENSO provide a favorable background for the occurrence of cold waves over India. During TYPE1 cold wave events, a low-level cyclonic anomaly generated over the Indian region as an atmospheric response to the equatorial convective anomalies is seen advecting cold temperatures into India and maintaining the cold waves. In TYPE2 cold waves, a cyclonic anomaly generated over west India anomalously brings cold winds to northwest India causing cold waves only in those parts.

Anatomy of Indian heatwaves.

India suffers from major heatwaves during March-June. The rising trend of number of intense heatwaves in recent decades has been vaguely attributed to global warming. Since the heat waves have a serious effect on human mortality, root causes of these heatwaves need to be clarified. Based on the observed patterns and statistical analyses of the maximum temperature variability, we identified two types of heatwaves. The first-type of heatwave over the north-central India is found to be associated with blocking over the North Atlantic. The blocking over North Atlantic results in a cyclonic anomaly west of North Africa at upper levels. The stretching of vorticity generates a Rossby wave source of anomalous Rossby waves near the entrance of the African Jet. The resulting quasi-stationary Rossby wave-train along the Jet has a positive phase over Indian subcontinent causing anomalous sinking motion and thereby heatwave conditions over India. On the other hand, the second-type of heatwave over the coastal eastern India is found to be due to the anomalous Matsuno-Gill response to the anomalous cooling in the Pacific. The Matsuno-Gill response is such that it generates northwesterly anomalies over the landmass reducing the land-sea breeze, resulting in heatwaves.