Analyzing the Hydroclimatic Teleconnections of Summer Monsoon Rainfall in Kerala, India, Using Multivariate Empirical Mode Decomposition and Time-Dependent Intrinsic Correlation (original) (raw)
Related papers
Journal of Climate Change, 2015
Climatic time series data are often nonlinear and non-stationary and hence the use of traditional techniques may not be suitable for their analysis. This research is focused on the monthly series of air temperature and precipitation data of 25 years from 1988 to 2012 at Langtang Meteorological Station (LMS), Kyangjing in Langtang River basin, Nepal to extract multi-scale cycles and trends. To address the non-linearity and non-stationary of these time series, we used Empirical Mode Decomposition (EMD) method. EMD decomposed LMS temperature and precipitation series into different oscillatory modes called Intrinsic Mode Functions (IMFs) and residue called trend. The extracted IMFs are subjected to Fast Fourier Transform (FFT) to determine their average period along with their power density. There exist oscillations of 1 year, 3.13 years, 6.25 years, 8.33 years and 12.5 years in temperature data. Among these cycles, only 1 year cycle is distinguished from Gaussian white noise at 95% confidence level. The air temperature at LMS, Kyangjing reflects monotonic positive trend till 2006 but remains as nearly steady state around 3°C from the end of 2006. Similarly, the precipitation data is embedded with cycles of 6 months, 1 year, 2.08 years, 2.27 years and 8.33 years of which only the first three are statistically significant at 95% confidence level. The precipitation shows a mixed trend with decreasing pattern till mid 1990s, increasing pattern till mid 2000s and again decreasing pattern till 2012. One year cycle is dominant in both the time series data. The above results reflect that temperature and precipitation fluctuates on various time scales. The effect of the changes in temperature and precipitation has already been manifested in the form of melting glaciers in this region. The causes for these oscillations might be related to phenomenon like Quasi-biennial Oscillation (QBO), solar activity, El Nino, monsoon climate dynamics and other local characteristics of the basin.
Analysis of Rainfall and Temperature Data Using Ensemble Empirical Mode Decomposition
Data Science Journal
Climatic variables such as rainfall and temperature have nonlinear and non-stationary characteristics such that analysing them using linear methods inconclusive results are found. Ensemble empirical mode decomposition (EEMD) is a data-adaptive method that is best suitable for data with nonlinear and non-stationary characteristics. The average monthly rainfall and temperature data for a selected region in South Africa are decomposed into intrinsic mode functions (IMFs) at different time scales using EEMD. The IMFs exhibit an inter-annual to inter-decadal variability. The influence of climatic oscillations such as El-Niño Southern Oscillation (ENSO) and quasi-biennial oscillation (QBO) is identified. The influence of temperature variability on rainfall is also shown at different time scales. Based on the results obtained, the EEMD method is found to be suitable to identify different oscillations in the rainfall and temperature data.
The singular value decomposition technique (SVD) is used to analyze the covariability between anomalous northeast monsoon (NEM) rainfall in Malaysia and the large-scale anomalous sea surface temperature (SST) in Indian Ocean, Pacific Ocean and seas surrounding the Maritime Continent. The SVD analysis reveals two significant coupled modes of covariability with the first dominant mode explaining ~75% while the second coupled mode explaining ~15% of the total covariance. The first coupled mode highlights the covariability between anomalous NEM rainfall in East Malaysia and anomalous SST associated with the biennial oscillation type (BO-type) of the El Nino-Southern Oscillation (ENSO). The second coupled mode highlights the covariability between anomalous NEM rainfall in West Malaysia and anomalous SST associated with the low frequency type (LF-type) of ENSO. Overall, the BO-type and LF-type of ENSO define two distinct regimes of different behaviour of anomalous NEM rainfall in Malaysia East Malaysia and West Malaysia regions. During the BO-type of ENSO, East Malaysia region is mostly affected while during the LF-type of ENSO, the impacts are mostly confined in West Malaysia region.
Modeling Earth Systems and Environment, 2017
different lower modes of PDO series with that of maximum and minimum temperature of the different regions depicted the possible association of PDO with the temperature regime of southern India. This association is further investigated with a recently developed running correlation analysis method namely Time Dependent Intrinsic Correlation (TDIC), which deciphered a long range positive correlation between the PDO and minimum temperature series at inter annual mode of 7 years to inter decadal mode of 15 years at all the three regions of southern India. The association between PDO and maximum temperature at inter annual ranges is positive at EC and IP regions. The association between PDO and maximum temperature at inter decadal scale show regional differences, with no association at WC region, negative association at IP region and positive association at EC region.
