The Relationship between Indian Ocean Dipole (IOD) 🌊 and Temperature ☀️ in Asia

Author: Shammunul Islam
Course: Climate Data (CLIM680) (Fall 2023), Department of Atmospheric, Oceanic & Earth Sciences, George Mason University
Contact: sislam27@gmu.edu

Introduction

Due to climate variability and climate change, the world is continuously affected by different extreme events, especially associated with high temperature. Asia is affected by many calamities caused by high temperature driven events. Indian Ocean plays a critical role in the atmospheric dynamics of this region, particularly, Indian Ocean Dipole (IOD), an oscillation of temperature in Indian Ocean equivalent to ENSO in central and eastern tropical Pacific Ocean. In this project, it is investigated whether there is any relationship between different phases of IOD and temperature, if there is, which areas have these relationships, and whether these are statistically significant. Along with this, it was also investigated how the temperature in Asia reponds to different phases of IOD.

For doing this study, APHRODITE Water Resources project data for daily mean temperature is chosen as they have daily temperature data from 1962 upto 2015 in a gridded format with a high spatial resolution of 0.5 degree by 0.25 degree. For sea surface temperature, NOAA OISST V2 SST data is used as it has 1 degree by 1 degree spatial resolution and temporal coverage starting from the end of 1981..

Data

Temperature Data – The APHRODITE Water Resources project

The APHRODITE Water Resources project is working on a series of precipitation products and this project completed its first phase in 2010 and embarked on its second phase in 2016, partnering with Hirosaki University, Kyoto University, and Chiba University.

Daily mean temperature values are provided by their AphroTemp_V1808 at 0.50 and 0.25 degree grid resolutions. These readings are produced by extrapolating information from meteorological stations dispersed throughout the targeted area. After the release of AphroTemp_V1204R1, surface data from nations including India, Bhutan, Thailand, and Myanmar was included. The accuracy of the temperature readings in South Asia in the most recent version has been improved by this inclusion, along with updated interpolation methods and climatic data. For a better understanding of this dataset, you can refer to this documentation

The major characteristics of the dataset
NOAA OI SST V2 High Resolution Dataset

This gridded dataset NOAA OI SST V2 High Resolution Dataset can be found at this link. I used monthly mean of sea surface temperature dataset hosted in Hopper.

Code description or data analysis

Temperature analysis

First of all, air temperature (in Celsius) is plotted over Asia using different Python libraries, including Cartopy and Matplotlib. Detailed step-by-step methods with code can be found in this notebook1. But, the code for each of the figures displayed here can also be seen by clicking on the button placed above each figure.

Now, let’s see if the temperature changed in 2015 compared to 1961 by plotting the temperature anomaly notebook1.

June, July, August, September (JJAS) temperature

This is normally the summer season and here, I have investigated whether temperature increased in Asia in this season. Let’s look at the temperature evolution over the period from 1961 to 2015 notebook1.

Now, let’s have a look at JJAS temperature anomaly over this period notebook1.

We see that it changed over time but it is not clear whether it has a consistent increase or decrease or no change.

Now, let’s have a look at standard deviation of temperature over Asia to see how it varies in different areas of Asia notebook1.

We can see that standard deviation of temperature varies over Asia. Although it seems that the standard deviation increases as we south.

Climatology of temperature in Asia

Now, let’s have a look at the climatology of temperature in Asia Notebook2.

Indian Ocean Dipole – Sea Surface Temperature Analysis

Indian Ocean Dipole (IOD) is a coupled ocean and atmosphere phenomenon in the equatorial Indian Ocean similar to ENSO that affects the Indian Ocean and has three phases. It is characterized by the difference in sea surface temperature between two regions: a western pole in the Arabian Sea (western Indian Ocean) and an eastern pole in the eastern Indian Ocean south of Indonesia. During a positive phase, warm waters are brought up to the western part of the Indian Ocean, and in the eastern Indian Ocean, cold, deep waters rise to the surface. This normally means a higher temperature anomaly in the western area or box relative to the eastern pole or box. In the negative phase of IOD, this is reversed.

Computing IOD using Dipole Mode Index (DMI)

The Indian Ocean Dipole (IOD) phenomenon is commonly identified and measured using the Dipole Mode Index (DMI). The general procedure to compute the DMI from sea surface temperature (SST) data is as follows:

  1. Choose the tropical Indian Ocean regions relevant to IOD. Usually, the regions needed to compute the IOD:

  1. Calculate the average SST anomalies over these two regions over time.

  2. Subtract the eastern region SST anomaly from the western region SST anomaly to get the DMI.

Please go to this notebook to find step by step instruction on doing the above steps with codes Notebook3.

Below is the time series plot of DMI along with temperature anomalies in eastern and western boxes or regions. If you click on the button, you will see from the very start how to use Python to calculate these anomalies for these two regions, how to calculate DMI, and finally how to plot them.

Now, a DMI value of more than or equal to +0.4 is defined as a positive IOD, and a negative value of -0.4 or less is defined as a negative IOD (Source: NOAA). The values between -0.4 and +0.4 are defined as the neutral phase of IOD.

Now, using this definition, we can calculate all these phases of IOD. Below, you can see an animation of the time evolution of IOD (DMI) over time according to its different phases. Again, all the codes can be seen by clicking on the button below or by going to [notebook3] notebook3.

Calculate Composites

Now, we will look at temperature anomalies for different phases of the IOD that we just computed. Note that, for the temperature dataset, all the time values are reported for the last day of a month, while for sea surface temperature (and so for IOD), the data is reported on the first day of every month. As the time coordinates are close but not exact for temperature data and SST data (IOD phase data), we will have to use the sel method with the method='nearest' parameter to select the nearest available time points. The composite is shown below, and you can find all the code for it as well.

