Dry Season, Wet Season: A Look at Climate Trends in Bangladesh from 1901–2023
Like many other countries near the equator, Bangladesh endures two starkly different periods of weather activity every year. The dry season (known for its sweltering temperatures and dusty air) and the wet season (often accompanied by potentially treacherous monsoon rains) have been hallmarks of the Bangladeshi way of life for centuries, but there may be cause for concern about how climate change, as well as the ecological impacts of human and technological innovation, may affect the lives of the country’s residents.
Bangladesh sits in a precarious position geographically. In the north are the Himalayan mountains, where runoff generated by melting snow can lead to flooding in the neighboring regions and potentially surge in response to global warming. In the south is Cox’s Bazar Beach, well-known as one of Bangladesh’s premier tourist attractions and the longest uninterrupted beach in the world, which is extremely vulnerable to coastal erosion and rising sea levels. In addition, as industrial work and infrastructure development continue to progress throughout the country, air quality suffers due to heavy pollution and dry season dust. Hotter temperatures could lead to even more arid conditions and even higher potential for public health crises (respiratory diseases and other pollution-influenced ailments are already problematic for Bangladeshi citizens).
Considering these various factors and potential consequences, the Bangladeshi government faces an important question: should more resources be invested into determining whether there is any evidence to suggest that climate patterns have already begun to shift substantially enough to warrant the implementation of safety precautions for Bangladesh and its citizens? The answer will indicate how a nation should prioritize its efforts to avoid potential health, economic, and ecological crises over the next few decades.
Records of standard weather metrics (ex: temperature, precipitation, humidity, air quality) are the best place to start looking at potential deviations from the norm. Determining whether or not there are any observable differences in the climate throughout recent history will either justify a call to action or whether further research is required.
This exploratory analysis is based on a dataset of weather data recorded over more than century by NASA and the Humanitarian Data Exchange, compiled by Yakin Rubaiat and published as a CSV file on Kaggle. Each observation includes the monthly average temperature (in degrees Celsius), monthly average rainfall (in millimeters), the month (as a number from 1 to 12) in which each observation was recorded, and the corresponding year.
Data Cleaning and Preparation
Before diving into the data, a few steps were taken to clean it in preparation for visualization and analysis. The first step was to relabel the column (variable) names of each observation for consistency and clarity. Although the units for temperature and rainfall were mentioned in the dataset’s Data Card on Kaggle (similar to a README file), they were not apparent in the data itself. Columns like tem and rain were renamed to temp_c and rain_mm to avoid any confusion. Second, a column displaying the name of the month in which each observation was recorded was added to the data for the sake of improved readability; this column was later used to create the visualizations which are color-coded by month. Finally, the .info() method was used to ensure that there were no null values or unexpected datatypes in the dataset.
Visualizations of the Data
The first two visualizations looked at the overall changes in monthly rainfall and temperature over the entire period contained in the dataset (1901–2023). Because of the way the dataset was set up, it was difficult to graph data in a chronological fashion with the year on the x-axis and the desired measure for each graph (rainfall or temperature, respectfully) on the y-axis. As a workaround, each graph was created with the index on the x-axis and rainfall or temperature on the y-axis. Because the x-axis on these graphs ran from 0 to 1474 (the total number of records in the dataset) instead of 1901 to 2023, the axis values were removed to avoid misconceptions about discrepancies across different visualizations.
There are clearly some local maxima and minima that can be discerned from these visualizations. In addition to the overall temperature fluctuations over time, it is also helpful to look at the temperature fluctuations of each month over time. Anomalies in climate patterns can then be discovered by comparing unusual observations to what can normally be expected at different times of the year. The next two visualizations show this alternate view, with lines for different months color-coded for differentiability.
Some differences in trends based on month are more visible with these visualizations, but they can be made clearer by highlighting each month’s temperature fluctuations over time in comparison to the others. This change is implemented in the two visualizations below.
Results and Discussion
In these visualizations, the most interesting insights can be drawn from the rainfall records over the relevant time period (most of the temperature trends are quite consistent). Further analysis would be required to determine how statistically significant the anomalous observations are (such as the large peaks in rainfall during August and June), but the initial impression is that these findings are worthy of a deeper dive.
To build on these findings, steps should be taken to address the limitations of the current dataset. First, more detailed data should be acquired in order for more custom aggregations and analyses. As collected, the temperature data has already averaged, but it is unclear whether any statistical measures, such as removal of outliers, were taken before calculating those averages. The methodology of how the monthly rainfall was calculated is also not apparent, nor is the methodology of how the data was collected. Although averages are acceptable for an initial analysis like this one, more specific temperature and rainfall data, especially if collected from different regions of the country, would decrease the likelihood that localized anomalies are being concealed.
In addition, other data, such as monitoring of climate change indicators (ex: ozone layer, greenhouse gas emissions, sea levels) and consequences of environments and surroundings (ex: spread of air- and water-borne diseases, natural disasters, displacement) over time, should be compared with this data to check for any correlations between periods of abnormal climate activity and societal impacts.
Reference
The code used to perform data cleaning, manipulation, and visualization of the Bangladeshi weather dataset can be found at the link below.