Meteorology and Atmospheric Physics, 2018
Rainfall is a vital hydrologic variable that has a direct and significant impact on the economic development of monsoondominated state of Kerala in southern India. An effective approach providing accurate prediction of rainfall makes it possible to take preventive and mitigation measures against natural disasters. In this study, the ensemble empirical mode decomposition (EEMD)-artificial neural networks (ANN)-multiple linear regression (MLR) hybrid approach is used to forecast the southwest monsoon (SWM) rainfall of Kerala. The EEMD of SWM rainfall of Kerala resulted in a set of orthogonal components of specific periodic scale. The non-linear components are identified and separately modeled using ANN and rest of the components are modeled using linear regression to get their values at a specific time t. Finally, the predicted modes are recombined to get the forecasts of a generic time t. The SWM rainfalls of 1871-1972 are used for model calibration and forecasts are made sequentially for 1973-2014 period, which clearly demonstrated its efficacy in handling non-linear part of SWM rainfall data with a predictive skill of 0.65 for validation data. Further, by considering a dataset of 1961-2014 period, this study has investigated the possible teleconnection of SWM rainfall of Kerala with the Indian Ocean dipole (IOD) using the cross-correlation and EEMD-based time-dependent intrinsic correlation (TDIC) analyses. Apart from the strong correlation in the trend component, the analysis has proved the dominancy of negative association of IOD with SWM of Kerala in different process scales with strong positive association at localized time spells. The forecasting strategy demonstrated in the study and the evidence of IOD-SWM rainfall link are an amendment to the efforts for improving the predictability of SWM rainfall in Kerala.
A coupled model study on ENSO, MJO and Indian summer monsoon rainfall relationships
Meteorology and Atmospheric Physics, 2003
In this paper, we have tried to understand the ENSO, MJO and Indian summer monsoon rainfall relationships from observation as well as from coupled model results. It was the general feeling that El-Niñ no years are the deficient in Indian monsoon rainfall and converse being the case for the La-Niñ na years. Recent papers by several authors noted the failure of this relationship. We find that the model output does confirm a breakdown of this relationship. In this study we have seen that a statistically defined modified Indian summer monsoon rainfall (MISMR) index, a linearly regressed ISMR index and dynamical Webster index (WBSI), shows an inverse relationship with ENSO index during the entire period of integration (1987 to 1999). It is also seen from this study that the amplification of the MJO signals were large and the ENSO signals were less pronounced during the years of above normal ISMR. The MJO signal amplitudes were small and ENSO signals were strong during the years of deficient ISMR. It has been noted that here is a time lag between the MJO and ENSO signal in terms of their modulation aspect. If time lag is added with the ENSO signal then both signals maintain the amplitude modulation theory. A hypothesis is being proposed here to define a relationship between MJO and ENSO signals for the entire period between 1987 and 1999.
Scientific Reports, 2018
Several cases of failure in the prediction of Indian Summer Monsoon Rainfall (ISMR) are the major concern for long-lead prediction. We propose that this is due to the temporal evolution of association/ linkage (inherent concept of temporal networks) with various factors and climatic indices across the globe, such as El Niño-Southern Oscillation (ENSO), Equatorial Indian Ocean Oscillation (EQUINOO), Atlantic Multidecadal Oscillation (AMO), North Atlantic Oscillation (NAO), Pacific Decadal Oscillation (PDO) etc. Static models establish time-invariant (permanent) connections between such indices (predictors) and predictand (ISMR), whereas we hypothesize that such systems are temporally varying in nature. Considering hydroclimatic teleconnection with two major climate indices, ENSO and EQUINOO, we showed that the temporal persistence of the association is as low as three years. As an application of this concept, a statistical time-varying model is developed and the prediction performance is compared against its static counterpart (time-invariant model). The proposed approach is able to capture the ISMR anomalies and successfully predicts the severe drought years too. Specifically, 64% more accurate performance (in terms of RMSE) is achievable by the recommended time-varying approach as compared to existing time-invariant concepts. Spatio-temporal variability of rainfall has significant economic and societal impacts, particularly for agriculture based countries. For instance, India receives more than 80% annual rainfall in four monsoon months (June-September) and substantial fluctuations are noticed in the annual food grain production due to the vagaries of monsoon. The association can be stated by the fact that the dips (such as 2002 and 2014) and peaks (such as 2013 and 2016) in the food grain production correspond to below normal and above normal monsoon rainfall respectively 1. Naturally, considerable efforts including long-lead monsoon prediction needs to be made towards identifying and adopting strategies to deal with crisis situations. Due to