From the figure below, we can see that China has a negative temperature anomaly when the IOD is positive and a positive temperature anomaly when the IOD is negative. We can also observe that in Kazakhstan, there is positive response to temperature, that is, with the increase of IOD or SST, the temperature (anomaly) increases and with decreasing SST, temperature decreases Notebook3.

For convenience, we can also look at how different is positive or negative phase compared to neutral phase for temperature. This is also in accordance with the composite plot we have just seen Notebook3.

Is the mean difference between Positive IOD and Neutral IOD significant?

Now, we will check whether the mean difference between Positive IOD and Neutral IOD is statistically significant or not.

Hypothesis Testing

We will test the hypothesis using a two-sided test based on the Student’s t-Test, for the null hypothesis that two independent samples have identical average (expected) values. This test assumes that the populations have identical variances by default.

The following plot (and the accompanying code) shows the locations where the value of t-Test is statistically significant or where the P-value is less than 0.05 Notebook3.

For all the points or locations where the mean temperature during positive IOD is significantly different than neutral phase of IOD is shown with dots. Now, we will look at a single such point, here a point at 75 degree East and 22.5 degree North. We will now visually inspect how IOD and temperature anomaly at this point move together over time Notebook3.

Correlation analysis

Now, we will look at the correlation between DMI (IOD) and temperature anomalies over Asia. Look at the code below Notebook3:

From the above map, we can see that the northeast and northwest of China, south of India, Kazakhstan, Uzbekistan, and east of Mongolia have a high positive correlation coefficient value with DMI, which means that as the positive anomaly increased as defined by DMI, the temperature in these areas also increased simultaneously. Similarly, as temperature anomaly decreases or DMI increases, the temperature also decreases. We can also observe that for Philipines, and Papua New Guinea, the correlation coefficient is negative, meaning that if IOD goes up, temperature goes down, and if IOD goes down, temperature goes up in these areas.

Now, we can check the statistical significance of this correlation Notebook3.

We see that in southern India, Kazakhstan, and the south-eastern and south-western parts of China, IOD and temperature anomalies have a statistically significant positive correlation. While in Papua New Guinea and the Philippines, there is statistically significant negative correlation.

Regression analysis

In this part,we will examine how the temperature anomaly over Asia is explained by IOD. So, we will regress the DMI index on temperature anomalies and see if the variation in temperature anomalies can be explained by DMI. We will also look at which areas or countries have a clear signal from DMI, as indicated by a statistically significant p-value or a p-value less than 0.05. We will also see if the pattern of correlation between DMI and temperature anomalies is observed for the regression coefficient. It is usual to expect to have a similar pattern Notebook3.

Similar to the correlation significance map, we can see that the northeast and northwest of China, south of India, Kazakhstan, Uzbekistan, and east of Mongolia have a high positive regression coefficient, which means that as DMI increases or as sea surface temperature (SST) increases, temperature in these areas also increases. Similarly, as SST decreases, temperature also decreases in these areas. We can also observe that for the Philippines, the north-west part of Thailand, the northern part of Vietnam, and Papua New Guinea, the value of the regression coefficient is negative and significant. For these areas, if SST goes up (positive IOD), temperature increases, and likewise, if SST goes down (negative IOD), temperature goes up in these areas.

Results

This project gives us some insights into the dynamics between the Indian Ocean Dipole (IOD) and temperature variations across Asia. Here are the key findings:

Temporal Temperature Variations: We observed temperature changes over time from 1961 and 2015. This was evident from the temperature anomaly plots and time series plots, indicating a shift in temperature patterns over the decades.

Seasonal Analysis: Focusing on the June to September period (JJAS), a crucial season for Asian climate, we saw an increasing trend in temperature evolution from 1961 to 2015 as found in the line chart.

IOD’s Influence on Regional Temperatures: This project work investigated how different IOD phases - positive, negative, and neutral - impact regional temperatures. The results from composite maps and statistical tests (like t-tests) underscored the significant influence of IOD on temperature anomalies in specific Asian regions.

Correlation and Regression Analyses: We conducted correlation and regression analyses to quantify the relationship between IOD and temperature anomalies. The findings indicated strong correlations in certain areas, affirming the impact of IOD on regional climatic conditions. This was further reinforced by regression analyses, revealing the extent to which temperature variations could be attributed to changes in IOD. Northeast and Northwest of China, south of India, Kazakhstan, Uzbekistan, and east of Mongolia respond positively, while the Philipines, north-west part of Thailand, northern part of Vietnam, and Papua New Guinea respond negatively to IOD change or signal.

Summary

We saw from the analysis that the Indian Ocean Dipole plays a crucial role in influencing temperature patterns across Asia. Using data analysis, visualization, and statistical testing, we saw how differently it affects temperature across Asia. In the future, this work can be expanded by including more variables, such as precipitation patterns and their interaction with IOD, as well as ENSO and how IOD and ENSO together influence both temperature and rainfall in Aisa. This will give us a more holistic understanding of the regional climate system.

I faced many challenges in dealing with these datasets because the temperature dataset and the SST dataset from NOAA uses different time index which made it challenging for me to subset by common time index (especially, when calculating composites). Further, Dask XArray was also a bit challenging as sometimes due to a lack of my klnowledge of it, my code took forever to execute. But, at the end, everything worked and I learned a lot and can’t wait to move ahead and use more sophisticated statical analyses and also ML and DL models with these data to more thoroughly understand how IOD plays a role in Asia. I also plan to include atmospheric analysis or understanding of climate dynamics in this region to conduct a more thorough research coupled with quantitative analysis.

Note: To run the notebooks, please follow the step-by-step guidelines provided in the first page of the repo or you can also look at the README.md file. Detailed instruction on setting up with the file environment.yml is provided there.