considerable complexity in the evolution of ISMR, the long-lead prediction remains as a challenging task, particularly in a changing climate 2-4. A large number of studies have analyzed the inter-annual variability of ISMR 5 ; however the long-term climate fluctuations which modulate such variability are still not clear. The long-range forecasting of ISMR was started more than a century ago. These are broadly grouped into statistical 6-18 and dynamical 19-25 forecasting approaches. Despite advancement in physical understanding and development of advanced statistical models the forecast failed in recent years too. Some major issues and inherent problems in the statistical models such as variation in the predictor-predictand relationship over time and conditional dependence among the predictors shows the necessity for constant scrutiny and update in the models 2,12,26-29. The changes can be brought out in different ways, such as the use of new predictors, changing the combination of the predictors or lags, updating the model parameters, etc. Predictor selection is an important aspect of statistical modeling and two climatic indices strongly influencing the variability of ISMR are El Niño-Southern Oscillation (ENSO) and Equatorial Indian Ocean Oscillation (EQUINOO) 10,13,15,30-32. The concept of utilization of ENSO and EQUINOO as predictor lies in the hydroclimatic
Empirical mode decomposition analysis of climate changes with special reference to rainfall data
Discrete Dynamics in Nature and Society, 2006
We have used empirical mode decomposition (EMD) method, which is especially well fitted for analyzing time-series data representing nonstationary and nonlinear processes. This method could decompose any time-varying data into a finite set of functions called "intrinsic mode functions" (IMFs). The EMD analysis successively extracts the IMFs with the highest local temporal frequencies in a recursive way. The extracted IMFs represent a set of successive low-pass spatial filters based entirely on the properties exhibited by the data. The IMFs are mutually orthogonal and more effective in isolating physical processes of various time scales. The results showed that most of the IMFs have normal distribution. Therefore, the energy density distribution of IMF samples satisfies χ 2 -distribution which is statistically significant. This study suggested that the recent global warming along with decadal climate variability contributes not only to the more extreme warm events, but also to more frequent, long lasting drought and flood.
MAUSAM
The spatial variability of northeast monsoon (NEM) rainfall of peninsular India and Sri Lanka is studied using Empirical Orthogonal Function (EOF) analysis based on monthly / seasonal rainfall data for the months of October, November, December, January and for the season October-December (OND) for the 107 year period of 1900-2006 over nine sub-regions defined for the study based on climatology and geography. Monthly / seasonal rainfall series over these nine sub-regions are subjected to EOF analysis and 2-3 significant Principal Components (PCs) are identified for each case. Each PC is then linked to physical modes known to be associated with NEM using correlation and compositing techniques. For the OND season and for all the four individual months, the first PC explaining maximum variance of 49-64% in the spatial rainfall distribution is identified with the overall NEM strength. The second and the third PCs are identified with rainfall due to passage of synoptic scale systems such ...
Asia-Pacific Journal of Atmospheric Sciences, 2019
The Western Ghats (WG) of India are basically north-to-south oriented mountains with three distinct meteorological divisions. These mountains exhibit the characteristic features of precipitation and distribution during the summer monsoon season and possess latitudinal variations. It is a well-known fact that sea surface temperature (SST) combined with the El Niño-Southern Oscillation (ENSO) enacts a predominant role in the precipitation over the entire Western Ghats during the summer monsoon season. Whereas the Niño regions affect the variability of the Western Ghats' precipitation in an asymmetric relationship. Nevertheless, the simulation of precipitation has been evidenced to be difficult. The current study attempts to predict the seasonal precipitation over the coastal region and the Western Ghats of Karnataka. The relationship between summer monsoon precipitation (SMP) and SST is examined up to eight seasons by conducting the correlation analysis with three seasons that lag before the onset of the monsoon season. The significant and positively correlated lagged Niño indices with the SMP index are identified as the predictors. The selected predictors are used for predicting the SMP by using statistical models, the multiple linear regression model and the artificial neural network (ANN) model. The statistical models are based on the combined lagged indices and the principle component as the predictor. The results of the statistical models on comparison suggest that neural network models have a better predictive skill than the linear regression models. Neural network models with combined lagged indices being used as predictors are slightly better, but a few more climatic parameters must be verified and the usage of this method on other meteorological divisions of the West Coast of India needs to be further investigated.